Background

Insects are associated with a plethora of microorganisms partitioned in different tissues of the host. Particularly, these bacterial symbionts can be distributed in the gut, hemolymph and even in the reproductive organs of the insect host [1]. An example of multi-tissue colonizer bacteria are some members of Spiroplasma.

The genus Spiroplasma comprises wall-less, motile, and helical bacteria of the class Mollicutes. These bacteria are mainly associated with insects but there are also reports of Spiroplasma in arachnids, crustaceans and plants [2, 3]. In insects, Spiroplasma can be vertically transmitted from females to offspring [4]. Transmission to plants involves an insect vector, as has been observed with the plant pathogen S. citri which is transferred to plants by leafhoppers and psyllids [5]. Spiroplasma associated with insects consist of mutualists, commensals, male-killing reproductive parasites and pathogens [6]. There are a few reports of pathogenic spiroplasmas in insects, like S. apis and S. melliferum, which produce lethal infections in honeybees [7, 8]. Nonetheless, most reported spiroplasmas are beneficial and may be considered facultative symbionts [3]. In addition to lethal pathogenicity, two other phenotypes are induced by spiroplasmas in insects such as protection against parasites (wasps and nematodes) and male-killing. Both phenotypes are produced by Spiroplasma poulsonii in Drosophila [9,10,11]. Protection against parasites has been associated with the presence of genes encoding ribosomal inactivating proteins (RIPs) in S. poulsonii genome. These RIPs are toxins that damage the ribosomes of Drosophila parasitic wasps and nematodes [12, 13]. Likewise, a plasmid-encoded protein (Spaid) seems to be involved in D. melanogaster male-killing phenotype produced by S. poulsonii [11, 14].

Molecular and phylogenetic classification using either the 16S rRNA or single-copy gene markers have split Spiroplasma spp. into four major clades. Three clades are composed of the formally described Spiroplasma: Citri-Chrysopicola-Mirum (CPM), Apis, and Ixodetis. The remaining clade (Mycoides-Entomoplasmataceae) contains species from the genera Mycoplasma, Mesoplasma and Entomoplasma, that have lost the helical cell morphology (thus they are not named Spiroplasma) [15]. Spiroplasma genome size ranges between 0.7 to 2.2 megabase pairs (Mbp). To date, 32 genomes of Spiroplasma spp. are available. However, only two belong to the Ixodetis clade, one is from the parasitic wasp Cephus cinctus [16] and the other is from the African monarch butterfly Danaus chrysippus [17], which makes this clade the least represented at the genomic level, in spite of being present in many insects [18].

Up to now, there are few functional genomic studies of Spiroplasma, two of them used quantitative reverse transcriptase PCR (qRT-PCR) to analyze the change of expression in key genes of S. citri [19, 20]. Additionally, there are two total transcriptional studies (RNAseq) of spiroplasmas, one analyzed the differences in culture growth of S. diminutum and S. taiwanense isolated from mosquitos [21], and other described the transcriptional profile of S. poulsonii in Drosophila (hemolymph) and in culture medium [22]. Differential gene expression was evidenced in all studies under contrasting conditions. Nonetheless, to the best of our knowledge, there is no transcriptome analysis of spiroplasmas colonizing different insect host tissues and the genes involved in the symbiotic interaction. Moreover, there is no transcriptional study of S. ixodetis.

Dactylopius (Hemiptera: Coccoidea: Dactylopiidae) is a scale insect that feeds on the sap of prickly pear, mainly from the genus Opuntia [23]. Dactylopius is the main source of carminic acid, a red dye used in cosmetics, drugs, food, and textile industry, achieving economical relevance [24]. The genus Dactylopius comprises more than 10 species, but only D. coccus is used for the extraction of carminic acid. This species was probably domesticated in Mexico more than one thousand years ago, selected for increased production and quality of the pigment compared to the other ‘wild’ species (e.g., D. opuntiae) [23, 25, 26]. Previous reports using metagenomic, metatranscriptomic and culture approaches showed diverse bacteria and fungi in both the domesticated and wild carmine cochineals [25, 27,28,29,30]. Within the microbial community present in Dactylopius, we reported a nitrogen-fixing β-proteobacterium Candidatus Dactylopiibacterium carminicum [30, 31] as well as two different strains of Wolbachia (wDacA and wDacB) [29], and fungi [32]. As part of the metagenome and metatranscriptome surveys from D. coccus, we also reported the presence of Spiroplasma sequences [31]. However, there is no study of the phylogenetic classification, genomic capabilities, and the putative roles of Spiroplasma bacterium within Dactylopius. Here we present the analysis of three Spiroplasma ixodetis metagenome-assembly genomes (MAGs) independently recovered from the domesticated D. coccus females and males, of two different populations each, as well as from a population of the wild cochineal D. opuntiae (females) metagenomes. Additionally, we used the previously reported metatranscriptome data from the gut, hemolymph, and ovary of D. coccus [31], to analyze the expression of genes putatively involved in the symbiotic interaction between S. ixodetis and Dactylopius. Genomic and metatranscriptomic analyses here presented, suggest that S. ixodetis may use the type IV secretion system (T4SS) to interact with the carmine cochineal. Moreover, the incompleteness of genes encoding virulence factors indicates that S. ixodetis could be considered as a non-pathogenic symbiont.

Results and discussion

New Spiroplasma ixodetis symbiont is present in multiple Dactylopius spp. metagenomic samples

Mollicute-related MAGs were recovered in metagenomic assembled and binned samples from adult males and females of the domesticated carmine cochineal D. coccus from two different populations, as well as in the metagenome from adult females of the wild cochineal species D. opuntiae. Analysis of single-copy gene markers showed that these MAGs exhibited high completeness (> 92%) and no apparent contamination (Table 1). Mollicute MAGs of D. opuntiae females, D. coccus females and males were placed into 258, 286 and 353 scaffolds, respectively (Table 1). The estimated genome size of these MAGs ranged from 1.32 to 1.9 Mbp. Around 1215 to 1371 coding sequences (CDS) were identified within these genomes (Table 1). Like in other Mollicutes [33], the Mollicutes-related MAGs from the Dactylopius spp. had low G-C % content (~ 24%, Table 1). The complete sequences of the 16S rRNA gene were obtained for all three MAGs (Table 1). Phylogenetic reconstruction of these 16S rRNA sequences placed the Dactylopius Mollicute-related MAGs within the Spiroplasma genus in the Ixodetis clade. (Fig. 1a). Additionally, 16S rRNA sequence from D. coccus and D. opuntiae MAGs showed 99% nucleotide identity with the S. ixodetis Y32 16S rRNA sequence from the western black-legged ticks (Ixodes pacificus) and to other S. ixodetis (Additional file 1 Fig. S1). We further refer to Dactylopius associated Spiroplasma as S. ixodetis DO (from D. opuntiae, female), S. ixodetis DCF (from D.coccus, female) and S. ixodetis DCM (from D. coccus male), respectively.

