Improved A. ceylanicum gene predictions

Analyzing gene function depends crucially on the quality of gene predictions [60,61]. We thus repredicted A. ceylanicum protein-coding genes with BRAKER2 [62], merging these repredictions (“version 2.0” or “v2.0”) with our original gene predictions (“version 1.0” or “v1.0”) to produce a hybrid v2.1 set of 33,190 protein-coding genes. Recently, Uzoechi et al. have also repredicted A. ceylanicum genes using methods similar to ours [54]; their new gene set is also substantially better than earlier predictions (94.5% completeness), though less complete than our v2.0 or v2.1 gene sets (Table 1). Having predicted the v2.1 gene set, we annotated its predicted protein products with predicted N-terminal signal sequences [63], conserved protein domains [64,65], orthologies to genes in related nematode species such as N. americanus [66], and Gene Ontology (GO) terms describing biological and molecular functions [67,68]; annotations are listed in S1 Table.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Protein-coding gene predictions for A. ceylanicum.

https://doi.org/10.1371/journal.pntd.0014106.t001

Identifying ES proteins

To collect ES proteins from adult A. ceylanicum hookworms, we isolated hookworms from hamster hosts 20 days after infection in two independent experiments, incubated the hookworms in protein-free culture medium for three days, concentrated and purified their supernatants, subjected their proteins to mass spectrometry, and mapped the resulting protein spectra to a nonredundant set of protein sequences. By this means, we identified 565 genes encoding ES proteins observed in at least one ES collection (1.7% of all genes), and 350 genes encoding ES proteins in both independent collections (1.05% of all genes; S1 and S2 Tables). Uzoechi et al. and Wong et al. have also identified genes encoding ES proteins from adult A. ceylanicum [54,55]; mapping their ES genes onto our v2.1 gene set, we find 955 ES genes from their analysis (2.9% of all genes), of which 444 are identical to ES genes from ours (1.3% of all genes; 79% of our ES genes). This overlap is 27-fold greater than chance (two-tailed Fisher test, p-value = 0; S3 Table), and shows high reproduciblity of ES genes in A. ceylanicum. Conversely, each ES gene set has unique members; they collectively have 1,076 genes (S2 Table).

Out of the 565 ES genes identified, we consider two subsets for detailed analysis here: either categories of ES genes for which we can show statistically significantly enhanced overlaps with other gene categories of interest (S3 Table), or individual instances of ES genes for which we have found specific published information in the literature which indicates that an individual ES gene may have a biologically informative function or effect.

A. ceylanicum ES genes encode possible host-parasite interaction proteins

Two-thirds of our ES genes (380) were predicted to encode classically secreted proteins with N-terminal signal sequences, 4.6-fold above genome-wide background (q = 4.5•10^-181); the remaining one-third of ES genes (185) encode proteins that are presumably non-classically secreted (Tables 2 and S3) [69]. This mix of predominant but not universal N-terminal signals also exists in ES proteins from A. caninum, N. americanus, and H. contortus (S4-S6 Tables) [44,45,47], and may reflect nonclassical secretion of ES proteins through extracellular vesicles [70,71]. For ES genes that appear to encode non-classically secreted proteins, an alternative possibility remains that these genes have 5’ exons that encode signal sequences but that have been overlooked. We have no way to completely exclude this possibility without extensive manual curation of their gene structures [72,73]. However, there are two lines of evidence that incline us to think that many, if not all, predictions of non-classical ES protein secretion are correct. First, the gene predictions we used in this study were based on extensive RNA-seq data sets [59,74,75] which drove BUSCO completeness from 87.8% to 95.1% (Table 1), and thus are likely to have detected most 5’ exons. Second, noticeable fractions of ES protein-coding genes without known N-terminal signal sequences are observable not merely in our own data for A. ceylanicum (33% of 565 ES genes without a predicted signal sequence), but also for ES protein-coding genes in three other strongylids: A. caninum (63% of 315 ES genes), N. americanus (53% of 197 ES genes), and H. contortus (40% of 844 ES genes; all three ES data sets are analyzed in S5 Table). So, if the apparent non-classical secretion of ES proteins is an artifact due to exon mispredictions, this artifact is pervasive not only for our own protein-coding gene predictions of A. ceylanicum, but also for gene predictions in three other parasitic nematode genomes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Protein domains overrepresented in A. ceylanicum ES genes.

https://doi.org/10.1371/journal.pntd.0014106.t002

Compared to the whole A. ceylanicum proteome, the 565 ES genes disproportionately encoded several protein families with plausible roles in host-parasite interactions (Tables 2 and S3 and Fig 3) [40]. These included five types of proteases (aspartyl, astacin, cysteine, serine, and metallopeptidase) that may digest host tissues and blood [7,18,76–78], along with three types of protease inhibitors (Kunitz, TIL, and TIMP) that may protect A. ceylanicum from native or host proteases; both proteases and protease inhibitors may also counteract host immune responses [79,80]. Five ES genes encoded glutathione S-transferase, an enzyme thought to detoxify free heme generated during hemoglobin proteolysis and other toxins [40,81]. Sixteen ES genes encoded Stichodactyla helianthus toxin (ShK)-related proteins, which might suppress host T cells [82,83]; and nine ES genes encoded C-type lectin proteins, which might mimic mammalian immune proteins (enabling immunosuppression) or dissociate cells (enabling tissue invasion) [25,26,59,84,85].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Groups of ES genes encoding overrepresented protein domains.

This UpSet plot [240] shows, for the set of 565 ES genes, occurrences of 16 protein domains from the Pfam database that are overrrepresented in that set (Table 2). The x-axis lists individual protein domains, with the numbers of ES genes that encode them; abbreviations for protein domains are given in Table 2. The y-axis lists categories of ES genes that encode a particular set of overrepresented protein domains, with the numbers of ES genes in each category. Dots in the matrix show each category’s encoded domain or domains. Although most ES genes only encode one overrepresented protein domain, instances encoding two domains do exist, and are shown by connecting two dots with a line. Not shown are an additional 242 ES genes that lacked any of these 16 overrepresented Pfam domains. Most ES genes encode only a single protein domain, and all domains are unique to some ES genes; for instance, 109 ES genes encode only CAP domains while 22 ES genes encode only ASPR domains. A minority of ES genes encode two domains at once: for instance, 29 ES genes encode both CAP and ASPR domains, and four ES genes encode both ShK and either Astacin or CAP domains.

https://doi.org/10.1371/journal.pntd.0014106.g003

One quarter of the 565 ES genes (140) encoded secreted venom-like allergen proteins with CAP or SCP/TAPS domains [24]; in hookworms, these are called Ancylostoma secreted proteins, activation-associated secreted proteins, or ASPs [24,86,87]. One ASP, neutrophil inhibitory factor in A. caninum (Acan-NIF), has been experimentally shown to block neutrophil adhesion [88,89]; another ASP, hookworm platelet inhibitor in A. caninum (Acan-HPI), has been shown to prevent blood clotting [90]. We observed ES genes encoding orthologs of both these ASPs (Acey-NIF-B and Acey-HPI) which may themselves be immunosuppressant and antithrombotic proteins (S1 Table). However, this leaves the other 138 ES ASP genes with only conjectural functions.

ES genes also encode three other multigene families of secreted proteins that are suspected to be involved in infection, although their molecular roles remain largely uncharacterized: ASP-related (ASPR), transthyretin-like (TTL), and secreted clade V proteins (SCVP). ASPRs were first observed in A. ceylanicum as a divergent subfamily of ASPs transcriptionally upregulated immediately after host infection [59], and were later observed in ES proteins from N. americanus [44]. TTL proteins were previously observed as a major component of ES proteins from H. contortus [47] and as multigene families in several parasitic nematodes [91–93]; one TTL protein of H. contortus (HcTTR) antagonizes goat IL-4 in vitro, and thus may be immunosuppressive in vivo [94]. SCVPs were observed in A. ceylanicum as a novel family of secreted proteins upregulated in young hookworm adults that is conserved in C. elegans and expanded in hookworms and other strongylids [59].

