RESEARCH | REPORT
the continuum (17). The fraction of the photon far—would likely depend on a and b, but they The analogy between electron emission from
energy that exceeds the sum of the adiabatic are equal for the different pathways that inter- the H2 molecule and a classical double-slit ex-
double ionization energy of H2 (31.03 eV) and fere under a certain emission angle a and do periment suggests that the birth time delay
the KER is shared between the two electrons. not influence our measurement. could be interpreted as the travel time of the
The symmetry of this energy sharing is a photon from one molecular center to the other,
measure for the strength of the Coulomb We compared our experimental findings with which is up to 247 zs for the average bond
interaction among the two electrons, which two simple models. First, we assumed that t′b is length of H2. Our experimental results support
has the potential to destroy the single-particle given by the time difference with which a point this picture, but studies targeting more-complex
quantum interference pattern (12). Therefore, of constant phase of the photon wave hits the molecules and applying more-sophisticated the-
we restricted our investigation to fast elec- two centers oretical models are necessary to further unveil
trons that carried >96% of the excess energy the scope of birth time delay. This work can
(fig. S2). For such fast electrons, double-slit t′b ¼ cosðbÞ R ð3Þ function as a benchmark for such studies.
interference effects are well established on c
the single-particle level (11, 18). Corresponding REFERENCES AND NOTES
slow electrons of the double ionization pro- where c is the speed of light. In the case of
cess had <4% of the excess energy and are not cos(b) = 0, the photon wave hits both centers 1. M. Schultze et al., Science 328, 1658–1662 (2010).
shown here (19). of the molecule simultaneously, and there can- 2. M. Isinger et al., Science 358, 893–896 (2017).
not be any birth time delay between electrons 3. A. L. Cavalieri et al., Nature 449, 1029–1032 (2007).
In Fig. 2A, we display the electron angular emitted from one or the other center because 4. Z. Tao et al., Science 353, 62–67 (2016).
distribution of those fast electrons for the aver- the outgoing waves are exactly in phase. On 5. J. Vos et al., Science 360, 1326–1330 (2018).
age internuclear distance of 0.74 Å (purple line). the other hand, cos(b) = ±1 resembles the max- 6. S. Beaulieu et al., Science 358, 1288–1294 (2017).
The results show a rich structure, which—as imum possible travel time of the photon from 7. J. M. Dahlström, A. L’Huillier, A. Maquet, J. Phys. B 45, 183001
expected—resembles the interference pattern one molecular center to the other. For this case,
of electrons emerging from a double slit. Figure the expected birth time delay was ±247 zs for (2012).
2B displays this measured interference pattern R = 0.74 Å. Between these extreme cases, the 8. H. D. Cohen, U. Fano, Phys. Rev. 150, 30–33 (1966).
of the fast electron as a function of the inter- birth time delay showed a linear dependence 9. J. Fernández, O. Fojón, A. Palacios, F. Martín, Phys. Rev. Lett.
nuclear distance. The results show how the on cos(b). For comparison to the experimental
number of interference fringes increased and data, the blue line in Fig. 3C resembles this 98, 043005 (2007).
how the angular separation of the maxima simple model. Note that the model agreed with 10. S. E. Canton et al., Proc. Natl. Acad. Sci. U.S.A. 108,
decreased with increasing internuclear dis- the prediction from equation 12 of (20), if one
tance R, which affirms the double-slit nature neglects the ionization potential. 7302–7306 (2011).
of the electron emission. 11. D. Akoury et al., Science 318, 949–952 (2007).
Second, the red line in Fig. 3C shows the 12. M. Waitz et al., Phys. Rev. Lett. 117, 083002 (2016).
The data shown in Fig. 2 were averaged over result from a more refined model. This model 13. This interferometric technique is insensitive to the length of
all orientations of the molecular axis with re- accounted for the fact that—other than in the
spect to the light propagation and the light’s optical double slit—in photoionization, the two the used synchrotron light pulses up to 100 ps.
polarization plane. Thus, the results must be interfering waves are not simply spherical. At 14. J. Ullrich et al., Rep. Prog. Phys. 66, 1463–1545 (2003).
symmetric. To search for possible shifts of the the high photon energy used here, the atomic 15. T. Weber et al., Nature 431, 437–440 (2004).
interference fringes resulting from birth time nondipole effect tilts the electron angular 16. M. S. Schöffler et al., Phys. Rev. A 78, 013414 (2008).
delays, we inspected the interference pattern distributions from each center slightly in the 17. C. Siedschlag, T. Pattard, J. Phys. B 38, 2297–2310
of the fast electron for different angles b be- forward direction with respect to the photon
tween the photon propagation direction and propagation (21). This fact led to an additional (2005).
the molecular axis in Fig. 3 for the subset S angular shift of the interference pattern that 18. D. A. Horner et al., Phys. Rev. Lett. 101, 183002 (2008).
(see Fig. 2B). Figure 3A shows the measured slightly increased the slope of the red line com- 19. The emission of the second electron is mediated through
fringes for molecules aligned parallel to the pared with the blue one. The red line obtained
light propagation direction (see fig. S3 for a from considering molecular photoionization in the Coulomb interaction and is therefore subject to the
corresponding polar plot). The zeroth-order the time domain was in line with the prediction retardation of the electromagnetic field. This possible
maximum of the distribution was displaced to of calculations of molecular photoionization in additional delay of the double ionization process is
the right, which suggested the existence of a the frequency domain if nondipole effects are symmetric with respect to the inversion of the two centers
birth time delay. To confirm this assumption, included in full (22). This model is in reason- of the molecule. Therefore, our interferometric technique
we depicted the interference fringes as func- able agreement with the experimental results, is blind to this absolute delay.
tion of cos(b) in Fig. 3B. The histogram shows but more theoretical work including electron- 20. G. L. Yudin, S. Chelkowski, A. D. Bandrauk, J. Phys. B 39,
a clear dependence of the central fringe on the electron correlation is needed to clarify the L17–L24 (2006).
photon direction. For a quantitative analysis, deviation. 21. S. Grundmann et al., Phys. Rev. Lett. 124, 233201 (2020).
we determined cos(a0)—i.e., the angular posi- 22. S. Chelkowski, A. D. Bandrauk, Phys. Rev. A 97, 053401
tion of the central maximum—for each row of We have shown that the birth of a photo- (2018).
the histogram in Fig. 3B through a Gaussian fit electron wave from a molecular orbital did not 23. S. Grundmann, Zeptosecond Birth Time Delay in
(red curve in Fig. 3A). Figure 3C shows the re- occur simultaneously across the molecule. With Molecular Photoionization, version v1, Zenodo (2020);
sults of these fits and the corresponding birth an electron interferometric technique, we ob- http://doi.org/10.5281/zenodo.3899920.
time delays (according to Eq. 2) on the right served the resulting birth time delay, which
vertical scale. The birth time delay might be was imprinted as a phase difference between ACKNOWLEDGMENTS
interpreted as a nondipole Wigner delay be- the parts of the wave launching from the two
tween photoionization from different locations sides of the H2 molecule. The observed effect We acknowledge DESY (Hamburg, Germany), a member of
of the spatially extended molecular orbital. is general and does not only alter molecular the Helmholtz Association HGF, for the provision of experimental
However, the usual Wigner times—treated photoionization, but it is also expected to be facilities. Parts of this research were carried out at PETRA III,
entirely within the dipole approximation so relevant for electron emission from solids and and we thank J. Seltmann and K. Bagschik for excellent support
liquids. during the beam time. S.G. is grateful for discussions on the
topic with A. Kheifets and H. Kremer. Funding: We acknowledge
support by BMBF and by DFG. K.F. acknowledges support
by the German National Merit Foundation. Author contributions:
M.S.S., T.J., K.F., N.S., A.P., L.K., S.E., M.W., S.G., D.T., L.Ph.H.S.,
R.D., and F.T. designed, prepared, and performed the experiment.
S.G. performed the data analysis. S.G. and M.K. created the figures.
All authors contributed to the manuscript. Competing interests:
None declared. Data and materials availability: Data presented in
this study are available on Zenodo (23).
SUPPLEMENTARY MATERIALS
science.sciencemag.org/content/370/6514/339/suppl/DC1
Materials and Methods
Supplementary Text
Figs. S1 to S3
References (24–26)
28 April 2020; accepted 26 August 2020
10.1126/science.abb9318
Grundmann et al., Science 370, 339–341 (2020) 16 October 2020 3 of 3
RESEARCH
THERMOGALVANICS and S3). Relative to Fe(CN)63–, Fe(CN)64– has
a higher charge density and takes part in
Thermosensitive crystallization–boosted liquid stronger interactions with Gdm+ (18). If Gdm+
thermocells for low-grade heat harvesting is added to the LTC system, Fe(CN)64– crystal-
lizes on the cold side (top) and then sponta-
Boyang Yu1*, Jiangjiang Duan1*, Hengjiang Cong2, Wenke Xie1, Rong Liu1, Xinyan Zhuang1, Hui Wang1,
Bei Qi1, Ming Xu3, Zhong Lin Wang4, Jun Zhou1† neously precipitates and redissolves on the hot
Low-grade heat (below 373 kelvin) is abundant and ubiquitous but is mostly wasted because present side (bottom). As a result, a low local concen-
recovery technologies are not cost-effective. The liquid-state thermocell (LTC), an inexpensive and tration of Fe(CN)64– near the cold electrode en-
scalable thermoelectric device, may be commercially viable for harvesting low-grade heat energy if its hances the reduction reaction of Fe(CN)63– →
Carnot-relative efficiency (hr) reaches ~5%, which is a challenging metric to achieve experimentally. We Fe(CN)64–, and a high local concentration of
used a thermosensitive crystallization and dissolution process to induce a persistent concentration Fe(CN)64– near the hot electrode also en-
gradient of redox ions, a highly enhanced Seebeck coefficient (~3.73 millivolts per kelvin), and hances the oxidation reaction of Fe(CN)64– →
suppressed thermal conductivity in LTCs. As a result, we achieved a high hr of 11.1% for LTCs near room Fe(CN)63–. We term this system a thermosen-
temperature. Our device demonstration offers promise for cost-effective low-grade heat harvesting. sitive crystallization–boosted LTC (TC-LTC).
A vast amount of low-grade heat (below of our LTC therefore substantially decreases We directly measured and simulated the con-
373 K) is distributed in industrial pro- centration ratio profile of Fe(CN)64–/Fe(CN)63–
cesses (waste heat), the environment and is comparable to that of current power in the TC-LTC under various temperature dif-
generation technologies, indicating its potential ferences (Fig. 1C and figs. S4 and S5). Fe(CN)64–
(solar-thermal and geothermal energy), is almost completely crystallized on the cold
and the human body. Because present- for inexpensive and highly efficient harvesting side (293 K), resulting in a low [Fe(CN)64–/
of low-grade wasted heat energy. Fe(CN)63–]cold ratio of ~0.02. In contrast, the
day energy recovery technologies are not cost- crystals rapidly dissolve with the increase in
effective, this heat is wasted (1). Thermoelectric The pristine LTC consists of two electrodes
devices convert heat to energy without moving and electrolyte containing a redox couple (Fig. temperature on the hot side, boosting the
[Fe(CN)64–/Fe(CN)63–]hot ratio to ~0.94 at 343 K.
parts, audible noise, or greenhouse emissions 1A). The electrodes and electrolyte are readily Therefore, a huge DCr is built between the two
(2). The performance of thermoelectric devices electrodes, which increases with increasing
is estimated by a device figure of merit (Z = available and do not require complex manu-
Se2s/k), which depends on the Seebeck co- facturing. The 0.4 M K3Fe(CN)6/K4Fe(CN)6 temperature difference (DT). Because Se is
efficient (Se), electrical conductivity (s), and aqueous electrolyte with Se ~ 1.4 mV K–1 has synergistically driven by DS and DCr, the
thermal conductivity (k). Among thermo- been studied extensively as the benchmark voltage output for the TC-LTC is intensively
electrics, traditional solid-state thermoelec- for LTCs (13–22). When a temperature gradient
tric cells are the most intensively studied. is built between two electrodes, the balance of enhanced (Fig. 1D and fig. S6). The maximum
Se value of 3.73 mV K–1 is more than 2.5 times
However, their efficiency near room temper- reversible reactions between the redox couple that of the LTC (1.4 mV K–1). Furthermore, the
ature has only modestly progressed because of is broken, and the reactions at the electrodes
the well-known strong interdependence of Se, Se of the TC-LTC is orientation-dependent and
s, and k in solid thermoelectric materials. This will incline to the opposite direction, thus achieves the maximum value when the TC-
is especially problematic for low-cost materials causing a potential difference. According to
free of rare elements (3–10). Alternatively, our theoretical analysis (23), the potential dif- LTC is laid with the cold electrode above the
liquid-state thermocells (LTCs) offer more ways ference under a temperature difference (namely
to decouple the interdependence of Se, s, and Se) is associated with the solvent-dependent hot electrode (fig. S7), further proving that
k (11). Furthermore, LTCs are inexpensive, entropy difference (DS) between the redox
scalable, and potentially commercially viable anions (24) and the concentration ratio dif- thermosensitive crystallization, precipitation,
ference (DCr) between the hot and cold sides
for harvesting low-grade heat energy if the of the LTC. and dissolution processes induce a persistent
Carnot-relative efficiency (hr) reaches ~5%
(12). That value has been challenging to In general, the concentration gradient is concentration gradient and thus highly en-
achieve even with ideal laboratory devices
(13). We achieved an hr of 11.1% for LTCs near thermodynamically unstable and spontane- hance the Se of thermocells. In addition, we
room temperature by using a thermosensitive ously decays into a homogeneous and ther- verified that the precipitates do not directly
crystallization process that highly enhances modynamically stable state (Fig. 1A and fig. react on the bottom electrode (fig. S8), and the
Se and suppresses k without scarifying s S1). Namely, DCr is equal to zero at stable state
synergistically. The cost-performance metric in the 0.4 M K3Fe(CN)6/K4Fe(CN)6 aqueous crystallization dissolution kinetics of precip-
electrolyte. Therefore, the Seebeck effect for
1Wuhan National Laboratory for Optoelectronics, Huazhong the LTC is only driven by DS. Specifically, itates are quick enough to ensure the continu-
University of Science and Technology, Wuhan 430074, Fe(CN)64– with a small solvation entropy is
China. 2College of Chemistry and Molecular Science, spontaneously oxidized into Fe(CN)63– with a ous redox reaction in TC-LTC. The high voltage
Engineering Research Center of Organosilicon Compounds high solvation entropy by releasing an elec-
and Materials, Ministry of Education, Wuhan University, and current output can therefore hold constant
Wuhan 430072, China. 3School of Materials Science and tron to the hot electrode, and the electron
Engineering, Huazhong University of Science and Technology, through an external circuit is consumed at the during continuous operation for ~100 hours
Wuhan 430074, China. 4Beijing Institute of Nanoenergy and cold cathode by the reduction of Fe(CN)63– to
Nanosystems, Chinese Academy of Sciences, Beijing, China. Fe(CN)64– (Fig. 1A). (fig. S9). We also investigated the optimized
*These authors contributed equally to this work. addition amount of Gdm+ at which the largest
†Corresponding author. Email: [email protected] In contrast to this paradigm, we achieved DCr is achieved, which is ~3 mol liter–1 (Fig. 1E
and fig. S10). If excess Gdm+ is added, DCr
the Seebeck effect enhancement driven by declines as a result of the accompanying crys-
both DS and DCr by using guanidinium tallization of Fe(CN)63– (fig. S11).
cations (Gdm+) to selectively induce Fe(CN)64–
crystallization (Fig. 1, A and B, and figs. S2 To achieve the largest enhancement of the
Seebeck effect, the cation additives need to
meet two characteristics: (i) the strong ability
to induce crystallization of Fe(CN)64–, and (ii)
the high thermosensitive solubility of Fe(CN)64–-
associated crystals (Fig. 2A). Small monovalent
cations, such as Li+, Na+, K+, and NH4+ (strongly
hydrated cations) (25, 26), cannot induce crys-
tallization of Fe(CN)64– (Fig. 2C) and thus produce
no enhancement effect on Se. Divalent cations,
such as Ca2+ and Mg2+ (strong cations) (27), can
Yu et al., Science 370, 342–346 (2020) 16 October 2020 1 of 5
RESEARCH | REPORT
A e− Cold cathode e− e− e− Cold cathode tively (Fig. 2D and fig. S15). According to
e− Crystallization thermodynamic theory, DG = DH – TDS; that
e− Guanidinium is, the thermosensitivity of the crystal is as-
Inducing sociated with its enthalpy change (DH) and
Low voltage High voltage entropy change (DS) during the dissolution
crystallization process. In general, crystals with a small DH
and a large DS will achieve a substantial
Fe(CN)64− e− Dissolution Gibbs free energy decrease (DG) at a small
Fe(CN)63− e− increase of the temperature and possess a
e− Crystal high thermosensitivity. The crystal induced
by Gdm+ contains the most hydrated water
B e− Hot anode e− e− Hot anode molecules and possesses the highest struc-
tural complexity. The highly hydrated crystal is
Liquid-state thermocell Thermosensitive crystallization–boosted LTC loose and has low lattice energy (28). The cor-
(LTC) (TC-LTC) responding DH during the dissolution process
is small. The highly complex crystal will also
C 1.0 Cold electrode (293 K) cause a large DS during the dissolution process.
Hot electrode
0.8 We measured the thermoelectric performance
[Fe(CN) 4 ]/[Fe(CN) 3 ] of the TC-LTC with a typical planar cell (Fig. 3A).
66 0.6 In brief, a plastic cell is filled with an electrolyte
containing crystals and sealed first with com-
Inducing 0.4 mercial carbon fabric paper (thickness ~600 mm)
crystallization and then with graphite plates, which serve as
the electrodes. The porous carbon fabric paper,
K3Fe(CN)6 Fe(CN)64 - 0.2 having high specific surface area, enhances the
/K4Fe(CN)6 associated current density (13–15, 19). The temperature
electrolyte 0.0 gradient in the TC-LTC is controlled by an
crystals 0 electrical heating plate (on the bottom) and
a water-cooled plate (on the top). In the steady
10 20 30 40 50 state, crystals precipitate to the bottom, and the
T (K) residual transparent electrolyte is on the top
(Fig. 3A). Owing to the crystallization-induced
D0 E 4 enhancement of the Seebeck effect, the elec-
trical output for the TC-LTC is much higher
-50 1.0 than that for the pristine LTC (Fig. 3B and fig.
S16). The maximum power density (Pmax) for
1.4 mV K 1 [Fe(CN) 4 ]/[Fe(CN) 3 ] 0.8 the TC-LTC is 17.7 W m–2 at a DT of 50 K, which
66 3 is more than five times the Pmax of the pristine
V (mV) S (mV K 1) LTC (3.1 W m–2). The enhancement effect is
oc 0.6 e further verified by using other electrode ma-
terials including graphite sheet, pin graphite
-100 sheet, graphite felt, and carbon cloth in the
LTC and TC-LTC, which demonstrates the
-150 LTC 0.4 2 universality of our strategy (fig. S17). Addition-
TC-LTC 0.2 ally, the thick precipitate layer in the TC-LTC
-200 Simulation clearly suppresses thermal convection of the
0 3.73 mV K 1 0.0 liquid electrolyte (20, 29). The effective thermal
0 conductivity (keff) for the TC-LTC is therefore
10 20 30 40 50 1234 1 substantially lower than that of the pristine
T (K) GdmCl (mol L 1) 5 LTC (Fig. 3C and fig. S18). Thermal convection
in the pristine LTC intensifies at a higher tem-
Fig. 1. Crystallization-inducing enhancement of the Seebeck effect in the TC-LTC. (A) Schematic of perature, and the corresponding keff increases
guanidinium cations (Gdm+) inducing Fe(CN)64– crystallization and enhancement of the Seebeck effect in the from 0.67 to 1.64 W m–1 K–1. By contrast, the
keff for the TC-LTC is maintained at ~0.4 W
0.4 M K3Fe(CN)6/K4Fe(CN)6 system. (B) Photographs of the 0.4 M K3Fe(CN)6/K4Fe(CN)6 electrolyte before and m–1 K–1 over the tested temperature range. In
after the addition of Gdm+. The diameter of each bottle is 2.7 cm. (C) Fe(CN)64–/Fe(CN)63– concentration ratio at the addition, a temperature difference of ~50 K in
the TC-LTC holds constant during a contin-
cold and hot electrodes with increasing temperature difference DT. The temperature of the cold electrode was uous operation of ~10 hours (fig. S19), which
indicates that the low thermal conductivity of
controlled at 293 K. (D) Open-circuit voltage (Voc) of the LTC and TC-LTC at different values of DT. The simulated TC-LTC is stable for long-term operation. In
contrast to the traditional solid-state thermo-
result (dashed line) is consistent with the experimental result. The Seebeck coefficient (Se) is calculated from the electric cells (TECs), the electrical conductivity
slope of the Voc-DT curves. (E) Fe(CN)64–/Fe(CN)63– concentration ratio in the 0.4 M K3Fe(CN)6/K4Fe(CN)6 electrolyte for the LTC/TC-LTC is difficult to directly mea-
(293 K) and corresponding Se values with the addition of Gdm+ at different concentrations. Error bars in (C) to (E) sure because of its complex dependence on the
denote SD from repeated measurements for three times at the same temperatures and Gdm+ concentrations.
induce crystallization of Fe(CN)64– (Fig. 2C), S12, A to C, and S14). Among them, the crystals
but their crystals have little temperature- formed by Gdm+ exhibited the highest ther-
dependent solubility, at least in the temperature mosensitivity, and the corresponding TC-LTC
system achieved the highest Se (Fig. 2, A and B,
range we are interested in (Fig. 2B and figs. S12, and fig. S14). To further explain the thermo-
D and E, and S13). As a result, they also have sensitivity difference of crystals induced by
no enhancement effect on Se (fig. S14). Only different cations, we analyzed the chemical
large monovalent cations, such as Gdm+,
tetraethylammonium (Tea+), and 1-ethyl-3- structures of three typical crystals induced by
methylimidazolium (Emim+)—weak cations, Gdm+, Emim+, and Ca2+ with single-crystal x-ray
weakly hydrated—possess the two character-
istics and enable the enhancement of Se (figs. diffraction; these are K2[C(NH2)3]2Fe(CN)6·6H2O,
K4Fe(CN)6·3H2O, and K2CaFe(CN)6, respec-
Yu et al., Science 370, 342–346 (2020) 16 October 2020 2 of 5
RESEARCH | REPORT
Fig. 2. Various additive-induced enhancements of A Crystallization B 10
the Seebeck effect in the TC-LTC. (A) Optimized 4 None
8 Ca2+
Se for the TC-LTC induced by different cation Fe(CN)64− 6Normalized solubility Emim+
additives. According to their ability to induce Gdm+
crystallization of Fe(CN)64–, cation additives are Thermosensitivity
classified into two categories: Dark cyan indicates 3 Sensitive
Se (mV K−1) Slightly Sensitive
that no crystallization was observed, whereas red 2 Insensitive 4
indicates crystallization of Fe(CN)64–. Furthermore,
1 2
according to the temperature-dependent solubility
(thermosensitivity) of Fe(CN)64–-associated crystals, 0 0
293 303 313 323 333 343
cation additives are classified as “sensitive” (solid Temperature (K)
star), “slightly sensitive” (half-solid star), and Blank K+ NHLi ++ + Mg 2+ Ca 2+ D
Na + + +
“insensitive” (open star). (B) Change in the normal- Gd 4m ETmieam
ized solubility of Fe(CN)64–-associated crystals C
induced by Ca2+, Emim+, and Gdm+ with increasing
temperature. (C) Photographs of the 0.4 M K3Fe(CN)6/ H
K4Fe(CN)6 electrolyte with different additives. The C
diameter of each tube is 1.3 cm. (D) Chemical N
O
structure of the crystal K2[C(NH2)3]2Fe(CN)6·6H2O K
induced by Gdm+. Error bars in (A) and (B) denote Fe
SD from repeated measurements for three times at
the same temperatures and cation additive concentrations.
