Coleoptera genome and transcriptome sequences reveal numerous differences in neuropeptide signaling between species

Definition of neuropeptide

The definition of neuropeptide is sometimes ambiguous as in principle any peptide from the nervous system could be called a neuropeptide. In this manuscript neuropeptide is defined as a peptide or protein, that is, either released into the hemolymph, directly on a target tissue, or within the nervous system to regulate cellular activity by interaction with a specific cell surface receptor, usually a GPCR. A large number of such neuropeptides has been identified by biological activity on target tissues and/or by directly activating their receptors, while others been identified only by their homology to known neuropeptides. Some neuropeptides have been identified solely on the basis of being produced after proteolytic processing of proteins of unknown function or even only on the basis of the strong likelihood that their putative precursors could be processed by neuroendocrine convertases into neuropeptides. The latter are hypothetical neuropeptides only and are more properly called putative neuropeptides. Indeed, a recent analysis of the precursor of the salivary gland salivation stimulating peptide from Locusta migratoria suggests that it is not a neuropeptide after all (Veenstra, 2017). These putative neuropeptide precursors have been included here, even though no physiological effects have been described for these peptides and their receptors are unknown. On the other hand, I have not included the putative antidiuretic peptide identified from Tenebrio (Eigenheer et al., 2002). The definition given above does not exclude it from analysis, but it is almost certainly derived from a cuticle protein (CAA03880). Although there is a one amino acid difference between the C-terminus of the reported sequence of this cuticle protein (Mathelin et al., 1998) and the antidiuretic peptide that was sequenced, when constructing a Tenebrio transcript with Trinity using the various RNAseq sequence read archives (SRAs) from this species the C-terminus of this proteins was found to be completely identical to the antidiuretic peptide. There are no structure activity data with regard to its antidiuretic activity and it is unclear which protease is responsible for cleaving it from the rest of the protein. This makes it difficult if not impossible to reliably predict which other proteins might be the precursors of similar antidiuretic peptides.

Sequence data

Genome assemblies were downloaded from NCBI (https://www.ncbi.nlm.nih.gov/genome) using the following direct links: https://www.ncbi.nlm.nih.gov/genome/?term=Tribolium; https://www.ncbi.nlm.nih.gov/genome/?term=Pogonus; https://www.ncbi.nlm.nih.gov/genome/?term=Nicrophorus; https://www.ncbi.nlm.nih.gov/genome/?term=Aleochara; https://www.ncbi.nlm.nih.gov/genome/?term=Oryctes; https://www.ncbi.nlm.nih.gov/genome/?term=Coccinella; https://www.ncbi.nlm.nih.gov/genome/?term=Harmonia; https://www.ncbi.nlm.nih.gov/genome/?term=Dendroctonus; https://www.ncbi.nlm.nih.gov/genome/?term=Hypothenemus; https://www.ncbi.nlm.nih.gov/genome/?term=Anoplophora; https://www.ncbi.nlm.nih.gov/genome/?term=Leptinotarsa; https://www.ncbi.nlm.nih.gov/genome/?term=Aethina), with the exception of the genomes of Hycleus cichorii and Hycleus phaleratus, which were obtained from the GigaScience repository (http://gigadb.org/dataset/100405) and those of Photinus pyralis, Aquatica lateralis, and Ignelater luminosus, which were downloaded from http://fireflybase.org/firefly_data.html. When available, protein sequences predicted from the various genomes were also downloaded from NCBI or the two websites mentioned using the same direct links indicated above. Predicted proteins for Hypothenemus hampei were obtained from https://genome.med.nyu.edu/coffee-beetle/cbb.html, and those for Aleochara the Animal Ecology department of the Free University of Amsterdam (http://parasitoids.labs.vu.nl/parasitoids/aleochara/data.php). For Pogonus chalceus the published transcriptome was useful (Van Belleghem et al., 2012). To facilitate reading, species will be identified by their genus name throughout this paper. In the case of the Hycleus species, this will refer to Hycleus phaleratus. There are also two Harmonia genomes, but these are from the same species and they showed no differences in the genes coding neuropeptides. Four other Coleopteran genomes are publicly available, however, they are not yet officially published and for this reason were not fully analyzed here. Those are Sitophilus oryzae, Diabrotica viriginfera, Onthophagus taurus, and Agrilus planipennis.

Pogononus belongs to the Adephaga, all the other species to the Polyphaga suborder. The genera Coccinella, Harmonia, Hypothenemus, Dendroctonus, Anoplophora, Leptinotarsa, Aethina, Hycleus, Tenebrio, and Tribolium all belong to the infraorder Cucujiformia. As will be seen this group shares certain neuropeptidome characteristics that are absent from the other Polyphaga as well as Pogonus.