Although we have focused on analyzing our own ES gene set, we have also identified statistically enriched protein domains for the 955 adult ES genes identified by Uzoechi et al. and Wong et al. (both involving Washington University, and thus collectively labeled “WashU”) and for their 511 ES genes not found in our data set (“WashU-only”; S3 Table). Domains preferentially enriched in WashU-only ES genes include aminopeptidases, fatty acid- and retinol-binding (FAR) proteins, SXP/RAL-2 proteins, and C-type lectins (Table 3). Aminopeptidases may cooperate with the five other protease types noted above to enable digestion and immunosuppression. FAR proteins are a nematode-specific protein family that may suppress immune responses in plant or animal hosts by sequestering lipid signaling molecules [95–97]. The SXP/RAL-2 protein family includes homologs of an immunodominant hypodermal antigen found in several parasitic nematodes: Ac16, in the dog hookworm A. caninum [98]; both As16 and As14, in the roundworm Ascaris suum [99–101]; Ani s 5 in the ascaridoid Anisakis simplex [102]; and Ov-RAL-2 in the filarial parasite Onchocerca volvulus [103]. Among the C-type lectins enriched in WashU-only ES genes, we observed two genes encoding structural mimics of mammalian mannose receptor (Table 4) [59]; in mammals, this receptor binds soluble secreted molecules from the whipworm Trichuris suis and from A. suum, and these A. ceylanicum mimics might thus be immunomodulatory [104]. Conversely, domains preferentially enriched in our ES genes included CAP, ASPR, cysteine protease, astacin, Kunitz protease inhibitor, ShK, TTL, and SCVP domains (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Protein domains differentially overrepresented in WashU-only ES genes or our ES genes. Protein domains overrepresented in WashU-only A. ceylanicum ES genes but not in our A. ceylanicum ES genes.

https://doi.org/10.1371/journal.pntd.0014106.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Individual A. ceylanicum ES genes of interest.

https://doi.org/10.1371/journal.pntd.0014106.t004

Other ES genes we observed were potentially relevant to infection despite not being from overrepresented families (Table 4). One encodes a macin homolog that may protect hookworms from being infected by bacteria, and thus enable their long-term residence in the human gut [105,106]. Three ES homologs of A. caninum anticoagulant protein [107] could act with Acey-HPI to prevent clotting during blood feeding. Seven other ES genes encode possible immunosuppressants: two ES apyrases and one ES adenosine deaminase could suppress both blood clotting and inflammation by hydrolyzing extracellular ATP [108–110]; phospholipases A2 and D, like FAR proteins, could suppress lipid signals in immune responses [111]; the ES macrophage migration inhibitory factor Acey-MIF-1, previously observed to bind the human MIF receptor CD74 and promote monocyte migration, could alter cellular immune responses [112,113]; and an ES deoxyribonuclease II could hydrolyze extracellular DNA nets secreted by neutrophils to trap and kill large pathogens [114].

Because ES proteins are exposed to the human host’s immune system and are likely to promote infection, ES proteins are promising targets for an anti-hookworm vaccine. Twelve ES genes encoded proteins that have already been shown to give partial protection as vaccines in laboratory mammals, or to be homologous to such proteins (Table 4): the cysteine protease Acey-CP-1 [115]; the M13 peptidases Acey-MEP-6 and Acey-MEP-7 [116,117]; the glutathione S-transferase Acey-GST-1, homologous to A. caninum’s Acan-GST-1 [118] and N. americanus’ Na-GST-1 [119]; a homolog of the immunodominant hypodermal Ac-16 antigen of A. caninum [98]; two homologs of apyrases in Teledorsagia circumcincta and H. bakeri [120,121]; two homologs of fatty acid- and retinol-binding proteins in A. ceylanicum [122] and Onchocerca volvulus [123]; one homolog of Brugia malayi calreticulin [124]; one homolog of Onchocerca volvulus fructose-1,6-bisphosphate aldolase [125]; and one homolog to A. suum enolase [126]. In addition to these strict orthologs, three instances of astacin, ASP, and TTL proteins (all overrepresented among ES proteins) have given partial protection as vaccines: Ac-MTP-1 of A. caninum, Acan-ASP-2 of A. caninum, and HcTTR of H. contortus [127–129]. Finally, two WashU-only ES genes encoded homologs of an extracellular Argonaute (exWAGO) protein in H. bakeri that binds small RNAs, is imported by mouse cells during infection with possible immunomodulatory effects, and confers partial protection as a vaccine in mice [130]; a third WashU-only ES gene encoded the aspartic protease Acey-APR-1, homologous to the vaccine subunit Na-APR-1 from N. americanus [131]. These results show the vaccine potential of ES proteins, with hundreds of other ES proteins remaining untested as vaccine subunits.

Over half of A. ceylanicum ES genes are functionally conserved in related blood-feeding parasitic nematodes: they do not merely have orthologs, but ES gene orthologs (Tables 5, S1, and S7). Out of 565 A. ceylanicum ES genes, 269 (47.6%) had ES gene orthologs in the closely related hookworm A. caninum [45], 21-fold more than expected by chance (p = 6.82•10^-301). For the more distantly related hookworm N. americanus [44], 141/565 ES genes (25.0%) had ES gene orthologs (18-fold over chance; p = 1.78•10^-137); for the still more distantly related blood-feeding H. contortus [47], 180/565 ES genes (31.9%) had ES gene orthologs (8.9-fold over chance, p = 4.10•10^-120). Considering all three species at once, 350/565 ES genes (61.9%) had some ES ortholog (12-fold over chance, p = 0). We found similarly high conservation for the 860 ES genes described by Uzoechi et al. (S7 Table), and overrepresented ES gene orthologies between A. ceylanicum and N. americanus were also observed by Uzoechi et al. [54]. These orthologies of A. ceylanicum ES genes in other species are predominantly to ES genes themselves (i.e., genes experimentally shown to encode ES proteins in these other species). When ES genes in other species are set aside, the remaining orthologies of A. ceylanicum ES genes fall to, or even below, background frequencies. For A. ceylanicum ES genes without known ES gene orthologs, orthology to A. caninum dropped to 1.6-fold over chance (p = 1.31•10^-29); in N. americanus, to 1.1-fold below chance (p = 0.22); and in H. contortus, to 2.4-fold below chance (p = 2.05•10^-30). Thus, the A. ceylanicum ES gene set predominantly encodes ES proteins not only in A. ceylanicum itself but also in related parasitic nematodes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Orthologies of ES genes of A. ceylanicum to those of related parasitic nematodes.

https://doi.org/10.1371/journal.pntd.0014106.t005

Identification of intestinal and immunoregulated genes

Because the hookworm intestine is an important source of ES proteins, we set out to characterize intestine- and non-intestine-biased expression of hookworm genes. We infected hamsters with A. ceylanicum hookworms, allowed infections to proceed for either 12 or 19 days, collected hookworms from hamster small intestines, extracted hookworm RNAs, and performed RNA-seq with 36.9 to 55.8 million reads per biological replicate (S8 Table). Hookworms at the 12-day infection stage were young adults that were too small for us to dissect further, so RNA-seq was done on whole 12-day hookworms. By 19 days, hookworms had grown into mature adults large enough to dissect, so we separated them into intestinal and non-intestinal tissues before RNA-seq. In addition, we hypothesized that hookworms would respond to the mammalian host’s immune system by increasing their expression of proteins that counteract the mammalian immune system. To detect hookworm genes regulated by the state of the host’s immune system, 12-day and 19-day infections were performed both in normal hamster hosts (nonDEX) and in hamsters whose immune systems had been suppressed by dexamethasone (DEX). We analyzed levels and changes of gene expression both for our RNA-seq data and for published A. ceylanicum RNA-seq data from adult male intestine, male adults, and female adults (Tables 6 and S1 and Figs 4-5). We defined a gene as significantly changing its activity between two biological conditions (e.g., intestinal versus non-intestinal 19-day tissues, or intestinal nonDEX versus intestinal DEX) if the gene changed its expression by at least 2-fold (log2FC ≤ -1 or ≥ 1) with a false discovery rate (p-value corrected for multiple hypothesis testing) of no more than 0.01 (FDR ≤ 0.01). For immunoregulated genes, it is important to note that upregulation of genes in a normal host is equivalent to downregulation of genes in a dexamethasone-immunosuppressed host; we have defined such genes as positively immunoregulated (and their reverse as negatively immunoregulated), but their regulation was observed by virtue of downregulation with dexamethasone.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Numbers of A. ceylanicum genes showing differential RNA expression (significant changes of gene expression between biological conditions).

https://doi.org/10.1371/journal.pntd.0014106.t006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Gene expression in A. ceylanicum.

