Both host and diet shape bacterial communities of predatory mites

Microbial communities, derived from food, ambient, and inner, can affect host ecological adaption and evolution. Comparing with most phytophagous arthropods, predators may have more opportunities to develop specific microbiota depending on the level of prey specialization. To explore how diet sources affect host microbial communities and vary across predator species, we considered 3 types of predators from Phytoseiidae (Acari: Mesostigmata): polyphagous (Amblyseius orientalis Ehara, Neoseiulus barkeri Hughes, and Amblyseius swirskii Athias‐Henrio), oligophagous (Neoseiulus californicus McGregor), and monophagous (Phytoseiulus persimilis Athias‐Henriot) predatory mites. The polyphagous species were fed on 2 types of diets, natural prey and alternative prey. By using 16S rRNA sequencing, we found that diet was the main source of microbiota in predatory mites, while there was no clear pattern affected by prey specialization. Among 3 polyphagous predators, host species had a larger impact than prey on microbial composition. Unlike A. orientalis or N. barkeri which showed consistency in their microbiota, prey switching significantly affected β‐diversity of bacterial composition in A. swirskii, with 56% of the microbial alteration. In short, our results confirmed the substantial influence of diet on host microbial construction in predatory species, and highlighted species differences in shaping the microbiota which are not necessarily related to prey specialization.


Introduction
Host-associated microbial communities can contribute to host ecology and evolution (Kennedy et al., 2020).The microbiome may be influenced by a variety of factors.Host deterministic selection is generally regarded as a strong force in shaping and stabilizing the bacterial communities of animals (Deb et al., 2019).Diet has also been Correspondence: Guo-Shu Wei, College of Plant Protection, Hebei Agricultural University, Baoding 071000, Hebei Province, China.Email: weiguoshu03@aliyun.com;Bo Zhang and Xuenong Xu, Institute of Plant Protection, Chinese Academy of Agricultural Sciences, Beijing 100193, China.Email: zhangbo05@caas.cnand xuxuenong@caas.cnshown to affect the gut microbiome, directly since food may inoculate bacteria to insect gut, or indirectly by promoting the growth of specific bacteria (Chouaia et al., 2019).The higher microbiota diversity associated with a wider diet width could be perceived as an adaptive trait (Brunetti et al., 2022).
Diet diversity could affect the diversity of microbial communities in generalist arthropods (Kudo et al., 2019).Particularly in phytophagous insects like Lepidopteran caterpillars, the microbial communities present at the surface of food plant influence host microbiota (Hammer et al., 2017).The specialist with narrow diet width may consistently acquire a specific set of microbes from food and form a long coevolutionary process with core microbiota (Russell et al., 2009).Generalist hosts are more likely to stochastically sample a wider range of environmental microbes associated with a variable diet in comparison with dietary specialists (Yun et al., 2014;Shukla et al., 2016).
Compared with phytophagous arthropods, predatory arthropods have a more diverse diet and they are considered to bring few microbes in the gut or entire body (Hammer et al., 2019) or lack core microbiota in a relatively simple gut (Tiede et al., 2017).Changes in the microbiome of spider guts show the pronounced fluctuations in the microbiome associated with prey variation; spider microbial assembly is dictated by the consumed prey, and different prey taxa remodel host microbiota, resulting in abundant shifts of rare microbial taxa in the spider's gut (Kennedy et al., 2020).
As a group of economically important biocontrol agents, predatory mites are commonly classified into several diet preference types, including: Type I, specialized predators of only Tetranychus mite; Type II, selective predators of tetranychid mites and a few other preys; as well as Type III, generalist predators without prey preference (McMurtry & Croft, 1997;McMurtry et al., 2013).Due to the diverse diets, simple structure of their guts and fast transition of ingested substances, predatory mites are considered to change their microbiome frequently in the environment (Pekas et al., 2017).However, to our knowledge, little research has been undertaken on microbiota in response to diet switching in predatory mites.Here we consider whether predator mite microbiota come mainly from the prey they consumed or from inherent species differences which may relate to their prey width.
In this study, we selected a specialized predator (Phytoseiulus persimilis Athias-Henriot), and a selective predator (Neoseiulus californicus McGregor) feeding on general Tetranychus prey (Tetranychus urticae Koch), as well as 3 generalist predators (Amblyseius orientalis Ehara, Neoseiulus barkeri Hughes, and Amblyseius swirskii Athias-Henrio) feeding on 2 respective preys of T. urticae and Carpoglyphus lactis.By investigating the bacterial community composition of predatory mites and their preys, we explored the bacterial communities of predatory mites from the perspective of host specialization and diet variation, addressing the question of how diet influences the microbial community in predatory arthropods.

