Building a platform for predicting functions of serine protease-related proteins in Drosophila melanogaster and other insects

Abstract

Serine proteases (SPs) and serine protease homologs (SPHs) play essential roles in insect physiological processes including digestion, defense and development. Studies of insect genomes, transcriptomes and proteomes have generated a vast amount of information on these proteins, dwarfing the biological data acquired from a few model species. The large number and high diversity of homologous sequences makes it a challenge to use the limited functional information for making predictions across a broad taxonomic group of insects. In this work, we have extensively updated the framework of knowledge on the SP-related proteins in Drosophila melanogaster by identifying 52 new SPs/SPHs, classifying the 257 proteins into four groups (CLIP, gut, single- and multi-domain SPs/SPHs), and detecting inherent connections among phylogenetic relationships, genomic locations and expression profiles for 99 of the genes. Information on the existence of specific proteins in eggs, larvae, pupae and adults is presented to facilitate future research. More importantly, we have developed an approach to reveal close homologous or orthologous relationships among SPs/SPHs from D. melanogaster, Anopheles gambiae, Apis mellifera, Manduca sexta, and Tribolium castaneum thus inspiring functional studies in these and other holometabolous insects. This approach is useful for tackling similar problems on large and diverse protein families in other groups of organisms.

Graphical Abstract

1. Introduction

Serine proteases (SPs) of the S1A subfamily play important roles, ranging from digestion of dietary proteins to blood coagulation, complement activation, fertilization, and tumor metastasis in mammals (Foley and Conway, 2016; Wojtukiewicz et al., 2016). In arthropods, some SPs also form cascade pathways to establish dorsoventral polarity of embryos, mediate immune responses in various life stages, and cause hemolymph coagulation in horseshoe crabs (Kanost and Jiang, 2015; Kawabata and Muta, 2010). SPs digest foods and process protoxins in the midgut of most insects, including agricultural pests and human disease vectors (Rukmini et al., 2000; Zhu-Salzman and Zeng, 2015). Due to their crucial roles in various physiological processes, SPs and their non-catalytic homologs (SPHs) have been actively investigated in several species that are research models and/or agricultural pests (e.g. Drosophila melanogaster, Tribolium castaneum, Manduca sexta, Helicoverpa armigera, Tenebrio molitor), and insects of medical relevance (e.g. Anopheles gambiae, Aedes aegypti). SPHs are enzymatically inactive due to the lack of catalytic residue(s) but some of them (e.g. Masquerade) have known regulatory functions, including somatic muscle attachment, axonal guidance, post-mating responses, and prophenoloxidase (proPO) activation by proPO activating proteases (PAPs) (Kwon et al., 2000; Murugasu-Oei et al., 1996; Yu et al., 2003). POs catalyze the formation of reactive compounds and melanin that kill and sequester pathogens and parasites (Zhao et al., 2011). Upon infection, SPs also generate active cytokines such as Spatzle and stress responsive peptides that bind receptors (e.g. Tolls) to trigger intracellular signal transduction (Cao et al., 2015a; Schrag et al., 2017). Many members of the SP cascades in insects contain a disulfide-stabilized structure named clip domain and these SPs and SPHs are abbreviated as CLIPs and classified in subgroups A–E (Kanost and Jiang, 2015; Veillard et al., 2016).

In the past two decades, technological advances have greatly facilitated the acquisition of massive amounts of sequence data from living organisms, which led to the discovery of many genes homologous to those with known functions. For instance, there are 176 SP-related genes in the human genome (Puente et al., 2003) and, despite all the efforts to understand their physiological roles, only a small portion of them have been characterized at the functional level. Similarly, the holometabolous insects D. melanogaster (Ross et al., 2003), A. gambiae (Cao et al., 2017; Christophides et al., 2004), Apis mellifera (Zou et al., 2006), T. castaneum (Zou et al., 2007), and M. sexta (Cao et al., 2015b) possess 204, 337, 58, 168, and 193 SP-related genes in their genomes, respectively. Investigators routinely construct neighbor-joining trees to study evolutionary relationships among all members of the protein family in a single insect, since it takes a short time for the computation. Analyses of sequences from multiple species based on neighbor joining are not that trustworthy even if comparisons are limited to a subgroup of these proteins, such as CLIPs (Waterhouse et al., 2007). Thus, systematic classification and comparison of SP-like proteins across different orders of insects, which demand hundreds of highly reliable sequences and serial phylogenetic analyses, have not yet been reported. Another difficulty in such phylogenetic studies stems from the deeply rooted evolution of SP and SPH genes. Branching near roots of the trees is particularly problematic, as often indicated by low bootstrap values. The large numbers of SP-like genes and their high degrees of sequence variations in insects suggest that multiple series of gene duplication and sequence divergence occurred at different rates in some lineages of insects or subgroups of SP-related proteins. These complications interfere with the correct assignment of orthologous relationships among homologs in insects, the largest group of animals. Orthology is crucial for reliable extension of functional information from model species (e.g. D. melanogaster, M. sexta) to other insects of practical importance. This problem also exists for discovering such relationships in other large, diverse protein families (e.g. zinc finger proteins, Leu-rich repeat proteins, serine esterases) from various groups of living organisms.

To address these issues, we selected five species of holometabolous insects with broad biological significance, extensively updated the information on SP-related proteins in three of the five, and classified the resulting protein sequences into three to four large groups on the basis of domain structures, expression patterns and initial phylogenetic analyses. After that, the protein sequences from the five species were subjected in individual groups to sequential Bayesian analyses to reveal phylogenetic relationships. To better understand the functional data from Drosophila, we uncovered close correlations among chromosomal locations, phylogenetic tree positions, and expression profiles of the SPs and SPHs. Their mRNA levels and protein abundance in different tissues or life stages were extracted from published datasets to further assist the selection of genes for targeted reverse genetic analyses in the future. Functional data from M. sexta and the other species were integrated through hypothetical orthology to identify candidate genes for research in not only Drosophila but also mosquitoes (e.g. A. gambiae, A. aegypti), bees (e.g. A. mellifera), wasps (e.g. Nasonia vitripennis), moths (e.g. H. armigera), beetles (e.g. T. castaneum), and other insects. In summary, these explorations represent our newest efforts to organize and utilize a sizable body of information to expedite the functional elucidation of SP-related proteins based on their evolutionary relationships.

2. Materials and methods

2.1. Identification and annotation improvement of the SP-related proteins in D. melanogaster, T. castaneum, and A. mellifera

OPS Dmel-r6.09 was downloaded from FlyBase (https://flybase.org/). Domains in the Dmel-r6.09 sequences were identified on a local supercomputer using InterProScan5 v5.17 (Jones et al., 2014). Proteins containing a chymotrypsin-like (i.e. S1A subfamily) SP-related domain were extracted and combined. 376 RNA-seq files from the modENCODE project (Brown et al., 2014), 1.7 billion reads representing 30 different cDNA samples of whole insects at different life stages (Table S4), were downloaded from NCBI Sequence Read Archive (SRA). The reads were trimmed by Trimmomatic (Bolger et al., 2014) with the setting “SLIDINGWINDOW:4:20 LEADING:20 TRAILING:20 MINLEN:50” to remove low-quality bases and assembled by Trinity-2.3.2 (Haas et al., 2013), with the k-mer length of 25. SP-like sequences in Dmel-r6.09 were crosschecked and improved with the Trinity models. After removal of alternatively spliced and severely incomplete genes, the sequences were manually examined according to characteristic features of the S1A SPs, such as signal peptide and conserved regions. The SPs were compared with the newer OPS, Dmel-r6.14, to ensure all SPs were included. Similarly, T. castaneum and A. mellifera protein sequences were downloaded from NCBI RefSeq using accession numbers GCF_000002335 (Tcas5.2) and GCF_000002195 (Amel4.5). Domain(s) in the beetle and bee SP-like proteins were found using InterProScan5. To improve the SP/SPH gene models in Tcas5.2 and Amel4.5, all the RNA-seq data deposited before 6-1-2017 (6 projects of T. castaneum: PRJNA247821, PRJNA376868, PRJNA315762, PRJNA305354, PRJNA275195 and PRJNA266839; 5 projects of A. mellifera: PRJNA268450, PRJNA386067, PRJNA325930, PRJNA322249 and PRJNA275154) (Dippel et al., 2014; Greenwood et al., 2017; Kim et al., 2018; Manfredini et al., 2015; Mao et al., 2017; Stappert et al., 2016) were downloaded from NCBI SRA for read trimming and assembling with Trinity. The final outputs contained limited numbers of transcripts, possibly due to SNPs in the cDNA samples. As such, the intermediate files (data not shown) generated by Inchworm, a part of Trinity, were compared with the corresponding official gene models prior to manual improvement of the protein sequences. BLASTP searches against NR database of NCBI were also performed to improve the sequences, as some newly verified proteins were not updated in the RefSeq models.

