Endosymbiont diversity and prevalence in herbivorous spider mite populations in South-Western Europe

Abstract

Bacterial endosymbionts are known as important players of the evolutionary ecology of their hosts. However, their distribution, prevalence and diversity are still largely unexplored. To this aim, we investigated infections by the most common bacterial reproductive manipulators in herbivorous spider mites of South-Western Europe. Across 16 populations belonging to three Tetranychus species, Wolbachia was the most prevalent (ca. 61%), followed by Cardinium (12%–15%), while only few individuals were infected by Rickettsia (0.9%–3%), and none carried Arsenophonus or Spiroplasma. These endosymbionts are here reported for the first time in Tetranychus evansi and Tetranychus ludeni, and showed variable infection frequencies between and within species, with several cases of coinfections. Moreover, Cardinium was more prevalent in Wolbachia-infected individuals, which suggests facilitation between these symbionts. Finally, sequence comparisons revealed no variation of the Wolbachia wsp and Rickettsia gtlA genes, but some diversity of the Cardinium 16S rRNA, both between and within populations of the three mite species. Some of the Cardinium sequences identified belonged to distantly-related clades, and the lack of association between these sequences and spider mite mitotypes suggests repeated host switching of Cardinium. Overall, our results reveal a complex community of symbionts in this system, opening the path for future studies.

endosymbiosis, maternally inherited bacteria, multiple infections, reproductive parasitism, two-spotted spider-mite, Tetranychus urticae

INTRODUCTION

Terrestrial arthropods often harbor several bacterial microorganisms that occur in the cytoplasm of host cells, and are maternally inherited from mother to offspring through the egg cytoplasm (Moran, McCutcheon and Nakabachi 2008). In the last decade, intensive field surveys of distinct arthropod species revealed that the most common of these endosymbionts are Wolbachia, Rickettsia, Cardinium, Arsenophonus and Spiroplasma (Duron et al.2008; Weinert et al.2015). Whereas Wolbachia is present in most groups, Rickettsia and Cardinium have low incidence in the hexapoda relative to other groups, with the exception of hemiptera (Weeks, Velten and Stouthamer 2003; Zchori-Fein and Perlman 2004; Weinert et al.2015). These three bacterial genera are also extremely prevalent in chelicerates, especially in spiders, mites and ticks (Duron et al.2008; Weinert et al.2015; Zhang et al.2016a; Duron et al.2017).

The maternal inheritance of endosymbiotic bacteria has selected for a variety of phenotypic changes in their hosts, such as cytoplasmic incompatibility (CI, i.e. embryonic mortality in crosses between infected males and uninfected females), feminization (i.e. genetic males that develop as females), thelytokous parthenogenesis (i.e. asexual daughter development), or male killing (i.e. infected males are eliminated during embryogenesis or late larval instars). These reproductive manipulations, by promoting the production and fitness of infected daughters (i.e. the transmitting sex) via negative effects on the fitness of individuals not involved in the transmission, enhance symbiont transmission and allow their spread within host populations (Werren, Baldo and Clark 2008; Engelstadter and Hurst 2009). Moreover, such manipulations may also facilitate the transmission of other vertically transmitted symbionts co-occurring in the same host (Frank 1998; Engelstadter, Telschow and Hammerstein 2004; Vautrin et al.2008; Vautrin and Vavre 2009; Hellard et al.2015).

Although traditionally viewed as vertically transmitted symbionts, evidence of horizontal transmission events stems from incongruences between host and symbiont phylogenies (Schilthuizen and Stouthamer 1997; Vavre et al.1999; Raychoudhury et al.2009; Gerth, Rothe and Bleidorn 2013). Indeed, a few recent studies have provided direct evidence for horizontal transmission, and suggest parasitism and feeding as the two main routes of transfer. For instance, Wolbachia can be transmitted between whiteflies via parasitoids (Ahmed et al.2015), or to a parasitoid from its infected drosophila host (Heath et al.1999), Spiroplasma may be transmitted between two Drosophila species via ectoparasitic mites (Jaenike et al.2007; Brown and Lloyd 2015), and Rickettsia infecting Bemisia tabaci may be transferred to parasitoids, although it is not vertically transmitted in the latter (Chiel et al.2009). Similarly, Trichogramma wasps of the same or different species have exchanged symbionts when infecting the same butterfly egg (Huigens et al.2004). Symbionts may also be transmitted via the resource that their hosts feed on (but see Faria, Paulo and Sucena 2016). The most compelling evidence for this infection route comes from herbivorous arthropods, in which horizontal transmission via plant feeding has been shown for Rickettsia and Wolbachia in whiteflies (Caspi-Fluger et al.2012; Li et al.2017) and for Cardinium in leaf hoppers (Gonella et al.2015). However, the incidence of horizontal transmission for many vertically transmitted arthropod endosymbionts is still under debate and other studies do not corroborate these results (e.g. Gualtieri et al.2017).

Several studies have shown that herbivorous spider mites are often infected with Wolbachia and Cardinium (Gotoh, Noda and Hong 2003; Liu, Miao and Hong 2006; Gotoh et al.2007b; Xie, Chen and Hong 2011; Ros et al.2012; Suh et al.2015; Zhang et al.2016b) and occasionally with Spiroplasma (Zhang et al.2016b; Staudacher et al.2017) and Rickettsia (Hoy and Jeyaprakash 2005; Zhang et al.2016b). In spider mites, Wolbachia and Cardinium may have diverse fitness effects. In particular, they may cause different degrees of CI (Gotoh, Noda and Hong 2003; Gotoh, Noda and Ito 2007a; Gotoh et al.2007b; Ros and Breeuwer 2009; Xie, Chen and Hong 2011; Zhu et al.2012; Suh et al.2015; Xie et al.2016). Moreover, these bacteria have been found in the same individual host (Ros et al.2012; Zhu et al.2012; Zhao, Zhang and Hong 2013a; Zhao et al.2013b; Xie et al.2016), and multiple strains of these symbionts can be found in spider-mite populations, although levels of nucleotide diversity depend on the host species and/or populations (Gotoh, Noda and Hong 2003; Liu, Miao and Hong 2006; Gotoh, Noda and Ito 2007a; Gotoh et al.2007b; Yu et al.2011; Ros et al.2012; Zhang et al.2016b).

