Effects of Cadmium Stress on Bacterial and Fungal Communities in the Whitefly Bemisia tabaci

Heavy metal contamination is among the most prominent environmental problems in China, posing serious threats to both ecosystem and human health. Among the diverse heavy metal contaminants, cadmium is the most serious. The whitefly Bemisia tabaci is a cosmopolitan pest capable of causing severe damage to a broad range of agricultural crops, especially vegetables. At present, little is known about the effects of cadmium stress on B. tabaci, including on its bacterial and fungal communities. In the current study, we investigated the effects of cadmium on bacterial and fungal communities in whiteflies. Meta-barcode sequencing of the 16S rRNA gene revealed that the whitefly bacterial community contained 264 operational taxonomic units (OTUs) belonging to 201 known genera and 245 known species. The top five most frequent bacterial genera were Rickettsia, Rhodococcus, Candidatus Portiera, Candidatus Hamiltonella, and Achromobacter. Meta-barcode sequencing of the fungal ITS locus revealed that the whitefly fungal community contained 357 OTUs belonging to 187 known genera and 248 known species. The top five most frequent fungal genera were Wallemia, unclassified_f_Dipodascaceae, Apiotrichum, Penicillium, and unclassified_o_Saccharomycetales. Cadmium exposure reduced the fungal OTU richness but increased the bacterial Shannon and Simpson diversity indices in whiteflies. In addition, upon exposure to cadmium, the microbial community composition in whiteflies changed significantly, with increased prevalence of the bacterial genera Rhodococcus and Exiguobacterium and fungal genus Wallemia. Our results indicate that the whitefly microbiota likely contributed to their adaptation and resistance to cadmium and suggested that whiteflies may contain microbes that could help remediate cadmium contamination in natural environments and agricultural fields.


Introduction
Contamination by heavy metals is a common environmental problem and has been attracting increasing attention around the world [1]. In China, heavy metal contamination is among the most prominent environmental concerns [2,3]. In 2014, the China National General Survey for Soil Contamination estimated that 16.1% of the total land surface soil in China had excessive concentrations of heavy metals, with cadmium as the most common heavy metal contaminant [4].
Heavy metals can become hazardous to humans and the environment and can be accumulated through the food chain. Previous studies have shown that heavy metals can also be toxic to plants and invertebrates, including phytophagous insects, threatening the health of organisms across all trophic levels [5,6]. As an important component of the food web in ecosystems, plant-feeding insects are important media for heavy metal dispersal and accumulation in the environment. Indeed, many studies have shown that the accumulation of heavy metals can cause a variety of adverse effects on insects [7]. Interestingly, low levels of some heavy metals can have stimulatory effects, while high levels show almost the three treatment groups. Our data provide insights into the potential mechanism of interactions between whiteflies and their microbiota in the presence of cadmium. This research contributes to understanding the role of the microbiota in the whitefly's adaptation to heavy metals.

Death Rate of B. tabaci after Exposure to Different Cadmium Concentrations
We observed that as cadmium concentration increased, the death rate of B. tabaci also increased ( Table 1). The death rates of B. tabaci after 48 h of cadmium exposure were 74.6%, 65%, 41.35%, 26.05%, and 14.93% for cadmium concentrations of 120, 60, 30, 15, 5, and 0 µM, respectively ( Table 1). The regression equation of cadmium toxicity for B. tabaci was Y = −2.1 + 1.33X, with an estimated LC50 value of 37.93 µmol/L and an estimated LC10 value of 4.12 µmol/L. The estimated 95% confidence interval for LC50 was 35.11-40.97 µmol/L. The coefficient of correlation (R) between cadmium concentration and the whitefly death rate was 0.98 (p < 0.05).

