Factors Affecting Population Dynamics of Maternally Transmitted Endosymbionts in Bemisia tabaci

While every individual of Bemisia tabaci (Hemiptera: Aleyrodidae) harbors the primary symbiont (P-symbiont) Portiera, the infection frequencies of the six secondary symbionts (S-symbionts) including Hamiltonella, Arsenophonus, Cardinium, Wolbachia, Rickettsia and Fritschea vary greatly among different populations. To characterize the factors influencing the infection dynamics of the six S-symbionts in B. tabaci, gene-specific PCR were conducted to screen for the presence of the P-symbiont Portiera and the six S-symbionts in 61 (17 B and 44 Q biotypes) field populations collected from different plant species and locations in China. All individuals of the 61 populations hosted the P-symbiont Portiera, but none of them harbored Arsenophonus and Fritschea. The presence and infection rates of Hamiltonella, Cardinium, Rickettsia, Wolbachia and their co-infections Rickettsia + Hamiltonella (RH), Rickettsia + Cardinium (RC), Hamiltonella + Cardinium (HC) and Rickettsia + Hamiltonella + Cardinium (RHC) varied significantly among the 61 field populations; and the observed variations can be explained by biotypes, sexes, host plants and geographical locations of these field populations. Taken together, at least three factors including biotype, host plant and geographical location affect the infection dynamics of S-symbionts in B. tabaci.


Introduction
Bacteria commonly form intimate, symbiotic associations with insects. These bacterial endosymbionts occur in a diverse array of insect species and are usually passed from host to host by vertical transmission [1,2]. The endosymbionts of insects are divided into two groups: primary symbionts (P-symbionts) and secondary symbionts (S-symbionts) [1]. P-symbionts often produce/provide essential nutrients for their hosts and thus form an obligatory mutualistic relationship with their hosts, i.e., they depend on each other for survival. Examples include the bacterial symbionts that make nutrients for their insect hosts, such as Buchnera aphidicola in aphids, Carsonella ruddii in psyllids, and Tremblaya princeps in mealybugs [1]. By contrast, S-symbionts may not be required for host survival. Nonetheless, they may play important roles in their host's nutrition [3,4], genetic differentiation [5], adaptation to a wide range of food plants [6], defense against natural enemies [7][8][9], reproduction [10,11], sensitivity to heat stress and other environmental factors [12].
The whitefly Bemisia tabaci (Gennadius) (Hemiptera: Aleyrodidae) is one of the most destructive insect pests of numerous protected and field crops worldwide. It causes serious damage in many crops by direct feeding and by vectoring 111 plant viruses [13]. B. tabaci has long been thought to comprise morphologically indistinguishable biotypes that often differ in host range, fecundity, insecticide resistance, and/or transmission competency for begomoviruses [14,15]. Recent studies suggest that most of these biotypes represent genetically distinct cryptic species [16][17][18][19]. The B biotype of the Middle East-Minor Asia 1 and the Q biotype of Mediterranean group are among the most invasive and destructive biotypes [16,19,20]. B. tabaci was first recorded in the late 1940s in China, but the crop damages and loses caused by this insect had not been serious until the introduction of the B biotype in the 1990s [21]. In recent years, the Q biotype of B. tabaci has invaded China [22], and in many areas, even displaced the B biotype [23].
The P-symbiont Portiera infects every individual of any B. tabaci populations. By contrast, the infection frequencies of the six Ssymbionts vary greatly among B. tabaci populations [31][32][33][34][35][36]. Little, however, is known about the factors affecting the infection dynamics of S-symbionts in B. tabaci. In this study, we conducted a systematic survey of symbiotic bacteria in 61 field populations of B. tabaci collected from various crops and locations in China. The results obtained suggest that biotype, host plant species, and geographical location of B. tabaci are the major factors affecting the population dynamics of S-symbionts in B. tabaci.

