The existence of a continuous array of sympatric, host associated forms - from polymorphisms, through host races with increasing reproductive isolation, to good species - would be strong evidence of a stable route for sympatric speciation via host shift. However, assessing this spectrum of differentiation in sympatric, plant-feeding insects (a group in which sympatric speciation seems particularly likely) is difficult, in part because multiple, ambiguous host race definitions are currently in use, but also because estimates of current gene flow have been made in only a handful of likely cases. Here, we present a clear operational criterion of host races and apply it to 17 putative host race systems. Of these, only two could be unambiguously classified as host races, but a further seven are strong candidates. We conclude that sympatric speciation via host race formation seems likely, but caution that current data still does not rule out a discontinuity in the theoretically stable route from host-associated polymorphism to host-associated species.
The formation of host races - genetically distinct populations separated by host use but connected by gene flow - is an important step in models of sympatric speciation via host shift, which describe how host-specific selection combined with differences in host use can drive sympatric populations to speciate. The study of these taxa has attracted controversy, partly because of association with the historically contentious theory of sympatric speciation (see Tauber and Tauber 1989), but also because of ambiguity in current host race definitions (Diehl & Bush 1984). The literature has been reviewed several times, both by authors who believe sympatric speciation is common in parasites (Diehl & Bush 1984; Strong et al. 1984; Tauber & Tauber 1989; see also Mopper & Strauss 1998), and those who do not (Jaenike 1981; Claridge 1988). Many empirical studies identifying host races have since been published.
In order to reflect the main focus of current work, this review will concentrate on host race formation in phytophagous insects and its role in sympatric speciation. We here
- outline developments in the theoretical study of speciation via host shift
- operationally define the term ‘host race’ by a set of empirically testable criteria
- identify those biotypes so far investigated that meet our criteria, and those that do not, and
- summarize current empirical evidence for speciation via host shift
However, because host races have also been reported in a variety of other taxa and their study is relevant to many aspects of ecological and evolutionary biology, we begin with a brief survey of other topics related to host races.
Host races in non-insects Host race formation may be particularly prevalent within phytophagous insects, and has been most commonly studied in these taxa. Nonetheless, several recent papers have reported host races in other organisms. Female races (‘gentes’) of the common cuckoo Cuculus canorus, for example, are adapted to and prefer different avian hosts, but mate randomly with males (Marchetti et al. 1998) and do not seem to differ in mitochondrial or microsatellite DNA (Gibbs et al. 1996). The flatworm Schistosoma mansoni has sexual forms that parasitise both humans and the crepuscular or nocturnal rat Rattus rattus which may also be host races. S. mansoni isolated from different hosts can hybridise and backcross with each other in the laboratory, but genetically controlled differences in the time of day infective asexuals exit their aquatic host, the freshwater snail Biomphalaria glabrata, suggest that gene flow between the populations is restricted (Theron & Combes 1995, and references within). A number of host races have been described in parasitic plant and fungal populations, such as the mistletoes Arceuthobium tsugense (Nickrent & Stell 1990), Phoradendroncalifornicum (Glazner et al. 1988), P. tomentosum (Clay et al. 1985) and Viscum album (Zuber & Widmer 2000). The host race designation of these mistletoes is supported mainly by molecular or biochemical evidence of differentiation, and the rate of any gene flow remains to be investigated further. Similarly, patterns of microsatellite variation within the anther smut fungus Microbotryum violaceum reveal differentiation along host plant lines (Bucheli et al. 2000). Greenhouse cross-inoculation tests suggest restricted gene flow between M. violaceum strains on different hosts, but the extent of hybridisation between these populations in nature is uncertain.
Population control and conservation The existence and formation of host races can be an important consideration in both the control and the conservation of some taxa. Plans to slow the adaptation of pests to transgenic cultivars, for example, rely heavily on gene flow from populations in sympatric, transgene-free refuges slowing the spread of resistance alleles selected for on the transgenic host (Bourguet et al. 2000). Refuges normally consist of transgene-free plantings of the crop host, but wild host species could form part of the refuge of some generalist pests (Gould 1998). Prior quantification of gene flow between populations on cultivated and wild host species would then be necessary (Bourguet et al. 2000). For example, Bourguet et al. (2000) found that populations of the European corn borer Ostrinia nubilalis on cultivated maize Zea mays are genetically distinct from sympatric populations found on the wild host sagebrush (Artemisia sp), possibly as a result of non-random mating. Gene flow might also be reduced between populations on transgenic and non-transgenic plantations of the same crop. Gould (1994) suggests that even highly resistant individuals can be expected to suffer reduced growth rates when feeding on transgenic crops, and might experience allochronically mediated reproductive isolation from non-resistant populations on refuge crops.
Clarke and Walter (1995) argue that many biological control programmes believed to have employed different populations (‘strains’) of the same biological control agent in fact used partially or completely reproductively isolated taxa. Thus, the relative merits of various control techniques, e.g. using multiple species vs. single species of control agent, or introducing several vs. single populations of a particular agent, have been obscured. Meanwhile, Frey and Frey (1995) demonstrated that pheromone-mimic traps used to monitor the pest Quadraspidiotus perniciosus also collect males of the sibling species Q. zonatus. The study by Frey and Frey used traps containing artificial pheromones, which are known frequently to attract males of non-target species, but natural pheromone mediated cross attraction has been observed between cryptic host races of the larch budmoth Zeiraphera diniana (Emelianov et al. 2001) and is likely to occur between other host races that communicate via pheromones.
Conversely, the recognition of host races as distinct evolutionary units add to the debate concerning how populations are chosen for conservation. Many conservation policies have been concerned with species in the traditional ‘biological’ sense - groups that do not hybridise (O'Brien & Mayr 1991; Brownlow 1996; Crandall et al. 2000). Host races, however, contribute to diversity despite their appreciable gene flow, and may be incipient or actual species under some definitions. Moreover, Duffy’s warning that cryptic sibling species are individually ‘... likely to be more vulnerable to extinction through habitat destruction and disturbance...’ than single species (Duffy 1996) may also apply to host races.
