Inferring evolutionary patterns from the biogeographical ...

Blackwell Publishing LtdOxford, UKBIJBiological Journal of the Linnean Society0024-4066The Linnean Society of London, 2006? 2006
89?
541549
Original Article

MUTUALISM AND EXPLOITATION
B. ANDERSON

Biological Journal of the Linnean Society, 2006, 89, 541–549. With 1 figure

Inferring evolutionary patterns from the biogeographical
distributions of mutualists and exploiters

BRUCE ANDERSON*

University of Cape Town, Botany Department, Private Bag Rondebosch, 7701, South Africa

Received 16 September 2004; accepted for publication 15 January 2006

Exploitation may lead to the breakdown of obligate species-specific mutualisms. However, the mutualism between
Roridula (plants) and Pameridea (hemipterans) is often exploited by spiders. The aim of the present study was to
determine when the exploiters became associated with the Roridula-Pameridea mutualism. The phylogenetic and
geographical associations between Roridula and Pameridea are documented and the distribution patterns of Ror-
idula and exploiters are overlaid to see how closely they correlate. A geographical discontinuity in Roridulas’ range
divides both the host plants and associated hemipterans into two sister species so that each hemipteran species is
associated with a different plant species. This suggests that Roridula was associated with Pameridea before frag-
mentation/vicariance events split the genus, allowing allopatric speciation. By contrast, Roridula is only associated
with exploiters in parts of its current range. This suggests that exploiters are unable to traverse the disjunctions in
Roridulas’ distribution and that they only developed associations with the mutualism after vicariance events. It is
hypothesized that Pameridea and Roridula were closely associated for a long period before the invasion of nonmu-
tualists. The absence of associated nonmutualist species may have helped facilitate the evolution of an obligate inter-
action between Roridula and Pameridea. © 2006 The Linnean Society of London, Biological Journal of the Linnean
Society, 2006, 89, 541–549.

ADDITIONAL KEYWORDS: fragmentation – geographical mosaic – mutualism – specificity – sympatry –
vicariance.

INTRODUCTION Bruna, 2000). Unpredictability, mutualism break-
down, and multispecific interactions are thought to be
The evolution of mutualisms often involves the pro- primary factors responsible for the paucity of species-
duction of novel and/or limited resources, which may specific, obligate mutualisms observed in nature
be susceptible to exploitation. In the context of the (Thompson, 1994). Consequently, highly specific
present study, exploiters are organisms unrelated to mutualisms are sometimes thought to evolve in envi-
mutualists but that utilize resources meant for mutu- ronments with low species diversity (Howe, 1984)
alists. These so-called ‘exploiters’ may affect existing where interactions with additional species are mini-
mutualisms through competition, cheating or preda- mized. Alternatively, highly species-specific mutual-
tion, and the accumulation of exploiters may ulti- isms (e.g. fig–fig wasp, yucca–yucca moth, globe
mately cause the breakdown of mutualisms over flowers–fly pollinators) may evolve from previously
evolutionary time (Thompson, 1982). Diverse associ- parasitic relationships (Herre, 1989; Pellmyr, 1989;
ated invertebrate faunas may also dilute strong direc- Addicott, Bronstein & Kjellberg, 1990; Powell, 1992),
tional selection pressures on mutualists and this may which are commonly species-specific (Thompson,
result in a selective regime that is too diffuse for tight 1994).
coevolution between any two species (Hoeksema &
In the present study, the fauna (and their distribu-
*Current address: Department of Botany and Zoology, Stellen- tions) are documented that are frequently associated
bosch University, Private Bag X1, Matieland, 7602, South with the carnivorous plant Roridula (Anderson &
Africa. E-mail: [email protected] Midgley, 2003), which has a species-specific digestive
mutualism and a pollination mutualism with a hemi-

