Introduction

Insects embrace some 1 000 000 described species on the earth (Stork, 2003), and many of them host phylogenetically diverse microbial symbionts (Buchner, 1965; Bourtzis and Miller, 2003). Some symbionts like Wolbachia in diverse insects are of facultative nature, not essential for their hosts, and usually localized in a broad array of cells and tissues (Werren et al., 2008; Oliver et al., 2010). Other symbionts like Buchnera in aphids and Wigglesworthia in tsetse flies are of obligate nature, essential for survival and reproduction of their hosts, and often localized in specialized cells called bacteriocytes (Baumann, 2005; Moran et al., 2008).

In general, these symbionts are stably maintained through host generations by vertical transmission, but the evolutionary trajectories may be markedly different between the facultative ones and the obligate ones. In the facultative symbiotic association, the symbiont phylogeny is generally incongruent with the host phylogeny, indicating horizontal transfers and/or new acquisitions of the symbionts at considerable frequencies in their evolutionary course. In the obligate symbiotic associations, by contrast, the symbiont phylogeny often mirrors the host phylogeny, reflecting strict vertical symbiont transmission and host–symbiont co-speciation over evolutionary time. Furthermore, these obligate symbionts tend to exhibit remarkable evolutionary patterns such as AT-biased nucleotide composition, accelerated molecular evolution and reduced genome size, which are attributable to the stable and nutrition-rich habitat for the symbionts and also attenuated purifying selection because of small population size and strong bottleneck associated with the lifestyle of the vertically transmitted symbionts (Wernegreen, 2002; Moran et al., 2008).

Such obligate host–symbiont associations may be evolutionarily stable, but acquisitions and/or replacements of obligate symbionts must have taken place occasionally. For example, aphids, mealybugs, whiteflies and psyllids constitute a monophyletic group in the order Hemiptera, and are commonly associated with specific bacterial symbionts within the bacteriocytes: Buchnera in aphids, Tremblaya in mealybugs, Portiera in whiteflies and Carsonella in psyllids. However, these symbionts are phylogenetically distinct between the insect groups, suggesting acquisitions and/or replacements of the bacteriocyte-associated symbionts during their diversification (Baumann, 2005; Moran et al., 2008). Within aphids, although the majority of over 4000 species are associated with Buchnera, about 20 species representing three genera of the tribe Cerataphidini have lost Buchnera and acquired yeast-like fungal symbionts in their body cavity (Fukatsu and Ishikawa, 1992, 1996; Fukatsu et al., 1994; Hongoh and Ishikawa, 2000).

Weevils constitute the most species-rich metazoan family Curculionidae with over 51 000 described species worldwide (Farrell, 1998; McKenna et al., 2009), which provide an ideal opportunity to investigate the evolutionary processes underlying acquisitions, replacements and diversification of novel host–symbiont associations. The bacteriocyte-associated Nardonella symbionts are found from diverse weevil subfamilies including the Dryophthorinae and the Molytinae, have strictly co-speciated with their hosts for over 125 million years, and exhibit highly AT-biased nucleotide compositions and accelerated molecular evolution that are typical of ancient insect symbionts of obligate nature (Lefèvre et al., 2004; Conord et al., 2008; Hosokawa and Fukatsu, 2010). On the other hand, the bacteriocyte-associated Sodalis-allied symbionts are restricted to a few grain weevil species of the genus Sitophilus (Heddi and Nardon, 2005), and the bacteriocyte-associated Curculioniphilus symbionts were recently described from several seed-infesting weevils of the genus Curculio (Toju et al., 2010). Nardonella symbionts, Sodalis-allied symbionts and Curculioniphilus symbionts form distinct clades within the gammaproteobacteria, respectively (Toju et al., 2010). The weevil genera Sitophilus and Curculio constitute distinct lineages within the family Curculionidae: Sitophilus is placed within the Dryophthorinae whereas Curculio clusters with the Molytinae (McKenna et al., 2009). These phylogenetic patterns indicate that, in the evolutionary course of the weevils, the ancient bacteriocyte-associated symbiont Nardonella has experienced at least two independent replacement events: once by Sodalis-allied symbiont in the lineage of Sitophilus weevils, and once by Curculioniphilus symbiont in the lineage of Curculio and allied weevils. It is of interest what evolutionary patterns are observed in the phylogenetically distinct novel symbionts and their hosts, and how replacements to the novel symbionts have affected ecological traits of their hosts such as food plant utilization.