I Theory of speciation via host shift
The enormous number of sympatric, closely related insect species specialised on different host plants, (e.g Farrell 1998) has led several biologists to argue that allopatric vicariance would be an almost impossible mechanism to explain existing diversity in this group (Bush 1975; Price 1980; Bush & Smith 1998). The existence of host races, particularly in taxa containing sympatric host associated species, is strong evidence for speciation via host shift, because these systems form an important intermediate stage between host use polymorphism and ‘good’ species.
Earlier scepticism towards the possibility of sympatric speciation (e.g. Mayr 1963, see also Tauber and Tauber 1989 for review) has been largely overcome. Recent refinements to models of this process, e.g. the assumption that diverging traits are polygenic rather than monogenic, suggest a more readily achievable set conditions for its operation than previous studies have predicted (e.g. Tregenza & Butlin 1999). Dieckman and Doebeli (1999) model a population dependent on a single resource of unimodal distribution such as seeds of differing sizes. Individuals can differ in three traits; (i) an ecological character that determines resource use, such as beak size, (ii) a marker character upon which mate choice is based such as colour, and (iii) a mating character that determines whether mate choice is assortative, disassortative, or random with respect to the marker character. Each trait is controlled by several additive diallelic loci. The carrying capacity of the resource is highest when all individuals have an intermediate phenotype, but those with extreme phenotypes suffer less from density- and frequency- dependent competition for the resource. Initially, the population evolves towards the intermediate phenotype. The resultant competition amongst conspecifics then increases the relative strength of disruptive selection for individuals with extreme phenotypes. However, because of recombination, divergence leading to a bimodal phonotypic distribution cannot occur without a reduction in random mating. In the model, gene flow between forms with dissimilar phenotypes is reduced by a positive feedback loop initiated by an association (linkage disequilibrium) between particular ecological and marker alleles. The linkage disequilibrium favours individuals with the tendency to mate assortatively, leading to increased disequilibrium in the next generation, which in turn increases selection for assortative mating, until reproductive isolation is achieved. The model works at a recombination frequency of 50%. Kondrashov & Kondrashov (1999) showed a similar result though in their simulations a bimodal phenotype could be produced by the selective regime before any assortative mating.
It has been wellknown for some time that the need for linkage disequilibrium between ecological adaptation and mate-choice loci is eliminated if divergence in an ecological trait directly causes assortative mating via pleiotropy (Maynard Smith 1966, Rice 1984b; Rice & Hostert 1993, Dieckman & Doebeli 1999). The initial reduction in gene flow caused by disruptive selection aids divergence in traits subject to less intense disruptive selection, which can pleiotropically reduce gene flow further, leading to a positive feedback loop that ends in speciation (Rice 1984b; Rice & Hostert 1993) After divergence in the first trait the process is similar to the accumulation of pre- and post-mating isolating mechanisms believed to take place during allopatric or partially allopatric speciation (Rice and Hostert 1993).
Because their life cycle and mating behaviour are often strongly affected by their host plant, phytophagous insects are particularly likely candidates for sympatric restrictions to gene flow via pleiotropy. The trait believed to be involved is host preference; the partially reproductively isolated candidate populations are host races (Bush 1975; Bush 1994). The most straightforward route by which a host shift can reduce gene flow is via a system of mating on the host as in Rhagoletis pomonella (Feder et al., 1994). However, gene flow might also be reduced in less direct ways. For example, host plant chemistry can affect cuticular hydrocarbons (Stennett & Etges 1997), and these chemicals often play an important role in mate choice (Coyne et al. 1994; Ferveur 1997; Tregenza & Wedell 1997; Singer 1998). Host plant phenology may also influence the developmental timing of insects (Wood & Guttman 1982; Langor 1989), and seasonal isolation can be a powerful inhibitor of gene flow (Wood & Guttman 1982). Even more generally, hosts are usually locally clumped in space as well as in time, so that host choice is liable to cause spatially-mediated assortative mating (Emelianov et al. 2001).
The physiological changes necessary for a population to be founded on a new host may often be minimal. Phytophagous insect larvae from several genera can complete their development in the laboratory using hosts on which they are not found on in the wild (Smiley 1978). Recently described examples include beetles Oreina elongata (Coleoptera: Chrysomelidae; Ballabeni & Rahier 2000) and Dendroctonus ponderosae (Coleoptera: Scolytidae; Cerezke 1995), and leafminers Liriomyza helianthi (Diptera : Agromyzidae; Gratton & Welter 1998). Of course, the ability to survive on a novel host in protected conditions does not always translate into the ability to survive in the field, because ecological factors other than nutrition, e.g. levels of parasitism and predation, may be important. Nonetheless, these results suggest that in some cases all that may be required for initiation of a successful host shift is a genetic change in the oviposition preference of females.
Direct selection for assortative mating (reinforcement)
Direct selection for positive assortative mating due to hybrid disadvantage (reinforcement) is another potentially important force in sympatric speciation, because it increases the likelihood that partially reproductively isolated, genetically distinct populations will speciate (Noor 1999).
Many models of reinforcement have been solely concerned with populations meeting at clinal hybrid zones, areas where “... an allele typical of one taxon monotonically replaces an allele typical of another taxon along linear transects ...”(Cain et al. 1999). Because selection for assortative mating occurs only in the band of contact, these models suggest that the evolution of assortative mating alleles would be hindered by gene flow from the rest of the population (Barton & Hewitt 1981; Sanderson 1989; Butlin 1990). This theoretical difficulty, and a paucity of clear experimental evidence of reinforcement, has resulted in skepticism towards its operation (see Noor, 1999 for critical review).
