Who Is in There? Exploration of Endophytic Bacteria within the Siphonous Green Seaweed Bryopsis (Bryopsidales, Chlorophyta)

Associations between marine seaweeds and bacteria are widespread, with endobiotic bacterial-algal interactions being described for over 40 years. Also within the siphonous marine green alga Bryopsis, intracellular bacteria have been visualized by electron microscopy in the early ‘70s, but were up to now never molecularly analyzed. To study this partnership, we examined the presence and phylogenetic diversity of microbial communities within the cytoplasm of two Bryopsis species by combining fluorescence in situ hybridization (FISH), denaturing gradient gel electrophoresis (DGGE) and 16S rRNA gene clone libraries. Sequencing results revealed the presence of Arcobacter, Bacteroidetes, Flavobacteriaceae, Mycoplasma, Labrenzia, Phyllobacteriaceae and Xanthomonadaceae species. Although the total diversity of the endobiotic communities was unique to each Bryopsis culture, Bacteroidetes, Mycoplasma, Phyllobacteriaceae, and in particular Flavobacteriaceae bacteria, were detected in several Bryopsis samples collected hundreds of kilometres apart. This suggests that Bryopsis closely associates with well-defined endophytic bacterial communities of which some members possibly maintain an endosymbiotic relationship with the algal host.


Introduction
Marine macroalgal-bacterial associations range from beneficial, harmful or neutral, over obligate or facultative, to ecto-or endophytic interactions [1]. Elaborating the latter, endobiotic associations between marine macroalgal hosts and bacteria have been reported over the past 40 years. Besides reports of bacterial endosymbionts associated with red algal galls [2][3][4], endophytic bacteria have been microscopically observed in the vacuolar as well as cytoplasmatic regions of various bryopsidalean green algae, including Bryopsis, Penicillus, Halimeda, Udotea and Caulerpa [5][6][7][8][9][10]. These seaweeds are composed of a single, giant tubular cell and form an interesting biotic environment for bacterial communities. The giant cell contains millions of nuclei and chloroplasts in a thin cytoplasmic layer surrounding a large central vacuole. The cytoplasm typically exhibits vigorous streaming, enabling transport of nutrients, organelles and various biomolecules across the plant [11]. In Bryopsis 'bacteria-like particles' have been visualized in the cytoplasm by means of transmission electron microscopy in vegetative thalli as well as in the gametes, the latter suggesting vertical transmission of the endophytic bacteria [5]. This implies a stable and specific relationship between the algal host and its endobionts in which both partners may provide mutualistic ecological benefits. To date, the diversity of the intracellular microbial communities associated with Bryopsis remains unidentified. Up till now investigations of the bacterial endophytic diversity of siphonous macroalgae have been limited to Caulerpa species and revealed endosymbiotic Alphaproteobacteria with the potential to photosynthesize, detoxify and/or fix nitrogen [9,10]. The endophytic bacteria in Bryopsis may similarly possess ecologically significant functions and bioactive potential since Bryopsis is a substantial source of bioactive compounds such as therapeutic kahalalides which may be of bacterial origin [12,13].
In order to explore these algal-endophytic bacterial interactions, we previously developed a surface sterilization protocol for the complete elimination of bacterial epiphytes from the Bryopsis surface [14]. We showed that Bryopsis samples treated with a combined chemical and enzymatic approach (i.e. a mixture of cetyltrimethylammonium bromide (CTAB) lysis buffer, proteinase K and the bactericidal cleanser Umonium Master) remained intact after sterilization and showed no remaining bacterial fluorescence on their surface when stained with a DNA fluorochrome. Successful 16S rRNA gene DGGE analysis following this surface sterilization treatment showed that endophytic DNA was still present within the sterilized Bryopsis samples, allowing specific molecular processing of the endophytes [14].
In this study, we verified the presence of bacteria inside two Bryopsis species from the Mexican west coast by a combination of fluorescence in situ hybridization (FISH), denaturing gradient gel electrophoresis (DGGE) and clone libraries.

Ethics Statement
No specific permits were required for the described field studies, i.e. the collection of algal samples from the Mexican west coast, because marine algae are not included in the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES, http://www.cites.org/eng/disc/species.shtml). The authors confirm that the location is not privately-owned or protected in any way and that the field studies did not involve endangered or protected species.

