The pea aphid complex as a model of ecological speciation



    Corresponding author
    1. INRA, UMR BiO3P (Biologie des Organismes et des Populations appliquée à la Protection des Plantes), 35653 Le Rheu, Cedex, France
      Jean-Christophe Simon, INRA UMR BiO3P (Biologie des Organismes et des Populations appliquée à la Protection des Plantes), 35653 Le Rheu, Cedex, France. E-mail:
    Search for more papers by this author

    1. INRA, UMR BiO3P (Biologie des Organismes et des Populations appliquée à la Protection des Plantes), 35653 Le Rheu, Cedex, France
    Search for more papers by this author

  • Conflicts of interest: the authors have not declared any conflicts of interest.

Jean-Christophe Simon, INRA UMR BiO3P (Biologie des Organismes et des Populations appliquée à la Protection des Plantes), 35653 Le Rheu, Cedex, France. E-mail:


1. Host-specialised races of plant-feeding insects are particularly informative models in the study of ecological speciation, that is, the evolution of reproductive isolation through divergent natural selection. However, within the enormous diversity of phytophagous insects, the mechanisms of ecological divergence have been elucidated in few host race systems.

2. Here we review the literature covering speciation through host-plant specialisation in a well-studied model, the pea aphid complex, Acyrthosiphon pisum, which encompasses numerous biotypes that parasitise different legume host species worldwide.

3. Published results are consistent with ecologically promoted reproductive isolation. Divergent host-induced selection is pronounced across biotypes, and reflects genetic trade-offs preventing the optimal use of multiple host plants. While these genetic trade-offs may partly explain the unfitness of hybrids between biotypes, hybridisation occurring on plants is also limited by genetically-based host preference, and by selection against migrants that chose unfavourable hosts. The continuum of genetic divergence displayed by 11 races and species of the pea aphid complex suggests that host races constitute an intermediate step in the speciation process, and that host specialisation may indeed lead to complete speciation. Uncertainties remain on the contribution of non-ecological reproductive barriers to biotype divergence and on the physiological and molecular bases of host specialisation.


Speciation in phytophagous insects has earned considerable attention, because the extraordinary diversification of this class of animals may partly result from adaptive processes that reflect their specialisation to different host plants (Berlocher & Feder, 2002; Drès & Mallet, 2002). Plant-feeding insects thus provide us with the opportunity to assess how natural selection promotes divergence, and contributes to the Earth's biodiversity. The evolution of reproductive isolation as a result of divergent selection acting on populations in contrasting environments (here, host plants) is defined as ‘ecological speciation’ (Schluter, 2001; Rundle & Nosil, 2005). Although ecological speciation draws its essence from Darwin (1859) and its developments from the modern evolutionary synthesis (Dobzhansky, 1937; Mayr, 1942, 1963), quantifying the contribution of divergent selection to reproductive isolation has been carried out rather recently. Cases of parallel speciation, such as repeated divergences of marine and stream stickleback fishes in different regions (McKinnon et al., 2004), allow the influence of ecology and divergence time (drift) in reproductive isolation to become distinguished, and constitute rare evidence for ecological speciation. In the absence of such ideal settings, the specific mechanisms linking traits under divergent selection to pre- and post-zygotic reproductive isolation must be deciphered (Schluter, 2001). Adaptation to feeding on different seeds has been shown to promote divergence in Galapagos finches (reviewed in Grant, 1999), and retinal adaptation to different water depths in cichlids, in conjunction with variations in male nuptial colours, promote premating isolation and rapid adaptive radiation (Terai et al., 2006; Seehausen et al., 2008). In many small parasitic invertebrates, specialisation to different host species may involve reproductive barriers both at the pre- and post-mating stages (Coyne & Orr, 2004; Bolnick & Fitzpatrick, 2007).

Ecological reproductive barriers can develop between geographically isolated (allopatric) populations, but may also evolve between lineages that are not extrinsically prevented from gene exchanges (that is, in sympatry) (Schluter, 2001; Rundle & Nosil, 2005). Therefore, ecological speciation is somewhat independent to the classical modes of allopatric and sympatric speciation (see Coyne & Orr, 2004 for a review on speciation). However, ecological speciation has stronger connections with sympatric speciation, whose theoretical possibility has fuelled intense debates until recent empirical evidence was published (Via 2001; Bolnick & Fitzpatrick 2007). Indeed, models of sympatric speciation aim at identifying the genetic and ecological conditions favouring a linkage disequilibrium between loci under divergent natural selection and those controlling reproductive isolation, in the face of possible recombination (Via, 2001; Kirkpatrick & Ravigne, 2002; Bolnick & Fitzpatrick, 2007). A reduced number of loci (for a given strength of divergent selection) decreases the risk of maladaptive recombination (Fry, 2003). Likewise, if loci under divergent selection also control mate choice (for instance, if mating occurs within the selective environment), premating isolation between ecologically divergent lineages will more readily evolve (Kirkpatrick & Ravigne, 2002).