However, hybrid zones can also be mosaic, i.e. “... characterized by abrupt reversals of gene frequencies at diagnostic loci along linear transects ... caused by a patchy distribution of the differentiated taxa and their hybrids”(Cain et al. 1999). The results of a computer simulation conducted by Cain et al. (1999) support predictions (Harrison & Rand 1989, see also Guldemond & Dixon 1994; Howard 1986) that reinforcement is more likely in hybridising taxa that overlap more broadly than in a clinal hybrid zone. Host races clustered on different host plants are likely to form a patchy, mosaic hybrid zone throughout their area of sympatry, and reinforcement is therefore potentially important in these systems (Guldemond & Dixon 1994).
II. Defining host races.
Difficulties with current host race definitions
While new models of sympatric speciation have largely overcome objections to the possibility of speciation via host shift, differing opinions about what host races really are continue to cloud the literature. Populations described as ‘host races’ by some biologists might be regarded by others as sibling species or polymorphic populations. Here, we examine some of the most common definitions, and present an attempt to resolve the debate.
The earliest published definition of a host race of which we are aware of is from Bush (1969a)
“a population of a species living on and showing a preference for a host which is different from the host or hosts of other populations of the same species. Host races represent a continuum between forms which freely interbreed to those that rarely exchange genes. The latter may approach the status of a species, generally regarded as an interbreeding population reproductively isolated from all other such populations”
This definition, which emphasises the relationship between host races and other taxa (discussed in the next section) rather than their practical identification, has been followed by several others.
Mayr (1970) defines host races as:
‘‘non-interbreeding sympatric populations, which differ in biology but not, or scarcely, in morphology [and which are] prevented from interbreeding by preferences for different food plants or other hosts.’’
As in the previous definition, the above is not tailored for use in empirical studies. It does not, for example, suggest how biotypes that mate assortatively due to differences in plant preferences are detected, or what degree of morphological differentiation can be considered ‘scarce’. More importantly, as pointed out by Diehl and Bush (1984), under the widely used biological species concept the requirement that the biotypes be non-interbreeding (and lacking in morphological differentiation) in fact describes host associated sibling species.
Jaenike (1981) was the first to propose a definition of host races consisting of a set of experimentally verifiable criteria:
1. ‘‘[The populations] are sympatric, so that individuals in breeding condition in one population are within normal cruising range of those in another...”
2. ‘‘There must be a statistically significant genetic difference between the populations, suggesting, though not proving, that gene flow between them is not extensive’’
3. ‘‘The genetic difference (2) under consideration cannot be one that is directly related to host selection [unless] both males and females manifest genetic differentiation in host preference, and ... mating takes place on or near the host plant’’
4. ‘‘It must be shown that the genetic difference (2) is not solely the result of natural selection acting on the current generation of individuals’’
5. ‘‘Finally, if the above conditions are met it should be shown, if experimentally feasible, that the genetic difference between the two populations disappears over a period of generations when they are confined to breed on a single food type ... if the genetic differences between the two groups do not disappear ..., or if they do so initially only to reappear in subsequent generations ... then reproductive isolation between them in the field cannot be ascribed to differences in host preference. In this case the two groups represent distinct species, not host races.’’
Criteria (1), (2), and (4) of Jaenike’s definition have been incorporated into the new definition we propose below. However, we believe that the main distinction between host races and host-associated species is the occurrence of appreciable gene flow between the host races. While Jaenike’s definition does not exclude the possibility of gene flow between host races, neither does it explicitly require it, suggesting only that it is ‘not extensive’ (criterion (2)). Gene flow will affect the host race collapsibility of Jaenike’s criterion (5) (see Rhymer & Simberloff 1996), but is more easily verifiable than this criterion in empirical studies, whether using genetic markers or field studies of mating behaviour. Furthermore, in light of current theory Jaenike’s criterion (5) is unnecessarily stringent and difficult to test. Models of speciation via host shift predict that correlated effects of host preference may initially be the sole reason for assortative mating between host races, but also that assortativeness can later be strengthened by pleiotropic effects of a variety of behavioural and phenotypic differences, as well as by reinforcement.
The need for a consistent definition of the term was raised by Diehl & Bush (1984). The authors discussed several alternatives before proposing what is now perhaps the most widely quoted:
‘‘a population of a species that is partially reproductively isolated from other conspecific populations as a direct consequence of adaptation to a specific host’’
This definition concisely conveys the main features of a host race but is still not an ideal guide for empirical studies; it does not state the properties biotypes in the field must be shown to possess in order to fit it. For example, it is not obvious what would constitute adaptation to a specific host, given that in order for the initial host shift to occur, the only adaptation necessary is preference for a new host. Likewise, ‘partial reproductive isolation’, which potentially includes any trait under disruptive selection or affecting mate choice, is open to interpretation. Conversely, in other ways it seems too strict - theory suggests that reproductive barriers between host races are strengthened by indirect consequences of adaptation to different hosts.
Another set of criteria, incorporating some of Jaenike’s ideas, is proposed by Bush (1992):
1. ‘‘Individuals of different host associated populations in breeding condition must be sympatric’’
2. ‘‘Statistically significant genetic differences exist between these sympatric populations that are not directly related to host selection or solely the result of natural selection acting on a single generation’’
3. ‘‘Males and females exhibit genetic variation in host preference that results in assortative mating, i.e., mating occurs on the preferred host plant and host preference is under genetic control’’
4. ‘‘Males and females show host-associated tradeoffs in fitness’’
5. ‘‘There is no evidence of post-mating reproductive incompatibility. Hybrid incompatibility between host associated populations indicates they are sibling species, not host races’’
Again, there is no direct requirement for gene flow between the races in the wild. Bush’s criterion (5) seems unnecessary for the same reason as Jaenike’s criterion (5); current theory does not predict that host races (at least provided there is gene flow -- see below -- and especially those in later stages of divergence) need lack post-mating reproductive incompatibility completely. Another reason we ignore post-mating incompatibility is that there is no clear distinction between “post-mating” incompatibility and incompatibilities that are a direct pleiotropic result of host adaptation. Hybrid inviability might be due to faulty host choice, faulty detoxification of host-chemistry, or even asynchrony with host phenology. Similarly, it is unclear why the second part of criterion (3), “mating occurs on the host”, is stipulated, as we see no reason to exclude the possibility of divergent host use reducing gene flow in less direct ways.