2.2. Properties of the SPs and SPHs in D. melanogaster, T. castaneum, and A. mellifera

The SP-related sequences in each species were divided into SPs or SPHs based on the presence or absence of the His-Asp-Ser catalytic triad as described before (Cao et al., 2015b). Signal peptide was predicted using SignalP 4.1 (http://www.cbs.dtu.dk/services/SignalP/) (Petersen et al., 2011) and Signal-3L (http://www.csbio.sjtu.edu.cn/bioinf/Signal-3L/) (Shen and Chou, 2007). Some clip domains (as well as other structural modules) were identified by InterProScan5 while other clip domains were found through manual inspection of the sequences for a Cys doublet in the region close to the protease or protease-like domain (PD or PLD). SPs and SPHs with four additional Cys residues upstream of the doublet were designated CLIPs to indicate the presence of a clip domain (Kanost and Jiang, 2015). Residues 190, 216 and 226 (chymotrypsin numbering) (Perona and Craik, 1995) that form the primary substrate-binding pocket of each PD were identified in the aligned sequences for protease specificity prediction (Cao et al., 2015b).

2.3. Classification, multiple sequence alignment, and phylogenetic analyses of the SP-like proteins in D. melanogaster, A. mellifera, and T. castaneum

The SP-related proteins in the three species were first classified into C (for CLIP), M (for multi-domain), and G-S (for gut and other single domain) groups based on the presence/absence of clip domain(s) and then two or more non-clip domains. While the G-S groups remained intact in A. mellifera and T. castaneum, single domain SPs/SPHs in D. melanogaster were further separated into G and S groups based on their expression profiles and positions in the phylogenetic tree of G-S. Members of the G group tend to associate with feeding, high-level expression, a typical size of about 270 residues, some gene clusters on chromosomes, and certain positions in the phylogenetic tree (Cao et al., 2017; Cao et al., 2015b). Multiple sequence alignments of the entire proteins in the C, G and S-M groups from D. melanogaster were performed using MUSCLE (Edgar, 2004), a module of MEGA 7.0 (Kumar et al., 2016), under the default settings with maximum iterations changed to 1,000. Alignments of the individual groups were converted to NEXUS format by MEGA, and phylogenetic analyses were conducted using MrBayes v3.2.6 (Ronquist et al., 2012) under the default model with the setting “nchains=12”. MCMC (Markov chain Monte Carlo) analyses were terminated after the standard deviations of two independent analyses was <0.01 or after 10 million generations (SD < 0.05). FigTree 1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/) was used to display the trees. To identify orthologous or close homologous relationships among CLIPs, S-group SP(H)s, Gd-domain SPs/SPH, and CUB-domain SPs from the five species, the sequences in each group were analyzed using MUSCLE and MrBayes. Based on the initial analysis of the 247 CLIPs, sequences of CLIPAs, CLIPBs, CLIPCs, and CLIPDs from the five species were separately examined by the Bayesian method.

2.4. Chromosomal locations and expression profiling of the SP and SPH genes in D. melanogaster

For most of the SP-related genes, their genomic locations were available in the information lists of the Dmel-r6.09 models. Retrieved position data were plotted using ArkMAP 2.0 (http://www.bioinformatics.roslin.ed.ac.uk/arkmap/) and improved using Adobe Illustrator. After excluding data from cell lines or challenged insects (Brown et al., 2014), the remaining 30 and 26 additional datasets (Table S4) were downloaded from NCBI SRA, representing whole bodies or tissues of healthy D. melanogaster under 52 conditions. Reads were trimmed with Trimmomatic to remove adaptors and low quality bases with the setting “SLIDINGWINDOW:4:30 LEADING:20 TRAILING:20 MINLEN:50”. Transcript sequences of the SP-related proteins in Dmel-r6.09 were replaced with the improved ones (Section 2.1). FPKM (fragments per kilobase of transcript per million mapped reads) values for genes in different libraries were calculated using Bowtie2 2.2.3 (Langmead and Salzberg, 2012) and RSEM 1.2.15 (Li and Dewey, 2011). FPKM values in libraries from biological replicates were averaged to represent gene expression in those types of samples. Hierarchically clustered gradient heat maps of log₂(FPKM+1) values were plotted using the clustermap function of Seaborn (https://seaborn.pydata.org/), a Python data visualization library, with the average linkage method and Euclidean matrix.

2.5. Abundances of the SP-related proteins in D. melanogaster eggs, larvae, pupae, and adults

Label-free quantification (LFQ) values of the SPs/SPHs in whole insects at 14 life stages, each with four biological replicates (Table S5), were extracted from the published data (Casas-Vila et al., 2017). Proteins identified in at least 3 out of the 68 samples were kept. After normalization with 5×10⁶, which is close to 5,103,000 or the lowest non-zero LFQ, relative protein abundances, as represented by log₂(LFQ/5×10⁶ + 1) values, were subjected to hierarchical cluster analysis and plotted as a gradient heat map.

3. Results and discussion

3.1. Current status of the research area and outlines of our approach

Genome-wide investigations of insect SPs and SPHs have facilitated functional studies of this family of proteins after the genome sequence of D. melanogaster was first published in 2000. We described and analyzed sequences of about 200 SP-related proteins in the fruit fly (Ross et al., 2003), and these were further annotated by functional residue clustering (Shah et al., 2008). Similar studies yielded overviews of the S1A protease subfamily in the yellow fever and African malaria mosquitoes (Cao et al., 2017; Waterhouse et al., 2007), honeybee (Zou et al., 2006), red flour beetle (Zou et al., 2007), silkworm (Zhao et al., 2010), diamond back moth (Lin et al., 2015), and tobacco hornworm (Cao et al., 2015b). However, due to the large family sizes, most of the investigations focused on CLIPs (clip-domain SPs and SPHs) but it has been quite difficult to define orthologous relationships, for example, among the A. gambiae and D. melanogaster CLIPBs. Incorrect gene models and incomplete genome coverage also compromised the quality of those studies, leading to weak support for functional predictions. When the role of an SP is known in species A but no clear ortholog stands out in species B, the amount of experimental work required to validate the assumed function increases proportionally with the number of candidates. To deal with these problems, we first improved the gene models in D. melanogaster, T. castaneum and A. mellifera, separated the family into 3–4 groups with similarities in sequence and domain structure, performed serial phylogenetic analyses at the group and subgroup levels along with the homologs in A. gambiae and M. sexta, and predicted functions mainly based on the literature and closest possible homolog.

3.2. Overview of the SP-related protein family in the five insect species with known genomes

To provide a summary of the family, we have selected five species from four different orders of holometabolous insects: D. melanogaster (Diptera) for its wide-ranging genetic data (Veillard et al., 2016), A. gambiae (Diptera) for its medical implications and available results on CLIPs (Cao et al., 2017), A. mellifera (Hymenoptera) for its agricultural importance and a small set of the core SPs and SPHs (Zou et al., 2006), T. castaneum (Coleoptera) for its unique phylogenetic position, amenability to RNAi assays, and representation of the largest order of animals (Zou et al., 2007), and M. sexta (Lepidoptera) for its extensive biochemical data on SP-related proteins and closeness to agricultural pests (Kanost and Jiang, 2015). Since the M. sexta and A. gambiae SPs and SPHs were recently examined (Cao et al., 2017; Cao et al., 2015b), we only updated the information on SP-like proteins in the fly, bee, and beetle.

3.2.1. The fruit fly D. melanogaster

In contrast to AgamP4.5, in which 117 of the 337 gene models were improved using the MCOT method (Cao and Jiang, 2015), the same approach only resulted in six corrections in the Drosophila sequences, reflecting the high quality of the genome assembly and gene annotation. Compared with results of the initial analysis (Ross et al., 2003) of D. melanogaster SP/SPH genes, major improvements are made in four areas (Table S1): 1) we identified 52 new SP or SPH genes in the genome, making 257 the total count of SP-like genes, 2) all CLIPs but four have a signal peptide, an entire clip domain, and a full-length protease or protease-like domain (PD/PLD). cSP44 and cSP56 have a predicted transmembrane region rather than a signal peptide, whereas cSP54 and cSP229 lack both, 3) we compiled a complete set of information including the group-chromosome-tree-expression (G-C-T-E) characterization for easily locating the genes on the chromosomes, phylogenetic trees, and expression profiles, and 4) we separated SPs and SPHs into four groups related to their domain structure and putative functions: C (for CLIP), M (for other multi-domain), G (for gut) and S (for other single domain) groups of SP-related proteins. Among the 190 SP and 67 SPH genes (Table 1) in the fly genome, 52 encode CLIPs (34 SPs and 18 SPHs with at least one clip domain), 21 are in the “M” group of non-CLIP, multi-domain SPs (17) and SPHs (4), 65 gut and 119 other single-domain SPs/SPHs.

