Diversity and Regional Variation of Endosymbionts in the Green Peach Aphid, Myzus persicae (Sulzer)

: The green peach aphid, Myzus persicae , is globally distributed and an important pest of many economically valuable food crops, largely due to its ability to transmit plant viruses. Almost all aphids, including M. persicae , carry the obligate symbiont Buchnera aphidicola , which provides essential amino acids that aphids cannot obtain from the phloem of plants themselves. Many aphids also harbor facultative (secondary) endosymbionts, which provide beneﬁts under speciﬁc ecological conditions. In this study, we screened for secondary endosymbionts in M. persicae , with a particular focus on Australian populations where this species is growing in status as a major agricultural pest. We compared 37 Australian M. persicae populations with other populations, including 21 ﬁeld populations from China and 15 clones from the UK, France, Italy, Greece, USA, Spain, South Korea, Chile, Japan and Zimbabwe. No secondary endosymbionts were identiﬁed in M. persicae samples outside of China, despite samples covering a wide geographic range and being collected from several host plant families. We detected two secondary endosymbionts ( Rickettsia , Spiroplasma ) in Chinese samples, although diversity appeared lower than detected in a recent study. We also found very high clonal diversity in Chinese samples based on DNA microsatellite markers in comparison with lower clonal diversity from Australia. These patterns may indicate a higher diversity of secondary endosymbionts (and clonal diversity) in the native range of M. persicae when compared to its invasive range.


Introduction
Bacterial symbionts are widespread in insects, and their symbiotic associations range from obligate mutualism to facultative parasitism [1]. In aphids, the primary (also called obligate) symbiont Buchnera aphidicola provides essential amino acids that aphids cannot obtain from the phloem of host plants themselves [2,3]. Secondary endosymbionts (also called facultative) can influence ecologically important traits in aphids, including resistance to parasitoid wasps [4], tolerance to heat stress [5], and host plant utilisation patterns [6]. An association between endosymbiont persistence and colonization of new plant species has been described by Henry et al. [7] in the pea aphid, Acyrthosiphon pisum (Harris) (Homoptera: Aphididae). Endosymbionts are spread within populations through vertical transmission, whereby the bacteria are passed from the mother directly to her offspring (which is the main pathway) and through horizontal transmission, for example through host feeding and via parasitoids (e.g., [8]).
The green peach aphid, Myzus persicae (Sulzer) (Hemiptera: Aphididae), is globally distributed and one of the most economically important aphid crop pests [9]. It has a host Diversity 2023, 15,206 2 of 14 range of more than 400 plant species [10], multiple methods of causing plant damage [11], and widespread resistance to insecticides [12,13]. In addition, M. persicae is an important vector of many plant viruses, that can cause considerable harm to host plants [14]. Myzus persicae is thought to be of Chinese origin, similar to its primary host, the peach, Prunus persica, and has invaded many countries on every continent, except Antarctica [15,16]. Studies of genetic variation within Chinese M. persicae populations collected from their primary host plant P. persica using DNA microsatellite markers have shown that the genetic structure of these populations involved a split into a southern group and a northern group divided by the Yangtse River. However, the historical demography of M. persicae in China remains unknown [17]. The subsequent invasion of this species has included Australia, where M. persicae was first detected in 1893 [18], and where it is now found in every state and territory. Singh et al. [15] recently used a high-quality chromosome-scale genome assembly with resequenced genomes of 127 globally sampled M. persicae to provide evidence of migration/gene flow between Australia and some populations in Europe and Asia, suggesting multiple incursions.
Artificial endosymbiont infections in aphids can be generated through microinjection of hemolymph across aphid strains or species [19,20], and there is increasing interest in the use of such endosymbiont transfections in pest control. There may be opportunities to release aphid strains generated through transinfection with favourable traits and with endosymbionts able to spread these through wild populations. When applying endosymbiont technology, it is important to understand the status of natural infections in populations of the target aphid, particularly when secondary endosymbionts can vary in incidence between geographic locations as demonstrated in the pea aphid [21]. In M. persicae, a few studies have reported that secondary endosymbiont infections may be very rare [22,23], but Xu et al. [24] observed a high diversity of secondary endosymbionts in several Chinese M. persicae populations.