© 2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 89, 541–549 541

542 B. ANDERSON

pteran (Dolling & Palmer, 1991; Ellis & Midgley, 1996; Synaema at the northern localities. The crab spiders
Anderson & Midgley, 2002; Anderson, Midgley & have been shown to compete with Pameridea for prey
Stewart, 2003). The plant family Roridulaceae is a and also consume the hemipterans (Anderson & Mid-
monophyletic group consisting of a single genus and gley, 2002). Populations with high spider densities
two species (Obermeyer, 1970) endemic to South have lower hemipteran densities and, consequently,
Africa. Based on molecular phylogeny, the genera these Roridula are less efficient at obtaining nitrogen
most closely related to Roridula are from the family (Anderson & Midgley, 2002). By affecting hemipteran
Sarraceniaceae and include the carnivorous genera densities, it is thought that spiders may influence the
Darlingtonia, Heliamphora, and Sarracenia (Chase reward quantity for Roridula. The development of the
et al., 1993; Conran & Dowd, 1993; Bayer, Hufford & mutualism between Roridula and Pameridea may
Soltis, 1996), none of which occur in South Africa. The have been hindered by the presence ofother exploit-
taxonomic and geographical isolation from its closest ative invertebrates on Roridula.
relatives suggest that Roridula is a palaeoendemic
genus. Palaeoendemics are systematically isolated The present study aims to document the existing
taxa with relatively ancient origins (Stebbins & Major, ranges of all fauna frequently associated with
1965). They are often ecological specialists and are Roridula. Fauna with broader distributions than the
perhaps ‘on the way to extinction’ (Stebbins & Major, distribution range of Roridula populations suggest
1965). Their distribution range is relictual and their facultatively associated invertebrates because they
endemic condition is the result of a formally more are not entirely dependent on Roridula for their sur-
extensive geographical distribution (Stebbins & vival. However, species with ranges that correspond
Major, 1965). Both species tend to be found in isolated, exactly with the entire Roridula distribution (or part
discrete populations. Populations can vary in size from of the distribution of Roridula) may be obligately asso-
between five and 2000 individuals. Usually, plants ciated with Roridula because this suggests that these
are fairly densely packed, frequently touching one species only occur in close association with Roridula
another. Because plants only recruit after fire, old (alternatively Roridula is obligately dependedt on
Roridula populations (> 10 years) are sometimes that species). Secondly, evidence is gathered to sup-
sparse and plants may be separated by up to 5 m. port/refute the hypothesis that the present Roridula
distribution patterns are due to vicariance events.
The only two species of hemipterans from the genus One of the rules of parasitism of Manter (1955) states
Pameridea (Miridae) are obligately associated with that if the same, or two closely-related species of host
Roridula (one on each species; Dolling & Palmer, exhibit a disjunct distribution and possess similar,
1991), suggesting that they too are palaeoendemics. obligately associated parasite faunas, then the host
Roridula plants capture large amounts of insect prey distribution must have been contiguous at some time
using sticky traps, although they have no digestive in the past. The same rule can be applied to Roridula
enzymes to digest this prey (Marloth, 1910; Lloyd, because it has an association with a hemipteran genus
1934; Ellis & Midgley, 1996). Instead, they have a that is obligately dependent on Roridula (Dolling &
digestive mutualism with Pameridea, which feed on Palmer, 1991). Finally, using phylogeographical pat-
the captured prey (Ellis & Midgley, 1996; Anderson & terns, the study aims to determine when relationships
Midgley, 2002). Hemipterans defecate on the leaves of evolved between Roridula and its associated fauna.
Roridula and nitrogen is absorbed through the thin,
noncontinuous leaf cuticle (Ellis & Midgley, 1996; MATERIAL AND METHODS
Anderson & Midgley, 2002). Roridula plants can
obtain up to 70% of their total nitrogen from hemi- To document the general distribution of Roridula,
pteran faeces (Anderson & Midgley, 2002, 2003) and approximate localities (within 5 km) were used that
hemipterans are unable to successfully complete their were acquired using information from the Bolus and
life cycles and reproduce without being on Roridula Compton Herbarium, Cape Town, South Africa. Exact
plants (B. Anderson, pers. observ.). Thus, the relation- localities were obtained from several sources, includ-
ship between plants and hemipterans appears to be ing searches by the author, biologists, farmers, and
species-specific and apparently obligate for both conservation authorities. Both exact and approximate
species. localities were mapped to obtain a broad distribution
pattern for the genus.
In addition to Pameridea, Marloth (1903, 1910) also
recorded a crab spider (Synaema marlothi) on Rorid- To document the invertebrate fauna, plants from 22
ula dentata populations in the south. In the northern populations were directly observed over the course of a
mountains, Marloth (1925) found a green spider on 4-year period and over most of the range of Roridula.
R. dentata that was probably the lynx spider Peucetia This was carried out by sweep netting, and making
(Oxyopidae) recorded by Dolling & Palmer (1991). close inspections of the Roridula plants for inverte-
Marloth (1903, 1910, 1925) did not find the crab spider brates and their nests. Each population was visited