Table 1.

Names and key features of the 257 serine protease-like proteins in D. melanogaster

name	G-C-T-E	activation cleavage site	specificity	domain	name	G-C-T-E	activation cleavage site	specificity	domain	name	G-C-T-E	activation cleavage site	specificity	domain
cSP1/Grass	CuAD	LSQR*VSNG	T(DGG)	cP	SP126a	GWJH	QSGR*IKGG	C(GGD)	P	SPH169a	SYWI	DEDQ*MLNM	na	H
cSP3/Ser7	CyAb	FSFR*LVGG	T(DGG)	cP	SP126b	GWJH	KDQR*IIGG	C(GGD)	P	SPH169b	SYWI	LDEK*LLYT	na	H
cSP4/SPE	CZAA	FADR*IFGG	T(DGG)	cP	SP129/ιTry	GIIH	ATGR*IIGG	T(DGG)	P	SP170	SRTI	PKDI*IVNG	C(SGS)	P
cSP5	CwAC	VTNR*IYGG	T(DGG)	cP	SP131	GWJH	FQNR*VING	C(GGD)	P	SPH173	SaqJ	QIRR*ETYG	na	H
cSP7/MP2	CpAA	FSNK*VYNG	T(DGG)	cP	SP134/Phae1	GchH	PEGR*VVGG	C(SGS)	P	SP175	SsnL	RGPR*IIAG	C(GGG)	P
cSP8	CuAB	GNPK*VSGG	C(HGG)	cP	SP135/Jon99Fii	GwGE	IQGR*ITNG	C(SVA)	P	SP176a	SNSL	ANSR*IVGG	T(DGG)	P
cSPH9	CsAA	VRWQ*RSND	na	cP	SP137/Jon25Bi	GCGE	INGR*IVNG	E(SVS)	P	SP176b	SNmL	SKTR*IVGG	T(DGG)	P
cSP10	CwAB	IRNR*IYDG	T(DGG)	cP	SP139/θTry	GIIH	REGR*IVGG	T(DGG)	P	SP178	SqUM	STTK*VISF	C(GGG)	P
cSP11	CwAD	AYNQ*ITKG	T(DGG)	cP	SP140	GAhG	PEGR*IING	C(GGS)	P	SPH182	SiQM	SSPK*FHGD	na	H
cSP12	CpAB	VESR*LLGG	T(DGG)	cP	SP141	GWJF	TKNR*IVGG	C(GGD)	P	SP183a	SNmL	RQGK*IFGG	T(DGG)	P
cSP14	CZAC	PLYR*MAYG	T(DGG)	4cP	SP143/Jon99Ciii	G4GE	IQGR*ITNG	C(SGA)	P	SP183b	SNSd	IQPR*IIGG	T(DGG)	P
cSP16	CZAB	PAYR*IFGG	T(DGG)	cP	SP145/Jon65Ai	GPGE	IEGR*ITMG	C(GGK)	P	SPH184	SjQM	FEML*ISGG	na	H
cSP18	COAA	YFKK*IVGG	T(DGG)	2cP	SPH146a	GWJH	PQGR*VIGG	na	H	SPH187	SBLM	ADFK*SIGI	na	H
cSP19	CXcC	STGR*IVGG	T(DGG)	cP	SPH146b	GWJH	FSSR*IVGG	na	H	SPH188	SRTI	PDHI*ITNG	na	H
cSP24/Easter	CsAA	LSNR*IYGG	T(DGG)	cP	SP147/Phae2	GchH	PEGR*VVGG	C(SGS)	P	SP189	SuSM	FHPR*IYNG	C(TGT)	P
cSP25/MP1	CpAA	FGDR*VVGG	T(DGG)	cP	SP150/Jon99Ci	G4GH	IEGR*ITNG	E(AVA)	P	SP190/Send2	G4GH	PEER*IIGG	C(AGG)	P
cSP26/Snk	CUBB	SVPL*IVGG	T(DGG)	cP	SP151	GRGE	AEGR*IVNG	E(SAS)	P	SP191	SSSM	HETR*VIGG	T(DGG)	P
cSP28/Psh	CzBD	LVIH*IVGG	T(DGG)	cP	SP152	GRGF	AEGR*IVNG	C(SGS)	P	SPH192a	SuSM	KFRR*VWGG	na	H
cSP31/Hayan	CzBA	LTVH*ILDG	T(DGG)	cP	SP153/Jon65Aii	GPGH	INGR*ITNG	C(GGK)	P	SPH192b	SuSM	PPVR*TLNK	na	H
cSP32	CHCC	RSNR*IVGG	T(DGG)	cP	SPH161	G6HH	PQGR*IAGG	na	H	SP193	SxSM	WGDA*LHRG	C(GGR)	P
cSP33/Spirit	CyBD	FFVS*VVGG	T(DGG)	cP	SP165/Jon66Cii	GmGG	MQGR*ITNG	C(SGA)	P	SPH194a/Spx1	SkTM	LSPR*IAGG	na	H
cSP34	CncA	QFPR*LTGG	T(DGG)	cP	SP171/Jon66Ci	GmGG	IEGR*ITNG	E(GVA)	P	SPH194b/Spx2	SkTM	LSPR*ITGG	na	H
cSPH35	CcDA	VGFK*ITGA	T(DGG)	cP	SP174	GWJF	LEHF*IVGG	E(GFD)	P	SPH195	SMLI	GDQR*IING	na	H
cSP36	CcaD	DQER*IVGG	T(DGG)	cP	SP177/Jon25Biii	GCGE	IEGR*ITNG	C(SVA)	P	SPH196a	SMIF	CNRT*TLGG	na	H
cSP38	CrAB	SRRK*PTKG	T(DGG)	cP	SPH179	G6HH	QSSR*LPAE	na	H	SPH196b	SMLI	AQSR*IIGG	na	H
cSP42	ChBB	TTPF*IVGG	T(DGG)	cP	SP180/γTry	GIIG	LDGR*IVGG	T(DGG)	P	SPH197	SDNK	RCGL*LTNG	na	H
cSP44	CHCC	KSGR*IVGG	T(DGG)	TMcP	SP181/Jon99Fi	GwGE	IQGR*ITNG	C(SVA)	P	SPH199	SnNL	LSPD*IVGP	na	H
cSP48	CLBA	STPF*IVGG	T(DGG)	cP	SP185	GWJH	LDNR*IVGG	C(GGD)	P	SPH200	SnQM	RAKR*LSSP	na	H
cSP54	CHbC	AQRR*IVGG	T(DGG)	cP	SP213/δTry	GIIG	LDGR*IVGG	C(GGD)	P	SPH201a	S1NK	LSND*IIFS	na	H
cSP56/Sb	CsCC	PETR*IVGG	T(DGG)	cP	SP215	GIIG	LDGR*IVGG	C(GGD)	P	SPH201b	S1NJ	PHQD*VFKE	na	H
cSPH58	CGDA	DNDKFPYS	na	2cH	SPH235	GAhH	IQPL*IIDG	na	H	SPH202/Intr	S2XM	IETL*LTDG	na	H
cSP59/Np	CHCC	PEPR*IVGG	T(DGG)	cP	SP15	SfjL	PSSY*IVGG	T(DGG)	P	SPH206	ShQL	YHQN*VVSI	na	H
cSP61	CZAb	PVFR*DRGA	T(DGG)	cP	SP17	SiMJ	IQKR*IVGG	T(DGG)	P	SPH207	SEKL	VQFN*VTEG	na	H
cSPH64	CdAB	VQGH*FYKG	na	cH	SP20	SiML	TLYK*IVGG	T(DGG)	P	SPH214	SmQM	RVKR*LSDG	na	H
cSPH66	CndA	KNPV*YVDG	na	cH	SP21	SbMJ	NVNR*IVGG	T(DGG)	P	SP216	SJNL	PTNR*IVGG	T(DGG)	P
cSP67	CHbB	LQKR*IIGG	T(DGG)	cP	SP22	SiMd	TRHR*IVGG	T(DGG)	P	SP217	SJNK	ISPK*IMHG	T(DGG)	P
cSPH69	CEED	FSFR*EEDT	na	cH	SP23	SuVN	EEIR*IVGG	T(DGG)	P	SP218	SJNL	TAMR*VVNG	T(DGG)	P
cSPH79/Mas	COdC	RRAR*VVGG	na	5cH	SP27	SUPL	HDDF*NGRS	T(DGG)	P	SP219	SJNK	VATR*IVRG	T(DGG)	P
cSPH93	CdFa	NGLQ*MVEG	na	cH	SP29	SpPJ	YVER*IFPN	T(DGG)	P	SP220	SJNN	IAFK*IIGG	T(DGG)	P
cSPH94	CbFD	GSPQ*VFGD	na	cH	SP37	SKOJ	SQFK*ILGG	C(NGG)	P	SP221	ShNL	FRIR*VIGG	C(YGG)	P
cSPH101	CpdB	AAPG*QASF	na	cH	SP39	SnMJ	DESR*IVGG	T(DGG)	P	SP223	SKNJ	NEEH*QAHI	T(VGG)	P
