Almost There: Transmission Routes of Bacterial Symbionts between Trophic Levels

Many intracellular microbial symbionts of arthropods are strictly vertically transmitted and manipulate their host's reproduction in ways that enhance their own transmission. Rare horizontal transmission events are nonetheless necessary for symbiont spread to novel host lineages. Horizontal transmission has been mostly inferred from phylogenetic studies but the mechanisms of spread are still largely a mystery. Here, we investigated transmission of two distantly related bacterial symbionts – Rickettsia and Hamiltonella – from their host, the sweet potato whitefly, Bemisia tabaci, to three species of whitefly parasitoids: Eretmocerus emiratus, Eretmocerus eremicus and Encarsia pergandiella. We also examined the potential for vertical transmission of these whitefly symbionts between parasitoid generations. Using florescence in situ hybridization (FISH) and transmission electron microscopy we found that Rickettsia invades Eretmocerus larvae during development in a Rickettsia-infected host, persists in adults and in females, reaches the ovaries. However, Rickettsia does not appear to penetrate the oocytes, but instead is localized in the follicular epithelial cells only. Consequently, Rickettsia is not vertically transmitted in Eretmocerus wasps, a result supported by diagnostic polymerase chain reaction (PCR). In contrast, Rickettsia proved to be merely transient in the digestive tract of Encarsia and was excreted with the meconia before wasp pupation. Adults of all three parasitoid species frequently acquired Rickettsia via contact with infected whiteflies, most likely by feeding on the host hemolymph (host feeding), but the rate of infection declined sharply within a few days of wasps being removed from infected whiteflies. In contrast with Rickettsia, Hamiltonella did not establish in any of the parasitoids tested, and none of the parasitoids acquired Hamiltonella by host feeding. This study demonstrates potential routes and barriers to horizontal transmission of symbionts across trophic levels. The possible mechanisms that lead to the differences in transmission of species of symbionts among species of hosts are discussed.