© 2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 89, 541–549

MUTUALISM AND EXPLOITATION 543

four times per year. By studying three representative (A. Dippenaar, pers. comm.) appeared in the study site
(representative in that between them they contained Pop6, after apparently being absent before (Fig. 1C).
all invertebrates commonly found in pilot studies) They reached densities of approximately one individ-
populations (Pop4, Pop16 and Pop20; Fig. 1A) in dif- ual on every fifth plant. One year later, the spider was
ferent parts of the range of Roridula, it was possible to not to be found. This spider was not recorded in the
observe all the invertebrate fauna that are known to southern R. dentata populations although it was
have close associations with Roridula. Host specificity recorded in the central populations and also the north-
was indirectly determined in these three populations ern populations (spanning disjunction B) where at
by ‘sweep netting’ vegetation during five field trips times it reached densities of several individuals per
over a 4-year period. Taxa that are specific to Roridula plant (e.g. 2.7 ± 3.2 in Pop17). Peucetia was absent
should only be found on Roridula plants. Using a from only a single population (Pop22) in the northern
sweep net, Roridula plants and ‘non Roridula’ plants populations (Fig. 1C). This spider probably has a wide
were independently sampled. In addition, the sweep distribution and the type specimen was found in the
net was also used to sample other glandular or hairy most south-eastern parts of the range of R. gorgonias;
plants (especially Elytropappus scaber) within 1 km of no mention of Roridula was made on the type speci-
Roridula populations. The sweep net was checked for men (A. Dippenaar, pers. comm.). Peucetia were occa-
invertebrates every 1 min, and the numbers of inver- sionally found in sweeps on noncarnivorous plants
tebrates captured (only those that commonly occur on (Table 1). When sweeping was confined to viscid and
Roridula) were noted. hairy plants (not Roridula), large numbers of
P. nicolae were captured (Table 1). They were espe-
RESULTS cially common on the plant E. scaber. When these spi-
ders were placed on Roridula, they were able to
Roridula is geographically divided into two species negotiate their way around the sticky traps.
(separated by approximately 70 km (disjunction A;
Fig. 1A). Roridula gorgonias occupies the southern The crab spider S. marlothi (Thomisidae) was found
and south-eastern mountains. Roridula dentata has a in large numbers on most of the populations in the
disjunct distribution, with the northern-most popula- southern and central range of R. dentata (Fig. 1D).
tions being very isolated from the rest (disjunction B; Sweep netting on noncarnivorous plants (including
Fig. 1A). other hairy and glandular plants) yielded no Synaema
specimens (Table 1). Synaema marlothi is abundant
The distribution of Pameridea perfectly matches on the Roridula plants, often attaining densities of up
the distribution of Roridula (Fig. 1B). Pameridea to seven adults per plant. Nests are usually made in
roridulae was only found in close association the axils of branches, close to the leaves. Spiders were
with R. gorgonias and Pameridea marlothi with observed to move quickly over the leaves to capture
R. gorgonias. Neither species was recorded on any prey ensnared by Roridula. Synaema was also
plants other than Roridula (Table 1). Occasionally, observed consuming Pameridea.
Pameridea were absent from recently burnt Roridula
populations (plants < 7 cm high). However, Pameridea An additional undescribed spider species was found
were always present approximately 1 year after seed- in the northern range of R. dentata (Fig. 1E). The spi-
ling germination, indicating that recolonization is not der could only be identified to familiy level by Dr Ansie
immediate. Large plants sometimes had several hun- Dippenaar: Araneidae (orb-web spiders). The spider on
dred adult Pameridea on them and more than ten-fold Roridula has lost the ability to construct an orb-web
as many nymphs. and instead relies on the sticky plants to catch prey for
them (B. Anderson, pers. observ.). Some other members
No spiders were associated with R. gorgonias dur- within the family also have secondarily reduced orb-
ing the first 3 years of study. However, after the third webs (Filmer, 1991). Spiders do construct nests that are
year of observation, the lynx spider Peucetia nicolae used as retreats and these are made on Roridula

Figure 1. A, map of Roridula distribution where labelled populations represent the populations studied, circles represent
Roridula dentata populations and triangles represent Roridula gorgonias populations. B, map of Pameridea distribution
on the populations studied where ‘Pm’ represents the presence of Pameridea marlothi on Roridula populations and ‘Pr’
represent the presence of Pameridea roridulae on Roridula populations. C, map of Peucetia nicolae distribution on the
populations studied where ‘Pn’ represents Roridula populations with Peucetia. D, map of Synaema marlothi distribution
on the populations studied where ‘Sm’ represents Roridula populations with Synaema. E, map of Araneid distribution on
the populations studied where ‘A’ represents Roridula populations with Araneids. F, map of Sphedanolestes sp. distribution
on the populations studied where ‘S’ represents Roridula populations with Sphedanolestes. A contour key and scale are
provide in Fig. 1A.

