Microbiome responses during virulence adaptation by a phloem‐feeding insect to resistant near‐isogenic rice lines

Abstract The microbiomes of phloem‐feeding insects include functional bacteria and yeasts essential for herbivore survival and development. Changes in microbiome composition are implicated in virulence adaptation by herbivores to host plant species or host populations (including crop varieties). We examined patterns in adaptation by the green leafhopper, Nephotettix virescens, to near‐isogenic rice lines (NILs) with one or two resistance genes and the recurrent parent T65, without resistance genes. Only the line with two resistance genes was effective in reducing leafhopper fitness. After 20 generations on the resistant line, selected leafhoppers attained similar survival, weight gain, and egg laying to leafhoppers that were continually reared on the susceptible recurrent parent, indicating that they had adapted to the resistant host. By sequencing the 16s rRNA gene, we described the microbiome of leafhoppers from colonies associated with five collection sites, and continually reared or switched between NILs. The microbiomes included 69–119 OTUs of which 44 occurred in ≥90% of samples. Of these, 14 OTUs were assigned to the obligate symbiont Candidatus sulcia clade. After 20 generations of selection, collection site had a greater effect than host plant on microbiome composition. Six bacteria genera, including C. sulcia, were associated with leafhopper virulence. However, there was significant within‐treatment, site‐related variability in the prevalence of these taxa such that the mechanisms underlying their association with virulence remain to be determined. Our results imply that these taxa are associated with leafhopper nutrition. Ours is the first study to describe microbiome diversity and composition in rice leafhoppers. We discuss our results in light of the multiple functions of herbivore microbiomes during virulence adaptation in insect herbivores.