Introduction
The occurrence of arthropods serving as hosts for bacterial symbionts is very common. Primary, obligate symbionts that provide essential nutrients lacking in the host's diet, are strictly maternally transmitted and show congruent phylogenies with those of their host group [1,2]. Facultative, secondary symbionts are also transmitted vertically, and promote their own transmission by contributing to host fitness or by manipulating the host's reproduction [3][4][5][6][7][8][9]. Phylogenetic trees of secondary symbionts are largely incongruent with those of their hosts. This, and the fact that the same secondary symbionts are sometimes found in distantly related hosts, is attributed to rare horizontal transmission events of the symbionts between species [1,10,11].
The routes of horizontal transmission are not very well known, although transmission via common host plants and/or common natural enemies has been hypothesized, and phylogenetic evidence for the latter has been provided [12][13][14]. Rare examples of experimentally demonstrated natural intra-specific horizontal transmission include Arsenophonus [15], Wolbachia [16] and a virus [17] in parasitoids, as well as transmission between mates of the same aphid species [18]. In contrast, documentation of interspecific transmission is almost non-existent. Huigens et al [16] showed horizontal transmission of Wolbachia between conspecifics of Trichogramma kayaki when developing within the same host. However, attempts to show inter-specific horizontal transmission of Wolbachia by the same mechanism, between Trichogramma species, resulted in loss of the symbiont from the recipient species within a few generations [19]. In lieu of more natural examples, some microinjection studies have been successful in establishing some new stable associations [20][21][22][23], yet others have been unsuccessful in establishing novel symbiont-host associations [24][25], suggesting limits to the ability of symbionts to colonize the germ line of some hosts. While elegant work has shown how Wolbachia colonizes the germ line of a Drosophila host following injection of cured individuals [26], why symbionts fail to become established is not understood.
The intimate interaction between hosts and their endoparasitoids would seem to provide opportunities for horizontal transmission of symbionts, as parasitoid larvae consume nothing but symbiont-contaminated food throughout their development. Yet, to our knowledge, there is no experimental evidence of permanent acquisition of arthropods' symbionts by their natural enemies, hence the notion that inter-specific horizontal transmission is a rare event.
Here we followed transmission routes of symbionts from their host -the sweet potato whitefly, Bemisia tabaci -to parasitoids. Bemisia tabaci (Gennadius) (Hemiptera: Aleyrodidae) is a minute insect that feeds on phloem sap of numerous host plants and is a major pest of agricultural crops [27]. Bemisia tabaci harbors a primary symbiont, Portiera aleyrodidarum that most probably produces amino acids lacking in the phloem diet [28]. This primary symbiont is located only within specialized cellsbacteriocytes -that are aggregated in two clusters called bacteriomes [1]. In addition, B. tabaci may harbor a variety of secondary symbionts: Arsenophonus, Cardinium, Fritschea, Hamiltonella, Rickettsia and Wolbachia (reviewed in [1,2]; [29]), whose function is yet mostly unknown. The B. tabaci colony used in our study carried only two of those secondary symbionts: Hamiltonella and Rickettsia. Hamiltonella is located inside the bacteriocytes with the primary symbiont, while the Rickettsia in our culture is dispersed throughout the hemocoel [30].
Bacteria of the genus Rickettsia (a-Proteobacteria) are best known as vector-borne agents of many vertebrate diseases. The more recent discoveries of Rickettsia in many different invertebrates, with diverse effects such as reproductive manipulation, heat tolerance and plant disease, suggest the disease-causing members represent a small portion of a much larger group [31]. The Rickettsia in B. tabaci is most closely related to the pea aphid Rickettsia, is found in all developmental stages of the whitefly, and is maternally transmitted [32]. Rickettsia is highly prevalent in B. tabaci populations [29], but its benefits to the host, if any, are not clear. As a matter of fact, Rickettsia was found to inflict some costs on fitness parameters of B. tabaci [33,34]. Hamiltonella (c-Proteobacteria) was described from the pea aphid, Acyrthosiphon pisum, where it occurs in various tissues both extra-and intra-cellularly and benefits its host by conferring resistance against parasitoids [5,6,35].
Bemisia tabaci is attacked by a wide variety of natural enemies, including parasitoids of the genera Eretmocerus and Encarsia (Hymenoptera: Aphelinidae) [36]. These two genera belong to two different sub-families: Eretmocerus, with 16 species recorded from B. tabaci [37], is in the Aphelininae subfamily; Encarsia, with 344 described species, of which 175 species attack whiteflies, is in the Coccophaginae subfamily [38]. Eretmocerus and Encarsia also differ markedly in their mode of development: Eretmocerus spp. lay a single egg under the host venter (i.e., between the host and leaf) and the first instar penetrates and develops within a vital cellular capsule inside the host [39]. Encarsia spp., in contrast, lay the egg directly into the body of their whitefly host [40].
In preliminary screening we found that two species of Eretmocerus, Er. eremicus (Rose & Zolnerowich) and Er. sp. nr. emiratus (Zolnerowich & Rose), were both highly infected with a Rickettsia that had the same 16S rDNA and citrate synthase gene sequences as the Rickettsia in their host, B. tabaci. Therefore the current study was initiated to address two key questions: 1) What (if any) are the routes of transmission of Rickettsia and Hamiltonella from B. tabaci to the whitefly's parasitoids? 2) Are symbionts that are acquired by the parasitoids then vertically transmitted to parasitoid offspring?

Insect colonies
1. Whiteflies. Two B. tabaci (biotype B) colonies were used for the study: one that carried Rickettsia (R + ) and one that did not (R 2 ). Rickettsia in these whiteflies was distributed throughout the hemocoel, the 'scattered' phenotype [30]. The presence/absence of Rickettsia was routinely monitored by diagnostic PCR, as described below. Additionally, the secondary symbiont Hamiltonella was established in all individuals of both colonies. Each colony was reared in a separate room at 2761uC, ca. 60% RH and 16:8 L:D. Both colonies have been maintained for over two years on cowpea plants (Vigna unguiculata var. California blackeye).
2. Parasitoids. Eretmocerus sp. nr. emiratus, Er. eremicus and Encarsia pergandiella were each reared separately on cowpea plants that were infested with R + B. tabaci nymphs as hosts, inside transparent ventilated plastic jars. Both sexes of Eretmocerus spp. develop as solitary, primary parasitoids, whereas Encarsia pergandiella is an autoparasitoid [41]; females are primary parasitoids of whiteflies and males are hyperparasitic, developing on conspecific or heterospecific immatures. Male En. pergandiella were thus produced by exposing Er. eremicus larvae and pupae to adult female En. pergandiella. All parasitoid cultures were kept in a climate-controlled walk-in chamber (2761uC, ca. 60% RH and 16:8 L:D).

