In this study, we found differences in the swarming behaviour of the molecular forms of An. gambiae
that help to explain their reproductive isolation (Tripet et al. 2001
; Diabaté et al. 2006
). A robust pattern of spatial segregation between swarms was found, revealing distinct form-specific mating units in sharp contrast to the mixed composition of the molecular forms indoors. Our results suggest that spatial swarm segregation in Mali is virtually complete, so it probably contributes strongly to the assortative mating between the forms. This mechanism of reproductive isolation could most easily be effective if females discriminate between swarms similarly to males. Some evidence in support of this hypothesis was obtained from analysis of 27 mating couples collected from swarms in Donéguébougou, all of which were of the same form. These results suggest that females also discriminate between swarms of their own versus the other form, although further study is needed to confirm this hypothesis. If intra-swarm recognition indeed plays a decisive role, it would be difficult to explain the sharp male segregation and the absence of ‘wrong’ females among couples collected from different swarms. Moreover, if males discriminate between swarms, and humans can use ground markers to correctly predict the form of the swarm, it is reasonable that females too can discriminate among swarms, especially because they are expected to incur a higher cost than males for cross mating. Assuming that the fitness of hybrid is reduced in nature, females are supposed to pay the highest cost in the case of cross mating, because they mate only once in their lifetime, whereas males can mate several times.
It is possible that a low rate of cross mating occurs during indoor mating, as suggested by the absence of form recognition in experiments conducted in natural huts (Dao et al. 2008
). Indirectly, it suggests that mate recognition does not operate well outside swarms. Dao et al. (2008)
found direct evidence for indoor mating only in an allopatric M population and proposed that in areas of sympatry, males and females of the S form depart houses before indoor mating starts. The absence of form recognition in tethered female experiments and in indoor mating provides additional evidence against the existence of within-swarm form recognition mechanisms in Mali.
In Burkina Faso, however, the absence of hybrids (Diabaté et al. 2006
), despite the relatively high rate of mixed swarms (approx. 15%), indicates that within-swarm form recognition must operate. Although the expected frequency of mixed swarms (by chance) in Burkina Faso is substantially greater than that observed (Diabaté et al. 2006
), we suggest that at least one additional within-swarm recognition mechanism is involved. Direct studies on the role of chemical and auditory signals will be rewarding (e.g. Gibson & Russell 2006
). The repeated failure of the tethered female experiment in an area of sympatry (Donéguébougou) as opposed to the allopatric M population in Niono (only 300 km away) probably reflects yet another difference in mating behaviour between populations and suggests that the importance of mechanisms of reproductive isolation may vary geographically.
The coexistence of the Bamako and Savanna chromosomal forms within the S molecular form in Mali and not in Burkina Faso (della Torre et al. 2001
) could contribute to this contrast in the mating behaviour between the two populations. However, because 99 per cent of the S form specimens collected from swarms in this study were of the Savanna chromosomal form, which is the only form found in Burkina Faso, this consideration cannot explain the differences. In both populations, the observed barriers operate primarily between Savanna and Mopti chromosomal forms.
Our results stress the role of ground markers as a determinant of swarm segregation in the molecular forms of An. gambiae
. Several studies on swarming insects have found that males aggregate at certain stations (Downes 1969
; Savolainen 1978
; Titmus 1980
; Charlwood et al. 2002
; Yuval 2006
). Consistent with our results, an allopatric S form population in Tanzania swarmed exclusively on bare ground (Marchand 1984
), whereas an allopatric M form population in São Tomé used patterns of contrast as marker (Charlwood et al. 2002
That only one An. arabiensis
male was collected from swarms, despite the fact that this species comprises 10 per cent of the indoor population, suggests that An. arabiensis
mates at specific sites not covered in our survey. Similarly, in Tanzania, no single pure swarm of An. arabiensis
was found in an area where An. arabiensis
and An. gambiae
coexisted (Marchand 1984
); however, swarms of An. arabiensis
could be seen in a village where An. arabiensis
was the only species present. The author concluded that in sympatry, An. arabiensis
changes its swarming behaviour or mates without swarming.
The extent of reproductive isolation within An. gambiae
has been the focus of much debate, although recent theoretical (Lehmann & Diabaté 2008
and references therein) and empirical (Turner & Hahn 2007
) studies have resolved many of the issues. Our data provide evidence that swarm segregation strongly contributes to the reproductive isolation of the two forms. The question remains as to how this isolation mechanism has evolved.
Recent studies suggest that divergent selection between the forms has acted on larval traits (Diabaté et al. 2008
). Larvae of the M form predominate in permanent larval habitats such as rice fields, whereas S larvae predominate in temporary puddles (Diabaté et al. 2002
; della Torre et al. 2005
). Larvae of the M form outperform S larvae in predator-rich habitats (i.e. permanent habitats), whereas S larvae outperform M larvae in the absence of predators (i.e. in temporary habitats; Diabaté et al. 2008
). We propose that M larvae are better adapted to avoid predators than S larvae, whereas the S larvae are better adapted for competition under low predator pressure (Diabaté et al. 2008
). Rundle and Nosil (2005)
, in their review on ecological speciation, stated that speciation is facilitated when genes under divergent selection cause reproductive isolation pleiotropically. The most convincing example is when reproductive isolation evolves as a direct consequence of habitat selection, assuming that individuals mate in their preferred habitat. The molecular forms of An. gambiae
do not mate near their preferred larval habitats, and it is therefore unlikely that the genes under divergent selection in the molecular forms also cause reproductive isolation. We presume that linkage exists between genes conferring adaptive differences at the larval stage and those that influence swarming site selection. The role of divergent natural selection in speciation has been demonstrated in many species, including Bombina
toads. Specifically, Bombina bombina
prefers semi-permanent ponds with a higher density of aquatic predators, rather than the temporary puddles typically used by B. variegata
. Similarly, behavioural differences in predator avoidance were reported between them in accordance with their habitat distribution (Kruuk & Gilchrist 1997
). The authors presumed that the differential adaptation to cope with predation pressure led to differential choice of habitat, and indirectly to preference for alternative breeding habitats.
Although no post-mating reproductive isolation has been found in the laboratory (Diabaté et al. 2007
), the fitness of hybrids in nature has not been tested. It is possible that hybrid inferiority contributes to reproductive barriers between the forms. In ecological speciation, post-zygotic isolation can arise when hybrids are not well adapted to either parental environment and, in effect, fall between the niches (Schluter 2001
; Rundle & Nosil 2005
Uncovering the ecological and genetic mechanisms involved in speciation is key to understanding how biological diversity is generated. Genetic differentiation between the molecular forms of An. gambiae
and its distribution across the genome has been extensively studied, but phenotypic differences between them, the evolutionary forces that generated divergence and the mechanisms that maintain their genetic isolation have only recently been addressed (Lehmann & Diabaté 2008
). Our study provides evidence that swarm spatial segregation strongly contributes to the reproductive isolation between the molecular forms of An. gambiae
in Mali, although this does not exclude the possibility that more than one mechanism of form recognition operates across the range of the molecular forms. This study extends our understanding of the behavioural components of the speciation process and may eventually facilitate the development of new strategies for vector control.