Introduction

The life cycle of the malaria parasite (Plasmodium spp) involves critical steps within the mosquito midgut, an environment it shares with bacteria. Various physical (for example, membranes, receptors), biochemical (for example, proteases, chitinases) and physiological (for example, immunological factors) barriers protect the insect against the parasite (for reviews see Sinden, 2002; Meister et al., 2004), but the largest losses in parasite titer occur when Plasmodium zygotes develop into ookinetes within the adult mosquito midgut. During this transition, the parasite may undergo more than two orders of magnitude loss in abundance (Vaughan et al., 1994). Although the negative impact of midgut bacteria on parasite development has been observed in the past (Pumpuni et al., 1993; Lowenberger et al., 1999; Gonzalez-Ceron et al., 2003), a symbiotic or consistent association between Anopheles gambiae s.l. and specific species of bacteria has not been described. Among other Anopheles species, An. stephensi has been reported to stably associate with bacteria of the genus Asaia (Favia et al., 2007). Such a relationship would make use of genetically transformed bacteria in paratransgenesis more efficient (Beard et al., 2002). Paratransgenics refers to the genetic transformation of a bacterium, typically a vector endosymbiont, to produce antiparasitic molecules, thus inhibiting parasite development (Conte, 1997; Beard et al., 2002). Since Anopheles mosquitoes are currently considered a dead end host for midgut bacteria that are believed to ‘perish when the adult mosquito dies’ (Riehle and Jacobs-Lorena, 2005), the current paradigm for paratransgenesis in Anopheles focuses on bacteria that opportunistically infect adult mosquitoes (Lindh et al., 2006; Favia et al., 2007; Riehle et al., 2007).

The most significant perturbation affecting mosquito midgut microbial communities occurs during metamorphosis from pupae to adults. For bacteria inhabiting the midgut, the process has been characterized as similar to ‘gut sterilization’ (Moll et al., 2001), which, if correct, would imply that environmentally derived bacteria that are present in larval midguts (through feeding) are greatly attenuated or are absent in adult midguts. It should be stressed, however, that microbial studies of the midgut of Anopheles are scarce, and have depended mainly on traditional culture-based techniques (Pumpuni et al., 1996; Straif et al., 1998; Gonzalez-Ceron et al., 2003; Lindh et al., 2005). One study has coupled culture- and PCR-based approaches to characterize An. gambiae s.l. midgut bacteria (Lindh et al., 2005). Therefore, the application of other nucleic acid-based tools, such as denaturing gradient gel electrophoresis (DGGE) (Muyzer et al., 1993), to study these systems is timely.

In this study, we provide evidence from PCR, DGGE and clone library analyses that adult field-captured An. gambiae s.l. mosquitoes in a Kenyan village are consistently associated with a Thorsellia anophelis lineage that was also detected in the surface microlayer of rice paddies. Unlike most members of the Enterobacteriaceae, this bacterium grows slowly in rich media, but is well adapted to the anopheline midgut because of its ability to utilize blood to enhance growth and its tolerance to alkaline conditions in the anopheline midgut.

Materials and methods

Study site and sample collection

The study site is situated within the Mwea Irrigation and Agricultural Development experimental station, located approximately 100 km northeast of Nairobi, Kenya (37° 20′ E and 0° 40′ S), at an elevation of 1159 m above sea level. The site falls within a region of low stable transmission of the malaria parasite, which infects around 20% of the human population (Mutero et al., 2000). A detailed description of the site has been reported elsewhere (Muturi et al., 2006). The experimental plots were laid out in a 0.4 ha area, which is situated within 100 m of Kariua, a village inhabited mostly by subsistence-level farmers living in houses with walls constructed of mud or wood. The experimental plots are arranged in eight blocks (rows) subdivided into eight plots (6.3 × 3.15 m), with each block hydrologically isolated using unidirectional inflow and outflow canals. Among the 64 plots, two adjacent rows (designated rows 1 and 2) were set aside for the activities described in this study. All the plots were exposed to natural colonization by mosquitoes. Previous fieldwork has shown that the most abundant anopheline species within Mwea Irrigation and Agricultural Development rice paddies is An. arabiensis (a sub-species of An. gambiae s.l.) (Mutero et al., 2000).