3.
Establishment of symbiont-free parasitoid colonies. Eretmocerus emiratus and Er. eremicus were fed on honey containing 50 mg/ml Rifampicin for 48 hrs and were then released on cowpea plants bearing R 2 B. tabaci nymphs for oviposition. This process was repeated for two consecutive generations. The infection status of the progeny was then checked with PCR and both species were found to be free of Rickettsia and Hamiltonella, therefore they were continuously reared on R 2 whiteflies under the conditions described above. Encarsia pergandiella was not treated the same way because neither Rickettsia nor Hamiltonella were detected in adult wasps after development in infected whiteflies.

Methodology
4. PCR analysis. To extract DNA, individual whiteflies or wasps were ground in a 3 ml droplet of proteinase K solution (20 mg/ml, Invitrogen). The droplet was then transferred into a tube containing 50 ml of sterile 10% Chelex beads (Sigma-Aldrich) in PCR water. The tubes were incubated at 37uC for 1 h, then at 96uC for 8 min and then kept at 220uC until analysis. Two microliters of the DNA lysate were used as a template for PCR reactions. The presence of Rickettsia was determined using specific primers for amplifying 16S rDNA gene fragments: 528F  and 1044R [5-GTTTTCTTATAGTTCCTGGCATTACCC-3]. PCR conditions were: 95uC for 2 min followed by 35 cycles of 92uC, 30 s; 60uC, 30 s; 72uC, 30 s, and final incubation at 72uC for 5 min. Screening for other B. tabaci symbionts, including Hamiltonella, was done using the primers and conditions described in [29]. Reactions were carried in a 10 ml volume containing 4 pmol of each primer, 0.01 mmol dNTP's, 16 ''Thermopol'' buffer and 0.4 units of Taq DNA polymerase (New England Biolabs). PCR products were visualized on 1.5% agarose gel using SYBR-Green (Cambrex Bio Science Rockland Inc.). To verify the identity of the PCR products, bands were eluted, DNA was purified (QIAquick gel purification kit, Qiagen) and sent for direct sequencing at the University of Arizona's sequencing facility. The resulting sequences were compared to known sequences using the BLAST algorithm in NCBI. Sequences from whiteflies and parasitoids were compared to one another using the BLAST 2 Sequence in NCBI.

Visualization of Rickettsia using Fluorescence In Situ Hybridization
(FISH) and Transmission Electron Microscopy (TEM). FISH of B. tabaci parasitized nymphs, and adult parasitoids was performed with Rickettsia-specific 16S rRNA DNA probes, as described in [32]. Stained samples were whole mounted and viewed under an IX81Olympus FluoView TM 500 confocal microscope (Tokyo, Japan). Reproducibility and controls were performed as described in the above reference (at least 20 individuals of each species). Samples of Er. eremicus females for TEM were prepared as described by [42] (n = 5 females).