Mite husbandry
Five predatory mites, including P. persimilis, N. californicus, A. orientalis, N. barkeri and A. swirski, had been fed by T. urticae on bean plants in the Laboratory of Predatory Mites, Institute of Plant Protection, Chinese Academy of Agricultural Sciences over 10 years.Alternative prey C. lactis feeding on yeast was provided by Shoubonong Biotechnology Company (Beijing, China).P. persimilis and N. californicus only survived on T. urticae rather than C. lactis.All mites were reared at 25 ± 1 °C, 70% ± 5% relative humidity, and L : D 14 : 10 photoperiod in incubators (MLR-352H-PC, Ningbo, China).

Diet treatments and sample collection
We selected 100 gravid adults of each of the above 5 predatory mites to lay eggs in small arenas (Yan et al., 2021).Approximately 200 eggs were collected for each species within 24 h.To reduce maternal and environmental effects, we disinfected the surface of predatory mite eggs by soaking in 0.5% sodium hypochlorite for 1 min and rinsing with sterile water 3 times.Eggs were maintained on plastic film (diameter = 7 cm) on top of a soaked sponge (13 cm × 13 cm × 4 cm) in a plastic box (17.5 cm × 17.5 cm × 7 cm) covered by a mesh cloth (diameter = 5 cm, 250 mesh).To explore the microbiota transient after switching prey, the eggs of generalists A. orientalis, N. barkeri, A. swirski were randomly divided into 2 colonies feeding on T. urticae or C. lactis, respectively.We supplemented prey items every 2 d.All predatory mites were fed for 3 generations before collecting adult individuals.For each of 8 groups of predators and 2 groups of prey, we sampled 5-8 biological replicates, each of which involved 15−20 adult mites of both sexes.Samples were thoroughly cleaned with sodium hypochlorite and preserved in 100% ethanol in a −20 °C freezer.
Raw fastq files were filtered by Trimmomatic version 0.38 (Bolger et al., 2014) and USEARCH fastaq_filter parameter in version 11 (Edgar & Flyvbjerg, 2015) to remove low quality reads.Merged sequences were then assigned to operational taxonomic units (OTU) using a 97% similarity cutoff in USEARCH UPARSE (Edgar, 2013).The unique sequences were denoised and de novo clustered into zero-radius operational taxonomic units (ZOTUs) with the unoise3 algorithm.The ZOTU table was normalized to the same number of reads per sample (20 000) to compare data from different samples.Processed sequences were mapped onto ZOTU sequences to calculate the presence and relative abundance of each ZOTU in every sample using the otutab command.The taxonomy was predicted for ZOTU sequences using the SINTAX classifier against RDP training set using confidence threshold of 0.8.

Data analysis
Both αand β-diversity metrics were calculated using the USEARCH package.The normalized abundances of ZOTUs for each sample were used to determine Shannon and Chao indices for each sample.Based on Shapiro-Wilk tests for data normality, we used Mann-Whitney tests or independent t-tests to examine the effect of diet types on microbial diversity in different mite species.Kruskal-Wallis tests and one-way analyses of variance with Tukey's multiple-comparison test were used to compare microbial diversity differences when mites with different diet widths fed on the same diet.For comparing community structure, Vegan 2.4 package was used to calculate the Bray-Curtis dissimilarity metric and we performed permutational multivariate analysis of variance (PERMANOVA), principal co-ordinates analysis (PCoA), and Anosim analysis.The core ZOTUs (definition: ZOTU abundance ≥0.001 per sample; each group had at least 3 replicated abundances greater than 0.001) of different predatory mites was used to build a phylogenetic tree in MEGA X with the neighbor-joining method.A likelihood score was edited on the interactive tree of life (iTOL) website (Letunic & Bork, 2007).UPGMA (Unweighted pair-group Method with Arithmetic Mean) was used to construct a cluster tree of the samples.Network analyses were performed using R and visualized in Gephi (Bastian et al., 2009).Spearman correlation scores were calculated and only robust (Spearman's r > 0.6 or r < −0.6) and statistically significant (P < 0.05) correlations were kept.Differential abundance analysis was conducted in DESeq2 to calculate read counts when predatory mites fed on different prey and the data were visualized in R (Love et al., 2014).We then used Source-Tracker (v1.0) based on Bayesian approach to estimate the microbiota source of mite species from their prey or progenitor (Knight & Kelley, 2011).

