Testing host-associated differentiation in a quasi-endophage and a parthenogen on native trees


Raul F. Medina, Department of Entomology, TAMU 2475, College Station, TX 77843, USA.
Tel.: +1 979 845 8304; fax: +1 979 845 6305; e-mail: rfmedina@tamu.edu


Host-associated differentiation (HAD) is the formation of genetically divergent host-associated sub-populations. Evidence of HAD has been reported for multiple insect herbivores to date, but published studies testing more than one herbivore for any given host-plant species pair is limited to herbivores on goldenrods. This limits the number of pair-wise comparisons that can be made about insect life-history traits that might facilitate or inhibit host-race development in general. Two traits previously proposed to facilitate HAD include endophagy and parthenogenesis. We tested for HAD in two herbivores, a quasi-endophagous caterpillar and a parthenogenetic aphid, feeding on two closely related species of hickories. We found that the quasi-endophage is panmictic, whereas the parthenogen exhibits HAD on their sympatric host plants, pecan and water hickory, at a geographic mesoscale. This is an important first step in the characterization of HAD in multiple insect herbivores using North American hickories, a host-plant system with many shared parthenogens.


Flowering plants, herbivorous insects and insect parasitoids together comprise more than 50% of the world’s species (Price, 1980; Godfray, 1994; Schoonhoven et al., 1998). Because insects often have narrow host breadths (Mitter et al., 1988; Godfray, 1994) and tight associations with their hosts, they are prime candidates for host-associated differentiation (HAD). HAD is the formation of genetically divergent host-associated sub-populations (Bush, 1969; Abrahamson et al., 2003). HAD has been proposed as a mechanism promoting adaptive radiation of host-associated lineages resulting over time in increased species diversity (Mitter et al., 1988; Funk et al., 2002; Stireman et al., 2006). HAD can be detected when genotypes cluster by host-plant species despite sampling from geographically separated populations (Berlocher & Feder, 2002; Stireman et al., 2006; Scheffer & Hawthorne, 2007).

Host-associated differentiation is a special case of ecological speciation (Schlutter, 2001; Rundle & Nosil, 2005) wherein the disruptively selective environments experienced by diverging populations of phytophagous insects are different host-plant species (Dres & Mallet, 2002). The ecological literature contains a growing body of HAD case studies (Dres & Mallet, 2002; Stireman et al., 2005; Sword et al., 2005; Vialatte et al., 2005; Lozier et al., 2007; Magalhaes et al., 2007; Dorchin et al., 2009; Peccoud et al., 2009) but these, with a single exception (Stireman et al., 2005), involve no more than one herbivore tested per host-plant pair. Many reports provide a plausible set of herbivore traits hypothesized to facilitate HAD but testing the relative importance of those traits will depend on studies which (i) test HAD for multiple herbivores on the same host-plant species pair and (ii) publish negative as well as positive results. Whereas there are several traits that could promote HAD, two herbivore traits proposed to facilitate HAD are addressed in this study; endophagy (Dreger-Jauffret & Shorthouse, 1992; Stireman et al., 2005) and parthenogenesis (Sunnucks et al., 1997; Dixon, 1998; Vialatte et al., 2005; Loxdale, 2008).

Endophagous insects are thought to be more prone to exhibit HAD because they are less exposed to selective regimes imposed by generalist predators and environmental conditions and because they are more likely to be selected by host-plant traits than exposed feeders (Cornell et al., 1998; Stilling & Rossi, 1998; Stireman et al., 2005). The only case of HAD tested in multiple herbivores on the same host-plant pair involves insects on sibling species of native perennial goldenrods (Stireman et al., 2005). In this system, HAD was found in two-thirds of the endophages tested but was not found in either of two tested exophages. HAD was found in six of eleven species tested and in insects belonging to different insect orders and different trophic levels (Waring et al., 1990; Abrahamson & Weis, 1997; Eubanks et al., 2003; Stireman et al., 2005, 2006).

