Jennifer A. Brisson, School of Biological Sciences, University of Nebraska, Lincoln, NE 68588, USA. Tel.: +402 613 4135; e-mail: firstname.lastname@example.org
Little is known about when, how or even if the wing development gene network elucidated in Drosophila is deployed in direct-developing insects. Here we identify the wing development genes (as determined in Drosophila) of the pea aphid (Acyrthosiphon pisum), which produces winged or unwinged adults in response to environmental cues. We find that the principal wing development genes studied in Drosophila are present in the aphid genome and that apterous and decapentaplegic exhibit duplications. We followed expression levels of 11 of these developmental genes at embryogenesis and across the nymphal instars. Six showed significant stage-specific expression level effects and apterous1 exhibited significantly different expression levels between winged and unwinged morphs, suggesting this gene acts proximately to realize polyphenic development.
Wings evolved in insects approximately 300–400 million years ago, either de novo from an outgrowth of the body wall (reviewed in Snodgrass, 1935) or by modification of existing limb branches (Wigglesworth, 1973; Kukalova-Peck, 1983). The evolution of wings occurred early (Fig. 1, based on information from Jockusch & Ober, 2004; Kjer, 2004) and flight was a successful innovation. Consequently, wing forms are known in most modern insect orders and wing morphologies have diverged greatly. Today, Odonata (dragonflies and damselflies) and Ephemeroptera (mayflies) best represent the four-winged ancestors. Insects in the remaining orders have folded wings, with various modifications to the fore and hind-wings such as the elytra of beetles or the halteres of flies.
The underlying developmental processes that produce adult wings have also diverged among insect orders. Holometabolous development, also known as indirect development or complete metamorphosis, involves a dramatic pupal reorganization between juvenile and adult phases. In holometabolous insects, the wing develops directly from the ectoderm or from sequestered imaginal discs (Truman & Riddiford, 1999; Jockusch & Ober, 2004). Most of what we understand about holometabolous wing development comes from decades of studies in Drosophila melanogaster (reviewed in Serrano & O'Farrell, 1997). In Drosophila, imaginal discs are set aside during embryonic development (Cohen, 1993). Patterning of these discs occurs during the larval instars (Bryant, 1978). During pupation, the discs evert and the wing appears similar to what will be seen in the adult, albeit in a compacted form (Turner & Adler, 1995).
In contrast, in hemimetabolous development, also known as direct development or incomplete metamorphosis, the insect traverses a series of juvenile stages that look like miniatures of the adult. There is no reorganizational pupal stage. In these insects, nymphs emerge from the embryonic stages with a wing bud that extends from the thoracic body wall. The wing bud develops slowly in each nymphal instar until the fully formed wing unfolds after the final molt to adulthood; there is no embryonically sequestered imaginal wing disc.
Hemimetabolous development is the evolutionarily ancestral form of ontogeny (Fig. 1; Jockusch & Ober, 2004), but we know little about the underlying molecular biology of wing patterning and development in hemimetabolous insects. An opportunity for investigating this issue presents itself in aphids, a clade of about 4400 species in the order Hemiptera. Most aphid species exhibit a wing polyphenism, meaning that a single aphid genotype can produce dramatically different winged or unwinged adult morphologies, depending on the environmental conditions the aphid experiences during development (reviewed in Müller et al., 2001; Braendle et al., 2006). As in other insect wing dimorphisms (reviewed in Zera & Denno, 1997), the unwinged morph specializes in reproduction while the winged form specializes in dispersal (Dixon & Howard, 1986).
Our first objective here was to identify wing development gene homologues in the recently sequenced pea aphid (Acyrthosiphon pisum) genome. A number of genes involved in the patterning, growth and differentiation of the wing in Drosophila have been identified, either through genetic studies or genome-wide microarray analysis (Campbell et al., 1993; Sturtevant & Bier, 1995; Weatherbee et al., 1998; Butler, 2003; Ren et al., 2005; Kiger et al., 2007). Many of these genes are highly pleiotropic across development so, although we will refer to these genes as ‘wing development genes’ for simplicity, we do so with this caveat. We depict a simplified network of the genes underlying wing development in Fig. 2.