A Electrolyte B 20 LTC C LTC
TC-LTC TC-LTC
Cooler 15 1.6
P (W m 2) 1.2
max (W m 1 K 1)
10
eff
0.8
5
Thermocouples Precipitate Electrodes 0.4
Heater 0 20 30 40 50 298 303 308 313 318
10 T (K) Temperature (K)
D 12 LTC E 12 Aqueous electrolytes This work F LTC
TC-LTC TC-LTC
9 Ref. 18 Ref. 32 0.4
6
3 Ref. 31 Ref. 21 0.3
9 Ionic liquid electrolytes 0.2
Ref. 33 Ref. 34
0.1
(%) (%) Optimized electrodes ZT
Ref. 13 Ref. 19 Ref. 14
r r
6 Ref. 15 Ref. 35 Ref. 30
Commercialization threshold
3
0 20 30 40 50 0 2.1 2.8 3.5 4.2 0.0 303 308 313 318
10 T (K) 1.4 S (mV K 1) 298 Temperature (K)
e
Fig. 3. Thermoelectric performance of the TC-LTC. (A) Photograph of a at different DT values, calculated according to Eqs. S2 to S4 (23).
single planar TC-LTC cell. The cross-sectional area of the cell is 2.6 cm2, and (E) Comparison of hr and Se values for various LTC systems obtained by
electrolyte and electrode optimization (table S2); a dashed line shows
the distance between the two electrodes is 1.5 cm. (B) Maximum power the predicted commercialization threshold, ~5% (12). (F) Comparison of the
density (Pmax) of the LTC and TC-LTC at different values of DT (the cold side figure of merit (ZT) of the LTC and TC-LTC at different temperatures.
is controlled at 293 K). Pmax values are the average values calculated from The temperature used for ZT is the average value of the cold-side
current-voltage curves (fig. S16). (C) Effective thermal conductivity (keff) temperature (controlled at 293 K) and hot-side temperature (dependent on
of the LTC and TC-LTC at different temperatures (statistical average DT). In (B), (C), (D), and (F), error bars denote SD from repeated
measurements for five times at the same temperatures.
values from the infrared images), measured using a steady-state method
(fig. S18). (D) Carnot-relative efficiency (hr) of the LTC and TC-LTC
Yu et al., Science 370, 342–346 (2020) 16 October 2020 3 of 5
RESEARCH | REPORT B4 100 C 100
80
A 3 120 60
2 75 40
TC-LTC module 1 20
90 0
D Powering electric fan 3
50
60
25
30
Voltage (V)
T (K)
Current (mA)
Power (mW)
00 0 12
0 100 200 300 400 500 0 Voltage (V)
Time (s)
E Powering LED array F Powering thermohydrometer G Charging mobile phone
Passive
voltage
booster
Cooler TC-LTC
module
Heater
Fig. 4. Electricity generation and demonstration of using a TC-LTC module to power electronic devices. (A) Photograph of a TC-LTC module containing
20 units in series. (B) Real-time voltage curves (black) of the module with an increase in DT (red). (C) Current-voltage curve (black) and corresponding power
output (red) at DT = 50 K. (D to G) The module can directly power various electronic devices, including (D) an electric fan, (E) a LED array, (F) a thermohydrometer,
and (G) a mobile phone using a passive voltage booster.
ionic conductivity of the electrolyte, electrical ating a useful voltage (>1 V) under a small devices including an electric fan, an LED array,
conductivity of the electrodes, and reaction and a thermohydrometer (Fig. 4, D to F, and
kinetics of the redox species at the electrodes temperature difference requires a challenging movie S1). Furthermore, the module enabled
(30). Therefore, a more comprehensive param- a smart mobile phone to be charged through
eter, the effective electrical conductivity (seff), integration of thousands or even tens of integration with a passive voltage booster
is used for the LTC/TC-LTC. This value is cal- thousands of units (10). Because of the high (Fig. 4G and movie S1).
culated from the slope of the current-voltage Se of 3.73 mV K–1, integrating fewer TC-LTCs
curve (12, 13). The seff of the TC-LTC increases enables a considerable voltage output, which Unlike previous LTC strategies that enhanced
from 23 to 36 S m–1 with increasing temper- Se but sacrificed seff (31), our thermosensitive
ature, which is nearly consistent with that of is beneficial to enhance the fill factor and re- crystallization–based TC-LTC strategy syner-
the pristine LTC (fig. S20). Overall, the strategy gistically enhances Se and reduces keff without
of crystallization in the TC-LTC synergisti- duce the manufacturing costs for the thermo- sacrificing seff. We achieved attractive single-
cally enhances Se and reduces keff without electric devices (36). To evaluate the potential cell performance for Carnot-relative efficiency
sacrificing seff. for application to low-grade heat harvesting, (11.1%) and current density (416 A m–2). We
obtained a high power density (6.86 W m–2)
The conversion efficiency is one important we estimated the cost-performance metric (CPM) for our module at a temperature difference
standard for evaluating thermoelectric devices. of 50 K (fig. S24 and tables S4 and S5). These
Owing to the high power density (electricity of various thermoelectric systems based on examples show that LTCs may be a viable
output) and low thermal conductivity (heat technology for harvesting low-grade waste
input), the highest hr for the TC-LTC is 11.1% their raw material prices (table S3 and figs. S21 heat. Our approach can be expanded to other
at a DT of 40 K, which is 18.5 times that of the LTC systems by using thermosensitive crystals
pristine LTC (Fig. 3D). In addition, the hr for and S22). Relative to inorganic solid-state ther- of corresponding redox species, as well as other
the TC-LTC substantially surpasses that for thermal energy–harvesting systems, including
other optimized LTC systems in the literature moelectric cells (ITECs) and organic solid- thermally regenerative electrochemical cycles
(13–15, 18, 19, 21, 30–35) and the predicted (37–39) and direct thermal charging cells (40),
commercialization threshold (~5%) (12) (Fig. state thermoelectric cells (OTECs), our TC-LTC which also require a redox system of high Se.
3E and table S2). Furthermore, the dimen-
sionless figure of merit (ZT) of ~0.4 for the system may be more cost-effective; its esti- REFERENCES AND NOTES
TC-LTC at 318 K is high relative to other mated CPM (~$1.56 W–1) is closer to those of
LTCs and comparable to that of most TEC 1. C. Forman, I. K. Muritala, R. Pardemann, B. Meyer, Renew.
systems (Fig. 3F and table S3). In addition, commercial power generation technologies Sustain. Energy Rev. 57, 1568–1579 (2016).
for a typical thermoelectric module, gener- (nuclear, $5.34 W–1; coal, $2.84 W–1; natural
gas, $0.98 W–1) (9). 2. J. He, T. M. Tritt, Science 357, 1369 (2017).
3. J. P. Heremans, Nature 508, 327–328 (2014).
Finally, we designed a TC-LTC module by
connecting 20 units (size ~1.4 cm × 5 cm × 2 cm)
in series to demonstrate viability for scale-up
(Fig. 4A and fig. S23). The module generated
a Voc of 3.1 V and a short-circuit current (Isc)
of 120 mA, and the corresponding Pmax was
96 mW under a DT of 50 K between the heater
and cooler plates (Fig. 4, B and C). Because of
the considerable power output we achieved, the
module could directly drive various electronic
Yu et al., Science 370, 342–346 (2020) 16 October 2020 4 of 5
RESEARCH | REPORT
4. R. Venkatasubramanian, E. Siivola, T. Colpitts, B. O’Quinn, 26. H. Ohtaki, T. Radnai, Chem. Rev. 93, 1157–1204 (1993). Science and Technology for their support. Funding: Supported by
Nature 413, 597–602 (2001). 27. Y. Marcus, G. Hefter, Chem. Rev. 106, 4585–4621 (2006). National Natural Science Foundation of China grant 51672097, the
28. J. Sohr, H. Schmidt, W. Voigt, Acta Crystallogr. C 74, 194–202 National Program for Support of Topnotch Young Professionals,
5. S. I. Kim et al., Science 348, 109–114 (2015). the program for HUST Academic Frontier Youth Team
6. B. Poudel et al., Science 320, 634–638 (2008). (2018). (2017QYTD11), and the Director Fund of WNLO (J.Z.). Author
7. J. Mao et al., Science 365, 495–498 (2019). 29. P. Yang et al., Angew. Chem. Int. Ed. 55, 12050–12053 (2016). contributions: J.Z. conceived the research; J.D., B.Y., and J.Z.
8. W. He et al., Science 365, 1418–1424 (2019). 30. T. J. Kang et al., Adv. Funct. Mater. 22, 477–489 (2012). designed the experiments; B.Y. and J.D. carried out the
9. S. LeBlanc, S. K. Yee, M. L. Scullin, C. Dames, K. E. Goodson, 31. H. Zhou, T. Yamada, N. Kimizuka, J. Am. Chem. Soc. 138, experiments; H.J.C. conducted the single-crystal x-ray
characterization; B.Y., J.D., H.C., W.X., R.L., X.Z., H.W., B.Q., M.X.,
Renew. Sustain. Energy Rev. 32, 313–327 (2014). 10502–10507 (2016). Z.L.W., and J.Z. analyzed the data; and J.Z., J.D., B.Y., Z.L.W., and
10. S. K. Yee, S. LeBlanc, K. E. Goodson, C. Dames, Energy Environ. 32. J. H. Kim et al., Sci. Rep. 9, 8706 (2019). M.X. wrote the paper. All authors discussed the results and
33. P. F. Salazar, S. T. Stephens, A. H. Kazim, J. M. Pringle, approved the final version of the manuscript. Competing
Sci. 6, 2561–2571 (2013). interests: The authors declare no competing interests. Data and
11. M. F. Dupont, D. R. MacFarlane, J. M. Pringle, Chem. Commun. B. A. Cola, J. Mater. Chem. A 2, 20676–20682 (2014). materials availability: All data are available in the manuscript or
34. K. Kim, H. Lee, Phys. Chem. Chem. Phys. 20, 23433–23440 the supplementary materials.
53, 6288–6302 (2017).
12. T. I. Quickenden, Y. Mua, J. Electrochem. Soc. 142, 3985–3994 (1995). (2018). SUPPLEMENTARY MATERIALS
13. H. Im et al., Nat. Commun. 7, 10600 (2016). 35. W. Qian, M. Cao, F. Xie, C. Dong, Nano-Micro Lett. 8, 240–246
14. L. Zhang et al., Adv. Mater. 29, e1605652 (2017). science.sciencemag.org/content/370/6514/342/suppl/DC1
15. M. S. Romano et al., Adv. Mater. 25, 6602–6606 (2013). (2016). Materials and Methods
16. Y. Mua, T. I. Quickenden, J. Electrochem. Soc. 143, 2558–2564 36. G. J. Snyder, A. H. Snyder, Energy Environ. Sci. 10, 2280–2283 Supplementary Text
Figs. S1 to S24
(1996). (2017). Tables S1 to S5
17. C. G. Han et al., Science 368, 1091–1098 (2020). 37. S. W. Lee et al., Nat. Commun. 5, 3942 (2014). Movie S1
18. J. Duan et al., Nat. Commun. 9, 5146 (2018). 38. Y. Yang et al., Proc. Natl. Acad. Sci. U.S.A. 111, 17011–17016 (2014). References (41–65)
19. R. Hu et al., Nano Lett. 10, 838–846 (2010). 39. Y. Ding et al., Energy Environ. Sci. 12, 3370–3379 (2019).
20. L. Jin, G. W. Greene, D. R. MacFarlane, J. M. Pringle, ACS 40. X. Wang et al., Nat. Commun. 10, 4151 (2019). 5 July 2020; accepted 25 August 2020
Published online 10 September 2020
Energy Lett. 1, 654–658 (2016). ACKNOWLEDGMENTS 10.1126/science.abd6749
21. G. Li et al., Adv. Mater. 31, e1901403 (2019).
22. T. Kim et al., Nano Energy 31, 160–167 (2017). We thank the reviewers whose comments have helped us to
23. See supplementary materials. improve the manuscript significantly; N. S. Xu, S. Z. Deng,
H. X. Deng, J. Tang, Y. Z. Pei, and J. Chen for their help; and the
24. S. Sahami, M. J. Weaver, J. Electroanal. Chem. Interfacial Center for Nanoscale Characterization and Devices, WNLO-HUST,
Electrochem. 122, 155–170 (1981). and the Analysis and Testing Center of Huazhong University of
25. Y. Marcus, Chem. Rev. 109, 1346–1370 (2009).
Yu et al., Science 370, 342–346 (2020) 16 October 2020 5 of 5
RESEARCH
ECOLOGY genetically distinct, reproductively isolated,
and analogous to species. Additionally, within
Species richness and redundancy promote each mutualist type, the strains have small dif-
persistence of exploited mutualisms in yeast ferences in genotype and phenotype (e.g., yield)
(Fig. 1, B and C), making them analogous to
Mayra C. Vidal1,2*, Sheng Pei Wang1, David M. Rivers3, David M. Althoff1*, Kari A. Segraves1* closely related, ecologically similar species that
would compete strongly with one another be-
Mutualisms, or reciprocally beneficial interspecific interactions, constitute the foundation of many cause of niche overlap.
ecological communities and agricultural systems. Mutualisms come in different forms, from pairwise
interactions to extremely diverse communities, and they are continually challenged with exploitation We also engineered cheater strains that
by nonmutualistic community members (exploiters). Thus, understanding how mutualisms persist provide no resources but consume either
remains an essential question in ecology. Theory suggests that high species richness and functional adenine (hereafter, “adenine cheater”) or lysine
redundancy could promote mutualism persistence in complex mutualistic communities. Using a yeast (“lysine cheater”). We call them cheaters be-
system (Saccharomyces cerevisiae), we experimentally show that communities with the greatest cause they are derived from the mutualists,
mutualist richness and functional redundancy are nearly two times more likely to survive exploitation simulating cheaters that share recent evolu-
than are simple communities. Persistence increased because diverse communities were better able tionary history with mutualists. Because no
to mitigate the negative effects of competition with exploiters. Thus, large mutualistic networks may lysine or adenine is available in the medium
be inherently buffered from exploitation. except for that released by the mutualists, the
cheaters cannot exist independently of the
M utualist communities are prevalent ness could potentially enhance mutualism mutualists, and community persistence criti-
in every ecosystem (1–3), forming the persistence under exploitation because rich cally depends on the presence of both types of
core of food webs and providing crit- communities have redundant species with overproducing mutualists. Thus, when one
ical ecosystem services. Like other similar functional roles (4, 5). Consequently, or both mutualist types went extinct, we
communities, mutualist communities if a mutualist goes extinct after exploitation, considered these communities as having
must be able to cope with constantly chang- the community can persist with fewer species collapsed. To test the hypothesis that species
ing conditions, but the factors that help main- because the remaining redundant mutualist richness enhances mutualism persistence,
tain their stability remain under debate (4–6). species still provide the commodities to sustain we created symmetrical communities that
Recent efforts to understand mutualistic com- the community (15, 16). Simultaneously, how- varied in richness with and without the two
munity dynamics by using network analysis ever, redundant species have similar niches types of cheaters (Fig. 1A). Communities were
suggest that high species richness could en- and may compete strongly with one another grown for 4 weeks, and we assessed their
hance persistence (4); however, experimen- for mutualistic commodities and/or other re- survival and community composition weekly
tal validations of this hypothesis are needed. sources. Theory suggests that coexistence of (1679 communities) (18).
Understanding the persistence of mutualist redundant mutualists is hindered by com-
communities is paramount for the manage- petition for the mutualistic commodity (17). Persistence of mutualistic communities was
ment and conservation of ecosystems (7), Competition could lead to removal of inferior highly dependent on community composition.
especially given the risk of species loss with competitors (17) and decreased species rich- All mutualist-only communities survived the
climate change (8). ness over time (5), and influence mutualism entire experiment. Communities with cheaters,
persistence. Considering these contrasting ef- however, went extinct at different rates depend-
To persist, mutualisms need to resist exploi- fects of species redundancy on communities, ing on the type of cheater. The lysine cheater
tation by organisms that use the exchanged in this study we experimentally test how led to the collapse of 55% of the mutualistic
commodities of the mutualism without provid- mutualist species richness and functional re- communities, whereas the adenine cheater
ing anything in return (9). These exploiters can dundancy contribute to mutualism persistence caused <5% collapse (Fig. 2A), thus demonstrat-
be unrelated to the mutualists or they can be with and without cheaters. ing that the effect of cheaters on mutualism
mutualistic species or individuals that have persistence is context dependent. This con-
defected from the mutualism (“cheaters”) (10). We created a synthetic mutualism using text dependency may help explain why some
Although there is debate about whether exploi- brewer’s yeast, Saccharomyces cerevisiae, by cheaters, but not others, can have strong nega-
tation has strong negative fitness consequences engineering asexual strains to overproduce tive effects on natural mutualistic commun-
in many mutualisms (11), exploitation can change either lysine or adenine but not produce the ities. The negative effect of the lysine cheater
the structure of communities (3, 12, 13), leading other resource (Fig. 1A) (18). Adenine is re- on community persistence was buffered by
to local species loss (14). Despite the possible quired for cell division and lysine for cell growth, the higher initial number of mutualist strains
negative effects of exploitation, exploiters or making these nutrients essential for yeast in the community, as persistence rates nearly
cheaters are present in virtually all mutualistic fitness. Because the overproduced nutrients doubled in the richest communities (Fig. 2A).
communities; thus, how mutualistic commu- are released into the medium and are freely Because the richest communities were not
nities are buffered from the effects of exploita- available, the mutualism cannot involve sanctions independently replicated, as the mutualist
tion is unclear. or partner choice, which are mechanisms used strains were sampled from a pool of eight total
in some mutualisms to restrict cheating, for strains, we also tested for changes in persistence
Similar to the proposed effect of species example, by controlling the amount of com- by excluding the eight-strain communities,
richness on mutualism persistence (4), rich- modities exchanged or by avoiding interactions thus eliminating communities that were not
(19). Thus, this mutualism is similar to common, independently replicated. This analysis con-
1Department of Biology, Syracuse University, Syracuse, NY diffuse mutualisms such as many generalized firmed that species richness buffers mutualist
13210, USA. 2Biology Department, University of Massachusetts pollination systems (20). We genetically engi- communities that are experiencing substan-
Boston, Boston, MA 02125, USA. neered the strains to function ecologically tial negative effects from cheaters (c2 = 17.96,
*Corresponding author. Email: [email protected] (M.C.V.); as different species; as such, the strains are df = 2, P = 0.0001).
[email protected] (D.M.A.); [email protected] (K.A.S.)
The results suggest that species richness is
an important component of persistence with
cheaters, yet increasing richness also adds
Vidal et al., Science 370, 346–350 (2020) 16 October 2020 1 of 5
RESEARCH | REPORT
Fig. 1. Representation of symmetrical yeast communities and yield dif- types (blunt-ended dashed lines). Besides mutualist-only communities, we
ferences among strains used to build the communities. (A) Symmetrical created communities with a lysine cheater (red) that competed with the
mutualist communities. (Top left) Simplest community, with one strain of adenine mutualists for lysine and communities with an adenine cheater
adenine (Ade) mutualist (top, green) that releases adenine into the medium, (purple) that competed for adenine with the lysine mutualists. (B) Yield at
which is taken up by the lysine (Lys) mutualist (bottom, blue) that releases 24 hours of growth for the adenine overproducing mutualists (AdeOP) and the
lysine, which is used by the adenine mutualist. Although these strains lysine cheater. (C) Yield at 24 hours of growth for the lysine overproducing
are mutualists, they compete for other resources (blunt-ended line). We mutualists (LysOP) and the adenine cheater. Yield was measured when
added pairs of mutualist types to create symmetrical communities of up to strains were growing alone in complete medium. Letters represent Tukey’s
eight strains (top right) in which there was also competition within mutualist honest significant difference (HSD) comparisons.
functional redundancy of mutualists. To dis- creasing redundancy of either mutualist type, suggest a key role for functional redundancy
entangle the effects of species richness and as the communities did not experience collapse in mutualism and that having a greater number
redundancy, we compared the persistence of (Fig. 2B). Similarly, changes in redundancy of of redundant mutualist species that compete
asymmetrical communities with and without the lysine mutualists that do not compete for with a strong cheater increases the likelihood
the lysine cheater. We created mutualist-only mutualistic commodities with the lysine cheater that the mutualism will persist despite the
communities and a replicate set including the led to no change in persistence, as these com- negative effects of cheaters.
lysine cheater in which one adenine mutualist munities suffered ~65% collapse regardless
was matched with either two, three, or four of the initial number of strains. By contrast, Although we found that functional redun-
lysine mutualists, as well as the converse (1431 increases in functional redundancy of the dancy can buffer the negative effects of a
total communities) (18). adenine mutualists that compete with the strong cheater (Fig. 2B), redundancy can be
lysine cheater led to a 25% increase in com- disadvantageous as well because similar mutu-
Results from the asymmetrical communities munity persistence when we compared com- alist species should also compete strongly with
showed that the positive effect of species rich- munities with two versus three or four adenine one another (17). Our results show that starting
ness on community persistence was driven by mutualists (from 43 to 67% survival) (Fig. 2B). species richness had a negative effect on
mutualist functional redundancy, and this Thus, functional redundancy of the mutualist individual strain retention in all multimutu-
redundancy was critical when mutualists use type that directly competes with the cheater alist communities (Fig. 3). Despite that, when we
the same mutualistic commodity as the lysine for the mutualistic resource had a notable examined the final composition of surviving
cheater. For the mutualist-only communities, impact on community persistence. These results communities, we observed that coexistence
there was no change in persistence, with in- among mutualists usually occurred in at least
Vidal et al., Science 370, 346–350 (2020) 16 October 2020 2 of 5
RESEARCH | REPORT
Fig. 2. Effect of species richness and functional redundancy on
community persistence. (A) Symmetrical communities with mutualists
only and with the adenine cheater had high survival, whereas communities
with the lysine cheater had 40 to 75% survival rate, depending on the
starting number of mutualistic strains (c2 = 27.47, df = 3, P < 0.0001).
(B) Asymmetrical communities containing only mutualists had high survival.
By contrast, communities with the lysine cheater and variable numbers
of strains of the adenine mutualist (AdeOPs) increased community
persistence from ~40 to 70% as the number of adenine mutualists
increased (c2 = 25.29, df = 2, P < 0.0001; excluding the most diverse
communities that were not independently replicated: c2 = 17.78, df = 1,
P < 0.0001). Communities with the lysine cheater and increasing numbers of
lysine mutualists (LysOPs) did not differ (c2 = 2.94, df = 2, P = 0.23). Points
on graphs represent mean ± SE.
Fig. 3. Effect of richness on strain loss in persistent communities. in asymmetrical communities (F2,992 = 100.9, P < 0.0001); within each community
(A) Mutualist strain retention in symmetrical communities decreased with the type, all pairwise comparisons differed with increasing richness (P < 0.001) except
starting number of mutualistic strains (mutualists only: t = −14.47, df = 1328, for communities with increasing number of lysine mutualists (light red and light
P < 0.0001; with lysine cheater: t = −14.8, df = 1328, P < 0.0001, with blue). In these communities, there was no difference in strain loss with starting
adenine cheater: t = −23.87, df = 1328, P < 0.0001). Communities with cheaters numbers of four or five mutualists, regardless of the presence of the lysine
had greater loss than communities with mutualists only (Tukey’s test: mutualists cheater (with cheater: z = 0.52, n = 123, P = 0.859; without cheater: z = −0.69,
only versus with lysine cheater: t = 3.06, df = 1328, P = 0.006; mutualists n = 285, P = 0.765). Points on graphs (A) and (B) represent mean ± SE.
only versus with adenine cheater: t = 6.02, df = 1328, P < 0.0001; with (C) Coexistence of mutualists as a proportion of communities, showing the
lysine cheater versus with adenine cheater: t = 2, df = 1328, P = 0.11). number (No.) of strains of adenine mutualists (AdeOPs) and lysine mutualists
(B) Strain retention decreased with the starting number of mutualist strains (LysOPs) retained to the end of the experiment for each community type.
half of the communities and was even more negative effect on community survival caused more intense than for adenine because lysine
frequent among lysine mutualists and in by the lysine cheater but not the adenine cheater availability is delayed but adenine is readily
mutualist-only communities (Fig. 3C). In suggests that there is likely a difference in the available. Lysine is stored in vacuoles and is
addition, strain loss was more pronounced intensity of competition for the mutualistic released as the lysine-producing mutualists
when either cheater was present (Fig. 3), commodities being exchanged. This idea is die, whereas adenine is continuously secreted
likely because cheaters removed mutualis- further supported by our finding that strain by the adenine mutualists (21). Consequently,
tic commodities without contributing any loss was higher among redundant adenine lysine availability was nearly unmeasurable for
resources to the environment. mutualists than among redundant lysine mu- the first 48 hours, whereas adenine availability
tualists (Fig. 3B). Furthermore, in contrast to increased over time (Fig. 4A). Thus, this dif-
Together, the results suggest that competi- the adenine mutualists, the lysine mutualists ference in resource availability might be leading
tion for the shared resources among mutualists were less likely to be excluded from commun- to stronger competition for lysine than for
and between mutualists and cheaters is driving ities (Fig. 3C). Competition for lysine could be adenine, both among adenine mutualist strains
the patterns of strain loss. For instance, the
Vidal et al., Science 370, 346–350 (2020) 16 October 2020 3 of 5
RESEARCH | REPORT
Fig. 4. Mechanism of strain loss and community collapse. (A) Estimates that could not produce adenine or lysine, and the only lysine or adenine
of lysine and adenine produced by mutualists and cheaters over time (compared available was that produced by the mutualists. OD, optical density. (B) Starvation
with the first time point: 8 hours adenine mutualists and lysine cheater, resistance of adenine mutualists and the lysine cheater. The lysine cheater
12 hours lysine mutualists and adenine cheater). Production of adenine by was more resistant to 48 hours of starvation than the adenine mutualists
mutualists increases over time (F4,35 = 40.4, P < 0.0001), whereas production (Tukey’s HSD between lysine cheater and AdeOPs had P < 0.05). (C) Starvation
did not differ for cheaters and lysine mutualists (LysOPs: F2,20 = 2.5, P = 0.107; resistance of lysine mutualists and adenine cheater. All strains had similar
adenine cheater: F4,5 = 2.36, P = 0.186; lysine cheater: F2,3 = 0.71, P = 0.56). starvation resistance at 48 hours (F4,10 = 0.72, P = 0.6). Points on graphs represent
Production was measured indirectly by assessing the growth of a test strain mean ± SE.
as well as between the adenine mutualists and dominant in 25% of the communities, and alists and lysine cheater was as likely as the
the lysine cheater. these communities eventually went extinct exclusion of the cheater for most of the sur-
(table S3). By contrast, communities contain- viving communities (fig. S5). Thus, the outcome
One important competitive trait for the ing the adenine cheater shifted in favor of of competition among mutualist strains and
lysine cheater and adenine mutualists would the mutualists, and the cheater was even- between mutualists and the lysine cheater was
be starvation resistance that would allow sur- tually excluded, possibly because of a com- context dependent and was not solely predict-
vival during periods when lysine is limiting. petitive trait other than starvation resistance able on the basis of the identity of the mu-
Strains that have more individuals surviving (e.g., yield) (Fig. 1B). For the mutualist-only tualist strains in the communities. These results
a period of starvation would have higher initial communities, the ratio of lysine and adenine show that although competition is an impor-
population density when the resource becomes mutualists remained constant. Thus, differ- tant factor in all communities, competitive
available, leading to priority effects. If lysine ences in starvation resistance appear to be exclusion alone does not determine the per-
cheaters are more resistant to starvation than linked to shifts in population ratio favoring sistence of mutualist communities that are
the adenine mutualists, it could explain the the lysine cheater, ultimately resulting in com- exploited.