Algal material
Five Bryopsis specimens were collected in February 2009 along the Pacific Mexican coast at different sites located between Mazunte Beach (Oaxaca, southwest Mexico) and Playa Careyero (Nayarit, central Mexico) (Figure 1). These five samples were classified in two different species with samples MX19 and MX263 representing Bryopsis hypnoides J.V. Lamouroux and MX90, MX164, and MX344 representing Bryopsis pennata J.V. Lamouroux var. leprieurii (Kützing) Collins and Hervey individuals. After sampling, living specimens were rinsed with sterile seawater and transferred to the laboratory in plastic vessels containing a small amount of sterile seawater. In the laboratory, clean apical fragments of the Bryopsis specimens were isolated and cultured in sterile 1 x modified Provasoli enriched seawater [15] at 23uC under 12h:12h (Light:Dark) conditions with a photon flux rate of 25-30 mE m 22 s 21 . This isolation procedure was repeated for several months until the Bryopsis cultures were free of eukaryotic contamination. Thus, the Bryopsis isolates were kept in culture for eight months prior to molecular analyses in October 2009. After isolation, all five unialgal Bryopsis cultures were maintained in the laboratory under the culture conditions described above.

Fluorescence in situ hybridization
Unialgal Bryopsis thalli were fixed in 4% paraformaldehyde and 0.25% glutaraldehyde in 50 mM PIPES (piperazine-N,N9-bis(2ethanesulfonic acid)) buffer, pH 7.2 for 2 hours. After dehydration through a graded ethanol series from 30% to 80%, ethanol was subsequently replaced by LR white embedding medium (London Resin, UK). Samples were loaded in gelatine capsules and allowed to polymerize at 37uC for 3 days. Semithin sections were cut using glass knives on a Microm HM360 microtome (Microm International GmbH, Germany) and collected on Vectabond-coated (Vector Laboratories, USA) slides. In situ hybridization was performed as described by Daims et al. [16] with 200 mL formamide per mL hybridization buffer, an incubation of 90 min at 46uC, and the universal bacterial Cy3-labelled EUB338 probe mix [17]. Algal DNA and cell wall counterstaining was performed by adding a mix of 4,6-diamidino-2-phenylindole (DAPI) and calcofluor to the sections for 7 min in the dark at room temperature. Sections were mounted in AF-1 antifadent (Citifluor, UK) and viewed with an Olympus BX51 epifluorescence microscope fitted with a DAPI/FITC/TRITC triple band filter. The Bryopsis specimens were not surface-sterilized prior to hybridization due to potential morphological losses.

Surface sterilization, DNA extraction and PCR
To identify the endophytic bacterial diversity, approximately 2 grams (ww) of each unialgal Bryopsis sample was surface-sterilized as previously described [14] prior to a total DNA extraction using a CTAB protocol modified from Doyle and Doyle [18]. These extracts, containing both algal and bacterial DNA, were subjected to rbcL and 16S rRNA gene PCR amplifications following protocols outlined in Hanyuda et al. [19] and Lane [20] with, respectively, primer pairs 7F/R1391 and 27F/1492R. All obtained PCR amplicons were purified using a Nucleofast 96 PCR clean up membrane system (Machery-Nagel, Germany) according to the manufacturer's instructions.

Cloning and DGGE
To determine the bacterial diversity, purified 16S rRNA gene amplicons from the algal extracts were cloned using the pGEMH-T Vector System (Promega Benelux, The Netherlands). For each Bryopsis sample a clone library of 150 clones was prepared, the diversity of which was examined via short fragment sequencing (see below). For dereplication, clones' short sequences were grouped into the same operational taxonomic unit (OTU) when having $97% similarity. From each OTU, representative clones were selected for full length (61450 bp) 16S rRNA gene sequencing (see below). Clone libraries' coverage was verified by DGGE analysis of each Bryopsis DNA extract and its representative clones. A V3 PCR with primers F357-GC/R518 and subsequent DGGE analysis were carried out as described previously [14], with a denaturing gradient of 45-65%. DGGE banding patterns were normalized using the BioNumerics 5.1 software (Applied Maths, Belgium). DGGE bands from the algal extracts which showed no correspondence with OTU band positions were excised from the polyacrylamide gel following Van Hoorde et al. [21] and sequenced (6150 bp) as described below.