From an ecological perspective, sympatric speciation can be viewed as ecological speciation not aided by geography, although demonstrating this phenomenon relies, by definition, on geographical patterns of divergence rather than ecological data (Coyne & Orr, 2004). Currently, excluding past allopatry is only possible in particular settings (small and closed environments, e.g. Barluenga et al., 2006; Savolainen et al., 2006), which expose little of the frequency of sympatric speciation in nature. However, the genetic and ecological features of focus taxa, as well as patterns of gene flow, may give insights into the reality of ecological speciation in sympatry.

Phytophagous insects provide good material for such investigations, as some species comprise ‘host races' (Bush, 1969), a term that defines sympatric populations adapted to different host plants and undergoing appreciable gene flow (hybridisation at a rate 1% or more; Drès & Mallet, 2002). Partial reproductive isolation, not just sympatry, is crucial in studying ecological speciation, as fully formed species may be isolated by ecological, behavioural differences and by intrinsic genetic incompatibilities, so that the contribution of ecology in their divergence could no longer be substantiated (Via, 2009). Among many candidate models, only a few insect species have been shown to encompass genetically and ecologically differentiated populations that still entertain appreciable gene flow (Drès & Mallet, 2002; Mallet, 2008). Similarly, the various aspects of ecological speciation, such as the nature of divergent selection, its link to and various forms and levels of reproductive isolation, are understood and demonstrated in few host race systems. Populations of the maggot fly Rhagoletis pomonella, which historically initiated their divergence after shifting from hawthorn to apple (Walsh, 1867), constitute the prominent example (Bush, 1969). Here we focus on another well-studied model, the pea aphid, Acyrthosiphon pisum, a worldwide pest on legumes.

Generalities and historical aspects

The pea aphid, Acyrthosiphon pisum (Homoptera: Aphididae), generally colonises the lower sides of leaves, buds and pods of legumes (Leguminosae), ingesting phloem sap with its stylets. More than 20 legume genera are known to harbour A. pisum throughout the world (Eastop, 1971) (Table 1), but since its host range has not been investigated specifically, potential hosts cover hundreds of plant species. As with most aphids, A. pisum reproduces by apomictic viviparous parthenogenesis in the warm season. The increasing night length in autumn then triggers the production of a single generation of sexual individuals. After fertilisation, females lay recombinant eggs, which constitute the resistant form to winter frost in aphids (Dixon, 1998). A new generation of parthenogenetic females hatches from eggs in early spring. Wingless forms, which compose the majority of aphid colonies, typically spend their lives on the same host plant, but winged parthenogenetic forms produced under high density may colonise new plants. Sexual females are wingless but some lineages can produce either wingless or winged males, or both, depending on their genotypes at a sex-linked locus (Braendle et al., 2005; Knäbe, 1999).

Table 1.  Known host species of the pea aphid complex (Eastop, 1971) and the biotypes they harbour in Western Europe.
Clade*GenusSpeciesAssociated biotypeSubspecies
  1. All plant species belong to the legume family, except Capsella, which only harbours pea aphids under dry conditions (Eastop, 1971)

  2. *Based on Wojciechowski et al. (2004). IRLC, inverted repeat-lacking clade.

  3. Based on Peccoud et al. (2009a).

  4. Obsolete designation based on Eastop (1971) and Müller (1980); subspecies is A. p. pisum unless specified otherwise.

  5. §Based on unpublished data (J. Peccoud).

CesalpinoideaeCassiaunspecified A. p. spartii
GenistoideaeCytisusscopariusSpecies ‘A’A. p. spartii
 Genistamonspessulana A. p. spartii
 Spartiumjunceum A. p. spartii
RobinoidaeaLotuscorniculatusHost race ‘C’ 
  pedunculatusHost race ‘D’ 
  Three other species  
 Securigeravaria§Probable biotype§ 
 Robiniapseudocacia A. p. spartii
IRLCOnobrychisviciifoliaProbable biotype§ 
 Trifoliumpratense, repens, dubium, campestre, resupinatum§Host race ‘E’ 
  Four other species  
 Melilotusalbus, officinalisHost race ‘F’ 
 MedicagolupulinaHost race ‘J’ 
  sativaHost race ‘K’ 
  Two other species  
 Ononisrepens, spinosaSpecies ‘B’A. p. ononis
 LathyruspratensisSpecies ‘I’ 
  Five other species  
 ViciacraccaHost race ‘H’ 
  faba, sativa, hirsuta, tetrasperma§, ervilla§Host race ‘G’ 
  Two other species  
 LensculinarisHost race ‘G’ 
 PisumsativumHost race ‘G’A. p. destructor