severity of their impact on mutualist commu- munity collapse.
nities. To test this hypothesis, we grew the strains Our results provide evidence for the fea-
alone and measured starvation resistance to Community persistence increased with rich- sibility of the coexistence of functionally
the mutualistic commodity that they require. ness regardless of strain composition (figs. S1 redundant species in multimutualistic com-
At 48 hours, 85% (±3.5 SE) of the lysine cheater and S2), suggesting that the patterns of com- munities. In non-neutral models that assume
population survived lysine starvation, whereas munity survival were not driven by the presence niche differentiation, coexistence occurs either
only 47.3% (±4.6 SE) of the adenine mutualist of competitively superior strains. As more mu- when intraspecific competition is stronger than
populations survived (Fig. 4B). In comparison, tualist strains are added, there is an increased interspecific competition, when there is a
the lysine mutualists and the adenine cheater probability that one of those strains will be trade-off between colonization and competi-
had similar starvation resistance at 48 hours competitively superior to the lysine cheater. tive abilities, or when there is spatial or tem-
(Fig. 4C); however, these strains probably do Consequently, we tested (i) whether commun- poral heterogeneity in resource availability
not starve for adenine because adenine is con- ities containing the superior adenine mutual- coupled with trade-offs in competitive abilities
tinuously released in relatively high quantities ist competitor were more likely to survive, (ii) for different resources or for environmental
(Fig. 4A). We hypothesized that the superiority whether the strongest mutualist competitor tolerances (22, 23). The mutualist species that
in starvation resistance of the lysine cheater was numerically dominant in the surviving we used in our experiments closely resemble
would cause shifts in composition ratios toward communities, and (iii) whether coexistence one another and were growing together in a
the lysine cheater, as would be expected under with the cheater was rare. These tests showed mixed, homogenous, closed environment. These
a model of priority effects. To test this, we that survival and abundance of different adenine conditions should promote competitive exclu-
assembled a small set of pairwise mutualist mutualist strains varied from one community sion, yet we commonly observed coexistence
communities with and without cheaters and to the next (figs. S3 and S4), and there was no among mutualists as well as between mutu-
quantitatively tracked the population size of specific strain that dominated all of the com- alists and cheaters. Johnson and Bronstein
each species. Lysine cheaters quickly became munities. In addition, coexistence of the mutu- (17) suggested that coexistence can be facilitated
Vidal et al., Science 370, 346–350 (2020) 16 October 2020 4 of 5
RESEARCH | REPORT
in multimutualist communities if mutualist competition for mutualistic commodities [e.g., 14. C. E. Christian, Nature 413, 635–639 (2001).
species can partition the shared mutualistic (24)]. Second, for mutualist communities ex- 15. J. Memmott, N. M. Waser, M. V. Price, Tolerance of pollination
commodity as well as another nonmutualistic periencing exploitation, the fate of the com-
resource. The redundant mutualist strains in our munity is determined largely by the effect of networks to species extinctions. Proc. R. Soc. London Biol. Sci.
system likely have trade-offs in their competitive cheaters on mutualist population dynamics 271, 2605–2611 (2004).
ability for different resources, and this, com- [e.g., (25)]. Cheaters that use resources more 16. S. D. Allison, J. B. Martiny, Proc. Natl. Acad. Sci. U.S.A. 105
bined with temporal resource heterogeneity efficiently or that better survive periods of (Suppl 1), 11512–11519 (2008).
due to mutualistic resource production and low resource availability will have an advan- 17. C. A. Johnson, J. L. Bronstein, Ecology 100, e02708 (2019).
nonmutualistic resource consumption, may tage over mutualists. Third, in complex, diffuse 18. Materials and methods are available as supplementary materials.
allow coexistence of multiple mutualist strains. mutualistic networks, regulatory mechanisms 19. J. L. Sachs, U. G. Mueller, T. P. Wilcox, J. J. Bull, Q. Rev. Biol.
How the temporal dynamics of competition such as host sanctions, partner choice, and 79, 135–160 (2004).
and resource availability drive coexistence in positive partner feedbacks [e.g., (26,27)] may 20. N. M. Waser, L. Chittka, M. V. Price, N. M. Williams, J. Ollerton,
mutualisms requires further experimenta- not be required to explain community stabil- Ecology 77, 1043–1060 (1996).
tion. Our results, however, highlight the impor- ity. Regulatory mechanisms are unlikely to work 21. W. Shou, S. Ram, J. M. Vilar, Proc. Natl. Acad. Sci. U.S.A. 104,
tance of context dependency in determining in diffuse mutualisms because these commun- 1877–1882 (2007).
coexistence in multimutualist communities ities have many species that differ in life history, 22. D. Tilman, “Interspecific competition and multispecies coexistence”
with and without exploitation. behavior, and the benefit they provide. Com- in Theoretical Ecology: Principles and Application, R. May,
petition, however, is one mechanism that is A. R. McLean, Eds. (Oxford Press, ed. 3, 2007), pp. 84–97.
Together, our results show that species rich- universal across species and communities and 23. P. Chesson, Annu. Rev. Ecol. Evol. Syst. 31, 343–366 (2000).
ness can ameliorate mutualistic community offers a general framework to explain the 24. E. I. Jones, J. L. Bronstein, R. Ferrière, Ann. N. Y. Acad. Sci.
collapse caused by cheaters. Mutualist func- stability of diverse types of mutualistic com- 1256, 66–88 (2012).
tional redundancy allowed the mutualism to munities under exploitation. In the face of 25. J. N. Holland, D. L. DeAngelis, Ecology 91, 1286–1295
persist even with extinction of some mutu- inevitable competition among redundant (2010).
alist species. The positive effect of functional mutualist species and exploiters, mainte- 26. E. T. Kiers et al., Science 333, 880–882 (2011).
redundancy was pronounced for mutualist nance of high richness in natural systems is 27. O. Pellmyr, C. J. Huth, Nature 372, 257–260 (1994).
species that directly competed with cheaters necessary to promote persistence of mutual- 28. M. C. Vidal et al., Data from: Species richness and redundancy
for the most limiting mutualistic commodity. ist communities. promote persistence of exploited mutualisms in yeast. Dryad
Although only one type of cheater markedly (2020); https://doi.org/10.5061/dryad.pc866t1m6.
affected community survival, both had nega- REFERENCES AND NOTES
tive effects on the communities in terms of ACKNOWLEDGMENTS
species loss, possibly because they reduce the 1. J. Albrecht et al., Nat. Commun. 5, 3810 (2014).
availability of the mutualistic resources. Thus, 2. J. Bascompte, P. Jordano, Annu. Rev. Ecol. Evol. Syst. 38, We thank J. Bronstein, T. Anneberg, A. Curé, and D. Luna for
our results show that cheaters in general can comments on earlier versions of the draft. Two anonymous
have negative effects on mutualist communi- 567–593 (2007). reviewers provided insightful, detailed comments that greatly
ties even when they do not cause community 3. E. Toby Kiers, T. M. Palmer, A. R. Ives, J. F. Bruno, improved the manuscript. We thank W. Shou for providing the
collapse. overproduction mutants and S. Erdman for access to equipment.
J. L. Bronstein, Ecol. Lett. 13, 1459–1474 (2010). We thank C. Moore, C. Ritchie, and M. Ritchie for insightful
In terms of understanding the persistence 4. A. James, J. W. Pitchford, M. J. Plank, Nature 487, 227–230 (2012). discussions. Funding: This research was funded by NSF-DEB
of multispecies mutualistic communities, our 5. U. Bastolla et al., Nature 458, 1018–1020 (2009). 1655544. Author contributions: K.A.S. and D.M.R. modified the
results suggest three key findings. First, ex- 6. E. Thébault, C. Fontaine, Science 329, 853–856 (2010). yeast strains. M.C.V., S.P.W., D.M.A., and K.A.S. designed the
ploitation can have strong negative effects on 7. K. R. S. Hale, F. S. Valdovinos, N. D. Martinez, Nat. Commun. experiments and synthesized media. M.C.V., S.P.W., and K.A.S.
multimutualist communities by affecting com- conducted the experiments and collected data. M.C.V. and
munity persistence and species loss. However, 11, 2182 (2020). S.P.W. analyzed the data. M.C.V. wrote the first draft, and M.C.V.,
the negative effects of cheaters are context de- 8. M. C. Urban, Science 348, 571–573 (2015). S.P.W., D.M.A., and K.A.S. contributed equally to editing.
pendent and vary greatly with the strength of 9. J. L. Bronstein, Ecol. Lett. 4, 277–287 (2001). Competing interests: The authors declare no competing interests;
10. O. Pellmyr, J. Leebens-Mack, Am. Nat. 156 (S4), S62–S76 (2000). Data and materials availability: Data are available in (28).
11. E. I. Jones et al., Ecol. Lett. 18, 1270–1284 (2015).
12. M. C. Vidal, S. F. Sendoya, P. S. Oliveira, Ecology 97, SUPPLEMENTARY MATERIALS
1650–1657 (2016). science.sciencemag.org/content/370/6514/346/suppl/DC1
13. J. Genini, L. P. C. Morellato, P. R. Guimarães Jr., J. M. Olesen, Materials and Methods
Figs. S1 to S5
Biol. Lett. 6, 494–497 (2010). Tables S1 to S4
References (29–35)
26 March 2020; resubmitted 26 June 2020
Accepted 26 August 2020
10.1126/science.abb6703
Vidal et al., Science 370, 346–350 (2020) 16 October 2020 5 of 5
RESEARCH
SIGNAL TRANSDUCTION AP-3 and HOPS are distinct events. The label-
ing of mTORC1-related proteins was strongly
The GATOR–Rag GTPase pathway inhibits mTORC1 reduced in the absence of HOPS, but not after
activation by lysosome-derived amino acids ablation of other tethering complex genes
(Fig. 2, A and B, and fig. S4, A and C). The
Geoffrey G. Hesketh1, Fotini Papazotos1, Judy Pawling1, Dushyandi Rajendran1, James D. R. Knight1, loss of labeling of mTORC1-related proteins
Sebastien Martinez1, Mikko Taipale2,3, Daniel Schramek1,2, James W. Dennis1,2,4, Anne-Claude Gingras1,2* upon AP-3 or HOPS disruption is not likely
due to their dissociation from lysosome mem-
The mechanistic target of rapamycin complex 1 (mTORC1) couples nutrient sufficiency to cell growth. branes, because they remained associated with
mTORC1 is activated by exogenously acquired amino acids sensed through the GATOR–Rag guanosine an alternative lysosomal BioID “sensor,” GTP-
triphosphatase (GTPase) pathway, or by amino acids derived through lysosomal degradation of protein locked RAB7A (fig. S5). Collectively, these data
by a poorly defined mechanism. Here, we revealed that amino acids derived from the degradation of support a model (fig. S6) in which VAMP7 and
protein (acquired through oncogenic Ras-driven macropinocytosis) activate mTORC1 by a Rag GTPase– VAMP8 partly coexist in a common late en-
independent mechanism. mTORC1 stimulation through this pathway required the HOPS complex and dosomal domain from which VAMP7 is traf-
was negatively regulated by activation of the GATOR-Rag GTPase pathway. Therefore, distinct but ficked (by AP-3) to a specific late endocytic
functionally coordinated pathways control mTORC1 activity on late endocytic organelles in response to domain where it encounters mTORC1-related
distinct sources of amino acids. proteins through a HOPS-dependent process.
This model was supported by analyzing gene
T he mechanistic target of rapamycin com- with each “sensor” captured >800 proximity coessentiality across published genome-wide
plex 1 (mTORC1) coordinates cell growth interactors at a false discovery rate (FDR) ≤1% CRISPR screens in cancer cell lines (17, 18),
with nutrient sufficiency. Exogenous relative to negative controls (Fig. 1, A and B; which revealed correlations between multiple
amino acids (AAs), taken up through figs. S1C and S2, A and B; and table S1, A to C). HOPS- and mTORC1-related components, sug-
cell surface transporters, promote mTORC1 Although VAMP7 and VAMP8 associated with gesting common functions (19) (fig. S7). We
lysosomal recruitment and activation (1). AAs a large common set of proteins, each also en- therefore aimed to explore the functional con-
derived from exogenous protein, acquired riched sets of proteins representing their dis- tribution of the HOPS complex to mTORC1
through Ras-driven macropinocytosis and tinct trafficking itineraries (Fig. 1, C to F; fig. activation.
trafficked to lysosomes for degradation, also S2, C and D; and tables S2, A and B, and S3, A
activate mTORC1 and can fuel cell growth in and B). VAMP8 enriched plasma membrane, By promoting macropinocytosis, activated
the absence of exogenous free AAs (2–7). Al- cell junction, and endosomal recycling pro- Ras can drive exogenous protein uptake (20),
though the AA transporter SLC38A9 mediates teins, whereas VAMP7 enriched late endocytic resulting in the generation of lysosome-derived
arginine-dependent efflux of protein-derived trafficking complexes, including AP-3 and the AAs that can fuel cell growth (2) and activate
essential AAs (EAAs) from lysosomes (8–10), homotypic fusion and vacuole protein sorting mTORC1 in the absence of exogenous free
whether this leads to mTORC1 activation (HOPS) tethering complex (Fig. 1, D and E). In AAs (3). Human embryonic kidney 293 (HEK293)
through the canonical Rag guanosine tri- addition, multiple proteins known to regulate and engineered HEK293 cells expressing an in-
phosphatase (GTPase) pathway (11) remains mTORC1 function were recovered with VAMP7- ducible oncogenic KRASG12V allele (HEK293G12V;
unknown. wt to a greater extent than with VAMP8 or fig. S8A) were deprived of AAs, then supple-
VAMP7-mut (Fig. 1F; VAMP7-mut more closely mented with exogenous free Leu or with exo-
We developed a pair of “sensors” for proximity- resembled VAMP8, fig. S2E). This is consistent genous protein [bovine serum albumin (BSA),
dependent biotinylation (BioID) (12) charac- with AP-3 mediating the delivery of VAMP7 to added alone or in the presence of Gln]. mTORC1
terization of the surface proteomes of late a HOPS- and mTORC1- containing late endo- was readily reactivated [measured by immuno-
endocytic organelles in living cells, using VAMP7 somal domain. blotting for ribosomal protein S6 kinase
(vesicle-associated membrane protein 7) and 1 (S6K1), phosphorylated at Thr389] in both
VAMP8 (Fig. 1A). VAMP7 and VAMP8 are two To validate the role of AP-3 in localizing cell types by Leu, and in HEK293G12V by the
related R-SNARE (soluble NSF attachment pro- VAMP7 to this domain, we performed VAMP7 synergistic addition of BSA and Gln, with par-
tein receptors) proteins that traffic through BioID after CRISPR-Cas9–mediated ablation tial activation by Gln alone and no activation
late endocytic organelles (13). An N-terminal of genes encoding individual complex sub- by BSA alone (Fig. 3A). Steady-state mTORC1
longin domain in VAMP7 binds adaptor pro- units in pooled cell populations [referred to activity in complete medium (fig. S8B) or after
tein complex 3 (AP-3) and enables its localiza- here as KO-BioID (fig. S3); estimated KO effi- Leu stimulation alone (fig. S8C) was mini-
tion to lysosomes (14–16), the site of mTORC1 ciencies ranged from 85 to 99% for all expe- mally affected by HOPS ablation (Fig. 3B and
activation. Green fluorescent protein (GFP)– riments (table S4)]. Upon ablation of the AP-3 table S4), and there was a moderate reduc-
tagged wild-type VAMP7 (VAMP7-wt), a VAMP7 d-subunit (AP3D1), VAMP7-wt BioID more tion in mTORC1 activation by EAAs (Fig. 3, C
mutant deficient in AP-3 complex binding (15) closely resembled VAMP7-mut and VAMP8 and D). By contrast, there was a marked re-
(VAMP7-mut), and wild-type VAMP8 all local- (fig. S4A), with gains in plasma membrane duction in mTORC1 activation in cells treated
ized to late endocytic organelles (VAMP7-mut proteins (fig. S4B) and loss of HOPS and with BSA and Gln (BSA+Gln; Fig. 3, E and F),
and VAMP8 also partially localized to the mTORC1 proteins (Fig. 2, A and B), confirm- suggesting reduced activation by lysosome-
plasma membrane; fig. S1, A and B). BioID ing that AP-3 mediates VAMP7 lysosomal derived AAs.
localization. By contrast, ablation of individ-
1Lunenfeld-Tanenbaum Research Institute, Sinai Health System, ual HOPS subunits had a milder impact on Lysosome-derived AAs have been suggested
Toronto, ON, Canada. 2Department of Molecular Genetics, total VAMP7 proximity interactions (Fig. 2, A to activate mTORC1 through the Rag GTPases
University of Toronto, Toronto, ON, Canada. 3Donnelly Centre and B, and fig. S4, A to C), and a subset of (3), but whether this is mediated by the GAP
for Cellular and Biomolecular Research, Toronto, ON, Canada. proteins sensitive to AP-3 disruption was in- toward Rags (GATOR)–Rag GTPase (here;
4Department of Laboratory Medicine and Pathobiology, sensitive to HOPS disruption (Fig. 2C and fig. GATOR-Rag) pathway (Fig. 4A) has not been
University of Toronto, Toronto, ON, Canada. S4D), suggesting that VAMP7 trafficking by thoroughly explored. HOPS and GATOR2 are
*Corresponding author. Email: [email protected] evolutionarily related protocoatomer family
complexes (21–23), which together exclu-
sively possess C-terminal RING domains within
Hesketh et al., Science 370, 351–356 (2020) 16 October 2020 1 of 6
RESEARCH | REPORT
“Cloud” of reactive = Biotinylated Organelle “sensors” Soluble negative control
biotinyl-AMP lysine BirA*
VAMP7-specific BirA* Longin BirA* Longin BirA* VAMP8-specific
SNARE SNARE
SNARE Trafficking SNARE
via AP-3 VAMP8
VAMP7 BirA*
Trafficking
Common compartment
Cytoplasm
Lumen
64 32 16 8 4 2 P-value Term ID Term name
log 2 Enriched with VAMP7 0.000000532 CORUM:6389 HOPS complex
2 4 BLOC1-BLOC2 complex
8 0.000039 CORUM:654 SOG complex
spectra 16 AP3 adaptor complex
32 GATOR2 complex
VAMP7-wt only (115) 64 0.00051 CORUM:6753 mTOR signaling pathway
Endocytosis
Total SAINT 0.000257 CORUM:59 SNARE interactions ...
filtered preys Lysosome
(FDR ≤ 1%) Common (722) VAMP8 0.00611 CORUM:6216
0.000202 KEGG:04150
0.00094 KEGG:04144
VAMP8-wt only (239) 0.00351 KEGG:04130
0.00658 KEGG:04142
“Class C core” VAMP7 log Enriched with VAMP8 0.05 CORUM:933 SCRIB-APC complex
2 0.00000885 KEGG:04520 Adherens junction
VPS11 0.00037 KEGG:04360 Axon guidance
VPS39 VPS16 spectra 0.00145 KEGG:05205 Proteoglycans in cancer
VPS41 VPS18 0.00184 KEGG:04530 Tight junction
HOPS VPS33A 0.0106 KEGG:04810 Regulation of actin cytoskeleton
0.0132 KEGG:05412 ARVC
(VPS3) TGFBRAP1 CORVET 0.0216 KEGG:05165 Human papillomavirus infection
VPS8
VAMP7-enriched Common VAMP8-enriched
VVVAAAMMMPPP778---wwmttut
VVVAAAMMMPPP778---wwmttut
VVVAAAMMMPPP787---wwmttut
VVVAAAMMMPPP877---wwmttut
VVVAAAMMMPPP778---wwmttut
VPS11 TSC1HOPS mTORC1-related BORCS5 VPS8 ERBB2 plasma membrane / recycling
VPS16 RPTORcomplex BORCS6
VPS18 RRAGA PLEKHM2 CORVET EGFR
VPS41 RRAGC SH3GLB1
VPS33A LAMTOR1 TBC1D9 ATP6V1B2 FRS2
ARFGAP1
AP3D1 MIOS IGF2R CASK
AP3B1 WDR24 FYCO1
AP3B2 WDR59 GATOR2 KIF1A TFRC DLG1
AP3S1
AP3S2 SZT2 VPS13C SLC1A5 SCRIB
FNIP1 STARD3
0 SLC38A9AP-3 STK11IP RAB11FIP1
NPC1complex PLEKHM3
TMEM106B VPS45 RAB11FIP2
Average spectra TM7SF3
50 (or above) MFSD8 CLINT1 RAB11FIP5
STX7 EPB41L1
STX8 EPB41L2
VTI1B EPB41L5
SNX2
SNX5
GOLGA5
≤ 1% ≤ 5% > 5% Relative
FDR abundance
Fig. 1. VAMP7 and VAMP8 access distinct endosomal domains. VAMP7 and VAMP8, respectively (see table S2, A and B). The black dashed
(A) Establishment of late endosomal BioID “sensors.” N-terminally BirA*-FLAG- diagonal line indicates equal abundance for both baits, with relative fold
tagged VAMP7 (a wild-type, VAMP7-wt, or AP-3 binding–deficient mutant enrichment indicated by gray dashed lines. (D) g:Profiler analysis of CORUM
VAMP7-mut) and VAMP8 traffic through both common and specific late complexes and KEGG pathway terms for VAMP7- and VAMP8-enriched proteins
endocytic compartment surfaces where they allow biotinylation of proximal [as indicated in (C); see table S3, A and B]. (E) Organization of HOPS and CORVET
proteins. (B) Venn diagram of proximity interactions with FDR ≤1% for VAMP7-wt multisubunit tethering complexes. The CORVET-specifying subunit VPS8 and
and VAMP8-wt baits (see table S1, A and C, for complete list). (C) Bait versus the HOPS-specifying subunit VPS41, and in some instances VPS39, are studied
bait plot of abundance (average spectral counts plotted as log2 values) for all here; “class C core” refers to the subunits common to both HOPS and CORVET.
proximity interactions (FDR ≤1%). Enriched proximity interactions (≥2-fold (F) Dot plots (columns show baits) of selected proximity interactors (rows; see
enriched and with ≥10 spectral counts) are color-coded in green and red for legend inset and table S6, A and B, for complete dataset).
Hesketh et al., Science 370, 351–356 (2020) 16 October 2020 2 of 6
RESEARCH | REPORT
= AP-3 Complex or pathway components = CHEVI
= HOPS = mTORC1-related = CORVET
log 2 2 2
2 4 4
4 8 8
spectra 16 16
8 32 32
AP3D1 KO VPS8 KO VPS8 64 VPS33B KO 64
16
AP3D1 32
64
log2 VPS33B
spectra Scramble
Scramble Scramble 2
4
VPS18 KO 2 VPS41 KO 2 VPS33A KO 8
4 4 16
8 8 32
16
32 VPS41 16 VPS33A 64
64
32
64
VPS18
Scramble Scramble Scramble
CORVET VPS8 CORVET VPS8 C CORVET VPS8
CHEVI VPS33B HOPS CHEVI VPS33B HOPS CHEVI VPS33B HOPS
AP-3 AP3D1 AP-3 AP3D1 AP-3 AP3D1
KO Scramble VPS11 VPS16 VPS39 VPS41 VPS33A Scramble VPS11 VPS16 VPS39 VPS41 VPS33A Scramble VPS11 VPS16 VPS39 VPS41 VPS33A
VPS18 VPS18 VPS18
VPS8
VPS33B CORVET mTORC1-related TSC1 TSC BORCS5
VIPAS39 CHEVI RPTOR mTORC1 BORCS6
HOPS RRAGA Rag/ PLEKHM2
VPS11 RRAGC Ragulator SH3GLB1
VPS16 AP-3 LAMTOR1 TBC1D9
VPS18 GATOR2 ARFGAP1
VPS41 MIOS
VPS33A WDR24 KICSTOR FYCO1
AP3D1 WDR59 FLCN KIF1A
AP3B1
AP3B2 SZT2 VPS13C
AP3S1 FNIP1 STARD3
AP3S2 SLC38A9 STK11IP
NPC1 PLEKHM3
TMEM106B
= sgRNA target TM7SF3
MFSD8
Fig. 2. mTORC1 access by VAMP7 is AP-3 and HOPS dependent. (A) Bait HOPS composition. (B) Dot plots of proximity interactors annotated by color
versus bait plots of VAMP7-wt BioID profiles comparing the scramble knockout coding in (A). Proteins that are components of defined complexes are indicated
condition (x axis) to different ablation conditions (y axis; display as in Fig. 1C). (see Fig. 1F for dot plot legend inset). (C) Dot plot of selected preys, highlighting
Members of the indicated protein complexes or pathways are colored as shown differential behavior between AP-3 and HOPS ablation conditions. Darker shade
in the inset legend. CHEVI is a trafficking complex that comprises VPS33B and indicates AP-3– but not HOPS-dependent proximity interactors; lighter shade
VIPAS39 and is used here as a negative control; see Fig. 1E for CORVET and indicates proximity interactors that depend on both AP-3 and HOPS.
three of their subunits (fig. S9A). We therefore treated with BSA+Gln (fig. S11A). This poten- activation by EAAs, as did ablation of the
tested if HOPS and GATOR2 share a func- tiated signal was strongly blunted by concom- GATOR2 subunit MIOS (Fig. 4C). Similar re-
tional as well as evolutionary relationship. KOs itant HOPS ablation [Fig. 4B; effect also seen sults were observed after ablation of LAMTOR1,
of individual GATOR2 or GATOR1 subunits in VPS18 KO clones (fig. S11B] and did not ap- the membrane-anchored subunit of the Ragu-
(fig. S9B) displayed delayed mTORC1 inactiva- pear to result from increased protein-derived lator complex that docks the Rags on the lyso-
tion when deprived of AAs in both HEK293G12V AA production (fig. S11, C to E) (6). some surface (fig. S12, A to C). The potentiated
and HEK293 cells (fig. S10, A and B), with signal was abolished in cells treated with
GATOR1 acting as a strong negative regulator, GATOR complexes control the activation either Torin-1 or rapamycin (fig. S12D) and in
as expected (11). Contrary to the effect of HOPS state of the Rag heterodimer (11). Consistent cells in which RPTOR (regulatory-associated
disruption, GATOR2-ablated cells displayed a with this, ablation of RagA+RagB or RagC+ protein of mTOR) is depleted (fig. S13), con-
marked increase in mTORC1 activation in cells RagD (or all Rags) strongly potentiated firming that mTORC1 activity is responsible
mTORC1 activation by BSA+Gln but blunted
Hesketh et al., Science 370, 351–356 (2020) 16 October 2020 3 of 6
RESEARCH | REPORT
AB = HOPS complex with BSA alone, although the ability of BSA
HEK293G12V HEK293 to synergize with Gln was maintained (Fig. 4D
VPS11 and figs. S13A and S15). Signal potentiation
Pre-starve x 3 hr ++++++ ++++++ KO VVVVVVSVVAVcAPPPPPPPPPr3SSSSSSSSaM33411831DPm161398317blABeVPS18with BSA stimulation was blocked by concom-
Leu x 15 min -+- - - - -+- - - - VPS41 itant ablation of VPS18 with GATOR-Rag path-
BSA - - ++ - - - - ++ - - 100 VPS33A way disruption, whereas potentiation with Gln
x 4 hr Gln - - - ++ - - - - ++ - VAMP7 stimulation alone was not (Fig. 4D and fig. S17).
unfed - - - - -+ - - - - -+ 100 These data suggest that the GATOR-Rag path-
75 way negatively regulates mTORC1 activation
S6K1 100 by protein-derived AAs, with AA production
75 pThr389 75 from exogenous or endogenous sources of
protein being HOPS dependent and HOPS
S6K1 25 independent, respectively.
total
To further explore how Gln may enable Rag-
C Starve x 3 hr EAA x 20 min D EAA independent mTORC1 activation by lysosome-
derived AAs, we treated cells with a glutaminase
# * inhibitor (CB-839), which partially reduced
# mTORC1 activation by BSA+Gln, suggesting
that Gln may be acting primarily as a nitrogen
KO VVVVVVAVVSVcAPPPPPPPPPr3SSSSSSSSaM48331311DPm119383617blABe 1.0 # source (fig. S18) (30). We therefore tested
*** whether any of the other 20 proteinogenic
S6K1 Relative * AAs can enable Rag-independent mTORC1
intensity 0.5 activation (either in combination with BSA
or alone). In addition to Gln, multiple AAs
pThr389 enabled Rag-independent mTORC1 activation
(Fig. 4E, fig. S19 and S20, and table S7). AAs
0.0 resulting in strong Rag-independent mTORC1
activation (including Q, S, N, T, A, P, and G)
S6K1 SVVcrVAPVVVVPVaAVPSPPPPSPm3M3P3SSSSSb33DPS114l3117B6818e19A largely aligned with those previously reported
to synergize with EAA in activating mTORC1
total (31), suggesting that these observations may
be linked.