Cyclic parthenogenic insects, such as aphids, may also develop host-associated lineages faster than sexually reproducing insects because favourable mutations can become fixed more quickly when sex is limited (Hartl, 1972; Lynch, 1984; Neiman & Linksvayer, 2006). Most aphids are holocyclic apomictic parthenogens (Blackman & Eastop, 1994; Loxdale, 2008) restricting sex to a single season. Thus, prior to sex, rapid succession of asexual generations can amplify and accelerate the response to selection (Lynch & Gabriel, 1983; King, 1993; King & Murtaugh, 1997; Vialatte et al., 2005; Loxdale, 2008). If parthenogens occupy different host-plant species, and sex is initiated in response to host-plant-mediated cues, then a difference in the timing of sexual reproduction may be a consequence of ecological differences between host-plant species (Guldemond & Mackenzie, 1994; Serra et al., 1998). To date, HAD has been reported in several aphid species (Akimoto, 1990; Guldemond et al., 1994; Vanlerberghe-Masutti & Chavigny, 1998; Via & Hawthorne, 2002; Simon et al., 2003; Brunner et al., 2004; Vialatte et al., 2005; Lozier et al., 2007; Peccoud et al., 2009).

Testing for HAD in an insect community feeding on the same two host-plant species is an optimal way to examine the role of endophagy and parthenogenesis in promoting HAD because multiple herbivores can be selected for study within a system based on the presence/absence of these traits without the confounding factors involved when comparing insects from different host-plant study systems (Stireman et al., 2005).

We have selected a host-plant study system that we think maximizes the possibility of finding HAD. First, the system is native (i.e., native tree and insect species). Whereas rapid evolution of host races has been documented in native insects feeding on introduced plants (Bush, 1969; Feder et al., 1999; Strauss et al., 2006), the increased evolutionary time of insect community and host-plant interaction afforded by native systems should increase the probability of finding HAD. Second, the host-plant species chosen are trees. HAD, to our knowledge, has been documented in arthropods feeding on annual plants in only a rare number of cases; Spodoptera frugiperda, Sitobion avenae, Nilaparvata lugens and Ostrinia nubilalis feeding on cultivated grasses (Dres & Mallet, 2002; Martel et al., 2003; Vialatte et al., 2005). In contrast, trees offer relatively more stable, long-lived genotypes to which many generations of arthropods can adapt (Edmunds & Alstad, 1981; Mopper, 2005; Magalhaes et al., 2007), and most cases of HAD are from such perennial systems (Dres & Mallet, 2002; Magalhaes et al., 2007). The native trees pecan Carya illinoinensis Koch and water hickory C. aquatica Michx (Fagales: Juglandaceae) have been selected as host plants for this study. These deciduous trees share a large and diverse native insect fauna (Table 1).

Table 1.   A list of insects shared by pecan and water hickory. This list is not comprehensive.
Insect speciesEndophagousParthenogenetic
  1. The X’s denote those species exhibiting parthenogenesis and endophagy (B. Ree, Texas A&M University, College Station, Pers. Comm.; A. Dickey Pers. Obs.).

 Actius luna  
 Amorpha juglandis  
 Cameraria caryaefoliellaX 
 Catocala agrippina  
 Catocala maestosa  
 Citheronia regalis  
 Cydia caryanaX 
 Datana integerrima  
 Gretchena bollianaX 
 Hyphantria cunea  
 Malacosoma disstria  
 Megalopyge opercularis  
 Satyrium calanus  
 Stigmella juglandifoliellaX 
 Aphelinus perpallidus  
 Periclista marginicollis  
 Pteromalus sp.X 
 Clavaspis crypta  
 Empoasca fabae  
 Goes pulcherX 
 Gypona octolineata  
 Metcalfa pruinosa  
 Phylloxera devastatrixXX
 Phylloxera notabilisXX
 Phylloxera texanaXX
 Velataspis mimosarum X
 Melanocallis caryaefoliae X
 Monellia caryella X
 Monelliopsis pecanis X
 Curculio caryaeX 

This study is just the first step in testing HAD in a community of phytophagous insects. Insects from Table 1 were selected for this study based on the presence of two traits of interest, endophagy and parthenogenesis in addition to their relative abundance at multiple study sites, the relative ease of collecting them in the field and the relative ease of rearing them to adulthood in the laboratory.

Host-associated differentiation was tested using cluster analyses of amplified fragment length polymorphisms (AFLPs) (Vos et al., 1995). AFLPs are anonymous dominant molecular markers, which are advantageous for HAD studies because they are often neutral, can be generated quickly, are cost effective, and can be used to determine both the structure of populations and the assignment of individuals to populations (Sword et al., 2005; Falush et al., 2007; Meudt & Clarke, 2007).