The subdivision of the pea aphid into entities according to their host plant precedes any modern theory of speciation. Since many related species of aphids are best identified by their specific host range, entomologists probably considered pea aphids feeding on new legume genera as undescribed species. Among the dozens of species names that have been attributed to the pea aphid since its original description by Harris as Aphis pisum in 1776, a few remain as subspecies designations (Table 1). Subspecies generally define geographical populations showing minor morphological differences, but the term does not appear to have a single rigorous definition in aphids (Müller, 1986). For instance, pea aphids from trees and shrubs (Robinia, Cytisus, Spartium, and Cassia, see Table 1) are lumped into the subspecies A. pisum spartii Koch, based on their sole presence on woody plants, and not on shared morphological traits (Eastop, 1971). The subspecies A. pisum destructor Johnson usually refers to lineages growing on Pisum and typically presents winged males and green morphs exclusively (Müller, 1980). Since host plants, wings, and colours are not reliable indicators for species delineation, a recently revised taxonomy proposed that only two pea aphid subspecies (or species) distinguishable by minute and partially overlapping variations in the number of hairs on some of their appendices: A. (pisum) ononis Koch, associated with Ononis, and A. (pisum) pisum, feeding on the remaining host range (Blackman & Eastop, 2006). Recent studies have identified 11 host-associated races and species of the pea aphid complex (Table 1). We collectively refer to them as ‘biotypes’, a generic term that applies to any degree of genetic or phenotypic divergence below or at the species level (Drès & Mallet, 2002).

Besides taxonomical aspects, host adaptation and speciation in the pea aphid were first investigated in its Palaearctic native range by Fritz Paul Müller, who hypothesised sympatric divergence through host shifts in several aphid species complexes (Müller, 1985a). Müller first demonstrated the existence of specialised pea aphid populations on various sympatric wild and cultivated legume species (Müller, 1962), and their separation by partial post-zygotic barriers (Müller, 1971). In the light of more recent works, we evaluate the hypothesis of speciation by host specialisation in the pea aphid complex. Specifically, we address three main questions: (1) does divergent selection act against gene flow, (2) does it reduce initial gene flow (hybridisation), and (3) has speciation become complete in the pea aphid complex?

Divergent selection in pea aphid biotypes

Pronounced host specialisation and fitness trade-offs

Parthenogenetic lineages of pea aphid biotypes typically suffer dramatically reduced longevity, biomass and fecundity when transferred on ‘alternate’ hosts (i.e. plant species favourable to other biotypes), as measured under laboratory conditions (Müller, 1962; Via, 1991b; Ferrari et al., 2008; Peccoud et al., 2009a). Pea aphids from North American red clover (Trifolium pratense) and alfalfa (Medicago sativa) constitute the canonical example of this pronounced host specialisation (Via, 1991b) (Fig. 1). Variations in performance (fitness) across host species are mostly genetically based and distributed between, rather than within biotypes, as evidenced by strong effects of biotype × test plant interactions on performance (Via, 1991a; Ferrari et al., 2008). Consistently, performance is mildly affected by conditioning pea aphids on alternate hosts (Via, 1991a; McLean et al., 2009) or on a common favourable medium (Ferrari et al., 2008; Peccoud et al., 2009a). Divergent selection in pea aphids, through their specialisation to different host species, is thus clearly established and very strong, with fitness reductions on alternate hosts often exceeding 50%.

Figure 1.

Mean fitness of pea aphid lines when tested on alfalfa and red clover. Circles represent means of subpopulations collected from two alfalfa fields (A1 and A2); squares are means of subpopulations from red clover (C1 and C2). Modified from Via (1991b).

However, documenting divergent selection at the population level provides no evidence for its role in speciation. Adaptation to different environments may be a consequence of reproductive isolation, rather than a cause (Futuyma & Moreno, 1988). Here, ecological specialisation would simply reflect linkage disequilibrium among ecologically relevant mutations that have accumulated between more anciently diverged populations (Fig. 2A). Ecological speciation generally involves fitness trade-offs, that is, constraints against the optimal use of multiple environments by the same genotype (Fig. 2B) (reviewed in Schluter, 2000). Put differently, an adaptation to a given environment is detrimental in alternate ones, so that no generalist genotype could evolve by recombination. Here, divergent selection is seen at the gene level. It may express the antagonistic effects of alleles in two environments and is defined as antagonistic pleiotropy. Alternatively, close linkage between alleles with opposite fitness effects across environments could constitute a fitness trade-off, provided it escapes from recombination. In pea aphids, genetic trade-offs in host use could be investigated through the genetic mapping of performance of F2 hybrids between North American populations from red clover and alfalfa (Hawthorne & Via, 2001). This analysis has revealed several groups of antagonistic or closely linked quantitative trait loci (QTL) with opposite effects on host performance, constituting a rare documented example of genetic trade-off in performance across two environments.

Figure 2.