E Starve x 3 hr [BSA + Gln] x 4 hr F BSA + Gln
We revisited the role of additional lysosomal
VAVVVSVVVVVcAPPPPPPPPPr3SSSSSSSSaM33811314DPm133186917blBAe trafficking proteins in the activation of mTORC1
KO 1.0 by BSA+Gln. HOPS function is coordinated
with Rab7 (32), and ablation of RAB7A re-
S6K1 Relative 0.5 * * * * * * duced activation of mTORC1 by BSA+Gln (and
pThr389 intensity partially reduced activation by EAAs; fig. S21).
0.0 Rab7 ablation both delayed the starvation re-
S6K1 sponse and inhibited activation by BSA+Gln,
total SVVcrVAPVVVVPVaAVPSPPPPSPm33MP3SSSSSb33DPSl141311B78e61891A indicating that Rab7 likely underlies an im-
portant regulatory point for mTORC1 acti-
Fig. 3. mTORC1 activation by exogenous protein–derived AAs requires the HOPS complex. vation (33). We tested additional genes that
(A) Immunoblots of HEK293G12V or HEK293 cells starved in AA-free Dulbecco’s modified Eagle coordinate Rab7 function (some of which cor-
related with HOPS through coessentiality analy-
medium–F12 medium for 3 hours and then treated as indicated (p85 and p70 splice variants are sis; see fig. S7). Ablation of vacuolar fusion
observed for S6K1). (B) Immunoblots of HOPS proteins in HEK293G12V KO cell pools (see table S4 for protein CCZ1 homolog [a Rab7-GEF (34)] re-
KO efficiency). (C) Immunoblots of HEK293G12V KO cell pools starved of AAs for 3 hours and duced activation, whereas TBC1 domain family
member 5 [TBC1D5; a putative Rab7-GAP (33)]
stimulated with EAAs (1×; see materials and methods) for 20 min (see fig. S8D for complete blot, had a partial effect (Fig. 4F). Disrupting the
including starvation condition). (D) Quantification of S6K1-pThr389 immunoblots from three independent Retromer complex [by ablating VPS35 or
VPS29], which is coordinated with Rab7
experiments (see fig S8E for replicates) done as in (C). Values are means ± SD relative to scramble function to mediate late endocytic recycling
(*P < 0.01, #P < 0.05; n = 3; two-tailed paired t test). (E) Immunoblots of HEK293G12V KO cell pools starved (35), reduced activation by BSA+Gln and in-
duced a delayed starvation response, indicat-
of AAs for 3 hours and stimulated with BSA plus Gln for 4 hours. (F) Quantification of three independent ing a role in this pathway.
experiments done as in (E); see (D) for details. (BSA = 3%, Gln = 2 mM). Taken together, our results demonstrate that
GATOR-Rag pathway activation negatively reg-
for the phosphorylation events detected. Signal To better distinguish whether the effects ulates Rag-independent mTORC1 activation by
potentiation requires AA sensing by the GATOR- observed were due to BSA-derived AAs and/ lysosome-derived AAs, although a mechanis-
Rag pathway, as ablation of the sensors for or the Gln contribution, we explored these tic explanation for this negative regulation is
Leu (the Sestrins SESN1, SESN2, and SESN3) stimuli separately. Gln activates mTORC1 currently lacking (3, 6). How Rab7 function is
(24) or Arg (CASTOR1 and CASTOR2) (25) through a Rag-independent mechanism (28, 29), coordinated with late endosomal trafficking
also potentiated stimulation with BSA+Gln, likely involving mTORC1 activation through complexes, including HOPS and Retromer,
albeit with differing magnitudes (fig. S14). By autophagy (30). In HEK293G12V cells (but not
contrast, ablation of folliculin (FLCN), which HEK293; Fig. 3A), Gln treatment induced
promotes an active Rag dimer by a mecha- gradual but transient mTORC1 activation (fig.
nism independent from the GATOR complexes S16A). The Gln signal was unaffected by HOPS
(26, 27), had no potentiation effect (fig. S13). (VPS18) depletion, but was reduced upon ab-
These results were confirmed with another lation of the autophagy-activating kinases
mTORC1 substrate, the eukaryotic transla- ULK1 and ULK2, and was strongly potenti-
tion initiation factor 4E binding protein 1 (fig. ated by ablation of RagA and RagB, LAMTOR1,
S15). These data support a model in which or all three Sestrins (fig. S16, B and C). Thus,
activation of the GATOR-Rag signaling axis mTORC1 activation by Gln is also negatively
negatively regulates Rag-independent acti- regulated by activation of the GATOR-Rag path-
vation of mTORC1 by exogenous protein– way. GATOR-Rag pathway disruption also po-
derived AAs. tentiated mTORC1 activation in cells stimulated
Hesketh et al., Science 370, 351–356 (2020) 16 October 2020 4 of 6
RESEARCH | REPORT
A B Starve x 3 hr
Extracellular Intracellular KO Starve x 4 hr [BSA+Gln] x 4 hr
Amino Anabolic
acids
Sensors GATOR2 GATOR1 Active pathways SVVSNNNNMMMMMMIIIIIIccPPPPPPrrOOOOOOSSaaRRRRSSSSSS11LLLLmm882222++++++bbll++++++eeVSVSNNccSSSVPPSVPPrrccccSSPPrrRRrr11LLSS88221188
mTORC1 Catabolic
? Active Rag pathways
dimer
Lysosome-derived
amino acids Ragulator
Lysosome surface
C Starve x 3 hr Exposure S6K1
pThr389
Starve x 3 hr [BSA+Gln] x 2 hr EAA x 15 min
S6K1
KO SVVSVSRRRRRRRRRMMMIIIcccaaaaaaaaaPPPrrrgggggggggOOOSSSaaaAAAAAASCCCSS111mmm++++++888+++bbblllBBBBBBDDDeee+++CCC+++DDD 100 total
35 VPS18
35 S6K1 MIOS
48 pThr389 E NPRL2
S6K1 GGGGGAAAAAHTTTTTOOOOOOPRRRRR12112S GGGGGAAAAAHTTTTTOOOOOOPRRRRR11221S
total
+ +
RagA + +
+ +
RagC
GAHHTOOOPPR2SS GAHHTOOOPPR2SS Starve x 3 hr
+ LEU x 20 min
Starve x 7 hr Starve x 3 hr BSA + [AA] x 4 hr
D Starve x 3 hr - G A V L I M F YWN Q S T H K R DE C P
[BSA] x 4 hr [BSA+Gln] x 4 hr [Gln] x 4 hr Starve x 4 hr Scramble S6K1
KO pThr389
RagA+B
KO
Exposure
LLLLLLLLVSSVVSVSRRRRccccAAAAAAAAaaaaPPPPrrrrggggSSSSMMaMMMaMMaaMAAAA1111TTTTTTTTmmmm++++8888OOOOOOOObbbbllllBBBBRRRRRRRReeee11111111++++VVVVPPPPSSSS11118888
Exposure
LLLTTTVSVVVVSVSVVVCCCcccAAABBBPPPPPPPPPCCCrrrCSCSSSCSSSSSZZZMMaMaa111231113223111TTTmmm895885599DDDOOObbblll555RRReee111
KO F Starve x 3 hr
Starve x 3 hr [BSA+Gln] x 4 hr EAA x 20 min
S6K1
pThr389 KO
S6K1 S6K1
total pThr389
11 LAMTOR1 S6K1
total
Fig. 4. The GATOR-Rag pathway negatively regulates Rag-independent complex components). (B to F) Immunoblots of HEK293G12V KO cell pools
mTORC1 activation by lysosome–derived AAs. (A) mTORC1 activation [either single or single guide RNA (sgRNA) combinations as indicated] starved
through the GATOR-Rag pathway: Sensing of free AA induces the regulation and stimulated as shown. Additional blots and quantification associated
of the GATOR2-GATOR1 module, which controls the activation status of with (E) are shown in fig. S19 (BSA = 3%, Gln = 2 mM, EAA = 1×). Abbreviations for
the Rag GTPases, enabling recruitment of mTORC1 to the lysosome the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe;
and activation. Positive (+) and negative (−) pathway regulators and effect G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser;
on downstream pathways are indicated (see fig. S9B for an expanded view listing T, Thr; V, Val; W, Trp; and Y, Tyr.
to control the switch between Rag-dependent ship between HOPS and GATOR2 complexes 5. S. M. Davidson et al., Nat. Med. 23, 235–241 (2017).
and Rag-independent mTORC1 activation will may underlie related but divergent roles in 6. M. Nofal, K. Zhang, S. Han, J. D. Rabinowitz, Mol. Cell 67,
require further study. Similarly, whether the activating mTORC1 in response to distinct
HOPS complex contributes to mTORC1 acti- nutrient inputs. 936–946.e5 (2017).
vation purely through its function as a tether, 7. R. M. Perera et al., Nature 524, 361–365 (2015).
or whether it regulates mTORC1 more directly REFERENCES AND NOTES 8. M. Rebsamen et al., Nature 519, 477–481 (2015).
[as proposed for the yeast HOPS subunit Vam6/ 1. R. L. Wolfson, D. M. Sabatini, Cell Metab. 26, 301–309 (2017). 9. S. Wang et al., Science 347, 188–194 (2015).
Vps39, suggested to be a GEF for the yeast RagA 2. C. Commisso et al., Nature 497, 633–637 (2013). 10. G. A. Wyant et al., Cell 171, 642–654.e12 (2017).
and RagB paralog Gtr1 (36)], is also unknown. 3. W. Palm et al., Cell 162, 259–270 (2015). 11. L. Bar-Peled et al., Science 340, 1100–1106 (2013).
We propose that the evolutionary relation- 4. J. J. Kamphorst et al., Cancer Res. 75, 544–553 12. K. J. Roux, D. I. Kim, M. Raida, B. Burke, J. Cell Biol. 196,
(2015). 801–810 (2012).
13. J. P. Luzio, P. R. Pryor, N. A. Bright, Nat. Rev. Mol. Cell Biol. 8,
622–632 (2007).
Hesketh et al., Science 370, 351–356 (2020) 16 October 2020 5 of 6
RESEARCH | REPORT
14. P. R. Pryor et al., Cell 134, 817–827 (2008). 32. H. J. K. Balderhaar, C. Ungermann, J. Cell Sci. 126, 1307–1316 of Canada Undergraduate Student Research Award and
15. H. M. Kent et al., Dev. Cell 22, 979–988 (2012). (2013). BioTalent Canada. We thank the Glowinsky-Sandler family for
16. I. B. Schäfer et al., Nat. Struct. Mol. Biol. 19, 1300–1309 financial support. Author contributions: Conceptualization:
33. A. Kvainickas et al., J. Cell Biol. 218, 3019–3038 (2019). G.G.H., A.-C.G.; Investigation, G.G.H., F.P., J.P., D.R., J.D.R.K.,
(2012). 34. M. Nordmann et al., Curr. Biol. 20, 1654–1659 (2010). S.M., M.T.; Supervision: D.S., J.W.D., A.-C.G.; Writing: G.G.H.,
17. R. M. Meyers et al., Nat. Genet. 49, 1779–1784 (2017). 35. K. E. Chen, M. D. Healy, B. M. Collins, Traffic 20, 465–478 A.-C.G. Competing interests: The authors declare no
18. M. Costanzo et al., Science 353, aaf1420 (2016). competing interests. Data and materials availability: All MS
19. E. Kim et al., Life Sci. Alliance 2, e201800278 (2019). (2019). files used in this study were deposited at MassIVE
20. D. Bar-Sagi, J. R. Feramisco, Science 233, 1061–1068 36. M. Binda et al., Mol. Cell 35, 563–573 (2009). (http://massive.ucsd.edu) and have been assigned the following
accession number: MSV000084159. The scored proximity
(1986). ACKNOWLEDGMENTS interactions associated with quantitative values are
21. S. Dokudovskaya et al., Mol. Cell. Proteomics 10, 006478 available for searching at prohits-web.lunenfeld.ca (project 43).
We thank C. Wong and B. Larsen for assistance with mass All code will be made available upon request.
(2011). spectrometry; J. P. Luzio and D. Owen for VPS33A antiserum; and
22. R. Algret et al., Mol. Cell. Proteomics 13, 2855–2870 F. Zhang for the px459v2 plasmid (Addgene plasmid no. 62988). SUPPLEMENTARY MATERIALS
We thank J. P. Luzio, D. Owen, L. Wartosch, J. Brumell, and J.-Y. Youn science.sciencemag.org/content/370/6514/351/suppl/DC1
(2014). for critical reading of the manuscript and discussions. Funding: Materials and Methods
23. M. P. Rout, M. C. Field, Annu. Rev. Biochem. 86, 637–657 This work was supported by Canadian Cancer Society Research Figs. S1 to S21
Institute Innovation and i2I grants (CCSRI grants 704301 and Tables S1 to S7
(2017). 705938 to A.-C.G.) and a Canadian Institutes of Health Research References (37–51)
24. R. L. Wolfson et al., Science 351, 43–48 (2016). Foundation Grant (CIHR FDN 143301 to A.-C.G). Proteomics
25. L. Chantranupong et al., Cell 165, 153–164 (2016). work was performed at the Network Biology Collaborative Centre View/request a protocol for this paper from Bio-protocol.
26. R. E. Lawrence et al., Science 366, 971–977 (2019). at the Lunenfeld-Tanenbaum Research Institute, a facility
27. K. Shen et al., Cell 179, 1319–1329.e8 (2019). supported by Canada Foundation for Innovation funding, by the 10 August 2019; resubmitted 18 April 2020
28. J. L. Jewell et al., Science 347, 194–198 (2015). Government of Ontario, and by Genome Canada and Ontario Accepted 27 August 2020
29. D. Meng et al., J. Biol. Chem. 295, 2890–2899 (2020). Genomics (OGI-139). A.-C.G. is the Canada Research Chair 10.1126/science.aaz0863
30. H. W. S. Tan, A. Y. L. Sim, Y. C. Long, Nat. Commun. 8, 338 (Tier 1) in Functional Proteomics. G.G.H. was funded by a Basic
Research fellowship from Parkinson Canada, and F.P. was
(2017). funded by a Natural Sciences and Engineering Research Council
31. J. Dyachok, S. Earnest, E. N. Iturraran, M. H. Cobb, E. M. Ross,
J. Biol. Chem. 291, 22414–22426 (2016).
Hesketh et al., Science 370, 351–356 (2020) 16 October 2020 6 of 6
RESEARCH
INORGANIC CHEMISTRY Paramagnetic resonance spectroscopies estab-
lish that 2 indeed exhibits a low-spin (S = 1/2)
Structural and spectroscopic characterization of an Fe(V) state. The X-band electron paramagnetic
Fe(VI) bis(imido) complex resonance (EPR) spectrum of 2 (20 K) exhibits
an anisotropic g-tensor that is roughly axial,
Jorge L. Martinez1, Sean A. Lutz1, Hao Yang2, Jiaze Xie3, Joshua Telser4, Brian M. Hoffman2, g3 = g|| = 1.910 < g⊥ ~ (g1, g2) ~ 2.07 (Fig. 2B
Veronica Carta1, Maren Pink1, Yaroslav Losovyj1, Jeremy M. Smith1* and fig. S19), indicative of a metal-based spin,
High-valent iron species are key intermediates in oxidative biological processes, but hexavalent complexes and the spectrum does not show the resolved
apart from the ferrate ion are exceedingly rare. Here, we report the synthesis and structural and spectroscopic 14N splitting characteristic of iminyl radicals.
characterization of a stable Fe(VI) complex (3) prepared by facile one-electron oxidation of an Fe(V) bis(imido) (24, 25) The g-values are distinctive among the
(2). Single-crystal x-ray diffraction of 2 and 3 revealed four-coordinate Fe centers with an unusual “seesaw” few known low-spin d3 metal-ion centers but
geometry. 57Fe Mössbauer, x-ray photoelectron, x-ray absorption, and electron-nuclear double resonance are consistent with the g-values calculated for
(ENDOR) spectroscopies, supported by electronic structure calculations, support a low-spin (S = 1/2) d3 Fe(V) Fe(V) by density functional theory (DFT), as
configuration in 2 and a diamagnetic (S = 0) d2 Fe(VI) configuration in 3. Their shared seesaw geometry is
electronically dictated by a balance of Fe-imido s- and p-bonding interactions. discussed below.
I ron in oxidation states Fe(IV) or higher provides the purple bis(imido) complex H2B This assignment of a metal-based odd-electron
plays a major role in oxidative reactions. (MesIm)2Fe(=NMes)2, 2 (Fig. 1). The molecu-
lar structure of 2 determined by XRD reveals orbital is confirmed by Q-band (35 GHz) CW
Such species are key intermediates in the a four-coordinate iron center bound by the
electron-nuclear double resonance (ENDOR)
catalytic cycles of many metalloenzymes, bidentate bis(carbene)borate and two imido
measurements. ENDOR provides a nuclear
including those involved in the selective ligands (Fig. 2A). The Fe=N bond distances
magnetic resonance (NMR) spectrum of only
oxidation of hydrocarbons. Fe(IV) complexes, [1.676(1) Å and 1.684(2) Å] are slightly lon-
nuclei that interact with the electron spin
particularly those with multiply bonded oxo, ger than in other structurally characterized
bis(imido) iron complexes (15, 18). However, through hyperfine coupling. This interaction
imido, and nitrido ligands, are now well these examples are all three-coordinate, and
established (1), but still higher oxidation states between the Fe(V) and the two coordinated
of iron are little known. Although invoked in consequently have shorter Fe=N distances 14N atoms determined from their ENDOR
some reactions (2), only a few Fe(V) oxo (2–7),
imido (7–9), and nitrido (10–12) complexes have than four-coordinate 2. The imido ligands in spectra (Fig. 2C and fig. S18) reveals elements
been spectroscopically characterized. Three 2 are slightly bent, as characterized by the
Fe–N–C angles [160.6(1)° and 171.9(1)°]. The of both geometric and electronic structure.
structurally characterized and spectroscopi- most noteworthy structural feature is the seesaw
These two atoms give a single ENDOR response,
cally observed Fe(V) complexes are known: a geometry at the iron center, characterized by a
four-coordinate nitrido (13), a five-coordinate large N–Fe–N angle [135.48(8)°] and dictated by demonstrating their magnetic equivalence,
alkylidyne (14), and a three-coordinate bis(imido) the electronic structure of the metal (see below).
(15). The venerable ferrate ion, [FeO4]2–, is an as expected from the molecular structure.
isolable example of Fe(VI) (16); a six-coordinate The structural metrics support an Fe(V) oxi-
Fe(VI) (S = 0) nitrido complex was spectro- Analysis of the two-dimensional (2D) field-
scopically characterized in frozen solution but dation state with no evidence of imido ligand
oxidation to the iminyl radical, N•=Mes (19–21). frequency pattern of ENDOR spectra collected
was neither isolated nor structurally charac- The Fe=N bond lengths in 2 are ~0.1 Å shorter across the EPR envelope yields a common 14N
terized (17). than those observed in Fe(III) iminyl com- hyperfine coupling tensor, A(14N) = [37, 11, 13]
plexes, whereas the imido C–N bond lengths MHz, and nuclear quadrupole coupling tensor
Here, we report the synthesis of thermal- of 2 are slightly longer than those for iminyl
ligands. There is also no evidence for bond (see the legend to Fig. 2). The near-axial hyper-
ly stable, four-coordinate Fe(V) and Fe(VI)
alteration in the aryl groups of the imido fine coupling is characteristic of spin density
bis(imido) complexes lacking the conventional ligands. Indeed, the C–C bond distances for the in a 14N 2pp orbital, which we take to be the
ligands (e.g., F–, O2–) of high valency, as well as aryl group are similar to those for bis(carbene) 2pp orbital perpendicular to the plane of the
Fe(II) arylamido complexes (18, 22, 23). mesityl ring (see below). The 2pp N→Fe charge
their structural characterization by single-crystal donation into a partially occupied Fe orbital
creates 2pp-spin density on the two nitrogen
x-ray diffraction (XRD). Spectroscopic charac- atoms, and for such a nitrogen p orbital, the
maximum 14N hyperfine tensor component
terization and electronic structure calculations lies along the orbital axis. Using the g-tensor
coordinate frame as our reference, the orienta-
illuminate the covalent bonding forces that tion of A, its axial symmetry, and even the
stabilize their unusual “seesaw” geometries. magnitudes of the tensor components are in
Reaction of the high-spin (S = 3/2) Fe(I) excellent agreement with those suggested by
complex H2B(MesIm)2Fe(h2:h2-dvtms), 1,
{H2B[MesIm]2 = dihydrobis[1-(2,4,6-trime- the DFT calculations below (table S11 and fig.
thylphenyl)imidazol-2-ylidene]borato, dvtms = S32). The magnitude of the largest 14N hyperfine
1,3-divinyl-1,1,3,3-tetramethyldisiloxane} with
two equivalents of the organoazide MesN3
1Department of Chemistry, Indiana University, Bloomington, Fig. 1. Synthesis of high valent iron bis(imido) complexes, 2 and 3; R = 2,4,6-Me3C6H2.
IN 47405, USA. 2Department of Chemistry, Northwestern
University, Evanston, IL 60208, USA. 3Department of
Chemistry, University of Chicago, Chicago, IL 60637, USA.
4Department of Biological, Physical and Health Sciences,
Roosevelt University, Chicago, IL 60605, USA.
*Corresponding author. Email: [email protected]
Martinez et al., Science 370, 356–359 (2020) 16 October 2020 1 of 4
RESEARCH | REPORT
Fig. 2. Characterization data for 2.
(A) X-ray structure showing thermal
ellipsoids at 50% probability. H-atoms are
omitted and the bis(carbene)borate
ligands are shown as lines. C/Fe/N
are shown as black/orange/blue ellipsoids,
respectively. (B) X-band EPR spectrum
(toluene, 20 K). Conditions: 9.378 GHz,
modulation-amplitude 10 G. Fit (red) uses
S = 1/2, g = [g1, g2, g3] = [2.010, 1.995,
1.910], linewidths (hwhm/Gaussian) 40, 40,
65 MHz. (C) 2D field-frequency pattern
of 35 GHz CW 14N ENDOR for 2; only n+ is
observed. Simulation of n+ (red): g; A(14N) =
[37, 11, 13] MHz; P(14N) = (0.8, –1, 0.2) MHz;
(a,b,g) = (20°, 18°, 0°) for both A and P.
