We modeled the insect proteome by selecting the subset of Drosophila protein sequences with homology to predicted genes in all insect-species studied here. Similarly, we defined the subset in Drosophila common to the eukaryote species studied here: mosquito, silkworm, beetle, honeybee, human, nematode and yeast. Because at this time it is not possible to definitively determine the eukaryote and insect proteomes, estimates are useful for comparative assessments. Our protein sets were derived from a collection of 13,525 protein sequences established for Drosophila melanogaster, which we reduced to 10,018 orthologous groups; proteins with significant similarity were considered as singletons in our processing, since paralogs may have arisen after speciation.

To determine the proteins in the Drosophila orthologous groups common to all insects studied here, called the insect core set, we used predicted proteins from insect genome sequences and EST sequences. We obtained 1346 orthologous groups from the intersection of the whole genomes of five holometabolous insects (see Methods). One aspect of our approximation is to use homologs to Drosophila proteins to characterize proteomes, implicitly assuming that function follows structure. This could contribute to differences in our characterization from the actual proteome, but it does not significantly detract from our use of the characterizations. We discuss further implications of our approximation in more detail below.

Using the insect-core protein set, we removed proteins with significant similarity to any genome sequence in yeast, human, and nematode (see Methods). The remaining 154 orthologous groups (with 360 proteins) form the insect-specific set, and 73 of these groups are represented in the hemimetabolous insect locust ESTs [see Additional file 1]. The insect-specific set contains proteins with homology evidence to all insects studied here; in addition, these sequences are without significant similarity to the non-insect species. Since we are interested in genes and proteins in insects which developed in insects after their divergence from other eukaryotes, we searched entire non-insect eukaryotic genomes in alignments with the insect-core proteins in order to exclude remnants of common ancestral genes. To refine the insect-specific proteins, we removed proteins with similarity to non-insect proteins in the NCBI protein database as described in Methods (Figure 1). This reduced the 360 candidate insect-specific proteins to the final insect-specific set consisting of 51 proteins [see Additional file 2].

We found 466 proteins with homology to all eukaryotes considered in this study using methods similar to those above [see Additional file 3].

As the eukaryotes used in this study are all opisthokont, this set of proteins should be properly considered opisthokont core proteins. Many of these eukaryotic core proteins – the opisthokont core proteins – are involved in housekeeping or general metabolic processes. We also defined 1850 proteins as Drosophila specific by eliminating proteins homologous to other insect proteins as discussed in Methods (Figure 2).

We categorized proteins in the eukaryote (466 groups in opisthokont) and insect-specific sets (154 groups) using high-level gene ontology categories with results shown in Figure 2. In both the eukaryote and insect-specific sets, metabolic proteins constituted the highest fraction, 25% and 20%, respectively. Disproportionately represented categories are interesting to consider for candidate proteins that confer distinguishing characteristics. In the eukaryote/opisthokont set, genes responsible for processes such as cell division, cell motility, cell cycle, reproduction and cellular process are more highly represented by factors from about two to twenty. These proteins and their respective functional categories may distinguish insects less from eukaryotes/opisthokont than those proteins in categories that have a significant representation in the insect-specific set and are underrepresented in the eukaryotic/opisthokont set. These more highly represented categories in the insect-specific set are: larval development (2% in opisthokont, 4% in insect); defense response (0 in opisthokont, 6% in insect); and stress respone (0.2% in opisthokont, 6% in insect). What's more, a significant number of the insect-specific proteins were found to be related to pheromone/odorant binding proteins (OBP), insect cuticle proteins, and proline-rich proteins [see Aditional file 2].

Our analysis of the eukaryote/opisthokont core and insect-specific protein sets was based on functional categories representative of high-level GO designations. Metabolism is the largest category of our eukaryotic/opisthokont core and of the insect-specific proteins. Significantly larger categories for the insect-specific proteins relative to the eukaryote core are stimulus and defense response (Figure 3.). A representative insect-specific gene in the stimulus response category is PedIII/CG11390 which has been reported to function in sensory perception 9. In the eukaryote/opisthokont core proteins, the more highly represented insect-specific categories are not pronounced fractions thereby highlighting the insect-specific proteins as candidates for specialized roles. In the eukaryote/opisthokont core, other housekeeping processes such as cellular division, cell cycle and cellular organization processes constitute a larger fraction of the total protein set. The disproportionate distribution of the eukaryote/opisthokont core and insect-specific sets may be at the very foundation of insect evolution. It is important to note that the disproportionate distributions of functional types of proteins between insects and eukaryotes/opisthokont may be caused to some degree by the methodology; the small number of proteins in the insect-specific core may be caused by the limited number of insect genomes used, artificially underrepresenting the insect proteome. However, assuming an approximately representative distribution of unrepresented proteins makes it unlikely that the overrepresented categories are invalid.