© 2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 89, 541–549

544 B. ANDERSON

22 21
20 19
18

Disjunction B 16
14
15
17 12
13

33o

10

11
98

Disjunction A

34o 23

4

7
1

A 18o Pm 20o Pn
Pm Pn
Pm Pn
Pm Pn
Pm

Pm Pm Pm Pn Pn Pn
Pm Pm Pn
Pm Pn
Pm Pn

Pm Pn
Pm Pn
Pm

Pr Pr Pn
Pr
Pr
Pr Pr Pr

B C

Figure 1. Continued

© 2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 89, 541–549

MUTUALISM AND EXPLOITATION 545

A
AA
A

Sm Sm
Sm
Sm
Sm

Sm
Sm

Sm

D E

SS
S

F

Figure 1. Continued

© 2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 89, 541–549

546 B. ANDERSON

Table 1. Catch per unit effort (standard deviation) of Pameridea, an unidentified Araneid, Synaema, Peucetia, and
Sphedanolestes sp. on three different Roridula populations, random shrubs, and viscid plants surrounding Roridula study
sites

Population Species Roridula Random Viscid shrubs

Pop4 Pameridea 175 (53) 0 (0) 0 (0)
Pop4 Araneid 0 (0) 0 (0) 0 (0)
Pop4 Synaema 0 (0) 0 (0) 0 (0)
Pop4 Peucetia 0 (0) 0 (0) 0 (0)
Pop4 Sphedanolestes 0 (0) 0 (0) 0 (0)

Total time Pameridea 10 min 120 min 0 min
Pop16 Araneid 86 (33) 0 (0) 0 (0)
Pop16 Synaema 0 (0) 0 (0)
Pop16 Peucetia 0 (0) 0 (0) 0 (0)
Pop16 Sphedanolestes 6.9 (3.3) 0.1 (0.2) 3.6 (2.8)
Pop16 2.4 (1.7) 0 (0) 0 (0)
Pameridea 0 (0)
Total time Araneid 150 min 150 min
Pop20 Synaema 10 min 0 (0) 0 (0)
Pop20 Peucetia 75 (32) 0 (0) 0 (0)
Pop20 Sphedanolestes 11 (6.4) 0 (0) 0 (0)
Pop20 0.1 (0.2) 3.7 (3.0)
Pop20 0 (0) 0.01 (0.1) 0.01 (0.1)
3.4 (2.8)
Total time 0.2 (0.4) 180 min 180 min

10 min

The total time spent sampling in these habitats is included.
Pop, population.

plants, in the axil of a leaf or small branch. The dis- observ.). Their associations with Roridula appear
tribution of this spider is geographically restricted and extremely transitory and they are apparently faculta-
was found on 80% of the northern populations tively associated with Roridula.
(Fig. 1E). On these populations, it was fairly common,
with up to ten adults per plant. This Araneid was not DISCUSSION
found on plant species other than R. dentata (Table 1).
HOW DEPENDENT ARE ASSOCIATED FAUNA ON
A species of hemipteran belonging to the family RORIDULA?
Reduviidae (Sphedanolestes sp., J. Maxen, pers.
comm.) was also discovered on 60% of the northern Pameridea, S. marlothi, and the undescribed Araneid
populations of R. dentata (Fig. 1F) at very low densi- have to date only been found on Roridula plants sug-
ties. The reduvid was not found on any R. dentata gesting that they are host specific. This supports the
populations from the central and southern regions suspicions of Dolling & Palmer (1991) that Pameridea
(Fig. 1F). Nor was it found on R. gorgonias (Fig. 1F). It is a host-specific hemipteran. The lack of web con-
was observed laying its eggs on the stems of Roridula; struction in the Araneid suggests that this species has
however, the adults were only able to move very slowly secondarily lost the ability to build webs because of its
over the sticky leaves where they hunted prey caught specialized life on Roridula. By contrast, Lynx spiders
by Roridula. Other species of reduvid were frequently (Peucetia) were frequently captured when sweep net-
caught by Roridula. The reduvid was also found occa- ting viscous vegetation (other than Roridula). Peucetia
sionally by sweep netting noncarnivorous plants is most likely an opportunistic species with preadap-
(Table 1). tations that allow them to walk on Roridula plants (as
well as other glandular plants) and exploit the rich
Several other spiders were sporadically found on resources. Similarly, some reduvids (Sphedanolestes
R. dentata including two species of jumping spider sp.) were also captured on noncarnivorous plants, sug-
(Salticidae), another crab spider (Synaema sp. Thom- gesting that they too are only facultatively associated
sidae), a rain spider (Palystes sp., Heteropodidae), a with Roridula. Note that these results are not conclu-
brown button spider (Latrodectus geometricus, Theri- sive as it is often difficult to demonstrate that an
idae). None of these spiders were collected more than organism is not in an area.
three times on Roridula plants and they are fre-
quently found in other localities (B. Anderson, pers.