Gene activity is shown for 8,130 A. ceylanicum genes with significant differences in expression between biological conditions, with biological replicates used for differential gene expression analysis on the x-axis and individual genes on the y-axis. Expression levels are in TPM (log10). Genes sharing similar patterns of expression are split into 10 clusters. Biological replicates are: young 12-day hookworm adults from normal hosts (YA_12dpi_noDEX); young 12-day adults from immunosuppressed hosts (YA_12dpi_DEX); dissected 19-day intestines from mature hookworm adults in normal hosts (intestines_19dpi_noDEX); dissected 19-day intestines in immunosuppressed hosts (intestines_19dpi_DEX); dissected 19-day non-intestinal tissues in normal hosts (non.intestines_19dpi_noDEX); dissected 19-day non-intestinal tissues in immunosuppressed hosts (non.intestines_19dpi_DEX); adult male intestine (WashU_male_intestine); adult males (WashU_adult_male); and adult females (WashU_adult_fem). On the x-axis, biological replicates are labeled as ‘condition_X’ where X is ‘01’ through ‘04’. The adult male intestine, adult male, and adult female data in Fig 4 is a reanalysis of A. ceylanicum RNA-seq data previously published by Wei et al. [74] and Bernot et al. [75]. Additional heatmaps of gene expression for all 33,190 genes and all replicates are given in S1 Fig.

https://doi.org/10.1371/journal.pntd.0014106.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Overlaps between groups of differentially expressed A. ceylanicum genes.

This UpSet plot shows the 12 most abundant overlaps of differential gene expression types (Table 6). The x-axis lists types of differential gene expression, and the numbers of genes showing them. Types are abbreviated as follows: “Female”, female-biased, determined from previously published data; “Intest”, intestine-biased, 19-day; “Male”, male-biased, determined from previously published data; “N_12d_imm”, 12-day negatively immunoregulated; “N_int_imm”, 19-day negatively immunoregulated intestinal; “N_nin_imm”, 19-day negatively immunoregulated non-intestinal; “Non_int”, non-intestine-biased, 19-day; “P_int_imm”, 19-day positively immunoregulated intestinal; and “P_nin_imm”, 19-day positively immunoregulated non-intestinal. The y-axis lists categories of genes that have a particular combination of differential gene activities, with their gene numbers; dots in the matrix show each category’s type or types of expression patterns. Although many genes have only one type of differential expression, instances with two different types do exist and are shown by connecting two dots with a line. Not shown are an additional 25,302 genes that did not show any types of differential gene expression. The most common instance of genes with two differential expression patterns is a set of 957 genes with both male-biased and positively immunoregulated intestinal expression. The next instance is a set of 225 genes with both positively immunoregulated intestinal and non-intestine-biased expression; although they are immunologically upregulated in the intestine, they are more strongly expressed in non-intestinal tissues.

https://doi.org/10.1371/journal.pntd.0014106.g005

By these criteria, in 19-day adults we observed 1,670 genes with intestine-biased expression and 1,196 with non-intestine-biased expression, the remaining 30,324 genes being intermediate (Table 6 and Figs 4 and 6). Of the intestine-biased genes, 723 (43.3%) were orthologous to H. contortus genes with intestine-biased expression (4.4-fold above background; p = 8.18•10^-306; S9 Table). For comparison, when we reanalyzed published RNA-seq data for male intestine versus whole male adults [74,75] as part of the same computation, we identified 635 genes with intestine-biased expression, of which only 180 were also found in our intestine-biased gene set (S9 Table). By chance alone, one would expect to observe 32 genes shared by both intestinal sets; the observed overlap was 5.6-fold above background (p = 9.93•10^-85) but limited (only 28% of the 635 genes were shared). The observation of over 700 genes with intestine-biased expression conserved between strongylids supports our larger gene set. The disagreement of these intestine-biased gene sets could have one or more causes: different protocols for dissection and RNA extraction; different background tissues for comparison to intestinal RNA-seq (non-intestinal here, versus whole adults in the previous data); different sexes (mixed-sex here, versus males previously); and different strains of A. ceylanicum (HY135 here, versus Indian previously), which could cause different rates of RNA-seq read mapping onto our HY135 reference genome (S10 Table) [132–134].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Differential expression of intestine-biased versus non-intestine-biased genes.

This volcano plot [242] compares the expression of A. ceylanicum genes (shown as dots) in 19-day (adult) intestine versus 19-day (adult) non-intestine, with gene expression being assayed by RNA-seq of dissected intestines versus dissected non-intestinal tissues. For each gene, the x-axis shows the fold change (i.e., ratio) of its expression levels in intestine versus non-intestine; intestine-biased genes are positive and non-intestine-biased genes are negative. The y-axis shows the statistical significance (corrected for multiple hypothesis testing as a false discovery ratio, or FDR) of each gene’s fold change. Both axes have logarithmic scales (log2 for fold changes on the x-axis, -log10 for FDRs on the y-axis). In our analysis, a gene is considered to be significantly intestine-biased if its fold change is ≥ 2 with an FDR of ≤0.01, and significantly non-intestinally biased if its ratio is ≤ 0.5 with an FDR of ≤0.01; dashed lines are shown for these two criteria. Genes encoding ES proteins are plotted as red dots. Other non-ES genes that are intestine-biased are plotted as gold dots; non-intestine-biased ones are plotted as green dots; and non-biased ones are plotted as dark gray dots. ES genes are found in all expression categories, including genes with non-intestinally biased expression; this is consistent with previous observations that ES proteins can be secreted not only by intestine, but also by cephalic/pharyngeal glands or cuticle [39,42,43].

https://doi.org/10.1371/journal.pntd.0014106.g006

When comparing A. ceylanicum in normal (nonDEX) versus immunosuppressed (DEX) hosts, we detected substantial changes of gene expression in 19-day hookworm intestinal tissues, but almost no changes of gene expression either in 12-day young adults or in 19-day non-intestinal tissues. We observed 1,951 genes upregulated in 19-day nonDEX intestine versus 19-day DEX intestine (5.9% of all genes), along with 137 downregulated genes (0.41%; Table 6 and Fig 7). In contrast, we observed only 32 genes positively or negatively immunoregulated either in 12-day young adults or in 19-day non-intestinal tissues (Table 6 and Fig 8). One explanation for this pattern is that only 19-day intestinal tissues are directly exposed to the host’s immune system through blood feeding. Although 12-day young adult hookworms inhabit the small intestine, they have only just started feeding on blood; up to that point, they instead probably eat mucosal cells [135,136]. By 19 days, hookworm adults have been feeding on blood for up to a week; however, most of their bodily contact with this blood (and, thus, with the host’s immune system) is through the lumen of their intestine (although cephalic or pharyngeal glands may interact with their host as well).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Differential expression of positively immunoregulated intestinal versus negatively immunoregulated intestinal genes.

This volcano plot compares the expression of A. ceylanicum genes assayed by RNA-seq of 19-day (adult) intestines from hookworms in normal hamster hosts (with normal immune systems) versus hookworms grown in dexamethasone-treated (immunosuppressed) hamster hosts. Aside from different conditions being compared, the x-axis, y-axis, and criteria for significance are as in Fig 6. Genes encoding ES proteins are plotted as red dots. Other non-ES genes that are positively immunoregulated in A. ceylanicum intestines are plotted as gold dots; negatively immunoregulated intestinal ones are plotted as green dots; and non-biased ones are plotted as dark gray dots. ES genes are prominently visible among genes with positively immunoregulated intestinal expression.

https://doi.org/10.1371/journal.pntd.0014106.g007

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Differential expression of positively immunoregulated non-intestinal versus negatively immunoregulated non-intestinal genes.