Mosquito and water samples were collected in June and July 2005. Mosquito larvae and rice paddy water were collected during a fallow period when the plots had been continuously flooded for 6 days prior to sampling. Surface microlayer (SML) samples of two randomly selected plots from each row (designated plots A and B) were collected using a stainless steel mesh screen measuring 400 cm2, with a mesh size of 1.25 and 0.36 mm wire diameter as described previously (Agogue et al., 2004). Mosquito larvae and bulk water from four plots (plots A, B, C and D; two plots per row) were sampled with a 350 ml plastic dipper, using standard dipping techniques (Service, 1993). Since the larval abundance was low during the time of sampling, mosquito larvae from the remaining 12 plots were also collected and pooled according to the row. Among the pooled water samples from each plot, 300 ml (from 30 pooled dips of SML water) and 500 ml (from 20 pooled dips of bulk water) aliquots were placed in sterile Nalgene bottles (Nalge Nunc International, Rochester, NY, USA) and transported on ice immediately to the laboratory. All water and larval samples were processed within 24 h of collection. Female An. gambiae s.l. fourth instar larvae were selected morphologically (Gillies and Coetzee, 1987) and rinsed three times in filter-sterilized phosphate-buffered saline (137 mM sodium chloride, 2.7 mM potassium chloride and 10 mM sodium phosphate dibasic/potassium phosphate monobasic, pH 7.4) prior to dissection. The larvae were dissected using aseptic techniques, with the aid of a dissecting microscope (Zeiss Stemi 1000; Carl Zeiss Inc., Jena, Germany). Each midgut sample subjected to DNA extraction consisted of 30 larval midguts pooled together in sterile phosphate-buffered saline and corresponded to larvae collected from two adjacent rows of plots.

Indoor resting mosquitoes from two houses (designated house 1 and house 2) approximately 100 m apart were captured in the morning hours starting around 0700 hours using the pyrethrum spray collection method (Service, 1993). Immediately after collection, mosquitoes were transported to the laboratory and adult female An. gambiae s.l. mosquitoes were selected based on morphology (Gillies and Coetzee, 1987). The samples were stored at −20 °C in phosphate-buffered saline solution containing 30% glycerol and 1% Tween 80. Dissection and midgut collection of adult mosquitoes was carried out as described above. Each pooled sample of 30 midguts corresponded to adult mosquitoes collected from one house. Greater than 50% of the extracted midguts were derived from blood-fed anophelines.

DNA extraction, Bacterial PCR and DGGE conditions

DNA from all samples was extracted using the UltraClean soil DNA isolation kit (MO BIO Laboratories Inc., Carlsbad, CA, USA) following the manufacturer's instructions. Each water sample consisted of pellet obtained from a 250 ml volume centrifuged for 20 min at 5100 g. Each midgut extract consisted of 30 pooled midguts suspended in 100 μl phosphate-buffered saline. DNA extractions were performed at the International Centre for Insect Physiology and Ecology laboratory in Nairobi, Kenya. DNA samples were transported to the USA for subsequent analyses. PCR reactions were performed in a Mastercycler ep thermocycler (Eppendorf North America, Inc., Westbury, NY, USA). PCR targeting the domain Bacteria and DGGE conditions were based on optimized conditions reported previously using primers 954f(GC) and 1369r (Yu and Morrison, 2004), except for the following modifications: PCR reactions (50 μl) contained 1 × PCR buffer, 200 μM deoxyribonucleotide triphosphates, 500 nM of each primer, 1.5 mM MgCl2, 1.25 U of Ex Taq DNA polymerase (Hotstart version; Takara Mirus Bio, Madison, WI, USA) and 10 ng template DNA. The thermocycling program included an initial denaturation step at 94 °C for 2 min followed by 10 cycles of touchdown PCR (denaturation at 94 °C for 30 s, annealing at 61 °C for 30 s with 0.5 °C decrement per cycle and extension at 72 °C for 1 min). The touchdown cycles were followed by 25 cycles of regular PCR using 56 °C as the annealing temperature. Amplicons of the expected size (456 bp, excluding the GC-clamp) were gel purified (MinElute gel extraction kit; Qiagen Inc., Valencia, CA, USA) following the manufacturer's instructions and resolved by DGGE (Dcode Universal Mutation detection system; Bio-Rad Laboratories, Hercules, CA, USA) using previously described conditions (Yu and Morrison, 2004). All DGGE gels included two lanes of standards (DGGE Marker II; Wako Chemicals USA Inc., Richmond, VA, USA). The DGGE bands were stained using SYBR Gold (Molecular Probes Inc., Eugene, OR, USA) and the images were captured using a Gel Logic 100 imaging system (Kodak Molecular Imaging Systems, New Haven, CT, USA). All PCR-DGGE steps were performed twice to determine the reproducibility of the results.