Our second objective was to determine whether some of these developmental genes are expressed differentially between winged and unwinged morphs. In the pea aphid wing polyphenism, unwinged asexual females, when crowded, produce genetically identical daughters that grow up to be winged (Sutherland, 1969). To investigate the proximate mechanisms underlying the wing polyphenism in aphids, here we examine expression levels of a subset of the developmental genes considered above, specifically engrailed (en), hedgehog (hh), decapentaplegic (dpp), apterous (ap), wingless (wg) Distalless (Dll), homothorax (hth), and Ultrabithorax (Ubx). We chose these genes in particular because they are involved in major wing patterning events, such as anterior–posterior (A-P) patterning genes (en, hh, dpp) (Tabata et al., 1992; Basler & Struhl, 1994; Tabata & Kornbert, 1994), dorsal-ventral (D-V) patterning genes (ap, wg, Dll) (Cohen et al., 1992; Campbell et al., 1993; Kim et al., 1995), a wing hinge development gene (hth) and a Hox gene (Ubx) (Struhl, 1982). We examined their expression levels in RNA extracted from whole individuals (minus embryos) across development. Thus, any differences between the two morphs may reflect their role in wing development and/or their role in other processes.
Pea aphid orthologs of wing patterning genes
Portions of the wing development gene network are conserved among holometabolous insect orders (Carroll et al., 1994; Weatherbee et al., 1999; Abouheif & Wray, 2002); are they also conserved between holometabolous and hemimetabolous insects? We began our investigation into this question by searching the pea aphid genome for orthologues of genes known to be involved in wing patterning in Drosophila and another recently sequenced insect, the red flour beetle Tribolium castaneum. In particular, we used a previously generated list of wing development genes to guide our search (Table S13b of Richards et al., 2008). We found every wing development gene that we looked for in the pea aphid genome.
As can be seen from Table 1, two wing development genes exhibit duplications. apterous (ap), also known as LIM-2 in vertebrates, is a homeodomain (HD) containing transcription factor that has two protein interaction LIM domains (Cohen et al., 1992). We aligned the amino acid sequences from the HD and LIM domain regions of the two pea aphid ap paralogs with ap orthologues from Drosophila melanogaster, Tribolium castaneum[which also has two copies of ap, Richards et al. (2008)], and Apis mellifera. The HD region remains extremely well conserved across these taxa and the LIM domain regions also show considerable conservation (Fig. 3A). To this alignment, we added the closely related LIM-HD-containing genes tailup (tup), Arrowhead (Awh), and LIM3 from these same taxa (see Experimental Procedures section for protein IDs; we did not identify an A. mellifera tup). The pea aphid ap paralogs clearly fall into a clade with ap copies from the other species (Fig. 3B). We conclude from this tree that ap1 and ap2 are ap homologues as opposed to homologues of other LIM-HD containing genes.
Table 1. List of wing development genes in the pea aphid
Genes in bold text are those used for gene expression level analysis in this study.
cubitus interruptus (ci)
No Ref. Seq.
No Ref. Seq.
No Ref. Seq.
No Ref. Seq.
Daughters against dpp (dad)
No Ref. Seq.
No Ref. Seq.
spalt major (sal)
No Ref. Seq.
No Ref. Seq.
No Ref. Seq.
No Ref. Seq.
Sex-combs reduced (Scr)
No Ref. Seq.
No Ref. Seq.
ventral veins lacking (vvl)
Drosophila dpp, the homologue of the vertebrate transforming growth factor-β family, encodes a secreted signalling molecule (Padgett et al., 1987). The pea aphid genome contains four putative dpp copies. As Shigenobu et al. (2010) report, the dpp paralogs fall into two classes, with dpp1 in one class and dpp2-4 in the other. Figure 1 of Shigenobu et al. (2009) illustrates the relationship among the pea aphid dpp paralogs and dpp homologues from other insects.