This volcano plot compares the expression of A. ceylanicum genes assayed by RNA-seq of 19-day (adult) non-intestinal tissues from hookworms in normal hamster hosts (with normal immune systems) versus hookworms grown in dexamethasone-treated (immunosuppressed) hamster hosts. Aside from different conditions being compared, the x-axis, y-axis, and criteria for significance are as in Fig 6. Genes encoding ES proteins are plotted as red dots. Other non-ES genes that are positively immunoregulated in A. ceylanicum non-intestinal tissues are plotted as gold dots; negatively immunoregulated non-intestinal ones are plotted as green dots; and non-biased ones are plotted as dark gray dots. Although their numbers are much smaller, a visible subset of ES genes also have positively immunoregulated non-intestinal expression.

https://doi.org/10.1371/journal.pntd.0014106.g008

Because nematodes lack corticosteroid receptor homologs [137], our analysis assumes that dexamethasone acts solely on the host immune system, with no direct effects on A. ceylanicum. However, when free-living larvae of the intermittently parasitic nematode Strongyloides stercoralis were treated in culture with dexamethasone by Rodpai et al., 433 S. stercoralis genes showed altered RNA expression [138]. Moreover, it has been speculated that dexamethasone administered to S. stercoralis-infected mammals may act directly on the nematodes to make them hyperinfective [139,140], perhaps through binding the S. stercoralis ecdysone receptor homolog [141]. Thus, it is possible that dexamethasone causes at least part of its transcriptional effects in A. ceylanicum by acting directly on the hookworm. That being said, some differences between S. stercoralis and A. ceylanicum may make this possibility less likely. First, the drug dosage applied to S. stercoralis by Rodpai et al. was ~ 3.6 mg of dexamethasone per 25 mg of agar, whereas the drug dosage in our experiments was 48-fold lower (3.0 mg of dexamethasone/ 1,000 mg of hamster). Second, dexamethasone has been shown through heterologous glucocorticoid-responsive transgenes to reach and affect all parts of the nematode body when fed to C. elegans [142]; in contrast, the transcriptional response of A. ceylanicum to dexamethasone treatment is almost entirely confined to its intestine. Thus, although we cannot rule out the possibility of direct transcriptional regulation by dexamethasone, we think the most likely mechanism of dexamethasone action on hookworms is by altering the immune system of their hosts.

Intestine-biased genes encode likely components of food digestion, detoxification, and absorption

Whereas ES genes overwhelmingly encoded proteins with N-terminal signal sequences alone, intestine-biased genes were modestly but significantly enriched for several predicted types of secreted or transmembrane proteins (Tables 7 and S9). These included secreted proteins, proteins predicted to have one transmembrane anchor sequence after the N-terminal signal peptide, and multi-transmembrane proteins. Intestinal genes disproportionately encoded protein families relevant to digesting, detoxifying, and absorbing food (Table 7 and Fig 9): aspartyl, cysteine, aminopeptidase, and metallopeptidase proteases [7,18,76–78]; UDP-glucoronosyl or UDP-glucosyl transferases, ecdysteroid kinase-like enzymes, and ABC transporters [81,143,144]; and major facilitator and sugar transporters [145,146]. Intestinal genes also disproportionately encoded histones; this raises the question of whether mitosis persists in the adult hookworm intestine, as has been observed or inferred for A. suum and H. bakeri [147–149]. In addition to these well-studied protein families, another notably overrepresented family was Strongylid L4 Proteins (SL4Ps), first observed in A. ceylanicum as a novel family of non-classically secreted proteins upregulated in fourth-stage (L4) larvae [59]. Analyzing previously published H. contortus intestinal RNA-seq data (S11 Table), we observed that SL4P genes were also disproportionately represented among intestine-biased H. contortus genes, along with several of the better-characterized protein families implicated in digestion, detoxification, or absorption (e.g., cysteine proteases, UDP-glucoronosyl/glucosyl transferases, and major facilitator transporters; S12 Table). Moreover, the C. elegans genes numr-1 and numr-2 encode SL4P proteins that are intestinally expressed [150]. We conclude that SL4Ps have conserved intestinal expression in hookworms and other nematodes, and that (as in C. elegans) their function might be to counteract toxic environmental stresses [151,152].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Protein domains overrepresented in A. ceylanicum genes with intestine-biased expression.

https://doi.org/10.1371/journal.pntd.0014106.t007

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Overlaps between groups of intestine-biased genes encoding overrepresented protein domains.

This UpSet plot shows, for the set of 1,670 intestine-biased genes, occurrences of 17 Pfam domains overrrepresented in that set (Table 7). The x-axis, y-axis, and matrix are as in Fig 3, except that the genes being considered are intestine-biased rather than ES protein-coding. Abbreviations for Pfam domains are given in Table 7. Not shown are an additional 1,451 intestine-biased genes that lacked any of these 17 overrepresented Pfam domains. Although most domains are solitary, three domains that may be associated with detoxification (APH, DUF1679, and EcKL) are encoded in 13 intestinally biased genes. Another 13 intestinally biased genes encode two or three histone-related domains (Histone, H2A_C, and CBFD), which might be associated with mitotic cells in the intestine of adult parasitic nematodes [147–149].

https://doi.org/10.1371/journal.pntd.0014106.g009

Immunoregulated genes encode signal transduction, male-associated, host-parasite, and ES proteins

The 1,951 positively immunoregulated intestinal genes were modestly, but significantly, enriched for encoding secreted proteins, proteins with a single transmembrane anchor, or both: in other words, possible secreted or cell-surface proteins (Table 8). Overrepresented protein families included possible signal transduction components: protein kinases, protein phosphatases, and SH2 scaffold proteins [153–155]. Other overrepresented families included homologs of nematode motile sperm proteins (MSPs) [156], along with two other classes of proteins associated with MSPs: MSP fiber 2 proteins (MFP2s) [157] and DUF236 proteins [158]. Although MSPs are indeed hallmark proteins of nematode sperm cells and are usually assumed to be entirely specific to sperm, there are in fact MSPs expressed in somatic cells and required for their mobility. Two different MSP homologs in Caenorhabditis elegans have cell and axonal migration RNAi phenotypes in male linker cells and hermaphroditic neurons, with one of them being expressed in linker cells [159,160]. Moreover, msp genes are expressed in C. elegans ADL chemosensory neurons [161], and are evolutionarily retained even in parthenogenetic nematode species lacking sperm [162]. Thus, nematodes express some msp genes somatically, which fits our observation here of msp, mfp2, and duf236 gene expression in dissected intestines of adult A. ceylanicum. In addition, positively immunoregulated intestinal genes disproportionately encoded protein families with possible roles in host-parasite interactions. These families included: astacin, leishmanolysin, metallopeptidase, and trypsin-like proteases; TIMP protease inhibitors; ShK-related proteins; ASPs and ASPRs; SCVPs; and immunoglobulin domain-containing proteins (Table 8). While most of these families were also enriched among ES genes, leishmanolysins, trypsins, and immunoglobulin domains were uniquely enriched here. Despite being expressed intestinally, only one of these immunoregulated genes had intestine-biased expression; instead, 258 were non-intestine-biased (3.67-fold over chance; p = 5.51•10^-78; S13 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 8. Protein domains overrepresented in positively immunoregulated intestinal A. ceylanicum genes.

https://doi.org/10.1371/journal.pntd.0014106.t008

Genes with types of immunoregulation other than positive intestinal had fewer notable protein families (S13 Table). The 137 negatively immunoregulated intestinal genes disproportionately encoded cysteine proteases and saposins, with saposins being uniquely enriched in this gene set (36-fold over chance; q = 9.94•10^-3). At least four C. elegans saposins have antimicrobial activity in vitro [163–166], as does a saposin from the strongylid parasite T. circumcincta [167]; conversely, one saposin in A. ceylanicum has no antibacterial activity but does lyse blood cells in vitro [168]. Negatively immunoregulated hookworm saposins might have either or both functions. Of the 26 positively immunoregulated non-intestinal genes, 15 encoded ASP proteins with CAP domains (46-fold over chance; q = 7.61•10^-19); 10 had non-intestine-biased expression (10.7-fold over chance; p = 1.12•10^-8). Both gene sets disproportionately encoded secreted proteins; neither set had prominently sex-biased expression.