DGGE bands were detected and analyzed using Quantity One software (version 4.6.2; Bio-Rad Laboratories, Hercules, CA, USA). We considered only bands with intensities exceeding 2.0% of the most intense band in each lane for subsequent analyses. Lanes were compared using Dice's similarity coefficient and cluster analysis was performed using unweighted pair group method with arithmetic means. Taxon richness was determined from the number of bands in each lane, while the Shannon-Weaver diversity index, H, was calculated from H=−∑(pi)(log pi), where p is the proportion of an individual band's intensity (peak height) relative to the cumulative peak height (Atlas and Bartha, 1998). The identities of the most intense DGGE bands were determined by excising the bands, reamplifying the PCR products using primers 954f and 1369r and sequencing the purified PCR products through the University of Michigan's DNA Sequencing Core facilities.

T. anophelis clone libraries, sequencing and phylogenetic analysis

Partial sequences of the T. anophelis 16S rRNA gene were amplified from the environmental DNA extracts described above using primers designated as Tha226f (5′-TTCGGGCCTCACGCACTAG-3′) and Tha1293r (5′-ATCCGGACTTTGAGG-3′), based on Escherichia coli numbering. The expected product size including the primers is 1101 bp. The primer sequences perfectly matched the 16S rRNA gene of T. anophelis CCUG 49520T and DGGE band-derived Thorsellia sequences, while each primer contained at least three mismatches to closely related Gammaproteobacteria sequences (see below). PCR was carried out as described above, except the extension times were increased to 1 min and 15 s.

Four clone libraries, each consisting of 48 clones, derived from T. anophelis 16S rRNA gene amplified by Tha226f and Tha1293r were constructed using a cloning kit (TOPO TA cloning kit for sequencing, Invitrogen, Carlsbad, CA, USA) and, respectively, were derived from pooled amplicons generated from SML water, bulk water and adult midguts from houses 1 and 2. The clones were initially screened by colony PCR (Gussow and Clackson, 1989), followed by digestion of cloned sequences using the restriction enzyme MvnI (Roche Diagnostics Corp., Indianapolis, IN, USA). Digests were loaded on 2% agarose gels to evaluate restriction patterns. Inserts from 41 clones redundantly representing seven unique restriction pattern types were sequenced bi-directionally through the University of Michigan's DNA Core Sequencing facilities. The 33 unique sequences generated from 41 clones were deposited in GenBank under accession numbers EF434748–EF434780.

DNA sequences were assembled and aligned using BioEdit software version 7.0 (Hall, 1999) freely available at http://www.mbio.ncsu.edu/BioEdit/bioedit.html. The sequences selected for phylogenetic analysis consisted of the 33 unique T. anophelis sequences generated in this study, T. anophelis CCUG 49520T and 42 other insect-associated and free-living representatives of the Gammaproteobacteria. Compressed branches in Figure 3 represent the following genera: insect-associated Arsenophonus: a symbiont of Wahlgreniella nervata, a symbiont of Bemisia tabaci, a symbiont of Austaliococcus greville, a symbiont of Diaphorina citri and a symbiont of Glycaspis brimblecombei; free-living Proteus: P. vulgaris and P. hauseri; representatives of Vibrio: V. splendidus and V. cholerae; and Pseudomonas aeruginosa. The compressed branches in Figure 3 corresponding to the Pasteurellaceae consist of Haemophilus sp oral clone BJ021, Actinobacillus indolicus and Pasteurella multocida. The accession numbers of these taxa as well as other insect-associated and free-living representatives of the Enterobacteriaceae are detailed in Supplementary Information.

