Whitefly Endosymbionts: Biology, Evolution, and Plant Virus Interactions

Simple Summary Whiteflies feed on plant sap and cause many problems on agricultural crops around the world. Whiteflies have endosymbiotic bacteria in cells inside their bodies that help them to feed on plants. Therefore, the sweetpotato whitefly feeds on a wide range of plants including many vegetable crops, and there are different types of these bacteria which do not all occur in the same whitefly population. In this paper, we focus on endosymbiotic bacteria that are associated with different biotypes of the sweetpotato whitefly with emphasis on their biological characteristics, diversity, and their interactions with whitefly-transmitted plant viruses. This information will be useful to the scientific community for the development of strategies to disrupt these bacteria and provide better whitefly control. Abstract Whiteflies (Hemiptera: Aleyrodidae) are sap-feeding global agricultural pests. These piercing-sucking insects have coevolved with intracellular endosymbiotic bacteria that help to supplement their nutrient-poor plant sap diets with essential amino acids and carotenoids. These obligate, primary endosymbionts have been incorporated into specialized organs called bacteriomes where they sometimes coexist with facultative, secondary endosymbionts. All whitefly species harbor the primary endosymbiont Candidatus Portiera aleyrodidarum and have a variable number of secondary endosymbionts. The secondary endosymbiont complement harbored by the cryptic whitefly species Bemisia tabaci is particularly complex with various assemblages of seven different genera identified to date. In this review, we discuss whitefly associated primary and secondary endosymbionts. We focus on those associated with the notorious B. tabaci species complex with emphasis on their biological characteristics and diversity. We also discuss their interactions with phytopathogenic begomoviruses (family Geminiviridae), which are transmitted exclusively by B. tabaci in a persistent-circulative manner. Unraveling the complex interactions of these endosymbionts with their insect hosts and plant viruses could lead to advancements in whitefly and whitefly transmitted virus management.


Introduction
Whiteflies (Hemiptera: Aleyrodidae) are globally significant agricultural pests and virus vectors causing direct and indirect damage to crops with estimated losses totaling billions of dollars (US) annually worldwide [1]. The notorious whitefly species Bemisia tabaci (Gennadius) is of particular Table 1. Taxonomic, biological, and genomic characteristics of the primary and secondary endosymbionts of cryptic Bemisia tabaci. Genome information was obtained from NCBI (https://www.ncbi.nlm.nih.gov/) from the earliest and most complete genome assembly for each endosymbiont where available. When the sequence publication reference was not given, the BioProject accession number was provided. In addition to its primary endosymbiont, B. tabaci can also harbor secondary endosymbionts, of which bacteria in seven genera including Hamiltonella, Arsenophonus, Rickettsia, Wolbachia, Cardinium, Fritschea, and Hemipteriphilus (Table 1) have been identified. These facultative endosymbionts demonstrate a variety of relationships with their host, ranging from parasitism, which may be highly subject to environmental conditions [24,25], to mutualism, with their products incorporated in host metabolism pathways [26]. Enterobacterium Ca. Hamiltonella defensa (hereafter Hamiltonella) may be more the latter type given its involvement in essential amino acid biosynthesis pathways and production of B vitamins [26,27]. However, it is also suggested to be a nutritional parasite competing with Portiera for host derived resources [25]. Arsenophonus, another enterobacterium, may also have a more mutualistic type of relationship with B. tabaci as it too is suggested to help provide essential nutrients [12,28]. Both Hamiltonella and Arsenophonus appear to be fixed within B. tabaci populations [28]. In contrast, the alphaproteobacteria Rickettsia and Wolbachia, class Rickettsiales, are generally not fixed in populations and may represent more dynamic, parasitic-like associations with B. tabaci [13,14,29]. Two groups of the genetically diverse Rickettsia genus have been found associated with B. tabaci: Rickettsia Bellii, which includes many pathogenic strains [30][31][32], and, recently, Rickettsia Torix [23]. Another alphaproteobacterium in the class Rickettsiales that is not fixed in B. tabaci populations, Ca. Hemipteriphilus asiaticus (hereafter Hemipteriphilus), is closely related to a mite-borne human pathogen in the genus Orientia [17,33]. Hemipteriphilus so far appears to be restricted to the bacteriocytes, but its function and effect on host fitness is unknown. Candidatus Cardinium hertigii (Bacteroidetes) has a reduced and dynamic genome that lacks robust cofactor and amino acid biosynthesis capability and, therefore, is not suspected to be important to host nutrition metabolism [15,34]. The final known secondary endosymbiont, Ca. Fritschea bemisiae (hereafter Fritschea) (Parachlamydiales), occurs in the bacteriocytes, but little is known about its function in B. tabaci [16]. Altogether, much remains to be elucidated about the specific functions of secondary endosymbionts, especially when considering the diverse makeup of endosymbiont communities in B. tabaci.