Sequencing
RbcL genes, DGGE bands as well as short and full length 16S rRNA genes were sequenced on an ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems, USA) by means of the BigDyeH xTerminator TM v.3.1 Cycle Sequencing and Purification Kit (Applied Biosystems, USA) according the protocol of the supplier. Primers used were, respectively, 7F/R1391 [19], F357/R518 [21], BKL1 [22] and T7/SP6 (Promega Benelux, The Netherlands). Sequences obtained were assembled in BioNumerics, compared with nucleotide databases via BLAST and chimera-checked using Bellerophon [23]. The bacterial 16S rRNA gene and Bryopsis chloroplast 16S rRNA gene and rbcL sequences were submitted to GenBank under accession numbers JF521593-JF521615 (see also Table 1).

Phylogenetic analyses
Two sets of alignments, made using MUSCLE [24], were considered for phylogenetic analyses. The first one, consisting of a concatenated chloroplast 16S rRNA gene and rbcL dataset, was used for the creation of a Bryopsis phylogram. A second set of alignments was assembled to assess 16S rRNA gene phylogenetic relationships between the Bryopsis-associated bacterial endophytes and known bacterial species, including BLAST hits and algaeassociated bacteria described in the literature. The most suitable model for phylogenetic analysis was selected using the AIC criterion in jModelTest [25]. Subsequently, the Bryopsis host and bacterial datasets were analyzed by means of the maximum likelihood (ML) algorithm in PhyML v3.0 [26] under a HKY + G4 model via the University of Oslo Bioportal website (http://www. bioportal.uio.no//). Reliability of ML trees was evaluated based on 100 bootstrap replicates. Output ML trees were subsequently visualized in Mega 4.0 [27] and edited with AdobeH IllustratorH CS5.

Fluorescence in situ hybridization
To confirm the observation of endogenous bacteria in Bryopsis made by Burr and West [5], Bryopsis sections were hybridized with the universal bacterial EUB338 probe mix labelled with Cy3. Figures 2A-C depict clear binding of the red fluorescent probe mix to bacterial rRNA present throughout the cytoplasm; both in the outer layer next to the cell wall, which contains most of the organelles except the chloroplasts (Figures 2A-C), as well as in the inner chloroplast layer immediately adjacent to the vacuole ( Figures 2B-C). These hybridization results demonstrate the presence of metabolically active bacteria within the Bryopsis cytoplasm. Since the Bryopsis thalli were not surface sterilized before fixation, the EUB338 probe mix also hybridized with epiphytic bacterial rRNA on the cell wall ( Figures 2B-C).

Bacterial diversity within Bryopsis algae: DGGE
Coverage of the clone libraries was verified by comparing DGGE community profiles of the different Bryopsis DNA extracts with the banding pattern of clones from their respective OTUs, including representative clones with 16S rRNA gene chloroplast and chimeric sequences. As shown in Figure 3 the OTUs DGGE bands overlap well with the individual bands of the MX extracts' DGGE profiles, indicating adequate clone library coverage. MX samples 19, 164 and 344, however, all showed one band in their DGGE profile not represented by an OTU band. Consequently, these three DGGE bands (A, B and C, respectively) were excised and sequenced. The sequence of DGGE band A showed 100% similarity with the chimeric sequences detected in MX sample 263, not unexpected given its corresponding band position with clone MX263.66. DGGE band B was identified as forming part of the OTU-2 cluster with 100% sequence similarity with clone MX19.9, whereas DGGE band C showed no correspondence with any bacterial OTU detected. Hence, the latter DGGE band was assigned to a new OTU, i.e. OTU-7. BLAST searches revealed that this OTU-7 is closely related to Labrenzia species isolated from the green seaweed Ulva rigida (FN811315), crustose coralline red algae (HM178529) and the dinoflagellate Karlodinium micrum (HM584720).   Figure 4 we can deduce that Flavobacteriaceae (OTU-1), Mycoplasma (OTU-2), Bacteroidetes (OTU-3) and Phyllobacteriaceae (OTU-4) species were present in more than one Bryopsis sample examined. Even though the endobiotic community members were to a certain extent similar, the total diversity of the endophytic community was unique to each Bryopsis sample. None of the Bryopsis samples harbored the same number or range of bacterial endophytes.