The five insects with whole genomes are all holometabolous and might not be representative of all insects. At present, a complete genome sequence for hemimetabolous has not been sequenced, most likely because hemimetabolous insects often have large genomes (more than 2 gigabases) 10. Fortunately, 45,474 high quality EST sequences from the hemimetabolous insect migratory locust permit us to perform analysis with all insects 11. We determined the insect-specific orthologs in the locust ESTs to arrive at a collection of six sets of insect-spectific proteins. Our analysis found the functional distribution of the orthologous proteins in of the six insects to be similar with the functional distribution of the largest set from the five holometabolous insects [see Additional file 2].

We have noted above, the computed insect-specific protein dataset is an approximatation dependent on available genome sequence. Inclusion of additional genomic data could alter the protein set. The lack of many representative outgroups might causes false positives, i.e. some proteins might be inaccurately included in our list. For example, the gene CG6895 related to immune function is identified as an insect-specific gene in this study, but its homolog was recently reported in the sea urchin 12. Improved quality of genome sequences and gene annotations for the insects used in this study will improve the accuracy of our computed proteins sets 1314.

A considerable number of the 51 insect-specific proteins were found to be related to insect cuticle proteins and pheromone/odorant binding proteins (OBP) [see Additional file 2]. Molting and metamorphosis are crucial processes in the developmental history of the insects involving cuticular proteins. Cuticular proteins are involved in important composite structural materials for insect cuticles, which provide protection, support, and locomotion; these prevent water loss via a wax layer, provide sites for waste product deposition, and protect from ultraviolet radiation 15. Olfaction is essential to insect survival and reproduction, such as in location of food sources and mate selection. These olfactory driven behaviors contribute significantly to the ability of insects to adapt to the environment. The odorant-binding proteins, which compose the insect olfactory system, are involved in the recognition of odorants of plants by insects 1617. The pheromone binding proteins (PBP), abundantly present in the sensillum lymph of pheromone-responsive antennal hairs, are thought to be important in the recognition and discrimination of species-specific pheromones 1819. The olfactory system in insects evolved as a remarkably selective and sensitive system, approaching the theoretical limit for a detector. Even a single pheromone molecule is enough to elicit impulses at the olfactory neuron 2021. The large number of odorant and olfactory proteins in the insect-specific set suggests that in the evolution and diversification of insects, communication and adaptation with the environment played key roles in shaping their morphological and physiological characteristics.

Other insect-specific proteins in our insect-specific set have been found essential to development through experimental procedures 22232425, supporting our insect-specific proteome characterization. Moreover, these have been found to be active in insects and are of interest for evolutionary reasons including their suspected roles in diversification. For example, the gene sinuous (CG10624), which is active in tracheal system development, can partially rescue the tracheal defects of sinuous mutants 22. The Exuperantia (Exu) protein in our insect-specific set is the earliest factor known to be required for the localization of bicoid mRNA to the anterior pole of the Drosophila oocyte. Exu is highly enriched in the sponge bodies; mutation of exu in Drosophila may result in defection of embryonic development 23. Larval serum proteins (Lsp), another type of protein in the insect-specific set, belonging to the hemocyanin superfamily. This family is thought to function as storage proteins that provide amino acids and energy during non-feeding periods of immature and adult development 2425.

It is widely accepted that all insects have arisen from a common ancestor that diverged from an aquatic arthropod more than 390 million years ago, and that they coevolved with a specific plant group 26. Homologs to the insect-specific proteins should be present in the ancestor and be conserved by natural selection. To test this, we analyzed the ratio of the number of nonsynonymous substitutions per nonsynonymous site (Ka) to the number of synonymous substitutions per synonymous site (Ks) for the insect-specific proteins in Drosophila; in this analysis eukaryote/opisthokont core proteins and Drosophila specific proteins were used as controls. The high percentage of insect-specific proteins have a Ka/Ks ratio lower than 0.5 (Figure 4) suggesting negative selection in these proteins 27. As non-synonymous changes are more likely to be deleterious, under negative or purifying selection pressure, these substitutions were eliminated in functionally active proteins, which may have provided a steady protein complement for insects 28. Furthermore, the higher Ka/Ks ratio of insect-specific proteins is on average greater than that of the eukaryote/opisthokont core proteins. This may reflect the later appearance of insect-specific set,relative to proteins in the common eukaryote ancestor.