Effects of Exposure to Low and High Concentrations of Cadmium on Bacterial Communities in Whiteflies
The bacterial community compositions in whiteflies exposed to the CK (no cadmium in diet), LD (0.186 mg of cadmium per kg of diet food), and HD (1.71 mg of cadmium per kg of diet food) treatments were investigated by 16S rRNA amplicon sequencing. Our data contained a total of 457,045 reads with an average read length of 407 bp. The obtained sequences were clustered into 264 bacterial OTUs. These bacterial OTUs belonged to 21 phyla, 201 genera, and 245 known species (Tables S1 and S3). Rarefaction curves from the original sequence data suggested that all 12 samples approached saturation, indicating that the sequencing depth for each sample was sufficient to uncover most bacterial diversity within individual samples ( Figure S1A).
Quantitative differences were found among the three treatment groups in the mean numbers of OTUs (Sobs) within each sample (Tables 2 and S3). However, none of the pairwise comparisons among the three treatment groups showed that their differences in Sobs were statistically significant ( Figure 1A, Tables 2 and S3). Similarly, no statistically significant difference was found among the three treatment groups for the ACE and Chao1 indices for their bacterial communities. However, the HD treatment group showed significantly higher Shannon diversity than the CK group (Table 2). In addition, both the LD and HD treatments had a higher Simpson's diversity index than the CK treatment, consistent with the greater overall bacterial diversity in the LD and HD groups than in the CK group.   The bacterial OTUs among the 12 samples were further analyzed through a PCoA based on pairwise Bray-Curtis distances of bacterial OTU compositions. The PCoA showed consistent differences in bacterial OTU compositions among the three treatment groups, with the CK samples clearly separated from HD samples and with the LD samples in between the CK and HD samples (ADONIS test, p < 0.05) ( Figure 1B). At the phylum level, OTUs belonging to Proteobacteria, Actinobacteriota, Firmicutes, and Bacteroidota dominated all three treatments ( Figure 1C). At the genus level, the OTUs of the CK group mainly belonged to the genera Rickettsia (80.67%), Rhodococcus (13.54%), Achromobacter (4.04%), Candidatus_Portiera (1.50%), and Candidatus_Hamiltonella (0.19%); the OTUs of the LD group were mainly in the genera Rickettsia (73.57%), Rhodococcus (18.84%), Achromobacter (5.00%), Candidatus_Portiera (2.18%), and Candidatus_Hamiltonella (0.30%); and the OTUs of the HD group mainly belonged to the genera Rickettsia (69.62%), Rhodococcus (20.21%), Achromobacter (6.06%), Candidatus_Portiera (3.62%), and Candidatus_Hamiltonella (0.36%) ( Figure 1D). Figure 1E shows the distributions of the 30 most abundant genera among the three treatment groups. While the genera Rickettsia, Rhodococcus, Achromobacter, and Candidatus_Portiera dominated all three treatment groups, their relative frequencies among the three groups differed ( Figure 1D,E).

Effective Number of Sequences in Selected Bacterial Genera and OTU Distributions
To further compare the effects of cadmium on bacterial diversity in B. tabaci, we selected the dominant genera in the total sample and compared their effective number of sequences among the CK, LD, and HD treatments. The results revealed significant differences among the CK, LD, and HD groups in the effective number of sequences in the genera Rickettsia, Rhodococcus, and/or Exiguobacterium ( Figure 2). Specifically, the CK group had a more effective number of sequences in the genus Rickettsia than the LD and HD groups. In contrast, the LD and HD groups had more effective numbers of sequences in the genera Rhodococcus and Exiguobacterium than the CK group. Overall, the effective number of sequences in these three genera in the LD group were in between the CK and the HD groups ( Figure 2). A similar pattern was observed for the other three common genera Achromobacter, Candidatus_Portiera, and Candidatus_Hamiltonella, but those differences were statistically insignificant (Figure 2). At the OTU level, there were 100 shared OTUs among the three treatments, and the number of unique OTUs of CK, LD, and HD were 26, 17, and 28, respectively ( Figure S2). The detailed similarities and overlaps of bacteria at three taxonomic levels (phylum, genus, and OTU) among the three treatments are shown in Figure S2. three taxonomic levels (phylum, genus, and OTU) among the three treatments are shown in Figure S2.