Immunoregulated genes have significantly male-biased expression

Multicellular parasites such as hookworms are generally studied to understand mechanisms of infection common to both sexes, and indeed we collected both ES proteins and RNA from mixed-sex populations. However, males of the African tick Rhipicephalus appendiculatus excrete immunoglobulin-binding proteins that the males themselves do not seem to need, and that are instead required for coinfecting female ticks to feed on host blood efficiently [169]. This instance of sexual cooperation in a parasite made us wonder whether there existed male-biased or female-biased genes among our immunoregulated gene set. Using published RNA-seq data from A. ceylanicum adult males and females [75], we observed that 3,135 and 1,547 of our 33,190 predicted A. ceylanicum genes had male- or female-biased expression, respectively (Table 6 and Fig 5). Going on to check for overlaps of these gene sets with our immunoregulated gene set, we found that 50.1% of our positively immunoregulated intestinal genes (977/1,951; 5.3-fold over background; p = 0) were also male-biased, while only 11.6% of them (227/1,951; 2.5-fold over background; p = 3.34•10^-38) were female-biased (Table 9 and S13 and Fig 10).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 9. Gene categories overrepresented in A. ceylanicum 19-day immunoregulated genes.

https://doi.org/10.1371/journal.pntd.0014106.t009

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10. Overlaps between A. ceylanicum 19-day immunoregulated genes and other gene groups.

This UpSet plot shows, for 2,122 genes that in some way are immunoregulated in adult (19-day) A. ceylanicum, their 16 most abundant overlaps with ES genes or other gene expression sets (Table 9). Note that because only adult immunoregulated genes are considered here, the numbers of other gene categories (e.g., 1,003 male-biased genes in this plot) are lower than for the entire A. ceylanicum genome (3,135 genes; Table 6). Types are abbreviated as follows: “ES_genes”, encoding ES proteins; “Female”, female-biased, determined from previously published data; “Intest”, intestine-biased, 19-day; “Male”, male-biased, determined from previously published data; “N_int_imm”, 19-day negatively immunoregulated intestinal; “N_nin_imm”, 19-day negatively immunoregulated non-intestinal; “Non_int”, non-intestine-biased, 19-day; “P_int_imm”, 19-day positively immunoregulated intestinal; and “P_nin_imm”, 19-day positively immunoregulated non-intestinal. The y-axis lists categories of genes that belong to a particular combination of gene types, with their gene numbers; dots in the matrix show each category’s type or types. The largest overlap has 950 genes with both male-biased and positively immunoregulated intestinal expression; the largest overlap involving ES genes has 107 genes that also show positively immunoregulated intestinal expression.

https://doi.org/10.1371/journal.pntd.0014106.g010

Such predominantly male-biased expression, along with the enrichment of three male-associated gene families (MSP, MFP2, and DUF236) in this gene set, raises the question of whether the changes in hookworm intestinal gene expression we observed with different host immunological backgrounds were actually due to our having harvested a higher proportion of male hookworms from normal than immunosuppressed hosts, leading to a male skew in putatively immunoregulated genes. We see two reasons why such bias is not sufficient to explain our observations. First, we only observed extensive changes of gene activity in dissected intestines from 19-day adults, while observing much smaller changes in non-intestinal tissues from those same adults (Table 6 and Figs 5, 7, and 8). If apparent immunoregulation of genes had actually been due to overlooked male bias in the hookworms we collected from normal hosts, one would expect to see at least as much (if not more) changed gene expression in non-intestinal tissues (which contained most, if not all, of the gonadal tissue from adults) as we saw from dissected intestines. The absence of significant immunoregulation in non-intestinal tissues (or in our 12-day whole animals) is inconsistent with such an artifact. Second, positively immunoregulated intestinal genes are not a simple subset of male-biased genes, as would be expected if male selection accounted for immunological changes of gene expression: 747 of these 1,951 immunoregulated genes had no sex bias, and 227 of them had female-biased expression. We conclude that the transcriptional immunoregulation we describe here, though dominated by male-biased genes, is nevertheless real. This suggests that male and female A. ceylanicum have different responses to the host immune system, and that such differences might account for the observed male-bias from our mixed-sex A. ceylanicum 19-day intestinal RNA-seq data.

The host immune system affects ES gene expression

Protein products of ES genes are thought to affect the host immune system, but whether the host immune system affects ES genes has been unclear. Comparing the ES gene set with immunomodulated gene sets, we observed significant overlaps (Table 9 and Fig 10). Out of 1,951 positively immunoregulated intestinal genes in 19-day hookworms, 153 also encoded ES proteins (27.1% of ES genes; 4.6-fold over chance; p = 1.19•10^-59). Of 26 positively immunoregulated non-intestinal genes, 13 were also ES genes (29-fold above chance; p = 7.47•10^-17); such expression might reflect synthesis and secretion of ES proteins by cephalic/pharyngeal glands [42].

The 153 positively immunoregulated intestinal ES genes disproportionately encoded astacin proteases, TIMP and TIL protease inhibitors, ShK-like proteins, ASPs, ASPRs, and SCVPs (Table 10 and Fig 11). Of these genes, 69 had ES gene orthologs in A. caninum (20-fold enrichment; p = 6.00•10^-72), 24 in N. americanus (11-fold enrichment; p = 2.35•10^-18), and 24 in H. contortus (4.4-fold enrichment; p = 1.23•10^-9; S14 Table). As noted above, all of the above protein families may affect immunomodulation in hookworm hosts; in addition, astacin has been shown to enable tissue invasion in vitro by A. caninum [77]. Other immunoregulated ES genes also encoded proteins with possible functions in immunoregulation, antithrombosis, or digesting host tissue (S1 Table): the ASPs Acey-NIF-B [88,89] and Acey-HPI [90], two apyrases [109,110], deoxyribonuclease II [114], hyaluronidase [170,171], and superoxide dismutase [172,173].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 10. Protein domains overrepresented in positively immunoregulated intestinal A. ceylanicum genes that also encode ES proteins.

https://doi.org/10.1371/journal.pntd.0014106.t010

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 11. Overrepresented protein domains encoded by positively immunoregulated intestinal genes that also encode ES proteins.

This UpSet plot shows, for the set of 153 positively immunoregulated intestinal ES genes, occurrences of seven Pfam domains overrrepresented in that set (Table 10). The x-axis, y-axis, and matrix are as in Fig 3, except that the genes being considered are not only ES protein-coding but also positively immunoregulated in the 19-day intestine. Abbreviations for Pfam domains are given in Table 10. Not shown are an additional 53 positively immunoregulated intestinal ES genes that lacked any of these seven overrepresented Pfam domains. The domain patterns are a subset of those observed for ES genes alone (Fig 3), and represent possible virulence factors in the ES gene set.

https://doi.org/10.1371/journal.pntd.0014106.g011

Ethics statement

All animal experiments were carried out under protocols approved by the University of Massachusetts Chan Medical School. All housing and care of laboratory animals used in this study conformed to the NIH Guide for the Care and Use of Laboratory Animals in Research (18-F22) and all requirements and all regulations issued by the USDA, including regulations implementing the Animal Welfare Act (P.L. 89–544) as amended (18-F23).

Sample procurement, preparation and storage

Cultures, infections, and collections of A. ceylanicum followed published methods [59,194]. For each A. ceylanicum infection in support of RNA-seq, we purchased Syrian golden hamsters (Mesocricetus auratus) of the HsdHan:AURA strain at 3–4 weeks of age. Hamsters were provided with food and water ad libitum. For immunosuppression experiments, we immunosuppressed half of each set of hamsters by injecting them with dexamethasone (3 mg/kg) twice per week throughout the duration of the experiment [195]; the other half were given mock injections. After the first two injections (one week), both immunocompetent and immunocompromised hamsters were infected at approximately four to five weeks of age with A. ceylanicum. Infections were allowed to progress for 12 or 19 days, after which we euthanized the hamsters and collected hookworms from small intestines dissected from the hamsters. Dissected hamster intestines were put into Hank’s Balanced Salt Solution (pre-warmed to 37°C); worms were picked quickly from the dissected tissues by hand. Collected worms were snap-frozen in liquid nitrogen and stored at −80°C until use. For A. ceylanicum infections in support of ES protein mass spectrometry, we followed a similar protocol but without injections, with two collections of A. ceylanicum at 20 days after infection, and without snap-freezing of dissected hookworms. Stages of A. ceylanicum selected for RNA-seq or proteomics are based on previously described stages of growth in golden hamsters [196]. For hookworm intestinal-specific studies, from each triplicate set of 19-day post-infection hookworms isolated from hamsters as above, we dissected both hookworm intestinal and hookworm non-intestinal tissue, and extracted RNA from the dissected tissues; for the triplicates of smaller 12-day post-infection young adult worms, such dissection was not possible, so we extracted RNA from whole 12-day young adult worms.