Table 1 General genomic features of Spiroplasma ixodetis DO, DCF and DCM compared to othersinsect-associated Spiroplasma
Full size table
Fig. 1
figure 1

Phylogenetic position of Spiroplasma symbionts associated with Dactylopius cochineals. Maximum-likelihood trees of 16S rRNA genes (a) and 168 concatenated single-copy gene markers (b). Scale bars indicate 10 and 20% estimated sequence divergence, respectively. See Additional file 2 Data set 1 for the accession numbers of the sequences used in these analyses

Full size image

To further classify and compare the S. ixodetis from Dactylopius cochineals with other spiroplasmas, we performed a pan-genome analysis using 30 public Spiroplasma genomes from the Apis, Ixodetis, Chrysopicola, Poulsonii, Citri and Mirum phylogenetic clades (Additional file 2 Data set 1). The pan-genomic analysis showed 11,045 gene clusters and 168 of them corresponded to single copy core-genes within the Spiroplasma pan-genome. Robust phylogenomic analysis using the 168 single-copy core genes confirmed the position of spiroplasmas from Dactylopius within the Ixodetis clade (Fig. 1b). Average amino acid identity (AAI), using the same 168 core single-copy genes, showed that S. ixodetis DO, DCM and DCF shared identities of ~ 99.5% with the genomes of Spiroplasma ixodetis symbiont of the wheat stem sawfly (WSS) Cephus cinctus [16] and ~ 98.5 with the Spiroplasma endosymbiont of the African monarch butterfly Danaus chrysippus [17] (Additional file 2 Data set 1). AAI analysis also showed that S. ixodetis in both female and male D. coccus populations were 100% identical (Additional file 2 Data set 1). Moreover, S. ixodetis of D. opuntia showed a 99.78% AAI value in comparison to S. ixodetis of both D. coccus populations (Additional file 2 Data set 1). This result suggests that there are slight variations in the bacterial genomes between the two Dactylopius species (i.e., D. coccus and D. opuntiae), although there is no apparent difference between spiroplasmas of the two populations (males and females) of D. coccus. Similar variations in core gene identity (ANI ~ 99.5) have been reported between the sister species strains of S. poulsonii (sMel and sHy), symbionts of Drosophila melanogaster and D. hydei, respectively, and it has been linked to a host adaptation process [34]. Likewise, genomic differences between the S. ixodetis strains of Dactylopius may result from adaptation to different host species (i.e., D. opuntiae and D. coccus).

To find out if S. ixodetis was present in other Dactylopius, we further analyzed the metagenomes of D. coccus (females), from commercial samples of Mexico and Peru, previously reported by Campana et al., [35]. No complete S. ixodetis MAGs were recovered after assembly and binning analyses of these D. coccus metagenomes. Nonetheless, we were able to identify long (> 900 base-pairs [bp]) and highly similar (BLASTn identity > 90% and e-value < 0.0005) contigs to those from the S. ixodetis DCF genome in both the Mexican (n = 56) and Peruvian (n = 20) metagenomic assemblies. The above confirms that S. ixodetis is distributed not only in different species of Dactylopius, but also in different populations of D. coccus. This indicates that this bacterium can form a seemingly symbiotic relationship with the carmine cochineal and may not be a merely sporadic association. As in nature, frequencies of Spiroplasma spp. in their insect host are variable [36, 37], further studies are required to elucidate the prevalence of S. ixodetis in other Dactylopius species/populations.

Reduced number of genes for amino acid and carbon metabolism were found in S. ixodetis DO, DCM and DCF

To gain information about the general metabolic profiles of S. ixodetis and their role in the interaction with Dactylopius, a comprehensive comparative genomic analysis was performed between S. ixodetis DO, DCM and DCF and other Spiroplasma (n = 30) genomes recovered from diverse environments (i.e., vertebrates, plants, and other insects). Analysis of the clusters of orthologous groups of proteins (COG) showed that S. ixodetis DO, DCM and DCF have fewer genes associated with transport and metabolism of amino acids compared to other Spiroplasma spp. genomes (Fig. 2). Spiroplasma species are auxotrophs for most of the essential amino acids and require multiple transporters to obtain them from the host [38]. A smaller number of genes coding for ABC transporters of peptides and oligopeptides as well as fewer genes involved in the biosynthesis of amino acids were detected in S. ixodetis genomes from Dactylopius in comparison with other Spiroplasma spp. (Fig. 2 and Additional file 3 Data set 2). Most of the Spiroplasma obtain ATP through arginine metabolism [39, 40]. However, and contrasting with the previous S. ixodetis WSS genome [16], no arginine biosynthetic genes were found in S. ixodetis of Dactylopius (Additional file 3 Data set 2).