The genetics of ecological specialisation to either of two environments ‘square’ or ‘round’. Vertical grey rods represent chromosomes, and horizontal black segments represent alleles (mutated genomic regions) affecting performance (fitness). A ‘+’ sign denotes the beneficial effect of an allele (relative to the other allele) on performance in the corresponding environment, square or round. A ‘–’ sign denotes a deleterious effect. (A) If large genomic regions affect performance in one environment only, a generalist genotype may evolve by recombination. (B) A genetic trade-off in fitness across environments will prevent the evolution of generalist genotypes, because adaptations to both environments cannot merge in the same genome. This impossibility may result either from antagonistic pleiotropy or close linkage (see text).

When considering the pea aphid host range as a whole, the definition of trade-offs with regards to the ability to use different host species becomes more intricate. Does an adaptation to a particular host species reduce performance on all other hosts of the complex? It has been known at least since Müller (1962), and confirmed in recent studies (Ferrari et al., 2008), that growth on Vicia faba (the broad bean), and possibly other annual Vicia (J. Peccoud, pers. obs.), is generally higher than on the natal plants themselves. Universally favourable host plants may counteract ecological divergence on other plants and promote gene flow (Ferrari et al., 2008). However, before reaching such a conclusion, one needs to consider the possibility of hybridisation on these supposedly universal hosts, which requires assessing other ecological factors of divergent selection (see hereafter). With the exception of universally favourable hosts, negative correlations between performances on natal and alternate plants are generally observed between biotypes (Ferrari et al., 2008; Peccoud et al., 2009a). However, these negative correlations simply illustrate the poor performances on most alternate hosts and do not demonstrate the existence of genetic trade-offs in host use. Indeed, deleterious mutations with no effect on the home plant (Fig. 2A) would yield the same results.

The physiological bases of fitness trade-offs in pea aphids have not yet been identified, and are far from evident. Although trade-offs certainly prevent biotypes from converting phloem sap of different hosts into offspring at a rapid rate, it is unknown whether particular nutrients or plant defences condition host specialisation. These uncertainties contrast with identified phenological or morphological adaptations in other models where fitness trade-offs are more easily understood. For example, such trade-offs affect the diapause length of Rhagoletis pupae developing in different fruits (Filchak et al., 2000), rostrum size matching fruit thickness in host-associated populations of the soapberry bug (Carroll et al., 1998), or the beak shape in Geospiza finches feeding on various seeds (Grant, 1999). In the pea aphid, it is suggested that amino acid composition of the host plant's sap may influence performance, but the physiology of host specialisation remains largely unresolved (Sandström & Pettersson, 1994; Douglas, 2003).

Divergent selection and hybrid unfitness

Hybrid unfitness (post-zygotic isolation) is central in ecological speciation, as it may represent the action of divergent selection against gene flow. In pea aphids from different host affiliations, first-generation hybrids typically demonstrate intermediate performance on their parents' home plants, as measured during the parthenogenetic phase (Müller, 1971; Via et al., 2000) (Fig. 3A). Hybrid performance remains highly variable among the few crosses that have been undertaken (Müller, 1971), making it difficult to quantify the importance of post-zygotic factors in overall reproductive isolation between biotypes. Moreover, hybrid performance has never been tested in the sexual phase. With this limitation in mind, post-zygotic isolation appears to be strong enough to maintain high genetic differentiation despite significant hybridisation in the field (up to 9% of outcrossing; Peccoud et al. (2009a).

Figure 3.

(A) Mean individual lifetime fecundity of two specialist genotypes, and of their F1 hybrids, from the alfalfa and red clover host races of North American pea aphids (adapted from Via et al., 2000). (B) Possible components of hybrid unfitness (see text). Symbols have the same meanings as in Fig. 2, squares correspond to clover and circles to alfalfa.

How much of this hybrid unfitness can be explained by ecological factors (Fig. 3B)? Under a hypothesis of ecological post-zygotic isolation (Schluter, 2001), the antagonistic effects of alleles inherited from both parents would explain hybrid unfitness. Another possible cause would be the combination of mutations with deleterious effects on the parents' alternative hosts, but with no effect on their home plants. Under a hypothesis of non-ecological, or intrinsic, post-zygotic isolation, genetic incompatibilities with no relation to host use would affect hybrids in any environment. Here, mutations fixed at different loci would negatively interact [Dobzhansky–Müller incompatibilities; see Coyne and Orr (2004) for a review].