Next, we suggest an alternative set of criteria by which host races may be identified.
Host races are sets of populations which
1.a. Use different hosts in the wild
1.b. Consist of individuals thatexhibit ‘host fidelity’, i.e. a preference for their natal host over alternative hosts
Host fidelity suggests that biotype differentiation (4.a) is not solely the result of host associated selection within a single generation
2. Coexist in sympatry in at least part of their range
Completely allopatric or parapatric populations would be classical geographic races. Sympatry in this sense does not exclude some within-locality clustering of hosts
3.a. Display a correlation between host choice and mate choice
Populations can only maintain genetic differences (4.a) across generations if mating is assortative.
3.b. Undergo actual gene flow with an appreciable frequency ( 1% per generation), and backcross
It is this feature alone that distinguishes host races from host associated species. Evidence for this gene flow may be obtained directly, via mark-recapture studies and observation of mating behaviour, or indirectly, via detection of linkage disequilibria between host-associated marker loci in populations on a single host (Barton et al., 1988).
4.a. Are genetically differentiated at more than one locus
Unlike members of a polymorphic population, host races consistently maintain correlated allelic differences at multiple loci. There are thus strong linkage disequilibria in mixed samples of host races. In contrast, for disequilibria to be used as evidence for reduced gene flow in (3.b.), they must be within host associated populations. The loci involved need not be marker loci; if it can be shown that multiple correlated genetic differences exist for morphology, mate choice (see also 3.a.), and host survival, such that few intermediates exist, a host race designation may be hypothesised.
4.b. Are spatially and temporally replicable, i.e. are more genetically differentiated from populations on another host in sympatry (and at the same time) than at least some geographically distant populations on the same host.
When populations have been shown to meet criteria (1) and (4.a.), evidence of at least a fleeting correlation of genetic differences and of differences in host use has been obtained. In populations also shown to meet criterion (3.a.), these correlations are very likely to be stable over time (i.e. in the face of gene flow). Indirect evidence for the stability of these correlations should, whenever possible, be obtained by showing that genetic differentiation between host races in sympatry is greater than in at least some allopatric populations on the same host (4.b.). In addition, because genetic drift could lead to short term differentiation between forms that is not being maintained across generations, it is important to have evidence that observed differentiation is stable over time.
Finally, members of different host races are likely to (5.a.) have a higher fitness on their natal than their alternate host and (5.b.) produce hybrids that are less fit than parental forms. For genetic differentiation to be maintained in the face of gene flow, there must almost always be some selection against moving to the other host and/or breeding with native populations on that host. However, host races in the earliest stages of divergence might be under only weak disruptive selection. Thus, these criteria (5a,b) are suggestive rather than diagnostic of host races, and will be harder to elucidate. We therefore do not believe that including them in our definition is helpful.
It may be convenient to use a much shorter form of this definition:
Host races are genetically differentiated, sympatric populations of parasites that use different hosts, and between which there is appreciable gene flow. Host races in the continuum from polymorphism to species
Host races are an intermediate stage between polymorphic populations and full species, and are difficult to define, in part because definitions of species are still contested. The most widely quoted idea of species in evolutionary biology is the biological species concept (Mayr 1963, 1970), which describes species as reproductively isolated populations between which there is zero or minimal gene flow. However, the continued usefulness of this concept has become somewhat uncertain, because many sibling taxa, normally considered species, are now known to undergo gene flow and hybridise at measurable rates (e.g. Grant & Grant 1992; Wang et al. 1997). A number of alternative species definitions, including the ecological, mate recognition, cohesion, and phylogenetic concepts have also been proposed (Van Valen 1976; Paterson 1985; Cracraft 1989; Templeton 1989) but this debate is not yet resolved (see Mallet 1995, 2001a). However, most species concepts can be viewed as ideas for the mechanisms by which separate clusters of genotypes originate or are maintained. Ecological concepts, for example, highlight the role of disruptive selection, while the biological concept emphasises the role of pre- and post-mating isolation. Phylogenetic concepts are more concerned with the history of origination. Here, we employ a ‘genotypic cluster’ criterion of species, which specifically allows for gene flow, non-monophyly and genetic differences at loci that are not necessarily fixed. Under this criterion, species are genotypic clusters separated in sympatry by correlated genetic differences at multiple loci (Mallet 1995; Feder 1998). The correlations between loci should be sufficient to cause a bimodal genotypic distribution such that two groups or ‘clusters’ of genotypes are identifiable, which are separated by intermediates rarer than the genotypes of the majority (Jiggins & Mallet 2000). However, our argument here is independent of the particular species concept employed: almost all taxonomically recognised species occurring in sympatry will also be separate genotypic clusters.
Correlations between alleles at different polymorphic loci (linkage disequilibria) can only be maintained between populations when disruptive selection is strong relative to inter-population gene flow and recombination of population-specific alleles. However, provided selection is strong enough, the movement of genes from one population to another via hybridisation and backcrossing need not be zero, and so, unlike the biological concept, the genotypic cluster definition allows for incomplete reproductive isolation between species.