Morphology, Localization, and Transmission
Observations of the structure of whitefly bacteriomes, also called mycetomes, and endosymbionts began with comparing those of B. tabaci and T. vaporariorum [35]. Bacteriomes, which are yellow-orange in color and easily observable in late immature stages, are specialized organs composed of large cells referred to as bacteriocytes, or mycetocytes, which house intracellular endosymbiotic bacteria [19,35]. Whitefly bacteriocytes were observed to contain large nuclei (sometimes multi-nucleate), mitochondria, ribosomes, and granular bodies, along with prokaryotic organisms of two described types, pleomorphic and coccoid [35]. While the structure of the more abundant pleomorphic type was observed to lack a distinct cell wall in both species, that of the coccoid type was different, with a much thicker cell wall in T. vaporariorum. These results were supportive of earlier 16S rDNA sequencing of whitefly bacteriomes [36] suggesting two endosymbiotic gram-negative eubacteria in B. tabaci. A follow up study comparing populations of B. tabaci MEAM1 (Middle East Asia Minor 1; previously, B biotype/B. argentifolii,) and NW (New World) group whiteflies (including A biotype and Jatropha biotype populations; see [37,38]) demonstrated differences in the cryptic groups [39]. Bemisia tabaci MEAM1 appeared as observed in Costa et al. [35], and the NW group populations harbored an additional coccoid microorganism with differences in distribution and frequency observed between the NW (A biotype) and Jatropha biotype populations. In both studies [35,39], the coccoid bacteria were sometimes observed in various states of degradation, indicating a possible mechanism of endosymbiont control by the host. Localization studies have since confirmed that pleomorphic Portiera is housed in the bacteriocytes, and secondary endosymbionts are located in the bacteriocytes or other host organs [13,40]. The early reports by Costa et al. [35,39] laid the groundwork for ultrastructural studies on the whitefly bacteriome, but few studies have followed up on this work. One exception, however, addressed the observed lack of cell wall in the primary endosymbiont, which is present in all other primary endosymbionts studied, providing evidence of such by electron microscopy [41].
Localization studies consistently place Portiera within the bacteriome; however, secondary endosymbionts exhibit more varied localization patterns [40]. Using fluorescence in situ hybridization (FISH) analysis, Gottlieb et al. [40] localized five of the seven known B. tabaci secondary endosymbiont genera in different life stages of MEAM1 and MED (Mediterranean) populations. Hamiltonella was identified in both MEAM1 and MED populations, and it was found in patches inhabiting only the bacteriome in all observed life stages (eggs, nymphs, and adults). Similarly, in the MED population Arsenophonus was located only in the bacteriome, but its distribution was more often surrounding the nuclei of bacteriocytes. Typically, Arsenophonus is observed exclusively in the bacteriome; however, Rana et al. [42] reported localization of this bacterium in the bacteriocytes, salivary glands, and midgut of B. tabaci Asia II. In eggs, nymphs, and adults of the MED population, Wolbachia was detected inside the bacteriome and concentrated along its circumference [40]. In some adult females, it was also found in the abdomen outside the bacteriome. Cardinium in the MED population and Rickettsia in MEAM1 and MED had wider distribution patterns than the other symbionts [40]. Cardinium was detected in the bacteriocytes in all stages, in higher concentrations in the abdomen and head in nymphs, in abdominal cells suspected by the authors to be fat cells in adult females, and widely distributed throughout the abdomen of adult males, excluding the rectal sac. Rickettsia interestingly had two very different distribution patterns, either strictly within the bacteriome concentrated around the circumference (confined phenotype) or, as previously observed [13], distributed throughout the body excluding the bacteriome (scattered phenotype) [40]. The specific phenotype was not correlated with the cryptic group nor collection location, and two populations (one MEAM1 and one MED) presented mixed phenotypes. Consistent with that observation, Shi et al. [43] found Wolbachia to have both confined and scattered phenotypes in B. tabaci AsiaII7. Similar localization patterns have been demonstrated for endosymbionts in other whitefly species, including Arsenophonus and Hamiltonella in T. vaporariorum [44] and Rickettsia, Wolbachia, and Cardinium in T. vaporariorum [45]. Thus far, bacteriocyte restriction has been observed for Fritschea in MEAM1 and NW2 [46] and Hemipteriphilus in China 1 [17]. Altogether, studies demonstrate that all secondary endosymbionts can co-occur with Portiera in the bacteriome and that Hamiltonella, Fritschea, and Hemipteriphilus are restricted to the bacteriocytes, while Arsenophonus, Rickettsia, Wolbachia, and Cardinium can occur in other whitefly tissues.