cSP115	CUBB	SQNL*LVGG	T(DGG)	cP	SP40	SnML	NVNR*IVGG	T(DGG)	P	SP224	SKOJ	GLYR*VING	C(SGG)	P
cSPH121	CGDD	YQLD*GYNN	na	2cH	SPH41	SJNK	QCGL*MREE	na	H	SP225	SKOK	YRAR*IDGG	T(DGG)	P
cSPH125	CbeD	TVEE*VVDQ	na	cH	SP43	S3PL	YTPL*IVGG	T(DGG)	P	SP43	S3PL	YVPN*IFGG	T(DGG)	P
cSPH128	CdED	HVNR*IGVG	na	cH	SP45	S3PI	SRPL*IVDG	T(DGG)	P	SPH227	SYWI	NSDN*IIAE	na	H
cSPH142/Scaf	CedA	TKPT*GVKD	na	3cH	SP62	SjOJ	FIPM*ITGG	T(DGG)	P	SP228	SXjL	STYR*MVGG	T(DGG)	P
cSPH156	CdFD	TERT*QPGG	na	cH	SP63	SnnL	RNPK*IVGG	T(DGG)	P	SP230	SBLI	PEER*IVGG	T(DGG)	P
cSPH166	CcDB	LRPL*GYKQ	na	cH	SP68	S3PI	YAPL*IIGG	T(DGG)	P	SP233	S5YL	ILPK*IVGG	T(DGG)	P
cSP229	CXAB	TTNR*VIGG	C(GGG)	cP	SP70	ShkJ	QDGR*IVGG	T(DGG)	P	SP234	SxpL	EDGK*IVNG	C(GGS)	P
cSPH231	CEED	FSFR*EEDT	na	cH	SP73a	SsNK	HRTR*IIGG	C(NGG)	P	SPH236	SiON	RIRR*VVGG	na	H
cSP232	COAB	TSNR*VVGG	T(DGG)	cP	SP73b	SsNK	SVPR*VKNG	C(NGG)	P	SP237	ShNL	YTYR*ITGG	E(YGL)	P
cSPH242	CoDA	FTLS*GVSQ	na	cH	SP74	SfNN	LRRR*ITGG	T(DGG)	P	SP238	ShNL	FRMR*IFGG	C(FGG)	P
SP6	GIIF	LDGR*IVGG	T(DGG)	P	SP82	ShLK	GDGR*IVGG	T(DGG)	P	SP239	SJNL	LSYK*IING	E(NGY)	P
SP46	GzHF	LNGR*VVGG	E(GSD)	P	SP84	SnTc	TDSY*AVGQ	T(DGG)	P	SP240	SJNK	ISER*SVNA	E(QGK)	P
SP47/Ser6	GzHH	LNGR*VVGG	E(GVD)	P	SP86	S5YN	IEPK*IVGG	T(DGG)	P	SP243	SKOJ	VREQ*ILGG	C(SGG)	P
SP50	GzHF	IEPR*IVGG	C(GGD)	P	SP92/Ser12	SBLL	SPER*IVGG	T(DGG)	P	SPH244	SdQJ	TTIK*INHY	na	H
SP51	GPGH	IDGR*ITNG	C(SVA)	P	SP95	SxYL	QQSR*IING	T(DGG)	P	SP245	Sund	VGGR*IVST	T(DGG)	P
SP52	GyGH	IDNR*IVSG	E(SVS)	P	SPH97	SOKJ	PNGL*VANV	na	H	SP246	SfNI	IRFM*ITGG	T(DGG)	P
SP57/ηTry	GIIH	SDGR*IVGG	T(DGG)	P	SP104	SQLL	PQER*IVGG	T(DGA)	P	SPH249	S1NK	HMER*INGS	na	H
SP65	GhgF	ISTH*IVGG	T(DGG)	P	SP105	SPTL	AMDR*IFGG	C(SGG)	P	SPH250	S1NK	FLEQ*NCGK	na	H
SP71	GnGH	VEPY*ITNG	C(SGA)	P	SP108	S5YN	DPGR*IING	T(DGG)	P	SPH252	S2XM	SKEP*VVTL	na	H
SP76λTry	GIIH	LDGR*IVGG	T(DGG)	P	SPH111	SDNL	LGPR*IVNG	na	H	SPH253	S2XM	QIVR*IINK	na	H
SP78	GPGE	PSGR*ITGG	C(SGG)	P	SP113	SBLI	WEQR*IIGG	C(TGG)	P	SPH254	S5YN	FQLK*MVAP	na	H-TM
SPH81	G6HH	PQGR*ILGG	na	H	SP118	ShkL	EESR*IIGG	C(SGG)	P	SPH255/Aqrs	S2XM	YNRE*ALEE	na	H
SP83	GdIE	PFGR*IVNG	T(DGG)	P	SP123	SPTN	GSLR*IMNG	C(LGG)	P	SP30	MnnJ	NMLK*IIGG	T(DGG)	2TSP-P
SP85/Jon74E	GnGH	IGGR*IAGG	C(SGS)	P	SP124a	SNSL	IQSR*IVGG	T(DGG)	P	SP49	MJNK	LIPK*IVGG	T(DGG)	PH
SP87	GpgF	SISR*VVNG	C(GSG)	P	SP124b	SNSL	RSPR*IVGG	T(DGG)	P	SP53/ModSP	MsoN	IKQF*SSGG	C(SGN)	Fig. 1
SP88/εTry	GIIE	LDGR*IVGG	T(DGG)	P	SP127	SPTL	VRTR*IAGG	C(GGG)	P	SP55	MVVI	QTPF*IHNG	C(GGG)	GdP
SP90/Ser8	GgIH	LGGR*IVGG	T(DGG)	P	SP130	SDNL	TAYR*IING	C(NGN)	P	SP60	MVVI	FSPL*QIGG	E(GGP)	GdP
SP91	GpgH	QMGR*VVNG	C(GSG)	P	SP132	StnJ	WTKK*FLAK	C(SGG)	P	SP72/Tequila	MmnN	REER*VVRG	T(DGG)	Fig. 1
SP96	GWJH	PETR*VIGG	C(GGD)	P	SP133	STUJ	SQER*VVGG	C(GGG)	P	SP75/Corin	MeiL	PSRR*IIGG	T(DGG)	Fig. 1
SP98/Jon65Aiii	GPGE	IEGR*ITNG	E(SVA)	P	SP136	SvkL	PDPR*IVGG	T(DGG)	P	SP77	MhNK	TRPK*ISGG	T(DGG)	PH
SP99	G6HF	VEPRI*VGG	E(GVD)	P	SP138	SBLI	IPER*IVGG	T(DGG)	P	SP80a	MFRc	ATTR*IANG	T(DGG)	CUBP
SP100/Jon44E	GGGH	IEGR*ITMG	C(SGA)	P	SP148a	SQLL	LPTR*IVNG	T(DGG)	P	SP80b	MFRK	FPNR*IANG	T(DGG)	CUBP
SP102/αTry	GIIE	LDGR*IVGG	T(DGG)	P	SP148b	SQLL	FPTR*IVNG	T(DGG)	P	SP89/Ndl	MknL	GDGR*IVGG	T(DGG)	Fig. 1
SP103	GAhE	ATGF*VING	C(SGS)	P	SP149	STUL	TPHR*IVGG	C(GGG)	P	SP120	MFRL	RIPR*IASP	T(DGG)	CUBP
SP106/Try29F	GcfF	LDGR*IVGG	T(DGG)	P	SP154	STUL	SPTR*INGG	C(GGG)	P	SPH144	MVVI	TTPL*IFQG	na	GdH
SP107	GbfF	LDGR*IVGG	T(DGG)	P	SP155	SsoJ	PDSR*IVNG	T(DGG)	P	SP186/Gd	MyVL	SLPS*ITRG	C(SGA)	GdP
SPH109/κTry	GIIH	PEGR*IIMG	na	H	SPH157	SEKL	VQFN*VTEG	na	H	SPH198	MyqL	SLPK*MSAP	na	Fig.1
SP110/ζTry	GIIH	PDGR*IVGG	T(DGG)	P	SP158	SzOL	AKLT*WWNY	C(HGG)	P	SP212	MVVL	TTPF*IVRG	C(GGG)	2Gd-P
SP112/Jon25Bii	GCGE	IEGR*ITNG	E(SVD)	P	SP159	SjSL	IQPR*IVGG	T(DGG)	P	SP222	MLNK	IALK*ITGG	T(DGG)	PH
SPH114/Sphe	G6HH	AQGR*IMGG	na	H	SP160	SxQM	FQFL*VTGG	T(DGG)	P	SPH241	MhNK	KTSE*NINF	na	HH
SP116/yip7	GPGE	ITGR*ITNG	C(SGA)	P	SP163/Send1	SBLI	PSER*IIGG	C(GGG)	P	SPH247	MLNJ	CGAP*ISNQ	na	HH
SP117/Jon65Aiv	GPGE	IGGR*ITGG	C(SGA)	P	SPH164	SBiM	SSNG*IYNG	na	H	SP248	MhNJ	PVPK*IISG	T(DGG)	PH
Sp119/βTry	GIIE	LDGR*IVGG	T(DGG)	P	SP167	SSSM	FQTR*VVGG	T(DGG)	P	SP251	MiNL	ITYR*VANG	T(DGG)	PH
Sp122/jon99cii	G4GE	IQGR*ITNG	C(SGA)	P	SP168/Sems	SSSM	YQTR*VIGG	T(DGG)	P