In order to better understand the diversity and regional variation of endosymbionts in M. persicae and to investigate the ecological and evolutionary factors that might influence the ability of endosymbionts to invade local populations, we characterized the secondary endosymbiont diversity in M. persicae populations and some laboratory clones using 16S metabarcoding and quantitative PCR. We undertook broad-scale sampling of aphids from different host plants throughout Australia, China and several other countries to provide baseline data for the exploitation of endosymbiont technology in M. persicae control. Our work suggests variation in the incidence of secondary endosymbionts between M. persicae within China and elsewhere.

Aphids
Thirty-two M. persicae samples were collected by direct searching for aphids from a variety of host plants from around Australia between 2019 and 2021, while several historical samples, which were collected from the field and had been placed in culture, were also included ( Table 1). Ten individuals from each sample were stored in 100% ethanol and frozen at −20 • C for later molecular analysis. The samples in culture were maintained as asexual lines established from a single female in the laboratory on single bok choy (Brassica napus subsp. chinensis) leaves that were placed in 60 mm petri dishes containing agar (1%). Petri dishes were kept in a controlled temperature cabinet at 11 • C, with a 16L:8D h photoperiod. Twenty-one aphid samples were collected from China (n = 2 to 25) from eight different plant hosts between 2016 and 2021. An additional 15 clones from Europe, the USA, Chile, Zimbabwe, South Korea and Japan (N = 5) were included in this study. These clones were established from aphids collected as part of a global study investigating the evolution of resistance in M. persicae [15], with these aphids being maintained as asexual lineages on individual wombok (Brassica rapa var. pekinensis) leaves in small plastic cups at 20 • C, with a 16L:8D h photoperiod prior to this study. A map displaying the location of all M. persicae samples used here is provided in Figure 1. from Europe, the USA, Chile, Zimbabwe, South Korea and Japan (N = 5) were included in this study. These clones were established from aphids collected as part of a global study investigating the evolution of resistance in M. persicae [15], with these aphids being maintained as asexual lineages on individual wombok (Brassica rapa var. pekinensis) leaves in small plastic cups at 20 °C, with a 16L:8D h photoperiod prior to this study.
A map displaying the location of all M. persicae samples used here is provided in Figure 1.

Clonal Assignment of M. persicae
Aphids from 27 Australian samples and 13 Chinese samples were genotyped to determine the clonal make-up of each sample. Two to 12 aphids from each population were genotyped using 10 previously described DNA microsatellite loci: M35, M37, M40, M49, M55, M63, M86, myz2, myz9 and myz25 [25,26]. DNA was extracted by homogenising individual aphids in a 200 µL solution containing 5% Chelex 100 resin (Bio-Rad Laboratories, Hercules, CA, USA) according to methods described previously [24]. Samples were centrifuged for 2 min at 20,800 g (Eppendorf Centrifuge 5417 C, Hamburg, Germany) and 2 µL of the supernatant was used as template in polymerase chain reactions. Loci were pooled into three groups, labelled with unique fluorophores (FAM, NED, VIC, and PET) and coamplified by multiplex PCR using a Qiagen multiplex kit and an Eppendorf Mastercycler S gradient PCR machine. Genotyping was subsequently performed using a 3730 capillary analyzer (Applied Biosystems, Melbourne, Australia) and product lengths were scored manually using GeneMapper version 4.0 (Applied Biosystems).

16S rRNA Gene Metabarcoding and Quantitative PCR
We used DNA metabarcoding to characterise the microbiome of some M. persicae samples listed in Table 1 (all Australian populations and a few Chinese populations). For these samples, individuals were pooled to provide sufficient DNA for next generation sequencing. Two replicate DNA extractions, each containing a pool of 5 individuals, were performed for each sample using a DNeasy ® Blood & Tissue kit (Qiagen, Hilden, Germany). Metabarcoding targeted the hypervariable V3-V4 region of the bacterial 16S rRNA gene and was carried out by Novogene (Novogene (HK) Co. Ltd., Hong Kong, China), using the universal primers 341F and 806R. Sequence analysis was performed using a standard QIIME2 pipeline [27]. Firstly, primer sequences were trimmed from reads with the cutadapt plugin. Sequence quality filtering and error correction, assembly of paired-end reads, and chimera removal were performed with the DADA2 plugin. DADA2 was then used to group reads into amplicon sequence variants (ASVs), which are analogous to Operational Taxonomic Units (OTUs) clustered at a 100% identity threshold. Background filtering was performed on a sample-wise basis. ASVs that made up less than 0.1% of the reads in a sample were removed from that sample. Taxonomic identity was assigned to ASVs with the classify-sklearn plugin, using a naïve Bayes classifier that had been trained against the SILVA 16S rRNA database [28] (release 132; 99% identity criterion). The identity of endosymbiont ASVs was further investigated with blastn searches (nr/nt database). LFN (low-frequency noise) filters were used to discard variants with low read counts and sequencing contamination (Yang et al., unpublished). An average of 283,084 reads per sample were retained after each of the quality filtering and assembly steps from our 16S metabarcoding data.