Quantifying expression levels of wing development genes across development in unwinged and winged morphs
All aphids in the parthenogenetic generations are born through viviparous reproduction as first instar nymphs containing wing buds. Only during the second instar do these buds degenerate in the unwinged morphs (Tsuji & Kawada, 1987; Ishikawa et al., 2008). In the winged morphs, the wing buds continue to slowly grow through the first three nymphal instars, and in the fourth instar the wing buds and flight muscle primordia are well developed.
We assayed gene expression levels of a subset of wing development genes during embryogenesis and the four nymphal instars (the genes in bold text in Table 1). For details on how we collected our samples, see the Experimental Procedures section. Aphids of the F1 clone, previously described in Braendle et al. (2005) and assayed for transcriptional differences between fourth instar nymphs and adults of winged and unwinged morphs (Brisson et al., 2007), were used for all experiments in this study. Thus all aphids were of the same genotype. Ovaries were removed from all nymphs prior to RNA isolation to avoid confounding effects of embryonic expression. We performed quantitative PCR (qPCR) on each of three biological replicates of each morph and each nymphal instar for the genes in bold text in Table 1. Figure 4 illustrates the results.
Gene expression of dpp and ap paralogs. As mentioned above, the pea aphid has two ap paralogs and four dpp paralogs. In order to determine whether the paralogs display different temporal expression levels or different expression levels between the two wing morphs, we performed qPCR on all six of these putative genes (see the Experimental Procedures section for specifics on primer design). dpp4 was not expressed or we designed our primers to amplify an untranscribed region: although the PCR worked efficiently on genomic DNA, no amplification was observed using cDNA. All other duplicates (ap1, ap2, dpp1, dpp2, dpp3) did have expression levels above background levels.
We wanted to compare the expression profiles of the two ap paralogs across development to determine if they were expressed in different manners. Because each paralog has its own expression scale, first we normalized each sample's expression level by subtracting the average expression level for that gene across the stages from the expression level for that sample and then dividing the resulting value by the standard deviation. We used a two-way anova examining stage and paralog to compare the expression profiles of the two ap paralogs across development. We detected a significant interaction term for the unwinged morph (P= 0.002) but not the winged morph. We found no difference among the dpp paralogs when analysed in the same manner.
Gene expression differences by developmental stage and wing morph. We used two-way anovas to determine whether gene expression levels were significantly different across the four post-embryonic time points, and whether this variation was affected by wing morph type. Dll, en, hh, hth, wg, and dpp1 all showed significance of developmental stage, ap1 and ap2 exhibited a significant stage by morph effect, and ap1 had a significant morph effect (see Fig. 4 for P-values for each gene).
We compared gene expression levels between the winged and unwinged morphs at each stage of development separately for ap1 and ap2 because they showed either morph or morph by stage effects in the two-way anovas. As described in the Experimental Procedures section, we were unable to distinguish winged from unwinged first or second instar nymphs from external morphology. The winged vs. unwinged samples for these stages were nymphs that resulted from crowding the mother vs. the uncrowded treatment, respectively. These samples likely contain the opposite morph at low numbers. Any expression differences between the two samples are therefore likely underestimates of true differences between winged and unwinged morphs at these two stages.
We detected ap1 at significantly higher levels in winged relative to unwinged first and second instar nymphs (1-sided t-test, P= 0.005 for each; P= 0.04 after Bonferroni correction). ap2 was not significantly higher in winged first instar nymphs after multiple comparison correction (1-sided t-test, P= 0.044; P= 0.352 after Bonferroni correction).
Correlation of gene expression levels across stages. To further explore the similarities and differences among the gene expression profiles between the unwinged and winged morphs, we determined all pairwise non-centered Pearson's correlations among gene expression levels across the developmental stages in the winged and unwinged morphs separately. The winged morph displayed a slightly more tightly correlated pattern of expression when compared to the unwinged morph, as evidenced by a t-test of pairwise correlations among gene expression values from the winged morphs vs. those from the unwinged morphs (2-sided t-test, P= 0.019).