Phylogenetic affiliations based on the neighbor-joining algorithm were evaluated using three different methods for distance matrix calculations: Kimura-2-parameter, Tajima-Nei and Tamura-Nei methods. All methods assumed different substitution rates among sites requiring an estimate of the γ parameter, which was calculated using the software package Diverge 2.0 (Gu, 1999) freely available at http://xgu.zool.iastate.edu. Distance matrix calculations and phylogenetic tree construction were performed using the software package MEGA 3.1 (Kumar et al., 2001) freely available at http://www.megasoftware.net. The robustness of each resulting tree was evaluated by bootstrap analysis based on 1000 replications. In addition to distance-based neighbor-joining trees, we also compared trees based on maximum parsimony algorithm in MEGA 3.1 and the maximum likelihood algorithm in Phylip 3.66 (Felsenstein, 2006). We have included the bootstrapped Tajima-Nei and maximum parsimony trees in Supplementary Figures.

T. anophelis growth measurements

T. anophelis CCUG 49520T cultures were maintained on agar slants containing trypticase soy agar supplemented with 5% sheep's blood (Blood agar, Remel, Lenexa, KS, USA). T. anophelis was grown overnight with shaking (all incubations were done at 30 °C) in trypticase soy broth (TSB; Difco Laboratories Livonia MI, USA) supplemented with 5% sheep's blood (Hemostat Laboratories, Dixon, CA, USA). This culture was used to inoculate duplicate flasks containing 150 ml of TSB and TSB supplemented with 5% sheep's blood. The growth of the shaken cultures was monitored by counting colony forming units from duplicate serial dilutions plated onto 5% sheep's blood agar plates (Remel, Lenexa, KS, USA). A similar protocol was followed to monitor the growth of T. anophelis in TSB maintained in either pH 7.4 or 9.1. The media were buffered at the appropriate pH using a final concentration of 0.1 M sodium carbonate-bicarbonate buffer following the proportions recommended by Delory (1945).

Results and discussion

Since the bulk of the current global burden of malaria is borne by countries in Africa and since flooded rice paddies serve as larval habitats and are a known risk factor for malaria transmission (Koudou et al., 2005), we selected a rice-growing region in central Kenya (Mwea) as our study site. Rice paddy water and An. gambiae s.l. larvae were collected from experimental rice plots. Since anopheline larvae are known to feed mainly on the water SML (Merritt et al., 1992), we collected both bulk water (from four plots) and water SML samples (from two of these four plots). An. gambiae s.l. female larvae (all identifications were accomplished morphologically) were collected from the same plots as well as from adjacent plots. The experimental rice plots selected for sampling were adjacent to a village, from where we selected two residences within 100 m of the rice paddies to collect adult female An. gambiae s.l. mosquitoes.

DNA extracted from water (bulk and SML) and midgut (larvae and adults) samples served as sources of templates for PCR targeting the 16S rRNA gene of the domain Bacteria. The results of PCR followed by DGGE on the different water and mosquito midgut samples are shown in Figure 1. The most similar bacterial community profiles within sample types (defined as water, larval midguts and adult midguts) were observed in the rice paddy water samples; there were no clear distinctions between bulk and SML water community profiles. Our bulk water samples consisted of the entire water column (including the SML), which may partially explain the similarities between bulk water and the SML. On the other hand, the least similar profiles were observed between two pooled samples of larval midguts, corresponding to larvae collected from two adjacent rows of plots. Cluster analysis grouped one sample of larval midguts (designated LM-2 in Figure 1) to rice paddy water samples, while the other sample (LM-1) was more similar to the adult midgut community profiles. These results are consistent with the bacterial communities in larval midguts as an intermediate stage between the aquatic habitat and the adult stage. Bacterial diversity as measured by the Shannon-Weaver diversity index, H, in the SML samples ranged from 2.48 to 2.72, which was slightly higher than those observed for bulk water (1.32–2.42). Bacterial diversity indices in all midgut samples were within the range of H values observed for water (larvae, H=2.26–2.63; adults, H=2.16–2.52).

Figure 1
figure 1

Cluster analysis of 16S rRNA gene-targeted DGGE profiles of bacterial communities in rice paddies and mosquito midguts. LM=larval midgut; AM=adult midgut; BW=bulk water; SW=surface microlayer water. The bands labeled A–E represent DNA that were sequenced and identified as T. anophelis. Band F failed to generate a reliable sequence, while band G in lane LM-2 had similar mobility to bands B and D but was identified as a species of Chitinophaga. The sources of the water, larvae and adult mosquitoes are the same as those described in Figure 2.