In this study, we collected 36 weevil species representing the genus Curculio and allied genera of the tribe Curculionini, and characterized their symbiotic bacteria on the basis of 16S ribosomal RNA (rRNA) gene sequences. Curculioniphilus symbionts were detected from 27 of the 36 weevil species, and we analyzed their molecular evolutionary patterns, host–symbiont co-speciation and correlation with food plants of their hosts. Then, we retrieved gene sequences of three Sitophilus weevil species and their Sodalis-allied symbionts from the DNA databases, and performed similar molecular phylogenetic analyses. Comparison of these newcomer bacteriocyte-associated symbiont lineages highlighted contrasting evolutionary trajectories and consequences of the novel symbionts that independently replaced the same ancient symbiont.

Materials and methods

Curculionini weevils

The Curculionini is a tribe of weevils within the subfamily Curculioninae (Coleoptera: Curculionidae). Among 18 genera of the tribe, the genus Curculio is the largest and embraces about 350 described species worldwide (Alonso-Zarazaga and Lyal, 1999; Hughes and Vogler, 2004a). In most of the species whose ecological information is available, their host plant range is limited to a small number of species within a single plant genus (Morimoto, 1960, 1962; Hayashi et al., 1984; Fujimoto, 2004; Hughes and Vogler, 2004b). In tropical and subtropical regions, diverse Curculionini weevils are associated with figs (Moraceae), while in Europe and North America, they are infesting hard seeds such as acorns and chestnuts (Fagaceae), hazelnuts and birches (Betulaceae) and pecans and walnuts (Juglandaceae) (Hughes and Vogler, 2004a, 2004b). In Japan, Curculionini weevils utilize host plants from such families as the Caprifoliaceae, Elaeagnaceae, Fabaceae, Lauraceae, Styracaceae, Theaceae and Ulmaceae, in addition to the Fagaceae, Betulaceae and Moraceae (Morimoto, 1960, 1962; Hayashi et al., 1984; Fujimoto, 2004; H Toju, unpublished; Y Notsu, unpublished). Although most species are seed feeders, some small-sized species feed on insect galls formed on Fagaceae or Salicaceae plants (Hayashi et al., 1984; Sugiura; Fujimoto, 2004). These gall feeders are currently classified to three genera Archarius, Curculio and Koreoculio (Fujimoto, 2004), and a molecular phylogenetic study suggested that the gall-feeding habit evolved at least twice in the Curculionini (Hughes and Vogler, 2004a).

Sampling and DNA extraction

In total, 36 weevil species representing 5 known and 1 unidentified genera of the tribe Curculionini were collected from Japan and Southeast Asia (Supplementary Table S1). The specimens were preserved in 99% ethanol or acetone (Fukatsu 1999). The specimens were individually subjected to DNA extraction using QIAamp DNA Mini Kit (Qiagen, Venlo, Netherlands). The quality of the DNA samples was confirmed by PCR amplification of a 0.8-kb fragment of mitochondrial cytochrome oxidase subunit I gene of the insect using primers C1-J-2183 and L2-N-3014 (Supplementary Table S2).

DNA cloning, genotyping and sequencing

A 1.5-kb fragment of bacterial 16S rRNA gene was amplified by PCR with primers 16SA1 and 16SB1 (Supplementary Table S2), and the PCR products were subjected to cloning, restriction fragment length polymorphism genotyping, and sequencing as previously described (Fukatsu and Nikoh, 1998). Accession numbers for the 16S rRNA gene clones are listed in Supplementary Table S3. Two mitochondrial genes (0.8-kb fragments of cytochrome oxidase subunit I; 0.4-kb fragments of cytochrome b) and three nuclear genes (0.4-kb fragments of elongation factor 1α, 0.7-kb fragments of 28S rRNA and 0.4-kb fragments of phosphoglycerate mutase) were amplified by PCR with primers listed in Supplementary Table S2, and cloned and sequenced as described above. Accession numbers for the sequences are listed in Supplementary Table S4.

Diagnostic PCR detection of Curculioniphilus symbiont

Specific primers Clp_Unv_456F (5′-GGGTTGTAAAGCACTTTCAGT-3′) and Clp_Unv_560R (5′-AYARRCCGCCTACGYACT-3′) targeting a 0.1-kb fragment of Curculioniphilus 16S rRNA gene were designed on the basis of the symbiont gene sequences determined in this study. Specificity of the primers was also confirmed by homology searches against the public DNA databases. Diagnostic PCR was conducted under a temperature profile of 95 °C for 10 min, 40 cycles of 95 °C for 20 s, 62 °C for 20 s and 72 °C for 30 s, and 72 °C for 2 min. A DNA sample of C. sikkimensis from which Curculioniphilus infection had been confirmed (Toju et al., 2010) was used as positive control.