All B. tabaci endosymbionts are vertically (transovarial, maternal) transmitted [7]. Costa et al. [47] described the process of vertical transmission of the bacteriocytes. In B. tabaci, a single bacteriocyte cell is transferred to the plasma of an individual developing oocyte [47,48]. This process is different in other whitefly species, in which several cells are transferred [19]. When secondary endosymbionts locate within tissues other than the bacteriome (e.g., Rickettsia), the endosymbiont will temporarily colocalize with Portiera and any other endosymbionts within the bacteriocyte [44,49]. As the bacteriocyte enters and moves to the center of the egg, the endosymbiont will then exit the bacteriocyte and inhabit the egg cavity, while the bacteriome-inhabiting endosymbionts remain in the bacteriocyte. Some secondary endosymbionts are also capable of horizontal transmission, which can be plant-mediated via feeding (e.g., Rickettsia and Wolbachia; [50][51][52]) or mechanical via parasitoids [53,54]. Evidence of horizontal transmission of secondary endosymbionts of B. tabaci has been supported by phylogenetic incongruence between host and endosymbionts [24]. Ahmed et al. [24] demonstrated a lack of congruence between Wolbachia and B. tabaci and partial congruence for Arsenophonus and Cardinium with B. tabaci, suggesting that Wolbachia may be more adapted to horizontal transmission events. Plant-mediated horizontal transmission has been demonstrated for Rickettsia, which established in phloem cells after whitefly feeding and was acquired by Rickettsia-free whiteflies feeding on the same cotton leaf [50]. Host plant mediation of horizontal transmission was also found for Wolbachia [51] and between B. tabaci species (MEAM1 and MED) for Rickettsia [52]. Further, parasitoid-mediated horizontal transmission of Wolbachia and Rickettsia has been documented. Wolbachia was stably acquired by whiteflies after non-lethal probing by Eretmocerus sp. nr. furuhashii (Hymenoptera: Aphelinidae) contaminated with Wolbachia from whiteflies harboring the bacterium [53]. Whiteflies that acquired Wolbachia via the parasitoid had enhanced fitness. Other work has revealed that the Rickettsia and Wolbachia detected in field collections of B. tabaci nymphs and the emerging Eretmocerus parasitoids were identical haplotypes, suggesting that parasitoid-mediated horizontal transmission may be occurring between these species in the field [54].