Like species, host races defined as above are clusters of genotypes separated by gaps. The only difference between members of the two groups lies in the extent of ‘actual gene flow’ -- the exchange of migrants and hybridisation they undergo (Mallet 2001b). Between host races actual gene flow is appreciable, but between species it occurs rarely or not at all. While any dividing level of gene flow is somewhat arbitrary because species and host races are part of a continuum, we believe a reasonable figure for practical purposes is about 1% per generation, approximately an order of magnitude higher than the rate of hybridisation between typical sympatric taxa recognized as ‘good’ species, but which are known to hybridise (Grant & Grant 1992; Mallet et al. 1998).
Host races and genotypic cluster species differ from polymorphic populations in the pattern, and often the number, of their differences. Host races differ at multiple loci, while morphs within a polymorphic populations may differ at only a single locus. If more loci are involved, their polymorphisms are more or less uncorrelated, so that individuals are placed in different groups depending upon the particular phenotype or locus examined. In contrast, when members of two host races or species are placed in groups according to multiple criteria, a bimodal distribution occurs (Mallet 1995, 2001a; Feder 1998; see also Jiggins and Mallet 2000).
Therefore, although host races are clearly distinct from polymorphic populations, they are simply a subset of genotypic cluster species. Distinguishing host races is useful because most systematists would not wish to name taxa that may have few fixed differences, and which exchange genes at a rate greater than 1% per generation. Host races also form an intermediate step in the continuum from polymorphism to species as generally recognized.
III. Case studies: steps in the continuum
Host race status has been suggested for many insect biotypes. Here, we attempt to identify those which are host races according to our criteria, categorising them as (i) single polymorphic populations, (ii) probable host races (iii) host races or (iv) sibling species. These classifications are a best estimate on the basis of current information, and may need to be reviewed as further relevant work is carried out. Our results are summarized in Table 1. The list of cases discussed is not exhaustive, and is biased towards those systems which have been tested for several of the criteria we include in our list. Host race status has also been proposed for a number of taxa consequently not mentioned here, many of which lie within the Aphididae (Thieme 1987; Tauber & Tauber 1989).
Cases of single polymorphic populations
(1). The mountain pine beetle Dendroctonus ponderosae (Coleoptera: Scolytidae) on lodgepole pine (Pinuscontorta) and limber pine (P.ponderosae)
Male-female mountain pine beetle pairs originating from different hosts are less likely to lay fertile eggs on lodgepole pine than pairs collected from the same species of pine, and, on limber pine only, progeny from mixed pairs have a lower average dry weight and fat content than those of insects collected from the same host species (Langor et al. 1990). However, the offspring of all cross types are fertile, differences in egg laying are not observed on limber pine, and neither development time nor mortality differs between any brood type on either host. Allozyme differentiation was initially reported (Sturgeon & Mitton 1986) but later shown to be caused by selection acting within single generations (Langor & Spence 1991). Field collected adults of the two populations do differ morphologically, but possible effects of within generation selection or phenotypic plasticity on morphology have not been examined (Langor & Spence 1991). Insects on lodgepole pine begin emerging approximately one week earlier than those on limber pine, but peak and late emergence on both hosts overlaps for about two months of the year (Langor 1989).
(2). Red or black headed biotypes of the fall webworm Hyphantria cunea (Lepidoptera: Tortricidae) on various hosts
The fall webwormconsists of polyphageous ‘red-headed’ and ‘black-headed’ larval forms that have overlapping host use patterns and may be sibling species (Jaenike & Selander 1980; McIntee & Nordin 1983; McLellan et al. 1991). Differentiation along host plant lines within (or between) colour pattern forms has not been intensively studied. However, no significant differentiation in allozyme frequency was found between populations of the red headed form on black walnut and black cherry (Jaenike & Selander 1980).
(3).The small ermine moth Yponomeuta padellus (Lepidoptera: Yponomeutidae) on hawthorn (Crataegus monogyna) and blackthorn (Prunus spinosa) Sympatric, host associated larval populations of the small ermine moth collected from a single site differed in allozyme frequency in 1978, 1979, and 1990 (Menken 1981; Menken 1982; Raijmann & Menken 2000), but supporting evidence that this differentiation is maintained in the face of gene flow, and is not simply due to within-generation selection is lacking. There is no evidence of pheromone mediated mate choice (Brookes & Butlin 1994b), and larvae from different hosts do not differ in their preference for or fitness on various hosts in the laboratory (Kooi et al. 1991). Other factors directly affecting the extent of actual gene flow, such as adult host plant choice and hybridisation rates, have not to our knowledge been investigated.
Cases of probable host races
(4). Ladybird beetles Epilachna niponica (Coleoptera, Coccinellidae) on thistle (Cirsium spp.) and E. yasutomii on blue cohash (Caulophyllum robustum)
Ladybird beetles Epilachna niponica on thistle and E. yasutomii on blue cohash differ in size, shape (Katakura 1981), and average time to adult development (E. niponica take approximately 35 days but E. yastomii need only about 30; Katakura and Hosogai 1994). Mating tends to occur on the host (Katakura et al. 1989), larvae of both species have a greatly reduced chance of survival to maturity on the alternate host (Katakura & Hosogai 1994), and, in the laboratory, adults prefer to feed on the host type they were collected from in the wild (Katakura & Hosogai 1994). Despite these differences the two forms hybridise in the laboratory without sex ratio distortion, and hybrids survive as well on either host as the native parental type (Katakura & Hosogai 1994). In the absence of host plants, mating between the two forms is random (Katakura & Hosogai 1994). Hybrids are intermediate in size, develop more slowly than E. yasutomii, and sometimes develop more rapidly than E. niponica(Katakura & Hosogai 1994). Because E. niponica and E. yasutommi have long and overlapping mating seasons, females of both types mate multiply, and there is no evidence of conspecific sperm precedence, allochronic isolation might not be extensive (Katakura & Hosogai 1994). However, the frequency of hybridisation and backcrossing in the wild has not to our knowledge been investigated.