RNA harvesting and sequencing

Total A. ceylanicum RNA was extracted from either whole worms or from dissected tissues as in Romeo and Lemaitre [59,197]. RNA-seq was done largely as in Srinavasan et al. [198]. RNA-seq libraries were built with Illumina’s TruSeq RNA Sample Prep Kit v2 executed according to manufacturer’s instructions, using 1 µg of total RNA for each sample. RNA-seq libraries were sequenced in single-end mode with read lengths of 50 nt (for all 19-day data) or 100 nt (for all 12-day data). Newly generated A. ceylanicum RNA-seq libraries are listed in S8 Table.

Protein harvesting

After 20 days of infection, A. ceylanicum hookworms were isolated from their hamster hosts and cultured in liquid culture medium for up to 3 days at 37°C in 5% CO₂. Liquid culture medium consisted of RPMI Medium 1640 (Gibco, Cat#11835–030) supplemented with 25 mM pH 7.2 HEPES buffer, 10 µg/ml amphotericin B (Gibco, Cat# 15290–026), and 100 U/ml penicillin/streptomycin (Gibco, Cat# 1570–063), with medium sterilized with a 0.22 µm filter. Fetal bovine serum was omitted from the medium to eliminate any added protein when quantifying ES proteins by BCA assay and to eliminate contamination for proteomics. After the first 24 hours, proteins excreted or secreted from the hookworms were collected by aspirating media only. The aspirated media were centrifuged to remove debris (1500 gs for 20 minutes, at 4°C); the resulting supernatant was concentrated 10-fold using a 3kD Amicon ultra-centrifugal filter (4000 gs for 45 minutes, at 4°C). After this first 10-fold concentration, PBS was added to match the initial volume, after which the media were reconcentrated by again being centrifuged (4000 gs for 45 minutes, at 4°C); we repeated this twice, for a total of three PBS rinses and reconcentrations. After the third PBS rinse/reconcentration, total protein was quantified using a Pierce BCA Protein Assay kit. Purified ES proteins were then frozen and stored at -80°C.

General computation

Where possible, we used mamba to install and run version-controlled software environments from bioconda [199]. For reformatting or parsing of computational results, we used Perl scripts either developed for general use or custom-coded for a given analysis. All such Perl scripts (named below with italics and the suffix “.pl”) were archived on GitHub (https://github.com/schwarzem/ems_perl). Internet sources (URLs) for other software are listed in S15 Table.

Nematode genomes, transcriptomes, coding sequences, and proteomes

For transcriptomic or proteomic analyses, published genome sequences, coding sequences, proteomes, and gene annotations of relevant nematodes were downloaded from WormBase [200] or WormBase ParaSite [184,193] (S16 Table). Published RNA-seq data of A. ceylanicum [59,74,75] and H. contortus [175] were downloaded from the European Nucleotide Archive (S11 Table). Alternative gene predictions for A. ceylanicum recently published by Uzoechi et al. [54] were obtained as a Generic Feature Format Version 3 (GFF3) annotation file (https://github.com/The-Sequence-Ontology/Specifications/blob/master/gff3.md) from Young-Jun Choi <choi.y@wustl.edu> that we have archived at https://osf.io/dxfsb. We extracted protein sequences from this GFF3 via gffread 0.12.7 [201] with the arguments ‘-g [input genome sequence FASTA file] -o/dev/null -C --sort-alpha --keep-genes -P -V -H -l 93 -y [output protein FASTA file] [input gene prediction GFF3 file]’.

Heligmosomoides species nomenclature

Recent genomic analysis has shown that the gastrointestinal parasitic nematode Heligmosomoides has two distinct species: H. bakeri and H. polygyrus [93]. This reinforced previous molecular phylogenetic observations indicating that H. bakeri is a terminal sister clade to H. polygyrus [202], and that both the morphology and the the host-specificity of H. bakeri are distinct from those of H. polygyrus [203–207]. Although laboratory strains of Heligmosomoides have often been described as H. polygyrus [208], it now appears likely that many or all of these strains have actually been H. bakeri. Thus, following previous suggestions for revised nomenclature [209], we refer exclusively to H. bakeri even when citing published work that was nominally done with H. polygyrus.

RNA-seq subsampling

We found that published A. ceylanicum RNA-seq data (S11 Table) were too extensive for gene predictions by BRAKER2 because of memory limitations. To make these data usable by BRAKER2, we selected a representative subset of them with khmer [210,211]. Because khmer requires “#/1” and “#/2” suffixes for paired-end reads, we retrofitted paired-end RNA-seq reads lacking such suffixes with retroname_fastq_reads.pl. We ran normalize-by-median.py from khmer on all data (paired and unpaired) twice, with the arguments ‘-k 31 -C 100 -M 100G’, and ‘-k 31 -C 30 -M 100G’; we then ran filter-abund.py from khmer on paired-end data with the arguments ‘--variable-coverage -C 2 [k-mer hash] [paired-end reads]’. We sorted khmer-filtered data into interleaved paired- and unpaired-end read files with paired_vs_unp_fastq.or.a.pl with the arguments ‘--r1 “#0\/1” --r2 “#0\/2”’.

Reprediction of protein-coding genes

To repredict protein-coding genes in A. ceylanicum, we ran braker.pl in BRAKER2 2.1.6 [62] on our published repeat-softmasked A. ceylanicum genome assembly [59], with the arguments ‘--genome [genome assembly FASTA] --prot_seq [collected proteomes FASTA] --bam [sorted mapped khmer-filtered paired-end read BAM alignment] --etpmode --softmasking --cores 48 --gff3’. BRAKER2 requires an input genome sequence to have its FASTA header lines previously stripped of comments; we did this with uncomment_FASTA_headers.pl. Running BRAKER2 also required us to install: AUGUSTUS 3.4.0 [212]; BamTools 2.5.2 [213]; cdbfasta 0.99 [214]; DIAMOND 2.0.9 [215]; GeneMark-ES/ET 4.68 [216]; and SAMtools 1.12 [217]. To guide BRAKER2 gene predictions, we used our khmer-subsampled paired-end subset of published A. ceylanicum RNA-seq data, along with predicted proteomes from the related nematodes Caenorhabditis elegans [200], H. contortus [218], and N. americanus [44]. After running BRAKER2, we renamed gene, transcript, and exon names in the GFF3 prediction file with a modified version of updateBRAKERGff.py (https://github.com/Gaius-Augustus/BRAKER/issues/416) and Perl one-line commands. To extract coding DNAs (CDS DNAs) and protein sequences from the renamed GFF3, we used gffread 0.12.7 [201] as above, with the additional argument ‘-x [output CDS DNA file]’.

This gave us a new gene set (“v2.0”) that was generally superior (in completeness as assayed by BUSCO, and nonfragmentation as assayed by count_fasta_residues.pl) to our original 2015 gene set (“v1.0”). However, we observed that some v1.0 genes encoded ES proteins (as determined by mass spectrophotometric mapping) but had no genomic overlap with v2.0 genes. To ensure no valid ES genes would be overlooked, we used BEDtools 2.30.0 [219] to identify v1.0 genes whose protein-coding exon sequences (CDSes) had no genomic overlap with v2.0 CDSes, as follows. We used get_gff3_gene_subset.pl to remove GFF3 v2.0 annotations that could not be translated by gffread, and then used Perl to extract CDS annotation lines from the GFF3 files of published v1.0 and gffread-translatable v2.0 gene predictions. We reformatted the CDS-subset files from GFF3 to BED with gff2bed in BEDOPS 2.4.41 [220]. We identified overlapping coordinates of the CDS-only BED annotation files for v1.0 and v2.0 genes with intersect from BEDtools 2.30.0 [219], using the argument ‘-loj’, and used Perl to extract a list of v1.0 genes having no CDS overlaps with v2.0 genes. Given this list of non-overlapping v1.0 genes, we used extract_fasta_subset.pl to extract their encoded CDS DNAs and proteins as subsets from the full published v1.0 CDS DNA and protein sets, and added these CDS DNA and protein subsets to our previous generated v2.0 CDS DNA and protein sequences. This gave us our final hybrid prediction set (“v2.1”) of CDS DNAs and proteins on which all further analyses in this paper were performed. The v2.1 CDS DNAs and proteome have been archived at https://osf.io/dxfsb.