Effects on Whitefly Fitness
Theoretically, secondary endosymbionts should provide some benefit to or increase the fitness of their hosts in return for the cost of harboring said endosymbiont. While studies have reported beneficial effects (e.g., [28,[55][56][57][58][59][60]) and negative effects (e.g., [61][62][63]) of harboring certain endosymbionts, there remains a paucity of information about their fitness effects in B. tabaci, particularly given the species' cryptic diversity and complex endosymbiont composition. To study the effects of specific endosymbionts on whitefly fitness, colonies that differ only in the presence or absence of a particular endosymbiont are ideally needed. In this respect, antibiotics have been used to obtain endosymbiont-free colonies. Endosymbiont elimination studies using antibiotic treatments of whiteflies have provided some clues as to the effects of endosymbionts on host fitness. Costa et al. [64] demonstrated that antibiotics have adverse effects on bacteriome size, number of bacteriome-inhabiting microorganisms, whitefly oviposition, adult growth, and progeny development. Antibiotic treatments of B. tabaci without identification of the specific bacteria reduced or eliminated demonstrated that antibiotics with different modes of action had differing effects on whitefly fitness. For example, antibiotics with action against bacterial protein synthesis (tetracycline and rifampicin) reduced host growth and development, while antibiotics with action against bacterial cell walls (penicillin, ampicillin, and lysozyme) did not have an observed effect on fitness [65]. Antibiotic selection is important because endosymbionts have differing susceptibilities [66][67][68]. Specific elimination of secondary endosymbionts via antibiotics has demonstrated positive or negative effects on host fitness [55,62,[69][70][71][72]. However, some work has shown that selective, stable elimination of secondary endosymbionts may not be feasible for some populations [73]. Further, gut bacteria may have more influence on whitefly adaptation to host plants than maternally inherited endosymbionts [74]. This highlights a key question and potential issue with antibiotic elimination studies; that is, how many generations after the antibiotic treatment and elimination of a particular endosymbiont is it appropriate to assume that the gut bacteria have reestablished in endosymbiont populations? To perform reliable comparative studies considering questions of endosymbiont effects on fitness, it may be key that the gut microbiota is as similar in composition as possible between compared populations. Perhaps the results of some elimination studies have been more attributable to gut bacteria than to endosymbionts.

Bemisia tabaci Cryptic Species
The diversity and evolution of whiteflies has been a subject of extensive interest since the discovery of an exotic B. tabaci in the United States and the subsequent revelation of B. tabaci as a cryptic species composed of multiple sibling species [75][76][77]. Bemisia tabaci is reported to comprise over 40 species variants, which are differentiated by 3.5-4% divergence in the sequence of their mitochondrial cytochrome c oxidase I gene [37,38,78,79]. However, the exact bounds of distinct species are yet indeterminate, particularly because of the difficulty in proving species separation as defined by the biological species concept in field populations of B. tabaci. Despite the varied and evolving interpretations of B. tabaci species status over the years, many studies have demonstrated biological differences among cryptic B. tabaci, including, among others, host range [80], insecticide resistance [81], virus transmission [82], and composition of endosymbionts [39].

Primary Endosymbiont Ca. Portiera aleyrodidarum
The symbiosis of whiteflies with their primary endosymbiont Portiera is estimated to have begun with a single infection in a common ancestor of psyllids and whiteflies [10,83]. Phylogenetic studies reveal a high level of congruence between Portiera and host phylogenies [10], a pattern commonly seen in coevolved endosymbiont-insect host associations [84][85][86][87]. Co-speciation is further demonstrated by genomic decay in endosymbiont genomes, characterized particularly by the loss of genes unrelated to the essential functions the endosymbiont provides the host [88]. This pattern is especially pronounced in Portiera in B. tabaci, although unlike most highly reduced primary endosymbiont genomes, it has extended non-functional intergenic regions and tandem repeats [89,90]. Compared to Portiera in T. vaporariorum, the B. tabaci endosymbiont has notably lost several genes involved in DNA replication, recombination, and repair [89]. While the Portiera genomes of T. vaporariorum and two other Aleyrodids, Aleurodicus dispersus Russell and Aleurodicus floccissimus (Martin, Hernandez-Suarez, and Carnero), are syntenic, indicating evolutionary stasis, the genome of B. tabaci Portiera has undergone extensive rearrangements [90].
A counter-intuitive finding in many studies on the genomes of obligate intracellular endosymbionts of sap-feeding insects is the loss of genes involved in the biosynthesis of the 10 essential amino acids [91][92][93], the main function of the primary endosymbiont symbiosis with the insect host. In B. tabaci and other phytophagous hemipterans, several of these genes have been lost or pseudogenized, therefore requiring complements of those genes to be expressed by either the host or another symbiont [9,26,27,90,94,95]. Comparison of four B. tabaci Portiera genomes revealed that approximately half of their metabolism genes are involved in amino acid biosynthesis [26]. However, all genomes lacked 12 essential amino acid genes and have at least 2 pseudogenes in these pathways. Genomic and transcriptomic analysis showed that these deficiencies were compensated by host genes (including horizontally acquired genes of prokaryotic origin), secondary endosymbiont genes (Hamiltonella in MEAM1), and other bacterial genes [26]. Whitefly-Portiera-Hamiltonella pathway complementarity was demonstrated for B. tabaci MED, as well [27]. The genomic reduction of Portiera appears to be both compensated for and facilitated by genomic complementation in the host and other symbionts [26,27].