Names, gene symbols (if available), G-C-T-E IDs, putative activation sites (*), predicted enzyme specificity (T for trypsin, C for chymotrypsin, and E for elastase) and its key determinants (in parentheses), as well as domain structures are enlisted. The 1^st letters in the G-C-T-E IDs are group names: C for CLIPs, G for gut, M for other multi-domain, and S for other single-domain SPs/SPHs. The 2^nd letters show chromosomal locations: A–Z and 1–6 for gene clusters and a–z for other genes (Fig. 3). The 3^rd letters indicate positions (A–Y and a–q) in the phylogenetic trees (Fig. 2). The 4^th letters mark major and minor expression assemblies A–N and a–d (Fig. 4). In domain structure, P stands for SP catalytic domain, H for non-catalytic SP-like domain, “c” for clip domain(s), and “TM” for transmembrane region. “CUB” and “Gd” domains are also indicated. More complicated domain structures are presented in Fig. 1.

Of the 257 proteins, 190 are predicted to be activated by cleavage after Arg(161)/Lys(29) by SPs with trypsin-like specificity, 26 after Phe(8)/Tyr(3)/Leu(15) by chymotrypsin-like SPs, and 11 after Ala(1)/Gly(2)/Val(1)/Ile(2)/Met(2)/Ser(3) by elastase-like SPs, and 5 after His by SPs with special features. Regarding catalytic specificity, 114, 61 and 15 of the 190 SPs have trypsin-, chymotrypsin-, and elastase-like properties, respectively. The dominance of trypsin-like SPs is also clear in the other insects (Cao et al., 2015b and 2017; Tables S2 and S3).

3.2.2. The honeybee A. mellifera

Compared with the initial analysis (Zou et al., 2006), there are major improvements in the honeybee SP/SPH sequences: 44 of the 57 SP-related gene models had been updated by BeeBase personnel; 31 of the updated and 2 newly identified genes were fully confirmed using the assembled RNA-seq data; 13 of the previously updated gene annotations were further improved in this work (Table S2). The genome contains 46 SP and 13 SPH genes encoding 14 SPs and 7 SPHs with at least one clip domain (C), 10 SPs and 3 SPHs with other domains (M), and 22 SPs and 3 SPHs with only a PD or PLD (G-S). In the third group, the shorter ones are more likely related to digestion whereas the longer ones may participate in processes that need regulatory region(s) for proper functioning.

3.2.3. The red flour beetle T. castaneum

The repertoire of 98 SP and 70 SPH genes in the beetle (Zou et al., 2007) is larger than that in the bee. After we improved the gene models using the RNA-seq data, the genome encodes 55 CLIPs (36 SPs and 19 SPHs), 20 SP-related proteins with other domains (M, 18 SPs and 2 SPHs), and 102 with one PD/PLD (G and S unseparated, 28 SPs and 39 SPHs with 231–270 residues; 24 SPs and 11 SPHs with 271–387 residues). With nine newly identified SP/SPH genes, a total of 177 SP-related genes (Table S3) are further examined below.

3.3. General structural features and classification of the SP-related proteins in D. melanogaster

Among the 52 CLIPs most have a single clip domain. The exceptions are cSP14, cSP18, cSPH58, cSPH121, cSPH142, and cSPH79, which contain 4, 2, 2, 2, 3, and 5 clip domains, respectively (Table 1). InterProScan5 only recognized 25 of the 64 clip domains in these CLIPs but it successfully identified 21 non-CLIP multi-domain SP-like proteins, including SP89/Ndl, SP186/Gd, SP53/ModSP, SP72/Tequila, and SP75/Corin (Fig. 1). There are 12 different types of regulatory domains (e.g. LDL, Sushi) in these proteins. SP55, SP60, SPH144, SP186/Gd, and SP212 have a similar domain structure of 1–2 Gd-PD/PLD; SP80a, SP80b, and SP120 have a CUB domain followed by a PD; SP49, SP77, SP222, SP248, and SP251 contain a PD and a PLD; SPH241 and SPH247 have two tandem PLDs. Functions of the CLIPs and other multi-domain SPs/SPHs are worth exploring in the future. The classification of gut (G) and single-domain (S) SPs/SPHs is not as clear-cut as CLIPs or multi-domain SPs/SPHs. Therefore, we took a combined approach (Section 2.3) and separated them into 57 gut SPs, 8 gut SPHs, 82 single-domain SPs, and 37 single-domain SPHs (Table 1).

Fig. 1.

Domain organization of 66 multi-domain SPs and SPHs in D. melanogaster. Signal peptide and other structural elements (see symbols in inset) were predicted. The schematic diagrams are not drawn to scale.

3.4. Phylogenetic relationships, chromosomal locations, and expression patterns of the D. melanogaster SP-related genes

The phylogenetic trees exhibit complex relationships among the CLIPs, gut, and single- and multi-domain SPs/SPHs (Fig. 2). In panel A, all 18 CLIPBs (cSP38 to cSP12, cSP1/Grass, cSP8) reside in branch A; all 7 CLIPCs (cSP42, 48, 115, 33, 26, 31, 28) constitute the entire B branch; 11 CLIPDs (cSP36, 232-18, 67-54, 34-19, and 44-32-59-56) are located in branches a, A, b, c and C, respectively; 16 CLIPAs (cSPH142/Scaf to cSPH125) are in branches d, D–F, and e. Note that the CLIP A–D classification system was developed on the basis of sequence alignments (Fig. S1) (Kanost and Jiang, 2015) and amended to include CLIPEs (Cao et al., 2017). In panel B, 26 gut SPs and 1 gut SPH are analyzed: 16 Jonahs (Jon25B_i–iii, Jon44E, Jon65A_i–iv, Jon66C_i, _ii, Jon74E, 99C_i–iii, Jon99F_i, _ii) on branch G and 11 trypsins (α–ι, λ, and κ/SPH109) on branch I. In panel C, two groups of multi-domain SP-like proteins are discovered: branch R has three CUB-domain SPs (80a, 80b and 120); branch V consists of Gd-domain SP55, SP60, SP186/Gd, SP212 and SPH144 (Fig. 1). Other close relationships occur in groups G, S and M, with large branches (e.g. G, I, L, N) formed as a result of extensive putative gene duplications.

Fig. 2.

Phylogenetic relationships of the 52 CLIPs (A), 65 GP(H)s (B), and other 140 SP(H)s (including 21 multi-domain proteins lacking clip domain) (C) in D. melanogaster. Complete sequences of the proteins in each group were aligned and a phylogenetic tree was constructed using MrBayes v3.2.6. Probability values for branches are indicated near the branching points, with “*” representing 100%. Those on red background with the capital letters (A–Y) represent branches (probability ≥55%) on the tree, and other branches with lower probabilities are assigned with lower case letters (a–q) on blue background. Group name (C for clip, G for gut, M for other multi-domain, S for other single domain SP(H)), chromosomal location (A–Z, 1–6, and a–z in Fig. 3), tree position (A–Y and a–q in this figure), and expression profile (A–N and a–D in Fig. 4) are used to generate G-C-T-E identification code for each SP-related gene. These IDs, in various colors depending on their categories, are listed before the corresponding systemic names.

