Integrated analysis of the sialotranscriptome and sialoproteome of the rat flea Xenopsylla cheopis

Abstract

Over the last 20 years, advances in sequencing technologies paired with biochemical and structural studies have shed light on the unique pharmacological arsenal produced by the salivary glands of hematophagous arthropods that can target host hemostasis and immune response, favoring blood acquisition and, in several cases, enhancing pathogen transmission. Here we provide a deeper insight into Xenopsylla cheopis salivary gland contents pairing transcriptomic and proteomic approaches. Sequencing of 99 pairs of salivary glands from adult female X. cheopis yielded a total of 7,432 coding sequences functionally classified into 25 classes, of which the secreted protein class was the largest. The translated transcripts also served as a reference database for the proteomic study, which identified peptides from 610 different proteins. Both approaches revealed that the acid phosphatase family is the most abundant salivary protein group from X. cheopis. Additionally, we report here novel sequences similar to the FS-H family, apyrases, odorant and hormone-binding proteins, antigen 5-like proteins, adenosine deaminases, peptidase inhibitors from different subfamilies, proteins rich in Glu, Gly, and Pro residues, and several potential secreted proteins with unknown function.

Graphical Abstract

Introduction

Fleas are small (2 – 10 mm) insects that belong to the order Siphonaptera which contains over 2,500 known species organized into 238 genera that can parasitize mammals and birds (1,2). In contrast to other insects, fleas are laterally flattened, wingless and are able to jump more than 100 times their body length (3). Fleas live in close association with their host, and, even though some flea species can have preferred hosts (e.g., rats for Xenopsylla cheopis), they can successfully feed on alternative ones in the absence of their primary hosts (4). While the adults possess specialized mouthparts that favors piercing of the host skin for blood acquisition, the larval diet is based on organic debris, including the feces of adults (3).

Xenopsylla cheopis, commonly referred as the “oriental rat flea”, is the main vector of the Gram-negative bacterium Yersinia pestis, the etiological agent of the bubonic plague responsible for three major pandemics that historically impacted the human population, (e.g., the Justinian plague and the black death) (5,6). Yersinia pestis is still a public health problem today, mainly in Africa (7–9), but cases are commonly reported around the world, including in the United States, South America and Asia (6). In addition to Y. pestis, fleas can also harbor other pathogens including Rickettsia spp. and Bartonella spp. imposing a challenge for veterinary and public health (1).

It is speculated that hematophagous behavior evolved independently more than 20 times in arthropods, and to overcome host hemostasis and immune response, blood feeders have developed a complex salivary concoction that interferes with blood clotting, platelet aggregation, vasoconstriction and modulates host immune responses (10). Over the last 20 years, advances in sequencing technologies have allowed a deeper insight into the salivary gland contents (sialome) of several hematophagous vectors, constituting the so-called ‘sialoverse’ and highlighting how diversified these concoctions can be (11). Further biochemical and structural studies of salivary proteins led to the postulation that blood feeders’ saliva always contains at least one anticlotting, one antiplatelet and one vasodilatory component (11). Yet, the biological activity of flea saliva has been less studied than that of blood feeding Diptera (e.g., mosquitoes, sand flies, black flies), triatomines or ticks, and only a handful of molecules have been functionally characterized (12–14).

The first sialotranscriptome of X. cheopis salivary glands shed light on its unique composition, revealing a large expansion of acid phosphatases that lack their putative catalytic residue and FS-H peptides that appear to be unique to fleas, in addition to the first CD39 apyrase identified in a hematophagous arthropod (15). Here we provide a deeper insight into X. cheopis salivary gland contents using new generation sequencing methods. We also performed a LC-MS/MS analysis of X, cheopis salivary gland peptides at 1-, 2-, or 3-days post feeding.

Materials and methods

Flea salivary gland dissection

Fleas used in this study were obtained from an establish colony maintained at the Rocky Mountain Laboratories. Adult female fleas were fed on neonatal mice at day 0 and intact salivary gland (SG) pairs were collected as previously described (15). SG (33 pairs) from fleas 1, 2- and 3-days post-feeding were pooled together in RNAlater solution (Invitrogen, USA) and used for RNA extraction. X. cheopis SG were also collected at the same timepoints and disrupted in 10 mM HEPES, 140 mM NaCl pH 7.4. These samples were centrifuged for 10 min at, 12,000 × g at 4°C and the supernatant was collected. The total protein concentration was determined with the PIERCE BCA Protein Kit assay (Thermo Fisher Scientific, USA). For each day after feeding 4 samples were obtained by pooling from 22 to 38 salivary glands pairs from adult female fleas. The use of mice for flea feeding was approved by the NIH Animal Care and Use Committee (ACUC protocol #2019-011).