Molecular phylogenetic analysis

Multiple alignments of the nucleotide sequences were generated using the program MAFFT v6.813b (Katoh and Toh, 2008), followed by elimination of ambiguously aligned nucleotide sites using GBlocks Server v0.91b (Castresana, 2000). Best-fit substitution models for the aligned sequences were selected using the program Kakusan v4 (Tanabe, 2011). Maximum likelihood phylogenies were inferred using the software Treefinder (Jobb et al., 2004) with the tool package Phylogears v1.5 (Tanabe, 2008), whereby parallelized tree search bootstrapping was conducted. Bayesian phylogenies were reconstructed using the program MrBayes v3.1.2 (Ronquist and Huelsenbeck, 2003).

Co-speciation analysis

The levels of topological congruence between the symbiont phylogenies and the host phylogenies were evaluated by a distance based, ParaFit analysis (Legendre et al., 2002) and a likelihood based, approximately unbiased test (Shimodaira, 2002). The distance matrices were calculated based on maximum likelihood trees using the wrapper program of ParaFit, CopyCat v1.00.14 (Meier-Kolthoff et al., 2007). Subsequently, a randomization test of congruence between the host and symbiont matrices were conducted (99 999 permutations). The approximately unbiased test compares the hypothesis of strict host–symbiont co-speciation to an alternative hypothesis allowing host shifts of symbionts. Under the hypothesis of strict co-speciation, a maximum likelihood phylogeny was reconstructed using Kakusan and Phylogears as described above after combining the sequence data of the hosts and the symbionts. The likelihood under this hypothesis was obtained by multiplying the likelihoods of all the nucleotide sites. Meanwhile, under the hypothesis allowing host shifts, a host phylogeny and a symbiont phylogeny were estimated independently; the likelihood under this hypothesis was obtained by multiplying the likelihoods of all the nucleotide sites of the hosts and the symbionts. The likelihoods under the two alternative hypotheses were compared based on the multi-scale bootstrapping of the log-likelihoods of respective nucleotide sites using the program CONSEL v0.1k with 10 sets of 10 000 bootstrap replicates (Shimodaira and Hasegawa, 2001).

Relative rate test

Relative rate tests were performed using the program RRTree (Robinson-Rechavi and Huchon, 2000) on the basis of 1257 unambiguously aligned nucleotide sites of 16S rRNA gene sequences with the maximum likelihood tree of Figure 1 used as a guide tree.

Figure 1
figure 1

Molecular phylogenetic analysis of Curculioniphilus symbionts from Curculionini weevils on the basis of 16S rRNA gene sequences. A maximum likelihood tree inferred from 1257 unambiguously aligned nucleotide sites under the GTR+G model is shown. A Bayesian analysis under the GTR+G model yielded substantially the same result. Bootstrap values (> 50%; 100 replicates) of the maximum likelihood analysis and posterior probabilities (> 0.50) of the Bayesian analysis are shown above and below each node, respectively. Sodalis-allied symbionts identified from the Curculionini weevils and Sitophilus weevils, Nardonella symbionts of Dryophthorinae and Molytinae weevils, and other insect symbionts and free-living gammaproteobacteria are analyzed together. Sequence accession numbers and AT contents of the nucleotide sequences are shown in brackets and parentheses, respectively. For each insect symbiont, name of its host insect is indicated in parentheses. P-symbiont, primary symbiont; S-symbiont, secondary symbiont.

Full size image

Accession numbers

The nucleotide sequences determined in this study were deposited in the DDBJ/EMBL/GenBank nucleotide sequence databases with accession numbers AB507712, AB573438–AB573634, AB604655–AB604677, AB604872–AB604873 and AB746368–AB746431 (for detail, see Supplementary Tables S1, S3 and S4).

Results

Cloning and sequencing of symbiont 16S rRNA genes from Curculionini weevils

In total, 36 species of Curculionini weevils were subjected to PCR, cloning and sequencing of bacterial 16S rRNA gene, of which 27 species yielded 16S rRNA gene sequences highly similar to the sequences of Curculioniphilus symbionts of Curculio sikkimensis, C. camelliae, C. dentipes and C. robustus (Toju et al., 2010), and some of them also yielded 16S rRNA gene sequences of such facultative symbionts as Sodalis, Wolbachia, Rickettsia and Spiroplasma. Meanwhile, the Curculioniphilus sequences were not detected in the remaining nine species, C. cerasorum, C. hachijoensis, C. koreanus, C. lateritius, C. maculanigra, C. morimotoi, C. ochrofasciatus, C. okumai and Archarius esakii (Supplementary Table S1). We repeated PCR of the bacterial 16S rRNA gene for the nine species. No PCR amplification was observed for A. esakii and C. cerasorum, whereas the other seven species yielded the PCR products, which were cloned and sequenced. A variety of bacterial sequences derived from facultative symbionts such as Sodalis, Wolbachia, Rickettsia, Spiroplasma and Lariskella, and also those from putative gut associates/contaminants such as Erwinia, Pseudomonas, Pantoea, Ochrobactorum, Enterobacter, Bradyrhizobium and Sphingomonas were identified, but no Curculioniphilus sequences were obtained (Supplementary Table S3).