Overview of sequencing results
A total of 3 058 293 reads were obtained from 49 samples, averaging 62 414 reads per samples after quality control.The rarefaction analysis indicated that sufficient sampling depth was achieved for all samples.Based on the threshold of 97% similarity and 0.1% relative abundance, the cluster analysis yielded 2 709 ZOTUs belonging to 21 phyla.In general, Firmicutes (43.37%) were the most abundant phylum in all samples, followed by Bacteroidetes (25.72%) and Proteobacteria (22.41%).
There was no obvious evolutionary relationship between core ZOTUs and feeding types of predatory mites (Fig. 3A).Most ZOTUs (51/91) were shared among 3 types of predators.Eight ZOTUs from polyphagous predators were enriched for the clade belonging to the Bacteroidia.Bacterial communities from all predators showed no strong clustering in the PCoA analysis (Fig. 3B).

Diet effect on bacterial communities of polyphagous predators
Although bacterial composition of 2 prey species was different, Clostridia was abundant and consistent in 3 types of predatory mite species regardless of preys (Fig. 4A, Table S1).When altering diet, body color and shape were changed in the 3 polyphagous predators (Fig. 4B).We further tracked bacterial source from either primary or alternative prey based on microbiota similarity.Predatory mites with different feeding habits obtained a similar proportion of bacteria from their prey (Fig. 4C).Different patterns were observed in the 3 polyphagous species (Fig. 4B, D-F).For example, although T. urticae provided 74% of microbes to A. orientalis, the alternative prey C. lactis only transferred 18% of microbes to this predator, while 68% of bacteria were retained from the original colony which fed on T. urticae and 14% came from unknown sources (Fig. 4D).Conversely, 56% of the bacteria of A. swirskii were from the new C. lactis prey, while only 10% was passed on from progenitors feeding on the primary prey, suggesting that microbiota changed more substantially in A. swirskii feeding on new prey compared to the other polyphagous predators (Fig. 4F).After diet change, bacteria in A. swirskii showed obvious diet-mediated differentiation (Fig. S1).
Diet change led to significant changes in the composition and content of bacterial communities (Fig. S2).The bacterial content from 5 orders changed significantly in A. orientalis, including reductions of Alteromonadales, Micrococcales, Xanthomonadales, and increases of Micrococcales and Selenomonadales (Fig. 5A, P < 0.001).The changing relative abundance of the different microbial groups was not consistent across the predators (Fig. 5B, C).

Effect of host and diet on network complexity
For the generalist predators, diet switching had a strong effect on bacterial diversity and network complexity.Common ZOTUs reduced from 71 to 63, resulting in network density changing from 0.40 to 0.49 and modularity from 0.27 to 0.20 after changing prey (Fig. 6A,  B, Table 1).This indicated that bacteria connectivity was  D, E, F) Amblyseius orientalis, Neoseiulus barkeri, Amblyseius swirskii feeding on different diets.PT, Phytoseiulus persimilis feeding on T. urticae; CT, Neoseiulus californicus feeding on T. urticae; OT, A. orientalis feeding on T. urticae; BT, N. barkeri feeding on T. urticae; ST, A. swirskii feeding on T. urticae; OC, A. orientalis feeding on C. lactis; BC, N. barkeri feeding on C. lactis; SC, A. swirskii feeding on C. lactis; T, prey T. urticae; C, prey C. lactis.closer among polyphagous species after feeding on C. lactis.When prey switched to C. lactis, the 67 common ZOTUs in A. swirskii was the lowest value among 3 polyphagous species, while the microbial network density of 0.58 and the average degree of 38.00 was the highest, even though the degree of modularity was not altered at 0.18 (Fig. 6C-E).It indicated that bacterial structure influenced by diet was only evident in A. swirskii rather than being common to all polyphagous species.