| INTRODUC TI ON
The phloem of grasses is consumed by a diversity of specialized insect herbivores. Although rich in sugars and nutrients, phloem is deficient in several amino acids that are the necessary building blocks of insect proteins (Gündüz & Douglas, 2009;Wan et al., 2014). Furthermore, grasses have evolved a range of defenses to deter herbivores and to reduce the fitness gains from phloem feeding. Many of these defenses are located in the phloem, including secondary chemicals and herbivore-induced defense hormones such as jasmonic (JA) and salicylic acids (SA) or enzyme-blocking molecules such as proteinase inhibitors (Behmer et al., 2013;Du et al., 2015;Fujita, Kohli, & Horgan, 2013). It is now apparent that phloem-feeding insects depend on a range of facultative or obligate symbiotic microorganisms, including bacteria and fungi, to overcome the nutritional deficiencies and defenses of phloem (Douglas, 1998;2009;Ferrater, Jong, Dicke, Chen, & Horgan, 2013;Hansen & Moran, 2014;Noda et al., 2012;Sasaki, Kawamura, & Ishikawa, 1996;Wan et al., 2014). Furthermore, a number of studies suggest that endosymbiotic bacteria and yeasts might determine host preferences among phloem feeders (Ferrari, Scarborough, & Godfray, 2007;Ferrari, West, Via, & Godfray, 2012;Ferrater et al., 2015). Moreover, based on evidence from manipulative studies with plant bugs (Megacopta spp.), Hosokawa, Kikuchi, Shimada, and Fukatsu (2007) indicated that symbionts can determine the comparative virulence of bug species on different host plants. Several authors have proposed that changes in symbiotic bacterial and yeast communities may also underlie the rapid adaptation by pest herbivores to resistant crop varieties (Chen, Bernal, Tan, Horgan, & Fitzgerald, 2011;Ferrater & Horgan, 2016;Ferrater et al., 2015;Lu et al., 2004;Tang, Lv, Jing, Zhu, & He, 2010;Wang, Zhu, Lai, & Fu, 2015;Xu et al., 2015). Knowledge of the role of the microbiome is therefore essential to fully understand the ability of herbivores to overcome host defenses-a major component of insect-plant coevolution (Stenseth & Smith, 1984).
Over the last 60 years, host plant resistance has received considerable research attention in the management of cereal pests, particularly phloem-feeding hemipterans. In Asia, plant resistance against rice planthoppers and leafhoppers is currently the principal focus of publically funded research into insect pest management (Horgan, 2018). However, the widespread and rapid adaptation by target herbivores to resistant hosts is a major challenge for crop breeders and seriously limits the effectiveness and durability of resistant crop varieties (e.g., aphids on wheat, planthoppers, and leafhoppers on rice [Haley, Peairs, Walker, Rudolph, & Randolph, 2004;Hirae, Fukuta, Tamura, & Oya, 2007;Horgan, 2018;Horgan et al., 2015Horgan et al., , 2018Vu et al., 2014]). Recent studies that examined the microbiomes of insect herbivores, particularly the rice planthopper, Nilaparvata lugens, reared-on or adapted to contrasting host plant genotypes have suggested that microbiomes change (structurally or functionally) during selection and possibly determine adaptation (Ferrater et al., 2013(Ferrater et al., , 2015Ojha, Sinha, Padmakumari, Bentur, & Nair, 2017;Wang et al., 2015). However, Ferrater et al. (2013) indicated that to date, most studies that seek associations between endosymbionts and herbivore virulence have only described microbiomes from two or three insect colonies or samples, each reared on a different host plant (e.g., Chen et al., 2011;Lu et al., 2004;Ojha et al., 2017;Tang et al., 2010;Xu et al., 2015). This design confounds microbiome variability between different herbivore populations and microbiome responses to host feeding. Furthermore, until recently, virulence studies of rice planthopper microbiomes have used rice varieties from diverse breeding backgrounds often without clear knowledge of the genetic mechanisms underlying resistance. For example, the varieties TN1 (susceptible), Mudgo (Bph1 locus), and ASD7 (bph2 locus) have been used in several studies of rice planthopper microbiomes despite widespread virulence of planthoppers to all three varieties (Chen et al., 2011;Wang et al., 2015;Xu et al., 2015). These factors make it difficult to relate changes in microbiomes to the selection potential of specific resistance genes. Recently, a number of research teams have developed near-isogenic rice lines that share common recurrent parents, but differ by containing specific gene loci introgressed through marker-assisted selection from different resistance donors (Fujita et al., 2013; and see Horgan et al., 2019).
By using such near-isogenic lines, studies of virulence adaptation can better associate changes in microbiomes with the effects of specific resistance genes without confounding background genetic effects (Horgan, 2018).
The present study examines the microbiomes of green leafhopper, Nephotettix virescens (Figure 1), colonies selected on nearisogenic rice lines (NILs) with one or two genes for resistance to Nephotettix cincticeps (see below), and on the susceptible recurrent parent, T65. Nephotettix virescens is a polyphagous leafhopper (Khan, Hibino, Aguiero, Daquioag, & Opina, 1991) that is a considerable pest of rice (Oryza sativa) in South and South-east Asia. The leafhopper occasionally causes mechanical damage to rice plants, but more importantly transmits tungro rice viruses (Azzam & Chancellor, 2002;Khan et al., 1991). Beginning in the 1970s, a number of studies indicated that leafhopper populations are variously affected by resistance factors present in traditional rice varieties (Fujita et al., 2013).
Although the mechanisms remain largely undetermined, these resistance factors reduce leafhopper feeding, nymph development, and egg laying (Asano et al., 2015;Horgan et al., 2019). Currently, more than 24 loci with genes for resistance to leafhoppers (Nephotettix spp. and Recelia dorsalis) have been identified (Fujita et al., 2013) and a number of these genes have been introgressed into the japonica rice variety T65 using marker-assisted selection (Horgan et al., 2019). In our experiments, we used monogenic near-isogenic rice lines (NILs) carrying either the GRH2 or GRH4 gene loci (henceforth GRH2-NIL and GRH4-NIL, respectively) with resistance to N. cincticeps, and a line carrying both genes together (henceforth GRH2/ GRH4-PYL [we use "PYL" as an abbreviation for the breeding term "pyramided," which indicates a breeding line or NIL with ≥2 resistance genes]). Both GRH2 and GRH4 were first identified from DV85 using N. cincticeps in phenotyping studies. Resistance mechanisms associated with GRH2 and GRH4 have not been fully elucidated; however, Asano et al. (2015) found that genes for several types of proteinase inhibitor and several genes of the cytochrome P450 family were expressed in a GRH2/GRH4-PYL infested by N. cincticeps.
Furthermore, genes associated with the production of volatiles, particularly sesquiterpenes, were upregulated during N. cincticeps attack (Asano et al., 2015). Previous studies have indicated that the pyramided line, GRH2/GRH4-PYL, is highly resistant to N. virescens (Horgan et al., , 2019. By continually monitoring the fitness of N. virescens on GRH2/ GRH4-PYL over successive generations, Horgan et al. (2018) noted that whereas partial virulence was achieved within ten generations (i.e., leafhopper feeding efficiency, survival, and weight gain equal to that on susceptible varieties), leafhoppers required several further generations of selection to achieve complete virulence (i.e., egg laying equal to that on susceptible varieties). We therefore used colonies only after 20 generations of selection to insure complete virulence adaptation in our study. We assessed the microbiomes of these adapted colonies to identify possible microbiome associations with virulence of the insect host. Horgan et al. (2018) have also shown that N. virescens colonies successively reared on the monogenic resistant lines (GRH2-NIL and GRH4-NIL) were capable of surviving and developing on the pyramided line GRH2/GRH4-PYL without any prior exposure to that line (henceforth phase I selection). Therefore, we expected that aspects of microbiome community structure would be similar among colonies reared on GRH2-NIL, GRH4-NIL, and GRH2/GRH4-PYL, but different from colonies reared on T65. Without prior evidence to suggest which bacterial taxa might influence virulence or how the microbiomes might react to host plant resistance, we focused on two response levels. Firstly, we examined the overall composition of the microbiomes from the different colonies. Secondly, we looked at the abundance and proportional representation of different taxa in the leafhopper samples to detect significant host plant effects. To further test the hypothesis, we examined changes in key bacterial taxa (identified after phase I selection) among leafhoppers when moved to novel resistant hosts for further selection (phase II selection). In particular, we predicted that switching leafhoppers from a susceptible to a more resistant host would produce changes in the abundance of the key OTUs identified from phase I. To relate potential changes in the microbiome to further adaptation, we continuously monitored the leafhopper colonies for survival, weight gain, and egg laying throughout phase II selection. We also assessed their ability to damage the host plant as populations became increasingly virulent. We discuss our results in light of current knowledge of the microbiomes of insect herbivores and of the mechanisms of virulence adaptation in phloem-feeding leafhoppers.