Secondary Endosymbionts
Secondary endosymbionts can be intricately involved in the essential metabolic processes of its insect host [26][27][28]. However, evidence of co-speciation is generally lacking for insect hosts and secondary endosymbionts. This is indicated by a lack of congruence in phylogenetic trees of host and secondary endosymbionts [7,10], by the wide variation and inconsistency in secondary endosymbiont-host associations (detailed for B. tabaci in a later section), and by evidence of multiple acquisitions of secondary endosymbionts [14,96]. The inconsistency of secondary endosymbiont-host associations is likely in part due to the nutritional and metabolic burden that they can place on the host and the extent to which these associations are parasitic or mutualistic [24,25].
Bemisia tabaci harbors the highest known diversity of secondary endosymbionts among phloem-feeding hemipterans [97]. Secondary symbiont infection patterns vary worldwide and regionally, with differences among populations and between native and invasive B. tabaci species. One of the earliest studies on endosymbiont composition in cryptic B. tabaci species demonstrated that US native NW populations had two distinct C-type (secondary) endosymbionts within their bacteriocytes whereas the invasive MEAM1 had only one [39]. This early report supported the hypothesis that the development of B. tabaci biotypes (now cryptic species) may have been facilitated by secondary endosymbionts and differences in their composition among the biotypes (see references in [39]). Since then, the prevalence, diversity, and phylogenetic relationships of endosymbionts in B. tabaci have been well documented from specific regions worldwide (e.g., [33,[98][99][100][101][102][103][104][105][106][107]) and on a global scale among a variety of cryptic species (e.g., [11,12,79,[108][109][110]).
Recently, using 16S rRNA sequences with associated geographical information retrieved from GenBank, a study on the global diversity of B. tabaci endosymbionts showed that three of the seven described genera of facultative endosymbionts are widespread in B. tabaci cryptic species [79]. Rickettsia, Wolbachia, and Cardinium were found in a wide diversity of B. tabaci species, including MEAM1, MED, and certain Asia, SSA (Sub-Saharan Africa), and NW cryptic species [79]. Of course, the specific endosymbionts present varied among and even within populations of these groups. According to this study, Hamiltonella and Fritschea were associated with only MEAM1, MED, and NW species, Arsenophonus was not associated with MEAM1, and Hemipteriphilus was associated with China 1 only. Further, phylogenetic analyses of the secondary endosymbionts demonstrated evidence of more than one genetic group for some endosymbionts (including 3 Rickettsia, 2 Arsenophonus, 2 Wolbachia, and 4 Cardinium) and only one group of Hamiltonella, which was conserved among MEAM1, MED, and NW species.
Another recent study, using PacBio sequencing of 16S rRNA bacterial genes from field populations of 23 cryptic species, also revealed wide genetic diversity in some secondary endosymbionts [110]. This study demonstrated 2 operational taxonomic units (OTUs) for Rickettsia, 5 OTUs for Arsenophonus, and 2 OTUs for Wolbachia [110]. One OTU each was found for Hamiltonella, Cardinium, Fritschea, Hemipteriphilus, and Portiera. Only one population of each B. tabaci cryptic species was sampled, however, Wang et al. [110] found Hamiltonella and, notably, Hemipteriphilus associated with more cryptic species than Kanakala and Ghanim [79] found. With the largely unpredictable nature of endosymbiont composition and diversity among B. tabaci cryptic species and populations, these studies highlight the improbability of secondary endosymbiont facilitated speciation of B. tabaci, the hypothesis proposed for cryptic B. tabaci early in the discovery of biotypes associated with different secondary symbionts.