Besides the sequences of the primary symbiont Curculioniphilus, the following 16S rRNA gene sequences were identified from the Curculionini weevils: Sodalis from C. hachijoensis and A. roelofsi (Figure 1); Wolbachia from C. hachijoensis, C. hilgendorfi, C. morimotoi, C. okumai, C. sp. 2, A. roelofsi and Koreoculio minutissimus (Supplementary Figure S1A); Rickettsia from C. aino, C. camelliae, C. elaeagni, C. hilgendorfi, C. kojimai, C. lateritius, C. sp. 1, A. pictus and K. minutissimus (Supplementary Figure S1B); Spiroplasma from A. albovittatus (Supplementary Figure S1C); and Lariskella from C. morimotoi and C. okumai (Supplementary Figure S1D). These symbiont sequences neither formed a well-supported monophyletic group nor reflected the host systematics, as previously reported in C. sikkimensis (Toju and Fukatsu, 2011).

Diagnostic PCR of Curculioniphilus symbionts

Diagnostic PCR detection of the Curculioniphilus symbiont was positive for 27 weevil species and negative for remaining 9 species, which was in agreement with the 16S rRNA sequencing results (Supplementary Table S1). As for species with two or more specimens, no intraspecific variation in Curculioniphilus infection was identified (Supplementary Table S1). Most of the weevil species utilizing figs, fagaceous acorns and insect galls were associated with the Curculioniphilus symbionts (4 of 4, 6 of 6 and 7 of 9, respectively), whereas only 1 of 5 weevil species feeding on birches was infected with the symbiont (Supplementary Table S1; Figure 2).

Figure 2
figure 2

Molecular phylogenetic analysis of Curculionini weevils. A maximum likelihood tree inferred from 2686 unambiguously allied nucleotide sites of two mitochondrial (cytochrome oxidase subunit I (COI), 775-bp; cytochrome b, 382-bp) and three nuclear (elongation factor 1 α, 367-bp; 28S rRNA, 737-bp; phosphoglycerate mutase, 425-bp) genes is shown. A Bayesian analysis yielded substantially the same topology. Bootstrap values (> 50%; 100 replicates) of the maximum likelihood analysis and posterior probabilities (> 0.50) of the Bayesian analysis are shown above and below each node, respectively. See Supplementary Table S4 for sequence accession numbers and Supplementary Table S5 for substitution models. The clades A, B and C are indicated on the right side. For each of the weevil species, host plant family and tissue type are shown in parentheses and brackets, respectively. Evolutionary history of their host plants is inferred and mapped on the phylogeny by colored branches. Pink shades indicate association with the Curculioniphilus symbionts.

Full size image

Phylogenetic analysis of Curculioniphilus symbionts

Molecular phylogenetic analyses based on the 16S rRNA gene sequences revealed that the Curculioniphilus symbionts of the 27 Curculionini weevil species formed a clade in the Gammaproteobacteria. The Curculioniphilus clade was allied neither to Nardonella symbionts of diverse weevils nor to Sodalis-allied symbionts of Sitophilus weevils (Figure 1).

Phylogenetic analysis of Curculionini weevils

Figure 2 shows the molecular phylogeny based on sequences of two mitochondrial and three nuclear genes of the 36 Curculionini weevil species, which was largely concordant with previous systematic studies based on morphological characters (Morimoto, 1960, 1962; Hayashi et al., 1984). For example, species feeding on fagaceous acorns with relatively large bodies (6–10 mm in body length) formed a monophyletic group (clade A), so did species characterized by flat dorsal sides of elytra and cylindrical rostra (clade B). In addition, species characterized by small body size (∼4 mm in body length) and insect gall feeding formed a monophyletic group (clade C). As for host plant usage, the following patterns were observed: (i) clade A members exclusively feed on fagaceous acorns; (ii) clade C members all live on insect galls; (iii) clade B members utilize various plant groups (birches, camellias, silverberries and others) as well as various plant parts (seeds, strobiles, flower buds and insect galls); and (iv) species feeding on figs constitute basal lineages.