Materials and methods

Study system

The genus Carya contains 13 North American species and five Asian species (Manning, 1978) of large deciduous trees. The genus has been present in North America for at least 34 million years based on evidence from fossilized fruits with extant species dating to the Pleistocene (Manchester, 1987). Pecan and water hickory are common to the river and creek bottoms in the hardwood forests of eastern North America (Fralish & Franklin, 2002) (Fig. 1). Pecan is the most economically important indigenous nut crop in the US (Grauke et al., 2003). Water hickory is a species closely related to pecan but unlike pecan has a flat, wrinkled and bitter nut (Stone et al., 1965). A detailed phylogenetic hypothesis does not exist for the genus Carya but within the genus, C. aquatica and C. illinoinensis have been grouped together in the section Apocarya along with three other species of North American hickories and five other species of Asian hickories (Thompson & Grauke, 1991). Grauke et al. (1987) documented phenological differences among sympatric Carya species in Louisiana and found no temporal overlap in pollen shed and pistil receptivity between pecan and water hickory with water hickory budding and flowering approximately 3 weeks later in the spring than pecan. Phenological differences between host plants have been shown to be important sources of ecological isolation of insect populations in sympatry (Komatsu & Akimoto, 1995; Feder & Filchak, 1999; Mopper, 2005), which could drive HAD. Despite phenological differences, pecan and water hickory hybridize in the wild (Stone et al., 1965; Grauke et al., 1987) but hybrids can be recognized by nut and bud phenotypes intermediate to parent species (Grauke et al., 1987; Thompson & Grauke, 1991). The genus Carya is very promiscuous, and other hybrids including either pecan or water hickory have been documented (Thompson & Grauke, 1991).

Figure 1.

 Distribution maps for (a) Genus Carya in North America and (b) Pecan and water hickory.

The yellow pecan aphid Monelliopsis pecanis Bissel (Hemiptera: Aphididae) is a holocyclic parthenogenetic exophage feeding on the lower surfaces of leaflet tertiary veins (Tedders, 1978). The pecan bud moth Gretchena boliana Granovsky (Lepidoptera: Tortricidae) is a sexually reproducing foliage feeder but early instars are inquilines, feeding inside of Phylloxera sp. galls (Mitchell et al., 1984) leading us to designate it as a quasi-endophage. Later instars fold leaves (Mizell & Schiffhauer, 1986). We have found and reared both herbivores commonly on both tree species and confirmed both our tree and insect species identity with systematists (G. Miller, USDA-SEL, L.J. Grauke, USDA-ARS, J. Brown, USDA-SEL). To the best of our knowledge, neither herbivore is recorded from another species of Carya.

Insect sampling

The study area is a four county area in central Texas (Fig. 2). Within this area both pecan and water hickory grow wild, and pecan is also planted both as an ornamental and as a crop tree. Populations of each tree species were sampled for target insects throughout the growing season. For each insect, a minimum of three populations were sampled for each tree species (Table 2) with the two sites furthest apart for each tree species separated from one another by at least 80 km. Our aim was to test the role of host-plant species in promoting reproductive isolation of insects while accounting for geography. As a starting point, we sought to capture between 15 and 20 individuals of each herbivore species from each tree species. After genotyping was completed, the data set was evaluated with the SESim method (Medina et al., 2006) to determine whether individual and marker sampling was adequate. If it was not, more molecular markers and/or more individuals could have been added to the project. Within this framework, we characterized at least three sites per host-plant species and sought to obtain herbivores from as many trees as possible within a site. We characterized sites and identified trees within each site, which we sampled regularly throughout the summers of 2007 and 2008. As new sites and trees were discovered, we incorporated them into our sampling effort. Maximizing the number of trees represented within a site was also the goal when sub-sampling individuals for AFLP work. Where possible, each individual genotyped is from a different tree (Table 2).

Figure 2.

 Location of pecan and water hickory populations sampled for insects within the central Texas study area.