This study aims to contribute to the knowledge of the symbiont community of spider mites, by characterizing the community of herbivorous spider-mite populations and their endosymbionts in South-West Europe, a relatively unstudied area. By unraveling the diversity and prevalence of endosymbionts in natural spider-mite populations, this study might provide some insight on potential facilitation in multiple infections and horizontal transfers among hosts.

MATERIAL AND METHODS

Spider-mite collection and population rearing

Populations of Tetranychidae mites were collected from September to December 2013 at several locations in Central Portugal (Table 1). In addition, two populations were collected from Southern Spain, and one population from Southern France for comparison with nearby countries. Each population in each location was collected from different leaves on different plants. Mite collection was done by placing infested leaves in closed paper bags. Subsequently, adult females (n = 25–600) from all populations were transferred to insect-proof cages containing bean cv. Contender seedlings (obtained from Germisem, Oliveira do Hospital, Portugal) immediately after collection from the field. After species identification for each of these populations, Solanaceae specialist spider mites (Tetranychus evansi) were transferred to tomato cv. Money Maker seedlings (obtained from Mr. Fothergill's Seeds, Kentford, UK). The colonies were then maintained in the laboratory under standard conditions (25 ± 2°C, 60% RH, 16/8 h L/D; Clemente et al.2016).

Table 1.

Tetranychidae mite populations collected in Portugal (P), France (F) and Spain (S) from September to December 2013.

Name	Species	Collection date	Location	Coordinates	Host plant
FR	T. urticae*	11/10/2013	Montpellier (F)	43.614951, 3.859846	Solanum lycopersicum
AlBe		09/11/2013	Almería (S)	36.855725, −2.320374	Solanum melongena
AlRo		09/11/2013	Almería (S)	36.855725, −2.320374	Rosa spp.
DC		10/09/2013	S. Domingos (P)	39.058742, −9.135427	Cucurbita pepo
DF		10/09/2013	S. Domingos (P)	39.058742, −9.135427	Phaseolus vulgaris
COL		08/09/2013	Colares (P)	38.799517, −9.448335	Phaseolus vulgaris
LOU		03/10/2013	Lourinhã (P)	39.248145, −9.276321	Solanum melongena
CH		31/10/2013	Casal Hortelão (P)	38.851962, −9.393918	Solanum lycopersicum
RF		04/11/2013	Ribeira de Fráguas (P)	39.366415, −8.851037	Solanum lycopersicum
AMP		18/11/2013	Aldeia da Mata Pequena (P)	38.534363, −9.191163	Datura stramonium
GH	T. evansi	03/10/2013	University of Lisbon (P)	38.757852, −9.158221	Solanum lycopersicum
QL		26/11/2013	Quinta das Lameiras (P)	39.085837, −8.991478	Datura stramonium
GRA		03/12/2013	Graça, Lisbon (P)	38.71392, −9.129537	Solanum lycopersicum
Assaf	T. ludeni	20/09/2013	Assafora (P)	38.904743, −9.408592	Datura stramonium
CVM		03/10/2013	Casal Vale do Medo (P)	39.248450, −9.294550	Ipomoea purpurea
Alval		05/11/2013	Alvalade, Lisbon (P)	38.75515, −9.14685	Ipomoea purpurea

*

Red form (or T. cinnabarinus).

Molecular species identification and screening for endosymbiont infection

From 0 to 100 days after collection, adult spider mite females (n = 11–16) were randomly picked from each colony and individually analyzed for the presence of reproductive manipulators and for species identification. Total genomic DNA was extracted from each individual spider mite using the Sigma-Aldrich protocol and materials (GenElute Mammalian Genomic DNA Miniprep Kit, Sigma-Aldrich, St. Louis, MO, United States). Total DNA was eluted in the final step with 20 μl RNase-free water (Qiagen NV, Venlo, The Netherlands) in order to increase the concentration of DNA obtained from these very tiny animals (c.a. 300 μm long). The same DNA from each individual was used to screen for the presence of potential endosymbionts and to identify the spider-mite species. These identifications were performed by PCR amplification and then sequencing of a fragment of the nuclear ribosomal DNA (rDNA) ITS2 (internal transcribed spacer 2) region, which is the most commonly used marker to distinguish among spider-mite species (Ben-David et al.2007; Hurtado et al.2008), as the mitochondrial DNA (mtDNA) Cytochrome Oxidase subunit I (COI) is more polymorph in spider mites (Navajas et al.1998). The presence of Wolbachia, Cardinium, Rickettsia, Arsenophonus and Spiroplasma in each sample was determined using genus-specific primers, and the mitochondrial DNA (mtDNA) Cytochrome Oxidase subunit I (COI) was amplified and sequenced for 75 samples chosen depending of their infection status with Cardinium and Wolbachia. The list of primers, for both hosts and symbionts, is given in Table S1 (Additional file 1 in the Supplementary data). For each of the genes under study, the PCRs were done along with both positive and negative controls, in a 10 μl reaction mixture containing 5 μl of 2X QIAGEN Multiplex PCR Master Mix (Qiagen NV, Venlo, The Netherlands), 2 μl RNase free water, 1 μl of each primer at 2 μM, and 1 μl of DNA solution. Amplification conditions were as follows: 15 min at 94°C, and then 35 cycles of 94°C for 30 s, annealing (see primer-specific temperatures in Table S1) for 1 min 30 s, 72°C for 1 min and a final step at 72°C for 10 min. PCR products were visualized on 2% agarose gel. To verify the identity of the amplified endosymbionts genes (for all Cardinium- and Rickettsia-infected individuals, and up to three Wolbachia-infected individuals per population haphazardly chosen within both single- and multiple-infected mites), the PCRs were repeated for positive samples in a 20 μl final volume reaction, the PCR products were purified using Qiagen protocol and materials (QIAquick PCR Purification Kit) and sent for sequencing (Stabvida, Caparica, Portugal). All ITS2 and COI genes were also sequenced using the same protocol. The chromatograms were checked manually using MEGA version 5.1 beta (Tamura et al.2011) and the resulting sequences were run against the non-redundant nucleotide database using the BLAST algorithm of the National Center for Biotechnology Information. New sequences have been deposited in GenBank (see Table S2 and S3 in the Supplementary data, Additional file 1).