Co-evolutionary analysis between Curculioniphilus symbionts and their weevil hosts

Figure 3 contrasts the phylogeny of the 27 Curculionini weevil species with the phylogeny of their Curculioniphilus symbionts. A distance-based co-speciation test showed that phylogenetic distances between operational taxonomic units were significantly congruent between the host phylogeny and the symbiont phylogeny (ParaFit method; P=0.00001). The analysis also identified significant co-speciating patterns in 20 of the 27 operational taxonomic units examined (Figure 3). Furthermore, the strict co-speciation hypothesis was not rejected by a likelihood-based test (P=0.093).

Figure 3
figure 3

Co-speciation analysis between Curculionini weevil hosts and their Curculioniphilus symbionts. (a) A maximum likelihood phylogeny of 27 Curculionini weevil species inferred from 2689 unambiguously aligned nucleotide sites (cytochrome oxidase subunit I (COI), 775-bp; cytochrome b (Cytb), 382-bp; elongation factor 1α (EF1α), 367-bp; 28S rRNA, 740-bp; phosphoglycerate mutase (Pglym), 425-bp). See Supplementary Table S5 for substitution models. (b) A maximum likelihood phylogeny of their Curculioniphilus symbionts inferred from 1458 unambiguously aligned nucleotide sites of 16S rRNA gene under the TVM+G model. In both trees, bootstrap values (> 50%; 100 replicates) of the maximum likelihood analysis and posterior probabilities (> 0.50) of Bayesian analysis are shown above and below each node, respectively. Host–symbiont associations are highlighted by lines between the trees, whose thickness indicates the significance of the association inferred by the ParaFit program (Legendre et al., 2002).

Full size image

Phylogenetic analysis of Sodalis-allied symbionts of weevils

Figure 4 shows the phylogenetic relationship between Sodalis-allied symbionts from Sitophilus weevils (nine symbiont sequences representing three host species), some Curculionini weevils (eight symbiont sequences representing three host species), and other insects on the basis of their 16S rRNA gene sequences. The symbionts of Sitophilus weevils did not form a well-supported monophyletic clade, and the symbiont phylogeny did not reflect the host systematics: the symbiont sequences from the same host species did not cluster but constituted distinct lineages in the phylogeny. Meanwhile, the symbionts of Curculionini weevils formed basal clusters in the phylogeny, and statistical supports for the groupings were generally low.

Figure 4
figure 4

Molecular phylogenetic analysis of Sodalis-allied symbionts from Curculionini weevils, Sitophilus weevils, and various other insects on the basis of 16S rRNA gene sequences. A maximum likelihood tree inferred from 1278 unambiguously allied nucleotide sites under the J1+G model is shown. A Bayesian analysis under the GTR+G model yielded substantially the same topology. Bootstrap values (> 50%; 100 replicates) of the maximum likelihood analysis and posterior probabilities (> 0.50) of the Bayesian analysis are shown above and below each node, respectively. Sequence accession numbers and AT contents of the nucleotide sequences are shown in brackets and parentheses, respectively. For each insect symbiont, name of its host insect is indicated in parentheses. P-symbiont, primary symbiont; S-symbiont, secondary symbiont.

Full size image

Co-evolutionary analysis between Sodalis-allied symbionts and their Sitophilus weevil hosts

Figure 5 contrasts the phylogeny of Sitophilus weevils with the phylogeny of their Sodalis-allied symbionts. No host–symbiont co-speciating pattern was recognizable in the weevil genus Sitophilus. A distance-based co-speciation test showed that phylogenetic distances between operational taxonomic units were incongruent between the host phylogeny and the symbiont phylogeny (ParaFit method; P=0.19).

Figure 5
figure 5

Phylogenetic comparison between Sitophilus weevil hosts and their Sodalis-allied symbionts. (a) A maximum likelihood phylogeny of six cytochrome oxidase subunit I (COI) gene sequences representing three Sitophilus species based on 467 unambiguously aligned nucleotide sites (J1+G, F81+G and TN93+G models for the first, second and third codon positions, respectively). (b) A maximum likelihood phylogeny of nine 16S rRNA gene sequences from the three Sitophilus species based on 1281 unambiguously aligned nucleotide sites under the J1+G model. In both trees, bootstrap values (> 50%; 100 replicates) of the maximum likelihood analysis and posterior probabilities (> 0.50) of Bayesian analysis are shown above and below each node, respectively. Host–symbiont relationships are indicated by dotted lines.