Functional Predictions of Bacterial Communities
The potential bacterial metabolic pathways (KEGG) and functions (for example, COG) in each sample were analyzed to better understand the putative roles of the bacterial community in B. tabaci. The relevant results are shown in Figure 3. The KEGG pathway predictions revealed that in all three treatment groups, the relative abundance of "metabolic pathways" was the highest, followed by "biosynthesis of secondary metabolites" and "microbial metabolism in diverse environments" ( Figure 3A). In addition, compared to the CK treatment, cadmium treatments increased the relative abundances of these three KEGG pathways, with the HD treatment showing greater increases than the LD treatment. Similarly, the relative abundances of the pathways "energy metabolism", "lipid metabolism", "metabolism of other amino acids", and "xenobiotics biodegradation and metabolism" also increased in the two cadmium-treated groups compared to the CK ( Figure 3B). The results of the COG functional predictions are displayed in Figure 3C. Here, the "function unknown" category was the most abundant in all samples, followed by "translation, ribosomal structure and biogenesis", and "energy production and conversion". However, there was a limited difference in the relative abundances of these predicted COG pathways among the three treatment groups ( Figure 3C).

Functional Predictions of Bacterial Communities
The potential bacterial metabolic pathways (KEGG) and functions (for example, COG) in each sample were analyzed to better understand the putative roles of the bacterial community in B. tabaci. The relevant results are shown in Figure 3. The KEGG pathway predictions revealed that in all three treatment groups, the relative abundance of "metabolic pathways" was the highest, followed by "biosynthesis of secondary metabolites" and "microbial metabolism in diverse environments" ( Figure 3A). In addition, compared to the CK treatment, cadmium treatments increased the relative abundances of these three KEGG pathways, with the HD treatment showing greater increases than the LD treatment. Similarly, the relative abundances of the pathways "energy metabolism", "lipid metabolism", "metabolism of other amino acids", and "xenobiotics biodegradation and metabolism" also increased in the two cadmium-treated groups compared to the CK ( Figure 3B). The results of the COG functional predictions are displayed in Figure 3C. Here, the "function unknown" category was the most abundant in all samples, followed by "translation, ribosomal structure and biogenesis", and "energy production and conversion". However, there was a limited difference in the relative abundances of these predicted COG pathways among the three treatment groups ( Figure 3C). Heatmap of predicted metabolism pathways. The inferred metabolic pathways included "xenobiotics biodegradation and metabolism", "nucleotide metabolism", "metabolism of other amino acids", "lipid metabolism", "energy metabolism", and "carbohydrate metabolism". (C) Relative abundances of predicted COG functions for bacteria in B. tabaci among the three treatment groups.

Fungal Community Compositions
The fungal community compositions in whiteflies under the CK, LD, and HD treatments were investigated by ITS amplicon sequencing. In total, we obtained 657,407 ITS sequences, and the average length of the ITS sequences was 189 bp. The obtained ITS sequences were clustered into 357 fungal OTUs. These fungal OTUs belonged to 11 phyla, 187 known genera, and 248 known species (Tables S2 and S3). Rarefaction curves from the original sequencing data suggested that all 12 samples reached OTU saturation, indicating that our sequencing depth was sufficient to uncover the total fungal diversity within each sample ( Figure S1B).
Quantitative differences were found among the three treatment groups in their mean numbers of fungal OTUs (Sobs), with the CK group having a greater number of fungal OTUs than both the LD and HD groups ( Figure 4A, Tables 2 and S3). Similarly, the CK group had a higher ACE index and Chao1 index than both the LD and HD groups ( Table  2). In addition, the LD treatment group had a higher ACE index than the HD treatment. Heatmap of predicted metabolism pathways. The inferred metabolic pathways included "xenobiotics biodegradation and metabolism", "nucleotide metabolism", "metabolism of other amino acids", "lipid metabolism", "energy metabolism", and "carbohydrate metabolism". (C) Relative abundances of predicted COG functions for bacteria in B. tabaci among the three treatment groups.