© 2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 89, 541–549

MUTUALISM AND EXPLOITATION 547

ARE PRESENT RORIDULA DISTRIBUTIONS DUE TO caused the disjunct distributions observed for the
VICARIANCE EVENTS? genus. Roridula gorgonias occurs in seeps and marshy
areas with permanent surface water (Carlquist, 1976;
Roridula distributions are very disjunct with large B. Anderson, pers. observ.), similar to the closely-
distances (50–70 km) separating the different species related Sarraceniaceae. Roridula dentata can be found
and regions within a species. Nevertheless, Pameridea on sandy ‘vlakte’ (flats) that are drier than
occurs on either side of these disjunctions. Disjunction R. gorgonias localities in summer although under-
‘A’ (Fig. 1) forms a species boundary, which can be used ground water seems plentiful (Carlquist, 1976; B.
to define the two Roridula and Pameridea species. Anderson, pers. observ.). In winter these areas are
Thus, there is zero gene flow across this distribution very wet (Carlquist, 1976; B. Anderson, pers. observ.).
gap. The gene flow study of Anderson et al. (2004) also
supports the argument that there is no migration HOW DID FRAGMENTATION OCCUR?
across this barrier. The most parsimonious explana-
tion for this distribution pattern is that the common The fact that speciation has occurred across disjunc-
ancestor of R. dentata and R. gorgonias was once more tion A for both Roridula and Pameridea suggests that
widely distributed and hosted a single Pameridea spe- vicariance occurred a long time ago and cannot be
cies. Large-scale fragmentation possibly allowed allo- attributed to contemporary human disturbance. It is
patric speciation to occur simultaneously in Roridula possible that some areas where Roridula once grew
and Pameridea. For a nonvicariance model to have cre- have become unsuitable for their growth due to cli-
ated these distribution patterns, both Roridula and matic change.
Pameridea would have had to disperse across a gene
flow barrier of (50–70 km) to founder new populations. Until approximately 3.2–2.5 Mya (late Pliocene),
Because available genetic data (Anderson et al., 2004) the Western Cape was characterized by a wetter sum-
suggest a zero gene flow between plant populations mer rainfall (Deacon, Jury & Ellis, 1992). Presently,
separated by more than a few kilometres, it is unlikely the Western Cape is typified by a winter rainfall and
that plant populations on either side of the disjunction hot, dry Mediterranean type summers. Biome shifts
could have been started through founder effects. Not associated with the change in climate may have con-
only would plants have had to disperse widely, but also strained Roridula to high lying areas, which receive
Pameridea would have had to disperse too, and colo- more rain. This may explain the affinity of Roridula
nize the new Roridula populations at the same time. for relatively high lying areas. A similar theory has
Because no migration occurs across disjunction A, a been invoked for the distribution of another rare
nonvicariance model explaining the disjunct distribu- palaeoendemic, Ixianthes retziodes (Scrophulari-
tion patterns can be rejected and, instead, it can be aceae), which has a very similar distribution pattern
proposed that Roridula populations were once much to R. dentata (Steiner & Whitehead, 1993, 1996).
more widespread. Ixianthes is also very closely associated with water
and is only found on the banks of rivers.
Similarly, a large distribution gap (∼50 km) sepa-
rates the northern R. dentata populations from the Alternatively, more recent climatic oscillations
central/southern populations. The genetic structure of could have facilitated vicariance and allopatric speci-
P. marlothi (Anderson et al., 2004) suggests that the ation in Cape plants (Midgley et al., 2001) and Rorid-
Pameridea on either side of this divide are incipient ula. In their climate change model, Midgley et al.
species and that gene flow between the northern and (2001) hypothesized that climatic oscillations within
central populations is absent. Allopatric divergence, the last two million years may have shifted biomes
due to vicariance in Roridula distribution, is the most along a north–south axis, which would also have
parsimonious explanation for the observed genetic forced taxa into topographic refugia. According to
and geographical structure on either side of this dis- their model, fynbos is presently refugial.
junction. Judging from the taxonomic status (real spe-
cies vs. incipient species) of Roridula and Pameridea WHEN DID ASSOCIATIONS WITH RORIDULA EVOLVE?
on either side of ‘disjunction A’ and ‘disjunction B’,
respectively, it can be assumed that ‘disjunction A’ Three distinct distribution patterns can be observed in
occurred before ‘disjunction B’. the associations between Roridula and their fauna.
The first pattern is the perfect correlation between the
Thus, Roridula populations must have been more distributions of Pameridea and Roridula. It is postu-
contiguous at one time. This is a common phenomenon lated above that Pameridea and Roridula speciated
in palaeoendemic species, whose distributions have allopatrically after a single vicariance event. This sup-
normally resulted from vicariance events (Stebbins & ports the hypothesis that Roridula and Pameridea
Major, 1965). Roridula is closely associated with moist were closely associated before the vicariance event
habitats and the shrinking of these habitats may have took place. Alternatively, the distribution patterns