Full size image

Molecular evolutionary aspects of Curculioniphilus symbionts and Sodalis-allied symbionts of weevils

The 16S rRNA gene sequences of the Curculioniphilus primary symbionts exhibited AT contents ranging from 48.1% to 50.8%, which were remarkably higher than the AT contents of free-living gammaproteobacteria around 45% (Figure 1). Meanwhile, the 16S rRNA gene sequences of the Sodalis-allied primary symbionts of Sitophilus weevils were around 45% in their AT contents, so were the sequences of the Sodalis-allied secondary symbionts of Curculionini weevils (Figure 4). The molecular evolutionary rate of the Curculioniphilus primary symbionts was significantly higher than the evolutionary rates of the Sodalis-allied primary symbionts of Sitiophilus weevils (P=0.0047), the Sodalis-allied secondary symbionts of Curculionini weevils (P<0.0001), the tsetse secondary symbiont Sodalis glossinidius (P<0.0001) and Escherichia coli (P=0.0293; Table 1). The molecular evolutionary rate of the Sodalis-allied primary symbionts of Sitophilus weevils was significantly higher than the evolutionary rates of the Sodalis-allied secondary symbionts of Curculionini weevils (P<0.0001) and the tsetse secondary symbiont S. glossinidius (P=0.0255; Table 1).

Table 1 Relative-rate test for comparing the molecular evolutionary rates of 16S rRNA gene sequences between Curculioniphilus symbionts of Curculionini weevils, Sodalis-allied symbionts of Sitophilus grain weevils and their free-living relatives
Full size table

Discussion

Curculioniphilus as the primary symbiont co-speciating with Curculionini weevils

We examined 36 weevil species of the tribe Curculionini, and identified 27 weevil species associated with Curculioniphilus (Supplementary Table S1). The phylogenetic relationship of the Curculioniphilus symbionts was congruent with the phylogenetic relationship of their host weevils (Figure 3). The host–symbiont co-speciation over evolutionary time must have been underpinned by stable vertical transmission of the bacteriocyte-associated symbiont through the host weevil generations via ovarial passage (Toju et al., 2010). The AT-biased nucleotide compositions and the accelerated evolutionary rates of 16S rRNA gene of the Curculioniphilus symbionts are also suggestive of long-lasting host–symbiont co-evolution (Figure 1; Table 1). Based on these results, we conclude that Curculioniphilus comprises the primary symbiont clade associated with the Curculionini weevils. As Nardonella is the ancestral primary symbiont of weevils (Lefèvre et al., 2004; Conord et al., 2008), it is conjectured that Curculioniphilus was acquired by the common ancestor of the Curculionini weevils and took over the original symbiont.

Occasional losses of Curculioniphilus in the evolutionary course of Curculionini weevils

Meanwhile, of 36 Curculionini weevil species we examined, 9 species are not associated with Curculioniphilus (Supplementary Table S1). The global host–symbiont co-speciation (Figure 3) and the local distribution of the Curculioniphilus-free species (Figure 2) indicate that Curculioniphilus has been occasionally lost in several host lineages independently, which highlights the dynamic evolutionary history of the endosymbiosis among the Curculionini weevils. For these Curculioniphilus-free species, whether the symbionts were simply lost or replaced by unidentified microbial associates deserves further analyses.

Evolutionary dynamics of Curculioniphilus among Curculionini weevils: possible relevance to biological roles of the symbiont and ecological aspects of the host

The evolutionary stability and the co-speciating pattern in the host–symbiont association (Figure 3) favor the hypothesis that Curculioniphilus has some biological roles for the host weevils, such as provisioning of nutritional components deficient in their food plants and/or detoxification of plant defense chemicals (Toju et al., 2010; Toju and Fukatsu, 2011). Meanwhile, considering the pattern of repeated evolutionary losses of Curculioniphilus (Figure 2), the necessity of the symbiont for the host weevils might be not stringent but could be mitigated/circumvented under certain ecological/environmental conditions. A candidate ecological factor relevant to the symbiont losses may be host plant switches in the Curculionini (see Figure 2), but the causal relationship is currently elusive and deserves more detailed studies. Notably, a recent study reported that, in the weevil Euscepes postfasciatus, even the ancient primary symbiont Nardonella is beneficial but not essential for the host: symbiont-free weevils were able to grow and reproduce, although they suffered retarded growth, smaller body size and reduced fecundity in comparison with normal symbiotic weevils (Kuriwada et al., 2010). Based on these evolutionary patterns and circumstances, we suggest that Curculioniphilus is certainly the bacteriocyte-associated primary symbiont of beneficial nature in the majority of Curculionini weevils, but may be not necessarily essential for the host insects under particular ecological conditions, which might have prompted the occasional losses of the symbiont in some host lineages.