Phylogenetic analyses of COI mitochondrial haplotypes

The phylogenetic relationships between Tetranychus spp. haplotypes were reconstructed using maximum-likelihood (ML) analyses based on COI mitochondrial nucleotide sequences (fragments of 392 unambiguously aligned sites; best-fitting evolutionary model: HKY + I). The evolutionary model that best fitted the sequence data was determined using the Akaike information criterion with the program MEGA (Tamura et al.2011). ML heuristic searches using a starting tree obtained by neighbor joining were conducted in MEGA (Tamura et al.2011). Clade robustness was assessed by bootstrap analysis using 1000 replicates.

Statistical analyses

Analyses were carried out using the R statistical package (v. 3.3.2). The different statistical models built to analyze the effects of species, or of population within species, on the prevalence of Wolbachia and of Cardinium are described in the electronic supplementary material, Table S4 (Additional file 1, Supplementary data). The general procedure for building the statistical models was as follows: time (number of days after collection from the field at which DNA extraction was performed), species and the status of infection of the females by Wolbachia or Cardinium were fit as fixed explanatory variables, whereas population was either fit as a random explanatory variable when analyzing the complete dataset, or as a fixed effect when analyzing the differences between populations within each spider-mite species. The effect of spider-mite species (models 1 and 5) on the proportion of Wolbachia- or Cardinium-infected females, a binary response variable (1: infected, or 0: uninfected), was analyzed using a generalized linear mixed model with a binomial distribution (glmer, lme4 package), and the effect of population within species (models 2 to 4, and 6 to 8) was analyzed using a generalized linear model with a binomial distribution (glm). When the variables species or population were found to be significant, a posteriori contrasts (Crawley 2007) were carried out by aggregating factor levels and testing the fit of the simplified model using ANOVA. For all analyses, maximal models were simplified by sequentially eliminating non-significant terms to establish a minimal model (Crawley 2007), and the significance of the explanatory variables was established using chi-squared tests (Bolker 2008). The significant Χ² values given in the text are for the minimal model (Crawley 2007). Finally, the effect of Wolbachia and Cardinium infection on sequence divergence among COI haplotypes, and, for Cardinium-infected individuals, the effect of Cardinium strains on COI sequence divergence were assessed by a hierarchical analysis of molecular variance (AMOVA; Excoffier, Smouse and Quattro 1992). The statistical significance of the variance components of the AMOVA was determined by nonparametric procedures using 10 000 random permutations. Full dataset are given in Additional files 2, 3 and 4 (Supplementary data).

RESULTS

Species identification: ITS ribosomal types and COI mitochondrial haplotypes

Amplification and sequencing of a portion of the ITS2 gene allowed assigning 3 out of the 16 populations collected to T. evansi, 3 populations to T. ludeni and the 10 remaining populations to T. urticae (red morph; Table 1). All sequences obtained from populations of T. urticae were strictly identical and matched (100% identity at the nucleotide level) a sequence obtained from T. urticae collected in southern Spain (e.g. GenBank: GU565314). Similarly, all sequences obtained for T. ludeni (GenBank: MF428439) were strictly identical and matched (99% identity) several sequences obtained for this species worldwide (e.g. New Zealand, Japan, India, or Israel; GenBank: KP744529, AB076371, KU759957 and DQ656451, respectively). For T. evansi, however, both ITS2 ribosomal type 1 (GenBank: FJ440674) and type 2 (GenBank: FJ440673; Boubou et al.2011) were found in the population ‘QL’, but only the type 1 in ‘GRA’ and ‘GH’.

Analysis of mitochondrial COI DNA sequences from 75 individuals revealed 64 polymorphic sites within 392 bases sequenced (Table S2), and allowed distinguishing eight different mtDNA haplotypes (each individual within the same haplotype sharing exactly the same COI sequence; Table 2; Fig. 1). In T. urticae, almost all mites collected in Spain and Portugal belonged to the same haplotype (GenBank: HM565901, and MF428440, respectively), and shared a high degree of similarity with each other (99% identity; the two sequences differed by one single nucleotide mutation; Table S2). One female from the population ‘DC’ collected in Portugal, however, belonged to a COI haplotype (GenBank: MF428441; Table 2) more similar to a haplotype found in the population ‘FR’ from France (GenBank: HM486513) than to the one from the Spanish populations (96% and 94% identity, respectively; Table S2). In the French population, two other haplotypes were found: the main haplotype found in Portugal (GenBank: MF428440), and another, highly similar to this one (99% identity; GenBank: HM565899). In T. evansi, two different haplotypes were found (GenBank: MF428442 and MF428443), which were strictly identical (100% identity, query cover: 91%; Table S2) to the COI haplotypes I and II, respectively, originally recorded in France and Spain (GenBank: FJ440676 and FJ440678; Gotoh et al.2009), but also previously found in Portugal (Boubou et al.2011). Conversely, only one haplotype of T. ludeni was found and was identical to a haplotype found in China (GenBank: GQ141911; Table 2). Importantly, one T. ludeni female from the population ‘Assaf’, and 4 females from all T. evansi populations (from both ITS types), harbored two different T. urticae COI haplotypes, but the reverse (i.e. T. ludeni or T. evansi COI haplotypes associated with T. urticae ITS type) was not observed (Table 2).