B. tabaci field populations
The 61 B. tabaci field populations used in this study were collected from 61 localities in 19 provinces of China in 2009 (Fig. 1). The 61 locations cover an area stretching from Hainan province (16u869N-110u479E) in the South to Heilongjiang province (44u839N-132u539E) in the North, and from Xinjiang Uygur autonomous region (40u789N-78u589E) in the West to Shanghai city (28u189N-126u469E) in the East. From each location, dozens of adults were collected alive, fixed in 95% ethanol, and stored at 220uC until analysis of symbionts. The locations, host plants, and biotypes of the 61 field populations are presented in Table S1.

Biotype determination
Genomic DNA was extracted from each of approximately 24 individual whiteflies (or 24 pairs) per population using the methods of Luo et al. (2002) [21]. Briefly, individual adult was homogenized in 20 mL of extraction buffer (50 mM Tris-HCl, pH 8.0, 20 mM NaCl, 1 mM EDTA, and 1% SDS) on a piece of parafilm. The extract was transferred to a microcentrifuge tube containing 2 mL of proteinase K and incubated at 60uC for 3 h. 78 mL of distilled water was added to the tube, which was incubated at 100uC for 5 min. After incubation, the extract (about 100 mL in total) was mixed well with 200 mL of absolute alcohol. The sample was then placed at 220uC for 2 h and centrifuged at 12,000 rpm for 20 min. The DNA pellets obtained were washed once with 75%  Table S1. The provinces of China in which no samples were collected are not included in this map. The vertical blue dot line indicates deletion of the central provinces between the western and the eastern provinces of China. Beijing, Tianjin and Shanghai cities are disproportionally enlarged. The filled blue and red circles represent B-and Q-biotype populations, respectively. doi:10.1371/journal.pone.0030760.g001 ethanol, air dried, dissolved in 20 mL of TE buffer (pH 8.0), and stored at 220uC for subsequent biotyping and PCR detection of symbionts. The biotype of every individual of each population was determined by PCR amplification of mitochondrial cytochrome oxidase I (mtCOI) gene fragment using biotype-specific primers [37]. The accuracy of such PCR-based biotyping was further verified by sequencing a ,620-bp mtCOI fragment of two individuals per population amplified using the primers C1-J-2195 and TL2-N-2819 [38].

Screening for the presence of symbionts and data analysis
Approximately 24 (or 24 pairs of) adults each of the 61 field populations ( Fig. 1; Table S1) were analyzed for the presense of the P-symbiont Portiera and the S-symbionts Hamiltonella, Arsenophonus, Cardinium, Wolbachia, Rickettsia, and Fritschea (Table S1). The presence of Portiera, Hamiltonella, Arsenophonus, Cardinium, Wolbachia, Rickettsia, and Fritschea in each individual adult was determined by PCR-gel analyses of 16S rRNA (Portiera, Hamiltonella, Cardinium, and Rickettsia), 23S rRNA (Arsenophonus and Fritschea) or wsp (Wolbachia) gene using genus-specific primers ( Table 1). PCR amplifications were performed in 25 mL reactions containing 2.5 mL 106PCR Buffer (Mg 2+ Plus), 2 mL dNTP mix (2.5 mM of each nucleotide), 0.5 mL of each primer (10 mM each), and 0.125 mL of TaKaRa Taq (5 U/ mL) (TaKaRa Biotechnology (Dalian) Co., Ltd). The PCR cycling conditions for detection of the seven symbionts were summarized in Table 1. The resultant PCR products were electrophoresed with the negative and positive controls of each symbiont on a 1.0% agarose gel in a 0.56TBE buffer and visualized by ethidium bromide staining. The identities of the expected bands for each of the seven symbionts were verified by subsequent sequencing of two independent clones derived from two different whitefly individuals. As a result, whether an individual adult was infected by any of the seven symbionts was determined based on the presence of the corresponding band.
The total numbers of tested and infected adults by the seven symbionts for each field population were recorded to calculate their infection frequencies by the formula: the number of individuals infected or co-infected by a given symbiont or a symbiont combination/the total number of individuals screened (usually n = 24) 6100%. The impacts of biotype, sex, host plant species, and geographical location on the infection frequencies of the symbionts were analyzed using the x2 tests, followed by multiple comparisons with Bonferroni corrections. The differences between the observed and expected ( = multiplication of infection frequencies of each symbionts) frequencies of multiple infections were compared by nonparametric x2 tests. All statistical analyses were performed with SPSS version 18.0 (SPSS Inc., Chicago, IL, USA).