RNA extraction, Illumina sequencing and analysis

RNA extraction of 99 pairs of X. cheopis salivary glands was performed with the RNAeasy RNA isolation kit (QIAGEN, USA) according to manufacturer`s specifications and analyzed with an Agilent 2100 Bioanalyzer (Agilent, USA). The library was constructed using the NEBNextUltraTM II Directional RNA Library Prep Kit and sequencing was performed in an Illumina Novaseq 6000 DNA sequencer (Illumina, USA). The fastq files were trimmed of the Illumina adapters and low-quality sequences (Q < 20) using TrimGalore (https://github.com/FelixKrueger/TrimGalore), merged into a single file and assembled using Abyss (16) with k values from 25 to 95 (with increments of 10) in single stranded mode and Trinity (17) in single stranded F mode. The assemblies from Abyss and Trinity were combined and filtered with the CD-HIT tool (18). Coding DNA sequences with open reading frames larger than 150 nt (CDS) were extracted based on BLASTp results to a subset of the non-redundant protein database and selected if fragments shared ≥70% similarity with a matching protein. Additionally, all open reading frames (ORF’s) starting with a methionine and having 40 aa in length were submitted to the signalP program (v. 3.0); those fragments having a signal peptide were mapped to the ORF’s, and the most 5’ methionine was selected as the starting methionine of the transcript coding for a putative secreted peptide (19). To assess the assembly quality, the BUSCO (4.1.3) benchmark for universal single-copy orthologs using the Arthropoda database (2020-09-10) was used (20). For annotation, we used an in-house program that scans a vocabulary of ~ 400 words and their order of appearance in the protein matches from BLASTp results, including their e-values and coverage. CDS quantification was performed by mapping the library reads to the extracted CDS using the RSEM tool (21), and the final annotated CDS was exported to a hyperlinked Excel spreadsheet.

Mass spectrometry analysis

X. cheopis salivary gland extract was prepared as follows: Each reaction contained 3 μg total protein in a final volume of 30 μl buffered with 50 mM HEPES pH 8.0. The protein was reduced with 5 mM DTT for 40 min at 37°C. The samples were cooled to room temperature and alkylated with 15 mM iodoacetamide for 20 min. Then, 200 ng of trypsin was added, and samples were incubated for 15 hours at 37°C. The pH was adjusted to approximately 2.5 with 10% TFA and samples were desalted and concentrated with Agilent OMIX10 tips (Agilent, USA). Peptides were eluted with 20 μl of 0.1% TFA/50% AcCN and dried under vacuum. The peptides were dissolved in 12 μl of 0.1% FA/3% AcCN and centrifuged at 18000×g for 5 minutes. The LC-MS experiment was performed using Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific, USA) coupled to EASY nLC 1200 nano-liquid chromatography system (Thermo Fisher Scientific, USA). Peptides were first bound to a PepMap C18 column (3 μm particle, 100 Å pore, 75 μm inner diameter, 2 cm length) then separated using an EASY-Spray analytical column (PepMap C18, 2 μm particle, 100 Å pore, 75 μm inner diameter, 25 cm length) using a linear gradient of 0 to 40% acetonitrile in water containing 0.1% formic acid for 100 min, followed by 40–80% for 5 min, 80% hold for 5 min, 80–0% for 5 min, and 0% hold for 5 min. Data were acquired with a standard data-dependent acquisition strategy, in which the survey MS1 scan was done at least every 3 sec with Orbitrap mass analyzer at 120,000 resolution, and the MS2 scans were done with a linear ion trap mass analyzer for multiply charged precursor ions isolated with a 1.6 m/z window using a quadrupole and fragmented by CID at 35% collision energy. The EASY-IC internal calibration was utilized for Orbitrap scans, and the dynamic exclusion period was set at 60 sec. Tandem mass spectra were analyzed using the PatternLab for proteomics 4.0 platform (22). A target-decoy database was generated using the deposited sequences from the Siphonaptera taxonomic group along with CDS obtained from the Illumina sequencing and searched with the Comet tool (23) implemented in PatternLab. The search space included all fully tryptic peptide candidates and carbamidomethylation of cysteine was used as a static modification. Data were searched with a 50 ppm precursor ion tolerance and a 0.4 Da fragment ion tolerance. The validity of the peptide spectrum matches (PSMs) generated by Comet was assessed using the Search Engine Processor (SEPro) module from PatternLab. A cutoff score was established to accept a protein false discovery rate (FDR) of 1% based on the number of decoys. Results were post-processed to only accept PSMs with < 10 ppm precursor mass error and proteins with a unique peptide. Normalized spectral abundance factors (NSAF) was used to represent relative abundance, and only proteins that were identified in at least 3 of the 4 biological replicates were considered.