Figure 1.

Phylogeny and distribution of Wolbachia and Cardinium (co)infections among T. evansi, T. ludeni and T. urticae mitochondrial COI haplotypes. The left panel shows the phylogeny of the 8 mtDNA haplotypes via maximum likelihood. Numbers on branches indicate percentage bootstrap support for major branches (1000 replicates). GenBank accession numbers are given between brackets. The scale bar is in units of substitutions/site. The right panel gives the distribution of symbiont infection observed with each mtDNA haplotype (black: Wolbachia-infected, n = 20; gray: Cardinium-infected, n = 7; dashed: Wolbachia-Cardinium coinfected individuals, n = 20; white: uninfected specimens, n = 28).

Table 2.

Mitochondrial COI alleles in spider mite populations. A total of 75 sequences were obtained from individual female spider mites. For each allele, we give the genbank accession number and the corresponding number of mites per population. We also give the total number of mites carrying each allele and the corresponding proportion relative to the total number of mites.

	Genbank accession	T. ludeni			T. evansi			T. urticae										Total number
Species	number [Ref]	Assaf	CVM	Alval	GRA	GH	QL	AlRo	AlBe	DF	LOU	COL	AMP	RF	DC	CH	FR	(proportion)
T. ludeni	GQ141911 [*]	3	5	5	–	–	–	–	–	–	–	–	–	–	–	–	–	13 (0.17)
T. evansi	MF428442 [**]	–	–	–	3	4	1	–	–	–	–	–	–	–	–	–	–	8 (0.11)
	MF428443 [**]	–	–	–	–	–	3	–	–	–	–	–	–	–	–	–	–	3 (0.04)
T. urticae	HM486513 [49]	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	1	1 (0.01)
	HM565899 [96]	–	–	–	1	1	–	–	–	–	–	–	–	–	–	–	1	3 (0.04)
	HM565901 [96]	–	–	–	–	–	–	5	5	–	–	–	–	–	–	–	–	10 (0.13)
	MF428440 [**]	1	–	–	1	–	1	–	–	4	4	4	4	5	4	5	3	36 (0.48)
	MF428441 [**]	–	–	–	–	–	–	–	–	–	–	–	–	–	1	–	–	1 (0.01)

*

Unpublished.

**

This study.

Inter- and intraspecific variation of endosymbionts prevalence and diversity

Overall, our sampling revealed infections by Wolbachia, Cardinium and Rickettsia within the three spider mites species, but none of the tested populations were found infected by Arsenophonus or Spiroplasma (Fig. 2).

Figure 2.

Prevalence of endosymbionts for each spider-mite population of (A) T. ludeni, (B) T. evansi, and (C) T. urticae. Each graph represents a population, in which the rows correspond to individual mites and the columns to their infection status by W: Wolbachia; C: Cardinium; and R: Rickettsia. White cells: uninfected; Gray cells: infected; Hatched cells: Cardinium and Rickettsia failed sequences but positive through PCR. None of the tested colonies were found infected by Spiroplasma or Arsenophonus.

Wolbachia was the most prevalent endosymbiont with an average of 60.7 ± 3.3% of infected females (Fig. 2). This prevalence differed significantly between host species (model 1, X²₂ = 5.92, P = 0.05), with no significant difference between infection prevalence in T. ludeni and T. evansi (31.6 ± 7.6% and 41.5 ± 7.8%, respectively; X²₁ = 0.24, P = 0.63), but the infection frequency in T. urticae being significantly higher than in these two species (74.3 ± 3.7%; X²₁ = 5.69, P = 0.02). Within species, no significant difference was found in Wolbachia prevalence among populations of T. evansi (model 2, X²₂ = 3.62, P = 0.16), but populations differed significantly for Wolbachia infection frequencies in T. ludeni and T. urticae (model 3, X²₂ = 11.51, P = 0.003, and model 4, X²₉ = 93.58, P < 0.0001, respectively). In T. ludeni, no Wolbachia-infected mites were found in ‘Assaf’, while both ‘Alval’ and ‘CVM’ harbored 46.2 ± 10.0% of infected mites. In T. urticae, the lowest prevalences were found in the populations collected in Spain (on average 14.8 ± 7.0%) and France (46.7 ± 13.3%; contrast between Spanish–French populations: X²₁ = 5.28, P = 0.02), whereas Wolbachia infection was fixed in five out of seven of the populations collected in Portugal, and the remaining two Portuguese populations (CH and DC) had an average prevalence of 83.3 ± 6.9% (contrasts between CH-DC and the other Portuguese populations: X²₁ = 13.05, P < 0.001; contrast between CH-DC and the French populations, X²₁ = 6.99, P = 0.008).

At the molecular level, no diversity of Wolbachia was found (Table 3). Of the 42 Wolbachia-infected mites for which the wsp gene was sequenced, 41 were strictly identical and matched (100% identity at the nucleotide level) several sequences available in GenBank for Wolbachia infecting several Tetranychus species worldwide (e.g. GenBank: DQ910771 and JX844806 in T. urticae, AB096218 and AF358417 in T. kanzawai, AB096226 in T. pueraricola, AY585711 in T. cinnabarinus) and previously attributed to members of the Wolbachia B supergroup and Ori subgroup (Gotoh, Noda and Hong 2003). Only one of the sequences obtained (GenBank: MF428430), from a Wolbachia-infected mites of the T. evansi population ‘GRA’, differed from the others (82% identity at the nucleotide level), but was highly similar (99% identity at the nucleotide level) to many sequences found in diverse arthropods species such as fruit flies (GenBank: FJ403332, KU870673, DQ834379, or EU116319), beetles (GenBank: KX436089), moths (GenBank: GU166594) or head lice (GenBank: AY596782). Note, however, that strains with similar wsp sequences can have very different MLST allelic profiles (Baldo et al.2006), but the amount of DNA obtained from each individual spider mite strongly restrained the possible number of PCRs and an in-depth study of Wolbachia variation within Tetranychus mites was beyond the scope of this study.