Arrow indicates continued change as field is
further raised toward g3. 11B features are
shown in blue; see the supplementary
materials for detailed spectra (fig. S22) and
analysis. Conditions: 35.041 GHz; RF scan speed, 1 MHz/s; modulation amplitude, 2 G; time constant, 32 ms; T = 2 K; (D) Solid-state zero-field 57Fe Mössbauer spectrum,
80 K. Black circles indicate the experiment; red line is the best fit. Parameters: d = −0.25 mm s−1, DEQ = 0.82 mm s−1.
Fig. 3. Characterization data for 3. (A) X-ray structure showing thermal [H2B(MesIm)2Fe(=NMes)2]+ (3), (pyrr2py)Fe(OEt2) (30), (pyrr2py)Fe(CH2PPh3)
ellipsoids at 50% probability. H-atoms are omitted and bis(carbene)borate (30), (pyrr2py)Fe(SCH2PPh3) (30), (pyrr2py)FeCl (29), and (pyrr2py)Fe=NAd
(29). See table S1 for data; (D) Normalized Fe K-edge XAS of 2 and 3. See
ligands are shown as lines. One component of the disordered bis(carbene) table S2 for fitting parameters; (E) High-resolution XPS Fe 2p spectrum of 3.
Black circles indicate the experiment; green and blue lines are the Fe 2p3/2
borate ligand is shown. C/Fe/N are shown as black/orange/blue ellipsoids, and Fe 2p1/2 components, respectively; and the red line is the resulting fit.
respectively. (B) Solid-state zero-field 57Fe Mössbauer spectrum, 80 K. Dark red and orange are the shake-up satellites; cyan is the background.
Black circles indicate the experiment; red line is the best fit. Parameters:
d = −0.48 mm s−1, DEQ = 1.25 mm s−1; (C) Isomer shift versus oxidation See table S5 for fitting parameters.
state for: H2B(MesIm)2Fe(I)(THF) (1), H2B(MesIm)2Fe(=NMes)2 (2),
Martinez et al., Science 370, 356–359 (2020) 16 October 2020 2 of 4
RESEARCH | REPORT
Fig. 4. Electronic structure
modeling. (A) Angular
dependence of the DFT-
calculated electronic energy of
the Fe(VI) model compound
[H2B(MeIm)2Fe(=NMe)2]+, 3′.
(B) Qualitative Fe-3d orbital
diagram showing the effects of
N–Fe–N bending and d-orbital
occupancies for Fe(VI) and
Fe(V). The z-axis lies on N–Fe–
N and C–Fe–C bisector; x-axis
is perpendicular to N-Fe–N
plane. Note that the 3dyz orbital
can engage in both s- and
p-bonding interactions with the
imido ligands, which are shown
separately for clarity. The
3dx2-y2 and 3dxz orbitals are at
higher energy because of
destabilizing interactions with
the s-donor orbitals of the
imido and bis(carbene)borate
ligands, respectively.
component corresponds to a semiempirically angle [143.4(1)°]. Because the structural data 7112.3(4) eV in 2 to 7113.5(4) eV in 3-BF4 (Fig.
3D). This 1.2-eV increase in energy is similar
estimated minority spin density on each nitro- provide no evidence for ligand-based oxidation
in 3-BPh4, they therefore implicate the Fe(VI) to that when the Fe(V) species [(Mecycla-
gen of only r(N) < 0.2 – 0.25 (26), which is oxidation state. mAc)Fe≡N]+ is oxidized to Fe(VI). (17) In ad-
compatible with the even smaller values of dition, the intensity of the pre-edge feature
Solid-state 57Fe Mössbauer spectroscopic
r(N) given by the DFT computations. The quad- increases upon oxidation, consistent with in-
rupole coupling constant, e2qQ/h = 2Pmax data support the Fe(VI) assignment of this creased 4p–3d mixing from shorter metal-ligand
~ −2 MHz is relatively small as the result of bond distances (31). The x-ray photoelectron
strong N→Fe s donation. (27) diamagnetic complex. The substantially de- spectrum (XPS) of 3-BF4 reveals that the Fe
creased isomer shift of 3-BF4 (d = −0.48 mm s−1 2p3/2 binding energy (710.34 eV) increases by
The solid-state 57Fe Mössbauer spectrum at 80 K) and increased quadrupole splitting 1.3 eV from 2 (Fig. 3E, table S5, and figs. S26
(DEQ = 1.25 mm s−1) are both consistent with to S28), as expected for oxidation at the metal
of 2 is an asymmetric doublet (Fig. 2D) the metal-centered oxidation (Fig. 3B). Although center (32). The narrow linewidths and low-
parameters of which (d = −0.25 mm s−1; DEQ = energy, low-intensity shake-up satellites are
0.82 mm s−1 at 80 K) are consistent with a high the isomer shift is not as negative as that of
tetrahedral (S = 1) [FeO4]2– (d = −0.81 mm s−1 also consistent with a diamagnetic Fe(VI) metal
oxidation state complex having low orbital for K2FeO4; fig. S17) (28), it is substantially center in a seesaw geometry (33). The struc-
lower than that observed for six-coordinate ture of 3-BF4 is maintained in solution, as sup-
degeneracy, as expected for the relatively low (S = 0) [(MecyclamAc)Fe≡N]2+ (d = −0.29 mm s−1) ported by multinuclear NMR spectroscopy
(17). As noted for 2, these comparisons are
symmetry structure. Although the parameters difficult because of the different geometries (figs. S6 to S8).
Complexes 2 and 3 are notable for their see-
differ from other Fe(V) complexes (3, 13), e.g., and electronic structures of these three Fe(VI)
[(TAML)Fe=O]− (S = 1/2), d = −0.42 mm s−1; saw geometries, with relatively large N–Fe–N
DEQ = 4.25 mm s−1; [PhB(tBuIm)3Fe≡N]+ (S = complexes. Because of this concern, we com- (q) and small C–Fe–C angles, similar to other d2
1/2), d = −0.45 mm/s; DEQ = 4.78 mm s−1, the pared the isomer shift of 3-BF4 with a series of bis(imido) and bis(oxo) complexes (tables S13
different coordination numbers and/or geom- structurally related iron complexes in a range
and S14). DFT calculations (B3LYP/def2-TZVP)
etries make direct comparisons difficult. of oxidation states (Fig. 3C). Because of the
on the truncated Fe(VI) model compound
Complex 2 has been characterized by other paucity of four-coordinate iron bis(carbene) [H2B(MeIm)2Fe(=NMe)2]+, 3′, with bulky
spectroscopies (see the supplementary materials). mesityl groups replaced by methyl, reveal a
borate complexes, we included bulky bis minimum energy for 3′ at q ~ 140° (Fig. 4A),
Unexpectedly, the cyclic voltammogram of matching the angle found for 3 by XRD (see
(pyrrolyl)pyridine complexes, which also sta- above). Similar calculations for the analogous
2 shows a reversible oxidative wave at low Fe(V) model H2B(MeIm)2Fe(=NMe)2, 2′, show
potential (E1/2 = –0.73 V vs Fc+/Fc; figs. S13 and bilize four-coordinate iron in a seesaw geom- its minimum energy at q ~ 135°, which matches 2
S14), suggesting a stable higher-valent state. etry (29, 30). Notably, we observed a linear (fig. S33). These results indicate that the ge-
Chemical oxidation with Fc+ (as Cp2FeBF4) correlation between isomer shift and oxida- ometries of 2 and 3 are electronically dictated.
afforded the dark blue diamagnetic complex
tion state for this series of complexes (Fig. 3C), These geometric preferences can be ex-
3-BF4 in 81% yield (Fig. 1). The molecular
structure of 3 has been determined by XRD for further supporting the Fe(VI) assignment for plained using a simple bonding analysis that
3-BPh4, prepared analogously using Cp2FeBPh4 3-BF4. A similar analysis of six-coordinate iron
(Fig. 3A). The seesaw coordination environ- cyclam complexes supported the Fe(VI) state considers the angular dependence of metal-
assignment for [(MecyclamAc)Fe≡N]2+, which
ment around iron in 2 is retained upon oxi- was characterized in frozen solution. (17) ligand orbital overlap. Because of the low
dation, with a slight contraction in the Fe=N
X-ray absorption spectroscopy (XAS) also
distances [1.630(2) Å and 1.638(2) Å], linear-
supports metal-based oxidation. The pre-edge
ization of the imido ligands [Fe–N–C 175.1(2)°
and 176.4(2)°], and increase in the N–Fe–N feature in the Fe K-edge XAS shifts from
Martinez et al., Science 370, 356–359 (2020) 16 October 2020 3 of 4
RESEARCH | REPORT
d-electron count in these complexes, the rel- ular orbital (LUMO) of 3 (Fig. 4B, blue) be- 27. E. A. C. Lucken, Nuclear Quadrupole Coupling Constants
ative energies and occupancies of the lowest- cause of s and p interactions between 3dyz (Academic, 1969).
and the imido ligands that destabilize the
lying metal-based orbitals and their covalent antibonding b2 orbital. 28. S. K. Dedushenko, Y. D. Perfiliev, M. G. Goldfeld, A. I. Tsapin,
Hyperfine Int. 136/137, 373–377 (2001).
interactions with the ligand-based orbitals are The combination of strongly donating bis
29. K. Searles et al., Angew. Chem. Int. Ed. 53, 14139–14143
expected to play the greatest role in determining (carbene)borate and imido ligands stabilize (2014).
Fe(V) and Fe(VI), respectively, in complexes 2
molecular geometry (Fig. 4B). The relevant iron- and 3, the unusual molecular structures and 30. D. Sorsche et al., Inorg. Chem. 57, 11552–11559 (2018).
based orbitals are the 3dz2-derived a1 orbital, properties of which are dictated by the elec- 31. N. Aliaga-Alcalde et al., Angew. Chem. Int. Ed. 44, 2908–2912
lying along the bisector of ∠N-Fe-N (q), and
the 3dxy and 3dxz orbitals, whose relative tronic structure requirements of their low (2005).
energies are determined by determined by q. d-electron count. This structural template 32. R. D. Feltham, P. Brant, J. Am. Chem. Soc. 104, 641–645
provides the basis for a distinct class of high-
The structure of 2 is determined by the (1982).
highest singly occupied molecular orbital valent iron complexes. 33. P. Brant, R. D. Feltham, J. Electron. Spectrosc. 32, 205–221
(SOMO), the 3dyz-derived b2 orbital (Fig. 4B,
green). Here, q = 135° minimizes overlap be- REFERENCES AND NOTES (1983).
tween the iron 3dyz and imido p-donor orbit- 34. J. L. Martinez et al., Data for: Structural and spectroscopic
als (fig. S34). Because overlap of the 3dyz and 1. A. R. McDonald, L. Que Jr., Coord. Chem. Rev. 257, 414–428
imido s-donor orbitals and of the 3dxy and (2013). characterization of an Fe(VI) bis(imido) complex, IUScholarWorks
imido p-donor orbitals is less sensitive to q, (2020); https://doi.org/10.5967/vty9-rd36.
the energy of 3dyz dictates the molecular 2. O. Y. Lyakin, K. P. Bryliakov, E. P. Talsi, Coord. Chem. Rev. 384,
geometry. This orbital description is quanti- 126–139 (2019). ACKNOWLEDGMENTS
tatively supported by the EPR/ENDOR results 3. F. Tiago de Oliveira et al., Science 315, 835–838 (2007). We thank the IU Nanoscale Characterization Facility for access
for 2. The computed SOMO reveals most of 4. M. Ghosh et al., J. Am. Chem. Soc. 136, 9524–9527 (2014). to the XPS, M. Warren and J. Wright for assistance with XAS
the spin to be on Fe, with p delocalization of a 5. M. R. Mills, A. C. Weitz, M. P. Hendrich, A. D. Ryabov, measurements at beamline APS 10-BM-A, B, and Y. Gao for
small spin density into the imido nitrogen 2pp assistance with experiments. Funding: Funding from the NSF is
orbitals normal to the mesityl plane: Mulliken T. J. Collins, J. Am. Chem. Soc. 138, 13866–13869 (2016). gratefully acknowledged by J.L.M., S.A.L., and J.M.S. (CHE-
6. R. Fan et al., J. Am. Chem. Soc. 140, 3916–3928 (2018). 1566258) and B.M.H. (MCB-1515981). This material is based upon
spin densities 0.756 on Fe and 0.134 on each 7. K. M. Van Heuvelen et al.., Proc. Natl. Acad. Sci. U.S.A. 109, work supported by the U.S. Department of Energy (DOE), Office
of Science, Office of Basic Energy Sciences under award nos.
imido nitrogen (fig. S31 and table S10). As 11933–11938 (2012). DE-SC0019466 (to J.M.S.) and DE-SC0019342 (to B.M.H.). The
noted above, the g-values are characteristic of 8. S. Hong et al., J. Am. Chem. Soc. 139, 8800–8803 (2017). XPS instrument at the IU Nanoscale Characterization Facility was
a metal-centered spin, and the DFT-calculated 9. S. Hong et al., J. Am. Chem. Soc. 139, 14372–14375 (2017). funded by the NSF (DMR MRI-1126394). Support for the acquisition of
g-tensor has components and orientation in 10. K. Meyer, E. Bill, B. Mienert, T. Weyhermüller, K. Wieghardt, the Bruker Venture D8 diffractometer through the Major Scientific
satisfactory agreement with the experimental Research Equipment Fund from the President of Indiana University
J. Am. Chem. Soc. 121, 4859–4876 (1999). and the Office of the Vice President for Research is gratefully
results (table S11). The DFT and EPR/ENDOR 11. G. Sabenya et al., J. Am. Chem. Soc. 139, 9168–9177 acknowledged. NSF’s ChemMatCARS Sector 15 is supported by the
analyses agree that the 14N-ligand hyperfine NSF Divisions of Chemistry (CHE) and Materials Research (DMR)
(2017). under grant no. CHE-1834750. Use of the Advanced Photon Source,
tensors have near-axial symmetry, with small 12. H. C. Chang et al., J. Am. Chem. Soc. 141, 2421–2434 an Office of Science User Facility operated for the DOE Office of
14N 2pp spin densities yielding small compo- Science by Argonne National Laboratory, was supported by DOE under
nents [calculated, A(14N) = (39.8, 11.4, 16.9) (2019). contract no. DE-AC02-06CH11357. Author contributions: J.L.M.
MHz]. Together, these results confirm the as- 13. J. J. Scepaniak et al., Science 331, 1049–1052 (2011). designed and performed the experiments, interpreted the results, and
signment of 2 as exhibiting an Fe(V) valency. 14. C. Citek, P. H. Oyala, J. C. Peters, J. Am. Chem. Soc. 141, assisted with writing the manuscript. S.A.L. designed and performed
the computational investigations and interpreted the results. J.T.
Similar analysis provides insight into the 15211–15221 (2019). and H.Y. collected and interpreted the EPR and ENDOR data. J.X.
larger q angle in 3, where the relative en- 15. C. Ni, J. C. Fettinger, G. J. Long, M. Brynda, P. P. Power, collected and interpreted the XAS data. V.C. and M.P. collected
ergy of the highest occupied molecular orbital and refined the x-ray data. Y.L. collected and interpreted the XPS data.
Chem. Commun. (45): 6045–6047 (2008). J.M.S. and B.M.H. discussed results and wrote the manuscript.
(HOMO) determines the molecular geometry. 16. V. K. Sharma, Coord. Chem. Rev. 257, 495–510 (2013). Competing interests: The authors declare no competing interests.
The a1 HOMO is most stabilized at q = 144.7°, 17. J. F. Berry et al., Science 312, 1937–1941 (2006). Data and materials availability: All x-ray structural data are available
which minimizes antibonding interactions 18. L. Wang, L. Hu, H. Zhang, H. Chen, L. Deng, J. Am. Chem. Soc. free of charge from the Cambridge Structural Database (CCDC
between the Fe 3dz2 and the imido s-donor 1990354-1990357). Computationally optimized structures as well
orbitals by placing the imido ligands in the 137, 14196–14207 (2015). as raw Mössbauer and XAS data has been deposited at IU Data CORE
3dz2 nodal cone (fig. S34). At this angle, b1 is 19. E. R. King, E. T. Hennessy, T. A. Betley, J. Am. Chem. Soc. 133, and are freely available (34). All other data are presented in the
anticipated to be the lowest unoccupied molec- main text or supplementary materials.
4917–4923 (2011).
20. D. A. Iovan, T. A. Betley, J. Am. Chem. Soc. 138, 1983–1993 SUPPLEMENTARY MATERIALS
(2016). science.sciencemag.org/content/370/6514/356/suppl/DC1
21. M. J. T. Wilding et al., J. Am. Chem. Soc. 139, 14757–14766 Materials and Methods
Figs. S1 to S40
(2017). Tables S1 to S18
22. X. Wang, Z. Mo, J. Xiao, L. Deng, Inorg. Chem. 52, 59–65 References (35–68)
(2013). 12 June 2020; accepted 31 August 2020
23. X. Wang, J. Zhang, L. Wang, L. Deng, Organometallics 34, 10.1126/science.abd3054
2775–2782 (2015).
24. Y. Dong et al., Chem. Sci. 11, 1260–1268 (2020).
25. E. Kogut, H. L. Wiencko, L. Zhang, D. E. Cordeau, T. H. Warren,
J. Am. Chem. Soc. 127, 11248–11249 (2005).
26. A. H. Cohen, B. M. Hoffman, J. Phys. Chem. 78, 1313–1321 (1974).
Martinez et al., Science 370, 356–359 (2020) 16 October 2020 4 of 4
RESEARCH
STRUCTURAL BIOLOGY manner. At a comparatively high concentra-
tion (50 nM), GS-6207 effectively inhibited
Structural and mechanistic bases for a potent reverse transcription. At the pharmacologi-
HIV-1 capsid inhibitor cally relevant concentration of 5 nM (3), the
inhibitor partly impaired viral DNA synthesis
Stephanie M. Bester1*, Guochao Wei1*, Haiyan Zhao2*, Daniel Adu-Ampratwum3, Naseer Iqbal2, and effectively blocked formation of 2-LTR
Valentine V. Courouble4, Ashwanth C. Francis5, Arun S. Annamalai1, Parmit K. Singh6,7, circles and integrated HIV-1 DNA. In line
Nikoloz Shkriabai1, Peter Van Blerkom2, James Morrison1, Eric M. Poeschla1, Alan N. Engelman6,7, with these results (Fig. 1C), 5 nM GS-6207
Gregory B. Melikyan5, Patrick R. Griffin4, James R. Fuchs3, markedly reduced viral DNA levels in both
Francisco J. Asturias2†, Mamuka Kvaratskhelia1† the cytoplasm and nucleus (fig. S1). At 0.5 nM,
GS-6207 inhibited integration without detect-
The potent HIV-1 capsid inhibitor GS-6207 is an investigational principal component of long-acting ably affecting reverse transcription. Although
antiretroviral therapy. We found that GS-6207 inhibits HIV-1 by stabilizing and thereby preventing 0.5 nM GS-6207 and 1 mM DTG similarly in-
functional disassembly of the capsid shell in infected cells. X-ray crystallography, cryo–electron hibited integration, the former failed to in-
microscopy, and hydrogen-deuterium exchange experiments revealed that GS-6207 tightly binds two crease 2-LTR circle formation, likely as a result
adjoining capsid subunits and promotes distal intra- and inter-hexamer interactions that stabilize the of concomitant inhibition of nuclear import
curved capsid lattice. In addition, GS-6207 interferes with capsid binding to the cellular HIV-1 cofactors (Fig. 1C). Results of cellular fractionation indeed
Nup153 and CPSF6 that mediate viral nuclear import and direct integration into gene-rich regions of support this interpretation of the population-
chromatin. These findings elucidate structural insights into the multimodal, potent antiviral activity of specific polymerase chain reaction assays (fig.
GS-6207 and provide a means for rationally developing second-generation therapies. S1). Relative to the dimethyl sulfoxide (DMSO)
control, 0.5 nM GS-6207 increased and decreased
L ong-acting antiretroviral therapy would ture in the presence of the inhibitor, identified viral DNA levels in the cytoplasm and nucleus,
substantially improve the care of people a number of CA mutations positioned near respectively (fig. S1). The multistep inhibition,
the potential inhibitor binding site that con- which depends on the concentration of GS-
living with HIV and would mitigate a ferred substantial resistance to GS-CA1 (1). 6207, is likely due to the inhibitor affecting the
However, the structural and mechanistic bases multifaceted roles of CA during virus ingress (4).
number of challenges including the ne- for how this class of compounds binds and
alters the biological functions of HIV-1 CA We considered the following two scenarios
cessity of daily administration of cur- remain unclear. to account for the observed inhibitions of viral
DNA replication intermediates: (i) GS-6207
rent HIV medications, suboptimal treatment We synthesized GS-6207 (Fig. 1A) and ex- could adversely affect functional disassembly
amined its antiviral activities. GS-6207 inhibited of the CA shell through stabilizing or desta-
adherence, and emergence of drug resistance. HIV-1 replication in peripheral blood mono- bilizing its architecture, which in turn would
nuclear cells (PBMCs) and various cell lines, adversely affect reverse transcription, nuclear
GS-6207 (Lenacapavir, Gilead Sciences) is the with EC50 values in the range of ~12 to 314 pM import, and integration; (ii) GS-6207 could
(Fig. 1B and table S1). PBMCs and MT4 T cells interfere with CA interactions with cognate
first-in-class long-acting ultrapotent HIV cap- were fully viable in the presence of 50 mM GS- cellular cofactors needed for nuclear import,
6207 (the highest concentration tested), indi- and/or could interfere with trafficking of pre-
sid (CA) inhibitor. Recently completed phase 1 cating a selectivity index of >106 (Fig. 1B). integration complexes inside the nucleus to
GS-6207 exhibited higher potency during preferred sites of integration.
clinical trials (NCT03739866) have suggested early (EC50 ≈ 55 pM) versus late (EC50 ≈ 314 pM)
steps of HIV-1 replication (Fig. 1B and table S1). To examine these possibilities, we imaged
that a 6-month dosing interval may be possi- Our subsequent efforts focused on understand- the effects of GS-6207 on incoming HIV-1 by
ing the structural and mechanistic bases for using single-particle detection of virus cores
ble. On the basis of these results, GS-6207 inhibition of incoming HIV-1 by GS-6207. colabeled with CypA-DsRed (a marker for CA)
and INmNG (IN fused to NeonGreen protein)
has advanced into phase 2/3 clinical trials To dissect HIV-1 post-entry infection steps (fig. S2) (5). GS-6207 substantially increased
targeted by GS-6207, we monitored viral DNA levels of virus cores in the cytoplasm, which
(NCT04143594/NCT04150068). Initial mech- intermediates, including total reverse transcripts, suggested a stabilizing effect of the inhibitor
two–long terminal repeat (2-LTR) circles (a (Fig. 1D). Conversely, GS-6207 inhibited the
anistic studies with GS-CA1, an archetypal surrogate for nuclear import), and integrated formation of IN puncta in the nucleus with
proviruses (the viral copy DNA incorporated concomitant inhibition of HIV-1 infection
predecessor of GS-6207, revealed its multistage into the host cell DNA) (Fig. 1C). In parallel, (Fig. 1, E and F, and fig. S3). These findings
mechanism of antiviral action (1). GS-CA1 we examined the effects of GS-6207 on viral indicate that GS-6207 stabilizes virus cores,
potently [half-maximal effective concentration DNA levels in the cytoplasm and nuclei of leading to their accumulation in the cyto-
infected cells (fig. S1). In control experiments, plasm and preventing nuclear import.
(EC50) = 87 pM] inhibited early steps of HIV- the reverse transcriptase inhibitor azidothy-
1 replication and also exhibited a second, less midine (AZT) impaired viral DNA synthesis, To explore the stabilizing role of the in-
whereas the integrase (IN) inhibitor dolute- hibitor on the CA shell, we conducted in vitro
potent (EC50 = 240 pM) antiviral activity during gravir (DTG) specifically blocked integration, assays with isolated HIV-1 particles (6). In the
virus egress. Molecular modeling studies pre- as evidenced by marked reduction of proviral absence of inhibitor, virus cores fully disso-
DNA and increased levels of 2-LTR circles. In ciated within 30 min, whereas picomolar
dicted that both GS-CA1 and GS-6207 bind contrast, GS-6207 affected multiple sequen- concentrations of GS-6207 markedly enhanced
tial steps of virus ingress in a dose-dependent the stability of native cores (Fig. 1G and fig. S4).
to the hydrophobic pocket formed by two Next, we tested the effects of GS-6207 on tu-
adjoining CA subunits within the hexamer (2). bular assemblies made in the presence of 2 M
HIV-1 genotyping, after selection in cell cul- NaCl (7) (Fig. 1H). The preassembled tubes
1Division of Infectious Diseases, Anschutz Medical Campus,
University of Colorado School of Medicine, Aurora, CO
80045, USA. 2Department of Biochemistry and Molecular
Genetics, Anschutz Medical Campus, University of Colorado
School of Medicine, Aurora, CO 80045, USA. 3Division of
Medicinal Chemistry and Pharmacognosy, College of
Pharmacy, The Ohio State University, Columbus, OH 43210,
USA. 4Department of Molecular Medicine, The Scripps
Research Institute, Jupiter, FL 33458, USA. 5Department of
Pediatrics, Infectious Diseases, Emory University, Atlanta, GA
30322, USA. 6Department of Cancer Immunology and
Virology, Dana-Farber Cancer Institute, Boston, MA 02215,
USA. 7Department of Medicine, Harvard Medical School,
Boston, MA 02115, USA.
*These authors contributed equally to this work.
†Corresponding author. Email: mamuka.kvaratskhelia@
cuanschutz.edu (M.K.); [email protected] (F.J.A.)