Full size image

The most significant finding of PCR-DGGE was the consistent detection of 16S rRNA of populations that appears predominant across two sets of adult midguts collected from different houses (the most intense bands in lanes AM-1 and AM-2, Figure 1). Based on their relative band intensities, these populations constituted around 40% of the total bacterial community. Alignments of sequences obtained from gel-excised DGGE bands from lanes AM-1 and AM-2 to sequences available in GenBank (http://www.ncbi.nlm.nih.gov/BLAST/) and the Ribosomal Database Project (RDP-II) (http://rdp.cme.msu.edu/) revealed the closest match as a recently described bacterium with proposed genus and species names, T. anophelis CCUG 49520T (Kampfer et al., 2006). Since no similar T. anophelis DGGE bands were evident in the water and larval midgut samples, the results of PCR-DGGE would appear to support the prevailing view that adult mosquito midguts are populated mostly by bacteria obtained opportunistically (Riehle and Jacobs-Lorena, 2005). We sought to confirm this by assaying specifically for T. anophelis 16S rRNA genes in the various environments using T. anophelis-specific PCR primers (Tha226f and Tha1293r), which were designed in this study (see Materials and methods). Because the reverse primer Tha1293r hybridizes to an internal region within the fragment amplified by the DGGE primers (954f and 1369r), we obtained only partial alignments (82%) between the DGGE sequences and the sequences obtained from Thorsellia primers.

The results of PCR using the T. anophelis-specific PCR primers are shown in Figure 2. T. anophelis DNA was consistently detected in adult midguts and water SML, suggesting that these midgut populations are derived through ingestion by larvae from their aquatic environments, and are not acquired post-metamorphosis. Their absence in larval midguts suggests that their levels here are below the detection limit of the PCR, due to the fast gut passage rate of material through mosquito larvae (Nilsson, 1986; Merritt et al., 1992), as well as the slow growth rate of T. anophelis (Kampfer et al., 2006), which we confirmed in this study (see below). It should be noted that the 30 midguts processed per mosquito sample correspond to less than 100 μl volume of sample compared to the 250 ml of water samples that were processed.

Figure 2
figure 2

16S rRNA gene PCR products generated using T. anophelis-specific primers. Lanes 1,2, plot A SML, lanes 3,4, plot B SML; lanes 5,6, plot A bulk water; lanes 7,8, plot B bulk water; lanes 9,10, plot C bulk water; lanes 11,12, plot D bulk water; lanes 13,14, adult midguts from house 1; lanes 15,16, adult midguts from house 2; lanes 17,18, larval midguts collected from row 1 containing eight plots that includes plots A and C; lanes 19–20, larval midguts from adjacent row 2 of plots that includes plots B and D; lane 21, no template.

Full size image

Four clone libraries each consisting of 48 cloned T. anophelis 16S rRNA gene amplicons derived from water (bulk and SML) and adult midguts (houses 1 and 2) were screened based on banding patterns generated by digestion with the restriction enzyme MvnI. Seven unique restriction types were identified and 41 clones were selected to redundantly represent each restriction type. Subsequent sequencing found that 33 of the 41 clones (80%) represented unique sequences. All unique sequences, including T. anophelis CCUG 49520T, exhibited pairwise sequence similarities ranging from 97.1% to 99.9%, based on an alignment of 1053 nucleotides. At this point, it is not possible to conclude whether all the obtained sequences correspond to a single species. However, this remarkable diversity of sequences suggests at least the existence of numerous strains of T. anophelis that can be resolved into at least four major lineages designated TA-1 to TA-4 in Figure 3. Sequences derived from rice paddy water comprised the majority of cluster TA-3 and the entirety of cluster TA-2, while the bulk of cluster TA-1 comprised midgut sequences. Excluding the single Thorsellia sequence available in GenBank, the highest BLAST alignment scores for all members of cluster TA-1 correspond to members of the genus Arsenophonus. Members of clusters TA-2 and TA-3 were also closely affiliated (by highest alignment scores) to members of the genus Pectobacterium or other unidentified Gammaproteobacteria. Cluster TA-4 consisted of a single water clone that was most closely affiliated with T. anophelis CCUG 49520T, which was originally isolated from Anopheles midguts in Lwanda, Kenya, approximately 550 km northwest of our study site (Lindh et al., 2005). Therefore, Thorsellia appears to be widely distributed in Kenya, although the geographic separation may influence the composition and distribution of strains. In Mwea, the more diverse collection of T. anophelis sequences occurred in rice paddy water: 24 clones representing all clusters TA-1, TA-2, TA-3 and TA-4 were detected at 21%, 46%, 29% and 4%, respectively. The midgut samples were less diverse, consisting of 17 clones represented by only two clusters, TA-1 (71%) and TA-3. The sequences derived from DGGE gels and clone libraries overlap only partially (82%) and should therefore be compared cautiously. However, based on these partial alignments, bands designated as B and D in Figure 1 are most closely similar to cluster TA-1 (zero to one base mismatch). The presence of ambiguous bases in the sequences derived from the other DGGE bands precludes reliable affiliation with the other TA clusters.