Table 3.

Occurrence and co-occurrence of Wolbachia wsp, Cardinium 16S rRNA and Rickettsia gtlA alleles. A total of 72 sequences were obtained from 80 positive samples through PCR. For each allele, we give the genbank accession number, the corresponding number of infected mites per population, and indicate the symbiont species involved in a multiple infection with superscripts (a: Cardinium, b: Wolbachia and c: Rickettsia). For each allele, we also give the total number of infected mites and the corresponding proportion relative to the total number of mites infected by the same endosymbiont.

Gene (fragment length)	Genbank accession	T. ludeni			T. evansi			T. urticae									Total number
	numbers [Ref]	CVM	Alval	GRA	GH	QL	AlRo	AlBe	DF	LOU	COL	AMP	RF	DC	CH	FR	(proportion)
Wolbachia wsp (524bp)	DQ910771 [72]	3(1a)	3	1	3(1a)	3(1a1c)	2(1a)	2(1a)	3	3	3(1a)	3(2a)	3(2a)	3(2a)	3(2a1c)	3(1a1c)	41 (0.98)
	MF428430 [*]	–	–	1	–	–	–	–	–	–	–	–	–	–	–	–	1 (0.02)
Cardinium 16S rRNA (325bp)	MF428431 [*]	–	–	–	–	–	1	–	–	–	–	–	–	–	1b	1b	3 (0.11)
	MF428432 [*]	1b,c	–	–	–	–	–	–	–	–	–	–	–	–	–	–	1 (0.04)
	MF428433 [*]	–	–	1	–	–	1	–	–	–	–	1b	–	–	–	–	3 (0.11)
	MF428434 [*]	–	2	–	2(2b)	–	1b	2(1b)	–	–	–	–	4(4b)	3(2b)	1b	2(2b)	17 (0.63)
	MF428435 [*]	1b	–	–	–	–	–	–	–	–	–	–	–	–	–	–	1 (0.04)
	MF428436 [*]	–	–	–	–	–	–	–	–	–	–	–	1b	–	–	–	1 (0.04)
	MF428437 [*]	–	–	–	–	1b	–	–	–	–	–	–	–	–	–	–	1 (0.04)
Rickettsia gtlA (431bp)	MF428438 [*]	–	–	–	–	1b	–	–	–	–	–	–	–	–	–	1b	2 (1.00)

*

This study.

The prevalence of Cardinium was 14.6 ± 2.4% on average (total of 32 infected mites over 219). However, we could obtain the 16S rRNA sequences only for 27 of 32 infected individuals: for five individuals we failed to amplify Cardinium through three successive PCR following the first diagnostic of infection. Two main possibilities could explain this failure: first, very low densities of Cardinium within these individuals could lead to PCR-based imperfect detection, as shown for other parasites (Zélé et al.2014). Second, we cannot totally exclude the possibility of contaminations or of unspecific amplifications of other bacteria when using 16S rRNA primers for Cardinium. For this latter reason, we did not consider these five individuals as infected in the following analyses. After correction, the prevalence of Cardinium was overall of 12.3 ± 2.2% (9.8 ± 4.7% in T. evansi, 10.5 ± 5.0% in T. ludeni, and 13.6 ± 2.9% in T. urticae) with no significant difference in prevalence among the three species (model 5, X²₂ = 0.32, P = 0.85; Fig. 2). Within species, no differences among populations of T. evansi and T. ludeni were found (models 6 and 7, X²₂ = 0.55, P = 0.76 and X²₂ = 3.25, P = 0.19, respectively), but Cardinium prevalences in populations of T. urticae differed significantly (model 8, X²₉ = 31.19, P < 0.001). Indeed, in three populations (‘COL’, ‘DF’ and ‘LOU’), none of the mites tested were infected, but a posteriori contrast analyses among the remaining (infected) populations revealed no significant difference in Cardinium prevalence (X²₆ = 6.62, P = 0.36).

At the molecular level, we found some degrees of inter-individual variation in the Cardinium 16S rRNA sequences (Table 3). Of the 27 Cardinium 16S rRNA sequences obtained, we could discriminate six different alleles, which, despite sharing a high degree of similarity (97%–99% identity at the nucleotide level), differed by single nucleotide mutations over 13 different nucleotide sites (Table S3). The most prevalent allele (GenBank: MF428434) was found in 17 Cardinium-infected mites over the three species (c.a. 63%; Table 3). The corresponding sequence is very close to several sequences found for Cardinium in diverse arthropod species, including spider mites (e.g. 99% identity at the nucleotide level with a sequence obtained in T. urticae in Japan; GenBank: AB241134; Gotoh, Noda and Ito 2007a). Three additional mites harbored highly similar Cardinium alleles (99% identity at the nucleotide level; GenBank: MF428435, MF428436, and MF428437, respectively; Tables 3 and S3). Interestingly, we also found alleles identical or highly similar (99% identity at the nucleotide level), to an allele previously found for Cardinium endosymbiont of Scaphoideus titanus, the American grapevine leafhopper (GenBank: AM042540; Marzorati et al.2006), for two T. urticae (GenBank: MF428431), and one T. ludeni (GenBank: MF428432) individuals, respectively. Finally, four individuals (one T. evansi and three T. urticae) harbored Cardinium sequences (GenBank: MF428433) with double nucleotide peaks at the nucleotide positions that differ between AB241134 and MF428434 (Table S3). This result suggests co-infection with two different Cardinium strains within these mites.