Table 2.   Site names, locations and number of insects genotyped per site.
SiteLocation (degrees decimal)Tree speciesBud moths genotypedYellow aphids genotyped
Sommerville Wildlife Management Area96.742 W
30.318 N
Water hickory2 (1 tree)0
USDA-pecan genetics96.434 W
30.517 N
Pecan10 (4 trees)5 (5 trees)
North College Station96.328 W
30.616 N
Pecan 5 (3 trees)
Tabor96.365 W
30.789 N
Water hickory 4 (4 trees)
Lick Creek Park96.222 W
30.561 N
Water hickory 4 (4 trees)
Navasota96.127 W
30.441 N
Water hickory10 (5 trees)0
Pecan7 (1 tree)0
Jewett96.147 W
31.346 N
Pecan 5 (3 trees)
Fort Boggy State Park95.979 W
31.189 N
Water hickory7 (2 trees)5 (3 trees)
Pecan1 (1 tree)0

Insects were reared to adulthood in the laboratory to make sure the genotyping was not complicated by parasitoids and to make sure the insects genotyped had developed completely on the host-plant species they were collected from. Pecan bud moths were collected as caterpillars during 2007 and 2008 from foliage and from inside Phylloxera sp. leaf galls and reared to adulthood in two ounce Cometware glasses (WNA, Covington, KY, USA). Fresh leaflets from the caterpillar’s host trees of origin were added weekly to the rearing glasses until pupation. Because we did not know if galls collected contained caterpillars until they emerged, survival data are not available for the period preceding emergence from galls. Following emergence from galls, survival of pecan bud moth was 100%. Yellow aphid nymphs were collected in 2008 from the underside of infested leaflets and reared to adulthood in 16 ounce Newspring DELItainer® (Pactiv Corp., Lake Forest, IL, USA) containers with fresh leaflets from their host tree of origin provided weekly. Nymphs were commonly collected from large, mixed species infestations of aphids and were not counted or identified to species until maturation. Thus, laboratory survival rates on the two hosts were unknown. When individuals of both species matured, they were frozen at −80 °C for genetic analysis or saved for vouchers.

DNA isolation and AFLP reactions

Comparable numbers of each insect species were genotyped from each tree species (Table 2). Whole genomic DNA was extracted from individual insects using a DNeasy blood and tissue kit (Qiagen Corp., Valencia, CA) following the manufacturer’s instructions. AFLP profiles were generated from ∼60 ng of DNA from each sample (Vos et al., 1995; Saunders et al., 2001; Gompert et al., 2006) using the following selective primer pairs: Mse1-CTC/EcoR1-AAC and Mse1-CAT/EcoR1-ACT. Polymerase chain reactions (PCRs) were run in GeneAMP® 9700 thermocyclers and diluted amplified selective products were submitted to fragment analysis on an ABI 3130 capillary sequencer with a co-loaded fluorescent (GeneScan 400HD [ROXDye]) size standard ladder (Applied Biosystems, Forest City, CA, USA) according to manufacturer’s instructions. Thermocycling conditions were as follows: the samples undergoing preselective amplification were held at 95 °C for 1 min followed by 20 cycles of 95 °C for 10 s, 56 °C for 30 s and 72 °C for 90 s followed by a hold at 75 °C for 5 min. For the selective amplification, samples were held at 95 °C for 30 s followed by 47 cycles of 95 °C for 10 s, 65–56 °C for 40 s and 72 °C for 90 s, followed by a hold at 75 °C for 5 min. The second temperature in the selective amplification cycle started at 65 °C and was lowered by 0.7 °C for the first 12 cycles until it reached 56 °C. Absence of contamination was assured by negative controls, and accuracy and repeatability of DNA fingerprints within species were verified by repeating all PCR steps for one individual from each species. For each insect and selective primer combination, resulting electrophenograms were examined and analysed using GeneMapper® 4.0 (Applied Biosystems) with the default allele calling threshold of 100 reflectance units. Selective primer combinations were then consolidated into a single 1/0 matrix for each insect species.

The SESim statistic (Medina et al., 2006) was calculated for each insect species to determine whether individual and molecular marker sampling was adequate for the HAD study. The two selective primer combinations produced 235 and 79 polymorphic loci for pecan bud moth and yellow pecan aphid, respectively, which gave SESim values of 0.034 and 0.038 for pecan bud moth and yellow pecan aphid, respectively. As Medina et al. (2006) showed that population structure began to break up because of inadequate sampling when SESim values were > 0.05, we determined that marker and individual sampling was adequate for our study.