© 2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 89, 541–549

548 B. ANDERSON

have arisen by secondary invasion throughout the limited distributions, it is likely that both the Araneid
present range of the plant. However, this explanation and S. marlothi have had shorter histories of close
requires repeated colonizations and invasions association with Roridula than Pameridea has had.
throughout the range of each species. It is therefore Similarly, Pellmyr & Leebens-Mack (1999) suggest
less parsimonious than the former hypothesis and that cheating lineages of yucca moth are one order of
hence can be rejected. Using similar reasoning, Pell- magnitude younger than mutualist lineages. Marr,
myr (1992) used distribution patterns of Trollius and Brock & Pellmyr (2001) suggest that the age differ-
their obligate fly pollinators (Chiastocheta) to infer ences between these lineages may be because the con-
evolutionary events in the two genera. He showed that ditions required for the evolution of cheating rarely
a monophyletic group of five Trollius were associated occur. Alternatively, the same authors suggest that
with Chiastocheta throughout their ranges, suggest- cheaters commonly arise but only a small fraction per-
ing that the common ancestor of that plant lineage sist. I concur with this and hypothesize that obligate
had been associated with the fly lineage before it mutualisms may need to be particularly well devel-
diversified. oped and stable to support exploitation by nonmutu-
alists. It is possible that associations with exploiters
The second distribution pattern involves Roridula have frequently caused the breakdown of obligate
and two host specific spiders (unidentified Araneid mutualisms that have not had enough time (time
and S. marlothi). Both of these species only occur on without exploiters being present) to evolve a high
R. dentata, with each occupying only a small part of degree of stability. In addition, selection mosaics due
the range of R. dentata. The absence of these species to differing invertebrate faunas may promote coevolu-
from R. gorgonias suggests that associations between tionary ‘hotspots’ and ‘coldspots’; see geographical
Roridula and spiders only evolved after the vicariance mosaic theory (Thompson, 1994, 1998, 1999; Gomulk-
event (disjunction A), which caused Roridula to speci- iewics et al., 2000; Thompson & Cunningham, 2002)
ate. Alternatively, these spiders could have had asso-
ciations with Roridula before the vicariance event and ACKNOWLEDGEMENTS
then gone extinct in all the R. gorgonias populations.
However, this explanation is not as parsimonious as Thanks to the late Dr I. Williams, Vogelgat staff,
the first because it requires repeated extinction events Fernkloof Municipality and Ceres municipality for
in all R. gorgonias populations (without Pameridea allowing me onto their land and for the use of their
going extinct as well). Using this logic, it is tempting facilities. Thanks to Willem Hanekom for sharing his
to suggest that spiders only evolved associations with knowledge on Roridula localities and to Louis
Roridula after the second vicariance event (disjunc- Hanekom and Theunis Hanekom for their hospitality.
tion B) because spider (Synaema and Araneid) distri- I also thank Mr Du Toit and Mrs van der Merwe for
butions are constrained by this disjunction in the allowing me to study Roridula on their properties. The
distribution of R. dentata. CNC was especially helpful in supplying vehicles,
devoting their time and their knowledge (especially
Finally, the distribution pattern of the lynx spider Marius Brand, Charl Du Toit, Pieter Viljoen, Koos and
(Peucetia) is not constrained by either disjunction ‘A’ or Mark Johns). Dr Ansie Dippenaar, John Maxen and
disjunction ‘B’, although it is not present in several of Ian Millar helped in the identification of spiders and
the Roridula populations. This distribution pattern hemipterans. Amrei, Drew, Ad, Arv, Eric and Jeremy
reflects the nonhost-specific nature of this species (i.e. accompanied me on many field trips. Thanks to the
disjunctions in Roridula distribution do not affect the NRF for funding and to two anonymous reviewers for
movement or presence of Synaema). Thus, the distri- their helpful comments.
bution pattern of Synaema does not appear to be influ-
enced by the distribution pattern of Roridula. From REFERENCES
the ‘haphazard’ distribution of Peucetia, it is only pos-
sible to say that this species is unlikely to be obligately Addicott JF, Bronstein J, Kjellberg F. 1990. Evolution of
associated with Roridula. It is difficult to determine mutualistic life cycles: Yucca moths and fig wasps. In: Gilbert
when this loose association first occurred. F, ed. Insect life cycles: genetics, evolution and co-ordination.
London: Springer-Verlag, 143–161.
CONCLUSION
Anderson B, Midgley JJ. 2002. It takes two to tango but
Distribution patterns suggest that Pameridea was three is a tangle: mutualists and cheaters on the carnivorous
closely associated with Roridula before vicariance plant Roridula. Oecologia 132: 369–373.
events caused speciation of the plants. Vicariance of
plant populations and the low dispersal capabilities of Anderson B, Midgley JJ. 2003. Digestive mutualism, an
Pameridea most likely caused cospeciation of the gen- alternate pathway in plant carnivory. OIKOS 102: 221–
era Roridula and Pameridea. Judging from their more 224.