Statistical analysis.

The Spearman correlation between proteomic and transcriptomic relative quantifications was performed using the Log2 (NSAF × 10⁴) and the Log2 FPKM. Heatmaps, scatterplot, and determination of the Spearman correlation coefficient (Rho) were performed using the R programming language (24), using the pheatmap (heatmap), ggplot2 (scatterplots) and stats (spearman correlation) packages. Phylogenetic trees were constructed using the maximum likelihood method (25) with MEGA X (26). The amino acid alignments were done using the clustal omega tool (27).

Data availability

The transcriptome data was deposited to the National Institute for Biotechnology Information (NCBI) under Bioproject PRJNA760518 and Biosample accession SAMN21216395. The raw reads were deposited to the Short Reads Archive of the NCBI under accession SRR15713648 and the CDS deposited to the Transcriptome Shotgun Assembly (TSA) under accession GJKK00000000. The raw proteomic data was deposit to the ProteomeXchange platform under accession number PXD028695.

Results and Discussion

Overall description of the sialotranscriptome

After primer removal and trimming of low-quality sequences, we obtained approximately 22 million reads. The de novo assembly generated 35,048 sequences from which 7,432 coding sequences (CDS) were extracted based on their homology with previously identified proteins or presence of signal peptide. The CDS were annotated and exported to a hyperlinked spreadsheet (Supplementary Table 1). Relative quantification of each CDS using the RSEM tool mapped around 63% of the reads which is in range with previous transcriptomes studies (28–30). The unmapped reads (~ 37%) could be from the 5` and 3` UTR of the CDS, non-coding RNA or any sequence that did not result in sufficiently large ORF’s (larger than 150 nt), or failed to produce matches to known proteins, or failed to code for a product with a signal peptide. Additionally, we also carried out the BUSCO analysis to evaluate our assembly quality. Using the Arthropoda database as reference resulted in a completeness of 77% (69.2 % single and 7.8% duplicate) with 5% of fragmented and 18% of missing sequences. For comparison, we conducted the same analysis using the C. felis genome assembly (31) and observed a completeness of 76.9% (76.9% single) with 12% of fragmented and 11.1% of missing sequences. These values are within range with other assemblies (32), reflecting no major bias in our assembly.

Functional classification of the 7,432 transcripts with a FPKM greater than zero revealed the presence of two major classes: a secreted class with 470 CDS accounting for 71.7% of all mapped reads and a protein synthesis machinery class with 455 CDS accounting for 15.1% of the mapped reads (Table 1). These data reinforce the role of salivary glands as a tissue specialized in producing secreted proteins. Further classification of the sequences belonging to the secreted class revealed that the transcripts of the acid phosphatases protein family are the most abundant in X. cheopis salivary glands (Table 2) with 27 CDS accounting for 63% of all mapped secreted reads. The second most represented group was the FS-H like peptides with 21 CDS accounting for 20% of the secreted reads, followed by proteins with unknown function. Fourteen additional protein families are shown in Table 2, which includes adenosine deaminases, apyrases, protease inhibitors (e.g., TIL, kunitz, serpins and pacifastins), antigen5-like proteins, mucins, odorant and hormone binding proteins.

Table 1:

Functional classification of coding sequences (CDS) from the X. cheopis sialotranscriptome.

Class	Number of Contigs	Number of FPKM	%
Secreted	470	1317205	71.737
Protein synthesis machinery	455	277339	15.104
Unknown	2787	120337	6.553
Signal transduction	713	21772	1.186
Transporters/storage	355	20764	1.131
Nuclear regulation	533	14736	0.803
Protein modification machinery	306	11930	0.650
Transcription machinery	395	7709	0.420
Metabolism, lipid	201	6749	0.368
Protein export machinery	145	5282	0.288
Metabolism, energy	70	4950	0.270
Cytoskeletal	128	4917	0.268
Proteasome machinery	197	4666	0.254
Metabolism, carbohydrate	162	4585	0.250
Oxidant metabolism/detoxification	69	2779	0.151
Transcription factor	91	2666	0.145
Metabolism, amino acid	40	2041	0.111
Metabolism, nucleotide	126	1987	0.108
Transposable element	61	1575	0.086
Immunity	29	781	0.043
Extracellular matrix/cell adhesion	57	736	0.040
Metabolism, intermediate	22	289	0.016
Storage	8	263	0.014
Nuclear export	12	108	0.006
Total	7432	1836166

Table 2:

Families of secreted proteins within the sialotranscriptome of X. cheopis.