(5). The spiraea aphid Aphis citricola (Homoptera: Aphididae) on satsuma (Citrus unshiu) and thunberg spiraea (Spiraea thunbergii)
The average emergence time of spiraea aphid populations on satsuma and thunberg spiraea differs by approximately one month, and is under genetic control (Komazaki 1986; Komazaki 1990). The two forms occur sympatrically, and laboratory-bred hybrids survive well on one of the hosts, thunberg spiraea, under field-cage conditions (Komazaki 1986). The potential for hybridisation between these forms in the field has not been investigated in detail, although Komazaki (1986) stresses that there is considerable (but incomplete) allochronic isolation of adult forms (Komazaki 1986).
(6). The aphid Cryptomyzus galeopsidis (Homoptera: Aphididae) on redcurrant (Ribes rubrum) and blackcurrant (Ribes nigrum) primary hosts
Populations of the aphid Cryptomyzus galeopsidis on redcurrant and blackcurrant primary hosts (the hosts where sexual forms reproduce) differ genetically, but will hybridise and backcross when housed together (Guldemond 1990b; Guldemond et al. 1994; Guldemond & Dixon 1994). However, the fitness of the single hybrid clone that was tested appeared reduced compared to its parents, as it produced fewer mature sexual females (Guldemond 1990b). There is some pre-mating isolation between blackcurrant biotype males and redcurrant biotype females, but hybridisation in the opposite direction occurs freely even when males can choose between females of both biotypes (Guldemond et al. 1994). Males do not appear to differentiate between pheromones of redcurrant and blackcurrant associated females (Guldemond & Dixon 1994). Populations on both primary hosts share a secondary host (the host where asexual forms produced later in the season feed), hemp nettle (Galeopsis tetrahit). Migratory forms of both biotypes tend to prefer their native host, although, in the case of the redcurrant biotype, this preference was not expressed by all clones that were tested (Guldemond 1990a).
(7). Rhagoletis pomonella (Diptera: Tephritidae) on hawthorn (Crategus mollis) and the ‘flowering dogwood fly’ on flowering dogwood (Cornus florida)
Although the two forms have recently been recommended for species status under a ‘nonstrict’ version of the biological species concept (Berlocher 1999), we use our criteria to tentatively place them in the host race category. The two biotypes are often sympatric, and exhibit only frequency differences at up to seven of 17 polymorphic allozyme loci. However, only some components of gene flow have been directly measured, and the results are inconclusive; the populations experience partial allochronic isolation, slight post mating isolation, and have differences in host preference (Berlocher 1999).
(8). The pea aphid Acyrthosiphon pisum (Homoptera: Aphididae) on alfalfa (Medicago sativa) and red clover (Trifolium pratense)
Populations of the pea aphid Acyrthosiphon pisum on alfalfa and red clover differ in allozyme allele frequency (Via 1999). The two forms will hybridise in the laboratory, and a comparison of semi-diagnostic allele frequency within new host plant fields to that in fields containing established populations suggested that approximately eleven percent of new migrants to clover fields are from alfalfa, and nine percent of migrants to alfalfa come from clover fields (Via 1999). However, the survival of aphids migrating to the alternative host is much lower than on the natal host (Via 1991a,b; Via 1999; Via et al. 2000), and the insects exhibit a strong preference for their natal host type (Caillaud and Via 2000). Furthermore, the frequency of alleles associated with the alternate host declined in new host plant fields throughout the season (Via et al. 2000). Several generations of successful reproduction by the parthenogenic migratory forms must take place before the sexual forms are produced (Caillaud and Via 2000). Genes affecting host choice and survival on each host map to the same chromosomal regions, indicating probable genetic trade-offs in host adaptation (Hawthorne & Via 2001), however the frequency of mating between sexual alfalfa and clover aphids in the wild remains unknown.
(9). The sawfly Platycampus luridiventris (Hymenoptera: Tenthredinidae) on two species of alder (Alnus glutinosa, A. incana)
Sympatric populations of this sawfly associated with the alder species Alnus glutinosa and A. incana differ in larval morphology, female oviposition preference (Heitland & Pschorn-Walcher 1992), and allozyme allele frequency (Herbst & Heitland 1994). Larvae of both types develop faster on their natal host (Heitland & Pschorn-Walcher 1992). The potential for hybridisation between A. glutinosa and A. incana associated populations has not to our knowledge been studied in either the field or laboratory.
(10). The goldenrod gall ballmaker Eurosta solidaginis (Diptera: Tephritidae) on two goldenrod species (Solidago altissima, S. gigantea)
Populations of Eurosta solidaginis using the different goldenrod species Solidago altissima and S. gigantea are significantly genetically differentiated (Waring et al. 1990), and have a higher survival on their natal than their alternative host (Craig et al. 1993; Craig et al. 1997). Mating is strongly assortative when host plants are present, but much less so when they are absent (Craig et al. 1993; Itami et al. 1998). Differences in emergence times also contribute to reproductive isolation (Craig et al. 1993). A direct estimate of the extent of gene flow between biotypes of Eurosta has been difficult to obtain, although in some conditions, such as high spring temperature, which reduces allochronic isolation, it appears particularly likely (Itami et al. 1998). The observation that approximately 3% of insects collected from the wild have the ‘intermediate’ oviposition preference also seen in laboratory hybrids suggests gene flow, as does the high fitness of some hybrids and backcrosses on particular host plant genotypes (Itami et al. 1998). Average hybrid fitness is however lower than that of pure forms (Craig et al. 1997), so backcrossing may be rare.
Cases of host races (11). The apple maggot fly Rhagoletis pomonella (Diptera: Tephritidae) on apple (Malus pumila) and hawthorn (Crategus mollis)
Apple and hawthorn infesting forms of Rhagoletis pomonella are perhaps the best-characterised pair of host races. The two forms differ in time of emergence (Smith 1988) and host choice (Feder et al. 1994). Differences in allele frequency at six allozyme loci have been maintained within a single sympatric site throughout eleven years of study (Feder et al. 1998; see also Feder et al. 1990; Feder et al. 1993) and there is substantial direct and indirect evidence for gene flow between the biotypes.