Data analysis

The AFLP phenotype data matrix for each insect species was analysed independently. Bayesian cluster analyses were executed in STRUCTURE 2.2 (Pritchard et al., 2007) using the recessive alleles model for dominant marker data assuming admixture and correlated alleles (Falush et al., 2007). Admixture is a general attribute of most species occurring in sympatry, and the “alleles correlated” model deviant has been shown to be the most sensitive to the presence of population structure in simulated data (Falush et al., 2003). STRUCTURE assumes that within a population, loci are in Hardy-Weinberg equilibrium and linkage equilibrium and assigns individuals to separate populations so as to eliminate violations of these assumptions. The output of STRUCTURE is the log probability of the data (X) given the number of clusters (K) assumed or [Ln Pr(X|K)]. Where parameter estimates indicated K > 1, the ad hocΔK statistic (Evanno et al., 2005) was used to predict the most likely number of clusters (K) in the data. This involves calculating the second-order rate of change of [Ln Pr(X|K)]. Evanno et al. (2005) showed that the K corresponding to a spike in this value accurately predicts the number of populations represented by the data. Additionally, the most probable number (K) of clusters present in the data was determined using Bayes’ law to calculate the probability of the number of clusters (K) given the data (X) or Pr(K|X); equation 4 in Pritchard et al. (2007). The model was run for 100 000 generations with a burn-in period of 10 000 generations for 20 iterations each from K = 1 to K = 7. Because pecan bud moth had a high number of alleles at low frequency, the parameter lambda was first inferred (Pritchard et al., 2007) and found to be 0.43. The default lambda of 1.0 was used for yellow pecan aphid.

AFLPsurv 1.0 (Vekemans, 2002) was used to estimate genetic diversity, percent polymorphic loci and Fst among host-plant species and among collecting sites using Bayesian analyses with a nonuniform prior distribution of loci (Zhivotovsky, 1999) and the estimation procedures of Lynch & Milligan (1994). The estimated Fst values were tested against the null hypothesis Fst = 0 using 9999 random permutations of the data. amova (Excoffier et al., 1992) and principal coordinates analyses (PCO) were conducted among host plants and among collecting sites within host plants using the software GenAlEx 6.2 (Peakall & Smouse, 2006). For pecan bud moth, the Somerville and Fort Boggy pecan samples were removed prior to the amova because of low sample sizes but they were retained in separate amova analyses testing host plant and site separately.


Pecan bud moth

Runs of STRUCTURE 2.2 indicated that the model parameter, alpha, fluctuated greatly over the course of a STRUCTURE run, an indication of a lack of population structure (Pritchard et al., 2007). Additionally, all individuals were assigned with relatively equal probability to both hypothetical populations when K was set to 2 (Fig. 3), further indication of a lack of population structure. The most probable number of populations was K = 1; Pr(K|X) = 1. Triangle plots (not shown) also produced single clusters, even when running STRUCTURE for K > 1.

Figure 3.

 Bayesian population assignment probabilities (y-axis) for pecan bud moth individuals (x-axis) collected from pecan and water hickory using the recessive alleles model for dominant marker data in STRUCTURE 2.2. All individuals are assigned with relatively equal probability to both hypothetical populations when two populations are assumed indicating a lack of population structure.

Expected heterozygosity was similar among sites and among host plants (Table 3), and Fst was not significantly different from zero by site (Fst = 0.0188, P = 0.152) or by host plant (Fst = 0.0017, P = 0.1425). Ninety-nine percentage of molecular variation occurred within collecting sites (Table 4), and there was no significant effect of host-plant species or collecting sites. PCO 1 and 2 explain 44.97% of the molecular variation and show no pattern by either host plant or collecting site (Fig. 4a).

Table 3.   Diversity statistics for pecan bud moth and yellow pecan aphid.
  1. N, sample size; #Loci, number of polymorphic amplified fragment length polymorphism loci genotyped for the species; #PL, number of polymorphic loci for a given collecting site or host plant; %PL, percent polymorphic loci; Hj, expected heterozygosity (Nei’s gene diversity) for a given collecting site or host plant.