The quality of these genomes is quite variable. Some have excellent assemblies and in addition numerous RNAseq SRAs making it possible to have high quality assemblies, others are much more limited. For example, the Aleochara assembly has no RNAseq data and only a limited amount of genomic sequences. In the case of Aleochara there is RNAseq data from a different species, Aleochara curtula (SRR921563, from the 1KITE project, Misof et al., 2014), which was helpful and it allowed in some case to reconstruct exons missing from the assembly using a combination of raw genome sequences and trinity. Nevertheless, it is still possible to ascertain the presence or absence of neuropeptide genes from this assembly.

In several instances the predicted complete coding sequences of some neuropeptides are incomplete. When there is little RNAseq data to deduce precursor sequences and a draft genome contains large and small gaps in the assembly such sequences are often incomplete and may well be incorrect in the parts that have been deduced. The Oryctes and Aleochara draft genomes suffer the most from these problems.

A complete list of all SRAs used is available in Table S1.

Presence of neuropeptide and receptor genes

Predicted neuropeptide precursors were preferentially obtained from the annotated genomes, but this was not always possible. On the one hand, small neuropeptide genes are often overlooked by automated annotation programs, even though progress has been quite impressive in that respect, on the other hand there are quite a few transcripts that are probably wrong. Thus many neuropeptide precursors were corrected or predicted de novo from RNAseq data by using the tblastn_vdb command from the SRA Toolkit (https://www.ncbi.nlm.nih.gov/sra/docs/toolkitsoft/) on one or more SRAs using the Tribolium neuropeptide precursors as query to extract reads that could potentially encode a homologous protein. Those reads were then assembled using Trinity (Grabherr et al., 2011) and transcripts that might encode neuropeptides or other proteins of interest were then identified using BLAST+ (ftp://ftp.ncbi.nlm.nih.gov/blast/executables/blast+/). Trinity produced transcripts were judged complete when the N-terminal of the predicted neuropeptide precursors had a signal peptide that could be identified as such by SignalP (Petersen et al., 2011) as implemented on the Web (http://www.cbs.dtu.dk/services/SignalP/) and had an in-frame stop codon at the C-terminus. For GPCRs the identification of the N-terminus is often more ambiguous as some sequences did not have an in-frame stop codon upstream from the putative ATG start codon. In those cases, perceived similarities with homologous GPCRs were used as criterium for completeness. However, for the GPCRs analyzed here the aim was not to obtain absolutely perfect sequences for each receptor, but rather to show whether or not it is present in a particular species.

When in a first round of analysis with the tblastn_vdb command incomplete sequences were obtained, partial transcripts were then used as query for the blastn_vdb command to obtain, where possible, the remainder of the putative transcripts. This process sometimes needed to be repeated multiple times. Transcripts could also be completed by using the assembled genomes, and in several instances no transcripts were obtained at all and only the genome was available. Although many genes were located on single genomic contigs, this was not always the case. In those cases either Trinity produced transcripts and/or individual RNAseq reads, or homology with other precursors from other species were used to confirm the continuity of these transcripts.

It may be noted here that not all trinity produced transcripts are copies of mRNA species of the genes their sequences seem to indicate. In a previous paper on the RYamide gene in Drosophila melanogaster we showed that the very large majority of RNAseq reads that correspond to the coding sequence of this neuropeptide in the various SRAs are not due to the transcription of the RYamide gene, but rather parts of 3′-ends of mRNA produced from genes located upstream (Veenstra & Khammassi, 2017). The RYamide gene is very little expressed in Drosophila melanogaster and mostly in only two neurons in the adult. This causes the RYamide transcript to be so rare that virtually every RNAseq read that contains part of the coding sequence of this gene is in fact an artifact and is not derived from an authentic RYamide mRNA. It is likely that RNAseq reads from other genes that are neither extensively expressed, such as is the case for many neuropeptide genes, may similarly be the product from a heavily expressed upstream gene, rather than from the neuropeptide gene in question. This problem was, for example, encountered with the PTTH gene from Hycleus and the Leptinotarsa periviscerokinin gene. Such RNAseq reads attest to the existence of the neuropeptide gene in question, but have not necessarily undergone the same splicing as the one that is imposed on the neuropeptide mRNA. This phenomenon explains why certain Trinity produced transcripts predict mRNA sequences that contain introns that have not been excised. Whereas in some cases such “false” transcripts can be discarded easily due to the presence of an in-frame stop codon, in other instances such stop codons are absent. Even though obviously such data reveal alternatively splicing, it is not at all clear that this alternatively splicing actually occurs in mRNAs produced from the neuropeptide gene. In other words, what at a first impression may look like very sloppy intron processing, may in fact reflect sloppy stop codon processing in a gene upstream where such sloppiness has no consequences. It is for this reason that I have made no effort to carefully analyze all alternative splice forms for neuropeptide precursors and only recorded those that seem authentic and physiologically relevant.