Modes of Virus Transmission
For us, one of the more exciting avenues of research on hemipteran endosymbionts is the elucidation of the interactions of these bacteria with the plant viruses transmitted by their host. Bemisia tabaci transmits viruses in three distinct manners: non-persistently, semi-persistently, and in a persistent-circulative manner [5,111]. A fourth mode of biological transmission by hemipterans, persistent-propagative, is not known to occur in whiteflies. Based on the duration of infectivity after vector acquisition and virus route of movement within the vector, viruses are grouped as circulative or non-circulative [111]. Non-persistent and semi-persistent, both non-circulative transmission types, describe interactions of the virus with its insect vector that are characterized by localization and retention of the virus within the vector's stylets or foregut before transmission. Whereas, persistent-circulative and persistent-propagative, both circulative modes of transmission, are characterized by virus acquisition into the vector's hemolymph via the filter chamber, midgut, or hindgut and circulation of the virus to the salivary glands for subsequent transmission, with propagative viruses having the additional capability to replicate in the vector. Bemisia tabaci is notorious for its persistent-circulative transmission of begomoviruses (family Geminiviridae), for which this species is the only demonstrated vector. Bemisia tabaci also transmits viruses in the genera Crinivirus (family Closteroviridae), Ipomovius (family Potyviridae), Carlavirus (family Betaflexiviridae), and Torradovirus (family Secoviridae) in a non-circulative manner [4,5,[111][112][113][114]. The manner by which plant viruses are transmitted may inherently determine, at least in part, whether or not there is an interaction between vector endosymbionts and their metabolites and the insect-transmitted virus. That is, circulative viruses may be more likely to have such interactions than non-circulative viruses, which are more transiently associated with the vector and are not acquired into the hemolymph.

Endosymbiont Expressed GroEL
The first reported study to suggest a possible role of endosymbionts in transmission of plant viruses was van den Heuvel et al. [115]. This study demonstrated that potato leafroll virus (genus Polerovirus, family Luteoviridae) binds to symbionin (homolog to Escherichia coli GroEL chaperonin protein; [116]) expressed by the primary endosymbiont of Myzus persicae (Sulzer) (Hemiptera: Aphididae). Additional studies with aphids and circulative luteoviruses further demonstrated this interaction, suggesting it as a means of virus protection from degradation in the hemolymph by the vector immune system and emphasizing the probable role of endosymbiont expressed GroEL-like protein in the virus transmission process [117][118][119].
The differences between luteoviruses and geminiviruses are many, but begomoviruses share a similar persistent-circulative mode of transmission, with a key barrier to transmission being persistence in the vector hemolymph subsequent to passing through the gut tissue [120]. While it was the primary aphid endosymbiont, Buchnera, that was demonstrated to express the GroEL-like chaperonin in M. persicae, and Buchnera is unrelated to the primary endosymbiont of whiteflies, a secondary endosymbiont related to Buchnera is present in some whiteflies [36]. This led researchers to investigate the possibility of whitefly endosymbiont GroEL-like chaperonin (previously symbionin; hereafter simply referred to as GroEL) expression and interaction with the begomovirus tomato yellow leaf curl virus (TYLCV). Morin et al. [18] demonstrated that GroEL was expressed by the coccoid secondary endosymbiont (C-type; [35,39]) of B. tabaci B biotype (MEAM1), that TYLCV showed an affinity for GroEL, and that TYLCV transmission was significantly (>80%) reduced with the introduction of anti-GroEL antiserum. Both insect transmissible (TYLCV-Israel) and non-transmissible (abutilon mosaic virus from Israel; AbMV-Is) begomoviruses interacted with GroEL in vitro, indicating the inability of whitefly transmission of AbMV-Is (and lack of AbMV-Is detection in whitefly hemolymph) is likely due to an inability to cross the gut epithelial barrier rather than a lack of affinity to B. tabaci GroEL [121]. Similarly, several luteoviruses bind to the GroEL chaperonins produced by both vector and non-vector aphid species in vitro [122,123]. Altogether, these studies show the importance of GroEL-begomovirus/-luteovirus interactions for virus protection in the vector hemolymph, where insect defenses may otherwise degrade the foreign invader, as one component of transmission competency, but other factors ultimately determine virus-vector transmission specificity.

