Discussion
Demographic expansion since the last glacial maximum
Our results clearly demonstrated that the demographic history of C. sp. 1 was tightly
coupled with the climate change since the LGM in Taiwan. Combined with results of
significantly negative Fu and Li’s D* and Fu’s FS values, a unimodal mismatch distribution
and a star-like genealogy indicated a historical demographic and spatial expansion of C. sp. 1
(Fu & Li 1993; Fu 1997; Rogers & Harpending 1992; Excoffier 2004). Most notably, results
of Bayesian skyline plot further delineated the tempo and magnitude of the post LGM
population expansion, which began approximately 27,000 years ago and then grew speedily
around 21,000 years ago, right after the LGM in Taiwan (about 24,000 years ago, Liew &
Chung 2001). Finally, the population of this species reached a stable effective population
size around 10,000 years ago, roughly corresponding to the rehabitation of subtropical and
warm-temperate species at lowland Taiwan around 14,000-12,000 years ago (Tsukada 1966,
1967). The tempo of demographic changes of C. sp. 1 coincides relatively well with the
transition of vegetation since the LGM, although our estimate of time for the onset of
expansion somehow predated the LGM, which could in part be due to the deviation of the
molecular clock employed in this study from the true clock.
The postglacial expansion of species at lowland Taiwan has been inferred. For example,
studies of an oak species, Cyclobalanopsis glauca (Hwang et al. 2003) and the Formosan
lesser horsehoe bat, Rhinolophus monoceros (Chen et al. 2006) have documented the genetic
evidence of historical population expansion. However, the historical demography inferred in
these studies was based on the oversimplified demographic models such as the simple
exponential growth model, and/or the tempo of expansion had not been determined.
Moreover, without a link between demographic changes and climate events, it is premature to
conclude that population expansions detected in these studies were driven by climatic changes.
Therefore, the result we presented here thus provides an unambiguous case to demonstrate the
correlation between climatic changes since the LGM and the demographic trend of a
subtropical species.
Because (sub)tropical regions used to be considered as glacial refugia for their relative
climate stability and thus, populations sustained in these areas were assumed to be relatively
intact during the severe glacial periods. However, the huge demographic change
(approximately 1000-fold) of C. sp. 1 represented in this study was not consistent with such
expectation. Although palynological data indicated that an elevational compression of
vegetation occurred during the last glacial period (Tsukada 1966, 1967; Liew & Chung 2001),
this magnitude of population expansion could not be solely explained by the vegetation
decompression since the LGM; we suspected that the population of C. sp. 1 as well as its host
plant and maybe other broadleaf evergreen plants at lowland Taiwan might have been highly
compressed and fragmented during the LGM.
Contemporary population structure
Consistent with that found in other fig-pollinating wasps (Molbo et al. 2004, Zavodna et
al. 2005a, van Vuuren et al. 2006), our results revealed little genetic differentiation among C.
sp. 1 populations according to low pairwise FST values, even over long geographic distances,
such as approximate 350 km in this study. Subtle geographic differentiation had been used
to suggest that fig-pollinating wasps were able to easily disperse over 20-500 km (Molbo et al.
2004, Zavodna et al. 2005a, van Vuuren et al. 2006). As contemporary population structure
was shaped by dispersal ability of a species and its demographic history (reviewed by Bossart
& Prowell 1998), little genetic differentiation of C. sp. 1 in Taiwan might be partially
attributed to the drastic post-LGM population expansion revealed in this study. Results of
mismatch distribution based on the spatial expansion model (Excoffier 2004) suggested that C.
sp. 1 experienced a spatial expansion, which was consistent with the result of no geographic
association among haplotypes revealed by the minimum spanning tree (Fig. 4). During the
process of expansion, C. sp. 1 had migrated into the newly available region with a high
migration rate (M = 9424.773); furthermore, a high effective population size since
approximately 10,000 years ago might prevented further population differentiation of C. sp. 1
(Zane et al. 2006).
On the other hand, results from the first-generation-migrants detection suggested that the
level of ongoing gene flow among populations of C. sp. 1 might be limited, since most
individuals (94%) were inferred as residents. It is consistent with the restricted dispersal of
the dioecious-fig pollinating wasp revealed in several ecological studies (Harrison 2003,
Yokoyama 2003, Harrison & Rasplus 2006).