| Plant materials
We obtained DV85 and T65 from the Germplasm Bank at the International Rice Research Institute (IRRI) in the Philippines. The resistant lines that we used, GRH2-NIL, GRH4-NIL, and GRH2/ GRH4-PYL, were BC 6 F 5 generations selected using Simple Sequence Repeat markers associated with the target loci during repeated backcrossing of the donor variety DV85 and the recurrent parent T65 (Horgan et al., , 2019. Seeds of the NILs were bulked-up in a screen-house at IRRI during the dry season when temperatures were coolest.

| Leafhopper colonies and Phase I selection
In this study, we used greenhouse colonies derived from five initial N. virescens populations. These were from original collections made in rice fields at Laguna, Batangas, Quezon, Rizal, and San Pablo (Luzon Island, Philippines). The colonies were each selected on the four rice lines (henceforth "natal" hosts) for 20 generations (5 populations × 4 natal hosts = 20 colonies). Details of selection and F I G U R E 1 The green leafhopper, Nephotettix virescens (photo courtesy of IRRI Knowledge Bank) the nature of the selected colonies (their ability to develop on natal hosts) are described in detail by Horgan et al. (2018). Briefly, selection was conducted as follows.
At two to three generations after field collection, the five populations reared on the rice variety TN1 were each split into four parts (with >200 adults each) and placed in insect cages (120 × 60 × 60 cm,  Horgan et al., 2018). For the purposes of the present study, we present the results of virulence monitoring from populations prior to selection and after 20 generations of Phase I selection. To assess adult and nymph survival, newly emerged gravid females (isolated as fifth instars to ensure they were unmated) and nymphs, respectively, were collected from each colony (5 adults or Egg laying was assessed by introducing mated, gravid females (2 females) to 20 DAS plants grown in size #0 pots under insect cages (dimensions as above). The females were allowed to oviposit for 5 days after which the plants were collected and dissected to count the eggs.

| Changes in virulence during phase I selection
All bioassays (for phase I and phase II [see below]) were conducted on a greenhouse bench (temperature: 25-37°C; 12 hr:12 hr, day:night) using a randomized block (=origin) design. Colonies of different origin were not assessed as a treatment during the selection studies but were used to gain increased external validity from the selection experiments (Ferrater et al., 2013).