Fungal Community Compositions
The fungal community compositions in whiteflies under the CK, LD, and HD treatments were investigated by ITS amplicon sequencing. In total, we obtained 657,407 ITS sequences, and the average length of the ITS sequences was 189 bp. The obtained ITS sequences were clustered into 357 fungal OTUs. These fungal OTUs belonged to 11 phyla, 187 known genera, and 248 known species (Tables S2 and S3). Rarefaction curves from the original sequencing data suggested that all 12 samples reached OTU saturation, indicating that our sequencing depth was sufficient to uncover the total fungal diversity within each sample ( Figure S1B).
Quantitative differences were found among the three treatment groups in their mean numbers of fungal OTUs (Sobs), with the CK group having a greater number of fungal OTUs than both the LD and HD groups ( Figure 4A, Tables 2 and S3). Similarly, the CK group had a higher ACE index and Chao1 index than both the LD and HD groups ( Table 2). In addition, the LD treatment group had a higher ACE index than the HD treatment. However, though the CK treatment group had an overall higher Shannon index and Simpson's index of diversity (1-D) than the LD and HD treatment samples, those differences were statistically insignificant (Table 2).
However, though the CK treatment group had an overall higher Shannon index and Simpson's index of diversity (1-D) than the LD and HD treatment samples, those differences were statistically insignificant ( Table 2).  Principal co-ordinate analysis (PCoA) based on Bray-Curtis distances between pairs of samples showed that the two main principal co-ordinates explained 76% of the total fungal diversity variation among the 12 samples (ADONIS test with 999 permutations, p < 0.05) ( Figure 4B). However, different from that for the bacterial community, the PCoA failed to separate fungal communities of the 12 samples into three distinct groups based on their cadmium treatments ( Figure 4B). Specifically, there were broad variations among replicates within each of the three treatment groups in their OTU distributions, leading to significant overlaps among the treatment groups. At the phylum level, two phyla, Basidiomycota and Ascomycota, dominated the fungal community in all 12 samples in the three treatment groups ( Figure 4C). Among the three treatment groups, the CK group had relatively more OTUs belonging to Ascomycota than the LD and HD groups. In contrast, the LD and HD groups had relatively more OTUs belonging to Basidiomycota than the CK group ( Figure 4C). At the genus level, in the CK group, the following five genera were the most abundant: Wallemia (56.97%), unclassified_f_Dipodascaceae (18.23%), Penicillium (4.74%), Apiotrichum (3.69%), and unclassified_o_Saccharomycetales (3.22%). In the LD treatment group, the top five dominant genera were Wallemia (70.66%), Apiotrichum (6.42%), unclassified_f_Dipodascaceae (5.68%), Penicillium (3.54%), and unclassified_o_Saccharomycetales (3.01%). In the HD samples, the top five most abundant genera were Wallemia (71.51%), unclassified_f_Dipodascaceae (14.93%), Apiotrichum (3.14%), unclassified_o_Saccharomycetales (1.74%), and Penicillium (1.25%) ( Figure 4D). At the OTU level, there were 47 shared OTUs among the three treatment groups, and the number of unique OTUs for the CK, LD, and HD samples were 164, 66, and 36, respectively ( Figure S3).

Fungi Functional Prediction Analysis
The results of the functional predictions of fungal communities in B. tabaci are presented in Figure 4E. Compared to the relatively well-developed approaches for functional predictions in bacterial communities, the predictions for fungal communities based on ITS sequences are still in the early stage of development. At present, only a few broad categories could be predicted for fungal community functions. Here, among our samples, the "undefined saprotroph" category was the most frequent, followed by "soil saprotroph", "animal pathogen", and "unknown". Among the three treatments, the HD group had the highest prevalence of the "undefined saprotroph" category. Interestingly, the LD group had a higher frequency of the "soil saprotroph" category than both the CK and the HD groups ( Figure 4E).