Quantitative PCR (qPCR) assays were used to screen for secondary endosymbionts in M. persicae samples which comprised a limited number of individuals and/or had insufficient DNA amounts for metabarcoding. DNA was extracted from individual aphids (variable numbers per population; Table 1)  Two primer sets were applied to amplify markers to confirm the quality of aphid DNA (β-actin as reference gene) [29] and the presence or absence of the target endosymbiont infection ( Table 2). Crossing point (Cp) values of three consistent replicate runs were averaged. Differences in Cp values between the actin and the target endosymbiont markers were transformed by 2n to produce relative endosymbiont density measures. Presence of the endosymbiont amplicon was confirmed by running a standard PCR using the same program and primers with qPCR, running 10 µL of PCR product on a 2% molecular biology grade agarose gel (Scientifix, Springvale, Australia) and observing a clear band with appropriate size for each primer pair. PCR products were then sent for Sanger sequencing (Macrogen, Inc., Geumcheongu, Seoul, South Korea). Sequencing chromatograms were examined and processed with Geneious 9.18 software (Biomatters, Inc., Auckland, New Zealand). The 16S rRNA gene sequencing data were analysed with the program Geneious. A phylogenetic tree was constructed with a neighbour-joining model applied to a genetic distance matrix with the Tamura-Nei2 model implemented with 1000 bootstrap replications in Geneious.

Results and Discussion
In this study, we examined endosymbiont diversity in 73 M. persicae samples from 18 different host plants, focusing on Australia but also including samples collected more widely. Importantly, this included samples from China, which is believed to represent the region where this species originated [17]. Genotyping across 10 microsatellite DNA loci identified 10 distinct clones among the Australian M. persicae populations sampled here and 47 clones among the Chinese populations sampled (Table 1). No secondary endosymbionts were detected within the 37 Australian samples (Table 3, Figure 2), even though individuals from Australian samples represented multiple clones as characterized by microsatellites, were sampled over a broad geographic range, collected from multiple plant families and included multiple color morphs. These patterns are consistent with the lack of secondary endosymbionts in the 15 clones from the UK, France, Italy, Greece, the USA, Spain, South Korea, Chile, Japan and Zimbabwe, and also from other studies of M. persicae outside of China [22,23,33]. We then screened 21 M. persicae populations from China using qPCR, the putative native range, in order to provide a comparison to invaded populations. Rickettsia was detected in four out of 21 samples, each of which was collected from tobacco, Nicotiana tabacum. We also detected Spiroplasma from a single population collected on radish, Raphanus sativus ( Figure 3, Table 3). The incidence of endosymbiont detection in the Chinese samples (5/21) differs from the incidence of detection in Australian samples (0/37) (Fisher's exact test, p = 0.004, IBM Statistics SPSS version 26), even though the sample size of clones from China was smaller (Table 1). A previous screen of diversity in secondary endosymbionts from Chinese populations also identified Rickettsia and Spiroplasma from samples of M. persicae. In addition, another five endosymbionts were detected including Wolbachia, which was abundant across the populations (Table 3) [24]. Although the sample sizes of field collections were not specified in this paper, it is unclear why we failed to detect these other endosymbionts and particularly Wolbachia in our field samples from China.  We then screened 21 M. persicae populations from China using qPCR, the putative native range, in order to provide a comparison to invaded populations. Rickettsia was detected in four out of 21 samples, each of which was collected from tobacco, Nicotiana tabacum. We also detected Spiroplasma from a single population collected on radish, Raphanus sativus (Figure 3, Table 3). The incidence of endosymbiont detection in the Chinese samples (5/21) differs from the incidence of detection in Australian samples (0/37) (Fisher's exact test, p = 0.004, IBM Statistics SPSS version 26), even though the sample size of clones from China was smaller (Table 1). A previous screen of diversity in secondary Figure 2. Secondary endosymbionts in M. persicae populations tested via 16S metabarcoding. Mean relative abundances are shown for each sample included within the study (calculated from 2 technical replicates). Known secondary endosymbionts are shown in color, while the primary endosymbiont, Buchnera aphidcola, and non-endosymbiotic bacteria are shown in grey. The color label on the top of the bar chart represents China, and blue represents Australia. Note that sample codes can be found in Table 1. The group of 'other non-endosymbionts' includes all the bacterial detected except for Buchnera and 8 secondary endosymbionts recorded in aphids (Serratia, Hamiltonella, Regiella, Rickettsia, Rickettsiella, Spiroplasma, Wolbachia and Arsenophonus).