Here we provide the first report of the wing development gene repertoire in a hemimetabolous insect, the pea aphid. Our overall objectives were to identify the wing development genes in the pea aphid genome, discover the expression levels of these genes across the embryonic and postembryonic stages of development, and determine whether any of these genes exhibit differential expression levels between winged and unwinged morphs.
Wing development genes in the pea aphid genome: ap and dpp exhibit duplications
Genes involved in wing development have been discovered working with holometabolous insects, primarily D. melanogaster. Guided by these previous studies, we searched the pea aphid genome for the wing development genes listed in Table 1. We found every gene for which we looked, suggesting that despite the large differences of developmental processes between holometabolous and hemimetabolous insects, the aphid genome contains all of the major components of wing development studied in Drosophila. We expected this result given that many of these genes have multiple roles in development and are therefore likely necessary for survival of the organism.
Two of these wing development genes had experienced duplications, with two ap and four dpp paralogs. As reported by the International Aphid Genomics Consortium (2010), the pea aphid genome shows a large number of gene duplications. The functional importance of these gene duplications remains unknown. The presence of only one ap copy in both Drosophila and Apis compared to two copies in the pea aphid and Tribolium suggests that there have been multiple ap losses within the holometabolous lineage, or that a more complicated pattern of gains and losses has occurred. Only further phylogenetic sampling will resolve this question.
What role do these paralogs play across the developmental stages of the pea aphid? In Drosophila, ap expression is found in the embryonic and larval nervous system (Cohen et al., 1992; Lundgren et al., 1995), embryonic muscles (Bourgouin et al., 1992), and larval and pupal imaginal discs (leg, wing, haltere, and eye-antennal) and brain (Cohen et al., 1992). The only reported adult expression is in the brain (Cohen et al., 1992). It is probably most appropriate to compare the larval and pupal stages of Drosophila to the nymphal stages of the pea aphid.
In the pea aphid, the legs, eyes and antennae complete their patterning embryonically, as these structures exist in a miniature form of that found in the adult when they are born. In contrast, at birth the wing buds are present only as a slight protrusion from the body wall visible histologically (Ishikawa et al., 2008). The embryonic expression of ap1 and ap2 could be in any or all of these structures as well as in the developing nervous system. Postembryonically, however, the best candidate expression domains are the developing wing buds and nervous system/brain. The two paralogs are expressed differently across development when we compared them within the unwinged morphs (but not the winged morph), suggesting the hypothesis that they are involved in different functions. Only further study via in situ hybridization to RNA will illuminate their expression patterns in each morph.
dpp, a secreted signalling molecule, is expressed in the embryonic mesoderm (Panganiban et al., 1990), in all of the larval and pupal imaginal discs (Masucci et al., 1990; Blackman et al., 1991), and the larval brain (Kaphingst & Kunes, 1994) of Drosophila. As with ap, the dpp expression levels we assayed in the postembryonic stages of the pea aphid are unlikely to be the result of leg, eye or antennal expression given that these structures are more or less fully formed as nymphs. As reported above, the pea aphid genome contains four paralogs of dpp. We failed to detect expression of dpp4. We designed our primers to amplify the 3′ end of the predicted dpp4 sequence, to an exon missing from the other dpp paralogs. It may be that this region is not transcribed (i.e. a poorly predicted exon), or it may be that this is a pseudogene. The remaining dpp paralogs are not expressed in statistically different manners across the stages within either the winged morphs or the unwinged morphs. However, we did observe much higher levels of dpp1 compared with either dpp2 or dpp3. This could have been attributable to different amplification efficiencies, but it does present the hypothesis that dpp1 is the pea aphid dpp paralog responsible for most previously described dpp functions, and that the others are either expressed in very low levels or in a very small portion of cells. Again, only detailed further examination will resolve this question. Interestingly, dpp3 is relatively highly expressed in embryos, but at barely detectable levels during the nymphal instars of both wing morphs. Therefore, if it is involved in wing determination so it may act embryonically.