The presence in RNAseq data of sequences that represent 3′-ends of primary mRNA sequences in which the polyadenylation signal has been ignored may lead to Trinity transcripts that are longer and extend into downstream genes. Thus many Trinity produced transcripts appear at first sight to lack a signal peptide, such transcripts were corrected by removing sequences judged to be extraneous based on sequence homology with other species and in the case of neuropeptide precursors on the presence of a credible signal peptide.

Absence of neuropeptide and receptor genes

The methodology described above allows one to demonstrate the presence of particular neuropeptides. However, when a particular gene is not identified in this way, it does not necessarily mean its absence from the species in question. When a gene is absent from the transcriptome, it may be simply because its expression levels are very low, as, for example, in the case of the previously mentioned RYamide gene from Drosophila melanogaster. If the gene is also absent from the genome assembly, it is possible that it is located in a part of the genome that did not make it into the genome assembly.

Neuropeptides act via receptors, most of which are GPCRs. In many cases, but certainly not all, GPCRs are specific for a particular neuropeptide. So if a neuropeptide gene is genuinely missing from a species, one should expect its receptor to have lost its function and no longer be subject to positive selection. Hence, its receptor is expected to be lost as well. So, when both a neuropeptide and its unique receptor(s) are absent from a genome assembly, it is a good indication that the particular neuropeptide signaling system has been lost from the species in question.

For receptors that may be activated by neuropeptides derived from different genes, this argument can not be used. For example, a Bombyx myosuppressin receptor can be activated by both myosuppressin and FMRFamides (Yamanaka et al., 2005, 2006), hence if either the myosuppressin or FMRFamide were lost, this receptor could still be present. Similar situations likely exists for other neuropeptides, for example, the CCHamides 1 and 2 (Hansen et al., 2011; Ida et al., 2012) or short NPF (sNPF) and NPF (Reale, Chatwin & Evans, 2004). Thus missing neuropeptide receptors can only be used to validate the absence of a neuropeptide ligand, if these receptors are activated exclusively by that ligand.

The loss of a gene can in principle only be demonstrated by flawless genome assemblies (they don’t exist), however, there is an alternative that is almost perfect. It exists in the analysis of the original genomic reads obtained for the assembly. When those reads are very numerous short reads, the chance that there is not a single read that covers the gene in question becomes extremely small and thus negligible. The only remaining problem than is the question, whether or not the gene in question can be reliably identified from a single short read. For the most GPCRs the answer to this question is yes, as the sequences of the seven transmembrane regions are strongly conserved and there are always a couple of them that one can unambiguously identify as being part of a particular receptor. Obviously, this might not work if all the individual transmembrane regions of a GPCR were coded by two exons interrupted by an intron. But this is not the case for the GPCRs analyzed here. An illustration of this method is provided as a in Fig. S1.

For the analysis of the absence of neuroendocrine convertase PC1/3 a similar procedure was applied. This was relatively easy, as this protein has a very well conserved primary sequence.

To demonstrate the absence of a particular neuropeptide gene in this fashion is much more difficult. First, many neuropeptide genes code for a single neuropeptide and the remainder of the precursor is often too poorly conserved to be recognized reliably in short genomic reads from species that are not closely related. Secondly, in some cases the sequence coding the peptide or its most conserved parts, may be interrupted by an intron. For example, the genomic sequences coding the NPF family all have a phase 2 intron in the triplet coding the Arg residue of the C-terminal Arg-Phe-amide, making identification of genomic sequences coding this peptide more difficult. A similar intron is present in the elevenin gene. Thirdly, some neuropeptides are not only small but are also made up of amino acids that have very degenerate codons, this is the case for sNPF. Finally, sometimes conserved amino acids in a particular neuropeptide are no longer conserved, as is the case for allatotropin in honeybees and other Hymenoptera (Veenstra, Rodriguez & Weaver, 2012). On the other hand, when dealing with a larger peptide, that is, structurally well conserved during evolution this would provide an additional argument for its absence.

Sequence comparisons

For comparing the sequences of various neuropeptides I have used Seaview (Gouy, Guindon & Gascuel, 2010) and the figures it produces. The different colors that are used to identify amino acid residues with similar chemical characteristics (acidic, basic, aromatic, aliphatic, etc.) provide good visualization of conserved amino acid sequences when absolute conservation of residues is limited.

General comments

Neuropeptides have previously been identified and sequenced for some of the species analyzed here. The sequences of the two sNPFs and the two AKHs that were identified from the potato beetle (Gäde & Kellner, 1989; Spittaels et al., 1996a) are exactly the same as those predicted from the genome. On the other hand there is no Leptinotarsa gene coding an allatotropin with the same structure as the allatotropin ortholog from Locusta migratoria, in spite of suggestions otherwise (Spittaels et al., 1996b) and this species has a proctolin gene that predicts a classical proctolin and not the previously reported [Ala¹]-proctolin (Spittaels et al., 1995a). Two other putative neuropeptides that were reportedly isolated from this species could neither be identified in any of the genomic or transcriptomic sequences analyzed here (Spittaels et al., 1991, 1995b).