| Leafhopper selection on GRH2/GRH4-PYL (phase II selection)
Adult leafhoppers (150-200 pairs) were collected from colonies that had been selected on GRH2-NIL and GRH4-NIL for 20 generations during phase I selection or were without exposure to any resistance loci (maintained for 20 generations on T65). The adults were transferred to cages (dimensions as above) with GRH2/GRH4-PYL as feeding plants (i.e., a novel natal host), allowed to oviposit, and nymphs allowed to develop. These new colonies (five for each original natal host, GRH2-NIL, GRH4-NIL, and T65 = 15) were continuously exposed to GRH2/GRH4-PYL for six generations. A series of control colonies that were continuously reared on their original natal hosts was also maintained (5 colonies × 4 original natal hosts = 20).
Therefore, 35 colonies were simultaneously maintained during the experiment.
At each generation, the colonies were monitored to assess nymph survival, adult survival, and egg laying on GRH2/GRH4-PYL using the bioassays described above.

| Extraction of sample DNA
Samples were homogenized in 2× CTAB extraction buffer and 2 µl of β-mercaptoethanol using a micropestle. The homogenized mixture was treated with 2.5 µl of ProteinaseK (20 mg/ml) and kept for lysis at 55°C for 1 hr. The samples were intermittently mixed every 15 min. After the lysis step, the samples were cooled briefly and an equal volume of chloroform:isoamyl alcohol (24:1) (500 µl) was added and centrifuged at 5,000 × g for 15 min at 4°C. After centrifugation, the top aqueous phase was carefully transferred to another 1.5-ml microtube. The top aqueous phase containing the DNA was precipitated by adding 500 µl of isopropanol and 50 µl of 3 M sodium acetate, mixed gently, and incubated at −20°C overnight. After overnight incubation at −20°C, the DNA was precipitated by centrifuging at 5,000 × g for 15 min at 4°C.
The supernatant was decanted and the pellet was washed twice with 70% ethanol and air-dried. The pellet was then dissolved in TE buffer of 50 µl + 1 µl of RNAse (20 mg/ml) and incubated at 37°C for 30 min.
After incubation, the DNA was precipitated using 10 µl of 3 M sodium acetate and 200 µl of absolute ethanol and incubated overnight at −20°C. The DNA was then precipitated by centrifuging at 5,000 × g for 15 min at 4°C. The supernatant was decanted, and the pellet was washed with 70% ethanol and airdried. The DNA pellet was then dissolved in 50 µl of TE buffer and quantified using 0.8% agarose gel and a NanoDrop 2000 UV-Vis Spectrophotometer.

| Confirmation and sequencing of bacterial symbionts using bacterial specific primers
The extracted sample DNA was confirmed for the presence of bacterial symbionts by PCR using bacterial specific primers (i.e., Univ-0008-a-S-19f and Univ-1528-a-A-17r). Details of primer pairs and their sequences are provided in Table 1. The PCR amplification was performed in a G-Storm GS1 Thermal Cycler. The 20-μl of PCR mixture contained 1× PCR buffer, 1.5 mM MgCl 2 , BSA 10 μg/ml, 0.2 mM dNTP, 5 μM forward and reverse primer each and 1 U Taq polymerase (Takara, Japan), and 100 ng of genomic DNA as a template. The PCR products were resolved in 1% agarose gels by means of electrophoresis at 150 V for 1 hr in 0.5 × Tris-borate-EDTA buffer.
The gels were stained with GelRed™ (Biotium) and photographed under ultraviolet light in a Bio-Rad Gel Doc™ XR + System. The amplified DNA fragment (50 ng/μl) was submitted to Eureka genomics service, USA, for Illumina-based sequencing of the V5 region.

| Sequence processing and clustering
Sequenced reads were quality filtered, dereplicated, abundance sorted, clustered, and chimera removed in the UPARSE pipeline (Edgar, 2013). Quality filtering of reads was conducted with a maximum expected number of errors of less than 0.5, which assumes that the average error of each base in each read is 0.5, and additionally, 8-bp barcodes at the 5′ end were removed, thus producing 143 high-quality base reads. Singletons were also removed. For filtering of chimeras, UCHIME (Edgar, Haas, Clemente, Quince, & Knight, 2011) was used against the RDP classifier training database (v9). Raw sequencing reads were mapped to the chimera filtered database at the 97% similarity threshold and were used to define the OTUs. A total of 264 OTUs along with the mock were obtained. OTUs were trimmed off to remove mock as well as OTUs with a total count of <10, and a final list of 227 OTUs were obtained. For the classification of the bacterial OTUs, megablast (NCBI Resource Coordinators, 2016) was used against the NCBI database. Taxonomy assignment was carried out through manual curation of the blast output with criteria based on the highest bit score, greater than 98% query coverage and at least 97% identity score. OTUs, which failed to suffice all of the criteria, were considered unclassified.