These are the main possible mechanisms of hybrid unfitness in pea aphids (without excluding other less common causes), but no dedicated study has focused on assessing their respective contribution to post-zygotic isolation. However, hybrids between North American host races would definitely suffer from the combinations of antagonistic alleles (Fig. 3B), which have been identified by quantitative genetics (Hawthorne & Via, 2001). Müller (1971) also reported normal growth of F1 hybrid lines on the universal host Vicia faba, despite extremely poor performance during the parthenogenetic phase of the same lines on their parents' hosts, T. pratense and C. scoparius. This observation suggests a strong ecological component in hybrid unfitness between the two host-associated lineages. In the pea aphid complex in general, we may probably exclude the combination of deleterious-only mutations in hybrids (Fig. 3B) as a sole cause of hybrid unfitness. As explained in Fig. 2A, deleterious mutations on alternate hosts would constitute a very inefficient barrier to gene flow, since higher generation hybrids could be purged from deleterious alleles and perform well on several host plants, contributing ultimately to population fusion into a generalist population (Berlocher & Feder, 2002). The high genetic differentiation measured between many host races despite substantial hybridisation (Peccoud et al., 2009a) argues against such a scenario.

In summary, the available results are consistent with ecological selection acting against hybrid pea aphids, but quantitative assessments of the ecological component of hybrid unfitness are lacking.

Multifarious divergent selection

It is proposed that a greater number of niche dimensions subject to divergent selection facilitate ecological speciation (Rice & Hostert, 1993; Nosil & Sandoval, 2008). Is divergent selection in pea aphids only promoted by the nutritional quality of their various hosts? In phytophagous insects, a common ecological factor of divergent selection, other than the nature of the food resource, is host phenology (Berlocher & Feder, 2002), which conditions resource availability and the insect's life cycle. Fruiting phenologies of apple and hawthorn thus select for different diapause lengths in parasitic pupae of the apple maggot, Rhagoletis pomonella (Filchak et al., 2000). In pea aphids, perennial hosts constitute permanent food resources and their phenologies less likely select different and incompatible parasite life cycles. One biotype however, exploits several annual host plants, in particular Pisum sativum and annual Vicia sp. (including Vicia faba), which may be unavailable during the cold season. Notably, this biotype is distinguishable from others by a very high frequency of winged males, which may disperse to perennial vetches (Vicia spp.) for reproduction (Müller, 1980; Frantz et al., 2009b). The locus determining male wing polymorphism in A. pisum (Braendle et al., 2005) may be under divergent selection, if winged males are less fit than wingless males in biotypes on perennial hosts, but are selected for alternation between summer and winter hosts.

Another important and peculiar aspect of ecological specialisation in pea aphids is their bacterial symbionts. In addition to the obligatory endosymbiont of aphids, Buchnera, the pea aphid harbours different types of facultative bacterial symbionts that are vertically inherited with high fidelity (Chen et al., 1996; Fukatsu et al., 2001; Moran et al., 2005). In that perspective, symbionts may be associated with genetic loci rather than a niche dimension. These bacteria influence the pea aphid's ecology, for instance by improving resistance to natural enemies (Oliver et al., 2003; Scarborough et al., 2005), and most notably, their prevalence in aphid lineages strongly differs across host-plant species (Leonardo & Muiru, 2003; Simon et al., 2003; Ferrari et al., 2004; Frantz et al., 2009a). Host association is particularly strong with the symbiont Regiella insecticola, which prevails in pea aphids from Trifolium species, and is rare elsewhere. The antibiotic suppression of R. insecticola was shown to critically alter performance of a Japanese genotype on white clover Trifolium repens, while performance on Vicia sativa, a seemingly universal host, remained unaffected (Tsuchida et al., 2004). Similar experiments conducted on Californian populations infesting white clover (Leonardo, 2004), and injection of R. insecticola in lineages initially deprived of this bacterium (Ferrari et al., 2007) did not correlate with the hypothesis that R. insecticola strongly controls host specialisation. For R. insecticola or any bacterial symbiont to induce divergent selection on pea aphids in the same way a locus operates (Fig. 2), it would have to confer an antagonistic fitness across different habitats. So far, no such effect has been demonstrated.

To summarise, beside the nature of the food resource, no other niche dimension has been clearly demonstrated as an agent of divergent selection in pea aphids associated with different host ranges.

From a theoretical viewpoint, the number of loci under divergent selection, rather than niche dimensions, influences the probability of ecological speciation in the face of gene flow (reviewed in Coyne & Orr, 2004; Bolnick & Fitzpatrick, 2007). In this context, divergent selection in pea aphids can be seen as ‘multifarious’, since QTL mapping analysis of fecundity on red clover and alfalfa have delineated several loci controlling host performance (Hawthorne & Via, 2001). A higher number of loci for a given overall strength of divergent selection is more prone to recombination (Fry, 2003), but is more likely to promote premating isolation among specialised population as a correlated response (Nosil et al., 2009).

Linking divergent selection to premating isolation

Premating isolation between lineages bearing different ecological adaptations is crucial in maintaining genetic and ecological divergence in sympatry, by preserving sets of co-adapted alleles that would otherwise recombine (Felsenstein, 1981). As premating isolation reduces initial gene flow ‘upstream’ (before fecundation), it may account for the main fraction of the overall reproductive isolation (Coyne & Orr, 2004). In aphids, as in other phytophagous insects that mate on plants, host specialisation will directly promote premating isolation through spatial isolation. In the pea aphid, which comprises only plant-feeding stages (except eggs), host specialisation could induce pre-mating isolation in two ways. One is the preference for different plants; the other is selection against immigrants that colonise unfavourable plants (Nosil et al., 2005).