The sequences of Tenebrio AKH, myosuppressin, three pyrokinins, DH37 and DH47 as predicted here from the transcriptome are identical to those reported previously, the only difference being that the transcriptome suggests a C-terminal amide for DH47 instead of the reported C-terminal acid (Gäde & Rosiński, 1990; Furuya et al., 1995; Furuya, Schegg & Schooley, 1998; Weaver & Audsley, 2008; Marciniak et al., 2013). The AKH sequences from Coccinella, Harmonia, and Nicrophorus (Neupert, 2007; Gäde, Šimek & Marco, 2015) are also identical to those deduced from the genome sequences.

The majority of the neuropeptide precursors seem quite similar in structure between the different species. Those will not be commented upon, but their sequences can be found in Table S2 and the Supplementary Figures. To facilitate interpretation of the data several figures include a simplified phylogenetic tree of the species analyzed. This tree is based on the extensive phylogenetic tree recently published for Coleoptera (Zhang et al., 2018). When I use the term closely related species in the text, this is short hand for species that are neighbors on the simplified phylogenetic trees.

Significant changes in peptide sequences

Pigment dispersing factor

The PDF present in Tribolium and other Cucujiformia has two more amino acid residues than the Drosophila peptide and differs from it especially in its C-terminal half (Fig. 1). This explains why it wasn’t identified in a previous study (Li et al., 2008). In the other Polyphaga it is more similar to the Drosophila peptide, but in Polygonus it is two amino acids shorter than in Drosophila.

Neuropeptide F

The structure of NPF has changed even more than that of PDF. It is relatively common for a Phe to residue change into a Tyr and vice versa and so the mutation of the C-terminal Arg-Phe-amide into Arg-Tyr-amide in most species studied here, is not unusual. More drastic is the presence of disulfide bridge in the N-terminal of NPF in the Cucujiformia and the mutation of the C-terminal Arg-Tyr-amide into a Pro-Tyr-amide in the two Curcuclionids. The primary sequence similarity of the predicted peptides to each other and other insect NPFs, as well as the characteristic phase 2 intron in the Arg residue (Pro in the Curcuclionids) of the C-terminal of these peptides confirm that these are true NPF orthologs (Fig. 2). I was unable to find an Oryctes NPF gene, although an NPF receptor seems to be still present in this species. Given the enormous structural variability of this peptide in Coleoptera it is not clear whether this is because the NPF gene was lost, or whether in this species the peptide has undergone even larger sequence changes.

ACP

AKH/corazonin-related peptide (ACP) is a peptide that has been lost independently at least three times and in those species where the gene is still present the predicted peptide sequences are quite variable (Fig. 3).

Calcitonin B

The Leptinotarsa and one of the Anoplophora calcitonin genes encode not only typical calcitonin peptides, but also a number of structurally very similar peptides that lack the disulfide bridge in the N-terminal portion of the molecule (Fig. 4).

Elevenin

Like ACP, elevenin has been lost independently at least three times and in those species where this gene is still present the predicted elevenin sequences are also very variable (Fig. 5).

Gene losses

Unambiguous gene losses

There are a number of instances in which genes could not be found in the assembled genome of a species. In six cases this concerns neuropeptides with a known and unique receptor which is also absent from the same genomes that lack the genes for the ligands. It was previously reported that the Tribolium lacks both ligand and receptor genes for allatostatin A, corazonin and leucokinin (Li et al., 2008; Hauser et al., 2008). The first two were found to be absent from all Coleoptera studied here, while both leucokinin and its receptor were found in Pogonus, the only species outside the Polyphaga suborder for which a genome is available. However, neither leucokinin nor its receptor was found in any of the other species. Leucokinin is also present in other species that do not belong to the Polyphaga suborder. As noted above both ACP and elevenin were lost independently at least three times, while natalisin was lost at least twice in Coleoptera (Fig. 6). Interestingly in Photinus there is still a remainder of the original calcitonin gene. It is clearly defective as it misses essential parts and it is no longer expressed, while the putative receptor (cf Veenstra, 2014) is completely gone. A similar situation occurs with the relaxin gene in Sitophilus oryzae; there also a remainder of non-functional relaxin gene is still present, but its putative receptor (cf Veenstra, 2014) is absent.

Gene duplications

AKH

When the putative Coleopteran AKH precursors are aligned it is evident that they consist of four different regions, the signal peptide, the AKH peptide sequence with its processing site consisting of the GKR triplet, a hydrophilic connecting peptide (C-peptide) and a more hydrophobic disulfide bridge containing sequence (Cys-peptide). It is noticeable that the sequences of the signal peptides, AKHs and the Cys-peptides are very well conserved (Fig. 7), albeit that there are a number of exceptions. The most glaring examples are the second putative Harmonia AKH gene, which obviously can not encode an AKH, and the putative Aethina AKH precursor that is predicted to have no functional signal peptide.