Family	Number of CDS	FPKM	FPKM (%)
Phosphatase	27	834470.19	63.323
FS-H family	21	265563.1	20.152
Unknown	359	136233.5	10.338
Glu-rich	1	48529.58	3.683
Apyrase	10	12686.84	0.963
Inhibitors	16	9125.67	0.692
Antigen5-like	3	8748.39	0.664
adenosine-deaminase	2	5325.85	0.404
Gly-rich	3	3738.54	0.284
Mucin	3	2283.18	0.173
Odorant binding	2	1303.99	0.099
Hormone binding	5	108.27	0.008
Protease	11	84.15	0.006
Pro-rich	2	16.5	0.001
ECA-like	4	15.67	0.001
Lipocalin	1	8.79	0.0007
Cys-rich	1	1.52	0.0001
Total	471	1317793

Overall description of the sialoproteome

In addition to the transcriptome analysis of X. cheopis SG, we also conducted a proteome study from the SG of fleas 1, 2 or 3-days post-feeding using the CDS generated by the de novo assembly as reference. After data filtering, peptides from 610 transcripts were identified (Supplementary Table 2). The heatmap plot of the Log2(NSAF×10⁴) from the 610 proteins identified by LC-MS/MS revealed that, although each replicate clustered within their biological sample (day 1, 2 or 3), the overall protein profile of X. cheopis salivary glands is broadly conserved at the different time points (Figure 1).

Figure 1:

Heatmap plot of the Log2 NSAF values of the 610 peptides identified by mass spectrometry of salivary gland homogenates (SGH) of X. cheopis that correspond to transcripts identified by sequencing. The “R” value represents the biological replicate number.

In accordance with the transcriptome data, the secreted class was also the most represented class by the LC-MS/MS analysis, accounting for 45 – 48% of X. cheopis SG contents (Table 3). The second most abundant group was the unknown class (~14%) followed by the protein synthesis machinery class (10 – 12%). Sixty of the 610 identified transcripts belonged to the secreted class. As expected, acid phosphatases were the most abundant protein family in the flea SG at all time points (Figure 2) with 12 CDS accounting for approximately 55% of all secreted NSAF. The second most abundant family was the protease inhibitors group with 5 CDS representing around 16% followed by the FS-H like proteins with 12 CDS accounting for approximately 13% of all secreted NSAF. Peptides from transcripts with high homology to adenosine deaminase, apyrases, Glu and Gly rich proteins, hormone and odorant binding proteins were also observed.

Table 3:

Average normalized spectral abundance factor (NSAF) of the 610 proteins identified by LC-MS/MS of X. cheopis salivary gland homogenate by their functional classification.

	Day 1		Day 2		Day 3
Class	Avg. NSAF	%	Avg. NSAF	%	Avg. NSAF	%
Secreted	0.3959	45.89	0.4150	48.63	0.4149	45.52
Unknown	0.1251	14.50	0.1190	13.95	0.1294	14.20
Prot. Synthesis	0.1054	12.22	0.0906	10.61	0.1148	12.59
Prot. Modification	0.0335	3.88	0.0336	3.94	0.0377	4.13
Cytoskeletal	0.0294	3.41	0.0275	3.22	0.0364	4.00
Signal Transduction	0.0293	3.40	0.0254	2.98	0.0289	3.17
Detoxification	0.0228	2.64	0.0205	2.41	0.0234	2.56
Transporters and storage	0.0225	2.61	0.0218	2.56	0.0204	2.24
Met/Lip	0.0181	2.09	0.0168	1.97	0.0179	1.97
Transcription machinery	0.0157	1.81	0.0161	1.89	0.0180	1.97
Met/Carb	0.0156	1.80	0.0179	2.10	0.0177	1.94
Proteasome	0.0114	1.33	0.0125	1.46	0.0138	1.51
Prot. Export	0.0087	1.00	0.0078	0.91	0.0096	1.06
Met/AA	0.0079	0.92	0.0078	0.92	0.0078	0.86
Met/Int	0.0045	0.53	0.0063	0.74	0.0050	0.55
Met/energy	0.0042	0.48	0.0051	0.60	0.0046	0.51
Nuclear regulation	0.0039	0.46	0.0019	0.22	0.0043	0.47
Transcription factor	0.0034	0.39	0.0023	0.27	0.0018	0.20
Met/Nuc	0.0021	0.24	0.0026	0.30	0.0022	0.24
Extracellular matrix	0.0015	0.18	0.0016	0.19	0.0000	0.00
Transposable element	0.0013	0.15	0.0008	0.10	0.0012	0.14
Immunity	0.0006	0.07	0.0003	0.03	0.0015	0.17
Total	0.8627		0.8533		0.9116