Bester et al., Science 370, 360–364 (2020) 16 October 2020 1 of 5
RESEARCH | REPORT
dissociated immediately upon exposure to Fig. 1. Multimodal mechanism of action of GS-6207. (A) Chemical structure of GS-6207. (B) Antiviral
a buffer containing 150 mM NaCl (<1 min; activities and cytotoxicity of GS-6207 (see also table S1). (C) Effects of GS-6207 on formation of total reverse
Fig. 1H, lane 2). In sharp contrast, addition transcripts, 2-LTR circles, and proviruses. Error bars indicate SD for three independent experiments.
of GS-6207 to preassembled CA tubes rendered (D) Effect of GS-6207 on the number of post-fusion HIV-1 cores in the cytoplasm (see also fig. S2).
these tubular assemblies highly resistant to low (E) Inhibition of nuclear import of HIV-1 (see fig. S3A). (F) Effect of GS-6207 on HIV-1 infectivity (see
ionic strength (150 mM NaCl) conditions. fig. S3B). eGFP, enhanced green fluorescent protein. (G) GS-6207 increases the stability of isolated
Strikingly, in the presence of GS-6207, tubu- HIV-1 cores in vitro (see fig. S4). Error bars in (D) to (G) represent SEM from four fields of view for a
lar CA assemblies remained stable even after representative experiment of two independent experiments (*P < 0.05, **P < 0.005 ***P < 0.0001).
96 hours of incubation under physiologically (H) Effects of GS-6207 on the stability of recombinant CA tubes. Only pelleted fractions of CA from each
relevant conditions (Fig. 1H). The stabilizing reaction are shown. CA tubes were assembled in 2 M NaCl in the absence (top) or presence of GS-6207
effects correlated with a GS-6207:CA ratio of (bottom) and then either directly pelleted (lane 1) or exposed to low ionic strength (150 mM NaCl) buffer for
~1:1 (fig. S5). increasing periods of time (0, 1, 4, 24, 48, and 96 hours; lanes 2 to 7) and then pelleted.
We tested the effects of the cellular CA CA1-NTD and CA2-NTD. GS-6207 establishes The interacting helices that predominantly
binding partner CypA on GS-6207’s activities. form the GS-6207 binding pocket include aH3
As expected (8, 9), the addition of increasing a hydrogen-bonding network with the side and aH4 from CA1-NTD, aH8 and aH9 from
concentrations of CypA resulted in effective chains of Asn57, Lys70, and Asn74 of CA1-NTD, CA2-CTD, and aH2* from CA2-NTD (Fig. 2C).
disassembly of the preformed CA tubes in the Ser41 of CA2-NTD, and Gln179 and Asn183 of Particularly noteworthy is that GS-6207 strongly
absence of the inhibitor (fig. S6). In sharp
contrast, GS-6207–stabilized CA tubes remained CA2-CTD (Fig. 2B, fig. S10, and table S3).
intact in the presence of CypA (fig. S6). Fur-
thermore, GS-6207’s antiviral activities remained
unaffected by depletion or overexpression of
CypA in Jurkat and MT4 T cells (fig. S7).
Next, we examined whether GS-6207 affects
CA interactions with the known cellular co-
factors Nup153 and CPSF6 needed for nuclear
import (10, 11). GS-6207 substantially reduced
binding of cellular Nup153 and CPSF6 to pre-
assembled CA tubes (fig. S8). Because CPSF6 is
also known to regulate integration site selec-
tivity (12, 13), we tested whether GS-6207
influences sites of HIV-1 integration. The in-
hibitor substantially reduced integration in
gene-dense regions and, conversely, enhanced
integration in lamina-associated domains
(fig. S9). These GS-6207–mediated effects on
integration targeting mimicked the CPSF6
depletion phenotype. However, the extent of
inhibitor-induced changes was less than those
seen with CPSF6 knockout, which suggests
that GS-6207 may not fully displace the cel-
lular cofactor. Taken together, our mechanistic
studies reveal stabilizing effects of GS-6207 on
viral cores, coupled with the ability of the in-
hibitor to interfere with CA binding to the cog-
nate cellular cofactors CPSF6 and Nup153.
To understand the structural basis for GS-
6207 interaction with CA, we solved a cocrystal
structure of the inhibitor bound to a prestabi-
lized CAA14C/E45C/W184A/M185A hexamer (14) (Fig. 2
and table S2). The high-resolution structure
(2.22 Å) revealed that GS-6207 binds in the
hydrophobic pocket formed by two adjacent
CA subunits (Fig. 2A and fig. S10) with a
stoichiometry of six GS-6207 compounds bound
per each CAA14C/E45C/W184A/M185A hexamer. GS-
6207 makes extensive van der Waals and
hydrogen-bonding interactions with CA1-NTD
(the N-terminal domain of CA subunit 1), CA2-
CTD (the C-terminal domain of CA subunit 2),
and CA2-NTD. Two ring systems, R3 and R4,
primarily drive the van der Waals interactions
with CA1-NTD and CA2-CTD (fig. S10). R1 and
R2 also provide additional interactions with
Bester et al., Science 370, 360–364 (2020) 16 October 2020 2 of 5
RESEARCH | REPORT
Fig. 2. Structural basis for GS-6207 interaction with CA hexamer. (A) X-ray crystal structure of GS-6207 hexamers (15, 18). The backbone of Nup153
(orange) bound to the prestabilized CAA14C/E45C/W184A/M185A hexamer (PDB ID 6VKV). GS-6207 binds at aligns along R1 and R3 of GS-6207, with Phe1417
the pocket formed by two adjoining CA subunits CA1 (light gray) and CA2 (pale yellow). Relative positions of of Nup153 closely superimposing on the di-
CA1-NTD, CA2-NTD, and CA2-CTD are indicated. (B) Cartoon representation of the structure indicating fluorobenzyl moiety (R3) of GS-6207 (fig. S23).
GS-6207’s interactions with the two subunits that form the binding pocket, CA1 and CA2. Hydrogen bonds Similarly, there is substantial overlap between
are denoted by black dashed lines. (C) The main helices (aH2*, aH3, aH4, aH8, and aH9) that interact GS-6207 and the main chain of CPSF6, with
with GS-6207 are indicated. (D) Reported resistance mutations (green) for GS-CA1 (1) are shown in the Phe321 of CPSF6 superimposing on R3 extremely
context of GS-6207 bound to CA1-NTD. well (fig. S24). Interestingly, the binding pockets
for Nup153 and CPSF6 are more open, with CTD
influences the conformation and relative posi- promised (fig. S21). The infectivity of N57S and aH9 being positioned farther away from NTD
tioning of aH9 of CA2-CTD with respect to aH4 K70A mutant viruses, which exhibited substan- aH4 than in the presence of GS-6207. In turn,
of CA1-NTD. For comparison, aH9 is seen to tial resistance to the inhibitor, was severely and the closer aH4-aH9 conformation imposed by
exhibit substantial conformational variation in considerably reduced, respectively. Q67H and GS-6207 creates steric clashes with Nup153 and
the absence or presence of different cellular pro- N74D, which exhibited lower levels of resistance, CPSF6 (figs. S23 and S24). Collectively, these
tein partners bound to CAA14C/E45C/W184A/M185A displayed wild-type HIV-1 infectivity (fig. S21). findings provide structural explanations for the
or native CA hexamers (figs. S11 to S20). displacement of Nup153 and CPSF6 by GS-6207
Structural comparison of GS-6207 with the (fig. S8).
Previously reported resistant mutations to substantially less potent HIV-1 CA inhibitor
predecessor compound GS-CA1 are within PF74 (15–17) revealed both similarities and To understand the structural basis for GS-
close proximity of the GS-6207 binding site marked differences (fig. S22 and table S3). 6207’s interactions with curved CA assemblies,
(Fig. 2D). Met66 is a key constituent of the The resemblance between the two compounds we used cryo-EM. GS-6207, but not a DMSO
hydrophobic pocket and forms strong van der is seen with respect to their interactions with control, stabilized preformed tubes and re-
Waals interactions with rings R3 and R4. The CA1-NTD. The phenyl R1 and R2 rings and sulted in well-defined tubular CA assemblies
M66I substitution had the most profound indole R3 ring of PF74 superimpose onto the at physiological salt concentration (Fig. 3A and
effects on loss of GS-6207 potency, reducing indazole (R2), difluorobenzyl (R3), and cyclo- figs. S25 and S26). Imaging these structures
activity by more than four orders of magni- pentapyrazole (R4) rings of GS-6207, respec- allowed us to obtain a 6.3 Å map for GS-6207
tude (table S4). N57S, Q67H, K70A, and N74D tively. However, unlike PF74, which only makes bound to A92E CA tubes (Fig. 3B, figs. S27
substitutions, which are expected to adversely limited hydrophobic contacts with CA2-CTD, and S28, and table S5); GS-6207 interacted
affect direct interactions of CA with GS-6207, GS-6207 establishes extensive hydrogen- similarly with wild-type and A92E CA tubes
reduced potency by factors of ~60, ~10, ~45, bonding and hydrophobic interactions with (fig. S25), and the latter protein was success-
and 14, respectively (table S4). Consistent with adjoining CA2-NTD and CA2-CTD (Fig. 2 and fully used for prior cryo-EM studies (8, 19).
a previous report (1), infectivity of the M66I figs. S10 and S22). A hexamer with pseudo–two-fold symmetry
mutant virus, which conferred the greatest ex- characteristic of CA tubes was readily iden-
tent of GS-6207 resistance, was markedly com- We also compared GS-6207 binding to known tified (Fig. 3, C and D) and was further re-
interactions of CPSF6 and Nup153 with CA fined by analyzing helical tube patches with
RASTR, a single-particle approach indepen-
dent of helical parameter determination (fig.
S29) (20). The mutually independent helical
and RASTR approaches produced equivalent
maps of a tube hexamer (fig. S30), further
validating the map’s accuracy (figs. S31 to
S33). Rigid-body docking of individual crys-
tallographic CA monomers in the presence
of GS-6207 could account for all features in
the cryo-EM hexamer, including the posi-
tions of well-defined a helices in the CTD
(Fig. 3D and figs. S30 and S34). Density cor-
responding to bound GS-6207 could be iden-
tified by segmentation of the helical or RASTR
cryo-EM maps (fig. S35). Thus, we were able
to obtain a model of the GS-6207–bound
tube hexamers under physiologically relevant
conditions.
Comparisons of our cryo-EM structure
with published cryo-EM– and cryo–electron
tomography–derived structures of CA hex-
amers from tubes and native HIV-1 particles
(19, 21) reveal the principal differences in
formation of curved hexameric lattices in
the absence and presence of GS-6207 (figs.
S36 to S41). Normally, CA CTDs move away
from the adjacent NTDs to accommodate
inter-hexamer contacts in the context of a
curved topology (19, 21). In sharp contrast,
GS-6207 strongly restricts changes in the
Bester et al., Science 370, 360–364 (2020) 16 October 2020 3 of 5
RESEARCH | REPORT
Fig. 3. Cryo-EM structure
of GS-6207–stabilized CA
tubes. (A) Cryo-EM image of
A92E CA tubes stabilized by
GS-6207 in 150 mM NaCl.
Inset shows a subset of the
averages obtained by 2D
clustering of tube segments.
(B) Cryo-EM map at
6.3 Å resolution from helical
processing of GS-6207–
stabilized A92E CA tubes.
(C) Diagram showing the
pseudo–two-fold symmetric
arrangement of monomers in
a tube hexamer. (D) Atomic
model of a hexamer in the
GS-6207–stabilized CA tube
generated by rigid-body
fitting of six copies of the
x-ray structure of a GS-6207–
bound CA monomer into the
RASTR map. (E) A portion
of a tube showing interactions
between seven hexamers.
Coloring corresponds to HDX
protection levels. The cyan,
green, and orange lines indi-
cate three helical directions.
(F) Close up of aH9-aH9
interactions involving a
central hexamer. All six H9
helices in the central
hexamer (dark blue) were
superimposed. aH9 helices in
neighboring hexamers are
shown in cyan, green, and orange, matching the coloring of helical directions in (E). The absence of true two-fold symmetry results in slight differences in the
positioning of the two helices along a specific helical direction. The visible side chain is Glu180.
CTD position with respect to the adjoining CA that also stabilizes virus cores (22)] showed rectly interact with GS-6207 (fig. S47), conferred
NTD, and requirements for establishing inter- strong protection (figs. S47 and S49) despite a partial resistance to the inhibitor (fig. S52).
hexamer interactions on a curved surface are lack of direct contacts with GS-6207. These
satisfied by repositioning of the comparatively findings suggest that GS-6207 stabilizes indi- Pliability of intra- and inter-hexamer in-
rigid GS-6207–bound CA monomers in each vidual CA hexamers. This notion is further teractions is essential for both proper assembly
hexamer (see movies S1 to S4; also compare supported by thermal shift assays, which show of the CA shell during virion maturation and its
movie S5 with movies S7 and S9, and movie S6 that GS-6207 substantially increases the melt- subsequent disassembly during virus ingress
with movies S8 and S10). Accordingly, NTD ing temperature of isolated CA hexamers (fig. (8, 23). GS-6207 disrupts this delicately ba-
aH4 and CTD aH9 from adjacent subunits are S50). Collectively these biochemical findings are lanced interplay by rigidifying the CTD confor-
farther apart and closer together in the ab- consistent with our cocrystal structure (Fig. mation and stabilizing both intra-hexamer and
sence and presence of the inhibitor, respec- 2), which shows that each GS-6207 connects aH9-aH9 inter-hexamer interactions (Figs. 2
tively (figs. S37 and S39). two adjoining monomers in a hexamer, with and 3 and fig. S47). These findings provide
the binding of six inhibitors resulting in a structural clues as to how GS-6207 inhibits
To further understand how GS-6207 affects more stable hexamer. functional disassembly of virus cores and blocks
tubular CA assemblies, we used hydrogen- incoming HIV-1 in infected cells (Fig. 1). Taken
deuterium exchange (HDX) (figs. S42 to S46). Strikingly, the strongest GS-6207–induced together, our results elucidate the structural and
HDX experiments revealed strong protection protections were seen in aH9 (figs. S47 and mechanistic bases for the multimodal, potent
in CA segments that directly interact with the S51), which suggests that the inhibitor stabilizes antiviral activity of GS-6207 and provide a plat-
inhibitor (figs. S47 and S48). Unexpectedly, we inter-hexamer aH9-aH9 contacts essential form for rationally developing improved long-
observed strong protection beyond the direct for curved lattice formation (Fig. 3, E and F) acting therapies.
inhibitor binding sites. The NTDs that form (19, 21). The E45A and E180A CA substitutions,
the inner hexamer core and provide the bind- which influence intra- and inter-hexamer We note that during the revision of the pre-
ing site for IP6 [a natural cellular cofactor of interfaces, respectively (23–25), but do not di- sent manuscript, an article describing clinical
targeting of HIV CA by GS-6207, which also
Bester et al., Science 370, 360–364 (2020) 16 October 2020 4 of 5
RESEARCH | REPORT
includes synthesis of the inhibitor and a crystal 18. A. J. Price et al., PLOS Pathog. 8, e1002896 (2012). G.B.M., P.R.G., J.R.F., F.J.A., and M.K. The entire project was conceived
19. G. Zhao et al., Nature 497, 643–646 (2013). by M.K. with all authors providing intellectual input and contributing to
structure of GS-6207 bound to CA hexamer 20. P. S. Randolph, S. M. Stagg, J. Struct. Biol. X 4, 100023 (2020). preparation of the manuscript. Competing interests: A.N.E. declares
(fig. S53), was published (26). 21. S. Mattei, B. Glass, W. J. Hagen, H. G. Kräusslich, J. A. Briggs, fees from ViiV Healthcare Co. for work unrelated to this project. No
other authors declare competing interests. Data and materials
REFERENCES AND NOTES Science 354, 1434–1437 (2016). availability: The cocrystal structure and cryo-EM–derived atomic
22. R. A. Dick et al., Nature 560, 509–512 (2018). model are deposited in the Protein Data Bank under accession
1. S. R. Yant et al., Nat. Med. 25, 1377–1384 (2019). 23. M. del Álamo, M. G. Mateu, J. Mol. Biol. 345, 893–906 numbers 6VKV and 6VWS, respectively. The EM maps are deposited in
2. K. Singh et al., Front. Microbiol. 10, 1227 (2019). the Electron Microscopy Data Bank under accession codes EMD-21423
3. E. Daar et al., paper presented at the Conference on Retroviruses (2005). and EMD-21424. DNA sequences for integration site mapping are
24. J. Shi, J. Zhou, V. B. Shah, C. Aiken, K. Whitby, J. Virol. 85, deposited in the National Center for Biotechnology Information Sequence
and Opportunistic Infections, Boston, 8 to 11 March 2020; Read Archive (NCBI SRA) under accession code PRJNA608802. All
www.croiconference.org/abstract/dose-response-relationship-of- 542–549 (2011). other data are available in the manuscript or the supplementary
subcutaneous-long-acting-hiv-capsid-inhibitor-gs-6207/. 25. B. M. Forshey, U. von Schwedler, W. I. Sundquist, C. Aiken, materials. Materials are available from M.K.
4. M. Yamashita, A. N. Engelman, Trends Microbiol. 25, 741–755 (2017).
5. A. C. Francis, M. Marin, J. Shi, C. Aiken, G. B. Melikyan, J. Virol. 76, 5667–5677 (2002). SUPPLEMENTARY MATERIALS
PLOS Pathog. 12, e1005709 (2016). 26. J. O. Link et al., Nature 584, 614–618 (2020).
6. A. C. Francis, G. B. Melikyan, Cell Host Microbe 23, 536–548.e6 (2018). science.sciencemag.org/content/370/6514/360/suppl/DC1
7. L. S. Ehrlich, B. E. Agresta, C. A. Carter, J. Virol. 66, ACKNOWLEDGMENTS Materials and Methods
4874–4883 (1992). Figs. S1 to S53
8. C. Liu et al., Nat. Commun. 7, 10714 (2016). We thank S. Rebensburg, P. Koneru, and other members of the Tables S1 to S5
9. M. Grättinger et al., Virology 257, 247–260 (1999). participating laboratories for their help with data analysis and valuable 1H, 13C, and HSQC NMR Spectra of GS-6207 and Synthetic
10. D. A. Bejarano et al., eLife 8, e41800 (2019). suggestions. All cryo-EM data were collected at the Anschutz Medical Intermediates
11. K. A. Matreyek, A. Engelman, J. Virol. 85, 7818–7827 (2011). Campus Cryo-EM Core Facility. Funding: Supported by NIH grants References (27–58)
12. V. Achuthan et al., Cell Host Microbe 24, 392–404.e8 (2018). R01 AI062520 and R01 AI143649 (M.K.), R01 AI157802 (M.K., F.J.A., Movies S1 to S10
13. G. A. Sowd et al., Proc. Natl. Acad. Sci. U.S.A. 113, and J.R.F.), U54 AI150472 (M.K., P.R.G., A.C.F., G.B.M., and A.N.E.), R01 MDAR Reproducibility Checklist
E1054–E1063 (2016). AI129862 (G.B.M.), R01 AI052014 (A.N.E.), and R01 AI77344 and DP1
14. O. Pornillos et al., Cell 137, 1282–1292 (2009). DA043915 (E.M.P.). Author contributions: The following authors View/request a protocol for this paper from Bio-protocol.
15. A. J. Price et al., PLOS Pathog. 10, e1004459 (2014). conducted experiments and interpreted the results: S.M.B. (x-ray
16. A. Bhattacharya et al., Proc. Natl. Acad. Sci. U.S.A. 111, crystallography); G.W., A.S.A., and J.M. (virology); N.S. (biochemistry); 26 February 2020; accepted 25 August 2020
18625–18630 (2014). H.Z., N.I., P.V.B., and F.J.A (cryo-EM); D.A.-A. and J.R.F. (medicinal 10.1126/science.abb4808
17. A. Saito et al., J. Virol. 90, 5808–5823 (2016). chemistry); A.C.F. and G.B.M. (cell imaging); V.V.C. and P.R.G. (HDX);
P.K.S. and A.N.E. (integration site sequencing). Separate aspects of
the study were designed and supervised by E.M.P., A.N.E., A.C.F.,
Bester et al., Science 370, 360–364 (2020) 16 October 2020 5 of 5
RESEARCH
PHYSIOLOGY (C18:2), an abundant essential FFA, indica-
tive of active lipolysis from adipose depots.
Comprehensive quantification of fuel use by the Many released amino acids were also essen-
failing and nonfailing human heart tial, reflecting muscle proteolysis. The total
release of carbons was 15% higher than up-
Danielle Murashige1*, Cholsoon Jang2*†, Michael Neinast1‡, Jonathan J. Edwards3, Alexis Cowan2, take, indicating a net loss of leg mass during
Matthew C. Hyman4, Joshua D. Rabinowitz2, David S. Frankel4, Zolt Arany1§ fasting.
The heart consumes circulating nutrients to fuel lifelong contraction, but a comprehensive mapping of In contrast to the leg, the heart obtained
human cardiac fuel use is lacking. We used metabolomics on blood from artery, coronary sinus, and most carbons from FFAs (Fig. 1E), accounting
femoral vein in 110 patients with or without heart failure to quantify the uptake and release of for >70% of carbons (1750 mM) extracted from
277 metabolites, including all major nutrients, by the human heart and leg. The heart primarily consumed the circulation [~99 mM FFA, similar to that
fatty acids and, unexpectedly, little glucose; secreted glutamine and other nitrogen-rich amino acids, previously measured by 11C-palmitate and
indicating active protein breakdown, at a rate ~10 times that of the leg; and released intermediates of positron emission tomography (7) equivalent
the tricarboxylic acid cycle, balancing anaplerosis from amino acid breakdown. Both heart and leg to ~1.4 mmol/min/g of FFA-derived carbons
consumed ketones, glutamate, and acetate in direct proportionality to circulating levels, indicating (8)]. This rate may be underestimated because
that availability is a key driver for consumption of these substrates. The failing heart consumed more FFAs can also accumulate in venous plasma
ketones and lactate and had higher rates of proteolysis. These data provide a comprehensive and when liberated from lipoprotein particles
quantitative picture of human cardiac fuel use. (9). Only a negligible portion of FA carbons
were converted to acylcarnitines and secreted
T he heart generates copious adenosine these metabolites in the artery (CA) versus vein (< 0.25 mM carbons, ~0.01%), indicating nearly
5′-triphosphate (ATP), almost all through (CCS or CFV for coronary sinus and femoral vein, complete oxidation of FFAs. Ketones accounted
oxidative phosphorylation in mitochon- respectively) (Fig. 1A), we identified statistically for ~15% of myocardial carbon uptake (Fig. 1E).
dria (1). A continuous supply of oxygen significant uptake or release of 117 metabolites Acetate, the most abundant short-chain FA
and nutrients is thus vitally needed. Heart across the leg and 65 across the heart (Fig. 1B produced by gut microbiota (10), accounted
failure is a leading cause of death worldwide, and fig. S1B). Our CFV/CA data confirmed all for ~2% of myocardial carbon uptake and was
and the failing heart is often described as an 10 previously reported metabolites taken up or avidly taken up by the heart (Fig. 1E and table
“engine out of fuel” that fails to use circulating released by the human arm without anesthe- S3). Acetate may be used as fuel and as an
nutrients to satisfy its metabolic demands (2). sia (3). We further identified an additional epigenetic modifier through histone acetyla-
Understanding how the heart handles fuels 107 metabolites that were significantly altered tion (11, 12).
during health and disease is thus foundational across the human leg (table S2).
to the rational development of new heart fail- The heart secreted most amino acids (Fig.
ure therapies. Table S3 lists cardiac- and leg-specific up- 1E and table S3), contributing to a large net
take and release of abundant, high-turnover negative nitrogen balance of 108 mM (P <
We measured arteriovenous (A-V) gradients fuels. Both organs efficiently extracted ketones, 0.0001) (fig. S4) and indicating active prote-
of circulating metabolites across the human acetate, and glutamate (up to ~67%). Although olysis. Glutamine and alanine were the most
heart and leg by simultaneously sampling the leg took up glucose, there was no net av- abundant amino acids secreted. Glutamine
blood from radial artery, coronary sinus, and erage glucose uptake by the heart, unlike prior release was nearly equimolar to glutamate up-
femoral vein (Fig. 1A). We enrolled 87 patients reports (4, 5), and cardiac glucose uptake did take, suggesting active exchange of glutamate
who were undergoing elective percutaneous not correlate with plasma insulin or insulin for glutamine for nitrogen removal. Lactate
catheter ablation of atrial fibrillation, had a resistance (fig. S2), perhaps reflecting the fasted uptake likely supports nitrogen release as
left ventricular ejection fraction (LVEF) >50%, state of patients. The heart consumed all free alanine through pyruvate transamination (13).
and had no history of heart failure (fig. S1A). fatty acid (FFA) species, whereas the leg pri- Amino acids with greater nitrogen content
Patient characteristics were representative of marily took up saturated FFAs and released (i.e., the ratio of nitrogen to carbon atoms)
the US middle-aged population (table S1). A unsaturated FFAs, likely from subcutaneous were strongly excreted (fig. S5A), further sup-
total of 600 known metabolites were mea- adipose depots (table S4 and fig. S3). The heart porting the notion that the heart actively
sured using liquid chromatography–mass spec- secreted most essential amino acids, indicat- releases nitrogen as amino acids (14–16). We
trometry, 277 of which were reliably detected in ing active proteolysis (table S3). calculated a net cardiac proteolysis rate of
the plasma. By comparing the abundance of 81 mM amino acid equivalents (AA-Eqs) (Fig. 2A;
To quantify the net contribution of each see the materials and methods for details of
1Perelman School of Medicine, Cardiovascular Institute, metabolite to carbon balance, we measured calculations), i.e., 0.06 mmol of AA-Eqs/min/g
University of Pennsylvania, Philadelphia, PA 19104, USA. absolute arterial concentrations of the most cardiac tissue. By comparison, the rate of
2Department of Chemistry and Lewis-Sigler Institute for abundant metabolites (tables S5 to S7) and proteolysis in the leg was 0.006 mmol of AA-
Integrative Genomics, Princeton University, Princeton, NJ used publicly reported concentrations for the Eqs/min/g. Cardiac proteolysis thus occurs at
08544, USA. 3Department of Pediatrics, Division of Pediatric remaining ones (6). From the concentrations ~10× the rate in the leg, equivalent to ~1.5 g of
Cardiology, Children’s Hospital of Philadelphia, Philadelphia, and molecular formula of each metabolite, protein over 12 hours, or just over 2% of that
PA 19104, USA. 4Division of Cardiovascular Medicine, we calculated the absolute carbon uptake or of cardiac protein (17, 18). Human cardiac pro-
Perelman School of Medicine at the University of excretion of each metabolite, measured as tein is thus actively consumed and produced
Pennsylvania, Philadelphia, PA 19104, USA. micromoles per liter of blood passing through across fasting-feeding cycles.