CCHamide 2

In Leptinotarsa the CCHamide 2 gene is duplicated (Fig. S14) and so is the CCHamide 1 receptor. One of the two CCHamide 2 gene codes for a typical CCHamide 2 neuropeptide, the second CCHamide 2 peptide lacks the typical C-terminal amide. Phylogenetic tree analysis of CCHamide receptors shows that the two Leptinotarsa CCHamide 1 receptors are more similar to one another than to the Anoplophora ortholog, thus suggesting that the duplication of this receptor is relatively recent (Fig. 8).

Exon duplications

Allatostatin CCC

In both Aleochara and Nicrophorus, but not in closely related Oryctes, the second and last coding exon of the allatostatin CCC gene has been duplicated allowing the production of two different allatostatin CCC transcripts (Fig. 9) predicted to produce slightly different allatostatin CCC peptides that both conform to the consensus sequence of this peptide (Veenstra, 2016b). In the other species only a single allatostatin CCC precursor was found (Fig. S21).

Allatotropin

The Pogonus allatotropin precursor is almost indistinguishable from the Hemimetabola sequences; it shares with them the N-terminal Gly-Phe-Lys and the remainder of its precursor is also very similar. However, in the other Coleoptera allatotropin sequences these characteristics have not been conserved (Fig. S22). On two occasions, the allatotropin precursor has acquired a second allatotropin paracopy, once by adding a second exon and a second time by adding an additional allatotropin paracopy directly next to the existing one (Fig. 10).

DH31

The DH31 gene shows considerable variation in its structure and the peptides it produces. Although in all species, it codes for the classical DH31 that is very well conserved (Fig. S24), in several species additional neuropeptides are encoded on alternatively spliced mRNAs that do not encode DH31. In its most basic form the gene produces a single transcript from three coding exons containing, respectively, the signal peptide, a conserved peptide that does not look like a neuropeptide, and DH31. In several species, one or two coding exons that code for alternative neuropeptides have been inserted between exons for the conserved peptide and DH31. This leads to alternative splicing in which different neuropeptides are made (Fig. 11). In Hycleus, Tenebrio, and Tribolium at least three different mRNAs are produced enabling precursors sharing the same N-terminal sequence but that have different C-termini encoding an Arg-amide, a short Pro-amide and the typical DH31 peptide, respectively. In contrast to DH31 itself, that has a very well conserved amino acid sequence, these alternative DH31 gene products lack well defined consensus sequences and are neither very similar to DH31 (Fig. S25).

In Pogonus there are two additional exons predicted from the trinity assembly of RNAseq sequences (Fig. 11). In two other Adephaga species, Gyrinus marinus and Carabus granulatus, the transcriptome assembly sequences corresponding to DH31 sequences lack sequences homologous to those two exons (GAUY02019591.1; GACW01024447.1).

DH37–DH47 or CRF-like diuretic hormones

The Tribolium DH37–47 gene has previously been reported to have three coding exons (Li et al., 2008), in which the first of those three is alternatively spliced to the second or the third one. This leads to the production of two CRF-like diuretic hormones, DH37 and DH47 which both been isolated and sequenced from Tenebrio, a species of the same family. Given the sequence similarity of DH47 (Fig. S26) and DH37 (Fig. S27) it seems likely that the last two exons arose by exon duplication. This gene structure seems to be common to the Cucujiformia, but in the other Polyphaga and Pogonus there are only two coding exons in which the last codes for a CRF-like hormone. In Aethina the DH37 coding exon has been duplicated once more, such that there are four coding exons in total and three different mRNAs are produced, each encoding different CRF-like peptides (Fig. 12).

Periviscerokinin (Capa)

The periviscerokinin genes are quite variable in Coleoptera. They can consist of several coding exons that all use phase 1 introns. This allows alternative splicing to produce a variety of different precursors from these genes. Although in some species RNAseq data confirm such alternative splicing, in many cases the number of total RNAseq reads for these genes is far too small to demonstrate alternative splicing. An important site of periviscerokinin synthesis is in the abdominal ganglia, from which RNAseq reads are generally only obtained when whole insects are used for RNA extraction. Thus while in many species only a single transcript can be documented, alternative splicing may well be common.

The number of coding exons for this gene in the species studied varies between four and seven (Fig. 13). The first coding exon contains the sequence for the signal peptide, the last for a hydrophilic C-terminal sequence of the precursor that is usually rich in acidic amino acid residues. The penultimate coding exon tends to be the largest and codes for subsequently a periviscerokinin, a tryptopyrokinin and a hydrophobic sequence. The variable number, one to four exons between the first and pentultimate coding exon, contain sequences for a pyrokinin-like peptide, although in Hypothenemus, the third one has lost this sequence.

sNPF

The sNPF precursor is very well conserved in Coleoptera (Fig. S9). In Anoplophora and Leptinotarsa, but not in closely related Dendroctonus and Hypothenemus or any of the other species studied here, partial duplication of the exon coding sNPF led to a gene having an additional coding exon. In Anoplophora the RNAseq data suggest the production of two alternatively spliced transcripts that code for either one or two sNPF paracopies. Although there is much more RNAseq data from Leptinotarsa, in this species there is only evidence for a single mRNA encoding two sNFP paracopies (Fig. 14).