It is possible that some detections in previously published work represents contaminants. Note that we set a sensitive cut-off at 0.1% to remove ASVs with low relative abundance and validated all our positive detections through PCR and Sanger sequencing, whereas previous work has relied on metabarcoding data and set a much lower cut off at 0.005%, with the exception of qPCR screening for Rickettsia. It is also possible that there is geographic or seasonal variation in the distribution of endosymbionts. Although we had multiple sampling points in China, there was only one that overlapped with the samples scored as Wolbachia infected by Xu et al. (M37103, Amygdalus persica, Beijing) [24]. Seasonal changes in Wolbachia infection have been documented in some insects [34] and may occur for other endosymbionts.
Nevertheless, the results from our survey combined with the data from Xu et al. [24] suggest that secondary endosymbionts are more common in the native range of this species than the introduced range, which is also consistent with our failure to detect secondary endosymbionts in a sample of the global clonal collection from Singh et al. [15]. Note that Singh et al. [15] also failed to detect any secondary endosymbionts in whole genome sequence data from >110 fully sequenced globally sampled clonal lines of M. persicae, although this method is less sensitive than metabarcoding of the microbiome.
The phylogenetic analyses of endosymbionts based on sequences obtained from both the 16S data and the Sanger data indicates that the Rickettsia we detected is not closely linked to Rickettsia from other aphids, but instead connects to other arthropod groups including weevils (e.g., Sitona obsoletusand and Liophloeus sp.) and green lacewings (Pseudomallada ventralis) ( Figure S1, Supplementary Data). However, the phylogenic analysis of the Spiroplasma sequences indicated that the secondary endosymbiont we identified was closely connected to an aphid-specific group, which was separate from the Spiroplasma sequences from Drosophila and several other insects, including the mealybug Antonina crawii ( Figure S2, Supplementary Data). Perhaps Rickettsia has a high rate of horizontal transmission which may explain its similarity across disparate taxonomic groups, but at this stage we are unaware of any literature on horizontal transmission in this group and additional molecular analyses based on more comprehensive sequence data are required. endosymbionts. Although we had multiple sampling points in China, there was only one that overlapped with the samples scored as Wolbachia infected by Xu et al. (M37103, Amygdalus persica, Beijing) [24]. Seasonal changes in Wolbachia infection have been documented in some insects [34] and may occur for other endosymbionts.
Nevertheless, the results from our survey combined with the data from Xu et al. [24] suggest that secondary endosymbionts are more common in the native range of this species than the introduced range, which is also consistent with our failure to detect secondary endosymbionts in a sample of the global clonal collection from Singh et al. [15]. Note that Singh et al. [15] also failed to detect any secondary endosymbionts in whole genome sequence data from >110 fully sequenced globally sampled clonal lines of M. persicae, although this method is less sensitive than metabarcoding of the microbiome.  Several reasons might explain the absence of secondary endosymbionts outside of China. Firstly, it may be the case that the colonization process of invasive M. persicae involved a low number of individuals and that by chance these lacked secondary endosymbionts. If this happened, we might expect lower nuclear genetic diversity in colonized regions (or lower clonal diversity in cases where there is no sex). Singh et al. [15] found that both host plant and geography play a significant role in partitioning genetic variation in M. persicae global populations. However, genetic divergence between Chinese M. persicae populations was not higher than divergence detected between Australian M. persicae populations. This may reflect limited sampling of Chinese clones by Singh et al. [15] given that they characterized only 9 opportunistically collected clones from China. Based on microsatellite data, we detected much higher clonal diversity in Chinese populations, with 47 clones detected from 53 individuals, compared to only 10 clones from 270 individuals detected in Australia (binomial test comparing two proportions, z = 14.6, p < 0.001).