Discrimination between host species in aphids essentially occurs at plant colonisation by winged morphs, which bear more developed sensory organs for host localisation (Powell et al., 2006; Lushai, 2002). Winged aphids can then initiate fast growing colonies, mostly composed of highly productive wingless forms (Dixon, 1998). In laboratory trials, winged and wingless forms of the pea aphids tend to reject alternate hosts (Caillaud & Via, 2000), even after conditioning on such hosts (Müller, 1985b) or on the universal Vicia faba (Ferrari et al., 2006). Variations in plant preference/rejection have thus an important genetic basis and correspond well to variations in host performance. Indeed, genetic variations in host choice are mostly distributed between host-associated populations, though phenotypic variations can be observed within the same biotype, and even the same genotype (Ferrari et al., 2006) leaving possible the accidental colonisation of unfavourable hosts.

In pea aphids, the overall good correlation between preference and performance on the same plant species denotes the link between divergent selection and premating isolation. This link is obviously adaptive, but is also favoured by a particular genetic architecture. The QTL mapping analysis of host acceptance and performance in North American populations (Hawthorne & Via, 2001) suggests close linkage between loci controlling both traits and/or a shared genetic determinism (pleiotropy). Such pleiotropy would be compatible with host detection and feeding behaviours of winged pea aphids, as they do not identify their home plant from a distance. Indeed, winged forms reject alternate hosts only after probing their parenchyma for several minutes (Caillaud & Via, 2000), following the typical behaviour of aphids on non-host plants (Powell et al., 2006). Nutrition on alternate hosts can be artificially triggered by epidermal extracts from the home plant (Del Campo et al., 2003), suggesting that food stimulants are used as chemical cues for host selection. Since parturition in aphids has been shown to rely on such stimulants (Powell et al., 2006), host recognition and adaptation could depend on the same molecules and, by extension, the same loci. Unfortunately, the genetic basis of differential sensitivity to food stimulants across biotypes of the pea aphid has not been quantified, as their feeding behaviour may have been conditioned by rearing on one of the proposed plant species (Del Campo et al., 2003).

As air currents may carry winged aphids over many kilometres (Dixon, 1998; Loxdale & Lushai, 2007), the dispersal of pea aphids, although not directly estimated, likely exceeds the distance between legume species. Host choice behaviour would thus strongly determine host specificity. Performance tests and genetic assignments of field-collected individuals indeed correlate with preference assays, and reveal that a small but significant fraction of aphids (of the order of 10%) from a given plant species are ‘migrants’, which may have developed on other hosts (Via, 1999; Ferrari et al., 2008; Peccoud et al., 2009a).

With the correlation between host performance and host choice being incomplete in the field, counter-selection by unfavourable plants comes into play. Contrary to other host race models in holometabolous insects, where adults use plants for oviposition and not for nutrition (reviewed in Drès and Mallet, 2002), host adaptation in pea aphids can strongly promote premating isolation. Moreover, selection by host plants is exacerbated by several generations of parthenogenetic reproduction separating host colonisation in spring and summer, and sexual reproduction in autumn. Its effects could be assessed by temporal field studies, which monitored increasing differences in host specialisation (Sandström, 1996) and in allelic frequencies (Via et al., 2000) between populations from alfalfa and red clover fields throughout the season of parthenogenetic reproduction.

On the perennial legume hosts, sexual reproduction of A. pisum may occur on the same plant as parthenogenetic reproduction (Eastop, 1971). However, premating isolation in pea aphid biotypes has not been directly monitored (that is, during the sexual phase), not even under laboratory conditions. It is only estimated that host preference and selection against immigrants account for 97% of reproductive isolation between North American populations on red clover and alfalfa (Nosil et al., 2005), an estimate that may be extrapolated to other biotypes showing comparable host specialisation (Ferrari et al., 2008; Peccoud et al., 2009a). Uncertainties still remain on the host fidelity of males that seldom feed when they are adults (J. Bonhomme, pers. comm.) and could promote higher hybridisation than predicted by female host specialisation. More importantly, the contribution of host specialisation to premating isolation is measurable only if other premating barriers, including behavioural isolation, are partial or non-existent (Coyne & Orr, 2004). In eight Western European biotypes of the pea aphid, there is good evidence that premating isolation is incomplete, as hybrids are detected at frequencies varying from 3% to 9% depending on biotypes (Peccoud et al., 2009a). Furthermore, the frequencies of unfit hybrids would decline throughout the season, thus underestimating the initial outcrossing rates. Evidence for mating between pea aphid biotypes, despite moderate migration between host species, demonstrates that host fidelity constitutes a fundamental reproductive barrier. The importance of host specialisation in the premating isolation of pea aphids appears even greater if we consider that reproductively compatible immigrants on a given host species may come from a broad array of hosts.