Paracopy numbers

Several insect neuropeptide precursors contain multiple copies of identical or very similar peptides. These typically include allatostatins A and B, calcitonin B, leucokinin, FMRFamide, pyrokinin, periviscerokinin, ETH, orcokinin A and B, RYamide, sulfakinin, and tachykinins. The number of such paracopies can vary between and even within species (Veenstra, 2010). The genes coding calcitonin B, leucokinin, pyrokinin, and periviscerokinin have already been discussed above. Allatostatin A has so far never been found in Coleoptera. ETH has usually two paracopies, but in the three species from the Elateroidea—Ignelater, Photinus, and Aquatica—the first copy has been lost (Fig. S28) and this is also the case in Aleochara and Hypothenemus. In all five of these species, the genome still contains coding sequences for both splice variants of the ETH receptor. The RYamide gene codes for two RYamide peptides in all Coleoptera (Fig. S29), except Anoplophora that lost this gene and its receptor, while the sulfakinin gene codes also for two paracopies in all species studied here (Fig. S30). The number of FMRFamide paracopies varies from four to six (Fig. S31), and from 5 to 9 for the NPLP1 precursor (Fig. S32) and for allatostatin B (Fig. S33) the numbers are from seven for the Curcuclionids Dendroctonus and Hypothenemus to eight for the other species.

Genes that seem very stable

It is fair to state that the number of changes in neuropeptide genes in Coleoptera is significant. This might obscure the fact that many other genes seem, as least as far as their sequences are concerned, remarkably stable, such is the case for CCAP (Fig. S35), SIFamide (Fig. S36), Sulfakinin (Fig. S30), GPA2 (Fig. S37), GPB5 (Fig. S38), FMRFamide (Fig. S31), hansolin (Fig. S39; Liessem et al., 2018), CNMamide (Fig. S40), ITG (Fig. S41), and PTTH (Fig. S42). The mRNA from the gene coding ion transport peptide (ITP) is generally alternatively spliced in two forms, ITP-A (Fig. S43) and ITP-B (Fig. S44). It has been reported that in Tribolium there is a third splice product (Begum et al., 2009), but such a form could not be detected for any of the other species studied here, including Hycleus or Tenebrio, two species closely related to Tenebrio.

Significant peptide sequence changes

In those Coleoptera where elevenin and ACP are (still) present neither the sequences of the peptides nor of those of their precursors are well conserved. Other neuropeptides have not only maintained the sequences of the neuropeptides themselves, but often those of the entire precursors, suggesting that those parts of the precursor that do not code for the biologically active peptides must have other important functions. It has previously been noted that as expression of the RYamide gene in Drosophila species decreases, the structure of both the neuropeptides themselves and their precursors are no longer well conserved (Veenstra & Khammassi, 2017). This could mean that the function of the peptide is becoming obsolete, which would facilitate its subsequent loss; use it or loose it. However, it is also possible that it is no longer needed in the large quantities that are necessary for discharge into the hemolymph. This might well be the case for Drosophila RYamide where the rectal papillae in Drosophila melanogaster are innervated by RYamide neurons. While the same neurons are present in Drosophila virilis, in that species—and many others—RYamide is also released from enteroendocrine cells, presumably likewise to stimulate the rectal papillae. The amount of RYamide that needs to be released from the midgut to reach sufficiently high hemolymph concentrations will be much larger than that made by the neurons that directly innervate the rectal papillae. This likely not only puts selection pressure on the peptide sequences but also on an efficient processing of their precursors. It is the latter that may explain why some insect neuropeptide precursors seem to be so well conserved. In Rhodnius and Tribolium ACP appears to be expressed exclusively in neurons (Hansen et al., 2010; Patel et al., 2013). This suggests that its large structural variability indicates a loss of physiological relevance in Coleoptera which may explain its loss from the genome on at least three occasions in this insect order. The same could also be true also for elevenin and it is tempting to speculate that when neuropeptide structures are no longer well conserved it either indicates the loss of their physiological relevance as a hormone or as both a hormone and a neuromodulator. For small neuropeptides that act on GPCRs replacement of a single critical amino acid may be enough to impair receptor binding. Insulin is larger and needs to bind simultaneously to two tyrosine kinases it may therefore be less sensitive to single amino residue replacements as long as the three-dimensional structure of the molecule is maintained and interaction with its receptor is still possible.