Bacterial diversity within Bryopsis algae: Phylogenetic analysis
A wide-range phylogenetic tree ( Figure 5, tree without compressed branches see Figure S1) was created, which includes bacterial OTUs determined in this study (clones and DGGE bands), significant BLAST hits (Table 1), type strains from the Bacteroidetes, Proteobacteria and Tenericutes division, and algaeassociated bacteria described in the literature (Table S1). As could be predicted from the BLAST maximum identity scores ( Table 1), none of the endobiotic bacterial sequences clustered tightly with cultivated bacterial type strains. Consequently, all endophytic bacterial OTUs derived from Bryopsis represent new species or genera which in some cases match previously sequenced unclassified bacteria. These OTU sequences, however, all showed at least 93% sequence similarity with their best BLAST hit which generally resulted in phylogenetic placements with good bootstrap support. Accordingly, all OTU-1 sequences formed a distinct and well-supported (98%) clade within the Flavobacteriaceae family and most likely represent a new genus given their low sequence similarities (87% at most) with Flavobacteriaceae type strains. The similarity among the five OTU-1 sequences, however, was 99.7%, suggesting all sequences belong to the same new Flavobacteriaceae genus even though they were derived from different Bryopsis samples collected several hundred kilometres apart. Likewise, the Bacteroidetes OTU-3 clones were virtually identical displaying 99.9% pairwise similarity. These OTU-3 clones, found in B. hypnoides samples MX19 and MX263, belong to a single clade (100% bootstrap support) of unclassified Bacteroidetes, but are distantly related to other unclassified Bacteroidetes symbionts. The OTU-2 clade, consisting of clones MX19.9 and MX263.1 and DGGE band B, fell into the genus Mycoplasma with 100% bootstrap support although these clones showed low levels of similarity (#90%) with Mycoplasma type strains. All OTU-2 sequences presumably belong to one and the same new Mycoplasma species (99.7 intra-OTU sequence similarity). The majority of the endophytic bacterial OTUs, however, were affiliated with the Proteobacteria phylum and belonged to the Alpha-, Gamma-and Epsilonproteobacteria. Particularly, OTU-5 and 6, both consisting of clones exclusively obtained from B. pennata var. leprieurii sample MX164, most probably represent a new genus of Xanthomonadaceae and a new Arcobacter species, respectively. OTU-4 and 7 are robustly affiliated (100% bootstrap support) with the Alphaproteobacteria class and belong to the Rhizobiales and Rhodobacterales, respectively. Despite the high sequence similarity of OTU-7 with algal-associated Labrenzia species, relatedness of DGGE band C with the Labrenzia alexandrii type strain (AJ582083) and an uncultured Labrenzia bacterium isolated from Caulerpa taxifolia (AF259594) lacks bootstrap support. The shortness of the DGGE band C sequence (6150 bp) and, consequently, the poor resolution within this clade, made it difficult to conclude whether OTU-7 represents a new Labrenzia species. Finally, OTU-4 is the only OTU containing clones derived from different Bryopsis samples in which the representative clones, i.e. clone MX19.12 and MX164.59, did not cluster together. This is in agreement with the 97% intra-OTU sequence similarity. Hence, both clones belong to the Phyllobacteriaceae clade with good bootstrap support (80%), but most likely represent two different new species or genera because of their low sequence similarities (96% at most) with Phyllobacteriaceae type strains.