Secondary symbionts of Curculionini weevils

A previous study reported that, in addition to the primary symbiont Curculioniphilus, several secondary symbionts including Wolbachia, Rickettsia, Spiroplasma and Sodalis are prevalent in natural populations of C. sikkimensis (Toju and Fukatsu, 2011). In this study, we demonstrated that diverse Curculionini weevils are also associated with these facultative secondary symbionts (Figure 4; Supplementary Figure S1A–C). In addition, from C. morimotoi and C. okumai, we detected Lariskella symbionts (Supplementary Figure S1D), which belong to a recently described facultative symbiont clade associated with stinkbugs, fleas and ticks (Matsuura et al., 2012). The finding that Sodalis-allied facultative symbionts are present in natural Curculionini populations is evolutionarily interesting, shedding light on the origin of the Sodalis-allied primary symbionts of Sitophilus weevils, as discussed later.

Sodalis-allied primary symbiont of Sitophilus weevils: lack of host–symbiont co-speciation

The weevil genus Sitophilus, consisting of 14 described species, is known for the notorious pest species S. oryzae, S. zeamais and S. granarius that infest stored crop products such as rice, wheat and maize (Plarre, 2010). The Sodalis-allied primary symbionts of the Sitophilus weevils, often referred to by the acronym SOPE or SPE (after Sitophilus (oryzae) primary endosymbiont), are localized in the midgut-associated bacteriome and also in the female ovaries, vertically transmitted through host generations via ovarial passage, and beneficial for host growth and reproduction (Heddi and Nardon, 2005). As most of the other Dryophthorinae weevils harbor the ancient primary symbiont Nardonella (Lefèvre et al., 2004), it has been suggested that the Sodalis-allied symbiont was acquired by an ancestor of the Sitophilus weevils, replaced the original symbiont Nardonella, and established as a new primary symbiont (Lefèvre et al., 2004; Heddi and Nardon, 2005). The symbiont replacement process in the Sitophilus weevils looks quite similar to that in the Curculionini weevils, but subsequent host–symbiont co-evolutionary trajectories are strikingly different between them: unlike Curculioniphilus, the Sodalis-allied primary symbionts exhibited no host–symbiont co-speciation (Figure 5), no bias in nucleotide compositions (Figure 1), and less accelerated evolutionary rates (Table 1) in the Sitophilus weevils. On account of the small number of Sitophilus weevil species examined and the relatively low statistical supports for the phylogeny of Sodalis-allied symbionts, the host–symbiont phylogenetic incongruence may be not robust statistically, but it is evidently different from the remarkable host–symbiont co-speciating pattern in the Curculionini weevils.

Evolutionary dynamics of Sodalis-allied symbiont in Sitophilus weevils: possible evolutionary processes underlying the promiscuous host–symbiont relationship

These evolutionary patterns suggest that the Sodalis-allied symbiont of Sitophilus weevils is, although bacteriocyte-associated and beneficial (Heddi and Nardon, 2005), similar to facultative secondary symbionts like Wolbachia in diverse insects and Serratia, Hamiltonella and Regiella in aphids, at least in some aspects (Werren et al., 2008; Oliver et al., 2010). This finding is striking in that a number of bacteriocyte-associated primary symbionts of beneficial nature, including Buchnera, Wigglesworthia and others, are generally co-speciating with their host insects over evolutionary time (Baumann, 2005; Moran et al., 2008). In theory, the promiscuous host–symbiont relationship in the Sitophilus weevils can be generated by either of the following processes: (i) horizontal transfers of the symbiont across different host lineages within the genus Sitophilus; (ii) horizontal acquisitions of the symbiont from outside, possibly from different insect sources; or (iii) repeated acquisitions from free-living symbiont pool in the environment, as reported in some stinkbugs (Kikuchi et al., 2007, 2011). Hereafter, we discuss these possibilities under the light of phylogenetic, evolutionary and other lines of evidence available to date.

Given that the scenario (i) applies, symbiont swapping within and between the host lineages would result in occurrences of genetically identical or close symbiont genotypes within and between the host insect species, which would generate a compact symbiont phylogeny with short terminal branches, as commonly observed with Wolbachia, Rickettsia, Serratia, Hamiltonella, Regiella and other facultative insect symbionts (Russell et al., 2003; Baldo et al., 2006; Weinert et al., 2009). Oddly, however, such phylogenetic patterns are not observed with the Sodalis-allied Sitophilus symbionts: the symbiont genotypes are remarkably divergent within and between the host weevil species, with long terminal branches giving a comb- or star-like appearance of the phylogeny (Figures 4 and 5). Hence, we suppose that the scenario (i) does not fit well to the observed data.