Field studies of several components of actual gene flow, including host preference and temporal co-occurrence of mature adults, have shown that the rate of exchange of migrants between the two populations is approximately 6% per generation (Feder et al. 1994). Linkage disequilibria between host associated loci within each form also suggests that gene flow is occurring (Barton et al. 1988; Feder & Bush 1991), although other factors may be involved; in some locations, disequilibria existed at loci that had similar frequencies within both biotypes (Feder & Bush 1991). Apple and hawthorn flies are likely to mate randomly when they encounter each other on the same host plant (Feder et al. 1994), and there is no evidence of an intrinsic reduction in hybrid fitness (Reissig & Smith 1978).
(12). The larch budmoth Zeiraphera diniana (Lepidoptera: Tortricidae) on larch (Larix decidua) and pine (Pinus cembra)
Populations of the larch budmoth Zeiraphera diniana on European larch (Larixdecidua) and cembran pine (Pinuscembra) co-exist in mixed stand forests of the French and Swiss Alps. They display heritable differentiation in a number of phenotypic traits, including female pheromone blend and male pheromone response (Baltensweiler 1977; Baltensweiler & Priesner 1988; Emelianov et al. 1995, 2001). Larch and pine associated moths also differ in allozyme frequency at three loci, and this differentiation has been stable since at least 1994 (Emelianov et al. 1995, 2001). There is no inter-population differentiation in allele frequency at 10 of the loci examined, but very strong differentiation at the other three. This highly heterogeneous pattern is consistent with gene-flow-with-selection models of divergence, which predict that some regions of the genome, where selection is strong relative to gene flow, will become strongly differentiated while others, at which the reverse is true, will not. Thus, the distribution of allozyme frequency differences provides indirect evidence of gene flow (Emelianov et al. 1995). Gene flow has been directly estimated at approximately 2.4% per generation (Emelianov et al. 2001) from the combined results of mate choice (Drès 2000) and field experiments on host choice and pheromone-mediated cross-attraction (Emelianov et al. 2001).
Cases of sibling species
(13). The apple maggot fly Rhagoletis pomonella (Diptera: Tephritidae) on hawthorn (Crategus sp.) and the blueberry maggot fly R. mendax on blueberry (Vaccinium corymbosum)
Populations of R. pomonella and R. mendax contain unique allozyme alleles at eleven loci (Feder & Bush 1989). A survey of 426 individuals of both types failed to reveal any putative hybrid genotypes, despite the co-occurrence of sexually mature adults on intertwined hawthorn and blueberry bushes (Feder & Bush 1989). Thus, the rate of actual gene flow between the forms appears to be considerably less that 1% per generation.
(14). The brown planthopper Nilaparvata lugens (Homoptera: Delphacidae) on weed grass (Leersia hexandra) and cultivated rice (Oryza sativa)
Brown planthoppers found on the weed grass Leerisa hexandra and on cultivated rice Oryza sativa display heritable differentiation in a number of traits, including mating call pulse repetition frequency. Although viable and fertile F1 hybrids can be produced (Heinrichs & Medrano 1984; Claridge et al. 1985), laboratory tests of mate choice suggest that the two populations do not hybridise in the wild; only a single putative hybrid was found in a crowded population cage containing males and females of both forms. Furthermore, when insects were played mating calls of members of their own population and of those found on the alternate host, both males and females responded only rarely, and with reduced vigour, to calls from members of the other population (Claridge et al. 1985).
(15). Treehoppers of the Enchenopa binotata (Homoptera: Membracidae) complex on various hosts (see Table 1)
Members of this treehopper complex have varying levels of allele frequency differentiation, and mate assortatively in the lab (Guttman et al. 1981; Wood & Guttman 1982). Low levels of hybridisation occurred in laboratory conditions in the absence of host plants, but there is no evidence that hybridisation also occurs in the wild, and considerable evidence for complete allochronic isolation in field conditions, due to the timing of egg hatch with host bud burst (Wood & Guttman 1982). Although the possibility that some of the forms may be host races has not been completely discounted, the biotypes so far discovered are most likely to hybridise at a rate of less than one percent per generation.
(16). The Muellerianella complex (Homoptera: Delphacidae). Muellerianella brevipennis, M. fairmairei, and M. extrusa on several grass species (see Table 1)
Members of this complex are, somewhat tentatively, placed in the species category. In the laboratory all three species mate assortatively with conspecifics (Booij 1982c). Hybrid broods from all interspecific crosses are much smaller than non- hybrid broods, female biased, and the males are predominantly infertile (Booij 1982c), although backcross broods were bred from some hybrid females (Booij 1982c). The calls by which M. brevipennis, M. extrusa, and M. fairmairei males communicate with potential mates differ (Booij 1982a) and although the extent to which this affects their long range cross attraction has not been directly investigated, acoustic behaviour strongly influences mate choice in closely related taxa (Booij 1982a; and references therein). Despite this, there is evidence that in some areas the forms may come into contact with each other, and hybridise. Hosts of different species are sometimes found in close proximity (Booij 1982b), and two putative M. fairmairei x M. brevipennis hybrid females have been collected from one such site (Booij 1982c). The balance of current evidence does however point towards a level of hybridisation of less than 1% per generation, and the very low chance of backcrossing also supports their species status.