We generated a v2.1 GFF3 as follows. From the published v1.0 GFF3, we extracted GFF3 annotations for these non-overlapping v1.0 genes first by running extract_parasite_GFF3_subset.pl and then by selecting ‘WormBase_imported’ annotation lines with Perl. We reformatted the resulting v1.0 non-overlapping GFF3 with AGAT 1.1.0 [221]. Likewise, we reformatted the gffread-translatable v2.0 GFF3 with AGAT. We then merged the two AGAT-reformatted GFF3s to produce a single v2.1 GFF3 with uniform formatting, which we have archived at https://osf.io/dxfsb.

Assessing quality of protein-coding genes

We determined general statistics of protein-coding gene products with count_fasta_residues.pl using the arguments ‘-e -t prot’; we obtained gene counts from maximum-isoform proteome subsets generated with get_largest_isoforms.pl using the arguments ‘-t parasite’ or ‘-t maker’. We determined and compared the completeness of predicted A. ceylanicum protein-coding gene sets with BUSCO 5.2.2 using the arguments ‘--lineage_dataset nematoda_odb10 --mode proteins’ [222], which tested a given proteome against 3,131 highly conserved single-gene orthologs in nematodes.

Gene reannotation

For the protein products of our v2.1 A. ceylanicum gene set, we predicted both N-terminal signal sequences and transmembrane alpha-helical anchors with Phobius 1.01 [63], reformatting results with tabulate_phobius_hits.pl. We predicted coiled-coil domains with Ncoils 2002.08.22 [223], reformatting results with tabulate_ncoils_x.fa.pl. We predicted low-complexity regions with PSEG 1999.06.10 [224] using the argument ‘-l’ and reformatting results with summarize_psegs.pl. We identified protein domains with Pfam 35.0 [183] database with hmmscan in HMMER 3.3.2 [225,226], using the arguments ‘--cut_ga’ to impose family-specific significance thresholds, ‘-o/dev/null’ to discard text outputs, and ‘--tblout’ to export tabular outputs; Pfam results were reformatted with pfam_hmmscan2annot.pl. We also identified protein domains with interproscan.sh in InterProScan 5.57-90.0 [65] using the arguments ‘-dp -iprlookup -goterms’, and reformatting results with tabulate_iprscan_tsv.pl. We generated Gene Ontology (GO) terms [67,68], EggNOG descriptions [227], and KEGG codes [228] with EnTAP 0.10.7-beta [229] using the argument ‘--runP’, and using selected UniProt [230] and RefSeq [231] proteome databases from highly GO-annotated model organisms (S16 Table); proteome databases were generated with makedb from DIAMOND 0.9.9 [215], itself bundled with EnTAP; annotations from EnTAP were reformatted from protein to gene annotation tables with cds2gene_EnTAP_annot.pl and cds2gene_annot.pl. We identified genes encoding possible antimicrobial peptides (AMP) by mapping predictions by Irvine et al. [106] from v1.0 to v2.1 of our gene predictions. We identified orthologies between A. ceylanicum and other nematodes with OrthoFinder 2.5.4 [66] using the arguments ‘-S diamond_ultra_sens -og’; results were reformatted with prot2gene_ofind.pl and genes2omcls.pl. For all analyses except OrthoFinder, full proteomes were used; for OrthoFinder, we used maximum-isoform proteome subsets generated with get_largest_isoforms.pl. An overall annotation table was constructed from both these and RNA-seq annotations (below) with add_tab_annots.pl.

Gene annotation for related parasitic nematodes

To make consistent comparisons of ES genes (or other categories of genes) between A. ceylanicum and other species, we annotated the protein-coding genes for three related parasites (A. caninum, H. contortus, and N. americanus) using the same methods as for A. ceylanicum above; the analyzed proteomes are listed in S16 Table.

Selection of ES gene sets from related parasitic nematodes for comparative analysis

We selected three published ES gene sets from A. caninum, H. contortus, and N. americanus for comparative analysis by the following criteria. They had to have either gene identification numbers (IDs) from public genome assemblies, or, if they did not have gene IDs from valid genomes, they needed to at least have expressed sequence tag (EST) IDs for which the original sequence data were publicly available so that they could be mapped onto genomes (and thus to modern genes with proper IDs). This requirement disqualified previously published ES gene sets for H. bakeri and N. brasiliensis. Second, the gene IDs needed to be correctly mapped from protein mass spectrometric data of one species onto genes of the same species. Remarkably, this requirement disqualified the ES genes from Angiostrongylus, which were generated from A. vasorum protein mass spectrometry data but were mapped onto genes of A. cantonensis and A. costaricensis. Third, the ES genes needed to be from parasitic nematodes that had evolved parasitism in common with A. ceylanicum hookworms. This criterion disqualified ES genes from the giant roundworm Ascaris suum or the whipworm Trichuris muris, which evolved parasitism independently of hookworms and other strongylids.

RNA-seq data

For A. ceylanicum, we generated biologically triplicated RNA-seq data for each biological condition. We also analyzed published RNA-seq data for A. ceylanicum [59,74,75] and H. contortus [175], listed in S11 Table. Before analysis, we Chastity-filtered new A. ceylanicum RNA-seq reads with quality_trim_fastq.pl using the arguments ‘-q 33 -m 50’. We then quality-filtered and adaptor-trimmed new A. ceylanicum RNA-seq reads with fastp 0.20.0 [232] using the arguments ‘--dont_overwrite --detect_adapter_for_pe --n_base_limit 0 --length_required X’, with X = 50 for new A. ceylanicum data. We quality-filtered and adaptor-trimmed previously published A. ceylanicum and H. contortus RNA-seq reads with fastp 0.20.0 using the arguments ‘--dont_overwrite --n_base_limit 0 --length_required X’ with X = 50 for A. ceylanicum.

RNA-seq expression values and significances

We quantified gene expression from RNA-seq data sets with Salmon 1.9.0, generating expression values in transcripts per million (TPM) and estimating mapped read counts per gene [233]. To prevent spurious mappings of RNA-seq reads, we used full selective alignment to a “gentrome” (a CDS DNA set, treated as a target for real mappings, combined with its genome, treated as a decoy for spurious mappings), followed by quantification using Salmon in non-alignment mode [234,235]. For Salmon’s index program, we used the arguments ‘--no-version-check --keepDuplicates -t [gentrome sequence] -d [decoy list]’; for Salmon’s quant program, we used the arguments ‘--no-version-check --seqBias --gcBias --posBias --libType A --geneMap [transcript-to-gene table]’, with ‘--unmatedReads’ used for single-end data. Results from quant.genes.sf output files were reformatted with make_salmon_TPM_slices.pl.

Heatmapping of RNA-seq data

To visualize and cluster gene expression values for RNA-seq replicates, we converted expression values of 0 TPM to empirical pseudozeros with assort_tpms.pl, with each replicate’s empirical pseudozero being defined as the lowest non-zero TPM expression value observed for all genes within that replicate. This allowed logarithmic transformation of zero expression values while defining these values in a replicate-specific way. We then converted the expression values of replicates to log₁₀(TPM) scores with log10_tsv_numbers.pl. Finally, we heatmapped and dendrogram-clustered all replicates with ComplexHeatmap 2.18.0 [236] using default parameters. Heatmapping showed individual non-dexamethasone (nonDEX) and dexamethasone (DEX) intestinal RNA-seq replicates that anomalously clustered with two DEX and nonDEX intestinal RNA-seq replicates, respectively (S1 Fig); these anomalous replicates were thus not used for differential gene expression analysis.