Fig. 2
figure 2

Comparative genomics of Spiroplasma associated with Dactylopius and other Spiroplasma. COG profiles of S. ixodetis DCF, DCM and DO (blue bars) compared to other Spiroplasma genomes (red bars). Blue arrows indicate enrichment of genes in the COG categories on S. ixodetis DCF, DCM and DO genomes. The mean ± SEM proportion of genes belonging to each COG category is shown

Full size image

A complete set of genes for glycolysis, fructose catabolism and the pentose-phosphate pathway were found in S. ixodetis DO, DCM and DCF genomes (Additional file 3 Data set 2). Trehalose, glucose and mannose are abundant sugars in insect hemolymph and most of the insect-associated spiroplasmas encode genes involved in the phosphotransferase system (PTS) to transport these sugars into the bacterial cell [38, 40, 41]. Accordingly, S. ixodetis DO, DCM and DCF showed genes encoding glucose (ptsG), fructose (fruA/B) and N-acetyl-glucosamine (nagE) PTS transporters (Additional file 3 Data set 2). Even though S. ixodetis DO, DCM and DCF display genes for maltose and cellobiose PTS systems, no other genes for oligosaccharides catabolism were found in these genomes (Additional file 3 Data set 2). Particularly, no genes coding for oligosaccharide breakdown and catabolism (e.g., glycoside hydrolases) were found in DO, DCM and DCF genomes (Additional file 3 Data set 2) suggesting that S. ixodetis is unable to utilize complex polysaccharides as a carbon source.

In Spiroplasma species that are insect pathogens, such as S. citri, S. apis and S. mellipherum, genes for trehalose utilization, including the PTS, are present [40,41,42]. However, similar to S. ixodetis of Dactylopius, genes for trehalose catabolism are absent or found as non-functional pseudogenes in the non-pathogenic species S. poulsonii, S. chrysopicola and S. syrphidicola [40, 43]. The putative inability to metabolize trehalose by S. ixodetis DO, DCF and DCM may limit spiroplasma growth in Dactylopius tissues as has been previously suggested in the Drosophila-S. poulsonii interaction [40].

S. ixodetis DO, DCM and DCF genomes have few virulence factors-encoding genes

Incomplete genes encoding toxin-like proteins and other putative virulence factors were found in the S. ixodetis DO, DCM and DCF genomes (Table 2). Ankyrin repeat domains are present in many virulence effector proteins [44]. Particularly, the ankyrin-repeat containing protein Spaid of S. poulsonii (locus SMSRO_SFP00290) contributes to the male-killing phenotype in D. melanogaster [11]. Incomplete homologs (> 50% identity, > 75% BLASTp coverage) of spaid gene were found in DO, DCM and DCF genomes (Table 2). Similarly, spaid homologs were reported in S. ixodetis WSS genome [16]. However, spaid sequences of S. ixodetis WSS lack the N-terminal signal peptide domains present in S. poulsonii Spaid protein and thus were classified as not functional proteins [16]. In addition to spaid, partial coding sequence, homologs to the ribosome-inactivating protein (RIP2) gene of S. poulsonii were found in S. ixodetis from the domesticated D. coccus but not from the wild species D. opuntiae (Table 2). RIP proteins of other spiroplasmas have been implicated in protection against nematodes and parasitic wasp in different species of Drosophila [45, 46]. Even though genomic results suggest S. ixodetis DO, DCM and DCF encode multiple toxin-like factors, most of these are incomplete or annotated as putative pseudogenes.

Table 2 Virulence factor-encoding genes present in S. poulsonii and homologous genes in Dactylopius-associated S. ixodetis
Full size table

In S. poulsonii and S. citri spiralin protein, encoded by spiA, spiB and spiC, acts as a lectin anchor for binding to glycoproteins present in the insect cells and is required for efficient colonization of the host [39, 47, 48]. No homologous gene of spiralin spiA, B or C was detected in S. ixodetis DO, DCM and DCF genomes (Additional file 3 Data set 2). Nonetheless, homologous genes encoding the adhesins SARP1 and SARP3 from S. poulsonii and S. citri were found in S. ixodetis DO and DCM genomes (Table 2). Adhesins like SARP1 and SARP3 as well as phosphoglycerate kinase (PGK) are involved in cell adherence and invasion of S. citri to the leafhoppers Circulifer tenellus and Circulifer haematoceps [49, 50]. All three S. ixodetis DO, DCM and DCF genomes encode PGK (Additional file 3 Data set 2) which in S. citri is a key factor for bacterial internalization into the host cells [51, 52].

Glycerol catabolic and transporter genes were found in S. ixodetis DO, DCM and DCF genomes (Additional file 3 Data set 2), particularly, the genes glpO in addition to glpF (membrane channel) and glpK (kinase) (Fig. 3). Synteny analysis showed similar genomic architecture of glpO, glpF and glpK genes in S. ixodetis DO, DCM and DCF genomes and in S. poulsonii¸ S. melliferum and S. culicicola (Fig. 3), which agrees with previous reports of glp gene organization in other spiroplasmas [53]. Glycerol metabolism may lead to the production of hydrogen peroxide (H2O2) by L-α-glycerophosphate oxidase (GlpO) [3, 54]. In the human pathogen, Mycoplasma mycoides subsp. mycoides, H2O2 has been linked to tissue inflammation and cellular damage [55, 56]. Thus, in insect associated pathogenic Spiroplasma spp. (i.e., S. taiwanense, S. culicicola and S. apis) the presence of glpO coding gene has been considered to be a virulence factor [57]. However, unlike other spiroplasmas, glpO and glpK genes in S. ixodetis DO, DCM, and DCF were incomplete and annotated as pseudogenes suggesting that they are unable to produce H2O2 from glycerol (Fig. 3).

Fig. 3
figure 3

Gene structure of the glycerol catabolic genes (glpF, glpO and glpK) present in different Spiroplasma spp. genomes. Color-empty blocks represent pseudogenes. Gray blocks show genes annotated as hypothetical (hyp) protein. Maximum-likelihood phylogenetic tree in the left shows the cladogenesis of Spiroplasma spp. genomes using the Glycerol-3-phosphate oxidase encoding gene glpO. MAFFT was used to align all glpO sequences from each genome and the phylogenetic tree, based on the LG + G4 substitution model obtained by ModelFinder, was calculated by IQtree with 1000 Bootstrap replicates for internal branch support. Scale bar indicates 5% estimated sequence divergence

Full size image

glpO, glpF and glpK genes are commonly recovered from pathogenic Spiroplasma like S. culicicola and S. taiwanense. In those bacteria n-glycerol 3-phosphate (G3P) is taken up by the transporters UgpA, UgpC and UgpE [57]. Even though no ugpA/C/E homologous genes were found in S. ixodetis genomes (Additional file 3 Data set 2), a glpT gene, encoding the G3P-transporter, was found in the DO, DCM and DCF genomes, but not present in other Spiroplasma genomes of insects (Fig. 3). Comparative analysis showed that this gene forms an orthologous cluster with the glpT of S. floricola, Mycoplasma yeatsii and Mycoplasma putrefaciens. This suggests that in S. ixodetis the G3P might be transported using a GlpT transporter instead of the UgpA/C/E system which is associated with pathogenicity.