Acquisition and maintenance of Rickettsia and
Hamiltonella in whitefly parasitoids. Wasps that developed on Rickettsiaand Hamiltonella-infected whiteflies were censused for infection. Using a fine needle, approx. 100 pupae of each wasp species were removed from leaves and placed in a glass vial with honey. Samples of the pupae were placed in 96% ethanol for diagnostic PCR. Newly emerged wasps were transferred to a new vial with honey and samples were placed in 96% ethanol. Subsequently, wasps were sampled and placed in ethanol on days 3, 6, 9 and 12 post-eclosion. Infection status was then determined by diagnostic PCR in 10-13 wasps of each species, at each time point.
7. Transmission of symbionts from B. tabaci to parasitoids. There are three likely routes by which symbionts can be transmitted from the whitefly host to its parasitoids: 1) the parasitoid larva acquires symbionts while feeding and developing in an infected host; 2) the adult female wasps acquire symbionts via host-feeding (piercing of the whitefly integument with the To test transmission of symbionts from B. tabaci to parasitoids, one female parasitoid was introduced to each leaf disk for 24 h, after which they were tested by PCR. From the emerging F 1 , one or two females from each replicate were used to continue to the vertical transmission experiment, while the rest of the cohort was tested by PCR (two-five from each cohort). The emerging F 2 were all collected and two-five from each cohort were tested by PCR. doi:10.1371/journal.pone.0004767.g001 ovipositor followed by consumption of host hemolymph); 3) adult wasps might acquire the symbiont via feeding on honeydew secretions of infected whitefly hosts. To test these pathways, cowpea leaf disks (30 mm diameter) infested with 30-50 R + B. tabaci nymphs (2 nd and 3 rd instars) were placed on 1% agar inside 35 mm Petri dishes and sealed with screen lids. One male and one female of R 2 wasps (cured Er. emiratus and Er. eremicus grown for six generations on R 2 whiteflies or R 2 En. pergandiella directly from the culture) were introduced onto each leaf disk for 24 hrs and were then collected to 96% ethanol for PCR analysis. The percentage of infection status of these adults was used to determine the acquisition of the symbiont via either host-feeding or feeding on infected honeydew (scenarios 2 and 3 above). For controls, wasps from the same sources were introduced onto leaf disks bearing R 2 whiteflies, and some wasps were placed directly in ethanol, without exposure to hosts. The leaf disks bearing parasitized whiteflies were then incubated for approximately two weeks until wasp progeny emergence and then two to five (at least one male and one female) wasps from each disk were collected and placed in 96% ethanol. An estimate of the percentage of symbiont acquisition via exposure during development (scenario 1 above) was determined by the infection status of this second group of wasps. Results were subjected to a chi-square test (JMP 6.1 software, SAS Institute). Figure 1 illustrates the set up of this experiment, as well as the vertical transmission experiment (#8, below).
To study whether symbiont acquisition via host feeding was permanent or transient, another experiment was carried out.
Here, approximately 50 R 2 wasps were introduced onto a plant infested with R + B. tabaci nymphs (each species on a separate plant). After 24 h the wasps were retrieved, half of them were transferred directly to 96% ethanol and the other half were kept in glass vials with honey for four days, and then also placed in ethanol. Twenty wasps of each species were screened for Rickettsia by PCR: ten wasps from the half that were transferred to ethanol immediately after the exposure to R + whiteflies, and ten from the half that were fed on honey after exposure.
8. Vertical transmission experiments. To study if Rickettsia and Hamiltonella are vertically transmitted between parasitoid generations, cowpea leaf disks bearing R 2 B. tabaci nymphs were prepared and parasitoids of three treatments were randomly assigned to them: 1) F 1 adults from the previous transmission experiment that were exposed to R + hosts during development, i.e. wasps that have been exposed to the symbionts for one generation only; 2) wasps that had been reared on R + hosts for many generations; 3) adults from the horizontal transmission experiment that emerged from R 2 hosts (control). One female parasitoid was introduced onto each leaf disk for 24 hrs and was then placed in 96% ethanol for PCR analysis. The leaf disks were incubated for approximately two weeks until progeny emergence and then two to five (at least one male and one female) progeny from each disk were collected and placed in 96% ethanol for PCR analysis. The set up of this experiment is illustrated in Fig. 1.

Acquisition and maintenance of Rickettsia and Hamiltonella in whitefly parasitoids
Almost all pupae of the three studied species carried Rickettsia and Hamiltonella (Er. emiratus-11 out of 12 infected; Er. eremicus -13/13; En. pergandiella females -10/10; En. pergandiella males-10/ 10). However, infection of adult wasps differed significantly between the two genera of wasps: Rickettsia did not persist in adults of the two Encarsia species, while adults of both Eretmocerus species were virtually all Rickettsia-positive, even 12 days after they had emerged and fed on honey only (sample size = 10 wasps; 9 or 10 tested positive in each sample). In contrast, all Encarsia and Eretmocerus adults were Hamiltonella-negative (0/10 tested for each species).