Pecan bud moth
 Collecting site
  Pecan genetics1023517172.80.17494
  Fort Boggy823515164.30.18357
 Host plant
  Water hickory1923515967.70.17237
Yellow pecan aphid
 Collecting site
  Pecan genetics5795670.90.24107
  College Station5794658.20.1869
  Lick Creek Park47949620.20309
  Fort Boggy57949620.20047
 Host plant
  Water hickory13795265.80.2006
Table 4.   Variance components (VC) and percent molecular variation because of host plant and collecting sites within host plant for pecan bud moth and yellow pecan aphid from amova. Significance testing was performed using 9999 permutations of the binary distance parameter ϕPT, an analog of Fst in GenAlEx 6.2.
Source of variation
Pecan bud mothYellow pecan aphid
VCPercentage of variationP-valueVCPercentage of variationP-value
Among host plants0.27010.16016.66571< 0.001
Within host among sites0.00000.6700.21510.314
Within collecting sites26.40699 6.52028 
Figure 4.

 Eigenvectors of principal coordinates (PCO) 1 (x-axis) and 2 (y-axis) for (a) pecan bud moth and (b) yellow pecan aphid. Symbol shapes denote collecting sites, and symbol colours denote host-plant species. PCO 1 separates the water hickory (black) and pecan (grey) populations of yellow pecan aphid.

Yellow pecan aphid

At K = 2, all individuals were assigned with high (> 99.7%) probability to one of the two populations (Fig. 5). Furthermore, the two populations detected by STRUCTURE corresponded exactly to host-plant species of origin. Ten loci were diagnostic for the pecan host race, and seven loci were diagnostic for the water hickory host race (Table 5). An additional six loci were strongly host-plant associated, being present in > 90% of individuals from the associated host and in < 10% of individuals from the host’s congener (Table 5). The second-order rate of change in K, (Evanno’s ΔK) peaked for K = 2 populations (data not shown) further indicating that two populations best explained the yellow pecan aphid data. The model produced gradually increasing values of LnPr(X|K) with K > 2 but in these cases, all individuals were assigned with high (> 86%) probability to one of the two host-plant-associated populations.

Figure 5.

 Bayesian population assignment probabilities (y-axis) for K = 2 populations of yellow pecan aphid individuals (x-axis) collected from pecan and water hickory using the recessive alleles model for dominant marker data in STRUCTURE 2.2. Two host-associated populations (light grey and dark grey) are indicated.

Table 5.   Frequency of the 23 most strongly host-associated amplified fragment length polymorphism loci in the pecan and water hickory host races of yellow pecan aphid. The selective primer combination used and the size of the DNA fragment in base pairs comprise the locus name.
LocusPecan (%)Water hickory (%)

Expected heterozygosity was similar among sites and among host plants (Table 3). The Fst among host plants was significantly different from zero (Fst = 0.5752, P < 0.0001) but among sites within host-plant species Fst was not significantly different from zero (pecan by site Fst = −0.0029, P = 0.6550; water hickory by site Fst = 0.0154, P = 0.3143). Seventy-one percentage of molecular variation occurred among host plants (Table 4) but only 1% of molecular variation occurred among sites within host-plant species (Table 4). PCO 1 and 2 explain 88.87% of the molecular variation with principal coordinate 1 strongly segregating aphids by host-plant species (Fig. 4b).


The parthenogen and the quasi-endophage

We conclude panmixia for pecan bud moth among populations sampled from pecan and water hickory. We also conclude that the yellow pecan aphid exhibits HAD and consists of at least two distinct host races, one feeding on pecan, and one feeding on water hickory. Host races are diagnostic at 17 of 79 (> 20%) of polymorphic loci indicating strong barriers to gene flow between them (Table 5). A possible reason for the presence of HAD in yellow pecan aphid but not in pecan bud moth could be parthenogenesis. If a mutant aphid colonizes a novel host that mutation can be amplified greatly over the course of a growing season. When males are produced, inbreeding is likely and will occur on the novel host (Dixon, 1998). This scenario may be particularly prevalent in monecious aphids (aphids that do not migrate to a second host plant for reproduction), and yellow pecan aphid is monecious. Sexual females are apterous and cannot leave the natal host. In contrast, males, while winged, may suffer a fitness penalty for leaving the natal host if they have the same mutation for host use as their mother. Male aphids inherit the entire genome minus one X chromosome and so are likely to inherit such mutations intact (Ward, 1991; Guldemond & Mackenzie, 1994). Unlike the parthenogenetic yellow pecan aphid, pecan bud moth reproduces sexually. Compared to parthenogens, sexually reproducing organisms experience more gene flow, which counteracts differentiation (Slatkin, 1973; Hendry et al., 2001; Nosil, 2009).