Discussion
Forty years after Burr and West [5] observed endogenous 'bacteria-like particles' in Bryopsis hypnoides, this is the first study to verify the presence of metabolically active endophytic bacteria inside the Bryopsis cytoplasm by means of the FISH technique. Mainly due to the intense background autofluorescence of algal cells, reports of successful FISH applications on macroalgae are limited to analyses of macroalgal surface-associated bacteria [28] and algal gall endosymbionts [4]. The use in this study of semithin algal sections and a triple band filter, however, made it possible to discriminate bacterial FISH signals from autofluorescence of algal pigments using standard FISH protocols in combination with epifluorescence microscopy. Even though Bryopsis samples were not surface-sterilized prior to hybridization to avoid potential morphological losses, the solid embedding at the start of the FISH protocol proved successful in immobilizing the epiphytes on the Bryopsis surface (data not shown). This prevented the detachment and potential spread of surface bacteria during sectioning. Consequently, our FISH results strongly suggest the presence of bacteria within Bryopsis cells.
In this study, the first insights are provided into the identity and phylogenetic diversity of endobiotic bacterial communities within Bryopsis. Despite the limited number of samples studied, our results indicate that Bryopsis harbors endophytic bacterial communities which are not very complex (i.e. only 7 bacterial OTUs detected), but taxonomically diverse including Arcobacter, Bacteroidetes, Flavobacteriaceae, Mycoplasma, Labrenzia, Phyllobacteriaceae and Xanthomonadaceae species. Although the composition of the total endophytic community seems unique to each Bryopsis culture, Bacteroidetes, Flavobacteriaceae, Mycoplasma and Phyllobacteriaceae species were detected in two or more Bryopsis samples. In particular OTU-1 Flavobacteriaceae species are present in all five Bryopsis cultures, which were collected from diverse sites along the Mexican west coast. Delbridge and colleagues [10] made similar observations when comparing the endosymbiotic communities within four different Caulerpa species. While the endosymbiotic communities seemed unique to each Caulerpa individual, all community members were photosynthetic Alphaproteobacteria.
Also within Bryopsis, Alphaproteobacteria appear well represented. This is not unexpected, since Alphaproteobacteria are frequently associated with macroalgae [1,29,30], an alliance which may be linked to dimethylsulfoniopropionate (DMSP) exchange [31]. Particularly OTU-7, belonging to the marine phototrophic and CO-oxidizing Labrenzia genus [32,33], is closely related to an uncultured bacterium reported by Meusnier et al. [29] in their study on the total bacterial community associated with Caulerpa taxifolia. Although Labrenzia species have not been reported as endophytes, the presence of Rhizobiales-specific proteins in L. aggregata [34] may hint at potential endosymbiotic features. The Rhizobiales order contains various well-known nitrogen fixing plant symbionts, mainly in terrestrial habitats but also in marine environments [35]. Moreover, Rhizobiales bacteria are common epiphytes on green [31,36], brown [37,38] and red [36] macroalgae; and a Rhodopseudomonas species with the potential to fix nitrogen was isolated from the inside of C. taxifolia [9]. Also within Bryopsis, Rhizobiales species seem to be well established as clones MX19.12 and MX164.59 (OTU-4) likely represent two different new Phyllobacteriaceae species or genera clustering together with, respectively, a free-living marine Phyllobacteriaceae bacterium [39] and a dinoflagellate-associated anoxygenic photosynthetic bacterial strain [40]. In addition, we amplified a Phyllobacteriaceae nitrogenase-like light-independent protochlorophyllide reductase gene (submitted to GenBank under accession number JN048464) from Bryopsis sample MX164 by the nifH protocol described by De Meyer et al. [41], supporting the above suggested relatedness of OTU-4 to photosynthetic bacteria.
Besides the presence of Alphaproteobacteria in three of the five Bryopsis cultures studied, endophytes from the Gamma-and Epsilonproteobacteria order seem restricted to a single Bryopsis sample. The latter endophytes (OTU-6) most likely belong to a new Arcobacter species within the Campylobacteraceae family. Arcobacter species are mainly known as potential human and animal pathogens, but have also been isolated from diverse marine environments including seawater surrounding seaweeds [42,43]. Despite their ecologically significant functions like nitrogen fixation, denitrification, sulfide oxidation and manganese reduction [42,44], they are not frequently reported as endobionts [45,46]. On the other hand, members of the Xanthomonadaceae family to which OTU-5 belongs, are well-known plant endophytes [47] and have previously been isolated from marine algae [38,48]. Since many Xanthomonadaceae species cause plant diseases, the high number of Xanthomonadaceae endophytes within Bryopsis MX164 could be a sign of infection. The alga, however, showed no visible disease symptoms (e.g. bleaching), indicating a neutral or beneficial relationship.
In the Bacteroidetes group, we found two distinct clusters (i.e. OTU-1 and OTU-3) of endophytic bacteria, one within the Flavobacteriaceae family and one belonging to unclassified Bacteroidetes. The Flavobacteriaceae endophytes (OTU-1) show an especially strong association with Bryopsis as evidenced by their occurrence in all five samples. The phylum Bacteroidetes, and in particular the family Flavobacteriaceae, forms one of the major components of marine bacterioplankton and mediates a substantial proportion of the carbon flow and nutrient turnover in the sea during and following algal blooms [49]. Moreover, many novel Bacteroidetes members, some of which were characterized as morphogenesis inducers [50], have been isolated from the surfaces of marine macroalgae [1]. Whereas Bacteroidetes bacteria are obviously common epiphytes on macroalgae, Meusnier and co-workers [29] suggested the existence of an endophytic Cytophaga-Flavobacteria-Bacteroidetes (CFB) bacterium within Caulerpa taxifolia. In addition, Bacteroidetes bacteria are well-known endosymbionts of amoebae, plant-parasitic nematodes and insects [51][52][53]. Phylogenetic analysis, however, revealed that the Bacteroidetes endophytes of Bryopsis are more closely related to bacteria tightly associated with corals and sponges [54][55][56] than to CFB sequences isolated from green [29,50], brown [38] and red [57,58] macroalgae.
Finally, three Bryopsis samples (i.e. MX19, 164 and 263) contained Mycoplasma sequences (OTU-2). Mycoplasmas are well-known human and animal parasites, but are also common members of the intestinal bacterial flora of fishes and abalones where they may provide nutrients to their hosts [46,59,60]. Moreover, the close affiliation of Mycoplasma sequences isolated from Bryopsis and abalone species is perhaps not at all surprising as the latter generally feeds on a broad selection of algae [61]. Also Huang and colleagues [59] postulated that the presence of Mycoplasma species in the intestinal microflora of the abalone Haliotis diversicolor could be algal-food related. Additionally, this bacterial link between Bryopsis and abalone species might be extrapolated to other marine gastropod mollusks, supporting the hypothesis of Rao et al. [13] that the production of therapeutic kahalalides by the sea slug Elysia rufescens as well as by its Bryopsis food could actually be performed through an associated microorganism. Indeed, it has been shown that several metabolites initially assigned to eukaryotes are in fact of microbial origin [1].
In summary, molecular analysis revealed, for the first time, that Bryopsis harbors relatively restricted but taxonomically diverse communities of endophytic bacteria. The presence of Phyllobacteriaceae, Bacteroidetes, Mycoplasma, and in particular Flavobacteriaceae endophytes in several Bryopsis samples collected hundreds of kilometres apart indicates a close association between these endophytes and Bryopsis plants. Even though these endophytic bacterial communities within Bryopsis cultures might not fully represent those that are present within the alga in its natural environment, the bacteria identified in this study are at least part of the natural Bryopsis endobiotic flora. Future investigations of Bryopsis algae in natural environments, however, are necessary to complete the Bryopsis-bacterial endobiosis picture.