It is interesting to see that some changes in Coleoptera neuropeptide precursors are similar to the those observed in other Holometabola. The allatotropin genes in Hemimetabola code for a single allatotropin paracopy, but in Lepidoptera the gene encodes various allatotropin-like peptides produced on alternatively spliced mRNAs (Taylor, Bhatt & Horodyski, 1996; Nagata et al., 2012). Whereas, the Pogonus allatotropin gene is quite similar to those of the Hemimetabola, in the Polyphaga suborder allatotropin genes coding for two paracopies emerged on two occasions. The sNPF gene in Hemimetabola is also very simple, but in Lepidoptera and Diptera, the gene codes for several paracopies. Again this evolved independently in Anoplophora and Leptinotarsa. If proctolin is indeed absent from Oryctes and the Coccinellids, this would be similar to what occurred in Hymenoptera and Lepidoptera. Thus in at least some cases neuropeptide evolution in different holometabolous insect orders seems to follow what look to be similar pathways, that is, increasing paracopies in neuropeptide genes that look like they never changed from the ancestral arthropod to cockroaches, decapods and chelicerates, or independently eliminating others, such as elevenin and ACP. This raises the question whether somehow complete metamorphosis is responsible for these changes.

Several neuropeptides contain a cysteine bridge structure constraining the structure of the peptide. This could have important effects on receptor binding and/or provide it protection against proteases degradation. It is surprising to see NPF gain a cysteine bridge in the Cucujiformia and some, but not all, calcitonins in Leptinotarsa and Anoplophora loose theirs. It would be very interesting to see the interactions of these peptides with their receptors in order to know what effects, if any, these structures have on receptor activation.

Gene losses

It appears that loss of a neuropeptide systems is not a very rare event and some neuropeptides are more easily eliminated than others. Thus, in Coleoptera some neuropeptides got lost repeatedly, that is elevenin and ACP each at least three times, natalisin twice and the dilp8 ortholog likely four times. Indeed, elevenin, ACP, calcitonin, corazonin, natalisin, dilp8, and relaxin were also found missing in other insect species (Huybrechts et al., 2010; Veenstra, 2014, 2019).

Although the loss of a neuropeptide signaling system may well have its origin in the degeneration of the peptide gene, for the reasons of gene sizes, it is as likely to start in the receptor gene. Not only are the receptor coding regions generally much larger than those of neuropeptides, but more often than not the total size of these genes is enormous. Thus the accidental elimination of a large piece of DNA after chromosome breakage and rearrangement may well be limited to sequences coding a receptor without altering any adjacent genes, while the elimination of a piece of the same size that touches a neuropeptide gene is more likely to affect also neighboring genes, thus increasing the likelihood of selection against such an event. Indeed in at least some instances, one can still find remnants of the ligand gene, while the receptor has vanished. Apart from calcitonin B in Photinus described here, the same phenomenon is observed with relaxin in Sitophilus and sulfakinin in the tsetse fly Glossina morsitans. In the latter species a highly degraded sulfakinin pseudogene is still recognizable, while both sulfakinin receptor genes have been lost (BLAST Data).

As discussed above if the sequence of a neuropeptide is no longer maintained it may indicate the loss of physiological relevance. This might be a useful criterium to look at duplicated neuropeptide genes as well. When the putative Coleopteran AKH precursors are compared it is evident that not all these precursors are well conserved. If the bulk of AKH precursor sequences is so well conserved, why are the others not? We have a good idea about what signal peptides and authentic AKHs should look like, and we can thus discard the truly aberrant genes from Aethina and Harmonia as obviously no longer functional AKH genes (Fig. 7). However, we do not know what the requirements are for a good Cys-peptide, as its function is unknown. Similar Cys-peptides have been found in other insect neuropeptide precursors, such as those of SIFamide and RYamide (Verleyen, Huybrechts & Schoofs, 2009; Veenstra & Khammassi, 2017). The conservation of the structure of such peptides implies that they are important—perhaps for assuring correct intracellular transport to the secretory granules of the neuropeptide precursors—and thus that those precursors that no longer have such a conserved Cys-peptide may be functionally impaired. Predicted AKH precursors that look like they might be defective are only found in species that also have an AKH gene with a well conserved AKH precursor. This suggests that AKH precursors that are not well conserved may have largely lost their functional significance and/or may be evolving into pseudogenes. It is interesting in this context that of the two Tribolium AKHs only the one that has the best conserved precursor sequence could be detected by mass spectrometry (Li et al., 2008). Similar arguments suggests that the copy of the Oryctes Bursicon A gene may well be on its way to become a pseudogene.

The predicted signal peptides of the proctolin precursors from Oryctes, Harmonia, and Coccinella seem to be perfectly normal as do other parts of the precursor that have been conserved since the last common ancestor of chelicerates and mandibulates. However, the predicted proctolin molecules have been mutated and these species all seem to have lost their proctolin receptors. It is a very puzzling and unresolved matter; as if proctolin is not the only biologically active peptide produced from the proctolin precursor or as if there is yet another proctolin receptor that remains to be identified.