Discussion
Overall, our study revealed that cadmium treatment had notable influences on both bacterial and fungal communities in whiteflies. For both the bacterial and fungal communities, the mean observed species richness (Sobs) decreased after cadmium treatment. However, the statistical significance of the changes differed between the bacterial and fungal communities. For the bacterial community, the decrease in Sobs was statistically not significant. In contrast, cadmium treatment significantly reduced the fungal Sobs. A similar pattern was observed for two other diversity indices, ACE and Chao1. In contrast, a reverse pattern was observed for the two remaining indices, where cadmium treatment caused significant increases in both the Shannon index and the Simpson's index of diversity for the bacterial community but not for the fungal community in whiteflies.
Multiple indices have been proposed to describe microbial diversity, and most indices take into account both OTU richness and the relative abundances of individual OTUs. However, different indices place different weights on the observed OTU richness and the relative abundances of OTUs [25]. For example, both Shannon and Simpson's indices estimate diversities by treating all OTUs equally and by considering both OTU richness and evenness, but with slight differences in their relative emphasis on either OTU richness (Shannon) or evenness among OTUs (Simpson). The observed increases in the Shannon and Simpson indices in the bacterial community in whiteflies after cadmium treatment suggest that the OTUs in the LD and HD treatments were overall more evenly distributed than in the CK treatment. In contrast, ACE and Chao1 estimate the expected species richness based on observed distributions of OTUs in each sample. Specifically, ACE estimates the expected species number by considering the relative frequencies of only OTUs with fewer than 10 individuals (i.e., <10 sequence reads in each sample in our study) [26]. For those with more than 10 individuals in the sample, only the presence or absence information is considered, while their exact frequencies are not considered in the ACE method. For the Chao1 OTU richness estimator, only the singletons and doubletons are used to estimate the predicted number of species in each sample [27]. Chao1 is frequently used for data sets skewed toward the low-abundance species. Our identification of a significantly higher number of expected fungal species in the CK samples than in the LD and HD samples, as indicated in the ACE and Chao1 indices, is also consistent with those observed for Sobs (Table 2). Indeed, the estimated fungal species richness by ACE and Chao1 were very close to those observed (Sobs) in all three treatment groups ( Table 2). In contrast, on average, the estimated bacterial OTU numbers per sample based on ACE and Chao1 in all three treatment groups were all higher than those observed (Sobs) ( Table 2). These results suggest additional undiscovered bacterial diversity in whiteflies in all three treatment groups.
In our study, Proteobacteria, Actinobateriota, Firmicutes, and Bacteroidota were the dominant bacterial phyla in all three treatment groups. Our results are similar to results obtained by Santos-Garcia et al., who demonstrated that Proteobacteria, Firmicutes, and Actinobacteria were the main bacterial groups of B. tabaci in pepper and watermelon [28]. Similarly, previous investigations reported that B. tabaci harbored Candidatus_Portiera, Rickettsia, Candidatus_Hamiltonella, Arsenophonus, Cardinium, and Wolbachia [21,22]. Bacteria of the genus Candidatus_Portiera are obligatory endosymbionts and have been reported in all B. tabaci analyzed so far [23,28]. Other symbiotic bacteria, for example, Candida-tus_Hamiltonella and Rickettsia, are facultative endosymbionts [22]. It is worth noting that while bacterial genera such as Rickettsia, Candidatus_Portiera, and Candidatus_Hamiltonella have been previously reported in B. tabaci, our study reported for the first time that bacteria in the genera Rhodococcus and Achromobacter are also common members of the microbial community in B. tabaci.
Bacteria of the genus Rhodococcus are Gram-positive, aerobic, and have high G+C contents. They are non-motile and non-sporulating [29]. Rhodococcus can survive in hostile environments, such as those contaminated with polychlorinated biphenyl [30], nitrophenols, trichloroethene [31], and polycyclic aromatic hydrocarbons [32]. Studies have shown that Rhodococcus can degrade and remove aromatic [33], nitrile [34], sulfuric [35], and nitrogenous compounds [36] in the environment. Indeed, Rhodococcus are commonly used to remediate environmental contamination and pollution [37]. In our study, Rhodococcus was significantly enriched in B. tabaci in the HD compared to the control group. As previously reported, Rhodococcus strains are often resistant to cadmium [38]. Our results also suggest that Rhodococcus likely contribute to cadmium resistance in B. tabaci, and its increased abundance in the LD and HD groups was potentially related to the increased cadmium concentration in its diet.
In our study, the whiteflies used in the experiments were kept on pepper plants for over a year. We suspected that Rhodococcus in these whiteflies was obtained from the host plant Capsicum and these bacteria might have helped in reducing the toxicity of secondary metabolites such as capsaicin in pepper plants to these whiteflies. In pepper plants, the biosynthesis of capsaicin involves several substances containing a benzene ring or sulfur. We hypothesize that Rhodococcus could convert these substances to others and block the synthesis of capsaicin to reduce its toxicity to B. tabaci. Controlled experiments with axenic whiteflies and detailed chemical analyses of gut contents are needed to test this hypothesis. In addition to Rhodococcus, we also found an increased abundance of Exiguobacterium in cadmium-treated samples over that in the CK group. As previously reported, strains of Exiguobacterium exhibited an intrinsic resistance to cadmium [39] and could also play a role in reducing cadmium toxicity to whiteflies.
In our samples, Rickettsia was among the most abundant facultative endosymbionts in B. tabaci. Interestingly, Rickettsia could synthesize lysine due to the horizontal transfer of genes from B. tabaci [40]. In B. tabaci, Rickettsia has shown to play a positive role in improving the adult survival rate, egg production, the female adult ratio among offspring, high temperature tolerance, and the development rate [41,42]. We observed that cadmium treatment significantly reduced the relative abundance of Rickettsia, likely a contributor to the increased mortality of whiteflies in the presence of cadmium.
The PCoA of bacterial communities separated the CK and HD treatment groups. However, the PCoA of fungal communities failed to separate the three treatment groups into distinct clusters. Overall, the difference in bacterial community compositions among the three treatment groups was greater than that of fungal communities. The results suggested that cadmium treatments had an overall greater influence on bacterial communities than on fungal communities. The compositional differences in bacterial communities among the three treatment groups resulted in the increased prevalence of several predicted major metabolic pathways among the three treatments. Those pathways included "lipid metabolism", "energy metabolism", "xenobiotics biodegradation and metabolism", and "metabolism of other amino acids", consistent with cadmium having a significant influence on the metabolic pathways of whiteflies.
In all three treatments, Wallemia was the dominant fungal genus in B. tabaci. This Basidomycete genus contains several strongly xerophilic, xerotolerant, and halophilic species [43]. Thus far, Wallemia species have been primarily reported from indoor and outdoor air, dry feed for animals, hypersaline waters, and salted or highly sugared foods, but rarely from insects [43]. Importantly, the majority of Wallemia species can produce toxins and contaminate food [43]. In addition, three species in this genus (W. sebi, W. muriae, and W. mellicola) have been associated with human health problems, such as subcutaneous/cutaneous infections and farmer's lung disease [43]. At present, while the role(s) of Wallemia in B. tabaci is unknown, the increased prevalence of Wallemia from~57% to >70% upon cadmium exposure in this cosmopolitan agriculture pest suggests potentially significant risks that these fungi can pose to human health [44]. Further study is warranted to investigate the interactions between Wallemia and whiteflies and to understand the epidemiology and health impact of these fungi on crops and humans.