Hamiltonella
Since the early work on GroEL-begomovirus interactions, the number of secondary endosymbionts discovered in whiteflies has increased from the two previously described C-types in B. tabaci MEAM1 and NW [35,39] to now seven different genera identified in different combinations and localizations in B. tabaci cryptic species. Studies to determine which secondary endosymbionts and to what degree are responsible for expression of GroEL correlated with virus transmission began with looking at populations of B. tabaci MEAM1 and MED from Israel. At that time, Rickettsia was known to occur in the Israeli MEAM1 and MED, Hamiltonella was identified in MEAM1 only, and Wolbachia and Arsenophonus were identified in MED only [13,98,124]. While all secondary endosymbionts, as well as the primary symbiont Portiera, expressed GroEL in the hemolymph of both MEAM1 and MED, only GroEL expressed by Hamiltonella in the MEAM1 biotype interacted with TYLCV resulting in transmission (Table 2; [124]). MED transmitted TYLCV at a significantly lower rate, and evidence of TYLCV-GroEL interaction in vitro was not found. Early on then, Hamiltonella was implicated in assisting begomovirus transmission via protection from the insect immune response, and other studies have emphasized this relationship and correlated this interaction with increased virus transmission by other B. tabaci cryptic species (MED in China and Brazil, respectively; [125,126]). In a comparison of the native NW2 B. tabaci with exotic MEAM1 and MED in Brazil, the Hamiltonella GroEL chaperone sequence of NW2 was found to be deficient in three amino acids [127]. Given the interaction of Hamiltonella GroEL with whitefly transmitted viruses, this was suggested as one possible factor affecting differential transmission of viruses, including a begomovirus and non-circulative crinivirus and carlavirus, by NW2 and MEAM1 in Brazil [127,128]. Interestingly, Bello et al. [126] demonstrated that MED B. tabaci with higher frequencies of Hamiltonella transmitted the non-circulative viruses cowpea mild mottle virus (genus Carlavirus) and tomato chlorosis virus (genus Crinivirus) at higher rates. Therefore, the Hamiltonella-plant virus interaction is not restricted to B. tabaci MEAM1, and it has implications in the epidemiology of begomoviruses worldwide as well as some non-circulative viruses.

Rickettsia
Bemisia tabaci-harbored Rickettsia has also been implicated in virus interactions within the vector (Table 2). Isoline colonies of Rickettsia+ MEAM1 B. tabaci, which also harbored Hamiltonella, acquired TYLCV at a higher rate and transmitted the virus with greater efficiency [129]. When whiteflies were highly infected with Rickettsia localized in the midgut, TYLCV virions were detected at higher titers in the filter chamber (along with the anterior midgut, where begomovirus virions cross into the hemolymph; [132]), likely where higher concentrations of the virus allowed increased translocation into the hemolymph. Interestingly, while vector infection with Rickettsia enabled increased virus transmission, the dynamic between the endosymbiont and virus appeared somewhat antagonistic with their spatial segregation in the gut and decreased levels of Rickettsia as virus was acquired [129]. Another interesting finding was that Hamiltonella levels increased in the Rickettsiastrain after acquisition of TYLCV but remained the same in the Rickettsia+ strain. While virus acquisition by the Rickettsiastrain was lower, perhaps Hamiltonella increased to facilitate transmission of those virions that crossed into the hemolymph. Further research demonstrated that acquisition of TYLCV induced an activation of immune system gene expression in Rickettsia-MEAM1 B. tabaci but had the opposite effect and induced a down-regulation of immunity-related genes in Rickettsia+ B. tabaci [130]. More recently, presence of the newly discovered species of maternally inherited Rickettsia, Candidatus Rickettsia Torix Bemisia tabaci (RiTBt), in B. tabaci Asia II 7 [23,110] slightly increased virus titer in the vector and transmission of cotton leaf curl Multan virus (genus Begomovirus, family Geminiviridae) after a prolonged acquisition access period [131]. RiTBt was localized to the bacteriocyte only, even after virus acquisition, and was therefore hypothesized to affect virus transmission via secretory proteins.