© 2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 89, 541–549

MUTUALISM AND EXPLOITATION 549

Anderson B, Midgley JJ, Stewart BA. 2003. Facilitated Manter HW. 1955. The zoogeography of trematodes of marine
selfing offers reproductive assurance: a mutualism between fishes. Experimental Parasitology 4: 62–86.
a hemipteran and carnivorous plant. American Journal of
Botany 90: 1009–1015. Marloth R. 1903. Some recent observations on the biology of
Roridula. Annals of Botany 17: 151–157.
Anderson B, Olivieri I, Lourmas M, Stewart BA. 2004.
Comparative genetic structures of Roridula and pameridea: Marloth R. 1910. Further observations on the biology of Ror-
cospeciation through vicariance. Evolution 58: 1730–1747. idula. Transactions of the Royal Society of South Africa 2:
59–61.
Bayer RJ, Hufford L, Soltis DE. 1996. Phylogenetic rela-
tionships in Sarraceniaceae based on rbcl and ITS Marloth R. 1925. The Flora of South Africa, Vol. 2. Cape Town:
sequences. Systematic Botany 21: 121–134. Darter Bros.

Carlquist S. 1976. Wood anatomy of Roridulaceae: ecological Marr DL, Brock MT, Pellmyr O. 2001. Coexistence of
and phylogenetic implications. American Journal of Botany mutualists and antagonists: exploring the impact of
63: 1003–1008. cheaters on the yucca-yucca moth mutualism. Oecologia
128: 454–463.
Chase MW, Soltis DE, Olmstead RG, Morgan D, Les DH,
Mishler BD, Duval MR, Price RA, Hills HG, Qui Y-L, Midgley G, Hannah L, Roberts R, McDonald DJ, Allsopp
Kron KA, Rettig JH, Conti E, Palmer JD, Manhart JR, J. 2001. Have Pleistocene climatic cycles influence species
Sytsma KJ, Michaels HJ, Kress WJ, Karol KG, Clark richness patterns in the greater cape Mediterranean region.
WD, Hedren M, Gaut BS, Jansen RK, Kim K-J, Wimpee Journal of Mediterranean Ecology 2: 137–144.
C, Smith JF, Furnier G, Strauss SH, Xiang Q-Y, Plun-
kett GM, Soltis P, Swensen SM, Williams SE, Gadek Obermeyer AA. 1970. Roridulaceae. In: Codd LE, DeWinter
PA, Quin CJ, Eguiarte LE, Golenberg E, Learn GH Jr, B, Killick DJB, Rycroft HB, eds. Flora of Southern Africa,
Graham SW, Barrett SCH, Dayanandan S, Albert VA. Vol. 13. Cape Town: Kirstenbosch, 201–204.
1993. Phylogenetics of seed plants: an analysis of nucleotide
sequences from the plastid gene Rbcl. Annals of the Missouri Pellmyr O. 1989. The costs of mutualism: interactions
Botanical Garden 80: 528–580. between Trollius europaeus and its pollinating parasites.
Oecologia 78: 53–59.
Conran JG, Dowd JM. 1993. The phylogenetic relationships
of Byblis and Roridula (Byblidaceae-Roridulaceae) inferred Pellmyr O. 1992. The phylogeny of a mutualism: evolution
from partial 18S RNA ribosomal sequences. Plant Systemat- and coadaptation between Trollius and its seed-parasitic pol-