Figure 2:

Heatmap plot of the Log₂NSAF values of the 60 proteins classified into the secreted class identified by mass spectrometry. Each column represents a biological replicate of X. cheopis SGH 1, 2- or 3-days post-feeding. CDS numbering and their functional class are shown. The gray squares represent proteins that were not identified in the sample.

A common question in transcriptomic studies is how the transcript relative quantification (i.e., FPKM, CPM or TPM) correlates to the actual protein concentration of a given sample. In the present study we observed a moderate and significant Spearman correlation between the Log₂ FPKM and the Log₂ of the average NSAF (Rho = 0.4961, p < 2.2×10⁻¹⁶) (Supplementary Figure 2). To further explore this correlation, we determined the Spearman correlation coefficient for each functional class. Statistically significant correlations were found for the detoxification (Rho = 0.775, p = 0.01), amino acid metabolism (Rho = 0.893, p = 0.007), protein modification (Rho = 0.288, p = 0.041), protein synthesis (Rho = 0.444, p = 5.36×10⁻⁶), secreted (Rho = 0.631, p = 2.3510⁻⁸ and unknown classes (Rho = 0.415, p = 1.83×10⁻⁸), (Figure 3 and Supplementary Figure 3). The high correlation observed for the detoxification and amino acid metabolism classes must be carefully interpreted, since both present a relatively small number of observations and accordingly, the correlation found could be biased. The protein modification class had a borderline significance value, which is in line with inspection of the scatter plot. Finally, moderate correlation was found for the unknown, protein synthesis and secreted classes (Rho values between 0.4 – 0.7). It is important to note that these classes are the three most represented ones, accounting for 93.3% of all mapped reads in the transcriptomic data (Table 1) and around 72% in the proteomic data (Table 2), suggesting that transcripts with high FPKM values tend to correlate to their NSAF counterpart. To inspect this hypothesis, we generated an in silico protein pattern considering the NSAF and FPKM of the 15 most abundant proteins identified in the proteome and transcriptome of X. cheopis salivary glands (Supplementary Figure 1A), in which a similar pattern was observed when comparing to the a protein electrophoresis from flea salivary glands (Supplementary Figure 1B).

Figure 3:

Scatter plots of the Log₂FPKM by the Log₂ of the average NSAF from the 610 transcript∷peptide matches, classified according to function. The Spearman correlation coeficient and the respective p-value is shown for classes with significant correlation (p < 0.05) and a linear model was fitted to the data.

In the following sections, we provide additional information regarding the major secreted salivary protein families found in this dataset.

An insight into the secreted salivary proteins of X. cheopis

Apyrases and adenosine deaminases

Apyrases are enzymes that hydrolyse ADP and ATP to AMP (33). They are classified into three subfamilies: the 5`-nucleotidase family, the cimex-type apyrases and the homologues of the human B cell antigen CD-39 (34). Apyrases are commonly found in the saliva of hematophagous vectors including ticks (35), kissing bugs (36), mosquitoes (37) and fleas (38,39). The previous transcriptome of X. cheopis SG reported the first CD-39 apyrase from a hematophagous arthropod. Additionally, the presence of a divalent cation-dependent apyrase activity was shown (15), indicating the presence of apyrases in the flea saliva. In the current dataset, in addition to the apyrase belonging to the CD-39 sub-family, we also identified several CDSs from the 5`-nucleotidase family with high levels of FPKM. Interestingly, only six CDS were identified by LC-MS/MS analysis, from which three 5`-nucleotidase family members were observed only at 3-days post-feeding. Two other 5`-nucleotidases and the CD-39 apyrase were found at moderate levels of NSAF in all time points (Figure 2). In addition to the apyrase sequences, we also identified two adenosine deaminase-like (ADA) proteins in the transcriptome and proteome of X. cheopis SG. Comparison of the NSAF values of the ADA-like proteins at different post-feeding days revealed a trend of increasing expression over time (Figure 2) suggesting that these proteins achieve their highest concentrations prior to the feeding period. The high abundance of both apyrases and adenosine deaminase-like proteins in the flea SG indicates the presence of high activity towards the complete degradation of ATP also proposed for other blood-feeding arthropods (40,41).