To determine whether these conserved genes appeared with low redundancy, we ascertained the number of paralogs in the insect-specific genes with the number of paralogs in the eukaryote/opisthokont core genes. Gene duplication is considered one of the principal mechanisms in generating new genes and redundant sequences of genes with the same function 28. Duplicated sequences of established genes often degrade to pseudogenes because purifying selection preserves essential coding sequence, while non-essential duplicates may lose function through random mutations favorable to natural selection. The relationship between duplicates and their functional ancestor is not fully understood. Some authors suggest that the stronger selective constraints on housekeeping genes relative to tissue specific genes is not due to their lower genetic redundancy 29. However, our results agree with the observation of constrained duplication since most of the eukaryote/opisthokont core and insect-specific proteins are without paralogs (Figure 5). This suggests that genes with established function may tend to avoid duplication, thereby tolerating fewer genetic perturbations. However, the insect-specific proteins are inclined to arise from genes producing a greater number of paralogs, which is in contrast to proteins in the eukaryote/opisthokont core. This may confer insect adaptation to changes in the environment. For example, CG16799 and CG6421 have been found to function in defense response; both arise from paralogous groups in Drosophila with ten and four members, respectively.

Our analysis suggests that our working set of insect-specific proteins had been shaped by strong natural selection, with environment as one of the selective influences.

The protein sets in this work were founded on 18,282 protein sequences of Drosophila melanogaster 30 obtained from Ensembl 31. Genes were predicted in genome sequences for Anopheles gambiae (mosquito) 32 and Bombyx mori (silkworm) 3334. Proteins of Tribolium castaneum and Apis mellifera 35 were obtained from HGSC36. Homologs to the insect protein sequences were isolated in annotated genomes of human 37, yeast 38 and nematode 39. We obtained the Anopheles gambiae (mosquito) genome annotated with 16112 proteins (anopheles-21.2b) from Ensembl. The annotated human genome sequence draft (hg17) was obtained from UCSC 40, the worm genome (celegans-21.116a) from Ensembl, and the yeast genome from Saccharomyces Genome Database SGD 41. Proteins where obtained for D. yakuba from FlyBase for use in Ka/Ks analysis. The locust (Locusta migratoria) UniGene collection with 12,161 ESTs and cDNA sequences was obtained from LocustDB 1142.

Sequence alignment was performed with BLAST 43 using the BLOSUM62 scoring matrix and default parameters. Gene prediction was performed using the gene-finder algorithm BGF used in BGI GeneFinder 44 based on GenScan 45 and FgeneSH 46.

We grouped homologous protein sequences into paralogous groups. Protein sequences were considered paralogous if their alignment had an E-value less than or equal to 1e-5 and the alignment covered 70% or more of one of the aligned proteins. We represented paralogous groups by the longest member in the group, with the size of the group determined by the number of unique sequences in it.

We defined protein sets based on Drosophila proteins in our processing pipeline to characterize proteomes. Similarity with genome sequences, predicted proteins, and ESTs was used to cull sets determined in the processing pipeline as described below. Thus, it is important to note that the various protein sets we computationally arrive at characterize insect and eukaryote proteomes through homology.

The insect core set was arrived at by selecting proteins in the Drosophila protein data set with similarity to mosquito and silkworm protein sequences predicted by genome analysis, and with similarity to the locust EST sequence data. Protein sequences for predicted genes in silkworm and mosquito were aligned against fruit fly using blastp 43 and considered homologous with an E-value cutoff of 1e-5 or less; in addition, we required that the length of the aligned sequences be within 70% of each other (Figure 5).

The insect-specific protein set was derived from the insect core set, where proteins without significant alignment to the genome sequences of human, nematode, or yeast were included (E-values of 1e-5 or less). In addition, sequences in the insect core set were retained for the insect-specific set if any alignment covered less than 30% of the insect protein sequence. The insect-specific proteins were further assessed against the NCBI protein database, retaining sequences without significant similarity and less than 30% alignment coverage with all non-insect proteins (Figure 5).

Proteins in the insect core set with an E-value cutoff of 1e-5 or less in alignments with each of the non-insect eukaryotes, and involving 50% or more of the insect protein in the alignments, were included in the eukaryote core protein set.

Functional annotations for proteins in each of the working insect proteomes were determined using the annotation tool Interproscan 47 and Gene Ontology nomenclature 48. GO terms were downloaded from Gene Ontology Consortium.

We selected the most similar orthologs to Drosophila melanogaster in the Drosophila yakuba proteome, YN00 49, to calculate Ka/Ks ratios.