| Data analyses
Because greenhouse conditions, particularly temperature, varied throughout the course of the selection and associated monitoring experiments, leafhopper development, and egg laying fluctuated between generations. We therefore present fitness results relative to leafhoppers feeding on the susceptible recurrent parent (T65) for phase I and relative to GRH2/GRH4-PYL-selected leafhoppers for phase II; however, full results are also included in the Tables S1-S8. Relative fitness of leafhoppers during selection was examined using univariate general linear models (GLM). Similarly, virulence on the respective natal hosts and on GRH2/GRH4-PYL was examined using GLMs after 20 generations of phase I selection. The factor "origin" (referring to the original collection sites in the Philippines: levels = Batangas, Laguna, Quezon, Rizal, and San Pablo) was initially F I G U R E 2 Details of selection experiments with green leafhopper colonies and microbiome sampling. Colonies were selected for 20 generations (G) on each of four rice lines and sampled for microbiome at 20 G. Selection on the natal hosts was continued for a further 6 G with microbiome sampling of T65-, GRH4-NIL-, and GRH2/GRH4-PYL-selected colonies at 26 G. Colonies that were switched to GRH2/GRH4-PYL after 20 G (phase II selection) were sampled for microbiome after 6 generations (simultaneously with 26 G phase I) Phase I 20G + 6 G Phase I 20 G + phase II 6 G

Primer
Sequence

Product size Blast hit
Univ-0008-a-S-19f GAGTTTGATCCTGGCTCAG 1,538 rRNA-16s ribosomal RNA Univ-1528-a-A-17r AAGGAGGTGATCCAGCC 1,538 rRNA-16s ribosomal RNA TA B L E 1 Primers used for confirmation of bacterial symbionts included in all models, but was removed where it had no significant effect. The covariate "rice plant weight" was also initially included in the analyses but removed where it had no effect.
The virulence of leafhoppers on GRH2/GRH4-PYL during each generation of phase II selection was examined using repeated measures GLM with generation as the repeated measure and original natal host (phase I selection) as the main factor. The factor "origin" was initially included in all models, but was removed where there was no significant effect.
The microbiomes of leafhoppers after Phase I and Phase II selection were initially examined at the community level. Prior to statistical analysis, data were square-root transformed (Council et al., 2016). Permutational analysis of variance (PERMANOVA: Anderson, 2001) was used to analyze differences in microbiota composition, at both the OTU and the genus levels. PERMANOVA is a robust semiparametric method that uses permutation techniques to calculate p-values. For Phase I-selected leafhoppers, the factors included in the PERMANOVA analysis were "origin" (random) and "natal" (fixed).
PERMANOVA pairwise tests were conducted to analyze differences between levels of statistically significant factors. The PERMDISP routine was used to examine homogeneity of dispersions (based on mean distance to group centroids), to ensure that dispersions were constant among groups (Anderson, 2006

Pre-selection
Generation 20 NS Exposed to natal hosts application of all GLMs. Where residuals were not normal or homogeneous, we transformed the data as indicated with the results.

| Virulence after phase I selection
Prior to selection, leafhoppers that were recently collected from the field sites were capable of developing on GRH2-NIL and GRH4-NIL, but had poor survival, development, and egg laying on GRH2/GRH4-PYL (Figure 3a,d; Table S1). After 20 generations of selection, all colonies had fully adapted to feed, survive, develop, and lay eggs on their respective natal hosts (Figure 3b,e; Table   S1). Furthermore, leafhoppers selected on both the monogenic (GRH2-NIL and GRH4-NIL) and pyramided (GRH2/GRH4-PYL) lines had a greater capacity to develop and lay eggs on the pyramided line after 20 generations compared to those reared continuously on T65 (Figure 3c,f; Table S1).  because of greater final (generations 4, 5, and 6) adaptation to survive on GRH2/GRH4-PYL by colonies that had been selected on GRH4-NIL during phase I (Figure 4d-f).