CAP superfamily

Composed of Cysteine-rich secretory proteins, Antigen 5, and Pathogenesis-related 1 proteins, the CAP family is widely distributed in different organisms (42). Members of the Antigen 5 subfamily are better described in hornets and wasps as they are the major allergens of vespid venom (43,44) but are frequently reported in sialome studies of hematophagous arthropods including ticks (28), sand fly (45), mosquitoes (46), and fleas (15,47). In the current data set we report 3 putative antigen 5-like transcripts, including one from the previous sialome of X. Cheopis, with a high range of FPKM levels (3.91 – 8056.82). Additionally, two of the transcripts were also identified by LC-MS/MS with low values of NSAF in almost all biological samples (Figure 2). Although commonly found in hematophagous vectors, the role of antigen 5 molecules is mostly unknown with only a few molecules functionally characterized. In the horsefly Tabanus yao, antigen 5-like proteins isolated from the fly salivary gland were shown to inhibit thrombosis, angiogenesis, and platelet aggregation (48–50). In Dipetalogaster maxima and Triatoma infestans, antigen 5-like proteins can act as an antioxidant and disrupt platelet aggregation induced by low concentrations of collagen (51), suggesting a role in blood acquisition.

FS-H-like peptides

Sometimes referred to as FS-I peptides, members of this family contain eight cysteine residues including a CxC motif at the C-terminus and one was originally identified as a potential antigen from C. felis saliva (52). Interestingly, FS-H peptides are not found in the sialomes of other vectors (e.g., mosquitoes, flies, kissing bugs, ticks) and appear to be exclusive to fleas (10). In the previous study of the X. cheopis sialome, 15 FS-H peptides were discovered (15) and despite their overall low homology, the conserved cysteine sequence motif is also found in the scorpion channel blockers and other insect defensins (53,54). Here we report a total of 20 CDS (Figure 4A), including 10 from the previous study. Members of this family presented a high range of FPKM values (7.61 – 38769.55), and the LC-MS/MS analysis identified 12 of them with variable values of NSAF (Figure 2).

Figure 4:

(A): Amino acid alignment of mature FS-H-like proteins identified in the X. cheopis sialome. The conserved cysteine residues are black-boxed. (B): Phylogenetic tree of the FS-H-like peptides constructed using Maximum likelihood. The number at the base of the branches indicates the concordance between 500 bootstrap replicates. Proteins previously identified are shown with their respective accession number and proteins identified by mass spectrometry analysis are shown in red.

FS50 was the first FS-H peptide to be functionally and structurally characterized in fleas, revealing the classic core βαββ structure found in the scorpion β toxins. Additionally, it was also shown that FS50 inhibits the Na_V1.5 sodium channel (13). Recently, the inhibitory action of a second FS-H peptide from X. cheopis, FS48, on K_V1.3 channels was described, and its anti-inflammatory effects were demonstrated in a mouse model (14). As both FS50 and FS48 lack the key residues responsible for the channel blocking activity, as described for the scorpion toxins, their channel blocking mechanism is not completely understood.

Phylogenetic analysis of all FS-H sequences from X. cheopis salivary glands resulted in the formation of 3 major clades (Figure 4B), with clade I containing 10 of the 25 CDS, including the potassium channel blocker FS48, clade II with 7 CDS and clade III with the remaining 8 CDS including FS50. The low bootstrap values observed in the base of each clade in addition to the low homology observed among members of this family suggests that the FS-H family is under a rapid expansion process generating peptides with different kinetic properties. The high abundance and variability observed within the FS-H family could ensure that different populations of channels are immediately blocked upon flea bite, potentially blocking host nociception (55) and immune signaling (56) that could be detrimental for the flea feeding success.

Protease inhibitors

Transcripts coding for serine and cysteine protease inhibitors are commonly found in the SG of hematophagous arthropods and usually are associated with modulation of hemostasis and host immune response (57). Here we identified serine peptidase inhibitors members of the Kazal, Kunitz, serpin, TIL (trypsin-like inhibitor) and pacifastin families in addition to a TIMP (tissue inhibitor of metalloproteinase) (Supplementary Table 2). With exception of two sequences, members of these classes presented low FPKM values (1.26 – 72.39). Mass spectrometry analysis identified 5 of the 16 reported transcripts belonging to the Kazal, serpin and pacifastin families in addition to two similar sequences previously identified and named XC-42 (ABM55431.1) and XC-43 (ABM55432.1) (15), which we recently characterized as thrombin inibitors (see below). Interestingly, the serpin and the pacifastin-like inhibitors showed a delayed expression pattern and were only observed at 3 days post-feeding with low levels of NSAF (Figure 2). In contrast, both XC-42 and XC-43 presented the highest NSAF levels during all time points.