Nevertheless, the potential for long distance dispersal for C. sp. 1 should be
acknowledged, as one third of putative first generation migrants were assigned to distant
origins (e.g. 350 km) with high probabilities. Moreover, results of phylogenetic analysis
revealed that mitochondrial haplotypes of three C. sp. 1 individuals from Lanyu Island (about
60 km off the southeast coast of Taiwan) were scattered within different clades in both ML
(Fig. 3) and minimum spanning tree (Fig. 4). Thus, our results indicated that these
individuals could have originated from Taiwan or other source populations, and that C. sp. 1
should be able to disperse over 60 km of open water. Long distance dispersal of
fig-pollinating wasps is considered to be achieved passively by wind (Ware & Compton
1994a,b) and some monoecious-fig pollinating wasps have been demonstrated to disperse
over substantial distances by trapping experiments (Harrison & Rasplus 2006); such
mechanism seems equally plausible for C. sp. 1.
However, origins of putative first generation migrants should be interpreted with caution.
For one, there may be intermediate populations from which the individual truly originated,
resulting in the overestimation of dispersal distances. Also, the low level of genetic
differentiation in terms of allelic frequencies among local populations may lead to
underestimation of first generation migrants.
Multiple fig-pollinating wasps of F. septica and its evolutionary implications
We found multiple genetically or morphologically distinctive taxa of fig-pollinating
wasps associated with F. septica on Taiwan and Lanyu Island. Mitochondrial data
demonstrated that the tentative C. b. jucundu was genetically distinct (7.8%) from the C. sp. 1,
and no evidence of gene flow between them was found based on microsatellite genotypes
(data not shown), indicating the presence of reproductive isolation. More intriguingly, C. sp.
2 from Lanyu Island also showed substantial mitochondrial differences from C. sp. 1 (mean
genetic distance = 7.2%) and was difficult to be discriminated from C. sp. 1 morphologically.
This case was similar to the cryptic fig-pollinating wasp species reported in the other studies
(Molbo et al. 2003; Haine et al. 2006), further indicating that coexistence of multiple
fig-pollinating wasp species would be a pervasive pattern across fig-wasp taxa worldwide as
suggested by Molbo et al. (2003). It is well known that morphological traits of different fig
wasp species associated with the same host may be similar due to the constraints of host
morphology (van Noort & Compton 1996), thus leading to the taxonomic difficulty; therefore,
the wasp diversity may be highly under-appreciated by assuming an exclusive one-to-one
host-pollinator relationship. As a result, our findings further challenge the traditional notion
of tight cospeciation between fig trees and their pollinating wasps as well as the theory
concerning the evolution and stability of mutualisms (Molbo et al. 2003; Haine et al. 2006).
In addition, C. bisulcatus in New Guinea was diverged from fig-pollinating wasps of F.
septica in Taiwan by 8.4-8.8 %, indicating that it should be another independent taxon.
Rasplus (1994) has pointed out that F. septica distributed from South-East Asia to Australia
(Queensland) (Zavodna et al. 2005a) is pollinated by different subspecies of pollinators in
different parts across its wide distribution, and suggested that high insularity and natural
fragmentation in South-East Asia may favor differential speciation in pollinators.
Furthermore, fragmentation caused by climatic and vegetational changes during glaciations
may further facilitate speciation of fig-pollinating wasps (e.g. Noonan & Gaucher 2005).
Populations of fig-pollinating wasps might become isolated and fragmented into different
glacial refugia and then reconnected during the interglacial periods. However, their
reproductive isolation might not be completed yet, leading to frequent hybridization between
closely related forms when secondary contact occurred (Brown & Lomolino 1998). Clearly,
future study must seek to identify hidden species of fig-pollinating wasps and interpret their
relationships in an evolutionary framework.
The two coexisting pollinating wasps in southern Taiwan, C. sp. 1 and the tentative C. b.
jucundu, showed distinguishable morphological traits, especially in the body color and the
shape of heads and antennas. With respect to the emergence time, both two pollinating
wasps emerged during the same period with the frequency peaking at two to four days after
the first individual emerged (data not shown). Thus, frequent coexistence of different
species within figs raised the interests concerning the ecological niche differentiation resulting
from competition within a confined and resource-limited environment, or behavior differences
(e.g. day-dispersing or night-dispersing activity, Harrison 2003) for compromising conflicts
arising from coexistence, all require further investigation.