Levels of gene flow and genetic divergence: on the way to speciation?

Advanced stages of divergence in the pea aphid complex

During the last decade, population genetic studies conducted using different types of nuclear markers on pea aphids from Eastern USA (Via, 1999; Via & West, 2008) and Western Europe (Simon et al., 2003; Frantz et al., 2006; Peccoud et al., 2009a) have revealed significant genetic differentiation among populations from different cultivated and wild plants, with comparatively reduced geographical structure on the same host plants. These results support reproductive isolation between populations adapted to different host species, but their levels of gene flow (hybridisation) and stages of speciation (host races or species) have long remained unclear (Drès & Mallet, 2002). Recent assessments of hybridisation rates in the field (Peccoud et al., 2009a) allowed at least eight host races within a single species to be defined in western Europe, in addition to three probable species within the complex (Table 1 and Fig. 4A). Although the levels of effective gene flow and overall reproductive isolation of the European host races remain undetermined, the correlation between genetic differentiation and hybrid proportion (Fig. 4A) suggests advanced divergence stages. Differentiation starts just below 50% (FST) in the less divergent biotypes, while no hybrid is found when it exceeds 70%. In addition, the absence of strong geographical structure within Western European populations (Peccoud et al., 2009a) indicates that effective gene flow between sympatric biotypes is lower than the migration levels between populations that are hundreds of kilometres apart on the same hosts.

Figure 4.

(A) Eleven Western European biotypes of the pea aphid (named from A to I, see Table 1) depict a continuum of divergence where hybridisation negatively correlates with genetic differentiation. Error bars represent 95% confidence intervals in hybrid proportion. (B) Pairwise ecological differentiation between biotypes, measured as differences in performance (growth in biomass) when tested on both host plants, does not correlate with genetic differentiation. Plots are modified from Peccoud et al. (2009a).

Ongoing speciation in the pea aphid complex

While studies conducted below the species level allow the identification of ecological and genetic factors promoting reproductive isolation (Via, 2009), they suffer the shortcoming that the actual focus of the work might not be speciation itself, but maintained polymorphism or even population fusion (Nosil et al., 2009). Obviously, any future course of evolution remains speculative, but two recent studies on pea aphids offer partial answer to this question.

Genetic divergence between American host races is notably higher in the chromosomal regions around the QTL controlling host acceptance and use, supporting very limited introgression (effective gene flow) at genomic regions linked to loci under divergent selection (Via & West, 2008). Here, host specialisation is confirmed to promote inter-population divergence, and to limit the effective inter-race recombination rates in large portions of the genome.

In Western European pea aphids, genetic differentiation does not suggest any clear boundary between the two regimes of divergence in sympatry- beyond the species level and below (Fig. 4A). Likewise, available biological data do not outline the different levels of divergence observed in the pea aphid complex. Proposed species on Cytisus scoparius and Ononis (species ‘A’ and ‘B’, Fig. 4A) readily hybridise under laboratory conditions, as do host races (Müller, 1971). Also, the most differentiated biotypes (species) do not show higher degrees of host specialisation (Fig. 4B). Consequently, a dynamic process of ongoing speciation or population fusion, in which biotypes occupy different stages, more readily explains the observed continuum of genetic differentiation.

A scenario of ongoing fusion contradicts the lack of correlation between ecological and genetic divergence (Fig. 4B), as one would expect less differentiated host races to become ecologically closer in the process. Under a scenario of ongoing divergence, the same absence of correlation would suggest that complete reproductive isolation does not involve increased divergent selection by host plants. However, several mechanisms may explain the completion of speciation. The pronounced host specialisation of races might have a constraining effect gene flow to the point that further genetic and phenotypic variations promoting complete reproductive isolation would accumulate via drift and local (not necessarily divergent) selection (Rice & Hostert, 1993; Berlocher & Feder, 2002). At the genomic level, such evolution would occur through ‘divergence hitch-hiking’, i.e. protected polymorphism in large regions surrounding loci controlling host adaptation and ecological reproductive isolation (Via & West, 2008; Via, 2009). In addition, hybrid unfitness may promote non-ecological premating barriers during the final stages of sympatric divergence, such as behavioural isolation, under a process known as reinforcement (reviewed in Servedio and Noor, 2003). Such a process may explain the absence of hybrids in the most differentiated pea aphid biotype associated with Lathyrus pratensis, despite very incomplete habitat isolation (Peccoud et al., 2009a).