*These authors contributed equally to this work. the heart or leg (Fig. 1, D and E). The leg
†Present address: Department of Biological Chemistry, University obtained ~90% of carbons as glucose and Suppression of branched chain amino acid
of California Irvine, Irvine, CA 92697, USA. ketone bodies, whereas it released most car- (BCAA) catabolism is implicated in maladaptive
‡Present address: Department of Chemistry and Lewis-Sigler bons as FFAs, lactate, and amino acids (Fig. 1D). remodeling in heart failure (19–21). We calculate
Institute for Integrative Genomics, Princeton University, Princeton, About half of the released FFA was linoleate that BCAAs contribute 105 mM carbon equiv-
NJ 08544, USA. alents to overall carbon use, or just under 5% of
§Corresponding author. Email: [email protected] total carbon combustion, similar to that mea-
sured in mice (22). Further reduction of BCAA
Murashige et al., Science 370, 364–368 (2020) 16 October 2020 1 of 5
RESEARCH | REPORT
A B C
Acetoacetate 15 Hypoxanthine 13
3-Hydroxybutyrate 3-Hydroxybutyrate
Glutamate C16:1 C14:1 2-Ketoisovalerate
Lactate Alanine Acetoacetate
Radial Coronary Gln
Artery (A) Sinus (CS) Glutamate 10 Succinate
C20:4 C18:1 C14:2
10 Lactate Indole-3-Pyruvate
Glucose
Femoral -log10(p*)Acetate C16:2 C18:2
Vein (FV) -log10(p*)C14:2
catheter Alanine
C20:0
Acetate
Glutamine
C18:0 C14:0 5
CA CCS Ketone 5
Amino Acid C16:0
C18:2 C19:3 Leucine
FFA
CA CCS Xanthine
Carnitine Cystine
other C16:2
CCS or CFV
CA <1: uptake -1.7-50.75 -0.5 0.0 0.5 -0.5 0.0 0.4 0.71.0
>1: release log2(CCS/CA) log2(CCS/CA)
CCS or CFV
CA Uptake
Release Uptake Release
DE
Uptake Release Uptake Release
1370 uM -554 uM 2.5 628 uM -88.1 uM
Amino Acid
Leg: Net Carbon Uptake FFA Glucose C18:2 Lactate C18:1 Glutamine
Amino Acid 3-hydroxybutyrate Lactate C18:2 Alanine
Carbon (mM) Glutamate Glutamine 2.0 3-Hydroxybutyrate Citrate/Isocitrate
2.0 c18:0 Alanine Ketone C16:0 Histidine
Acetoacetate Citrate/Isocitrate Lactate Cystine
Ketone Acetate C18:1 1.5 Acetoacetate Lysine
Serine C18:3 Glutamate Ornithine
1.0 C20:0 C20:3 1.0 FFA C18:3 Proline
C20:1 C16:0 C20:3 Serine
Glucose C22:0 Lysine 0.5 C18:0 Glycine
3-hydroxyisobutyrate C22:5 Acetate Arginine
C22:1 Arginine Heart: Net Carbon Uptake 0.0 C20:1 Sorbitol
Cystine Glycine TCA Cycle C22:5 Succinate
C19:0 Proline Carbon (mM) Amino Acid C22:4 Pyruvate
C22:2 Phenylalanine Leucine Asparagine
C24:1 Threonine -0.35 C16:1 Phenylalanine
Carnitine C2 Histidine C14:0 Methionine
C16:1 C20:2 Malate
C20:4 C14:2
Aspartate C14:0 C17:1 Indoxyl sulfate
2-Ketoisovalerate C24:0
Leg: Net Carbon Release 0.0 2-Ketoisovalerate Pyruvate C22:1 Aspartate
C24:0 Leucine Valine Indole-3-Propionate
FFA Carnitine C18:1 Asparagine Isoleucine 3-Hydroxyisobutyrate
Carnitine C3 C22:6 C22:0 Glutaric Acid
-1.0 Carnitine C16 C22:4 2-Ketoisocaproate Hydroxyindoleacetate
Carnitine C14:2 C15:0 C22:6 C18:4
Carnitine C14:1 C20:4 C20:0 Carnitine C3
C23:0 Methionine Tyrosine Methylmalonic Acid
C19:1 Valine C15:0 Carnitine C16:0
Carnitine C18:2 Isoleucine C22:2 Carnitine C18:2
Carnitine C18:0 Tyrosine C14:1 Carnitine C10
Carnitine C4 Methyl-Histidine C19:0 Carnitine C8
C17:1 C22:3 C23:0
C19:1 Carnitine C18:0
0.006 uM Succinate Threonine Carnitine C10:1
C14:1 C24:1 Carnitine C4
Carnitine C2 Carnitine C20
C20:2 C15:1
Carnitine C18:1 -0.008 uM
Malate C17:0
Carnitine C14:1
C14:2 Carnitine C14:2
Carnitine C5
-2.0 2-Ketoisocaproate Release 0.006 uM
C18:4
Amino Acid Aconitate
2-Hydroxyglutarate
TCA Cycle C17:0
C15:1
Lactate Carnitine C10
C22:3
Carnitine C8
-3.0 Carnitine C5
Carnitine C10:1
Carnitine C20
-0.004 uM
Fig. 1. Human A-V metabolomics reveal distinct fuel profiles of the heart and leg. (A) Blood was sampled simultaneously from the radial artery (A), coronary
sinus (CS), and femoral vein (FV), and metabolite uptake or release was determined. (B and C) Volcano plot of metabolite abundance in the FV (B) or CS (C) relative
to (A). P values were derived from one-sample Wilcoxon test and then Benjamini-Hochberg corrected (P*). Dotted line indicates P* = 0.05. (D and E) Net A-V carbon
balance across the leg (D) and heart (E) shown in order of greatest to least average absolute carbon uptake or release.
catabolism in the failing heart would thus have composed of 6% histidine, in contrast to most ation, the heart and leg release TCA cycle in-
only a small effect on overall accessible carbons proteins, which contain only 1 to 2% histidine termediates (Fig. 1, B to E). Tissues may export
for combustion, suggesting that BCAAs affect (24). Therefore, in addition to carrying oxygen, citrate, an allosteric inhibitor of FA oxidation
heart failure through other mechanisms. Histi- myoglobin may also serve as a reservoir of and glycolysis (26), to prevent excessive sup-
dine was the most highly secreted essential carbons. Other cardiac proteins may also con- pression of these pathways. TCA intermediate
amino acid (Fig. 2A and figs. S5A and S5B), tribute, as could sources exogenous to the secretion may also represent a means of clear-
suggesting that the proteins being degraded heart, e.g., through macropinocytosis of albu- ing excess TCA cycle four-carbon units produced
were histidine rich. Myoglobin constitutes 5 min or other plasma proteins (25). by amino acid breakdown. We calculated 140 mM
to 10% of cytosolic cardiac protein, is dispens- (0.1 mmol/min/g) anaplerotic carbon flux from
able for baseline cardiac function (23), and is Paradoxically, despite high reliance on the amino acid breakdown (Fig. 2B), well in excess
tricarboxylic acid (TCA) cycle for ATP gener-
Murashige et al., Science 370, 364–368 (2020) 16 October 2020 2 of 5
RESEARCH | REPORT
A Tyr B Anaplerotic carbon flux of the ~50 mM (0.04 mmol/min/g) loss of TCA
Thr intermediates to the circulation. This is a likely
Net uptake from 100 Val -33.0 underestimate of total anaplerosis because we
circulation Leu Ile could not quantify the contribution from pyru-
20.3 µM +20.9Ala vate carboxylase (27, 28). This suggests active
amino acid Glu Cys Lactate internal catabolism of TCA four-carbon units,
Met e.g., through malic enzyme.
Tyr +165.2 -12.1
His Calculating the contribution of all metabo-
Net amino acid +12.4Asn Pyruvate -38.8 lites to cardiac oxygen consumption and ATP
liberation +21.0Asp PC PDH production (Fig. 3, A and B) implied a cardiac
from Phe requirement of 3.0 mM plasma oxygen, equiv-
proteolysis Ser Arg +4.5Arg alent to 2.0 mM in whole blood after correc-
Val Pro +33.4 +7.4Pro tion for plasma-to-blood volume. Measured
50 Asn Thr -2.9 Oxaloacetate Citrate Glu oxygen consumption was 3.5 mM (~9.5 ml
Leu O2/g/min). Thus, ~60% of measured oxygen
Ile +11.9 +80.4 consumption was accounted for by the mea-
sured metabolite A-V gradients (Fig. 3A).
77.0 µM Asp Malate Isocitrate Gln +31.0 This is an underestimate because it does not
amino acid Lys account for FAs liberated from lipoproteins
+23.3 -88.1 (29). Thus, the remaining ~40% oxygen con-
sumption likely reflects a combination of
Gln Gln lipoprotein-derived FAs (LpFAs) that are
combusted by the heart and those that re-
Gly Glu place albumin-derived FFAs combusted by
+70.1 the heart. Consistent with this, unaccounted
Fumarate oxygen consumption correlated inversely with
Glu measured FA uptake (fig. S6, B and C). Other
0 Ala minor sources may include internal fuels such
Net amino acid Succinate Succinyl-CoA
Gln release into
-6.5 +36.8
circulation +15.1Val
Ala Arg 51.7 µM Uptake from circulation +16.3Ile
Contribution from proteolysis -0.3Met
Cys Pro amino acid Release to circulation
His Ser +5.72-Ketoisovalerate
Lys Phe
Anaplerosis from Amino Acids: 140.3 µM Carbon
Met
-50 TCA carbon lost to circulation: 48.5 µM Carbon
Asp
Fig. 2. Cardiac nitrogen release reveals net amino acid liberation from proteolysis. (A) Calculated
cardiac sources of free amino acids (uptake from circulation is shown in red, liberation from proteolysis in
gray) and released amino acids (shown in blue). Shading is proportional to the quantity of amino acid
uptake of secretion. (B) Calculated anaplerotic carbon input from amino acid consumption exceeds carbon
released as TCA cycle intermediates. Anaplerotic contribution from lactate through pyruvate carboxylase
(PC) could not be determined (dashed lines). All numbers are micromoles of carbon. Non-anaplerotic amino
acids (leucine) and amino acids not catabolized in heart (histidine, phenylalanine, and tyrosine) were
excluded. PDH, pyruvate dehydrogenase.
A 4.0 B Contribution to ATP rEF
pEF
3.54 mM 3.56 mM LpFA
3.0 + FFA LpFA
O2 (mM) + FFA
2.0 LpFA LpFA and
uncounted
and 41.9% FFA
uncounted FFA 34.9% FFA
1.0
FFA (36.5%)
(44%)
0.0 rEF <1%
pEF other
FFA Acetate <1% 5.0%
Lactate
Amino acid Ketone other 4.6% 6.7%
2.8% 6.4% Amino Acid Amino Acid
Lactate Other Ketone 16.4%
Lactate Ketone
Unmeasured (LpFA, FFA, other)
C Amino acid Ketone Lactate D Amino acid-derived
FFA * * nitrogen release
NS NS 0.6 0.2 *
2 0.2 1000
fraction of total O2 1 0.4 0.1 Nitrogen (µ M) 500
0.1 0.2 0
0.0 0.0
0 -500
pEF rEF -0.1 pEF rEF
0.0 pEF rEF
-1
-2 -0.1
pEF rEF pEF rEF
Fig. 3. Comparison of myocardial substrate use in patients with preserved versus reduced ejection fraction. (A) Calculated substrate-specific
contribution to total cardiac oxygen consumption. Average measured myocardial O2 consumption (DO2) is indicated above each bar. (B) Substrate-specific
contribution to cardiac ATP generation in patients with preserved ejection fraction (pEF) versus reduced ejection fraction (rEF). (C) Proportion of total
DO2 accounted for by the catabolism of each indicated substrate class in pEF versus rEF. (D) Net amino acid–derived nitrogen release in patients with pEF
versus rEF. *P < 0.05 by t test.
Murashige et al., Science 370, 364–368 (2020) 16 October 2020 3 of 5
RESEARCH | REPORT
as triglycerides (30), FFAs liberated from epi- (Fig. 3, B to D, and tables S5 to S8). Overall suggesting that consumption of these three
cardial adipose tissue, or oxygen consumption use of FFAs was suppressed. Increased plasma fuels is driven by substrate availability without
that does not generate ATP (e.g., conversion long-chain acylcarnitines have been associated the need for overt regulation. Moreover, their
of hypoxanthine to uric acid by xanthine oxi- with heart failure (35, 36), but tissue acylcarni- fractional uptake correlated strongly with each
dase) (table S7) (31). Combustion of glycogen tine amounts are decreased in failing hearts other (fig. S8, A and B), suggesting that it
is not thought to occur in the heart during (37). We detected no increase in secretion of depends on tissue perfusion. In patients with
fasting (32–34). Assuming that ~35% of LpFAs acylcarnitine by the failing heart (table S9), reduced LVEF, all three fuels were extracted
are released into the coronary sinus, we cal- indicating that (i) any putative defect in FA ~20% more by the heart (fig. S10, A and B),
culated that ~57% of cardiac ATP production consumption in heart failure occurred up- consistent with the longer transit times of
derives from FFAs, 6.4% from ketones, 4.6% stream of acylcarnitine production and (ii) a blood through the heart vasculature allowing
from amino acids, 2.8% from lactate, and ~28% tissue other than the heart increases systemic increased uptake. Thus, the increased con-
from LpFAs (for a total of ~85% from FAs) (Fig. acylcarnitine production. sumption of ketones and glutamate in heart
3B and table S8). failure reflects higher plasma concentrations
To gain insight into factors affecting fuel and lower rates of cardiac perfusion, as op-
We evaluated in a similar fashion 23 pa- choice, we looked for correlations between posed to an inherent change in cardiac capacity
tients diagnosed with cardiomyopathy with metabolite concentrations and their use as to combust these fuels.
an LVEF <40. Other than reduced LVEF, the fuel (Fig. 4A), between fractional extraction
cohort was similar to that with preserved of different metabolites (fig. S8, A and B), and In summary, we measured the uptake and
LVEF (table S1). Patients with reduced LVEF between fractional extraction and clinical or secretion of 277 circulating metabolites by
had nearly tripled consumption of ketones demographic parameters (fig. S9). Uptake of the human nonfailing and failing heart during
(16.4 versus 6.4%), doubled lactate consump- acetate, 3-hydroxybutyrate, and glutamate (but fasting. FAs were the predominant cardiac
tion (5.0 versus 2.8%), and a doubled rate not that of glucose, lactate, or FFA) was directly fuel source. Unexpectedly, we observed little
of net amino acid–derived nitrogen release proportional to circulating concentrations in glucose uptake, perhaps a response to fasting
(212 versus 106 mM), i.e., rate of proteolysis both heart and leg (Fig. 4, A to C, and fig. S10C), or to general anesthesia, limitations of our
A CCS - CA B Acetate C Acetate
Acetate 50 0
Ketone
Lactate 0
Glutamate
Glutamine -50
Glucose
C16:0
C18:0
C18:1
C18:2
C20:4
C22:6
FFA sum
Acetate -50 -100
Ketone
Lactate -100 pEF slope = -0.32 -150 pEF r 22== 0.25
Glutamate 0 rEF slope = -0.43 0 rEF r 0.29
Glutamine
Glucose 50 100 150 200 50 100 150 200
C16:0
C18:0 3-hydroxybutyrate 3-hydroxybutyrate
C18:1
C18:2 0 0
C20:4
C22:6 -200
FFA sum
Arterial concentration
CFV- C A -400 -400
Acetate pEF slope = -0.16 * -600 pEF r 22== 0.30
Ketone rEF slope = -0.23 -800 rEF r 0.62
Lactate
Glutamate -800 0 1000 2000 0 1000 2000
Glutamine 30
Glucose CA(µM)scaled Glutamate Glutamate
C16:0 30
C18:0
C18:1 0 CA(µM) 0
C18:2
C20:4 C- -50 C- -50
C22:6
FFA sum CS FV
greater slope less pEF slope = -0.22 -100 pEF r 22== 0.90
uptake uptake rEF slope = -0.32 rEF r 0.56
0.00 -100
-0.50 0.50 0 50 100 150 0 50 100 150
CA (µM)
r 2> 0.5 r2= 0 CA (µM)
Fig. 4. Cardiac uptake of acetate, ketones, and glutamate primarily depends upon circulating concentrations in pEF and rEF. (A) Relationship of A-V
metabolite gradient (CV – CA) with arterial concentration of indicated metabolites by linear regression. (B) CA versus uptake of indicated metabolites by the heart after
adjustment for acetate extraction [(CCS – CA)scaled; see the supplementary data]. *P < 0.05 by analysis of covariance. (C) CA versus uptake of the indicated
metabolites by the leg.
Murashige et al., Science 370, 364–368 (2020) 16 October 2020 4 of 5
RESEARCH | REPORT
study that were imposed by clinical necessity. 14. T. Takala, J. K. Hiltunen, I. E. Hassinen, Biochem. J. 192, study, in particular M. Gnap, K. Conn, L. Tomczuk, and
Uptake of ketones was substantial and accen- 285–295 (1980). L. Czerniawski, as well as members of the Arany and Rabinowitz
tuated in heart failure. Ketone consumption laboratories for insightful comments, in particular N. Yücel and
has been suggested to be protective in heart 15. H. Taegtmeyer, A. G. Ferguson, M. Lesch, Exp. Mol. Pathol. 26, S. Yang. Funding: D.M. was supported by the NHLBI (F30
failure (38), and our data suggest that delivery 52–62 (1977). HL142186-01A1) and the Blavatnik Family Foundation, C.J. was
of ketones to the heart should be readily supported by the American Diabetes Association (1-17-PDF-076),
achievable. There was evidence of cardiac pro- 16. O. I. Pisarenko, E. S. Solomatina, I. M. Studneva, Biochim. J.J.E. was supported by NIH 5T32HL007915, J.D.R. was supported
teolysis despite the availability of circulating Biophys. Acta 885, 154–161 (1986). by an NIH Pioneer grant (1DP1DK113643) and an NIH Diabetes
amino acids, which was markedly accentu- Research Center grant (P30 DK019525), and Z.A. was supported
ated in heart failure. Whether this increased 17. G. H. A. Clowes Jr., H. T. Randall, C.-J. Cha, JPEN J. Parenter. by the NHLBI (HL126797) and the NIDDK (DK114103). Author
proteolysis in heart failure is adaptive or mal- Enteral Nutr. 4, 195–205 (1980). contributions: This study was conceived and designed by D.M.,
adaptive will require further studies. The pre- C.J., J.D.R., D.F., and Z.A. Patients were identified and recruited
sent data provide a framework of fuel use in 18. V. R. Preedy, L. Paska, P. H. Sugden, P. S. Schofield, by D.F. and M.H. Samples were collected by D.M., M.H., and D.F.
the beating human heart and a foundation M. C. Sugden, Biochem. J. 250, 179–188 (1988). Mass spectrometry was performed by C.J. Data analysis was
for understanding aberrant cardiac metabo- performed by D.M. with input from C.J. and M.N. Quantification
lism in disease. 19. H. Sun et al., Circulation 133, 2038–2049 (2016). of metabolites by the YSI Biochemistry Analyzer was performed
20. T. Li et al., Cell Metab. 25, 374–385 (2017). by A.C. ELISAs were performed by J.E. Figures were prepared
REFERENCES AND NOTES 21. W. Wang et al., Am. J. Physiol. Heart Circ. Physiol. 311, by D.M. The original draft was written and revised by D.M., C.J.,
and Z.A. and reviewed by all authors. Competing interests: J.D.R.
1. M. F. Allard, B. O. Schönekess, S. L. Henning, D. R. English, H1160–H1169 (2016). is a member of the Rutgers Cancer Institute of New Jersey and
G. D. Lopaschuk, Am. J. Physiol. 267, H742–H750 (1994). 22. M. D. Neinast et al., Cell Metab. 29, 417–429.e4 (2019). the University of Pennsylvania Diabetes Research Center,
23. D. J. Garry et al., Nature 395, 905–908 (1998). cofounder of VL54, stakeholder in Colorado Research Partners,
2. S. Neubauer, N. Engl. J. Med. 356, 1140–1151 (2007). 24. A. E. Romero-Herrera, H. Lehmann, Proc. R. Soc. London Ser. B and consultant to Pfizer, Inc. The remaining authors declare no
3. J. Ivanisevic et al., Sci. Rep. 5, 12757 (2015). competing interests. Data and materials availability: All data are
4. Y. Mizuno et al., Metabolism 77, 65–72 (2017). 186, 249–279 (1974). available in the main text or the supplementary materials.
5. J. A. Wisneski et al., J. Clin. Invest. 76, 1819–1827 (1985). 25. C. Commisso et al., Nature 497, 633–637 (2013).
6. D. S. Wishart et al., Nucleic Acids Res. 46 (D1), D608–D617 26. P. B. Garland, P. J. Randle, E. A. Newsholme, Nature 200, SUPPLEMENTARY MATERIALS
(2018). 169–170 (1963). science.sciencemag.org/content/370/6514/364/suppl/DC1
7. L. R. Peterson et al., Diabetes 57, 32–40 (2008). 27. B. Comte et al., J. Biol. Chem. 272, 26125–26131 (1997). Materials and Methods
8. G. D. Hutchins et al., J. Am. Coll. Cardiol. 15, 1032–1042 (1990). 28. A. R. Panchal et al., Am. J. Physiol. Heart Circ. Physiol. 279, Figs. S1 to S10
9. R. H. Nelson, A. Prasad, A. Lerman, J. M. Miles, Diabetes 56, Tables S1 to S10
H2390–H2398 (2000). Captions for Data S1 to S2
527–530 (2007). 29. H.-L. Noh, K. Okajima, J. D. Molkentin, S. Homma, References (39–43)
10. R. J. Perry et al., Nature 534, 213–217 (2016). MDAR Reproducibility Checklist
11. M. L. Soliman, T. A. Rosenberger, Mol. Cell. Biochem. 352, I. J. Goldberg, Am. J. Physiol. Endocrinol. Metab. 291, Data S1 and S2
E755–E760 (2006).
173–180 (2011). 30. N. H. Banke et al., Circ. Res. 107, 233–241 (2010). View/request a protocol for this paper from Bio-protocol.
12. P. B. Taylor, C. C. Liew, Basic Res. Cardiol. 71, 27–35 (1976). 31. Y. Xia, J. L. Zweier, J. Biol. Chem. 270, 18797–18803 (1995).
13. K. J. Peuhkurinen, I. E. Hassinen, Biochem. J. 202, 67–76 (1982). 32. M. R. Laughlin, W. A. Petit Jr., R. G. Shulman, E. J. Barrett, 19 May 2020; accepted 25 August 2020
Am. J. Physiol. 258, E184–E190 (1990). 10.1126/science.abc8861
33. C. A. Schneider, H. Taegtmeyer, Circ. Res. 68, 1045–1050
(1991).