Gene expression levels between winged and unwinged pea aphids
Given that the pea aphid's wing buds develop slowly across the embryonic and post-embryonic stages of development, we had few a priori expectations regarding when the wing development genes would be expressed. We now have a preliminary map of the timing of their expression. Most genes showed both embryonic and post-embryonic expression. We observed a tendency of some genes, for example Dll, en, hth, and dpp1, to be expressed at higher levels during the earlier stages. Interestingly, we found a significant morph effect on ap1, and a significant morph by stage effect on ap1 and ap2 expression levels (Fig. 4). The strongest differences between the morphs were during the first and second nymphal instars, the periods in which unwinged morphs are discontinuing wing growth and may actively be degenerating their wing buds (Ishikawa et al., 2008). The ap paralogs could play a proximate role in wing polyphenism, with lower levels of either paralog in the unwinged morphs resulting in mispatterning of the D-V axis.
Overall, the expression levels of the wing development genes in the winged morphs are more tightly correlated than those in the unwinged morph. This suggests that wing development gene expression is deployed in a more coordinated fashion in wing morphs and that the wing development gene network deciphered in Drosophila may have similarities in the pea aphid.
An analogous case to the aphid wing polyphenism can be found in the caste polyphenism of ants, where wingless workers disrupt the wing developmental cascade at different points depending on the species (Abouheif & Wray, 2002). For example, in the ant species Pheidole morrisi, wg expression was not observed in worker wing discs but was present in the wing discs of winged castes (Abouheif & Wray, 2002). The aphid and ant wing polyphenisms are not homologous, but similar developmental mechanisms might achieve morph divergence in each case and therefore future studies might more precisely target wg. Here we observed a trend towards higher wg expression in winged morphs, although differences were not significant. Another candidate not examined here is spalt, which exhibits repressed expression in worker ant larvae.
Overall, this work provides initial insight into the wing development gene repertoire of the pea aphid and reveals that ap1 is expressed at different levels in the winged vs. the unwinged morphs. Our approach focused specifically on one downstream aspect of the polyphenism: wing development. As reviewed in Braendle et al. (2006), the polyphenism is determined embryonically, and the two adult morphs exhibit systemic differences that go well beyond the presence or absence of wings, including differences in overall morphology, behaviour and life history characteristics. More upstream, maternal and embryonic studies are required to reveal the molecular, mechanistic basis of the polyphenic switch.
Orthologue identification and tree building
To find pea aphid orthologues of wing development genes, we used the D. melanogaster protein sequence in a Blast search querying pea aphid Refseq or Gnomon predicted proteins from version 1.0 of the A. pisum genome. Hits with an e-value less than 1e-15 were examined, although the majority of the hits were much higher. If one was clearly more significant than the others, it was chosen as the putative pea aphid orthologue of the Drosophila sequence. The pea aphid protein was then used in a Blast search to query D. melanogaster proteins. If the Blast and the reciprocal Blast matched, then we assigned orthology to that sequence. In the case of apterous, we built a neighbor-joining similarity tree of protein sequences using the ClustalW program and bootstrapped it 1000 times to estimate confidence in the branches (Larkin et al., 2007). We gathered sequences from GenBank for ap (XP_970113, XP_001810983, XP_392622), Awh (XP_971202, XP_001121397, XP_001121365, XP_001949712), tup (XP_001944557, XP_001815525, XP_001815525), and LIM3 (Am XP_394135, Tc XP_973330, Ap XP_001944208).
Sample preparation for RNA extraction
Aphids of the F1 clone, previously described in Braendle et al. (2005), were used for all experiments in this study. One of the parentals of the F1 line is the LSR1 line. An inbred LSR1 line (LSR1.G1.AC) was used for genome sequencing, so there are limited numbers of polymorphisms between the F1 line and the pea aphid genome sequence.
Aphids were reared in an incubator at 19 °C with 16 light hours alternating with eight dark hours, in 15 mm Petri dishes, each with a leaf of Medicago arborea inserted into 3 ml of 1% agar containing 1 g/l Miracle-Gro. To collect winged and unwinged females, female asexual aphids were reared at low density (two to three aphids per plate) for three generations, initiating each generation with unwinged aphids. This was necessary because aphids exhibit a telescoping of generations in which an adult female contains within her ovaries embryos that contain embryos within their ovaries. Therefore, treatment of one generation can affect treatment of the next two generations. In the fourth generation, first instar females were reared in isolation to third-day adults and then placed in three groups of 10 in a 10 mm Petri dish containing a moistened piece of Whatman paper. The females were crowded for 24 h. For this line, this is sufficient to produce 80–100% winged offspring.