Other Secondary Endosymbionts
Fewer studies have focused on the roles of the other secondary endosymbionts in virus transmission. While, previously, Hamiltonella GroEL was the only endosymbiont-expressed GroEL thought to interact with begomoviruses in B. tabaci, Rana et al. [42] demonstrated that the secondary endosymbiont Arsenophonus of an Asia II population devoid of Hamiltonella expressed GroEL which interacted with cotton leaf curl virus (genus Begomovirus, family Geminiviridae) in vitro and in vivo ( Table 2). Unlike Hamiltonella and previous reports of Arsenophonus restriction to the bacteriocyte, Arsenophonus in this population of B. tabaci was also found in the salivary glands and midgut, two organs involved in the transmission process [42]. Another report on the effect of Arsenophonus on virus transmission found that fitness of an African SSA1-SG3 B. tabaci (cassava whitefly) population was negatively impacted by harboring Arsenophonus and Rickettsia (AR+) and that acquisition and retention of East African cassava mosaic virus-Uganda variant was lower in AR+ B. tabaci [63]. We are not aware of any described interactions of Portiera or secondary endosymbionts Wolbachia, Cardinium, Fritschea, and Hemipteriphilus with plant viruses in B. tabaci transmission.

Conclusions
Tri-trophic vector-endosymbiont-plant and vector-endosymbiont-virus interactions are indeed complex considering the cryptic B. tabaci complex with its numerous sibling species, several secondary endosymbionts inhabiting their hosts in different combinations, and transmission of plant viruses. Understanding all levels involved (vector, bacterial, and viral) and their activities in host interactions and virus transmission processes are key to developing novel pest management strategies targeting endosymbionts or their products. With the application of high throughput sequencing technology to the study of whitefly metagenomes, the issue of false negatives in PCR-based endosymbiont detection has been minimized, allowing a more complete picture of the microorganisms involved in the whitefly holobiont. While endosymbionts play important roles in whitefly-host plant interactions, other symbiotic bacteria, including gut bacteria, likely influence these interactions. It is central then to accurate research on endosymbiont roles in whitefly fitness and whitefly-host plant and -virus interactions to demonstrate the establishment of comparative populations whose microbiome compositions are, ideally, identical. Using antibiotic treatments to establish endosymbiont-free populations disrupts the microbiome; therefore, efforts to reestablish the microbiota (excluding the endosymbiont of interest) should be considered. Selective elimination of endosymbionts which are generally not fixed in populations, such as Rickettsia and Wolbachia, is possible without antibiotics; however, those which generally are fixed, such as Hamiltonella and Arsenophonus, present more of a challenge.
While many questions regarding the effects of particular endosymbionts on whitefly behavior and fitness remain, the same is true for endosymbiont influence in virus transmission. Studies have demonstrated the involvement of endosymbiont expressed proteins, such as GroEL, in the virus transmission process. Given this interaction and that the GroEL chaperone is highly expressed by intracellular endosymbionts, the question of coevolved relationships arises. Perhaps the relationship is one of chance, or maybe the protection of begomovirus virions against the vector immune system by secondary endosymbiont GroEL is a coevolved relationship which benefits both whitefly/endosymbiont and virus. With improvements on the resolution of in vivo endosymbiont transcriptomes, studying the effects of virus acquisition on endosymbiont gene expression using comparative transcriptomics could address this question. Whitefly gene expression [133] and behavior [134,135] are altered after feeding on TYLCV infected tomatoes. These observations along with the demonstrated endosymbiont-begomovirus interactions raise questions about how the endosymbionts are affected by or involved in these processes and if they may be targeted to reduce or prevent virus transmission. Given the evidence of such multitrophic interactions and recent advancements in genome-based biotechnologies, management of whiteflies and whitefly transmitted viruses via endosymbiont-targeted manipulation is possible. For example, endosymbiont-mediated RNA interference (RNAi) has been demonstrated for control of the western flower thrips (WFT) [136]. In that study, a cultivable symbiotic gut bacterium was isolated from the WFT and transformed with a dsRNA expression cassette targeting the WFT alpha-tubulin gene. Re-introduction of the gut endosymbiont and colonization in its host resulted in heritable target knockdown and high mortality in first instars [136]. While direct manipulation of maternally inherited primary or secondary endosymbionts presents more of a challenge due to their fastidious nature, endosymbiont genes that are involved in production of essential amino acids or carotenoids [21,26,27] or in virus transmission [18,121] may be targeted for specific B. tabaci or virus management. Disentangling these multitrophic relationships should unlock new ideas, leading to needed advancements in managing this global pest.