Altogether, the absence of genes for trehalose catabolism, the pseudogenization of virulence factors (i.e., spaid and glpO) as well as the presence of genes for insect cell colonization (i.e., the adhesins SARP1, SARP3 and PKG), suggest S. ixodetis is adapted to live as a non-pathogenic symbiont inside Dactylopius.

Comparative genomics revealed an enrichment of genes encoding the type IV secretion system (T4SS) in the Dactylopius associated S. ixodetis in comparison to other Spiroplasma spp. genomes

COG profile comparison between DO, DCM and DCF and other Spiroplasma spp. revealed a greater representation of the intracellular trafficking/secretion category in DO, DCM and DCF (Fig. 2). Secretion of macromolecular effectors (i.e., protein and nucleic acids) plays a central role in modulating the interactions between symbiotic (pathogenic and mutualistic) bacteria and their host [58, 59]. Similar to other spiroplasmas associated with insects (e.g., S. poulsonii) [40], S. ixodetis DO, DCM and DCF showed genes encoding the general secretion (Sec) system. Particularly, the genes secA, secY (missing in DCF), secG and secE were found in S. ixodetis from Dactylopius spp. (Fig. 4a). Other Sec coding genes such as the signal recognition GTPase (ffh and ftsY) and the translocase (yidC) were also present in S. ixodetis DO, DCM and DCF genomes (Additional file 3 Data set 2) suggesting DO, DCM and DCF use the Sec system to export proteins.

Fig. 4
figure 4

Differences in genes encoding secretion systems among Spiroplasma spp. genomes. Heatmaps showing the number of genes associated with (a) Intracellular trafficking, secretion, and vesicular transport COG category. (b) Heatmap showing the number of genes associated with the type IV secretion system (T4SS) among all the Spiroplasma spp. genomes. Colors at the top of the heatmaps indicate the Spiroplasma phylogenetic position of each genome

Full size image

Additionally, S. ixodetis DO, DCM and DCF genomes showed more virB4-D4 predicted genes, associated with the type IV secretion system (T4SS), than other Spiroplasma species associated with insects (Fig. 4a). A phylogeny using the VirB4 ATPase clustered the virB4 sequences of S. ixodetis DO, DCM, and DCF with other virB4 of Spiroplasma spp. (Additional file 1 Fig. S2), discarding a putative chimeric-assembly origin of these genes. Additionally, multiple virB4-D4 genes were found in scaffolds encoding plasmid-like coding sequences (i.e., soj and plasmid replication protein; Additional file 1 Fig. S3 and Additional file 3 Data set S2) in the S. ixodetis DO genome. Differences in coverage of genomic contigs have been used as a proxy to detect plasmids in bacteria [60] and plasmid-like scaffolds with virB4-D4 sequences of S. ixodetis DO showed a higher sequence coverage (> ~ 5000x) than the coverage for other scaffolds in the DO genome (Additional file 1 Fig. S3). Similarly, genes encoding VirB4 T4SS ATPase were detected in plasmids of S. kunkelii and S. citri, pSKU146 and pSci1–6, respectively [61, 62]. A maximum-likelihood phylogenetic analysis showed that the sequence of S. ixodetis DO virB4 (DO_DO_00343) cluster together with the virB4 genes from plasmids pSKU146 and pSci1–6 (Additional file 1 Fig. S4). All this evidence suggests that like in S. citri and S. kunkelii the T4SS of S. ixodetis DO is plasmid borne. Although the DCM and DCF genomes have copies of genes soj/parA, smc/spo0J and the replication protein COG5521, involved in plasmid segregation [63,64,65], no virB4-D4 genes were found in the plasmid-like scaffolds of these bacteria (Additional file 3 Data set 2).

Annotation using the TXSSdb database (see methods) revealed that other T4SS elements are present in S. ixodetis DO, DCM and DCF genomes such as genes coding for the type IV coupling protein (t4cp), the MOB relaxase and tra genes (Fig. 4b). These elements are typically used by bacteria in the conjugation process [66]. Previously, the presence of mob and tra coding genes in S. citri plasmid pSKU146 suggested that spiroplasmas are conjugative [62]. Having similar elements in DO, DCM and DCF indicates that these bacteria might be conjugative. Not all the components for the T4SS were found in S. ixodetis DO, DCM and DCF genomes (Fig. 4a-b), however, it has been suggested that Gram-positive bacteria, and specifically the cell wall lacking members of the Mollicutes, do not require as many T4SS components to secrete macromolecules as the Gram-negative bacteria [61]. Thus, S. ixodetis DO, DCM and DCF may have a functional T4SS. In Wolbachia, perhaps the most abundant insect intracellular-symbiont, multiple T4SS loci encoding Vir proteins were expressed during all stages of the host development from embryogenesis to adult male and female Drosophila flies [59]. Moreover, it has been considered that Wolbachia uses the T4SS to secret as many as 105 proteins with putative effector function to the host (Drosophila) cells [53]. Like Wolbachia and other symbiotic bacteria (e.g., Mesorhizobium loti) [67], T4SS could be used by S. ixodetis to interact with the host.