Differential gene expression analysis

Having generated RNA-seq readcounts, we used the exactTest function of edgeR 3.36.0 [237] to compute log₂ fold-changes (log2FC) and false discovery rate (FDR) significance values [238], to identify statistically significant changes of gene expression between pairs of biological conditions in our RNA-seq data. Dispersions were computed with edgeR’s estimateDisp function using the argument ‘robust=TRUE’. For this analysis, we used all non-anomalous replicates of our A. ceylanicum RNA-seq data (S8 Table) along with published A. ceylanicum RNA-seq data from whole fourth-stage larval (L4) males, whole adult males, whole adult females, and adult male intestine (S11 Table). To increase the statistical signal for differentially expressed genes, we used filter_minimum_readcounts.pl with the argument ‘10’ to remove genes which failed to achieve 10 mapped reads in any biological replicate before submitting readcounts to edgeR. An edgeR R script (S1 File) was generated with make_edgeR_classic_scripts.pl, and run in batch mode with R 4.1.3. In summarizing edgeR results, we defined a gene as being differentially expressed (e.g., with tissue-, sex-, or condition-biased expression) if edgeR scored it as being ≥2-fold more strongly or weakly expressed in one condition than another (either log2FC ≥ 1, or log2FC ≤ -1) with an FDR of ≤ 0.01. All other genes with intermediate expression ratios between two conditions (-1 < log2FC < 1), or with expression ratios that were determined with lower significance (FDR > 0.01) were not classified as having differential gene expression. We used edgeR’s exactTest with these criteria because it is well-adapted for analyzing small numbers of biological replicates [239].

UpSet plotting of gene set overlaps

To visualize overlaps between categories of genes, we used UpSet 2 [240,241] at the Visualization Design Lab at the University of Utah (https://vdl.sci.utah.edu/upset2), uploading our data to UpSet 2 via Multinet.app (https://upset.multinet.app).

Volcano plotting of RNA-seq data

To visualize changes of gene expression between two biological conditions in our RNA-seq data, we generated volcano plots [242] with EnhancedVolcano 1.24.0 (https://github.com/kevinblighe/EnhancedVolcano). R scripts (S2 File) were run in batch mode with R 4.4.3.

RNA-seq and differential gene expression analysis for H. contortus

These were performed with published RNA-seq data for dissected adult female intestine and whole adult female bodies (S11 Table) [175] and with predicted genes from the published chromosomal reference genome (S16 Table) [218], by the same methods as used for A. ceylanicum.

Constructing a protein set for mass spectrophotometric analysis

To combine our newly predicted A. ceylanicum proteins with alternative predictions by Coghlan et al. [185], we ran cd-hit-2d from CD-HIT 4.8.1 [243] with the arguments ‘-i [v2.1 predicted proteome] -i2 [alternative published proteome by Coghlan et al.] -o [distinct Coghlan protein sequences] -d 100 -c 1.0 -M 0 -T 1 -l 5 -s 0.0 -aL 0.0 -aS 1.0’. We then joined the v2.1 proteome with distinct Coghlan sequences to yield a nonredundant A. ceylanicum protein set for LC-MS/MS analysis.

Liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis

Peptides were analyzed with an Orbitrap Exploris 480 hybrid quadrupole-Orbitrap mass spectrometer coupled to an UltiMate 3000 RSLCnano liquid chromatography system (ThermoFisher Scientific). Peptides were loaded onto an Acclaim PepMap RSLC column (75 µm × 15 cm, 2 µm particle size, 100 Å pore size, ThermoFisher Scientific) over 15 min in 97% mobile phase A (0.1% formic acid) and 3% mobile phase B (0.1% formic acid, 80% acetonitrile) at 0.5 µL/min. Peptides were eluted at 0.3 µL/min using a linear gradient from 3% mobile phase B to 50% mobile phase B over 120 min. Peptides were electrosprayed through a nanospray emitter tip connected to the column by applying 2000 V through the ion source’s DirectJunction adapter. Full MS scans were performed at a resolution of 60,000 at 200 m/z over a range of 300–1,200 m/z with an AGC target of 300% and the maximum injection time set to ‘auto’. The top 20 most abundant precursors with a charge state of 2–6 were selected for MS/MS analysis with an isolation window of 1.4 m/z and a precursor intensity threshold of 5 × 103. A dynamic exclusion window of 20 s with a precursor mass tolerance of ± 10 ppm was used. MS/MS scans were performed using HCD fragmentation using a normalized collision energy of 30% and a resolution of 15,000 with a fixed first mass of 110 m/z. The AGC target was set to ‘standard’ with a maximum injection time of 22 ms.

Mass spectrometry data analysis

Thermo RAW files were searched against our nonredundant A. ceylanicum protein set using the SEQUEST algorithm in Proteome Discoverer 2.4 (Thermo). Data were searched for peptides with tryptic specificity with a maximum of two missed cleavages. The precursor mass tolerance was set at 10 ppm and the fragment mass tolerance was set at 0.02 Da. Search parameters included carbamidomethylation at cysteine (+57.021 Da) as a constant modification and the following dynamic modifications: oxidation at Met (+15.995 Da), acetylation at protein N-termini (+42.011 Da), Met loss at protein N-termini (-131.040 Da), Met loss+acetylation at protein N-termini (-89.030 Da). The Percolator node of Proteome Discoverer was used for PSM validation at a false discovery rate of 1%. Final results for both LC-MS/MS analyses are given in S17 Table.

Mapping ES mass spectrometry hits onto gene predictions

Our protein database for analyzing ES data included both our new v2.1 A. ceylanicum predictions and any predictions by Coghlan et al. that differed by even one amino acid residue from ours. We selected ES protein hits in S17 Table whose experimental combined q-value was ≤ 0.01; for each ES sample, we extracted their gene names as separate v2.1 and Coghlan genes, yielding four ES gene lists in all.

Because gene prediction in complex eukaryotes remains challenging [60] and because the Coghlan predictions were made on a different strain of A. ceylanicum (Indian strain, US National Parasite Collection Number 102954) than our laboratory strain used for genome sequencing and gene prediction (HY135), ES proteins whose mass spectrophotometic data fit a Coghlan prediction exclusively might do so only through slight differences in predicted gene structures or through strain-specific amino acid polymorphisms. To identify such cases where a Coghlan gene was equivalent to one of our v2.1 genes, we mapped the protein-coding exons (CDSes) of Coghlan genes onto our v2.1 gene set, as follows. We downloaded Coghlan gene annotations in GFF3 format and their corresponding A. ceylanicum genome assembly from WormBase ParaSite (S16 Table). We mapped the coordinates of Coghlan gene annotations onto our A. ceylanicum genome assembly sequence (i.e., lifted them over) with Liftoff 1.6.3 [244] using the arguments ‘-copies -polish -cds’. We extracted CDS coordinate lines from the resulting liftover GFF3 file with Perl, and did likewise for the CDS coordinates from the GFF3 of our v2.1 gene prediction set. We converted Coghlan and v2.1 CDS coordinate files from GFF3 to BED format with gff2bed in BEDOPS 2.4.41 [220]. We identified overlapping Coghlan and v2.1 CDSes with intersect in BEDtools 2.30.0 [219], using the argument ‘-loj’. We extracted columns 10 and 20 from the resulting intersection file with cut (https://www.gnu.org/software/coreutils/cut) using the argument ‘-f 10,20’, and processed the results with washu.cds_loj_umass.cds_genemap.pl using the positional arguments ‘[CDS-to-gene table] [BEDtools intersect columns 10 and 20 table]’; the CDS-to-gene map file was extracted from the FASTA headers of Coghlan and v2.1 proteomes via Perl.

We then used annot_es_v01.pl with our Coghlan-to-v2.1 gene map and our four ES gene lists to produce a unified list of ES-encoding v2.1 genes.

Mapping ES genes of Uzoechi et al. onto our gene predictions

Uzoechi et al. have recently identified ES-encoding genes in A. ceylanicum, using their own reprediction of our original 2015 gene set [54]. To compare their results to ours, we first mapped their results from ES transcripts to ES genes with map_tx_to_genes.pl. We then mapped their predicted gene set onto ours using methods similar to those for the Coghlan gene set. Because their repredictions were performed on our A. ceylanicum HY135 genome sequence rather than their A. ceylanicum Indian genome (as Coghlan predictions had been), we did not need to use Liftoff to map coordinates from one genome to the other. Subsequent mapping steps (CDS extraction with Perl; GFF to BED conversion with BEDOPS gff2bed; overlap determination with BEDtools intersect; extraction and summarizing of ES gene results with annot_es_v02.pl) were as above.

Statistical significance of overlapping gene sets

To identify significant overlaps between sets of A. ceylanicum genes, we used the Perl script motif_group_fisher.pl, which computed p-values from two-tailed Fisher tests with the Perl module Text::NSP::Measures::2D::Fisher::twotailed; for multiple hypothesis testing (e.g., comparisons to sets of protein domains) this Perl script also computed q-values via the qvalue program of the MEME 5.4.1 software suite [245].