In general, an efficient mechanism to inhibit blood coagulation is to target factor Xa or thrombin, highlighting their crucial role in the coagulation cascade. Accordingly, inhibitors of those enzymes have been widely identified in blood-feeding arthropods (58–60). We recently demonstrated that XC-42 and XC-43 are tight-binding thrombin inhibitors that interact with the catalytic and fibrinogen-binding sites of thrombin (61). Moreover, the crystal structure of the thrombin-XC43 complex revealed an intact XC-43 suggesting that these inhibitors are resistant to cleavage, a process commonly observed in competitive thrombin inhibitors (62,63). Based on their relative high expression in the flea SG, it seems that these molecules are the primary coagulation inhibitors deployed by X. cheopis. Interestingly, BLAST of XC-42 or XC-43 against the C. felis genome assembly (31) fails to return significant hits. Although there are no reports regarding the ability of C. felis saliva to interfere with host coagulation, it is reasonable to assume that at least one inhibitor is present. Together, these results suggest the presence of an unknown anticoagulant in cat flea saliva.

Odorant-binding proteins (OBP)

OBPs are small proteins with a hydrophobic binding pocket commonly found in the sensory organs and salivary glands of arthropods. In the sensory organs OBPs have a key role in insect chemoreception, contributing to the insect’s ability to locate food, hosts and mating partners (64). In the present study, we identified two transcripts with similarities to the C. felis odorant-binding protein 19d-like and 56a-like (Supplementary Table 1). Both sequences were also found in the LC-MS/MS analysis, in which CDS 4301 was found in higher amounts at all time points (Figure 2). Amino acid alignment of XC-4301 with other odorant-binding proteins from insects (Figure 5) revealed the presence of the conserved six cysteine residues found in classic OBPs, suggesting the presence of a putative binding pocket. Salivary OBPs are better described in mosquitoes, where they encode for a multi-gene family named D7 that are among the most abundant salivary proteins. From the functional perspective D7 proteins are described as kratagonists, molecules that can bind and, therefore, limit the availability of several signalling molecules relevant for host haemostasis, including serotonin, histamine, epinephrine, norepinephrine, and leukotrienes (e.g., LTB₄, LTC₄ and TXA₂) (65). Since moderate concentrations of XC-4301 were found at all time points in X. cheopis saliva, it`s possible that this protein also acts as a kratagonist. Further biochemical experiments are currently under way in order to better understand its role in flea biology.

Figure 5:

Amino acid alignment of mature XC-4301 and other insect odorant-biding proteins XP_026482252.1 from C. felis, EDS44686.1 from Culex quinquefasciatus, XP_039433538.1 from Culex pipiens pallens, XP_031634712.1 from Contarinia nasturtii, XP_035776018.1 from Anopheles albimanus, SAJ59035.1 from Triatoma brasiliensis, NP_001298183.1 from Stomoxys calcitrans and XP_001966896.1 from Drosophila ananassae. Conserved residues among the sequences are black-boxed.

The histidine acid phosphatase superfamily

Members of this superfamily contain a conserved His residue that is phosphorylated during the hydrolysis of phosphate monoesters (66) and are currently classified in two branches (67) corresponding to PFAM families PF00300 (branch 1) and PF00328 (branch 2). In the previous sialotranscriptome of X. cheopis SG, 10 CDS classified into the branch 2 were reported (15). Interestingly, all previously reported sequences displayed loss of the conserved catalytic His residue suggesting that catalytic activity is absent. In the current dataset we identified 25 CDS as members of this family including 6 CDS previously identified. The 19 new CDS were also classified into branch 2. In addition to novel sequences lacking the catalytic His residue, we identified four CDS containing the conserved RHG domain (Figure 6). Two of the putative catalytic CDS had low FPKM values (4.47 and 28.45) and were not observed in the mass spectrometry analysis. On the other hand, CDS 3334 and 213 presented moderate levels of FPKM (4200.83 and 4826.9, respectively) and both were also identified in the LC-MS analysis (Figure 2). Acid phosphatases were previously identified in the salivary glands of several blood feeding vectors including ticks (68,69), triatomines (70) and mosquitoes (71). Yet their role in the vector biology remains unclear.