Insect Strain
Bemisia tabaci MED (Mediterranean) was originally obtained from a pepper plant (Capsicum annuum L.) field in Changsha, China, in 2020. These whiteflies were reared on pepper cultivar Zhongjiao 4 in an artificial climate box (L14: D10, 26 • C, 80% relative humidity (RH)). The quality of our experimental populations was maintained and monitored according to a previously published method [45].

Toxicity of Cadmium Exposure to B. tabaci
To develop the bioassay for testing cadmium toxicity for B. tabaci adults, whiteflies were directly fed the solutions containing cadmium for 48 h in a special feeding facility. The original feeding solution consisted of 30% sucrose and 5% yeast extract (wt/vol). This was the cadmium-free feeding solution and was used in the control (CK) treatment. The solutions with different cadmium concentrations (5,15,30,60, and 120 µM) were prepared by dissolving anhydrous cadmium chloride (CdCl 2 , 99.99% pure, Sigma-Aldrich, St. Louis, MO, USA) into the original feeding solution. Each 200 µL feeding solution with different concentrations of cadmium was placed in each replicate sample of whiteflies in a designated feeding facility. There were four replications per treatment, and each replicate contained 60 adult whiteflies of mixed sexes. Finally, the treated B. tabaci adults were placed in an artificial climate chamber with a temperature of 26 • C, 80% RH, and a photoperiod of L14:D10. The toxicity of each cadmium concentration to B. tabaci was estimated based on the death rate of B. tabaci at 48 h after cadmium exposure.