Localization of Rickettsia
Examination of the symbionts' localization by means of FISH shows a concentration of Rickettsia in the center of the Eretmocerus spp larval body in what seems to be the parasitoid's digestive tract, as well as scattered signals outside of the larval body in the remaining whitefly hemolymph (Fig. 2A). Later on, in the pupal stage, Rickettsia is aggregated in a kidney (or oval) shape within the wasp larva, and is more distal, toward the tip of the abdomen (Fig. 2B). Looking at an image without fluorescence shows an identical kidney-shaped concentration of small, dark spheres that are likely meconia (fecal material, typically retained within the wasp body until late in development) (Fig. 2C). In En. pergandiella, Rickettsia signals can be seen along the digestive tract of the crescent-shaped third instar larva as well as outside of the larva (Fig. 3A). In the pupal stage, however, Rickettsia is clearly present only in the meconia, deposited before pupation on both sides of the pre-pupal wasp (Fig. 3B). These FISH results are consistent with the results of the acquisition and sustainability experiment. In particular, they support the finding that adult En. pergandiella that developed on R + whiteflies are not infected, and suggest that the detection of Rickettsia in pupal En. pergandiella by PCR is likely due to an extraction method that includes the whitefly cuticle and meconial pellets that surrounds the pupal wasp.
Electron micrographs of Er. eremicus reveal the presence of bacteria inside the ovaries, within follicular epithelial cells, but not within the oocytes (Fig. 4). Bacteria were also seen right outside the ovary, adjacent to the tunica propria, the ovarian envelope (Fig. 5). The germarium also shows bacteria among stem-, pre-follicle-, and nurse cell nuclei (Fig. 6). The determination that these bacteria are Rickettsia is supported by: 1) Denaturating gradient gel electrophoresis (DGGE) analysis of the bacteria present in Er. eremicus using general 16S rRNA primers that target most known bacteria. A single band, corresponding to Rickettsia was found in this analysis (data not shown). 2) Diagnostic PCR using specific primers designed for B. tabaci symbionts (Hamiltonella, Wolbachia, Cardinium, Arsenophonus and Rickettsia) showed bands only for Rickettsia in the Er. eremicus, as well as for the positive controls in all other cases (data not shown).

Transmission from B. tabaci to parasitoids
Eretmocerus. Approximately 30% of the uninfected (R 2 ) Eretmocerus adult wasps (from both species) that were exposed to R + whiteflies as adults were subsequently infected with Rickettsia ( Fig. 7A & 7B). The proportion of infected females was significantly higher than the proportion of infected males (Er. emiratus: 56% infected females vs. 6.7% infected males, x 2 32 = 8.8, P,0.01; Er. eremicus: 44% infected females vs. 11% infected males, x 2 43 = 5.8, P = 0.016), suggesting that host-feeding, in which females pierce hosts with their ovipositor and imbibe host hemolymph, is more likely a source of Rickettsia than feeding on honeydew (which both sexes do) or simple contact with contaminated insect surfaces. A much higher proportion of those wasps that developed inside R + whiteflies were infected: 84% of Er. emiratus and 93% of Er. eremicus emerged as Rickettsia infected wasps (Fig. 7A & 7B). Thus, transmission of Rickettsia from infected whitefly hosts to Eretmocerus occurred at the greatest rate during parasitoid development, and to a much lower extent via host feeding by adults. All of the controls, i.e. R 2 wasps that were not exposed to any hosts and R 2 wasps that were exposed to R 2 hosts, were Rickettsia-free. Rickettsia infections that were acquired by host feeding seemed to be largely transient, as the proportion of infected females decreased sharply four days after removal from hosts (Er. emiratus: 15/20 infected immediately after exposure to hosts, vs. 2/10 infected four days later; Er. eremicus: 19/20 and 1/ 10 infected at the two time points, respectively). In contrast with the pattern seen for Rickettsia, Hamiltonella was not detected in any of the Eretmocerus wasps that fed or developed on Hamiltonellainfected whiteflies (0/26 tested).
En. Pergandiella. Almost all (15 out of 16) of the adult females were infected with Rickettsia after exposure to R + hosts, compared to only one infected male (x 2 32 = 25.5, P,0.0001) (Fig. 7C). Rickettsia acquired by host feeding and exposure to honeydew was also transient in En. pergandiella female adults: 17/20 were infected after exposure to R + hosts, while 0/10 were infected four days later. None of the wasps that developed inside an R + host were infected. As was found in Eretmocerus, Hamiltonella was also not detected in any of the En. pergandiella wasps exposed to infected whiteflies (0/23 tested).

Vertical transmission
Eretmocerus wasps that developed inside R 2 whitefly hosts emerged as uninfected wasps, even when their mothers were infected throughout their lifetime (0/20 Er. eremicus, 0/35 Er. emiratus infected, Fig. 7 A, B). There was no difference between the two experimental treatment groups, i.e., wasps with multiple generations of exposure to infected whiteflies prior to the experiment, and wasps with a single generation of exposure (parents). These experiments provide no evidence of vertical transmission of Rickettsia. Vertical transmission was not tested in En. pergandiella because Rickettsia infection did not persist in the adults of this species.