Nation-wide survey of symbionts in the field populations of B. tabaci
A total of 61 field populations of B. tabaci collected from a wide range of host plants in 19 provinces of China during July-September 2009 were biotyped and tested for the diversity and infection frequencies of the P-and S-symbionts ( Fig. 1 and Table  S1). 44 out of the 61 populations belonged to the Q1 subclade of the Q biotype and the remaining 17 were of pure B biotype (Table  S1). All individuals of the 61 field populations had the P-symbiont Portiera, regardless of their collection sites, host plants, biotype and sex of the whitefly (Table S1).
Among the 6 S-symbiont species we examined, Arsenophonus was not detected in any of the 61 field populations (data not shown). PCR-gel analysis revealed that 24.6% of the 61 populations had the expected band for Fritschea (data not shown), but subsequent cloning and sequencing of the band eluted from gels confirmed that it was either Flavobacterium or Acidobacteria, not Fritschea (data not shown). Wolbachia was extremely rare and only detected in one B biotype (Code 6 in Table S1) and one Q biotype (Code 52 in Table S1) populations. Interestingly, both populations were collected from cotton and had the same infection frequency (4.2%) by Wolbachia (Table S1), but were geographically far from each other (Fig. 1).
Rickettsia, Hamiltonella, and Cardinium were present alone or together in more than 2 field populations ( Table S1). The infection frequencies of the three S-symbionts varied greatly among the 61 populations. Statistical analyses revealed that the plant species, geographical location, biotype and sex of the field populations were the four factors significantly contributing to the great variations in the diversity and infection frequency of the Ssymbionts in the B. tabaci field populations.

Impacts of B. tabaci biotype on the infection frequency of the S-symbionts
The B-and Q-biotype field populations differed significantly in their infection frequencies by the S-symbionts. All the 17 B biotype populations possessed at least one S-symbiont, whereas 4 (i.e. 9.1%; Codes 30, 37, 49, and 56 in Table S1) out of the 44 Q biotype populations did not harbor any of the 6 S-symbionts. The infection frequency of Wolbachia was very low in both biotypes and there was no significant difference between the B-and Q-biotype individuals ( Table 2). The two biotypes also had a similar percentage of Hamiltonella infection. The infection frequency of Rickettsia was significantly higher in the 17 B biotype populations (64.3% of a total of 456 B biotype individuals) than the 44 Q biotype populations (7.2% of a total of 1149 Q biotype individuals) ( Table 2). By contrast, the infection frequency of Cardinium was significantly greater in the 44 Q biotype populations (16.3%) than in the 17 B biotype populations (10.8%). Co-infection of whiteflies by Wolbachia and any other S-symbionts was not found in the two biotypes. The B biotype field populations also had significantly higher co-infection frequencies by Rickettsia + Hamiltonella (RH) (40.6%), Rickettsia + Cardinium (RC) (8.6%), and Rickettsia + Hamiltonella + Cardinium (RHC) (5.5%) than the Q biotype field populations ( Table 2). In the B biotype, the observed frequency of RH co-infection was significantly higher than its expected frequency, which was calculated by multiplying the infection frequencies of each symbionts. No differences were found between the observed and expected frequencies of RC, HC and RHC coinfections in the two biotypes (Table 2).