After the 24 h of crowding, the females were re-isolated in dishes to allow them to deposit nymphs. After 24 h, we dissected 10 first-instar nymphs, removing their ovaries and transferring the remainder of the body into 300 µl of Tri Reagent (Molecular Research Center, Inc., Cincinnati, OH). We collected no more than two nymphs from each female, so the nymphs were from a number of different mothers. Further, the nymphs of each sample were randomly selected from those deposited within that 24 h so there was no systemic biases in sampling due to deposition order. All remaining first instar nymphs were then placed individually in dishes with leaves and monitored for molting every day. Second instar nymphs were harvested in a similar manner on the first day of their respective instar. By the third and fourth instar, wing buds are outwardly visible. We collected only winged third and fourth instar nymphs, again on the first day after molting. Unwinged aphids were collected in the same manner, with the exception that their mothers were not crowded, and fourth instar nymphs without visible wing buds were collected. All aphids were harvested during the same time of day (between 1 and 2 pm) to avoid any effects of photoperiod on the collections. In total, we collected 10 individuals for each instar (four instars) and each treatment (crowded and uncrowded), with three replicates, for a total of 24 samples. In addition to the nymph collections, we also collected a sample of embryos by removing the ovaries from 10 3-day-old adult unwinged females that were uncrowded. This sample was included to provide an indication of the expression level of the genes of interest in the developing embryos.
RNA extraction and real time quantitative PCR
RNA was isolated from each sample using a phenol/chloroform extraction. We treated each sample with rDNaseI (Ambion, Austin, TX, USA) at 37 °C for 30 min. We synthesized cDNA using the High Capacity cDNA Reverse Transcription kit (Applied Biosystems, Foster City, CA, USA). We used real time qPCR to quantify the relative transcript accumulation between winged and unwinged morphs. Primers to target transcripts were designed with Primer3 (Rozen & Skaletsky, 2000) and are listed in Table 2. We did not design primers to span introns. 5 µl of cDNA or water (for negative control reactions) was added to a total PCR reaction volume of 25 ml containing 2X Power SYBR Green Master Mix (Applied Biosystems). Samples were run on a DNA Engine Opticon 2 Continuous Fluorescence Detection System (Bio-Rad, Hercules, CA, USA) for 10 min at 95 °C and then 40 cycles of 15 s at 95 °C followed by one minute at 60 °C. Primer specificity was verified using disassociation curve analysis.
Table 2. Oligonucleotide primers used for quantitative PCR
Designing primers to the specific paralogs of ap and dpp required extra attention. In the case of ap1 and ap2, the paralogs exhibited very little similarity at the nucleotide level in most regions and therefore it was not difficult to find primer pairs specific to the paralog of interest (see Supporting Information Fig. S1 for a nucleotide alignment of exons with the primers indicated).
We initially considered both actin and Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as our endogenous controls. We used the program GeNorm (Vandesompele et al., 2002, 2007) to assess their stability and thus utility as an endogenous control. Based on this assessment, we used GAPDH as our control. Therefore, each gene of interest was quantified in each sample relative to the expression level of GAPDH.
Thank you to Shuji Shigenobu for his help with annotation, to Ryan Bickel for comments on an earlier draft of this manuscript, and to three anonymous reviewers for their thoughtful critiques. JAB was supported by award number K99ES017367 from the National Institute of Environmental Health Sciences. This study was partly supported by a Grant-in-Aid for Scientific Research (B) (No. 18370007) and by a Grant-in-Aid for Challenging Exploratory Research (No. 20657004) of the Ministry of Education, Culture, Sports, Science and Technology of Japan to TM. AI was supported by a JSPS Research Fellowship for Young Scientists.