Supporting Information
Table S1 Bacterial 16S rRNA gene sequences isolated from algae (excluding BLAST hits) included in the phylogenetic analysis. (DOCX) Figure S1 A wide-range maximum likelihood tree showing the phylogenetic positions of endophytic clones and DGGE bands. Phylogenies were inferred from 16S rRNA gene sequences determined in this study (in bold), BLAST hits (see Table 1), Bacteroidetes, Proteobacteria and Mollicutes type strains, and algae-associated bacteria described in the literature (see Supplementary Table S2). The tree was generated in PhyML according the HKY + G4 algorithmic model. Bootstrap values above 50% are indicated at the branch nodes and the scale bar shows 10 nucleotide substitutions per 100 nucleotides. Asterisks denote sequences previously isolated from micro * -and macroalgae**. (TIF) Figure 5. A wide-range maximum likelihood tree showing the phylogenetic positions of endophytic clones and DGGE bands. Phylogenies were inferred from 16S rRNA gene sequences determined in this study (in bold), BLAST hits (see Table 1), Bacteroidetes, Proteobacteria and Mollicutes type strains, and algae-associated bacteria described in the literature (see Table S1). The tree was generated in PhyML according the HKY + G4 algorithmic model. Bootstrap values above 50% are indicated at the branch nodes and the scale bar shows 10 nucleotide substitutions per 100 nucleotides. Asterisks denote sequences previously isolated from micro * -and macroalgae**. The same phylogenetic tree without compressed branches is depicted in Figure S1. doi:10.1371/journal.pone.0026458.g005