34. G. Evans, J. Physiol. 82, 468–480 (1934).
35. T. Ahmad et al., J. Am. Coll. Cardiol. 67, 291–299 (2016).
36. W. G. Hunter et al., J. Am. Heart Assoc. 5, e003190 (2016).
37. K. C. Bedi Jr. et al., Circulation 133, 706–716 (2016).
38. R. Nielsen et al., Circulation 139, 2129–2141 (2019).
ACKNOWLEDGMENTS
We thank the staff of the University of Pennsylvania
Electrophysiology Section for their enthusiastic support of this
Murashige et al., Science 370, 364–368 (2020) 16 October 2020 5 of 5
Produced by the Science/AAAS Custom Publishing Office LIFE SCIENCE TECHNOLOGIES
new products
2D-Barcoded NMR Tube Scanner Sterile Deep-Well Plates
The Express Scanner from Ziath delivers an Porvair Sciences offers an extensive range of sterile deep-well micro-
industry-leading fast scan and decode time plates for sensitive biological and drug discovery applications. These
of under 3 s for a full rack of 2D-barcoded high-quality deep-well plates are made from virgin polypropylene and
nuclear magnetic resonance (NMR) sample are available in several different sizes, well shapes, plate formats (the
tubes. 2D-barcoded NMR tubes are usually most commonly used being 96- and 384-well plates), and volumes
difficult to read with a conventional scan- (240 µL–2.2 mL). With qualified low-extractables and low-leachables
ner, as the caps are on top of the tubes and characteristics, they contain no contaminants that can leach out and
most scanners are designed to read from affect stored sample or bacterial or cell growth. They are are precisely
underneath. Designed in collaboration with manufactured to ANSI/SLAS dimensions to ensure they are completely
a leading European NMR spectrometry lab, automation compatible. Porvair Sciences deep-well plates are designed
the Express Scanner operates on top of the sample tubes without with raised well rims to facilitate reliable heat-seal closure—critical for
applying any weight (which could damage the tubes). In addition, for long-term integrity of stored samples at –80°C. Used in conjunction
Bruker NMR spectrometer users, Ziath provides a dedicated slide with a support mat, the plates can be routinely centrifuged at up to
for their tube carriers that can only be inserted in one orientation— 6,000 g.
thereby conserving the origin position (“A1” position) each time. The Porvair Sciences
Express Scanner is supplied with a controller that regulates scanning For info: 800-552-3696
and data processing using Ziath industry-standard DataPaq software www.microplates.com
(version 3.17) or new DP5 web-enabled software that allows export-
ing of tube barcodes as tables or text, or in JPEG, XML, XLS, or JSON Dye for Flow Cytometry
file formats. For convenience, the unit can be separated and posi- Bio-Rad Laboratories announces the launch of StarBright Violet 515
tioned under your lab bench. (SBV515) Dye, the first of a new range of unique fluorescent nanopar-
Ziath ticles for use in flow cytometry. StarBright Dyes are conjugated to a
For info: 858-880-6920 growing range of antibodies and are compatible with most flow cy-
www.ziath.com tometers and experimental protocols. With an excitation maximum of
401 nm and emission maximum of 516 nm, SBV515 offers researchers
Water-Soluble Lipofuscin Autofluorescence Quencher improved brightness to resolve rare and low-antigen-density popula-
Lipofuscin autofluorescence is a common source of background in tions more easily. It has a narrow emission profile to reduce spillover
human and aged animal tissues, particularly brain and eye, which can into neighboring filters and minimal excitation by other lasers, making
make specific fluorescence imaging virtually impossible. Common it suitable for inclusion in multicolor panels. The dye is not susceptible
quenchers such as Sudan Black B have been used to reduce lipofuscin to photobleaching, delivers high lot-to-lot reproducibility, and is stable
autofluorescence, however these reagents often introduce nonspe- at 4°C with no loss of signal, ensuring consistent staining and reliable
cific red and far-red fluorescence. TrueBlack Plus is a next-generation results. It has been tested in many common staining buffers and does
lipofuscin autofluorescence quencher that offers the lowest far-red not require a special staining buffer, permitting easy integration into
background of all available quenchers, including Sudan Black B and our common experimental protocols without compromising performance.
original TrueBlack. In addition, TrueBlack Plus is water-soluble, making Bio-Rad Laboratories
it the only lipofuscin quencher that can be used in aqueous buffer For info: 800-424-6723
instead of ethanol. Quenching in phosphate buffered saline allows www.bio-rad-antibodies.com/flow-cytometry-starbrightviolet515.html
longer incubation times for thicker samples without shrinkage and can
be performed with hydrophobic stains. SARS-CoV-2 rRT-PCR Kit
Biotium The Quick SARS-CoV-2 rRT-PCR Kit is a real-time reverse transcrip-
For info: 510-265-1027 tion PCR (rRT-PCR) test for the qualitative detection of RNA from the
www.biotium.com severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which
is responsible for the coronavirus disease (COVID-19). It targets three
Primary Kupffer Cells different regions of the viral nucleocapsid (N) gene and a host-specific
BioIVT’s high-purity Kupffer cells are ideal for in vitro screening assays, target region (human ribonuclease P gene) to assess sample quality.
toxicity studies, disease models, and predicting inflammation-induced The kit also includes coefficient of variation (CV) Positive Control to
adverse drug reactions. Researchers often use these cells in plated enable assay performance monitoring as well as a No-Template Control
monocultures, cocultures, and liver microtissue configurations. The to confirm the absence of contamination in the reagents. It can be used
cells are cryopreserved immediately after isolation at passage zero. on purified RNA samples isolated from upper respiratory and lower res-
Their purity is assessed using flow-cytometric analysis of CD68+ cells, piratory systems. The Quick SARS-CoV-2 rRT-PCR Kit has high sensitivity
and their function is measured based on interleukin-6 release in with a limit of detection as low as 15 viral genome equivalent copies
response to lipopolysaccharide. Each vial of BioIVT’s high-purity Kupffer per reaction, a fast turnaround time of less than 1.5 h, and a simple
cells contains a minimum of 1 million viable cells. The cells can be workflow in which RNA is simply added to the kit reagents and directly
thawed and plated in coculture with BioIVT’s cryoplateable human analyzed. Due to its simplicity, the kit can be set up with a manual or
hepatocytes using InVitroGRO CP NPC Medium. automated system and is compatible with high-throughput platforms.
BioIVT Zymo Research
For info: 516-483-1196 For info: 888-882-9682
bioivt.com www.zymoresearch.com/products/quick-sars-cov-2-rrt-pcr-kit
Electronically submit your new product description or product literature information! Go to www.sciencemag.org/about/new-products-section for more information.
Newly offered instrumentation, apparatus, and laboratory materials of interest to researchers in all disciplines in academic, industrial, and governmental organizations are featured in this
space. Emphasis is given to purpose, chief characteristics, and availability of products and materials. Endorsement by Science or AAAS of any products or materials mentioned is not
implied. Additional information may be obtained from the manufacturer or supplier.
SCIENCE sciencemag.org/custom-publishing 16 OCTOBER 2020 • VOL 370 ISSUE 6514 369
online @sciencecareers.org FACULTY POSITION IN U.S. POSTAL SERVICE
STEM CELLS AND REGENERATIVE MEDICINE
Statement required by the Act of 12 August 1970,
We invite applications for faculty position at the Assistant or Associate Section 3685, Title 39, United States Code, showing the
Professor levels in the STaR Center at Baylor College of Medicine. ownership, management, and circulation of:
1–9. Science, Publication No. 0036-8075, is pub¬lished
We are seeking motivated investigators in all areas of stem cells and weekly on Friday, except the last week in December,
regenerative medicine including embryonic and adult stem cell biology, at 1200 New York Avenue, N.W., Washing¬ton, DC
organoids, developmental biology, cancer stem cells (hematologic, glioma,
and others). 20005. Date of fling: 29 September 2020. This is also
the address of the publisher, the editor, and the manag-
BCM is the premier medical school of Texas and has the top-ranked ing editor, who are, respectively, Bill Moran, Holden
Genetics and Cell Biology Departments in the U.S. based on NIH Thorp, and Monica M. Bradford.
funding. BCM has internationally recognized strengths in structural 10. The owner is the American Association for the
biology and biochemistry, a long-standing NIH human genome sequencing Advancement of Science, 1200 New York Avenue,
center, cutting edge Advanced Technology Cores, exceptional Ph.D. N.W., Washington, DC 20005. Stockholders: None.
graduate programs, a commitment to technology transfer for faculty 11. Known bondholders, mortgages, and other secu-
discoveries, and a rich history of translating basic science into clinical rity holders owning or holding 1 percent or more of
implementation. total amount of bonds, mortgages, or other secu¬rities:
None.
Applications received by November 1st, 2020 will receive priority.
12. The purpose, function, and nonproft status of this
Please send Cover Letter, CV and a two-page summary of research organization and the exempt status for federal in-come
interests as a single PDF file to: tax purposes have not changed during the preceding 12
[email protected] months.
http://www.bcm.edu/star/ 13–15. The average number of copies of each issue
during the preceding 12 months is (A) Total number of
Baylor College of Medicine is an Equal Opportunity/Affirmative Action/ copies printed: 57,397; (B) Paid circulation: 52,808;
Equal Access Employer (1) Paid/Requested outside-county mail subscriptions
stated on form 3541: 46,515; (2) Paid/Requested
Research Leaders and Principal in-county subscriptions stated on form 3541: 0; (3)
Investigators Sales through dealers and carriers, street vendors,
counter sales: 6,288. (4) Other classes mailed through
The Agency for Science, Technology and Research USPS: 5; (C) Total paid cir¬culation: 52,808; (D) Free
(A*STAR) is Singapore’s lead public sector science distribution: samples, complimentary, and other free
and technology agency. A*STAR has a strategic R&D agenda, driving use-inspired copies: 3,275; (1) Outside-county as stated on form
basic research, spearheading economic growth and advancing social well-being 3541: 2,776; (2) In-county as stated on form 3541: 0;
through scientific discovery, technological innovation and talent development. (3) Other classes mailed through the USPS: 2; (4) Free
A*STAR is consistently ranked world’s top 10 in the Reuters Innovative Insti- distribution outside of mail carrier or other means: 497;
tutions, and top 30 in the Nature Index of Top Government Institutions for the (E) Total free distribution: 3,275; (F) Total ¬distribu-
last 5 years. It averaged about 3,000 international publications annually in recent tion: 56,083; (G) Copies not distributed: 1,315; (H)
yeasrs, and filed over 2,000 patents in the past decade. At A*STAR, we Impact, To-tal: 57,398; (I) Percent paid and/or Re¬quested
Invest, and Inspire, to bring out the best in you. Circulation: 94.2%.
One Organisation, Multiple Careers Actual number of copies of single issue (9/25/2020)
A*STAR is seeking established and potential research leaders and principal
investigators in the following areas: • Artificial Intelligence & data science (e.g. published nearest to fling date are (A) Total number
machine learning) • Decarbonization & urban technology (e.g. CO2 capture and of copies printed: 54,671; (B) Paid circulation: 50,214;
utilization) • Diagnostics & therapeutics (e.g. gene therapy; immunotherapy; med- (1) Paid/Requested outside-county mail subscrip-tions
ical devices) • Neuroscience & cognitive analytics • Other emerging technologies stated on form 3541: 44,328; (2) Paid/Requested
(e.g. food security; industrial biotechnology; quantum engineering) in-county subscriptions stated on form 3541: 0; (3)
We are particularly interested in candidates who are creative and impactful in Sales through dealers and carriers, street vendors,
their works, able to deliver outcomes in a highly competitive environment, and counter sales: 5,881; (4) Other classes mailed through
share our ideals in delivering scientific & technological outcomes and benefits USPS: 5; (C) Total paid circulation: 50,214; (D) Free
to the society and its people. We value cross-disciplinary ideas in a strongly distribution: Samples, complimentary, and other free
collaborative setting. Suitable candidates will be appointed as Senior Scientists/ copies: 3,207; (1) Outside-county as stated on form
Principal Scientists (or Engineers) typically in one of the research institutes of 3541: 2,863; (2) In-county as stated on form 3541:
A*STAR, with responsibilities depending on experience, aptitude and research 0; (3) Other classes mailed through the USPS: 2; (4)
areas. Researchers with leadership experience may also be appointed as centre Free dis¬tribution outside of mail: Carrier or other
directors, programme managers or team leaders. At A*STAR, there are multiple means: 342; (E) Total free distribution: 3,207; (F) Total
career track options, from research to planning and from enterprise development distribution: 53,421; (G) Copies not distributed: 1,250;
to institutional management. A*STAR offers competitive remuneration and (H) Total: 54,671; (I) Percent paid and/or Requested
benefits, commensurate to experience and research qualities and potential. We Circulation: 94.0%.
are seeking research leaders and performers with at least 10 years of relevant I certify that the statements made above are correct and
experience (in academia or industry), but emerging talent with less experience complete. (signed) Bill Moran, Publisher.
are welcome to apply.
We invite all interested applicants to visit bit.ly/joinastar to submit your CV and
contact details by 7 November 2020. All enquiries can be directed to talent@
hq.a-star.edu.sg with the subject title “CAREER” For ongoing news and latest
research developments, visit www.a-star.edu.sg and https://research.a-star.edu.
sg/print-magazine/.
Independent Research Postdoctoral Fellow & Staff Scientist Positions online @sciencecareers.org
Fellowships Leading to Tenured
Faculty Positions at the One Brave Idea and the laboratories of Dr. Calum
John Innes Centre MacRae and Dr. Rahul Deo are seeking Postdoctoral
Fellows & Staff Scientists interested in studying the
The John Innes Centre (JIC), Norwich, UK is a world leading centre of excel- earliest molecular and phenotypic changes during
lence in plant and microbial sciences based on the Norwich Research Park. cardiovascular disease development in order to design
We are inviting applications from outstanding researchers who either hold or novel, more efficacious therapies and gain insight into
wish to apply for Independent Research Fellowships [such as a UKRI Future disease mechanisms. One Brave Idea combines
Leaders Fellowship (https://www.ukri.org/funding/funding-opportunities/ innovative participant phenotyping, large scale genetic,
future-leaders-fellowships/), or a Royal Society University Research Fellow- genomic and proteomic characterization, and
ship (http://royalsociety.org/grants/schemes/university-research/)]. Shortlisted mechanistic cell biology to enable novel discoveries and
candidates will be invited to give a seminar at a virtual Fellows Conference, drive their application to clinical practice.
which will be held on 01, 02 and 03 February 2021. During the conference,
you will be able to discuss your proposals, the development of your group The successful candidate would join our international
and your future career plans in depth with JIC Faculty in virtual one-to-one team of data scientists, staff scientists, postdocs and
meetings. MD/PhDs and will be part of the highly inclusive and
collaborative research community at the Brigham and
After the conference, we will select and mentor outstanding candidates in WomenÕs Hospital & Harvard Medical School. We have
writing Fellowship applications and/or offer the opportunity to move existing generous core-funding support and access to state-of-
Fellowships to the JIC. Candidates who win Fellowships will be considered the-art facilities and technology platforms.
for transfer onto tenure-track at 3 years, and if transferred, for tenure at 5
years. Considerable additional resources will be provided to Fellows by the WeÕre seeking candidates who are interested in
Centre. For further information, contact [email protected] combining interdisciplinary approaches to gain further
knowledge on the initial events happening at the earliest
Further details and particulars can be found at https://www.jic.ac.uk/ stages of cardiovascular disease using an integration of
vacancies/ zebrafish models and mammalian fixed and fresh
samples, including those collected through a high-
To apply: please e-mail a 2-page summary of your research plan, a copy of volume human translational program. The suitable
your CV and arrange for three letters of recommendation to be e-mailed candidates will apply both genetic and high-resolution
to [email protected] by Friday 04 December 2020. Before applying please imaging approaches with metabolic, transcriptional, and
read our Privacy Notice. chromatin profiling assays to elucidate the core
pathological mechanisms at play in cardiovascular
The John Innes Centre is a registered charity (No223852) disease.
grant-aided by the Biotechnology and Biological
Sciences Research Council and is an Equal Qualifications: PhD in molecular and cellular biology,
Opportunities Employer and supports flexible working. molecular genetics, or related field. Previous experience
with zebrafish research is preferred for postdoc
The University of Missouri (MU) invites applications for applicants. Applicants interested in the staff scientist
a tenure-track faculty position in the School of Medicine position should have 3-5 years of academic or industry
(SOM) and Department of Biochemistry. Responsibilities will experience. We are committed to support candidatesÕ
include contributions to medical, graduate and undergraduate transition to academic positions or industry.
education. We seek applications in areas of including gene/
genome regulation, gene delivery and transgenic technology, Interested candidates should email their CV and cover
synthetic biology, medicinal biochemistry, structural biology letter to Evan Wilson Ð [email protected]
and imaging, including development of new molecular markers and reporters.
Of particular interest are areas of fundamental science that might relate to For more information, please visit
emerging strengths in biomedical translation, precision health and/or electron www.onebraveidea.org
microscopy (see below). and
https://macraelab.bwh.harvard.edu/
Qualifcations: An earned doctorate and a record of research in an area of
biomedically-relevant biochemistry that complements departmental strengths.
Applications for this position at the Assistant Professor level are encouraged, but
candidates at a higher level will be considered if there is evidence of continuous
extramural research funding, high impact publications and excellence in teach-
ing and service expected of tenured faculty at a research-intensive university.
About MU and Biochemistry: MU is the fagship campus of the University
of Missouri System. Current strengths of the Department of Biochemistry
include medical biochemistry, plant biochemistry and biophysical, structural and
mechanistic biochemistry. The department forms a bridge between fundamental
molecular science and translation into biomedical and agricultural applications.
Biomedical advances will be anchored by the $225M NextGen Precision Health
Institute, opening Fall 2021, that will house an Electron Microscopy center with
multiple state-of-art instruments.
Application: To apply, visit https://biochem.missouri.edu/open-positions/,
click on Open Positions and submit: (1) a cover letter; (2) a curriculum vitae;
(3) a one-page summary of research accomplishments with two pages of future
plans; (4) a narrative of teaching philosophy and diversity inclusion; and (5)
names and contact information of four references. Review of applications will
begin on November 16, 2020 and will continue until the position is flled. For
information, contact Dr. Jack Tanner, Chair of the Search Committee (tannerjj@
missouri.edu).
The University of Missouri is an Equal Access, Equal Opportunity, and
Affrmative Action Employer. We are fully committed to achieving the goal
of a diverse and inclusive academic community of faculty, staff and students.
We seek individuals who are committed to this goal and our core campus
values of respect, responsibility, discovery and excellence.
online @sciencecareers.org
Science Careers helps you advance
your career. Learn how !
Register for a free online account on Download our career booklets, including
ScienceCareers.org. Career Basics, Careers Beyond the Bench,
and Developing Your Skills.
Search hundreds of job postings and
fnd your perfect job. Complete an interactive, personalized career
plan at “my IDP.”
Sign up to receive e-mail alerts about job
postings that match your criteria. Visit our Employer Profles to learn more
about prospective employers.
Upload your resume into our database and
connect with employers. Read relevant career advice articles from our
library of thousands.
Watch one of our many webinars on
diferent career topics such as job searching,
networking, and more.
Visit ScienceCareers.org
today — all resources are free
SCIENCECAREERS.ORG
PROFESSOR/NOOYI ENDOWED CHAIR online @sciencecareers.org
POSITION ID: F00119P
The College of Engineering, on behalf of the University of Texas at Arlington, invites applications for the Nooyi Endowed Chair position, with starting
employment date early in 2021. This position will be at the full professor level, with an appointment in the College of Engineering or the College of Nursing
and Health Innovation or in the College of Science with possible secondary appointments in a variety of supporting disciplines across the University. It is
preferred that the candidate’s primary appointment be in the Department of Bioengineering. The search is part of the University’s strategic hiring plan within
the four thematic research thrusts of UTA’s Strategic Plan Health and the Human Condition, Sustainable Urban Communities, Global Environmental Impact, and
Data-Driven Discovery. A key objective is to hire faculty members with outstanding qualifications who share the University’s core values of high standards of
excellence in teaching, innovative and collaborative research, and service, combined with fostering an open and inclusive environment.
In this prestigious role, this individual is expected to lead at least one of the university’s efforts to establish research clusters at the intersection of human health
and data-driven discovery. Major emphasis will be placed on creating and realizing potentials for a major bioinformatics-related research and education power-
house in North Texas by collaborating across UTA and with premier medical schools in DFW area. Indications of success will include a stimulating environment
that supports development of joint degree-granting programs in areas such as bioinformatics, medical informatics, or more broadly in biocomputation.
The University of Texas at Arlington is a Carnegie Research-1 “highest research activity” institution. With a student enrollment of close 60,000, UTA is rapidly
becoming one of the largest institutions in the University of Texas System. Guided by its Strategic Plan Bold Solutions | Global Impact, UTA fosters
interdisciplinary research and teaching to enable the sustainable megacity of the future within four broad themes: health and the human condition, sustainable
urban communities, global environmental impact, and data-driven discovery. UTA was cited by U.S. News & World Report as having the second lowest average
student debt among U.S. universities in 2017. U.S. News & World Report also ranks UTA fifth in the nation for undergraduate diversity. The University is a
Hispanic-Serving Institution and is ranked as the top four-year college in Texas for veterans on Military Times’ 2017 Best for Vets list.
The successful candidate will have demonstrated internationally recognized research programs in one or more areas including, but not limited to, big data
analytics, bioinformatics, molecular and computational modeling, and systems biology. The successful candidate will be expected to teach and develop graduate
courses in these areas, supervise graduate students, and serve on departmental, college and university committees. Applicants must have earned a Ph.D. or M.D./
Ph.D. degree in an engineering, science or health care related discipline and have a significant level of research and scholarship accomplishments commensurate
with the rank of full professor with tenure and Nooyi Endowed Chair.
Application Instructions
To apply, applicants should go to https://uta.peopleadmin.com/ and submit their application, including a cover letter, curriculum vitae, statements of research
and teaching objectives, and contact information for at least five references. Questions about the position should be addressed to [email protected].
Review of applications will continue until the positions are filled.
UTA is an Equal Opportunity/Affirmative Action Employer. All qualified applicants will receive consideration for employment without regard to their race,
color, national origin, religion, age, sex, disabilities, or sexual orientation.
myIDP: TENURE TRACK FACULTY POSITIONS
A career plan customized
for you, by you. DEPARTMENT OF PHYSIOLOGY
For your career in science, there’s only one The Department of Physiology at Wayne State University (WSU) School
of Medicine (SOM) (http://physiology.med.wayne.edu) in Detroit,
Features in myIDP include: Michigan invites applications for two tenure-track Assistant/Associate
Professor positions. We seek candidates that employ molecular, cellular
Exercises to help you examine your skills, interests, or systems approaches to explore research interests in cardiovascular,
respiratory or metabolic physiology/pathophysiology and biophysics.
and values. WSU SOM is a state-of-the-art research environment, rated in the top
third of all US Research Institutions by the Carnegie Foundation. The
A list of 20 scientifc career paths with a prediction Department of Physiology, has one of the most active research programs
among the basic science departments at WSU-SOM and is presently ranked
of which ones best ft your skills and interests. #40 out of ~120 Departments of Physiology in the USA. The start-up
package and salary are highly competitive.
Visit the website and start planning today!
myIDP.sciencecareers.org Candidates should hold a Ph.D., M.D. or equivalent from a relevant area.
The selected candidates are expected to establish an extramurally funded
In partnership with: active research program and participate in teaching medical and
graduate students. Please apply to https://jobs.wayne.edu/applicants/jsp/
shared/Welcome_css.jsp, posting 043648 and 043757, by uploading a
curriculum vitae, a detailed future research plan, and names/contact
information of three references. Please submit inquiries with a CV to
[email protected]. Review of applications will begin
after October 30, 2020 and continue until the positions are filled.
WSU is an affirmative action/equal opportunity employer and
encourages applications from women, people of color or other
underrepresented backgrounds.
WORKING LIFE
By John C. Ayers
My last drop
I was looking back in my diary, trying to find clues to why I was struggling with severe insomnia. I
had just begun to take new antidepression medications, and something wasn’t right. I’d experienced
insomnia before, and now I saw the common thread. In both cases, my psychiatrist had started me
on new medications and had recommended that I temporarily stop drinking alcohol. Suddenly it hit
me: The insomnia was a symptom of alcohol withdrawal. I was a functioning alcoholic. It was the
wake-up call I needed, and I’ve been sober ever since. But now I worry that others, facing the stresses
and sadness of the pandemic, may be starting down a similar path. Here’s my cautionary tale.
Alcohol had long been a respite for me from realizing how much I con-
me. During high school and into sumed. Afterward, I was frustrated
college, I drank heavily to cope and confused by my lack of control,
with anxiety. Part of me knew this but I wasn’t quite ready to admit I
wasn’t a healthy approach, but it had a serious problem.
seemed to work. When I discov- That changed a few months later
ered a love of geochemistry, I eased when I looked back on my diary
up on my drinking. On weekdays, and finally, with the help of my
I chose to study rather than go psychiatrist, named my problem.
to the bars. I still enjoyed drink- I immediately committed to absti-
ing on weekends, but it was social nence. The first 6 weeks were espe-
drinking—nothing I was concerned cially hard, but I got through them
about. Throughout grad school and by exercising regularly and spend-
my early years as a professor, I still ing time with my family. I was for-
sometimes drank too much. But it tunate that I was on a sabbatical at
didn’t cause problems. that time, which gave me space to
That started to change roughly focus on my health and recovery. I
11 years into my faculty position, “It was the wake-up call I needed, started to practice mindfulness and
when my father died. Devastated meditation and attend Alcoholics
Anonymous meetings. I also took
and I’ve been sober ever since.”by his loss, I began to suffer from
depression, which in turn led to time to learn about a new scientific
weight gain and sleep apnea. I became chronically sleep discipline and start a new collaboration, which got my cre-
deprived and could no longer think clearly, which made it ative juices flowing again and helped me rediscover my
challenging to meet the intellectual demands of my job. I thirst for research.
suffered from a short temper and strained relationships. Now, nearly 10 years later, I live with less stress, have
I started to self-medicate with alcohol, which reduced my healthier relationships, and am happier and more pro-
anxiety in the short term. But eventually I became so de- ductive. I still suffer from anxiety, but I find that regu-
pressed that I no longer tried to restrain my drinking. I lar exercise and meditation help me cope. When I attend
took up mixology as a hobby and started to drink cocktails conferences—at least, when I used to do so in person,
every night. before COVID-19—I avoid alcohol-centered events or de-
Years passed, and I still felt deeply unhappy. I decided cline the free alcohol tickets. Occasionally, I get odd looks
to see a psychiatrist, who began to treat me for chronic de- from colleagues, but they quickly understand when I tell
pression at first. It took me several more years to recognize them I’m a recovered alcoholic. No one I’ve confided in
I was an alcoholic. has made me feel bad.
An important clue came one morning when I awoke af- If you’re one of the many people who are currently strug-
ter an awards dinner at a conference feeling so hungover I gling in the midst of the pandemic, take it from me: Alcohol ILLUSTRATION: ROBERT NEUBECKER
wasn’t able to co-chair a session that morning as planned; I may make you feel better temporarily, but it’s not a healthy
had to ask colleagues to go on without me. I had vowed not way to cope with stress and anxiety. Ask for help instead. j
to drink too much. But my anxiety got the best of me. After
multiple bottles of wine were placed on the table in front of John C.Ayers is a professor at Vanderbilt University in Nashville,Tennessee.
me, I started to drink heavily, the conversation distracting Send your career story to [email protected].
374 16 OCTOBER 2020 • VOL 370 ISSUE 6514 sciencemag.org SCIENCE
Published by AAAS
The words you are searching are inside this book. To get more targeted content, please make full-text search by clicking here.
Science 16.10.2020
Discover the best professional documents and content resources in AnyFlip Document Base.
Search