S. ixodetis DCF differential gene expression in the gut, hemolymph, and ovary tissues of D. coccus

To detect genes seemingly involved in the interaction between S. ixodetis and Dactylopius, we compared the transcriptome of S. ixodetis DCF recovered from previously reported metatranscriptomic data of D. coccus gut, hemolymph, and ovaries [31]. After ribosomal sequences depletion, ca. 1 million metatranscriptomic reads mapped to S. ixodetis DCF genome in each of the D. coccus gut, hemolymph, and ovary metatranscriptomic samples (Additional file 1 Fig. S5a). Principal component analysis (PCA) of S. ixodetis DCF transcripts showed that the transcriptomes differed along with the ordination space by intrasample variation (y-axis 21% of variance) and tissue specificity (i.e., gut, hemolymph, and ovary; x-axis 30% of variance; Additional file 1 Fig. S5b). The above suggests that S. ixodetis DCF modulate gene expression depending on the allocation in the different host tissues. Pairwise comparisons between each tissue combination (i.e., gut vs. hemolymph, gut vs. ovary and hemolymph vs. ovary) identified 69 unique differentially expressed genes of S. ixodetis DCF across all comparisons and 46% matched with a coding product in the bacterial genomic annotation (Fig. 5a and Additional file 3 Data set 2). Among these genes, we detected that the conjugal transfer protein coding gene traG, as well as the virB4 and virD4 genes of the T4SS of S. ixodetis DCF, were up-regulated in the hemolymph and the ovaries in comparison to the gut (Fig. 5b). Additionally, other S. ixodetis DCF T4SS related encoding genes such as the type IV pilus inner membrane component pilM, the type four conjugation tfc, the relaxase mob and the virB3 were also expressed, but not differentially (p-value > 0.05), in the D. coccus gut, hemolymph, and ovary (Additional file 3 Data set 2). Expression of vir genes, particularly virB4, encoding for the T4SS of Wolbachia has been previously detected in the ovaries of the wasp Asobara tabida suggesting that T4SS is active and could participate in the Wolbachia-A. tabida interaction. It has been suggested that Wolbachia might use T4SS to export effectors for manipulating the wasp cell biology and oogenesis [59, 68]. Similarly, high expression of virB4 and virD4 genes in D. coccus ovary suggests that the T4SS of S. ixodetis DCF is biologically active. An elevated number of T4SS coding genes in the S. ixodetis DCF genome and the up-regulation of T4SS genes in the hemolymph and ovaries of D. coccus (Fig. 5a-b and Additional file 3 Data set 2), indicate that the T4SS might play an important role in the interaction between S. ixodetis DCF and the host.

Fig. 5
figure 5

Variation on gene expression of S. ixodetis DCF in the gut, ovary, and hemolymph of D. coccus. (a) Heatmap showing all the differentially expressed genes in each tissue comparison (i.e. gut vs hemolymph, gut vs ovary and hemolymph vs ovary). Color range in heatmaps indicates variation in log2 fold-change. Gray boxes indicate no differential expression observed after DESeq2 analysis (absolute log2 fold-change ≥0.58 and p-adjust value ≤0.05). hyp means genes annotated as hypothetical proteins. (b) Volcano plot of differential gene expression of S. ixodetis DCF in gut vs hemolymph, gut vs ovary and hemolymph vs ovary comparisons. Red dots show genes considered as differentially expressed after DESeq2 analysis absolute log2fold change > 0.58 (outer broken vertical lines) and p-adjust value ≤0.05 (−log10 p-value ≥1.3, broken horizontal line). Green dots show non-differentially expressed genes with an absolute log2 fold-change > 0.58 and p-adjust value > 0.05. Gray dots show non-differentially expressed genes with an absolute log2fold change < 0.58 and p-adjust value > 0.05 (non-differentially expressed). All annotated genes are represented, red dots without annotation legend represent hypothetical protein coding genes differentially expressed

Full size image

Metatranscriptome analysis also showed that S. ixodetis DCF genes encoding the Spaid and GlpO proteins, described as virulence factors, have low average levels of expression (≤ 42 fragments per kilobase of transcript per million mapped reads [FPKM]) and were not differentially expressed in any of the D. coccus tissues analyzed. Moreover, no expression of S. ixodetis RIP encoding gene was detected in any of the D. coccus metatranscriptomes (Additional file 3 Data set 2). In S. poulsonii, genes encoding the RIP and GlpO proteins were up-regulated in D. melanogaster hemolymph in contrast to in-vitro cultures, indicating an important role of these genes in the fly-S. poulsonii interaction [22]. The low expression values of these spiroplasma genes in all the Dactylopius tissues, as well as the fact that these genes are incomplete, suggests that genes encoding virulence factors might not have an important role for the D. coccus-S. ixodetis interaction.

A high expression of S. ixodetis DCF genes associated with reactive oxygen species (ROS) catabolism such as the sodA, encoding the superoxide dismutase, was observed in the D. coccus ovaries relative to the gut and hemolymph. Additionally, genes pdhC and pdhD, encoding the dihydrolipoyl dehydrogenase, were up-regulated in the hemolymph relative to the gut (Fig. 5a-b and Additional file 3 Data set 2). In the intracellular bacteria Mycobacterium tuberculosis the expression of pdhC and pdhD has been associated with resistance to host-reactive nitrogen intermediates [69]. ROS and reactive nitrogen species (RONS) play a critical role in the establishing, maintenance, and success of the symbiosis [70]. As other eukaryotes, insects control bacterial symbionts load in their tissues by producing RONS. Nonetheless, some bacteria, mainly mutualistic bacteria, can cope with the burst of host RONS by displaying an arsenal of antioxidant enzymes (e.g., superoxide-dismutase and peroxidase) [70]. For example, in the mutualistic association between the bobtail squid Euprymna scolopes and the luminous bacteria, Vibrio fisheri, bacteria can overcome the squid RONS by over expressing a peculiar NO/oxygen-binding (H-NOX) protein to catabolize the nitric oxide produced by the host. This allows V. fisheri to successfully colonize the highly oxidative light-organ tissue of E. scolopes [71]. Likewise, Wolbachia, possess a single sod gene that likely plays a fundamental role in the management of oxidative stress produced by insects’ cells [72]. In many insects, Spiroplasma reside as extracellular bacteria in the gut and the hemolymph, however, in other tissues such as the fat body and ovaries these bacteria are found as intracellular symbionts [4, 38, 73]. The up-regulation of S. ixodetis DCF sod gene in the ovaries compared to the gut indicates that like Wolbachia, Spiroplasma could use the superoxide-dismutase to manage the oxidative stress produced inside the insect ovarian cells. This strategy might allow S. ixodetis DCF to colonize the embryonic cells and be vertically transmitted to the Dactylopius offspring.