Figure 3
figure 3

Phylogenetic analysis representing 76 sequences of Gammaproteobacteria. Individual taxa belonging to Thorsellia, Arsenophonus, Proteus, Vibrio, Pseudomonas and Pasteurellaceae are listed in the Methods while all accession numbers and descriptions of other taxa are provided in the Supplementary Information. The numbers in parentheses after three of the four T. anophelis clusters represent the number of unique sequences in each cluster and the proportion of water sequences in the cluster. DNA distance and treeing methods used were the Tamura–Nei model and neighbor-joining method, respectively. The numbers are bootstrap values from 1000 replicates. Molecular clock scale was calibrated by the divergence between E. coli and S. typhimurium. Two trees offering expanded views of the Gammaproteobacteria lineages are presented in the Supplementary Information.

Full size image

Since mosquito larvae likely do not discriminate between these different taxa during feeding, these results are consistent with the following scenario: only certain T. anophelis strains (mostly cluster TA-1 and to a lesser extent TA-3) are able to survive and proliferate during and after metamorphosis, respectively. Based on its high frequency in the aquatic environment, cluster TA-2 represents a group that is most likely ingested, but does not survive during subsequent growth and development, because of digestive, excretory or gut remodeling processes.

T. anophelis sequences formed a distinct and robust clade within the class Gammaproteobacteria (Figure 3); it is currently classified under the family Enterobacteriaceae in both NCBI and RDP classification schemes. Although cluster TA-1 aligned consistently with members of Arsenophonus (above), phylogenetic analysis provided only weak bootstrap support for affiliating the entire Thorsellia clade with this genus. Surprisingly, our analyses (distance based, maximum parsimony and maximum likelihood) consistently grouped members of the Pasteurellaceae with Thorsellia, Arsenophonus and Proteus. While 16S rRNA gene sequences alone should not be relied upon to fully resolve the phylogenetic affiliations within the Gammaproteobacteria, it is interesting to note that some members of the Pasteurellaceae are α-hemolytic when grown on blood agar plates (Donachie et al., 1995), a trait that we report here to be exhibited by T. anophelis (see below). Additionally, we found T. anophelis CCUG 49520T to be oxidase positive (Oxidase test discs, Sigma-Aldrich, St Louis, MO, USA), a characteristic shared by members of the Vibrionaceae and Pasteurellaceae, but not Enterobacteriaceae.