| OTU assembly
Sequencing of 16s rRNA-V5 amplicons from the phase I 20 G, Phase I 20G + 6 G, and Phase I 20 G + phase II 6 G samples yielded 60,232-281,854 counts per insect sample. Counts were uniform among the samples. A total of 227 and 208 OTUs were assigned to phase I and phase II samples, respectively (total = 227). Full details of the OTUs and their corresponding taxa are presented in Table S3.

| Leafhopper microbiome after phase I selection
At the OTU level, both "location" and "natal" host had a significant impact on the microbiome community of leafhoppers (Pseudo-F = 18.966, p = .001 and Pseudo-F = 6.0432, p = .002, respectively). "Origin" had a greater effect than "natal" at F I G U R E 5 Nonmetric multidimensional scaling (MDS) showing ordination of the microbiome communities associated with leafhoppers after (a, b) phase I selection (phase I 20 G) with (c, d) relative changes in microbiomes after six further generations on the same natal host (phase I 20 G + 6 G, 0-0, 1-1, and 2-2) or after phase II selection (phase I 20 G + phase II 6 G, 0-2, and 1-2). Plots represent abundance data at (a) the OTU level, and (b) the genus level after phase I selection. Symbols indicate locations and labels indicate (a, b) natal hosts or (c, d) "transitions," 0-0, 0-2, etc., between host plant species, where 0 = T65, 1 = GRH4-NIL, and 2 = GRH2/GRH4-PYL. For further details, see Table S7 the community level as visualized in the MDS ordination plot ( Figure 5a). PERMANOVA pairwise test results indicated that microbiomes were significantly different between all locations (all showing p-values ≤ .05), except between Batangas and Laguna (t = 0.929, p = .557; Figure 5a). Pairwise tests between levels of the "natal" factor showed significant differences between T65 and all other rice lines (all with p-values ≤ .05). More than 80% of the pairwise cumulative dissimilarity between levels was significantly different for both "origin" and "natal" factors because of differences in the abundance of six OTUs (i.e., numbers 1, 8, 10, 12, 17, and 54) each identified as belonging to the Canditatus sulcia clade (Tables S4 and S5).

| Leafhopper microbiome after phase II selection
There were significant differences in the leafhopper microbiome between origins at the OTU level (Pseudo-F = 5.1337, p = .005), as shown in the MDS plot (Figure 5c). Pairwise tests showed statistically significant differences between San Pablo and every other collection site (all with p-values ≤ .05), as well as between Batangas and Laguna (t = 3.177, p = .031). These differences between origins were mostly attributed to five OTUs (no. 1,8,10,12,17) assigned to the Candidatus sulcia clade (Table S7).
Similar results were observed at the genus level, in which "origin" had a significant impact on the microbiome community (Pseudo-F = 5.487, p = .001; Figure 5d).

| D ISCUSS I ON
Our results depict a highly diverse leafhopper microbiome that in virulence adapted colonies were influenced more by population origin than by exposure to resistance genes. However, six bacterial taxa, including the obligate symbiont Candidatus sulcia clade, showed abundance patterns that reflected the virulence of the leafhopper hosts on GRH2/GRH4-PYL (after phase I selection).
These patterns were consistent during a second selection phase when their insect hosts were switched from T65 or GRH4-NIL to GRH2/GRH4-PYL for six generations (phase II selection).