Gene duplications

Gene duplications are a common phenomenon during evolution and most of these duplicated genes are subsequently lost (Lynch, 2007). Genes coding insulin-related peptides and adipokinetic hormones are repeatedly amplified in insects and in Coleoptera this also includes pyrokinin genes. Why is it that some neuropeptide genes regularly show increased numbers while others do so only rarely? When there are paralogous genes in a genome, this is the result of two independent processes, first duplication of the original gene and then maintaining both the original and its copy.

Just like elimination of receptor gene is likely facilitated by its large size, the initial duplication of a neuropeptide gene is probably more easily accomplished due to the smaller the sizes of the gene. It is striking in this context that both the insect AKH and insulin genes—two that are commonly amplified in insects—are generally very compact genes, have small introns and plausibly small regulatory regions (of that we only have some information from Drosophila, which is not necessarily a model for all insect species). The Coleoptera pyrokinin genes similarly look very compact, as the sizes of the introns between the coding regions are small. This is also true for the strongly amplified vasopressin genes in Locusta (Veenstra, 2014).

Such small sizes may not only favor the original duplication, but also make it much more difficult to eliminate the gene by gross chromosome reorganizations and this may explain the presence of amplified genes in a genome that are not as well conserved as others. However, in order to permanently maintain two paralog genes, there also needs to be an advantage to maintaining both copies. This is often achieved through neo- or subspecialization (Lynch, 2007). In Drosophila the different insulin genes do not all have the same temporal and spatial expression (Brogiolo et al., 2001; Liu et al., 2016), suggesting that subspecialization may be at least part of the reason these genes are maintained. I have previously suggested that in some cases it may be the need for massive amounts of neuropeptides that facilitates the maintenance of paralog neuropeptide genes (Veenstra, 2014). In the case of the Coleoptera pyrokinin genes this may well apply also, but there is maybe something else at play as well.

Pyrokinins, tryptopyrokinins, and periviscerokinins are structurally similar arthropod peptides that each act on specific receptors (Iversen et al., 2002; Rosenkilde et al., 2003; Cazzamali et al., 2005; Homma et al., 2006; Paluzzi et al., 2010; Paluzzi & O’Donnell, 2012; Jiang et al., 2014). The tryptopyrokinins are only present in insects, where they seem to play an important physiological role and they are absent from basal arthropods. In most insect species they are coded by two different genes, the pyrokinin and periviscerokinin genes. The pyrokinin genes codes for pyrokinins, often also for a tryptopyrokinin and rarely even a periviscerokinin. The periviscerokinin gene codes for periviscerokinins, often a tryptopyrokinin and rarely a pyrokinin. Whatever their exact roles or physiological functions, there appears to be a physiological need to be able to produce these three types of peptides, pyrokinins, tryptopyrokinins, and periviscerokinins independently from one another. In termites, crickets, stick insects, locusts, and cockroaches separate tryptopyrokinin genes have evolved that code for a tryptopyrokinin precursors containing multiple tryptopyrokinins. However, this has not happened in holometabolous insects.

The tryptopyrokinins are produced predominantly, if not exclusively, by neuroendocrine cells in the labial neuromere of the suboesophageal ganglion from either a pyrokinin or periviscerokinin precursor by mechanisms that are not understood. Although receptor ligand interactions in Tribolium suggests that at least in that species there may be no tryptopyrokinin specific receptor (Jiang et al., 2014), in the closely related Zoophobas atratus the periviscerokinin precursor is still differentially processed to produce predominantly a tryptopyrokinin and pyrokinin in the suboesophageal ganglion and periviscerokinins in the abdominal ganglia (Neupert et al., 2018). Interestingly, in Tribolium the tryptopyrokinin from the pyrokinin precursor is less active on the pyrokinin receptors than the one from the periviscerokinin precursor (Jiang et al., 2014) and so it may be no coincidence that the tryptopyrokinin from was lost from a number of Coleoptera pyrokinin genes.

Whereas the pyrokinin genes are often amplified in Coleoptera, the periviscerokinin genes are often sometimes partially amplified, that is, a periviscerokinin coding exon is duplicated. Adding an extra exon does not change the reading frame, since the introns defining the duplicated exons are of the same phase. This makes alternative splicing relatively easy. Duplication of this exon may be further facilitated by the presence of much larger introns than those that are found in the pyrokinin gene.

The size of receptor genes should make their segmental duplication a relatively rare event. It is thus interesting that in Leptinotarsa the CCHamide 1 receptor is duplicated and that both copies seem to be well expressed. Surprising and unexpected is that in the same species the CCHamide 2 neuropeptide gene is also duplicated. Although this does not constitute final proof for the evolution of a novel insect neuropeptide system in Leptinotarsa, it certainly is as close as one can get from sequence data alone. Both receptor and peptide have evolved significantly since their respective duplications (Fig. 8) and this very much suggests that Leptinotarsa has three separate CCHamide neuropeptide systems.