Impacts of plant species on the infection frequency of S-symbionts
Chi-square tests showed that host plant species significantly and differentially affected the infection frequencies of the B and Q biotypes B. tabaci by the S-symbionts (Tables 3 and 4). In the B biotype populations, Rickettsia was the most prevalent S-symbiont. It was detected in the B biotype whiteflies collected from all the 7 plant species (cucumber, tomato, cabbage, cotton, poinsettia, sweet potato, and bean; each representing 1 plant family) surveyed (Table S1 and 3). The infection frequency of Rickettsia was the highest in the B biotype populations collected from cotton (100%), followed by those from cucumber (73.3%), poinsettia (70.8%), and cabbage (67.7%), those from bean (41.7%) and sweet potato (33.3%), and those from tomato (12.5%) ( Table 3). Hamiltonella was less prevalent than Rickettsia in the B biotype populations and was not detected in the B biotype whiteflies from sweet potato (Table 3). Chi-square tests divided the B biotype whiteflies from the 6 plants into 2 significantly different groups. The B biotype populations from poinsettia (83.3%), cotton (61.5%), cucumber (57.5%) and cabbage (50.0%) had significantly higher Hamiltonella infection than those from tomato (20.8%) and bean (8.3%) ( Table 3). Cardinium was detected only in two B biotype populations, one from cucumber and another from cotton (Table S1). The infection frequency of Cardinium was significantly higher in the cucumber population than in the cotton population ( Table 3). As a result, the co-infection rates of RC, HC and RHC were higher (significant for RHC and HC) in the cucumber population than in the cotton population (Table 3). RH co-infection was significantly more prevalent in the cucumber, cabbage, cotton, and poinsettia populations than in the tomato, bean and sweet potato populations. Further nonparametric x2 tests revealed no significant differences between the observed and expected frequencies of RHC, RC and HC co-infection in the populations from all the seven plant species (Table 3).
Unlike in the B biotype populations, Rickettsia was the least prevalent S-symbiont in the Q biotype whiteflies (Table S1 and 4). It was found in the Q biotype populations from gerbera, Japanese hop, cotton, egg plant, squash, towel gourd, and cucumber, but not from tomato, pepper, poinsettia, and bean (Table 4). Rickettsia infection was the highest in the Q biotype whiteflies from Gerbera (75.0%), followed by those from towel gourd (26.1%), those from cotton (11.3%) and eggplant (10.1%), and those from squash (4.2%), cucumber (2.3%) and Japanese hop (1.4%). Hamiltonella and Cardinium were more prevalent and detected in the Q biotype populations from all the 11 plants but squash. Q biotype whiteflies from gerbera (87.5%), poinsettia (83.3%) and Japanese hop (76.4%) had the highest rate of Hamiltonella infection, followed by those from pepper (61.1%), eggplant (53.6%), cotton (50.6%), and bean (41.7%), those from towel gourd (34.8%) and tomato (32.3%), and those from cucumber (16.4%) and squash (0.0%). For Cardinium, the pepper populations (52.1%) had the highest infection frequency, whereas the squash (0.0%) and cotton (0.6%) populations had the lowest infection ( Table 4). The plant populations between the two groups had a more or less similar level of Cardinium infection. Further, Q whiteflies from gerbera and towel gourd had a greater frequency of RH (significant), RC and RHC co-infection than those from other plants (Table 4). For HC co-infection, Q whiteflies from gerbera, pepper and towel gourd had the greatest frequency, followed by those from bean, tomato, Japanese hop, and eggplant, and those from the remaining plants.
Nonparametric x2 tests showed no significant differences between the observed and expected frequencies of RH, RC and RHC coinfections in the populations from all the seven plant species ( Table 4). The observed frequencies of HC co-infection were significantly lower than its expected frequencies in the pepper and poinsettia populations (Table 4).

Impacts of geographical location on the infection frequency of S-symbionts
To test the effects of geographical location on the infection and co-infection of the 4 S-symbionts, we divided the collection sites into three geographical regions: 15uN-25uN, 25uN-35uN and  35uN-45uN (Fig. 1). The B biotype populations collected from the three regions had a similar level of Rickettsia and Hamiltonella infection frequency (Table 5). By contrast, Cardinium and Wolbachia were found only in the 35uN-45uN region (18.6%, significantly greater than 0.0% in the other two regions) and the 25uN-35uN region (0.7%, not significantly different from 0.0% in the two other regions), respectively. Consequently, RC, HC, and RHC coinfections of the B biotype whiteflies existed only in the 35uN-45uN region; the corresponding co-infection frequency was significantly greater than 0.0% in the two southern regions. In contrast, RH co-infection was prevalent in all the regions, but no significant difference in co-infection frequency existed between the three regions (Table 5). In addition, the observed co-infection frequencies of HC and RHC in the 35uN-45uN region and RH in the 25uN-35uN and 35uN-45uN regions were significantly greater than their expected frequencies.
In the Q biotype whiteflies, Rickettsia, Hamiltonella and Cardinium were present in the three geographical regions, whereas Wolbachia (1.1%) was detected only in the 15uN-25uN region ( Table 6). Hamiltonella was prevalent in the three geographical regions, with   15uN-25uN region, and the 25uN-35uN region ( Table 6). Double or triple co-infection rate of Q biotype whiteflies by the three S-symbionts ranged from 0.4% to 5.9% across the three regions, with no significant difference. There were no significant differences between the observed and expected frequencies of RH, RC and RHC in all the three regions. But the observed frequency of HC in the 35uN-45uN region was significantly lower than its expected frequency ( Table 6).