Sample Preparation
According to the results of cadmium toxicity for B. tabaci, the final concentrations of 0 (control, marked as CK), LC10 (the concentration of cadmium exposure that caused 10% death rate of whiteflies, at 0.186 mg/kg in artificial diet marked as LD), and LC50 (the concentration of cadmium exposure that caused 50% death rate of whiteflies, at 1.71 mg/kg in artificial diet, marked as HD) were selected for microbial diversity assessment. About 120 B. tabaci adults were included in each of the three treatments (CK, LD, and HD). Each treatment was performed four times. After 48 h, surviving whiteflies were collected, washed with 70% alcohol twice, rinsed with sterile water three times, flash-frozen in liquid nitrogen, and stored in a −80 • C freezer until their microbial diversities were investigated.

Amplicon Sequencing
Total DNA was extracted from the B. tabaci adults using the E.Z.N.A. ® DNA Kit (Omega Bio-tek, Norcross, GA, USA) following the operating instruction's guidance. The primers for the bacterial DNA amplification were 338F (5 -ACTCCTACGGGAGGCAGCAG-3 ) and 806R (5 -GGACTACHVGGGTWTCTAAT-3 ). The primers for the fungal DNA amplification were ITS1F (5 -CTTGGTCATTTAGAGGAAGTAA-3 ) and ITS2R (5 -GCTGCGTT CTTCATCGATGC-3 ). The PCR amplification, the Illumina MiSeq sequencing, and the analysis of sequencing data were performed following the procedures described in our previous study [14]. The raw reads have been deposited in the NCBI Sequence Read Archive (SRA) database (BioProject Number: PRJNA957946; sequence accession numbers SRR24235724-SRR24235735).

Statistical Analysis
The alpha diversity indices were measured by Optifit and the beta diversity indices were measured by Bray-Curtis distance matrices [46]. All the data that needed to be analyzed statistically were tested by one-way analysis of variance (ANOVA) with LSD's range test using SPSS (the Statistical Package for the Social Sciences, version 20.0, Chicago, IL, USA); p values < 0.05 were considered statistically significant. The diversity values were expressed as the mean ± standard error. Principal co-ordinate analysis (PCoA) and hierarchical clustering of bacterial and fungal communities were conducted based on the Bray-Curtis distance among samples. Heatmaps were generated using Heml (version 1.0). Bacterial and fungal functional groups in B. tabaci under different concentrations of cadmium were predicted by PICRUSt2 [47] and FUNGuild [48].

Conclusions
In our study, the influences of cadmium on the bacterial and fungal communities of whiteflies were investigated. Cadmium exposure reduced fungal community alpha diversity but increased the bacterial community Shannon and Simpson's diversity indices. We found that Rhodococcus and Achromobacter were among the dominant genera of the bacterial community in B. tabaci. Compared with the control, the HD treatment showed a significant enrichment of OTUs in the genera Rhodococcus and Exiguobacterium, consistent with their potential roles in the resistance to cadmium and in reducing cadmium toxicity to whiteflies. In addition, cadmium treatment increased the relative abundance of bacterial taxa involved in "lipid metabolism", "xenobiotics biodegradation and metabolism", and "energy metabolism" in B. tabaci. Of special interest is the high and increased abundance of the fungal genus Wallemia upon exposure to cadmium in whiteflies. The high relative abundances of Rhodococcus and Wallemia in our samples suggested potentially essential roles of these microbes for B. tabaci in China. Our results also suggested potential novel strategies for controlling B. tabaci by the targeted disturbance of Rhodococcus and/or Wallemia using chemical and/or biological agents to reduce their contributions to the survival and reproduction of whiteflies and consequently to limit the damage from the whitefly pests on crops. Furthermore, understanding the mechanisms of how Rhodococcus and/or Wallemia reduce cadmium toxicity in whiteflies could also help us develop strategies to minimize the negative effects of cadmium contamination in agricultural fields and improve food security and human health.