Conclusions

Microbial associations with insects are considered key to their biological success. In many cases, there are multiple microbes within a single insect, and that is the case of Dactylopius spp. which is associated with the diazotroph bacterium Candidatus Dactylopiibacterium carminicum and two Wolbachia strains. Here, we described that Spiroplasma ixodetis is also a member of the bacterial community of both the wild (D. opuntiae) and the domesticated (D. coccus) carmine cochineals. The presence of S. ixodetis in three independent metagenomic samples of different populations of Dactylopius from two species (D. coccus and D. opuntiae) as well as in the metagenomes of other two different D. coccus metagenomes from Peru and Mexico, indicates that S. ixodetis is commonly found within Dactylopius. However, since these metagenome studies were performed with pools of insects, we cannot conclude that each individual cochineal has Spiroplasma. In Drosophila not all individuals of same population are infected with Spiroplasma and the same phenomenon should be further explored in the carmine cochineals. The genome analysis showed that females and males from different populations of the domesticated cochineals carry the same S. ixodetis, with slight differences from that of the wild cochineals, suggesting a common origin of the symbiosis.

The Dactylopius associated S. ixodetis genomes showed pseudogenization or incompleteness of genes commonly coding for virulence factors in other Spiroplasma spp. As no complete virulence genes were detected, S. ixodetis in Dactylopius could be considered a non-pathogenic symbiont. Transcriptomic analyses of endosymbionts in insects are scarce. Here the transcriptomic analysis of S. ixodetis DCF showed that genes involved in the T4SS (e.g., traG, virB4 and virD4) were over expressed in hemolymph and ovary of the host D. coccus. Moreover, genes involved in oxygen-stress tolerance such as sodA were also over expressed in the ovary. These results suggest that a) S. ixodetis might use T4SS to interact with Dactylopius cells, and b) these bacteria can tolerate oxygen-stress produced inside the host cells. Finally, transcriptional differences of S. ixodetis among Dactylopius tissues suggest that this bacterium can respond and adapt to the different conditions present within the host tissues indicating that there is a strong interaction and molecular signaling in the symbiosis between S. ixodetis and the carmine cochineal Dactylopius.

Methods

Insect sampling

Females of Dactylopius coccus were obtained from Campo Carmín Company, Morelos, Mexico (18°44′46.7″N, 99°11′17.8″W) in April 2012. Males from a different population of D. coccus were obtained from a commercial farm in Tepoztlán, Morelos, Mexico (18°59′24.8″N 99°07′02.9″W) in February 2017. Females of wild cochineal D. opuntiae were collected from a cactus farm field in Morelos, Mexico (18°59′25.3″N 99°07′02.6″W) in May 2017. Insects were obtained from Opuntia spp. cactus and were transported together with their host plants to the laboratory for further experiments.

DNA extraction and metagenomic sequencing

Shotgun metagenomes were previously reported and obtained [29, 32] using different sequencing technologies (i.e., 454-GS-FLx, Illumina HiSeq2000–2500 and PacBio) of total DNA from guts, Malpighian tubules, ovaries and hemolymph of Dactylopius coccus females. Details of the insect dissection and DNA extraction procedures from different Dactylopius tissues are fully described in [29, 30, 32]. Briefly, for 454 sequencing, 20 D. coccus adult females were superficially disinfected with 70% ethanol, rinsed with sterile distilled water, and dissected with sterile forceps to remove the exoskeleton and guts. Cells in the hemolymph and debris were separated by centrifugation in a Percoll gradient as is described by [74]. Percoll phases were observed under a microscope, and those with cells were selected for DNA extraction. For Illumina HiSeq2000 sequencing guts, ovaries, and Malpighian tubules from 40 females were dissected using sterile forceps under a stereoscopic microscope. We tried to avoid tissue cross-contamination by rinsing the tissues several times with sterile phosphate-buffered saline to remove remnant hemolymph. These organs were pooled, suspended in sterile PBS, and macerated using a sterile plastic pestle. For PacBio sequencing, eight individuals were superficially disinfected with 70% ethanol as above. Guts and exoskeleton were removed with sterile forceps and hemolymph from all individuals was pooled for DNA extraction. In addition, metagenome sequences of wild Dactylopius opuntiae females and males of different D. coccus population were obtained by Illumina HiSeq2500 platform. For these last metagenomes, two independent pools containing the whole body of 10 D. opuntiae females and 30 D. coccus males were superficially disinfected as described above. Insects were macerated using sterile pestles and resuspended in sterile PBS. Independently of the sequence technology, DNA from all insects was extracted with the DNeasy Blood & Tissue Kit (QIAGEN) following the manufacturer’s instructions. Sequencing was performed at Macrogen Inc. (Korea) for 454 and Illumina HiSeq2000–2500. PacBio sequencing was performed at Duke University Genome Sequencing Core Facility (USA). In all cases, after base calling high-quality metagenomic reads (quality Phred score > Q30) were obtained by adaptor removal and trimming using the TrimmGalore 0.4.4 pipeline (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) with the --paired --q 30 options.