Infection by Rickettsia was rare. Indeed, this symbiont was found in only 6 of 219 individuals belonging to the three spider mite species (2.7 ± 1.1%; Fig. 2): 7.3 ± 4.1% of T. evansi individuals were infected, belonging to a single population, 2.6 ± 2.6% of T. ludeni individuals, also from a single population, and 1.4 ± 1.0% of T. urticae individuals, stemming from two populations. However, for four of the six infected individuals we failed to obtain the gtlA sequences of Rickettsia. After correction, the prevalence of Rickettsia was overall of 0.9 ± 0.6% among the three mite species (2.4 ± 2.4% in T. evansi, 0% in T. ludeni, and 0.7 ± 0.7% in T. urticae). The two sequences obtained, one in T. evansi, and one in T. urticae (GenBank: MF428438; Table 3), were strictly identical to a sequence previously found in T. urticae in China (GenBank: KP828066; Zhang et al.2016b). The most closely related sequences found in other arthropod hosts were from Rickettsia in the parasitoid wasp Pnigalio sp. (97% identity at the nucleotide level; GenBank: GU559856), and in B. tabaci (96% identity; GenBank: DQ077708).

Multiple infections and covariation of Wolbachia and Cardinium prevalence

Several individuals within the three mites species, and across different populations, were infected with more than one endosymbiont. We detected 25 mites (11.4 ± 2.2% in average) coinfected by Wolbachia and Cardinium, four mites (ca. 1.8 ± 0.9%) coinfected by Wolbachia and Rickettsia and one individual (ca. 0.5 ± 0.5%) infected by the three endosymbionts (population ‘CVM’; Fig. 2). Moreover, our analyses revealed a co-variation between the prevalence of Wolbachia and that of Cardinium, both symbionts affecting significantly the probability of infection by each other (model 1, X²₁ = 4.99, P = 0.03, and model 5, X²₁ = 4.56, P = 0.03, respectively; Fig. 3). However, no particular pattern of association between Cardinium 16S rRNA gene sequences, Wolbachia infection, mites species or population emerged (Table 3).

Figure 3.

Mixed logistic regression of individual probability of Wolbachia infection on the probability of Cardinium infection. The solid line represents the mean predicted values; the dark shaded area represents the range of the 95% confidence intervals based on the fixed-effects uncertainty only, while the light shaded area also includes the variance of a random factor ‘population’. The observed data at the individual level are not represented as they all take the values ‘0’ or ‘1’ for both Wolbachia and Cardinium infection. Open circles are the observed mean prevalence of Cardinium per population plotted against the observed mean prevalence of Wolbachia, while filled circles are their corresponding linear unbiased predictors.

Effects of Wolbachia and Cardinium infection on mtDNA variation

No particular pattern of association was found between the mitochondrial COI DNA sequences obtained from individual mites and their Wolbachia and/or Cardinium infection status (F_1,74 = 0.05; P =  0.07, and F_1,74 = 0.009; P =  0.26, respectively; Fig. 1), or between their mitochondrial COI DNA sequences and the Cardinium 16S rRNA type (F_1,75 = 0.22; P =  0.18, Table 4). For instance, the most prevalent Cardinium sequence obtained (MF428434; c.a. 63% of Cardinium infected-mites; Table 3) was found for all mitotypes, and, reciprocally, several mitotypes harbored different Cardinium sequences (Table 4).

Table 4.

Co-occurrence of Cardinium 16S rRNA and mitochondrial COI alleles of spider mites. Cardinium 16S rRNA and mite COI sequences were obtained from a total of 29 Cardinium-infected individual female spider mites. For each Cardinium 16S rRNA allele, we give the corresponding number of infected mites per COI allele (Genbank accession numbers [Ref]), and the species to which they belong between brackets. Te: T. evansi, Tl: T. ludeni, Tu: T. urticae.

	Mites’ COI
Cardinium 16S rRNA	HM486513 [49]	HM565899 [96]	HM565901 [96]	MF428440 [*]	MF428441 [*]	MF428442 [*]	GQ141911 [**]
MF428431 [*]	–	1 (Tu)	1 (Tu)	–	–	–	–
MF428432 [*]	–	–	–	–	–	–	1 (Tl)
MF428433 [*]	–	–	1 (Tu)	2 (Tu)	–	1 (Te)	–
MF428434 [*]	1 (Tu)	2 (Te-Tu)	3 (Tu)	9 (Tu)	1 (Tu)	1 (Te)	2 (Tl)
MF428435 [*]	–	–	–	–	–	–	1 (Tl)
MF428436 [*]	–	–	–	1 (Tu)	–	–	–
MF428437 [*]	–	–	–	1 (Te)	–	–	–

*

This study.

**

Unpublished.

DISCUSSION

In this study, we analyzed the prevalence and diversity of South-West European populations of spider mites and their endosymbionts. The populations collected mainly belonged to three species of spider mites: T. urticae, T. evansi and T. ludeni, the most prevalent being T. urticae. Our sampling effort was such that we did not pre-select the plants sampled. Consequently, this result may reflect the host range of T. urticae, which is larger than that of T. evansi and T. ludeni (Migeon and Dorkeld 2006–2017). However, our results contrast with those of another study conducted in the Eastern part of the Iberian Peninsula, near Valencia, where T. evansi is the most abundant species (Ferragut, Garzon-Luque and Pekas 2013). A study on invasion scenarios of T. evansi predicts that populations around Valencia stem from an invasion route that differs from that leading to the establishment of T. evansi in Portugal (Boubou et al.2012). Possibly, the invasive success of this species differs depending on the invasion event. The T. evansi clades that we recovered match those found earlier for the same region in Portugal (Boubou et al.2011).