Figure 6:

Amino acid alignment of human phosphatase (NP_001090.2) with selected X. cheopis phosphatases identified by de novo assembly. Proteins identified in the mass spectrometry analysis are shown in red. Residues found in the active center of the human enzyme are boxed in blue. Conserved residues are black-boxed and similar residues (70%) are gray-boxed.

Overall, the acid phosphatase CDS lacking the catalytic His residues presented high FPKM values, including the highest found (356,794.05 for CDS 6146), comprising the most abundant family of proteins in X. cheopis salivary glands. It was speculated that loss of the enzymatic activity might result in the phosphatase permanently bound to its substrate, limiting its availability (15) and potentially interfering with hemostasis (72,73). An insight into the sialoverse revealed a trend in which the most abundant salivary proteins in hematophagous vectors usually act as kratagonists, binding to small agonists such biogenic amines, nucleotides or eicosanoids (74). This is true for the D7 family of proteins in mosquitoes (65), lipocalins in ticks and kissing bugs (75,76) and the yellow protein in sand flies (77). Therefore, as these known kratagonist families are absent in the X. cheopis salivary glands and considering the high abundance and diversity of flea salivary acidic phosphatases missing their catalytic residues, it`s tempting to suggest that these pseudo-phosphatases are also kratagonists. Regardless of their role in flea biology, repeated exposure of mice to X. cheopis bites results in a host serum capable of recognizing the flea phosphatases, however no protective phenotype was observed (78). A similar study using C. felis, in which phosphatase-like sequences were also reported (31,47), resulted in a comparable immunoblot pattern (52). Together these studies illustrate that, although the acid phosphatases accounts for over half of the total protein mass of X. cheopis SG, they appear to be poor candidates for flea immune control.

Conclusion

The transcriptomic and proteomic analyses of X. cheopis salivary glands homogenates revealed its unique composition, highlighting the different pathways taken during the evolution of salivary gland supporting the hematophagic behavior. When compared to the sialome of other blood feeding vectors, specially to mosquitoes and ticks, the flea salivary gland contents present a lower complexity, with fewer protein families and the expansion of two major groups, the acid phosphatases lacking their putative catalytic residue and FS-H-like peptides that have not been identified in the salivary glands of other blood feeding vectors. Additionally, comparison between transcriptomic and proteomic quantification data indicates that high transcript expression level correlates with the actual protein concentration in the sample. Together, in the absence of X. cheopis genome sequences, the data presented here serves as an extended reference for the identification of pharmacological active proteins present in the flea saliva.

Supplementary Material

1

Supplementary figure 1: (A): In silico gel of X. cheopis SG using the average NSAF of the 15 most abundant proteins at each day and their respective FPKM value. (B): NuPAGE (4–12%) of X. cheopis SG homogenates (7.9 μg) from 1, 2 or 3 days post-feeding.

Supplementary figure 2: Scatter plot of the Log₂ FPKM by the Log₂ of the average NSAF from the identified proteins by mass spectrometry. The Spearman correlation coeficient (Rho = 0.4386 p < 2.2×10⁻¹⁶) shows a significant statistical correlation between the two relative quantification methods.

Supplementary Figure 3: Scatter plots of the Log₂ FPKM by the Log₂ of the average NSAF from 610 transcripts∷peptide matches, classified according to function. The Spearman correlation coeficient and the respective p-value is shown for classes with non-significant correlations (p > 0.05).

Significance.

The rat flea Xenopsylla cheopis is the main vector of Yersinia pestis, the etiological agent of the bubonic plague responsible for three major pandemics that marked human history and remains a burden to human health. In addition to Y. pestis fleas can also transmit other medically relevant pathogens including Rickettsia spp. and Bartonella spp. The studies of salivary proteins from other hematophagous vectors highlighted the importance of such molecules for blood acquisition and pathogen transmission. However, despite the historical and clinical importance of X. cheopis little is known regarding their salivary gland contents and potential activities. Here we provide a comprehensive analysis of X. cheopis salivary composition using next generation sequencing methods paired with LC-MS/MS analysis, revealing its unique composition compared to the sialomes of other blood-feeding arthropods, and highlighting the different pathways taken during the evolution of salivary gland concoctions. In the absence of the X. cheopis genome sequence, this work serves as an extended reference for the identification of potential pharmacological proteins and peptides present in flea saliva.

Highlights.

Over 7,000 transcripts were identified in the salivary glands of Xenopsylla cheopis.

Approximately 600 proteins were identified by mass spectrometry.

Transcriptome and proteome data indicates that the acid phosphatase is the most abundant salivary protein of X. cheopis.

Transcripts with high values of FPKM presented statistically significant correlation with their NSAF counterpart.