Binning, genome assembly and annotation

For genomic binning all metagenomic reads from D. coccus females (i.e., 454, Illumina and PacBio), D. coccus males and D. opuntiae females were independently mapped using Bowtie2 2.3.5 [75] versus a concatenated index of Wolbachia wDacA (LSYX00000000.1), Wolbachia wDacB (LSYY00000000.1) and Candidatus Dactylopiibacterium carminicum (NZ_MQNN00000000.1) genomes, previously reported bacteria present in Dactylopius metagenomes [29, 32], with the --end-to-end and --very-sensitive options. All non-mapped reads from each metagenome (D. coccus female, D. coccus male and D. opuntiae, respectively) were retrieved from the bam files using Samtools 1.7 [76] with the following options: samtools fastq -@ 20 -f 2. To maintain read parity after samtools procedure, all fastq files from each metagenome were analyzed by TrimmGalore 0.4.4 pipeline with the --paired option, this resulted in high-quality paired reads for all libraries. These reads, from each Dactylopius species and populations, were then independently assembled using idba-ud 11.1 [77] with the default parameters. The resulting assembled genomic contigs were binned into single metagenomic assembled genomes (MAGs) by MaxBin 2.2.1 [78] with the following parameter -min_contig_length 900 for minimum contig length. In each binning the metagenomic contigs were divided into two different MAGs and classified by checkM 1.1.2 pipeline [79] with the lineage_wf option for a broad taxonomical classification. In the three metagenomes analyzed, checkM detected the presence of a Mollicute related MAG in addition to an insect-related MAG. To improve the assembly, the original high-quality reads were mapped to the MAG classified as Mollicutes lineage in each metagenome using Bowtie2 with the --end-to-end and --very-sensitive options. All paired and mapped reads were once again retrieved using samtools as above and reassembled using idba-ud. Assembled contigs were compared to the previous MAG obtained by MaxBin2 and if any improvement on the total assembly length, N50 and N90 statistics was observed, this new assembly was considered further, if no improvement was observed the original MaxBin2 Mollicute MAG was used instead. Final genome assemblies were then polished by SSPACE [80] with bwa as aligner and Ciclator 1.4.1 [81] was used to find possible circular scaffolds. After the ‘best’ assembly was obtained, manual curation of the contigs in each MAG was done by performing a BLASTn [82] search against the ‘nr’ database (downloaded on September 2016) and manually removing those contigs matching to insect or any other no Mollicutes bacteria (> 90% identity, <1e-5 e-value, > 80% coverage). Protein translated sequences of the three Mollicutes MAGs, one for each metagenome, were retrieved by Prodigal 2.6.3 [83]. The quality (completeness, contamination, and strain heterogeneity) of this manual curated Mollicutes MAG was then assessed by the checkM [79] pipeline using the protein predicted sequences from Prodigal and the following parameters: checkm lineage_wf --reduced_tree --genes -t 8 -x faa. Total genomic features prediction and open reading frame (ORFs) annotation on each MAGs were performed using PROKKA v1.11 [84] with the following parameters: --gcode 4 --metagenome --rfam. Metabolic pathways were predicted using the GhostKoala tool from KEGG [85] and manually curated using the Bioconductor R package KEGGREST 1.26.1 and the KEEG_parser tool developed for this work (https://github.com/avera1988/KEGG_parser). The dbCAN2 pipeline [86] was used to detect genes coding for putative carbohydrate-active enzymes (CAZymes) in the Mollicutes MAGs using the dbCAN-database 7 (downloaded in January 2019). Additionally, all Mollicutes MAGs were annotated using the NCBI Prokaryotic Genome Annotation Pipeline PGAP-4.6 [87] for public availability. Correlations between PROKKA, KEGG, CAZy and PGAP locus identifiers of each MAGs are shown in the Additional file 3 Data Set S2.

Phylogenetic and comparative genomic analysis

Near-full-length (~ 1400-nucleotide [nt]) 16S rRNA gene sequences were retrieved from the annotations of the Mollicutes MAGs and compared to the NCBI “16S ribosomal RNA sequences” (accessed in February 2020) and Arb-SILVA (accessed in January 2020) databases using BLASTn to identify the nearest neighbors. MAFFT 7.3.10 [88] was used to align all sequences and a maximum-likelihood-based (ML) phylogenetic tree, based on the TVMe+R4 substitution model obtained by ModelFinder [89], was calculated by IQtree 1.6.12 [90] with 1000 Bootstrap replicates for internal branch support.

A total of 30 genomes (Additional file 2 Data set 1) across all the different Spiroplasma phylogenetic clades (i.e., Citri-Chrysopicola, Mirum, Apis, and Ixodetis) were downloaded from the NCBI “assembly” database (accessed in January 2020) and used for phylogenomic and comparative analyses. Pan and core genome analyses were conducted using GETHOMOLOGUES 2.0 [91], and OrthoMCL [92] was used for orthologue clustering (parameters: -A -c -t 0 -M -n 12). Single-copy genes from the core-genome of all Spiroplasma analyzed were parsed from the GETHOMOLOGUES pan-genome matrix results using custom Bash and Perl scripts (deposited in https://github.com/avera1988/Comparative_genomics).

For phylogenomics, a concatenated protein-based maximum-likelihood (ML) phylogenetic tree was constructed using the protein predicted sequences of 168 single-copy genes of Spiroplasma core genome. Protein sequences were concatenated and aligned using the multigenome2blocks pipeline [93]. ML phylogeny, based on the LG + F + R4 amino acid substitution model obtained by ModelFinder [89], was constructed using IQtree 1.6.12 [90] with 1000 Bootstrap replicates for internal branch support. Additionally, these 168 protein-coding genes were used to calculate the average amino acid identity (AAI) between Spiroplasma genomes using the AAI calculator from the “enveomics” collection tools [94]. AAI distance matrices were calculated and visualized by custom Perl and R scripts (deposited in https://github.com/avera1988/Comparative_genomics).

Finally, to compare functional profiles between Dactylopius associated and public Spiroplasma, all clusters of orthologous groups (COG) of proteins from each genome (Additional file 2 Data set 1) were annotated using the cdd2cog pipeline (https://github.com/aleimba/bac-genomics-scripts). Comparisons between proportions of COG genes from Dactylopius and non-Dactylopius associated Spiroplasma were performed using custom R scripts (deposited in https://github.com/avera1988/COG_differential_analysis). The number of genes on each COG category between Dactylopius and non-Dactylopius-associated Spiroplasma genomes were visualized by heatmaps in R 3.6.1.