ics and Evolution 188: 73–86. linators. Biological Journal of the Linnean Society 47: 337–
365.
Deacon HJ, Jury MR, Ellis F. 1992. Selective regime and
time. In: Cowling RM, ed. The ecology of Fynbos. Cape Town: Pellmyr O, Leebens-Mack J. 1999. Forty million years of
Oxford University Press, 6–22. mutualism: evidence for an Eocene origin of the yucca–yucca
moth association. Proceedings of the National Academy of
Dolling WR, Palmer JM. 1991. Pameridea (Hemiptera: Science of the United States of America 96: 9178–9183.
Miridae): predaceous bugs specific to the highly viscid plant
genus Roridula. Systematic Entomology 16: 319–328. Powell JA. 1992. Interrelationships of yuccas and yucca
moths. Trends in Ecology and Evolution 7: 10–15.
Ellis AG, Midgley JJ. 1996. A new plant-animal mutualism
involving a plant with sticky leaves and a resident hemi- Stebbins GLJ, Major J. 1965. Endemism and speciation in
pteran. Oecologia 106: 478–481. the Californian flora. Ecological Monographs 35: 1–35.

Filmer MR. 1991. South African spiders, an identification Steiner KE, Whitehead V. 1993. Has Ixianthes (Scrophular-
guide. Cape Town: Struik Publishers. iaceae) lost its special bee? Plant Systematics and Evolution
185: 7–16.
Gomulkiewics R, Thompson JN, Holt RD, Nuismer SL,
Hochberg ME. 2000. Hotspots, coldspots, and the geo- Steiner KE, Whitehead V. 1996. The consequences of spe-
graphic mosaic theory of coevolution. American Naturalist cialization for pollination in a rare South African Shrub,
156: 156–174. Ixanthes retziodes (Scrophulariaceae). Plant Systematics and
Evolution 201: 131–138.
Herre EA. 1989. Coevolution and reproductive characteristics
in 12 species of New World figs and their pollinator wasps. Thompson JN. 1982. Interaction and evolution. New York,
Experientia 45: 637–647. NY: John Wiley & Sons, 165.

Hoeksema JD, Bruna EM. 2000. Pursuing the big questions Thompson JN. 1994. The coevolutionary process. The Univer-
about interspecific mutualisms: a review of theoretical sity of Chicago, IL: Chicago Press, 376.
approaches. Oecologia 125: 321–330.
Thompson JN. 1998. The population biology of coevolution.
Howe HF. 1984. Constraints on the evolution of mutualisms. Researches on Population Biology 40: 159–166.
American Naturalist 123: 764–777.
Thompson JN. 1999. Specific hypotheses on the geo-
Lloyd FE. 1934. Is Roridula a carnivorous plant? Canadian graphic mosaic of coevolution. American Naturalist 153:
Journal of Research 10: 780–786. S1–S14.

Thompson JN, Cunningham BM. 2002. Geographic struc-
ture and dynamics of coevolutionary selection. Nature 417:
735–738.

© 2006 The Linnean Society of London, Biological Journal of the Linnean Society, 2006, 89, 541–549