Analysis of mitochondrial and nuclear DNA revealed a high COI diversity in comparison to that of ITS for T. urticae, as found earlier (Navajas et al.1998). Overall, these two markers yielded concordant species identifications in most individuals tested. However, we found mitochondrial DNA of T. urticae in the other two species, albeit at low frequencies. Screening contaminations seems unlikely, as mitochondrial and nuclear DNA were extracted from the same individuals, PCR controls were all correct and these unexpected ITS-COI associations occurred for two different mitotypes, and in four different populations (one of T. ludeni and the three T. evansi). Therefore, it seems that introgression may occur among these species, and this suggests that they co-occur on the same host plants and locations, although we did not detect species co-occurrence in this study. In laboratory studies using these species, viable hybrids have been found at very low frequencies as well (Sato, Alba and Sabelis 2014; Clemente et al.2016, but see Clemente et al.2018), and assortative mating was either low (Clemente et al.2016) or absent (Sato, Alba and Sabelis 2014). These findings are compatible with the possible introgression found here. Nevertheless, the low frequencies observed here might also be due to the detection of F1 hybrids only, as in most cases such hybrids are sterile (Clemente et al.2016). Curiously, however, this introgression seemed to be unidirectional, with the mtDNA of T. ludeni or T. evansi never being found in individuals with nuclear T. urticae DNA. It would thus be interesting to corroborate these results in future laboratory experiments, along with broader analyses of natural populations of these species.

Of the five endosymbionts screened, only Wolbachia, Cardinium and Rickettsia were found to infect spider mites, whereas none of the tested individuals harbored Spiroplasma or Arsenophonus. However, this is the first report of Wolbachia, Cardinium and Rickettsia in T. evansi and T. ludeni. Overall, Wolbachia prevalence was higher than that of the two other symbionts, even though it was lower in T. evansi and T. ludeni, and in non-Portuguese T. urticae populations. Such variable levels of Wolbachia prevalence across species and populations have already been reported in other regions worldwide (Gotoh, Noda and Hong 2003; Gotoh et al.2007b; Zhang et al.2016b). Interestingly, however, we found Wolbachia in the invasive species T. evansi. This runs counter other studies that have shown the loss of Wolbachia in invasive populations of fire ants (Shoemaker et al.2000), Argentine ants (Reuter, Pedersen and Keller 2005) and Australian thrips (Nguyen, Spooner-Hart and Riegler 2016). In invasive species, the loss of Wolbachia is expected because introduced populations are likely to harbor only a subset of the symbionts of their parental population(s), possibly because a bottleneck effect during invasion. Spider mites may thus be used to test hypotheses concerning the loss or maintenance of symbionts in invasive populations (see below). Cardinium was found in all three spider mites species, which corroborates previous findings of a relatively high occurrence of this symbiont in mites (Nakamura et al.2009; Zhang et al.2016a) compared with other arthropods groups (Weeks, Velten and Stouthamer 2003; Zchori-Fein and Perlman 2004; Duron et al.2008; Nakamura et al.2009). However, its prevalence was overall much lower than that of Wolbachia, a result not always found in spider mites (Zhang et al.2016b). The prevalence of Rickettsia was very low, as found by Zhang et al. (2016b), but this study is the first to report Rickettsia infection in T. ludeni and T. evansi. Finally, although Spiroplasma has been previously found in T. urticae (Enigl and Schausberger 2007), we did not detect it in our study, as reported by Zhang et al. (2016b). Similarly, we did not find Arsenophonus. Although the latter has been found in the Acari (Clay et al.2008; Dergousoff and Chilton 2010; Clayton et al.2015; Duron et al.2017), it has, to our knowledge, never been reported in spider mites.

The maintenance and spread of reproductive manipulators in nature, hence their prevalence, are mainly affected by the type and strength of reproductive manipulation, the fitness costs/benefits of the symbionts to their hosts, and the efficiency of vertical transmission (Werren, Baldo and Clark 2008; Engelstadter and Hurst 2009). Variability in any of these factors might explain the observed lower prevalence of Cardinium than that of Wolbachia. Both endosymbionts induce CI in spider mites (Gotoh, Noda and Ito 2007a; Gotoh et al.2007b; Ros and Breeuwer 2009; Zhu et al.2012; Zhao et al.2013b). Although some studies found that Cardinium induces a higher level of CI than Wolbachia in several mites species (e.g. T. pierci and Bryobia sarothamni; Ros and Breeuwer 2009; Zhu et al.2012), another showed that Cardinium may not induce CI in T. urticae (Gotoh, Noda and Ito 2007a). It is thus possible that its lower prevalence is due to low reproductive manipulation. In addition, Cardinium may exhibit imperfect maternal transmission, as found in thrips (Nguyen et al.2017), or in the parasitic wasp Encarsia pergandiella (although near-perfect in this species; Perlman, Kelly and Hunter 2008). Conversely, the high prevalence of a single Wolbachia strain found here may be due to this strain inducing low fitness costs, strong reproductive manipulation and having nearly perfect maternal transmission. An in-depth study of the fitness effects of these endosymbionts, along with their ability to manipulate the reproduction of their host in these populations would thus allow testing these hypotheses.