Presence of S. ixodetis in other D. coccus metagenomes from Mexico and Peru

Previously reported metagenomes of D. coccus collected from Mexico and Peru by Campana et al., [35] were used to detect the presence of S. ixodetis in different Dactylopius samples. Metagenomic reads were obtained by the fastq dump tool (https://github.com/ncbi/sra-tools) using the SRR1231828 and SRR1231831 accession numbers from the sequence-read archive (SRA) NCBI portal (https://www.ncbi.nlm.nih.gov/sra). Paired-reads of each metagenome (i.e., Mexico and Peru) were independently assembled using the idba-ud 11.1 [77] assembler with the default parameters. The assemblies resulted in 8085 (22,546,409 bp, N50 192,783 bp) and 10,617 (25,132,088 bp, N50 99,725 bp) contigs for the Mexican and Peruvian metagenomes, respectively. The resulting assembled metagenomic contigs were binned into MAGs by MaxBin 2.2.1 [78] with the following parameter -min_contig_length 900 for minimum contig length. The resulting MAGs were then classified by CheckM 1.1.2 pipeline [79] with the lineage_wf option for a broad taxonomical classification. In the case that no Spiroplasma-like MAGs were recovered, BLASTn searches were performed using as a database the index of S. ixodetis DCF genome and as query the total assembled contigs for each Mexican and Peruvian metagenomes with the following parameters: blastn -query contigs.fa -db DCF.fasta -max_target_seqs 1 -num_threads 8 -outfmt 6. Metagenomic contigs from both Mexican and Peruvian samples with an identity value > 90% and e-value < 0.0005 in the BLASTn searches were further classified as S. ixodetis-like contigs. To corroborate these contigs were taxonomically close to Spiroplasma spp. all S. ixodetis-like contigs recovered in each metagenome were annotated by BLASTx searches against the Uniref100 protein database (downloaded November 2020). Those contigs with a Spiroplasma match in the database were then scored as S. ixodetis contigs.

Genes encoding components of the type IV secretion system (T4SS) detection

Protein sequence predictions from each Spiroplasma genome were compared to the type IV secretion system (T4SS) HMM profiles in the TXSSscan database [95] using the “hmmscan” (parameters: hmmscan –cpu 40 –domtblout) tool from HMMER 3.1b2 [96] to identify genes encoding domains that comprised potential components of T4SS. Putative T4SS component domains were manually parsed from hmmscan result tables and compared with previous PROKKA-PGAP annotations of each Spiroplasma genome.

RNASeq and metatranscriptomic analysis

Metatranscriptomic data, previously obtained and reported by Bustamante-Brito et al., [31] from hemolymph, ovaries, and gut of female D. coccus cochineals, was used to analyze changes in the expression of Spiroplasma genes in different tissues of Dactylopius. Tissue dissection and RNA extraction procedures are fully described in [31]. Briefly, 30-s instar nymphs of D. coccus females were collected and superficially rinsed with ethanol 90% to remove the covering wax. Hemolymph was collected, pooled, and resuspended in 200 μl RNAlater (ThermoFisher) by doing a dorsal puncture in each insect with a 1 ml sterile syringe. After bleeding, all individuals were dissected under sterile conditions and the whole gut (including foregut, midgut, hindgut, and Malpighian tubules) and ovaries were independently collected, pooled, and resuspended in 200 μl of RNAlater (ThermoFisher). As for DNA extraction (see above), we tried to avoid tissue cross-contamination by rinsing the tissues several times with sterile phosphate-buffered saline. Total RNA was obtained from each pool of tissues (i.e., gut, hemolymph, and ovary) using the RNEasy kit (Qiagen) following the modification reported by Guerrero-Castro et al., [97]. High-quality RNA samples were used for cDNA strand-specific library preparation by the TruSeq Stranded Total RNA kit (Illumina), and ribosomal RNA (rRNA) present in the samples was removed with the RiboZero Removal kit for Bacteria (Illumina). Three biological replicates of each tissue were obtained, thus nine metatranscriptomic libraries were generated and sequenced using a single lane on an Illumina HiSeq4000 sequencer at Macrogen Korea, by a 100 bp read length pair-end protocol. Purity-filtered reads were adapter and quality (Phred score Q > 30) trimmed by TrimmGalore 0.4.4 pipeline with the --paired --q 30 options. Reads matching to rRNA sequences in each library were identified by Metaxa2 2.2–1 pipeline [98] using the -x T option and manually removed from the fastq files. To maintain read parity, all fastq files obtained after rRNA removal were subjected to a second round of filtering by TrimmGalore 0.4.4 with the --paired option and considered further.

Differential gene expression analysis and quantification

To extract all the reads of Spiroplasma from the metatranscriptomes, high-quality filtered reads of each library (gut, hemolymph, and ovary) were independently aligned against the Spiroplasma ixodetis DCF genome using Hisat2 2.1.0 [99]. Properly mapped reads were obtained using samtools 1.7 with the following parameters: samtools fastq -@ 12 -c 6 -F 4 -N. The number of Spiroplasma read counts per gene locus in each library (tissue) was then obtained and summarized by the RSEM 1.3.1 pipeline [100] with the following parameters: rsem-calculate-expression -p 12 --paired-end --bowtie2 --estimate-rspd --append-names --output-genome-bam using the S. ixodetis DCF genomic annotation. Total matrix count with all abundances form each tissue replicate was obtained with the abundance_estimates_to_matrix.pl script from Trinity 2.4.0 [101] and parsed by the tximport Bioconductor package [102] in R. For differential expression analysis, treatments were classified depending on the RNAseq library as gut, hemolymph, and ovary. The software DESeq2 [103] was used to detect differentially expressed genes using as contrast each library/tissue comparison (i.e., gut vs hemolymph, gut vs ovary and hemolymph vs ovary). Genes were considered differentially expressed if the adjusted p-value (Benjamini-Hochber [BH] multiple test correction [104]) was less than or equal to 0.05 and an absolute fold-change was above 1.5 (absolute log2 fold-change ≥0.58) per each gene in any of the particular contrast comparison analyzed. Evaluation of main differences in expression per library (i.e., tissue) was assessed by a principal component analysis (PCA). PCA was performed and plotted by plotPCA function from DESeq2 with the variance stabilizing transformation (vst) method. Volcano plots and heatmaps were used to display and visualize all differential expressed genes of S. ixodetis DCF in the different libraries using the phamp and EnhancedVolcano (https://github.com/kevinblighe/EnhancedVolcano) packages in R. Fragments per kilobase of transcript per million mapped reads (FPKM) values of S. ixodetis DCF genes in each transcriptomic sample were obtained by the “fpkm” function of the DESeq2 package in R. All scripts to reproduce the gene expression analysis are deposited in Github (https://github.com/avera1988/Spiroplasma_ixodetis_Dactylopius_RNAseq).