| Endosymbionts associated with leafhopper feeding and virulence adaptation
There is considerable evidence that bacteria and fungi that are facultatively or obligately symbiotic with rice planthoppers and leafhoppers are essential for normal insect development. These often ancient mutualisms have resulted in some symbiotic bacteria, including members of the Candidatus sulcia clade, losing key functional genes that would otherwise allow them to exist independently of the host insect (Bennett, McCutcheon, MacDonald, Romanovicz, & Moran, 2014).
Unlike planthoppers that require YLS for nitrogen metabolism, the N. virescens in our colonies did not possess YSL (J. B. Ferrater, unpublished data). However, they did harbor high densities of obligate symbiont bacteria assigned to the Candidatus sulcia clade. We did not identify these bacteria, but they possibly include strains of Candidatus sulcia muelleri that occur in the closely related N. cincticeps (Noda et al., 2012). These bacteria occur in bacteriocytes within the large bacteriomes of N. cincticeps. Noda et al. (2012) indicated that the bacteriomes include two types of bacteriocyte that each harbored different symbiotic bacteria: Candidatus sulcia muelleri in outer regions, and a further β-proteobacterial symbiont in the inner region of the bacteriomes. A further symbiont assigned to the α-proteobacterial genus Rikettsia also occurs in N. cincticeps (Noda et al., 2012). As with YLS F I G U R E 6 (a) The proportion of total OTUs represented by 16 prominent taxa (i.e., most abundant in phase I 20 G samples and occurring in samples from all five origins) with (b) the abundance of each OTU in the samples.  (Nan et al., 2016), symbiotic bacteria from the bacteriomes are transovarially transmitted from parents to offspring (Kobialka et al., 2016;Michalik et al., 2014). Furthermore, Candidatus sulcia likely plays a role in host nutrition; for example, Candidatus sulcia muelleri can synthesize essential amino acids (Bennett & Moran, 2013) and a recent genomic analysis of the closely related Candidatus arsenophorus nilaparvatae, which occurs in rice planthoppers, suggests that the bacterium plays a role in B vitamin synthesis (Fan, Lu, Ye, Yu, & Zhang, 2016).
In contrast to the often strong evidence that symbionts play a role in host nutrition, evidence for their role in virulence adaptation has been inconclusive. This is largely due to a lack of agreement between the results from similar experiments conducted by different research groups (e.g., YLS: Lu et al., 2004 andFerrater et al., 2015;Lu et al., 2004 andWang et al., 2015;Chen et al., 2011 andHorgan &Ferrater, 2017;endosymbiotic bacteria: Xu et al., 2015 andWang et al., 2015). For example, although Chryseobacterium  Xu et al. (2015) and Wang et al. (2015).
In our study, six bacterial taxa showed positive associations with the virulence of selected leafhopper colonies on the highly resistance GRH2/GRH4-PYL. This included colonies selected on the monogenic to the short selection period (six generations), which only allows partial adaptation to the novel host Rapusas & Heinrichs, 1990;Vu et al., 2014 TA B L E 2 (Continued) Dyella have been recorded as endophytes associated with rice seeds (Hardoim, Hardoim, Overbeek, & Elsas, 2012 (Ferrater & Horgan, 2016). Such bacteria may act as "decoys" that shift the target of induced host plant responses from herbivores to pathogens (Chung et al., 2013). Consistent trends in our results after phase I selection and the predictable changes during phase II selection suggest that the six taxa are associated with leafhopper feeding physiology; however, there is still no clear evidence that they determine virulence and virulence mechanisms remain to be elucidated.

| Mechanisms of leafhopper virulence adaptation and future research
Compared to planthoppers, leafhopper virulence adaptation to resistant hosts has received relatively little research attention.
Evidence suggests that leafhoppers can adapt to feed on resistant rice plants within five to ten generations of selection, but that they adapt more slowly (10-20 generations) to lay eggs on resistant varieties (Heinrichs & Rapusas, 1985;Horgan et al., 2018;Rapusas & Heinrichs, 1990). This suggests that the mechanisms that underlie feeding adaptation are different from those underlying adaptation to egg laying. Asano et al. (2015) indicated that resistance to leafhoppers in GRH2/GRH4-PYL is associated with genes for proteinase inhibitors and cytochrome P450s. Attacks by N. cincticeps on GRH2/GRH4-PYL also produced a strong induction of sesquiterpene volatiles (Asano et al., 2015). Virulence adaptation might therefore include short-term (within a few generations) desensitizing to F I G U R E 7 The relative abundance of key bacterial genera associated with the microbiomes of leafhoppers reared for 20 generations on four natal hosts. The genera are from the (a) Candidatus sulcus clade, (b) Bosea, (c) Dyella, (d) Mycobacterium, and (e) Sandaracinus. Data for (f) Dyanobacter are also indicated: The abundance of this genus was not associated with host plant, but the proportional representation by the taxon in the samples was significantly higher in leafhoppers on GRH2/GRH4-PYL than on T65 (see Table 2). Error bars are indicated, lowercase letters indicate homogenous groups (Tukey, p ≤ .05, N = 5 colonies)