These general predictions, coming from epidemiological modeling, may be tempered by environmental factors. Indeed, a large number of other factors have been identified, such as symbiont-host genotype interactions, host diet or temperature effects (e.g. Van Opijnen and Breeuwer 1999; Mouton et al.2007; Wilkinson, Koga and Fukatsu 2007; Gotoh et al.2007b; Pan et al.2013; Zhang et al.2013b; Boivin et al.2014; Suh et al.2015). For instance, Gotoh, Noda and Ito (2007a) showed that Wolbachia is much more sensitive to elevated temperature than Cardinium in T. pueraricola. In our study, the higher summer temperatures in Almeria (Spain; c.a. 26.7°C on average in August; http://www.aemet.es/es/serviciosclimaticos/datosclimatologicos/valoresclimatologicos?l=6325O&k=and) compared to that of the region of Lisbon (Portugal; c.a. 23.5°C on average in August; http://www.ipma.pt/pt/oclima/normais.clima/1981-2010/012/) might explain the lower prevalence of Wolbachia, but not that of Cardinium in the T. urticae populations collected in this region. The fact that a single Wolbachia wsp sequence was found in this region, which contrasts with other studies of spider mites in other regions (Gotoh, Noda and Hong 2003; Gotoh et al.2007b; Ros et al.2012; Zhang et al.2013a,b), may indicate specific adaptation of this strain to the environmental conditions of South-Western Europe. Host plants of herbivorous arthropods may also affect symbiont diversity and identity (e.g. Ferrari et al.2004; Ahmed et al.2010; Brady and White 2013; Guidolin and Consoli 2017; but see Ji et al.2015). Moreover, bacterial endosymbionts may also be conditional mutualists in some host species, conferring advantages under certain environmental conditions. For instance, they may compensate for the nutritionally deficient diets of herbivore hosts and promote dietary specialization by affecting their performance on plants, they may also increase their ability to tolerate high temperatures, or may protect them against natural enemies (reviewed in Haine 2008; Moran, McCutcheon and Nakabachi 2008; Frago, Dicke and Godfray 2012). Unfortunately, our sampling design, largely unbalanced with respect to these factors, does not allow testing their relevance for the system under study.

Interaction between symbionts within a single host may also be an important factor affecting their prevalence, either negatively (competition) or positively (facilitation) (Vautrin et al.2008; Vautrin and Vavre 2009; Hellard et al.2015). As previously found in several spider mites species in China (Liu, Miao and Hong 2006; Zhao et al.2013b; Zhang et al.2016b), and in Northern Europe (Ros and Breeuwer 2009; Ros et al.2012), in this study Cardinium prevalence was higher in populations also infected by Wolbachia, which may suggest facilitation between these endosymbionts. The maintenance of Cardinium in the spider mites populations collected in this study might thus be due to a hitchhiking effect, whereby Wolbachia-induced phenotypic changes favor their own transmission, but also that of other co-infecting symbionts (Frank 1998; Engelstadter, Telschow and Hammerstein 2004; Vautrin and Vavre 2009). Moreover, the effects of each symbiont on their host may be exacerbated in presence of the other. Indeed, several studies showed that Wolbachia can strengthen the induction of cytoplasmic incompatibility by Cardinium in several spider-mite (Zhu et al.2012; Zhao, Zhang and Hong 2013a; Zhao et al.2013b), or arthropods (White et al.2009) species (but see Nguyen et al.2017). Moreover, Ros and Breeuwer (2009) found a higher fecundity advantage conferred by Wolbachia-Cardnium co-infections in Bryobia sarothamni mites, than by Wolbachia or Cardinium single infections (although this result may be confounded by the host genetic background, as acknowledged by the authors). Studies of the fitness effects of both Wolbachia and Cardinium, in single and multiple infections in these populations would allow corroborating or rejecting this hypothesis.

Symbionts usually spread within species, but they may also cross species boundaries via two main routes. First, cytoplasmic introgression, through interspecific hybridization events, may favor the spread of vertically transmitted symbionts among species (Raychoudhury et al.2009; Jackel, Mora and Dobler 2013). The incongruences between the mites COI and ITS sequences obtained for some individuals in this study (see above), go in that direction. Second, intimate ecological associations, such as trophic interactions and plant feeding in particular, may provide horizontal transmission opportunities. It has been shown, for instance, for Rickettsia and Wolbachia in whiteflies (Caspi-Fluger et al.2012; Li et al.2017) and for Cardinium in leaf hoppers (Gonella et al.2015), and could explain the genetic diversity patterns observed here. Indeed, multiple events of horizontal transfer of Cardinium in the field would explain both its maintenance at low frequency within populations, the high sequence variation (i.e. potential strains) found for this symbiont, and the lack of association with the host mtDNA. In addition, phylogenetic similarity of donor and recipient host species should facilitate horizontal transmission (Boivin et al.2014), and incongruences between phylogenies of host and parasites, suggesting occasional horizontal transmission of Cardinium and/or Wolbachia between hosts, have been found at many occasions in spider mites (Yu et al.2011; Ros et al.2012; Zhang et al.2016b; but see Zhang et al.2013a). However, although these symbionts might have unexpected tolerance towards stressful conditions (e.g. Wolbachia can be maintained alive for at least a week in synthetic liquid medium out of the host cells; Rasgon, Gamston and Ren 2006), the acquisition of a symbiont by horizontal transfer might be transient and not subsequently vertically transmitted in the new host species (Chiel et al.2009). Further studies should thus be conducted to determine the occurrence of both intra- and interspecific horizontal transfers of these endosymbionts under controlled laboratory conditions in spider mites.

In conclusion, this study highlights the richness and complexity of the symbiont community in spider mites. Although the factors affecting the spread and persistence of these endosymbionts in host populations remain largely unexplored, this system has the potential to become a model system in the study of host-symbiont interactions.

SUPPLEMENTARY DATA

Supplementary data are available at FEMSEC online.

Acknowledgements

We are grateful to J. Moya-Laraño and F. Vavre for useful discussions and comments. We also thank D. Godinho, R. Ponce, and L. Rodrigues, for their help in spider-mite collection.

Authors’ contributions

Experimental conception and design: FZ, OD, MW, SM, IO; field collections: FZ, IS, SM; molecular analyses: FZ, IS; statistical analyses: FZ; phylogenetic analyses: OD; paper writing: FZ, SM, with input from all authors. Funding: SM, IO. All authors have read and approved the final version of the manuscript.

FUNDING

This work was funded by an FCT-ANR project (FCT-ANR//BIA-EVF/0013/2012) to SM and IO, and by an FCT-Tubitak project (FCT-TUBITAK/0001/2014) to SM and Ibrahim Cakmak.

FZ was funded through an FCT Post-Doc fellowships (SFRH/BPD/125020/2016). Funding agencies did not participate in the design or analysis of experiments.

Conflict of interest. None declared.