Locations of SP-like genes on chromosomes also reflect the history of family expansion (Fig. 3). There are 32 clusters (A–Z, 1–6) containing 3 to 13 genes in each. For example, all thirteen genes (trypsins α, β, γ, δ, ε, ζ, η, θ, ι, κ/SPH109, λ, SP6, and SP215) in cluster I belong to the G group and are responsible for digestion of dietary proteins (Davis et al., 1985). Likewise, clusters C, P and #4 contain 3, 4, and 3 Jonah genes that also participate in food digestion (Akam and Carlson, 1985; Carlson and Hogness, 1985). Except for SP186/Gd, all the Gd-domain SPs/SPH are in cluster V. Other gene clusters include H and Z for CLIPs, F for multi-domain, W and #6 for gut, and B, J, K and N for single-domain SP(H)s (Fig. 3). Interestingly, SPs/SPHs involved in development or immunity (e.g. Ndl, Gd, Snk, Ea, Mas, Sb, Tequila, ModSP, Grass, Psh, SPE, MP1) are often encoded by genes that are not members of clusters. Maintenance of the solitary state and, hence, proper dosage could be crucial for the conserved functions. A simpler explanation is that mutation in one member of a gene cluster with overlapping function may not produce a phenotype and, in other words, finding these mutants may be the result of a bias toward genes with unique function (e.g. defense).

Fig. 3.

Chromosomal locations of the 257 genes coding for D. melanogaster SP-related proteins. As indicated by the scale bar, positions of the genes are plotted in proportion on chromosomes, with “+” and “−” indicating positive and negative strands on the left and right, respectively. The G-C-T-E identification code for each SP-related gene is defined in the legend to Fig. 2, indicated along with the systemic name, and linked to its location by a straight line. Adjacent genes with high sequence similarities are grouped by lines in the same color and marked by a series of location IDs (A–Z and 1–6) on red background for different chromosomal segments. Regions in between are labeled with IDs on blue background (a–z).

Gene expression patterns associate with functions. Thus, we extracted and analyzed the publicly available RNA-seq data on D. melanogaster SPs and SPHs. The highest transcript levels occur in the gut group (Fig. 4B). Except for SP50, SP107, SP141 and SPH196a, all the SP-related transcripts in the G group are detected at high levels [log₂(FPKM+1): 8–16] in larval and adult gut tissues. The mRNA levels are so high in most cases that they are easily detected in the whole body samples at the feeding stages. The low expression in late embryo (16–24 h) may allow the 1^st instar to digest food right after hatching. In contrast, the production of digestive proteases ceases completely in the pupal stage until right before adult emergence (Carlson and Hogness, 1985). In comparison, SPs and SPHs in the S and M groups are expressed at low levels [log₂(FPKM+1): 1–7] (Fig. 4C). A group of genes with similar expression profiles is defined as an assembly here. Single-domain SP63/Send1 and SP190/Send2 are in assembly I; SPH255/Aqrs, SPH194a/Spx1, SPH194b/Spx2, SPH202/Intr and SP168/Sems are in assembly M. Sphinx1/2 are required for Toll activation (Kambris et al., 2006). Send1, Send2, Aquarius, Intrepid, Seminase, and others form a protein network required for sperm storage in mated females (Findlay et al., 2014; Schnakenberg et al., 2011). Multi-domain SP75/Corin and SP89/Ndl belong to assembly L; SP72/Tequila and SP53/ModSP to N. Detailed expression patterns of these and other related genes in the 52 datasets are shown (Fig. 4C). Among the CLIPs, cSP3, cSP61, cSPH93, and cSPH156 have unique expression patterns, whereas expression profiles of the other genes fall into assemblies A–D (Fig. 4A).

Fig. 4.

Transcript profiles of the 52 CLIPs (A), 65 GPs/GPHs (B) and 140 SPs/SPHs (including multi-domain SP-related ones) (C) in D. melanogaster. The mRNA levels in 52 kinds of tissue samples, as represented by log₂(FPKM+1) values, are shown in the hierarchically clustered gradient heatmap from blue (0) to maroon (≥10). The values of 0–0.49, 0.50–1.49, 1.50–2.49, … 8.50–9.49, 9.50–10.49, 10.50–11.49, … 15.50–16.49 are labeled in the color blocks as 0, 1, 2 … 9, A, B, … G, respectively. Groups of genes with similar expression patterns are shown in different colors. The branches on red background with the capital letters A–N represent reliable assemblies, while other branches are assigned with lower case letters a–D on blue background. Systemic names are listed on the right along with the G-C-T-E IDs (see definition in the legend to Fig. 2). Briefly, in the first position, C (in red) for CLIPs, G (in green) for gut SPs/SPHs, M (in pink) for other multi-domain SP-like proteins, and S (in blue) for other single domain SP(H)s. In the 2^nd to 4^th positions, chromosomal location (Fig. 3), tree position (Fig. 2), and expression assembly (Fig. 4) are shown in various colors as defined in the corresponding figures. The libraries names are abbreviated using: W, whole insect; E, embryo; L, larval; preP, prepupal; P, pupal; A, adult; M, male; F, female; h, hour; D, day; G, gut; FB, fat body; ID, imaginal discs, SG, salivary glands; C, carcass; AG-AmM, accessary glands of adult mated male; T, testes; O, ovaries of adult mated female; AvF, adult virgin female; MT, Malpighian tubules.

Gene duplication is a major mechanism for forming gene clusters. In this process, regulatory elements are often duplicated with the coding region, leading to similar profiles of expression of the gene copies unless mutations occur in these elements. We thus predicted and later observed a considerable correlation among phylogenetic relationships (T), chromosomal locations (C), and expression patterns (E) (Fig. 5). Of the 257 genes, 216, 147, and 249 form branches, clusters, and assemblies in the T, C and E groups, respectively, and 99 genes in 33 groups have T-C-E three-way consistency (Fig. 5, B and C). Such relationships are vital for future research on these genes by taking into account possible functional redundancy and coordinated regulation of gene expression. For instance, we may not observe any phenotype if we only knock out 1 or 2 of cSP32, cSP44 and cSP59.

Fig. 5.

Correlations of the D. melanogaster SP-related genes in clusters on chromosome (C), branches of phylogenetic tree (T), and assemblies of expression (E). (A) Venn diagram of numbers the genes located in clusters, branches, and assemblies of the entire S1A SP/SPH family. (B) Two- and three-way matches of the genes in the C(LIP), G(ut), M(ulti-domain), and S(ingle domain) groups. For each CTE triangle, numbers of 2- (C-T, T-E, or E-C) and 3- (C-T-E) way matched genes are indicated on the sides and center. Each match has at least two genes. For instance, in the G group, 42 C-T, 53 T-E, 37 E-C, and 37 C-T-E matches are found, involving a total of 65 SPs/SPHs. (C) A list of SP/SPH names with C-T-E matches in the C, G, M and S groups. Among the 52 CLIPs, 7 belong to three 3-way matches: cSPH69-231, cSP32-44-59, and cSP26-115. For the G group, there are ten C-T-E matches with 2–6 members in each and 37 proteins in total.

3.5. Orthologous relationships and function prediction of SP-related proteins from insects

Orthology disclosed by multiple sequence comparison and phylogenetic analysis is critical for inferring equivalence of function. To identify such relationships, we performed an initial phylogenetic analysis of all CLIPs from the fly, mosquito, bee, moth and beetle, and found that A. gambiae subgroup E differs drastically from subgroups A–D (Cao et al., 2017). After removing the CLIPEs, we generated a MrBayes tree of 247 CLIPs (Fig. S1), which exhibits a clear separation of clades A–D (probabilities or p: 0.95 to 0.99). We then aligned the sequences in each subgroup and constructed five trees using MrBayes v3.2.6 (Fig. 6). Three CLIPA branches and one CLIPB branch display 1:1:1:0:1 or 1:1:1:1:1 (Dm:Ag:Am:Ms:Tc, same species order below) orthologous relationships (p: 1.00). Lineage-specific expansions are common in subgroups A–C and E, suggesting that, in each cohort of descendants, at least one member is functional in that species to play their ancestor’s role. In comparison, seven 1:1:1:1:1, one 1:1:1:2:1, one 1:1:0:1:1, and one 2:3:1:2:2 orthologous sets (p: 1.00) constitute the entire subgroup D. We assume ancestors of this subgroup existed long before the radiation of holometabolous insects but do not understand why gene duplication seldom occurred. In spite of the complexity in parts of the trees (Fig. 6, A–C), our stepwise analysis of CLIPs yielded insights into their evolutionary relationships useful for guiding focused functional studies.

Fig. 6.