Impacts of sex on the infection frequency of Ssymbionts
Male and female adults of 2 B-(Codes 1 and 2 in Table S1) and 4 Q-biotype (Codes 19, 20, 22, and 23 in Table S1) field populations were separately screened for the presence of Ssymbionts. x2 tests indicated that sex significantly affected the infection frequency of Hamiltonella, but not Rickettsia or Cardinium ( Table 7). The average infection frequency of Hamiltonella was significantly higher in females (88.9%) than in males (12.5%) (x2 = 165.046, P,0.0001). The co-infection frequencies of RHC, RH, and HC, but not RC, were also significantly higher in females than in males (P,0.0001). No significant differences existed between the observed and expected co-infection frequencies of female and male whiteflies by RH, RC, and RHC. The observed frequency of HC was significantly lower than its expected frequency in female, but not male whiteflies (Table 7).

Discussion
Several recent studies have showed a 100% infection frequency of B. tabaci by the P-symbionat Portiera, but a 0-100% infection frequency by any of the six S-symbionts (Hamiltonella, Arsenophonus, Cardinium, Wolbachia, Rickettsia and Fritschea), depending on the populations tested [31][32][33][34][35][36]. To reveal the key factors that govern the infection dynamics of whiteflies by the six S-symbionts, we surveyed 61 field populations of (24 individuals or 24 pairs each) for the presence of the P-symbiont Portiera and the six S-symbionts. Consistent with the aforementioned studies, all individuals of the 61 field populations harbored the P-symbiont Portiera. Arsenophonus and Fritschea were not present in any of the 61 populations, whereas the presence and infection frequencies of Hamiltonella, Cardinium, Wolbachia and Rickettsia varied greatly among the 61 populations.
Our data from the 61 (17 B biotype and 44 Q biotype) field populations and all the prior studies [31][32][33][34][35][36] demonstrate that biotype or genetic group of B. tabaci is a key factor determining the infection dynamics of the six S-symbionts. However, there are noticeable discrepancies in the S-symbiont-host biotype associations or linkage disequilibrium revealed by these studies. For example, Chiel et al. (2007) and Gueguen et al. (2010) reported that the B biotype only harbored Hamiltonella (100%) and Rickettsia (64.0%), the Q1 subclade of the Q biotype only harbored Hamiltonella (close to 100.0%), Cardinium and Wolbachia, the Q2 subclade only harbored Rickettsia (74.0%), Wolbachia (33.0%) and Arsenophonus (87.0%), and the Q3 subclade only harbored Rickettsia (about 30.0%) and Arsenophonus (about 97%). By contrast, all of the six S-symbionts but Fritschea were detected in biotypes B and Q (i.e. Q1) in China during 2005 to 2009 [36], and the Q biotype in Table 5. Impacts of geographical location on the diversity and infection frequency of S-symbionts in the field populations of B biotype B. tabaci*.  Croatia [31] hosted Hamiltonella, Rickettsia, Wolbachia and Cardinium, but not Arsenophonus and Fritschea.
One explanation for the above discrepancies is that there are other important factors that have not been characterized but significantly affect the infection dynamics of the S-symbionts in B. tabaci. That the female adults of 2 B-(Codes 1 and 2 in Table S1) and 4 Q-biotype (Codes 19, 20, 22, and 23 in Table S1) field populations had significantly higher infection frequencies of Hamiltonella, RHC, RH, and HC than did their male adults (Table 7) suggests that sex of the hosts could be another factor impacting the infection dynamics of B. tabaci by the 6 S-symbionts. However, this female-biased infection rate is probably not a real infection difference between male and female adults, but a matter of a combination of lower symbiont titer in males and lower detection sensitivity of our PCR-gel analysis technique. Single whiteflies are small and ethidium bromide used in this study is not the most sensitive DNA stain. It is well known that symbiont infections are not easily detectable in arthropods [39]. Moreover, it is females, not males, that transmit the symbionts to their offspring, and males are smaller than females. Therefore, it makes some sense that the titer of symbionts in males would be lower and thus would be more difficult to be detected by our PCR-gel analysis technique. Another less likely explanation for the pattern of lower frequencies of symbiont infections in males would be male-killing effects of Hamiltonella, RHC, RH, and HC. The fact that the male-killing symbiont identified in the ladybird Cheilomenes sexmaculata shares the closest similarity to Hamiltonella defensa (98% nucleotide identity) of whiteflies [40] implicates that Hamiltonella of whiteflies could be a male-killing symbiont. More experiments are necessary to resolve these two possibilities.