Discussion
Interspecific horizontal transmission of facultative intracellular symbionts is believed to occur rarely, and little empirical evidence of such transfers exist [e.g. 20,21,23]. That horizontal transmission between species must have occurred, however, is amply demonstrated in phylogenetic studies that show little concordance between host and symbiont phylogenies [1,2]. The results presented here demonstrate distinct transmission patterns of secondary symbionts between trophic levels and reveal differences in those patterns between two closely related parasitoid host genera. Further, we show that Rickettsia that is ingested during wasp larval development may penetrate the host hemocoel and infect the ovaries, but do not appear to invade the developing oocytes (Figs. 4-6), preventing vertical transmission in the wasp.

Horizontal transmission
The variation we document in the transmission of Rickettsia from whiteflies to parasitoids highlights two possible views of horizontal transmission. From an evolutionary point of view (most often used in the symbiont literature), our results show no transmission of secondary symbionts from B. tabaci to parasitoids that result in a heritable infection. From a mechanistic point of view, however, we document the transmission of a microorganism from one individual to another, unrelated, individual within the same generation, a necessary precondition of a novel heritable infection in a population. Further, we show that symbionts acquired by feeding may be ultimately excreted (''contamination''), or invade the hemocoel and persist throughout the host lifetime, two distinct and sequential steps in the establishment of a long term association.
Rickettsia established a transtadial infection in Eretmocerus wasps, i.e. Rickettsia sustained in Eretmocerus from the larval stage through adulthood, but was not transmitted vertically. The FISH results indicate that Rickettsia was concentrated in the lower abdomen of the adult Eretmocerus wasps (Fig. 2C & 2D). The electron micrographs show that Rickettsia reached the ovaries of Eretmocerus but did not penetrate the germ line. Instead, it was found in the follicular epithelium surrounding the eggs and also in tissues abutting the ovaries (Figs. 4-6). The fact that Rickettsia is found within or in close proximity to the ovaries suggests that like other vertically transmitted bacteria, Rickettsia requires admission to the germ line for its spread and persistence in host insect populations. In their thorough study, Frydman et al [26] found that injected Wolbachia migrate and enter the Drosophila germline via the somatic stem cell niche in the germarium, from which follicular epithelial cells develop. Our results suggest, for Rickettsia at least, that invading the oocyte may require an adaptation distinct from the ability to find and invade the ovaries. Nonetheless, the inability of Rickettsia to invade the germ line of Eretmocerus may be a result of a defense mechanism of the latter.
The frequency of interspecific horizontal transmission in endosymbiosis of arthropods is clearly low and variable (excluding disease agents vectored by ticks etc). Possibly, the paucity of empirical studies conceals a number of unpublished negative results. Among published results, the frequency of Wolbachia horizontal transfer between Trichogramma species sharing a common host was 0-40% and the vertical transmission within the recipient species diminished within a few generations [19]. Similarly, Spiroplasma was horizontally transmitted between two species of Drosophila by an ectoparasitic mite vector but the subsequent vertical transmission was very low [43]. Variability of interspecific transmission success was also demonstrated in the study of Russell & Moran [25]: pea aphids were injected with three different symbionts that were obtained from other aphid species. Two symbionts -Hamiltonella and Arsenophonus -were successfully established and maintained for multiple generations in their new host, whereas the third one -Regiella -was not. Grenier et al [22] Figure 7. Horizontal transmission (from R + whiteflies to wasps) and vertical transmission (from R + wasps to progeny) of Rickettsia to males and females of Er. emiratus (top), Er. eremicus (middle) and En. pergandiella (bottom). 'P' are R 2 wasps that were exposed to R + whiteflies for 24 hrs (horizontal transmission via host feeding and/or honeydew), 'F 1 ' are their resulting progeny that developed in R + hosts (also horizontal transmission), and 'F 2 ' are progeny of F 1 that were exposed to R 2 hosts (vertical transmission). The numbers above the columns are the sample size, n, from which the proportion of infected wasps was calculated. See also Fig. 1  reported successful horizontal transfer of Wolbachia from one species of Trichogramma to another via microinjection, followed by stable vertical transmission, but the efficiency of this process was low. To the best of our knowledge, the only study that describes symbiont horizontal transmission from a host to its parasitoid is that of Heath et al. [44], in which Wolbachia was weakly transmitted (3.2%) from an infected Drosophila host to a parasitoid, and subsequently diminished within four generations. Compared to these studies, the efficiency of Rickettsia transmission from the host, B. tabaci, to Eretmocerus wasps was very high and yet no vertical transmission was observed. Our results therefore support the notion that invasion of the germ line may be the greatest challenge for symbionts invading novel hosts. Among parasitoids, maternally transmitted Rickettsia was so far only found in a leaf miner parasitoid, where it causes parthenogenesis. However, it is not known whether this symbiont is present also in its hosts, which may be indicative of inter-trophic horizontal transmission [45].