Phylogenetic trees of the CLIPAs (A), CLIPBs (B), CLIPCs (C), CLIPDs (D), and CLIPEs (E) in the five insects. Based on the initial analysis of 247 CLIPA–D’s (Fig. S1), entire protein sequences in each subgroup were aligned for building a phylogenetic tree using MrBayes v3.2.6. Probability values are indicated near the branching points, with “*” representing 100 and colored branches representing various sets of potential orthologous genes (probability >80, 3–5 species). The bold branches represent 1:1:1, 1:1:1:1 or 1:1:1:1:1 orthology, except for the set containing closely linked MsHP1a and MsHP1b. As shown in the inset, the protein names are in different colors.

Comparison of protein domain structures is another way to uncover hypothetical orthologous relationships among non-CLIP multi-domain SPs/SPHs, since their regulatory modules are likely associated with unique functions of the entire proteins. As shown in Table 2, nine orthologous groups (e.g. Nudel, Tequila, Corin, ModSP, Masquerade, Gd) are identified in these five insects with 1:1:1:1:1, 1:1:1:0:1 or a:b:c:d:e, where a–e are any integers from 1 to 6. While no functional data is available for any member of the SEA, TSP and CUB sets, future research should cast light on roles of these proteins with the conserved domain structures.

Table 2.

Domain organization of the non-clip, multi-domain serine proteases and their homologs in the five model insects

Name/feature	Domain structure	Dm	Ms	Ag	Am	Tc
Nudel	TM-3LDL-SP-2–4LDL-SPH-3/4LDL	Ndl/SP89	SP50	SP212	SP20	mSP19
Tequila	S-2/15CBD-LDL-SR-LDL-SR-SP	Tequila/SP72	-	SP213	SP23*	mSP7*
Corin	TM-Fz-2LDL-SR-SP	Corin/SP75	SP56	SP214	SP30	mSP15
/SEA	TM-SEA-Fz-2LDL-SPH	SPH198	SPH145	SPH216	SPH56	mSPH6
ModSP	S-4/5LDL-Sushi-Wonton-SP	ModSP/SP53	HP14a, HP14b	SP217	SP49	mSP3, 13
/TSP	S-2TSP-SP	SP30	SP55	SP218	SP38	mSP2
Masquerade	S-5/4Clip-SPH	Mas/cSPH79	SPH53	CLIPA15	cSPH41, 39	cSPH51
/CUB	S-CUB-SP	SP120, SP80a-b	HP27	SP201–6	SP28, 34	mSP9, 10
/Gd	S-1/2Gd-SP	SP186/Gd, 55, 60, 212, SPH144,	SP45, 138, HP19	SP207–10	SP46	mSP1, 4, 5, 11, 12
GRAAL	TM-SEA-EGF-5LDL-SR-SPH			SP220	SPH54	mSPH16
/SEA	TM-SEA-3LDL-Coil-SP				SP45
/SEA	TM-SEA-3LDL-SP-LDL					mSP8
/LamG	S-SP-Ig-2LamG			SP219
/Sushi	S-1/2/3Sushi-SP(H)		SP25, 112, SPH33
/2Sushi	S-2Sushi-SP-2Sushi-SP-2Sushi-SP					mSP14

CBD, type-2 chitin binding domain; CTL, C-type lectin domain; Coil, coiled coil region; CUB, a domain identified in complement 1r/s, uegf, and bmp1; EGF, a Ca²⁺-binding domain in epidermal growth factor; Gd, a conserved region first identified in DmSP186/Gd; Fz, frizzled domain; Ig, immunoglobulin domain; LamG, a domain in laminin γ-subunit; LDL, low-density lipoprotein receptor class A repeat; S, signal peptide; SEA, a domain identified in a sperm protein, enterokinase and agrin; SP, serine protease catalytic domain; SPH, non-catalytic serine protease homolog domain; SR, scavenger receptor Cys-rich domain; Sushi, Sushi domain, also known as CCP or SCR; TM, transmembrane region; TSP, type-1 repeat in thrombospondin-1; Wonton, a Sushi-like module first identified in M. sexta HP14 with six instead of four Cys residues.

^*:AmSP23 and TcmSP7 are more complex in domain organization than AgSP213 or DmSP72/Tequila. In between the last CBD and the first LDL of Tequila, AmSP23 has SR-CTL-Kringle-LDL-Apple domains whereas TcmSP7 has only LDL-Apple domains.

We have used the known SP-SPH pathways in D. melanogaster and M. sexta as templates to overlay their putative orthologs or orthologous groups and hypothesize the presence of similar systems in the other species, which likely stemmed from their last common ancestors existing before the radiation of holometabolous insects (Fig. 7). By inspiring research to test the overarching hypothesis, the current knowledge on these systems will be expanded and enriched. Panel A surmises that the Drosophila embryonic SP pathway for establishing dorsoventral polarity (Moussian and Roth, 2005) may be conserved in the five insects of complete metamorphosis. The detection of M. sexta SP50, SP45, HP28/SP30 and HP8 transcripts in adult ovaries and early embryos (Cao et al., 2015b) provided initial experimental support for the proposed SP cascade in the moth. Besides, if silencing of T. castaneum mSP19, mSP11/12, cSP44, or cSP136–138 gene expression in ovaries impairs embryo ventralization, the pathway may be functional in the beetle as well.

Fig. 7.

Predicted functions of the CLIPs and other multi-domain SPs and SPHs in the insect SP-SPH pathways. (A) The SP cascade that establishes the dorsal-ventral axis of D. melanogaster embryo is used as a template to identify ortholog sets for proposing similar pathways in the other insects. When an ortholog is not identified in the phylogenetic analysis, its closest homolog(s) are listed to assist functional exploration: for example, A. mellifera cSP9/10/14 under cSP26/Snk and T. castanusm cSP136–8 under cSP24/Ea. (B) Members of the SP-SPH network that mediates proPO and proSpätzle-1 activation in M. sexta are presented along with their orthologs or close homologs, as revealed by the serial phylogenetic analyses and domain structure comparison (Fig. 6, Table 2). The protein names are in blue (Dm), red (Ag), black (Ms), green (Tc), and brown (Am) fonts. Circles with a “+” sign represent positive feedback mechanisms, one being auto-activation of proPAP3 by PAP3 (Wang et al., 2014) and the other being indirect activation proHP6 by PAP1 (dashed arrow) and direct activation of proPAP1 by HP6 (Wang and Jiang, 2008). “?” indicates that the step (i.e. cleavage activation of proHP6 by uncut but active proHP1) is partially established (He et al., 2017).

The immune SP-SPH systems for proPO and proSpätzle activation in hemolymph have been understood to different degrees in several insects (Fig. 7B) (Kanost and Jiang, 2015; Park et al., 2010; Veillard et al., 2016). PO produces reactive compounds to kill and sequester pathogens; Spatzle elicits the Toll pathway to induce antimicrobial peptide synthesis. We propose that the putative protease orthologs may play similar roles: M. sexta HP14a/b corresponds to DmModSP, AgSP217, TcmSP3/13 and AmSP49; MsHP1a/b is closely related to DmcSP34, AgCLIPD1, TccSP52 and AmcSP8. While these initiation steps often involve one ortholog per species, subsequent steps of the pathways contain 1 to 9 SP/SPH paralogs, depending on the extent of lineage-specific gene expansion. For instance, M. sexta HP2, HP13, HP18a/b. HP21, HP22, SP33, and SP144 form one orthologous set. MsHP14a activates proHP2 in wandering larvae and pupae, and HP2 activates proPAP2 (He et al., 2018). HP21, activated by MsHP14a and activating both proPAP2 and 3 (Gorman et al., 2007; Wang and Jiang, 2007), has a functional overlap with HP2, but it acts mainly in hemolymph of the feeding larvae. In another example, Drosophila cSP24/Easter and cSP4/SPE activate Spätzle in embryos and adults (Jang et al., 2006; Mulinari et al., 2006), respectively, and so does their ortholog MsHP8 in larval hemolymph (An et al., 2010). It is possible that DmcSP3 and the other six CLIPBs in the same orthologous set (Fig. 7B) activate DmSpätzle1–6 in the same or different tissues or life stages. Likewise, functional redundancy may be extensive in the group of M. sexta SPH1 and the eleven A. gambiae CLIPA/E’s. While MsPAP1–3 all need the presence of a high M_r complex of SPH1 and SPH2 to generate highly active PO (Wang and Jiang, 2004), it is unclear if AgCLIPB9 also needs one or more of CLIPA4–7, 12–14, 26, 31, 32, E6, and E7 to generate active POs. MsPAP1 or AgCLIPB9 alone cleaved M. sexta proPOs at the correct peptide bond but yielded POs with a low specific activity (An et al., 2011; Gupta et al., 2005) and so did DmMP2 to DmPPO1 (An et al., 2013).