Diet is the main source of microbiota in predatory mite hosts
From our results of microbial source tracking and comparisons among different mites, diets are found to shape the microbiota in predatory mites.Consistent with many studies, diet can introduce different bacteria to the hosts (Anderson et al., 2012;Tang et al., 2012;Deb et al., 2019).In mammals, high-protein diet notably reduced microbial diversity and changed the microbial compositions at the phylum level (Li et al., 2022).In panda, butyrate-producing bacterium Clostridium butyricum was more abundant in gut during shoot-eating season than leaf-eating season (Huang et al., 2022).These implied that diets might most likely determine the host-specific intestinal bacterial community.In our study, all predatory mites possessed almost half of microbes from their prey T. urticae.When feeding on T. urticae, bacterial species increased in predatory mites were mostly of plant endophytic bacteria, indicating the microbial transferences in tritrophic relationship among plant, spider mites, and predatory mites.B) UPGMA (Unweighted pair-group Method with Arithmetic Mean) was used to construct the cluster tree of the samples, and the clustering results were combined with the relative abundance of species in phylum.PT, P. persimilis feeding on Tetranychus urticae; CT, N. californicus feeding on T. urticae; OT, A. orientalis feeding on T. urticae; BT, N. barkeri feeding on T. urticae; ST, Amblyseius swirskii feeding on T. urticae; OC, A. orientalis feeding on Carpoglyphus lactis; BC, N. barkeri feeding on C. lactis; SC, A. swirskii feeding on C. lactis; T, prey T. urticae; C, prey C. lactis.
In our hypothesis, we expected to see exactly independent microbial profiles in specialist mites, whereas only one distinct ZOTU was found in comparison with generalist mites, suggesting that a large part of microbiota was provided by diet, with the same feed as other mites.Although some microbes are reported to adjust host diet adaptation (Hosokawa et al., 2007;Tsuchida et al., 2011), we did not find any signal of diet-associated microbes in response to specific diet in predatory mites.Theoretically, impacts of the microbes on their hosts may range along a continuum from strong to weak dependence until there is no association.Fruit flies and stick insects were found to have less dependency on gut microbes for digestion (Shelomi et al., 2015;Erkosar et al., 2017), while this may also be possible in predatory mites.

Host species resulted in microbiota differentiation among generalist predators
In our study, we found that microbiota in polyphagous predators was only stable in A. orientalis and N. barkeri rather than A. swirskii.During the interaction of microbiota and hosts, various factors including stochastic, neutral, or selective processes could determine microbial community compositions of the host (Deb et al., 2019).Among them, host deterministic selection for specific diet-acquired microbiota indicated functional association and strong dependence (Kikuchi et al., 2012).We also highlighted the host-imposed selection of microbiota in predatory mites, so that different host species affected the core microbiota and network structure.For example, microbiota of A. swirskii was strikingly affected by their diet, suggesting fast replacement of microbiome in response to diet alteration.In predatory insects, dragonflies also do not have consistent diet-specific or stage-specific associations with gut bacterial communities, and dietary specialization and spatial variation in bacterial communities suggest a passive process (Deb et al., 2019).Although microbiota in A. swirskii could be changed with diet, host species itself is determinative to the microbial community after prey switching.Different predator species had various adaptations to alternative preys.

Core microbiota of predatory mites
Even though host-associated microbial communities generally present substantial fluctuations with different conditions (Hongoh et al., 2005(Hongoh et al., , 2006;;Moran et al., 2008), hosts often associate with a specific set of microbes as core microbiota (Turnbaugh et al., 2007;Hamady & Knight, 2009;Huse et al., 2012).The taxa and abundance of these core microbes are often stable and conserved.It can be considered as a host species-, genus,-or family-specific trait that may play key roles in determining host fitness and evolutionary potential (Hongoh, 2010;Brucker & Bordenstein, 2012;Tinker & Ottesen, 2016).Flavobacterium was detected stably in spider mites, and represent a core microbiome (Zhu et al., 2020).The relationship between core microbiota and host diets has been studied in insects.For example, Moraxellaceae, Enterobacteriaceae, and Pseudomonadaceae were highly prevalent in the specialist beetle species, while Rickettsiaceae was exclusively associated with generalist beetles (Blankenchip et al., 2018).In our study, Clostridia was abundant and consistent in   Note: OT, Amblyseius orientalis feeding on Tetranychus urticae; OC, A. orientalis feeding on Carpoglyphus lactis; BT, Neoseiulus barkeri feeding on T. urticae, BC, N. barkeri feeding on C. lactis; ST, Amblyseius swirskii feeding on T. urticae; SC, A. swirskii feeding on C. lactis; Polyphagous (C), all polyphagous predators feeding on C. lactis; Polyphagous (T), all polyphagous predators feeding on T. urticae.Average degree: the average number of connected edges per node; Diameter: the maximum distance between nodes with finite distance; Density: the ratio of the actual number of edges to the possible maximum number of edges; Modularity: the nodes classified according to the connection relation.
polyphagous predatory mites regardless of diets.It has been reported to degrade nitrogenous waste uric acid in termite guts (Thong-On et al., 2012), indicating the potential function of nitrogen fixation for host nutrition.
In short, our research revealed that both host species and dietary variation modify the microbial composition of predatory mites.Although predatory mites had stable core microbiota, microbiome fluctuation indicated that predatory mites had weak dependence on their microbial communities which were greatly influenced by diet.The association is closely related to predatory mite species.