Differences among hosts
Why does Rickettsia establish (even if for only one generation) in the Eretmocerus adults, whereas it appears to be completely excreted by the Encarsia? Encarsia embryos and larvae are in intimate, direct contact with the host's hemolymph throughout their development, whereas Eretmocerus become in contact with the host's hemolymph only in the third instar, due to the unique capsule in which the larval wasps reside [39]. Hence, our finding that Eretmocerus acquire Rickettsia while Encarsia do not is, at first, counterintuitive. It is possible that these differences relate to the timing of the deposition of the meconium, fecal material. In En. pergandiella the mid-gut and the hind-gut are not continuous in early development, when the larva is in a fluid environment, but join only at the end of the third instar stage. Subsequently, the prepupal wasp deposits the meconium, with Rickettsia in it, and then pupates [40]. Eretmocerus spp., in contrast, excrete the meconium only after the adult emerges, so meconia, with Rickettsia in them, are present in the body throughout metamorphosis. It may be that Rickettsia has the opportunity to invade new tissues during this phase, when tissues are breaking down and new ones are being built. This idea is supported by the observation that adult acquisition of the symbiont by consumption of honeydew or host hemolymph does not persist. Nevertheless, other routes of infection cannot be excluded: Rickettsia may get to the ovaries by crossing the larval mid-gut tissues, which in aphelinid larvae typically bear very few cells, no typical epithelium and no membranes (Dan Gerling, pers. comm.).

Differences between symbionts
Adult wasps of all species in our study acquired Rickettsia but not Hamiltonella from host feeding. One possible reason for that difference may be the localization of the two symbionts: Rickettsia is abundant and accessible in the host hemolymph consumed during the process of host-feeding while Hamiltonella is sequestered within the bacteriomes [30]. Another explanation is required, however, for why Eretmocerus, during their development inside a host, acquire Rickettsia but not Hamiltonella, since the parasitoid larvae consume the entire host contents before pupation. Indeed, it seems that Rickettsia is generally more prone to horizontal transmission (e.g. to mammalian hosts in the case of the disease agents, or to plants in the case of insect-vectored plant pathogens) than many facultative intracellular symbiont lineages. To date, Rickettsia has been found in many host lineages [2,31], whereas Hamiltonella has so far been revealed only in aphids, in whiteflies and in one psyllid species [2,10]. A possible mechanism for a greater propensity for horizontal transmission is greater symbiont mobility: while little is known about mobility in most symbiont lineages, some Rickettsia are able to move between cells and tissues using actin filaments [46,47]. We know nothing about the mobility of Hamiltonella, yet the fact that Hamiltonella reside in B. tabaci bacteriocytes where they are vertically transmitted along with the primary symbiont Portiera, suggests that Hamiltonella may have more limited mobility.
Naturally, an interesting question for further research would be to look for phenotypic effects of Rickettsia infection on the parasitoids. Preliminary results showed no differences between R+ and R2 Er. emiratus with regards to fecundity, longevity and sex ratio, thus more fitness parameters need to be explored to address this question (Chiel et al., unpublished results).
To conclude, our study is one of few empirical demonstrations of the routes and barriers to horizontal transmission of facultative symbionts. These data are especially relevant to the often repeated idea that parasitoids or predators may be instrumental agents for moving symbionts from one host lineage to the next. In fact, this notion has some phylogenetic support [e.g. 12,48], but in some cases, enemies have likely been wrongly diagnosed by PCR as being stably infected when the symbionts are simply present in the gut along with the prey or host material [49]. Our data suggest that host-parasitoid transmission may, nonetheless, be one way in which symbionts acquire new hosts. Given that the symbiont is on the doorstep of vertical transmission, it is not hard to imagine that some lineage might, with time, acquire an adaptation that improves the precision of cell targeting in this new host lineage to get the symbiont over the threshold. Lastly, our study underscores how little we currently know about the processes of dispersal of symbionts to new host lineages, and the within-host movement and germ-line invasion processes necessary for them to stay once they get there.