Genetic research on Drosophila immunity provided support for the hypothetical immune SP-SPH system (Fig. 7B). ModSP integrates signals from pattern recognition receptors and indirectly activates Grass and then cSP4/SPE (Buchon et al., 2009). Microbial proteases can activate cSP28/Psh that induces the Toll pathway via SPE and Spätzle (Issa et al., 2018). A wound in the integument induces cSP31/Hayan activation and melanization (Nam et al., 2012). The functions of Psh and Hayan are consistent with the central role of M. sexta HP6 in proSpatzle-1 and proPO activation (An et al., 2009 and 2010). The roles of cSP25/MP1 and cSP7/MP2 appear controversial: in vitro test indicates that MP2 is a PAP (An et al., 2013) but in vivo data suggest that MP2 activates proMP1 and MP1 activates proPO (Tang et al., 2006). This situation may reflect the limitation of our approach or, in contrast, the limitation of a genetic study in discerning the position or order of pathway members.

3.6. Levels of SP-related proteins in Drosophila whole body samples at different life stages

Protein abundances are useful information for functional studies, simply because the concerned proteins must exist at certain levels to act properly in a specific tissue or stage. This is particularly important if a large set of orthologs (e.g. DmcSP3, 4/SPE, 14, 16, 24/Easter, 25/MP1, 38, 61, 229) is in question. mRNA levels are informative and easier to get, but may not be well correlated with protein abundances. For instance, levels of 571 transcripts in M. sexta fat body corresponded with concentrations of their proteins in hemolymph with correlation coefficients of 0.41 (naive larvae) and 0.43 (immune challenged) (He et al., 2016). Therefore, we extracted LFQ values of SPs/SPHs from the published work (Casas-Vila et al., 2017) in whole insects at various life stages (Table S5) and displayed their relative abundances in the labeled heat map (Fig. 8). Among the 110 proteins identified, 44 are GPs/GPHs, 29 are other single domain SPs/SPHs, 30 are CLIPs, and 7 are non-CLIP, multi-domain SPs/SPHs. They account for 68%, 24%, 58%, and 41% of the G (65), S (119), C (52), and M (17) groups, respectively. Overrepresentation of the G and C groups indicates gut SPs/SPHs and CLIPs are present at higher levels than the average to be detected in whole insects by mass spectrometry. In contrast, lower abundances of the single-domain SP(H)s lead to the S group underrepresentation, which is in general consistent with the transcriptome data (Fig. 4). When we compared the mRNA and protein levels of specific genes, discrepancies were noticed. For instance, SP89/Ndl mRNA levels [log₂(FPKM+1): 1–4] are low and scattered (Fig. 4C) but its protein levels [log₂(LFQ/5×10⁶ + 1): 6–9] in adult females and eggs are among the highest (Fig. 8). Np/cSP59 mRNA (level: 1–5, Fig. 4A) is present in embryo, larva, pupa, but its protein (level: 6–8) is mainly found in pupa to affect bristle and wing development (Bridges et al., 1936). While the expression profiles of Np/cSP59, Sb/cSP56, and Masquerade/cSPH79 mRNA (levels: 1–6, Fig. 4A) closely resemble, the levels of Sb and Mas are below the detection limit in the whole insect. Fine dissection of tissue samples at various life stages and more sensitive proteomics analyses in the future should provide data on relative abundances of the SP-related proteins in this important model species.

Fig. 8.

Abundances of the 110 SP-related proteins in D. melanogaster at various developmental stages. Relative protein levels in the 16 egg, 16 larval, 20 pupal, 8 female adult, and 8 male adult samples at different time points, as represented by log₂(LFQ/5×10⁶ + 1) values, are shown in the hierarchically clustered gradient heat map from blue (0) to maroon (≥10). The values of 0–0.49, 0.50–1.49, 1.50–2.49, … 8.50–9.49, 9.50–10.49, 10.50–11.49, and 11.50–12.49 are labeled in the color blocks as 0, 1, 2 … 9, A, B, and C, respectively. Proteins identified in two or less of the 68 samples are eliminated. Due to high sequence identity, no distinction can be made in SP51-117, SP122-143, SPH195-196b, cSP7-10, cSP4-229, and cSPH69-231 pairs. The datasets or library names are abbreviated using: E, embryo; L, larval; P, pupal; M, male; F, female; L3c, crawling third instar larva; h, hour; D, day.

3.7. Conclusions

Serine proteases (SPs) and their catalytically inactive homologs participate in food digestion, reproduction, embryo development, innate immunity, and other insect physiological processes. We have verified a total of 191 SP and 66 SPH genes in the Drosophila genome, 52 more than previously reported. These include 52 CLIPs, 21 multi-domain, 65 gut, and 119 other single-domain SPs/SPHs. Revelation of the links among phylogenetic relationships, genomic locations, and expression patterns for 99 SP-related genes in 33 groups accounts for a substantial portion of the evolutionary history of this gene family in the model organism. mRNA and protein levels in different tissues or stages are presented as background information to help with predictions of potential function. In addition, we have elucidated hypothetical orthologous relationships of SPs/SPHs from D. melanogaster, A. gambiae, A. mellifera, M. sexta, and T. castaneum to inspire functional studies and information exchange among researchers studying these holometabolous insects and beyond.

Supplementary Material

Fig S1

Fig. S1. Phylogenetic trees of the 247 CLIPA–Ds in the five insect species. After removal of all 32 A. gambiae CLIPEs that deviate considerably (Cao et al., 2017), entire sequences of CLIPA–Ds in the five species were analyzed using MrBayes v3.2.6. Probability values (0–100) are indicated near the branching points, with “*” representing 100. Branches with the probability value of 100 are in various colors. Protein names and their grouping are marked in different colors, as shown in the insets.

Table S1

Table S1. Names and properties of the 257 SP-related proteins in D. melanogaster For each protein, listed are its systemic name [(c)SP(H)#], protein name (FBpp#), gene symbol (CG# or alias), gene name (FBgn#), group name (C for CLIP, M for other multi-domain, G for gut, or S for other single-domain SP/SPH), protein sequence, domain structure, comment, length, clip domain sequence (if available), putative activation cleavage site, predicted enzyme specificity of SP, gene information, chromosome number, exon number, chromosomal location, gene coordinates, +/− strand, and identifications of chromosome, tree and expression.

Table S2

Table S2. Features of the 57 SP-related proteins in A. mellifera For each protein, listed are its reference ID, systemic name [(c)SP(H)#], sequence, length, putative activation cleavage site, predicted enzyme specificity of SP, homolog, domain structure, group name (C for CLIP, M for other multi-domain, or S for single-domain SP/SPH expressed in gut or other tissues), and comment.

Table S3

Table S3. Features of the 168 SP-related proteins in T. castaneum For each protein, listed are its reference IDs, systemic name [(c/m)SP(H)#], protein sequence, length, domain structure, putative activation cleavage site, predicted enzyme specificity of SP, group name (C for CLIP, M for other multi-domain, or S for single-domain SP/SPH expressed in gut or other tissues), and note.

Table S4

Table S4. Levels of the 257 D. melanogaster SP/SPH mRNAs in different cDNA libraries Transcript profiles of the D. melanogaster SP-related genes in the 56 RNA-seq datasets representing 52 different conditions. FPKM values are shown in the heat map from cyan to white and then to red. The cDNA libraries are described in sheet 2.

Table S5

Table S5. Relative abundances of the 110 D. melanogaster SP/SPH proteins in different samples

Highlights:

• Extensively updated knowledge on serine protease-related proteins in the fruit fly, honey bee, and flour beetle;

• Inner links detected among phylogenetic relationships, genomic locations and expression profiles for 99 of the Drosophila genes;

• Putative orthologous relationships revealed among S1A members from the five holometabolous insects;

• Physiological roles predicted on the basis of orthologous relationships and functional data.

Abbreviations:

SP

serine protease

SPH

(non-catalytic) serine protease homolog

PD

SP catalytic domain

PLD

protease-like domain in SPH

CLIP

clip-domain SP or SPH

GP and GPH

gut serine protease and gut serine protease homolog

CUB

a domain in complement 1r/s, uegf and bmp1

FPKM

fragments per kilobase of transcript per million mapped reads

G-C-T-E

group-chromosome-tree-expression

Gd

a domain in Drosophila gastrulation defective

LDL

low-density lipoprotein receptor class A repeat

LFQ

label-free quantification

PO and proPO

phenoloxidase and its proenzyme

PAP

proPO activating protease

SEA

a domain in sperm protein, enterokinase and agrin

SRA

sequence read archive

TSP

thrombospondin