If the scenario (ii) applies, the Sodalis-allied primary symbionts of Sitophilus weevils must have been occasionally acquired from somewhere via repeated horizontal transfers, and the candidate sources are the Sodalis-allied facultative symbionts of other insects. For a long time, Sodalis glossinidius and allied insect symbionts have been known only from Glossina tsetse flies and Sitophilus weevils (Dale and Maudlin, 1999; Heddi et al., 1999). However, recent studies unveiled universal occurrences of Sodalis-allied symbionts in diverse insects such as bird lice (Fukatsu et al., 2007), louse flies (Nováková and Hypša, 2007; Chrudimský et al., 2012), weevils (Toju and Fukatsu, 2011), longicorn beetles (Grünwald et al., 2010) and stinkbugs (Kaiwa et al., 2010, 2011). In this study, notably, we demonstrated that Sodalis-allied facultative symbionts are commonly present in natural populations of Curculionini weevils, which constitute basal lineages in the phylogeny of Sodalis-allied symbionts of various insects including Sitophilus weevils (Figure 4). Therefore, it seems plausible, although speculative, that the weevil-associated Sodalis-allied facultative bacteria might have served as the evolutionary sources for the Sodalis-allied primary symbionts of the Sitophilus weevils. To confirm this hypothesis, future surveys should focus on whether Sitophilus and other Dryophthorinae weevils are associated with Sodalis-allied symbionts of facultative type.

In the context of the scenario (iii), the recent finding of a Sodalis-allied cultivable bacterium from human wound (Clayton et al., 2012) is of interest. Meaningfully, the human-derived Sodalis-allied bacterium was placed in the basal position of the phylogeny as Sodalis-allied symbionts of Curculionini weevils were (Figure 4). It should be noted that the tsetse-associated symbiont S. glossinidius can be cultured in cell-free media (Dale and Maudlin, 1999), suggesting the possibility of free-living lifestyle of some Sodalis-allied bacterial strains, which may potentially serve as environmental pool for symbiont acquisition.

Repeated acquisitions and replacements of Sodalis-allied symbionts in Sitophilus weevils: putative factors relevant to the symbiont evolutionary dynamics

In summary, the host–symbiont phylogenetic incongruence in the Sitophilus weevils may be best accounted for by repeated horizontal acquisitions and replacements of their Sodalis-allied primary symbionts, whose origins are weevil- or other insect-associated Sodalis-allied facultative symbionts, or free-living Sodalis-allied bacteria in the environment. The host–symbiont promiscuity might have been facilitated by the ‘beneficial but not essential’ nature of the symbiotic association. Previous studies reported that Sitophilus weevil strains experimentally deprived of the Sodalis-allied symbiont could be established and continuously maintained (Nardon, 1973), but these strains exhibited paler body color, soft cuticle, slower growth, reduced fecundity, and lower flight activity in comparison with normal symbiotic strains (King and Sang, 1959; Grenier et al., 1986, 1994; Nardon and Nardon, 1998). Also host plant difference might have affected the process of symbiont replacements. Lefèvre et al. (2004) reported that, distinct from S. oryzae, S. zeamais and S. granarius that live on monocot grain seeds, S. linealis feeding on dicot tamarind seeds was aposymbiotic, and argued the possibility of symbiont loss associated with the host plant shift.

Conclusion and perspective

In this study, we highlighted two remarkable evolutionary events in weevils, namely independent replacements of the ancient bacteriocyte-associated primary symbiont Nardonella by different microorganisms, Curculioniphilus symbiont in the tribe Curculionini and Sodalis-allied symbiont in the genus Sitophilus. Despite replacing the same original symbiont, these two lineages of newcomer symbionts exhibited strikingly different evolutionary patterns: host–symbiont co-speciation in the Curculiophilus–Curculionini association vs host–symbiont promiscuity in the Sodalis–Sitophilus association. In both weevil lineages, further symbiont losses have occurred occasionally, which might be relevant to food plant switching of the host weevils. The evolutionary origins of the Sodalis-allied primary symbiont in Sitophilus weevils might be, although speculative, either Sodalis-allied facultative symbionts associated with diverse weevils or free-living Sodalis-allied bacteria in the environment, whereas the evolutionary origin of the Curculioniphilus symbiont in Curculionini weevils is still elusive.

In describing the diversity of various host–symbiont relationships, we tend to use dichotomous categorizations for simplicity, such as primary symbiont vs secondary symbiont, obligate symbiont vs facultative symbiont, beneficial symbiont vs parasitic symbiont, co-speciating symbiont vs promiscuous symbiont, and so on. However, the real world is usually better described by continuity spanning the extremes, which is impressively illustrated by the endosymbiont diversity among the weevils.