Our survey data of the 61 field population collected from three different regions in China (15uN-25uN, 25uN-35uN, and 35uN-45uN) suggest that geographical location is another factor affecting the infection frequency of the S-symbionts in B. tabaci. The impacts of geographical locations are best exemplified by infection of the B biotype by Cardinium and Wolbachia only in the 35uN-45uN and the 25uN-35uN regions respectively (Table 5), significantly greater infection of the Q biotype by Cardinium in the 35uN-45uN region, and the increase of Rickettsia infection with decreasing latitude in the Q biotype (Table 6). This is consistent with the results of Ahmad et al. (2010), who showed that the infection frequency of Wolbachia in B. tabaci was related to country (location).
Our survey data of the 61 field populations collected from 7 (B biotype) or 11 (Q biotype) different plant species also indicate that host plant affects the population dynamics of the S-symbionts in B. tabaci. This is because the infection frequencies of Rickettsia, Cardinium, RH, and HC were all significantly different among the B and Q biotype field populations collected from different plant species (Table 3 and 4). This agrees with the findings of Chiel et al. (2007), who reported that Rickettsia and Arsenophonus had a significantly higher infection frequency in the Q populations collected from sage than those from all other host plants.
In summary, at least three factors including biotype or genetic group, host plant and geographical location affect the infection dynamics of S-symbionts in B. tabaci. Two of the three factorshost plants [6,[41][42][43][44][45] and collection sites [42,46]-as well as temperature [46][47][48] are known to affect the infection frequency of S-symbionts in other insects. Nonetheless, the host plant-and/or location-symbiont association patterns revealed from the field populations could be caused, at least partially, by factors other than host plants or latitude, particularly population inertia and stochasticity. In theory, these genetic (biotype, sex) and environmental factors (host plant and geographical location) could change the horizontal transfer frequency, vertical transmission fidelity, and/or relative fitness of these S-symbionts on B. tabaci. Further experiments are required to address how the four factors affect the population dynamics of the six maternally transmitted S-symbionts in B. tabaci.
This study showed that co-infections with Rickettsia and Hamiltonella are more common than expected in natural populations of B biotype B. tabaci (Table 2-3, 5). In theory, there are a number of possible mechanisms that can facilitate such endosymbiont co-infections. For example, the co-infecting endosymbionts may additively confer fitness advantages on the host, either by positively affecting different fitness parameters or by favoring overlapping fitness parameters in a synergistic way [49]. However, the co-infections by HC are significantly rarer than expected in natural populations of B-and Q-biotype B. tabaci (Table 4, 6-7), suggesting antagonistic interactions between the S-symbionts upon co-infection, possibly because of competition for limited resource and niche in the same host individual [30,50] or synergistic negative effects of co-infection on the host organism [8]. To confirm which of these hypotheses can better account for the coinfection patterns for each of the S-symbiont pairs, further experiments are needed.  Table 7. Impacts of sex on the diversity and infection frequency of S-symbionts in the 6 field populations (population codes 1, 2, 19, 20, 22, and 23 in