Endophagy has also been implicated in the propensity to show HAD and pecan bud moth, a quasi-endophage, was expected to show HAD because of its intimate larval association with Phylloxera galls. Despite this, the pecan bud moth population was found to be panmictic and did not consist of genetically differentiated host-associated populations. In contrast, yellow pecan aphid is not an endophage and yet does show strong evidence for pecan and water hickory host-associated populations.

Several authors have suggested univoltinism as one factor facilitating HAD because univoltine insects must be linked very tightly to the phenology of their hosts (Nyman, 2002; Mopper, 2005). For example, the apple and hawthorn host races of the univoltine apple maggot break diapause to coincide with the fruiting of their respective hosts as does their parasitoid (Feder & Filchak, 1999; Forbes et al., 2009). In contrast, neither of the insects under study is strictly univoltine; both are present and feeding on foliage throughout the growing season. However, yellow pecan aphid males are univoltine, only produced late in the season. Tedders (1978) found males from October 14th to November 22nd on pecan in Georgia. This type of sexual univoltinism is a general feature of the biology of aphids and may contribute to aphid susceptibility to HAD. Thus, if the timing of sex is coupled to differential host-plant phenologies, reduced gene flow between host races should be favoured over time (Stam, 1983; Butlin, 1990; Emelianov et al., 2003).

Some authors argue that enemy free space can be an important driver of diversification (Feder, 1995; Stireman et al., 2008). Unfortunately, the dominant predators of pecan bud moth are not known, and no parasitoids have been documented nor were any reared during this study. Thus, we can say very little at this time regarding the possibilities of third trophic level effects on pecan bud moth on either host-plant species. We reared two parasitoids from yellow pecan aphid, Aphelinus perpallidus (Gahan) and an unidentified braconid. Aphelinus perpallidus was the most common parasitoid accounting for > 99% of parasitized aphids. Documenting parasitism rates of yellow pecan aphid by this parasitoid is ongoing but preliminary data suggests that parasitism rates do not differ significantly between the two host races (likelihood ratio test χ21 = 0.016, P = 0.9; 13 parasitoids from n = 109 aphids on pecan, eight parasitoids from n = 71 aphids on water hickory). Thus, preliminary data suggest that neither host race of pecan aphid enjoys reduced attack, at least from A. perpallidus.

Pecan bud moth and yellow pecan aphid HAD tests represent an informative comparison because they are herbivorous insects on the same host-plant species pair in the same geographic location. However, this comparison is not without caveats, first, it is but a single data point and more species should be tested within this host-plant system to determine the extent to which generalities can be made. Second, this was a mesoscale geographic study and did not encompass the entire geographic range of the insects in question or their host plants. Therefore, our findings may not translate to larger geographic scales, or to other localities. Also, pecan bud moth feeds on foliage and inside buds in addition to Phylloxera gall tissue (Mizell & Schiffhauer, 1986). Thus, although it may be common to find pecan bud moths in Phylloxera gall tissue, perhaps it should not be considered an obligate endophage for the sake of comparison in the way that the inquiline beetle, Mordellistena, is on Eurosta goldenrod galls (Eubanks et al., 2003).

Aphids in general exhibit very strong host specificity with 99% of species considered specialists (Eastop, 1973). We found literature reports of host races or host-associated aphid genotypes in 18 aphid species (Table 6). These reports extend back over 150 years when Francis Walker recorded a host switch of Aphis persicae from sloe to peach following the introduction of the latter to England (Walker, 1850). Most cases of HAD in aphids involve food crops but recently, Peccoud et al. (2009) found 11 host races of the pea aphid Acyrthosiphon pisum throughout Europe, eight of which were specific to wild hosts. This suggests that genetic diversification of aphids on wild host plants is probably much greater than previously thought. Our results add further evidence that surveys of wild host plants are likely to result in the discovery and genetic resolution of increased numbers of host differentiated aphids.