The genus Arsenophonus is comprised mostly of members that are endosymbiotic to a variety of arthropod hosts, including insects and ticks (Gherna et al., 1991; Grindle et al., 2003). Artificial culture has been achieved only on insect cell lines (Hypsa and Dale, 1997; Dale et al., 2006). Different species of Arsenophonus from different hosts share >99% 16S rRNA gene sequence similarity, suggesting recent host acquisition (Dale et al., 2006). The divergences between members of the T. anophelis clusters are substantially greater, suggesting a more ancient time of divergence between sublineages. By calibrating the phylogenetic tree in Figure 3 to the divergence between E. coli and Salmonella typhimurium at 130±10 million years ago (Ochman and Wilson, 1987), we estimate the T. anophelis divergence to have occurred around 50–70 million years ago (the lower limit is based on calibrated maximum likelihood and neighbor joining trees, while the upper limit is based on minimum evolution trees). This matches the estimated divergence between the lineages of Anopheles at 70–85 million years ago (Krzywinski et al., 2006). This concordance raises the possibility that mosquitoes have played a role in the evolutionary diversification of Thorsellia and vice versa. Based on the scenario we describe above, we speculate that a lineage such as cluster TA-1, which predominates in adult An. gambiae s.l. midguts, is more widely disseminated geographically by being transported from and deposited back to the aquatic environment as a direct or indirect consequence of oviposition on water. Whether cluster TA-1 represents an Anopheles symbiont will require further study, but its affiliation with Arsenophonus suggests this possibility.

The possibility of a close association between T. anophelis and Anopheles mosquitoes is further strengthened by two characteristics of strain CCUG 49520T, that is, its alkali tolerance and its ability to utilize blood to enhance growth (Figure 4). The alkaline conditions in the mosquito midgut (maximum pH of 10.5 in larvae; pH of 8.0–9.5 in adults) have long been recognized (Dadd, 1975; Corena et al., 2005). Therefore, a mosquito midgut association would explain the evolutionary significance of alkali tolerance in T. anophelis, a trait that would be useless in rice paddy environments. In contrast to most Enterobacteriaceae, T. anophelis grows slowly in various media and carbon sources (Kampfer et al., 2006). In trypticase soy agar supplemented with 5% sheep's blood, we observed α-hemolysis and growth enhancement by a factor of 1.4 (generation time of 1.4 h with blood supplementation vs 2.0 h without blood). This enhanced growth rate is still relatively low compared to most Enterobacteriaceae grown in nutrient rich media. Therefore, although the physiological adaptations we describe here help explain the association of T. anophelis with adult mosquitoes that have acquired a blood meal, other factors may be involved in their apparent predominance in this environment. Research on Arsenophonus also suggests that the basis of symbiosis with its host is probably not based on nutrition but on still unknown, but critical, function(s) (Dale et al., 2006).

Figure 4
figure 4

Exponential phase of growth curves of T. anophelis grown in trypticase soy broth (TSB) with and without blood (a) and in TSB maintained at two different pH values (b). Each point represents one of the two independent replicates per sampling time (some points not visible because of overlapping symbols) and the curves were fitted by nonlinear regression using an exponential growth equation. Comparisons of curves from log-transformed data showed that the slopes in (a) are very significantly different (P<0.0001) while those in (b) are equal (P=0.5164). The difference in elevation (significant, P=0.0017) of curves in (b) was due to a longer lag phase in the pH 9.1 treatment. Curves and statistical tests were performed using Prism 4 for Windows (GraphPad Software Inc., San Diego, CA, USA).

Full size image

Another group has recently described an association between An. stephensi and an Alphaproteobacterium of the genus Asaia (Favia et al., 2007). High levels (41% of 16S rRNA gene copies using quantitative PCR; 90% of 16S rRNA gene clones) of this bacterium associated primarily with insectary-reared An. stephensi, although it was also detected at lower levels (5% of 16S rRNA gene clone libraries) in An. gambiae captured from villages in Burkina Faso,West Africa. None of the major DGGE bands sequenced in our study were affiliated to Asaia. On the other hand, Thorsellia has been detected by at least two independent studies in Kenya using culture-(Lindh et al., 2005; Kampfer et al., 2006) and PCR-based techniques (this study). We recognize the need to further evaluate the association between T. anophelis and An. gambiae (or other possible mosquito hosts). In this respect, the primers and environmental sequences generated in this study will be helpful in developing a quantitative molecular assay (for example, real-time PCR) that should supplement further strain/species collection and characterization. The identification of a T. anophelis lineage that is associated with the adult midgut, but can also be detected in the aquatic environments where larvae live provides a promising starting point to determine the feasibility of using this bacterium for paratransgenic control of Plasmodium. Ingestion of genetically transformed T. anophelis by mosquito larvae in the rice paddy environment, where larvae are concentrated, relatively immobile and accessible to human intervention, should provide a more efficient paratransgenic system compared to opportunistic acquisition of bacteria by adult mosquitoes.