(17). The fall armyworm Spodoptera frugiperda (Lepidoptera: Noctuidae) on corn (Zea mays) and rice (Oryza sativa)
Rice (Oryza sativa) and corn (Zea mays) associated Spodoptera frugiperda can produce viable hybrids in the laboratory (Pashley & Martin 1987; Whitford et al. 1988), but there is evidence of hybrid sterility in at least one cross direction (Pashley & Martin 1987). Both long range cross attraction and hybridisation at close range are highly (but incompletely) assortative (Pashley et al. 1992), the former at least in part because females call at different times of the night. The two taxa display frequency differentiation at several allozyme loci (Pashley 1989a), and diagnostic differentiation in an RFLP tandem repeat marker (Lu et al. 1992). Rates of hybridisation in nature have not been estimated, but they are assumed low.
IV. Empirical evidence for speciation via host shift
How common are host races?
There is insufficient evidence to resolve the host race status of 7 of the 17 pairs of taxa discussed above (case studies 4 – 10). Because of a continued shortage of detailed information about host races (Tauber & Tauber 1989), it is not possible to draw very definite conclusions about the frequency of these taxa in nature. The lack of data is most pronounced in relation to our criteria (3.a) and (3.b), those dealing with gene flow between the forms, which we argue is critical for distinguishing them from species. Members of most likely systems show differentiation along host plant lines, will interbreed in the laboratory, and have levels of genetic differentiation compatible with continuing gene flow, and while several components of gene flow have often been studied, direct or indirect estimates of actual gene flow have rarely been obtained
Nonetheless, host races have been confirmed in two of the 17 studies, approximately 13% of the sympatric, phytophagous insect populations in which the possibility has been investigated reasonably thoroughly. This conservative estimate seems likely to rise, because current data suggests the presence of 7 more (case studies 4-10)
The studies carried out to date have concentrated on insects in which the presence of host races seems particularly likely; in most cases there was prior evidence of host-associated differentiation between populations of a presumed sympatric species. New studies, benefiting from sensitive protein and DNA markers, suggest that cryptic, host-associated differentiation may be common in phytophagous insects. Taxonomic revisions of several taxa, by taking account of new behavioural, DNA or protein based characters, have detected sympatric, monophagous biotypes within presumed polyphageous species. Recent examples include bark beetles Dendroctonus brevicornis (Kelley et al. 1999) and fruit flies of the genus Blepharoneura (Condon & Steck 1997). Intensive studies of agricultural pests have also begun to uncover population substructuring along host plant lines (Shufran et al. 2000), the starting point of many studies described here. Thus, although the number of phytophage host races discovered so far is small, the number of insect systems that may conceal them is potentially very large, and, if the pattern observed in the data presented here is representative of other systems, host races may indeed be a common phenomenon.
Is the formation of host races likely to lead to sympatric speciation?
Did host races in sympatry today undergo their divergence during a period of isolation in allopatry? In the case of Rhagoletis pomonella on apple and hawthorn, at least, they did not. Historical records show that the apple host was introduced within the range of hawthorn, and it is extremely unlikely that the two were ever allopatric (Bush 1969; Bush et al. 1989; Bush 1994), though they may of course have been partially separated by patchy host-plant distribution. Another frequently quoted line of evidence that host races can form, and diverge to the point of speciation in sympatry, is the geographical distribution of phytophagous insect species (e.g. Tauber & Tauber 1989). An enormous number of host specialised insect species have evolved – many millions of beetle species alone (Farrell 1998). Many of these are sympatric, identifiable primarily by their host, and capable of producing viable hybrids in the laboratory. A sympatric origin for most of these species seems reasonable, because the alternative, that all were separated and diverged in allopatry, obtaining their present sympatric distribution as a result of secondary contact is hard to imagine for so many species (Tauber & Tauber 1989). Further support for speciation via host shift comes from comparisons of several host-associated biotypes in Rhagoletis. The fact that Rhagoletis pomonella and R. mendax, the apple and blueberry maggot flies, are good sympatric species, while host races of R. pomonella exist on apple and haw with the ‘flowering dogwood fly’/R. pomonella pair somewhat intermediate, provides a continuity argument that the formation of sympatric host races in this genus is likely to lead to speciation (Payne & Berlocher 1995, 1999; Feder et al. 1998; but see Barraclough & Vogler 2000 for a different view). Comparison between the Rhagoletissuavis and pomonella species groups shows that members of the first have an allopatric or parapatric distribution, and are not specialized on different hosts, whereas members of the latter, which are all sympatric, are also all restricted to different hosts (Bush & Smith 1998). Again, this argues for an important role of host shift in the sympatric formation of Rhagoletis species.
The main difficulty for the theory of sympatric speciation has always been to explain how selection can cause multilocus differentiation that is correlated with habitat use and mate choice, in the presence of gene flow. By providing a continuum of examples in which host-associated differentiation is maintained in spite of actual or probable gene flow, the studies discussed here show empirically that populations exchanging genes in sympatry can be stable. The existence of host races therefore show that a route to sympatric speciation exists.
Convincing evidence that host races exist, and are intermediates on the route to sympatric speciation has been provided by intensive studies of tephritid flies of the genus Rhagoletis. However, although the empirical studies reviewed here suggest the possibility that host races are common, there is not yet enough data to draw conclusions about the frequency of sympatric speciation via host shift. Current examples suggest a stable route for sympatric speciation via host shift, but the possibility of an unexpected hiatus in this route cannot be ruled out. In the majority of studies of potential host races discussed here, firm conclusions about the level of gene flow have not been reached. This measurement is however vital in unambiguously testing the host races status of likely systems according to the criteria suggested in this review, and is the best way of investigating the possibility of ‘gene flow gaps’ in the current array of sympatric host-associated forms. In addition, a number of insect biotypes which have been presented as host races seem most likely to represent simple polymorphisms or what would be generally accepted as sibling species. This confusion is likely to have arisen due to ambigous host race criteria.
We would like to thank Mike Claridge and Andrew Hingle for helpful discussions about this manuscript.
Ballabeni, P., and M. Rahier, 2000. Performance of leaf beetle larvae on sympatric host and non-host plants. Entomologia experimentalis et applicata