Table 6.   Aphids showing host-associated differentiation.
Aphid speciesHostsAuthors
Acyrthosiphon pergoniiAt least four host racesDixon (1998)
Acyrthosiphon pisumAt least 11 host racesPeccoud et al. (2009)
Acyrthosiphon solaniMultiple hostsDixon (1998)
Amphorophora sp.Rubus sp.Blackman et al. (1977)
Aphis fabaeVicia sp., Tropaeolum sp.Dixon (1998)
Aphis frangulaeMultiple hostsDixon (1998)
Aphis gossypiiAt least five host racesCarletto et al. (2009)
Aphis spireaCitrus sp., Spirea sp.Dres & Mallet (2002)
Cryptomyzus galeopsidisRibes sp.Dres & Mallet (2002)
Dysaphis crataegiCrataegus sp., Daucus sp.Dixon (1998)
Eriosoma yangiUlmus sp.Akimoto (1988)
Hyalopterus amygdalePrunus sp.Lozier et al. (2007)
Monelliopsis pecanisCarya sp.Current work
Myzus cerasiPrunus sp.Guldemond & Mackenzie (1994)
Myzus persicaeNicotania sp., multiple hostsMargaritopoulos et al. (2007)
Schizaphis graminumAt least three host lineagesAnstead et al. (2002)
Sitobion avenaeAt least three host lineagesVialatte et al. (2005)
Tetraneura yezoensisUlmus sp.Akimoto (1990)
Uroleucon sp.Centaurea sp., Cirsium sp.Guldemond & Mackenzie (1994)

Testing the importance to HAD of parthenogenesis per se will be challenging because there are other aphid traits proposed to facilitate HAD; host alternation, host specificity and the relative commonness of apterous sexuals (Dixon, 1998), in addition to univoltinism of sex (mentioned previously). Furthermore, apomictic parthenogenesis is likely the ancestral state of the Aphidomorpha (Dixon, 1998), and it is not common to other herbivorous taxa making it difficult to test for HAD in taxonomically diverse parthenogens as one can with taxonomically diverse endophages. Two possible exceptions to this could be thrips (Order Thysanoptera) and sawflies (Family Tenthredinidae), some of which have been shown to be parthenogenetic (Arakaki et al., 2001; Muller et al., 2004; Nault et al., 2006). There have been five species of thrips and three species of sawfly documented on pecan (Smith et al., 1996; Ree, Pers. Comm.). Pecan and water hickory share at least one sawfly species, Periclista marginicollis (Norton) but it is not parthenogenetic. In addition to aphids, future impetus should be given to uncovering taxonomically diverse parthenogens in this system if possible. If none are found, shared haplo-diploid herbivores such as mites (Order Acari), sawflies (Family Tenthredinidae) and thrips should be tested for HAD as haplo-diploidy is predicted to promote a level of recombination intermediate to that of parthenogenesis and sexual reproduction (Hartl, 1972).

Carya hickories: an ideal system for HAD investigation

The North American hickories selected for this study offer many advantages for HAD study: (i) they are native, (ii) they are long-lived trees, (iii) they are sympatric over a fairly large part of their native range, (iv) they are ecologically similar – both bottomland species and so can be found in mixed stands, (v) they are genetically similar – both are diploid as opposed to tetraploid Carya species, and (vi) they host many herbivores (over 400 species documented on pecan alone). The results from our first two shared species tested provide impetus to test additional parthenogens in this system. Pecan and water hickory share at least seven species of aphids, three of them endophagous phylloxerans (B. Ree, Pers. Com.; A. Dickey, Pers. Obs.). It will be interesting to see if the abundance of shared parthenogens will make this system a good counterpoint to the Solidago system, which has an abundance of shared endophages.


We thank two anonymous referees whose suggestions improved the quality of this manuscript. We gratefully acknowledge Gary Miller USDA-SEL, John Brown USDA-SEL and LJ Grauke USDA-ARS who confirmed the aphid, moth and tree identifications, respectively. We also appreciate access to collecting sites provided by Bobby Murphy, Wendel Legally and Pat Caraway. The Texas Parks and Wildlife Department provided access to Fort Boggy State Park and Sommerville Wildlife Management Area. College Station Department of Parks and Recreation provided access to Lick Creek Park. LJ Grauke USDA-ARS provided access to the USDA pecan genetics facility. This research was funded in part by the Department of Entomology at Texas A&M University.