Fig. 3
Fig. 3 (A) Phylogeny of bacteria from predatory mites.Venn-kit generated core zero-radius operational taxonomic units (ZOTUs) for each group.Core ZOTUs definition: ZOTU abundance ≥0.001 per sample; each group had at least 3 replicated abundances greater than 0.001.Rooted neighbor-joining phylogeny revealed host-specificity and relatedness among bacteria from several studied species.Branch colors illustrated the taxonomic classification for the bacterial ZOTUs.The taxonomic/trophic level category for the studied hosts was indicated by colors in concentric circles.I: core ZOTUs in Phytoseiulus persimilis; II: core ZOTUs in Neoseiulus californicus; III: core ZOTUs in polyphagous mites (Amblyseius orientalis, Neoseiulus barkeri, Amblyseius swirskii); I & II: the common core ZOTUs shared in P. persimilis and N. californicus; I & III: the common core ZOTUs shared in P. persimilis and polyphagous mites; II & III: the common core ZOTUs shared in N. californicus and polyphagous mites; I & II & III: the common core ZOTUs shared in all predatory mites.(B) UPGMA (Unweighted pair-group Method with Arithmetic Mean) was used to construct the cluster tree of the samples, and the clustering results were combined with the relative abundance of species in phylum.PT, P. persimilis feeding on Tetranychus urticae; CT, N. californicus feeding on T. urticae; OT, A. orientalis feeding on T. urticae; BT, N. barkeri feeding on T. urticae; ST, Amblyseius swirskii feeding on T. urticae; OC, A. orientalis feeding on Carpoglyphus lactis; BC, N. barkeri feeding on C. lactis; SC, A. swirskii feeding on C. lactis; T, prey T. urticae; C, prey C. lactis.

Fig. 4
Fig. 4 (A) Class-level bacterial composition of predatory mites and (B) Source model of mite microbiome.The proportions indicate percentage of microbiota passed from preys or original colonies.Color shades of the arrows indicate the varying proportions.(C) Potential sources of bacterial communities in 5 predatory mites from the prey Tetranychus urticae.PT, Phytoseiulus persimilis feeding on T. urticae; CT, Neoseiulus californicus feeding on T. urticae; OT, Amblyseius orientalis feeding on T. urticae; BT, Neoseiulus barkeri feeding on T. urticae; ST, Amblyseius swirskii feeding on T. urticae; ns, not significant.(D) The bacterial community potential sources of A. orientalis (OC) after diet switching.OT, bacteria retained from original colony; C, prey Carpoglyphus lactis; Unknow, the unknown source.(E) The bacterial community potential sources of N. barkeri (BC) after diet switching.BT, bacteria retained from original colony.(F) The bacterial community potential sources of A. swirskii (SC) after diet switching.ST, bacteria retained from original colony.One-way analyses of variance with Tukey's multiple-comparison test were performed to detect differences: lower-case letters indicate significant differences of the bacterial communities' potential sources among different predatory mites (P < 0.05).

Fig. 6
Fig. 6 Bacterial community connectivity and network complexity.Venn diagrams showing the shared and specific bacterial zeroradius operational taxonomic units (ZOTUs) when polyphagous mites feeding on different diets and bacterial co-occurrence networks based on shared bacterial ZOTUs of polyphagous mites.Different colors in the networks indicated different degrees of modularity.(A) Polyphagous mites feeding on Tetranychus urticae; (B) polyphagous mites feeding on Carpoglyphus lactis; (C) Amblyseius orientalis ZOTUs comparison feeding either on C. lactis (OC) or T. urticae (OT); (D) Neoseiulus barkeri ZOTUs comparison feeding either on C. lactis (BC) or T. urticae (BT); (E) Amblyseius swirskii ZOTUs comparison feeding either on C. lactis (SC) or T. urticae (ST).

Table 1
The characteristics of bacterial co-occurrence networks in different diet treatments.