Secondly, it may be that secondary endosymbionts have been lost in colonized regions. This might occur if there is imperfect maternal transmission and no selective advantage of individual aphids with these endosymbionts in newly colonized environments. Numerous phenotypes have been associated with aphid endosymbionts that could contribute to selective differences between host aphid lineages, including resistance to parasitoid wasps [4], tolerance to heat stress [5], and host plant utilisation patterns [6], although very few of these have been examined in M. persicae. One exception is the secondary endosymbiont Regiella, which has been associated with parasitoid resistance in M. persicae [35]. Interestingly the natural Regiella strain used in that study was collected from Bacchus Marsh, Australia on wild mustard, Hirschfeldia incana, in 2003 [36]. Clearly this was a serendipitous finding given that our much more extensive collection indicated an absence of Regiella in all 32 M. persicae field populations tested. We have also failed to collect this endosymbiont from aphids from this host plant in subsequent work (Yang et al., unpublished data). Perhaps the Regiella strain was a recent introduction into Australia which has been lost because of a selective disadvantage and/or transmission leakage. Or perhaps the abundance of secondary endosymbiont seasonally changes. In China, there was no association between the presence of secondary endosymbionts and host plant, but Regiella was commonly detected [24].
It is not yet clear if the loss of endosymbionts from the invaded range of a species is a general finding or specific to M. persicae. Perhaps the most extensively investigated comparison of native and invasive ranges involves endosymbionts in the mosquito Aedes albopictus. In this species, Yang et al. [37] showed that two Wolbachia strains were widely distributed across both its native and introduced ranges, with only minor population differences in endosymbiont frequency. In other species, variation in the distribution of endosymbionts has been found across populations, such as the pea aphid A. pisum [21], the spider mite Tetranychus truncates [38], and various plant parasitic nematodes [39]. These comparisons do not provide a contrast between invaded versus native range populations, but there are other instances (for instance in fire ants [40] and in thrips [41]) where there has been a loss of symbionts in the invasive range.
These results have implications for the introduction of secondary endosymbionts for pest control. This is an area of increasing interest, particularly from the perspective of blocking plant viral transmission [42,43] and changing pesticide susceptibility [44]. Based on the current work, a lack of endosymbionts in the invasive range of this species means that any secondary endosymbionts artificially introduced into M. persicae populations from this range will not have to compete with other secondary endosymbionts already present in field populations. Interactions among endosymbionts are known from other work [45] and could otherwise lead to a suppression of newly introduced endosymbionts in a native host. We have recently introduced two endosymbionts from other aphid species into M. persicae (Gu et al., under review) and into the oat aphid, Rhapalosiphum padi [46] from their invasive range. These have proven to be stable introductions, and we are currently exploring if they have favorable characteristics (such as blocking plant virus transmission or decreasing host fitness) that may make them suitable for introductions into pest aphid populations in Australia and elsewhere.
Endosymbionts have also been implicated in color variation among aphids [47], including in a recent study where the endosymbiont Rickettsiella viridis was transferred to M. persicae which modified the aphid body color from light to dark green (Gu et al. under review). Although color morphs were not specifically considered in this study, we did test individuals from Australia that varied considerably in color, ranging from light green, to orange, pink and red. Since we did not find secondary endosymbionts associated with these color morphs, it appears that environmental factors generate color variation in Australian M. persicae.

Conclusions
In summary, there appear to be differences in the diversity (number and variety) of secondary endosymbionts found in M. persicae from invaded and native range (China) populations. Secondary endosymbionts in colonized regions may have been lost due to imperfect maternal transmission, founder events and/or selection associated with different

Data Availability Statement:
The data presented in this study are available from the corresponding authors upon reasonable request.