Finally, recent phylogenetic data suggest a plausible scenario of post-Pleistocene diversification of the pea aphid complex, possibly promoted by climate warming and Neolithic agriculture, which may have increased the availability of potential host plants (Peccoud et al., 2009b). The recent origin of the pea aphid complex seems more compatible with ongoing radiation than population fusion, as the environmental conditions favourable to ecological divergence may not have drastically changed today. However, the timing of diversification does not exclude past episodes of allopatric divergence in pea aphid biotypes.


The pea aphid complex of host races and species presents many features suggesting repeated events of ecological speciation by specialisation to different host species (Table 2). In North American host races adapted to red clover and alfalfa, genetic divergence is higher in the genomic regions linked to loci underlying trade-offs in host use. In the pea aphid complex in general, comparable levels of host specialisation as in North American races suggest strong divergent selection and are consistent with the observation of many host-associated biotypes. Host specificity, mediated by strong differential preference and performance, largely contributes to premating isolation, limiting hybridisation that would otherwise certainly occur at high rates on shared host plants. And most remarkably, the pea aphid complex offers a unique overview of the process of speciation in the presence of gene flow, which perhaps reflects the imperceptible evolution of reproductive isolation. Such a setting provides us with the rare opportunity to assess the contribution of ecology to reproductive isolation at different stages of the speciation process. This involves testing for non-ecological reproductive barriers, which have constantly been overlooked in previous studies.

Table 2.  Main observations relevant to ecological speciation in the pea aphid complex.
Category of observationObservationReferencesRemark
  1. References: 1, Müller (1962); 2, Via (1991b); 3, Ferrari et al. (2008); 4, Peccoud et al. (2009a); 5, Hawthorne and Via (2001); 6, Via and West (2008); 7, Müller (1971); 8, Via et al. (2000); 9, Dixon (1998); 10, Via (1999); 11, Caillaud and Via (2000); 12, Ferrari et al. (2006); 13, Simon et al. (2003); 14, Frantz et al. (2006); 15, Peccoud et al. (2009b). Dixon (1998) is not specific to the pea aphid, but general characteristics of aphids are likely applicable to its biology.

Divergent selection
 Host specialisationGeneral across the complex1–4 
 Biotype × host plant interactions affect performance2, 3Indicates divergent selection between biotypes
 Genetic trade-offs in host adaptation (American host races)5Indicates divergent selection between alleles
 Traits under selectionFeeding performance1–4Divergent selection acts on most stages
 No other agent of divergent selection clearly identified  
 Moderate number of loci control host specialisation (American host races)5, 6Risk of maladaptive recombination
 Strength of selectionSelection is strong on most plants1–4Strong selection coefficients favour divergence
 No fitness reduction on some hosts (Vicia faba)3Possible gene flow in favourable habitats
 Post-zygotic isolationProbably general across the complex7, 8 
 Ecological hybrid unfitnessDocumented in two pairs of biotypes5, 7, 8 
 Effect of divergent selection on genetic differentiationIncreased genetic differentiation near QTL of host specialisation (American host races)6Indicates host-induced genetic divergence
 No correlation between genetic differentiation and the strength of divergence selection between biotypes4Possible non-ecological sources of divergence
Ecological premating isolation
 Link between host adaptation and mate choiceMating occurs on hosts9Loci under divergent selection control mate choice (pleiotropy)
 Colonisers and their offspring feed on host plants9 
 Link between host adaptation and host choicePhenotypic correlation between both traits across the complex10–12Habitat choice (and mate choice) is under selection
 Genetic correlation (close linkage or pleiotropy) in American host races5Reduces the risk of maladaptive recombination
 Known sources of premating isolationHost preference10–12 
 Selection against immigrants on unfavourable hosts7Selection directly enhances premating isolation
 Non-ecological premating isolationUnknown, at most partial in host races4, 7, 8 
 Geographical (macrospatial) isolationUnlikely at present time4 
Progression and history of speciation
 Genetic differentiationHigh between host ranges, much lower between regions (same host)4, 10, 13, 14Indicate advanced stages of speciation
 High in large portions of the genome (American host races)6 
 Gene flow (hybridisation)Moderate in eight host races, not detected in three possible species4 
 Time of divergenceVery recent, thousands of generations15Suggests an adaptive radiation and limits the length of allopatric phases

The second major aspect to address concerns the specific molecular and physiological bases of host specialisation, which remain unresolved. Only a behavioural, physiological, and biochemical analysis of nutrition (e.g. Mutti et al., 2008) could explain why optimal performance on several host plants is prevented, as well as the specific nature of host preference. The annotated genome of the pea aphid will complement the quantitative genetics (Hawthorne & Via, 2001) and population genomics of host races (Via & West, 2008) in order to pinpoint the loci controlling the host specialisation and to test hypotheses of speciation by divergent selection at the gene level.


We thank the organisers, and Hugh Loxdale in particular, of the symposium Insect Evolution Below the Species Level: Ecological Specialisation and the Origin of Species to celebrate Charles Darwin and Alfred Wallace, for inviting us to present this work and offering the possibility to contribute to this special issue.