Before staining, all samples were diluted (10-‐1 and 10-‐2) in sterile 4X PBS, and 100 µl of suspensions were filtered thought GTTP Isopore filters of 0.22 µm pore size and 16 mm diameter (Milipore). Filters were stored at -‐20ºC until use.
11.2. 4’-‐6-‐diamidino-‐2-‐phenylindole stain (DAPI)
In order to determine the number of cells present in each culture, a piece of filter was stained with 25-‐
30 µl of 4'-‐6-‐diamidino-‐2-‐phenylindole solution (1 µg ml-‐1)(Porter & Feig, 1980) during 1.5 min at room temperature. Then, washed with sterile MiliQ water and absolute ethanol and dried at room temperature and darkness. Finally, filter was mounted with a drop of Citiflour AF1 (Citifluor ltd) and covered in a microscope slide. Cells were quantified using a fluorescence microscope (Axio imager.A1, Zeiss) with filter set 49 (G 365, FT 395, BP 445/50, Zeiss). Counts are reported as means calculated from 15 randomly chosen microscope fields. Fifteen microscope fields of 1200 µm2 were the optimum number of fields with the lowest standard deviation. A number of fields > 15 did not produced significant modifications in the means and the standard deviations.
11.3. Fluorescence in situ hybridization (FISH)
To evaluate the integrity and ribosome containing of Salinibacter cells, a piece of filter was hybridized with EHB-‐412 monolabel probe (Antón et al., 1999; Antón et al., 2000). EHB-‐412 probe (5’-‐
TACGCCCCATAGGGGTGT-‐3’; 50 µg ml-‐1) was diluted in sterile MiliQ water to a final concentration of 1 µg ml-‐1. The hybridization was performed at 45% formamide and the hybridization buffer was prepared as follows: 360 µl 5M NaCl, 40 µl 1M Tris-‐HCl pH 8.0, 904. 5 µl formamide, 695.5 µl Mili Q water, and 2 µl of 10% SDS. Filters were placed on a clean slide and each was hybridized with 20 µl of hybridization mix (4 µl probe, 16 µl hybridization buffer). Slides were placed into a hybridization chamber and incubated at 46ºC during 2 h. Then, to eliminate the unspecific hybridizations, filters were immersed in a washing buffer (300 µl 5M NaCl, 1 ml 1M Tris-‐HCl pH 8.0, 500 µl 0.5 M EDTA, MiliQ water to complete a final volume of 50 ml, and 50 µl of 10% SDS) and incubated at 48ºC during 15 min. After washing, filters were dried at room temperature and darkness and then, stained with DAPI. Finally, hibridized cells were quantified using a fluorescence microscope (Axio imager.A1, Zeiss) with filter set 49 (G 365, FT 395, BP 445/50, Zeiss) for DAPI, and the HQ: Cy3 filter set (AF analysentechnik; HQ 545/30, Q 570 lp, HQ610/75).
Counts were reported as means calculated from 15 randomly chosen microscope fields and the percentage of hibridized cells was calculated based on the total of DAPI counts of each sample.
IV. RESULTS AND DISCUSSION
Results and Discussion: Chapter 1
-‐69-‐
CHAPTER 1: Intraspecific diversity and biogeography of S.ruber strains
1.1.Background
Growth of the extremely halophilic bacterium S. ruber (Antón et al., 2000) is constrained to relatively small water bodies in restricted areas on Earth. S. ruber has been isolated from different areas of the world, and in sites as diverse as Mediterranean coastal solar salterns (Peña et al., 2005) or the remote Andean Peruvian salterns of Maras at 3,380 m above sea level (Maturrano et al., 2006a). The extreme conditions and geographical isolation of its environments are optimal circumstances for observing allopatric speciation (Coyne & Orr, 2004; Whitaker, 2006). Preliminary analyses based on fingerprinting genomic traits, such as PFGE or RAPD, although indicating a certain incipient trend, did not render a clear cut geographical discrimination among isolates (Peña et al., 2005). In order to discern biogeographical patterns in S. ruber, ten strains from five different locations were selected to study, through MLSA, the intraspecific diversity within the same group. For this, twelve protein-‐coding genes, which had been observed as phylogenetically informative, were selected (Sória-‐Carrasco et al., 2007). In addition, a metabolomic approach by Ion Cyclotron Resonance Fourier Transform Mass Spectrometry (ICR-‐FT/MS) was performed to evaluate some phenotypic evidence for allopatric segregation of members of S ruber, by identification of phenotypic patterns of the chemical extracts of this strain collection (as detailed in Materials and Methods section).
1.2.Multilocus sequence analysis (MLSA)
Multilocus sequence analysis was applied to study the intraspecific diversity of S. ruber strains isolated from different geographic localizations. The obtained sequences of specific coding-‐protein genes from each strain were concatenated and analyzed to calculate the total number of synonymous or nonsynonymous substitutions. Finally, different phylogenetic reconstructions were applied in order to evaluate stability of the genealogies by including and excluding the 16S rRNA gene sequences.
1.2.1.Amplification of protein-‐coding genes
A total of ten S. ruber strains (Table 3) were selected in order to study their intraspecific diversity by MLSA. The selected strains were representative of three main geographic areas: Mediterranean (M8, M31, P13, P18, E3, E7), Atlantic (C9, C14) and Peruvian (PR1, PR3).
Results and Discussion: Chapter 1
-‐70-‐
The study was conducted using 12 coding-‐protein genes that were selected according to previous phylogenetic S. ruber studies (Soria-‐Carrasco et al., 2007).
DNA amplifications were performed following standard procedures as indicated in Materials and Methods. Amplified products of proper size were obtained with all primers tested and in most of the analyzed strains. Figure 12 shows examples of these amplifications in 10 tested strains, where the expected size of amplicons for groEL (1293 bp) and tuf (996 bp) genes can be seen. Excluding ES4 strain, all strains showed a good signal and mostly optimal concentrations to be sequenced (Fig. 12). In order to obtain a good concentration for sequencing purposes, DNAs yielding lower concentrations were re-‐
amplified (Fig. 12). PCR products were purified using standard procedures, quantified and, finally sequenced (see Materials and Methods).
Figure 12: Amplified products of coding-‐ protein genes in S. ruber strains. The picture shows groEL (A) and tuf (B) genes separated in 1.5% agarose gel and visualized by ethidium bromide staining. Size of amplicons and names of strains are also specified (M: Lambda DNA/PstI marker; NC: negative control).
1.2.2.Sequencing of protein-‐coding genes
Sequencing of amplicons was performed with the same primers used for gene amplification, and under the conditions explained in the Materials and Methods section. Only seven out of the twelve tested genes yielded good sequences and could be used for phylogenetic studies. The sequences obtained for each gene were checked and corrected using the program Sequencher v4.7 (Gene Codes Corp.).
Subsequently, the sequences were aligned using the ClustalX 1.83 program. The alignment was then manually improved. Finally, hypervariable or unalignable positions were removed using Gblocks program (http://molevol.ibmb.csic.es/Gblocks-‐_server.html).
The total number of substitutions independently of whether they were synonymous or nonsynonymous, as well as the number of insertions-‐deletions, were calculated for each gene (Table 8).
Results and Discussion: Chapter 1
present any type of substitution, corresponding to the most conserved gene in these strains. Insertions-‐
deletions were not present in any of the analyzed genes.
Results and Discussion: Chapter 1
-‐72-‐
1.2.3.Phylogenetic reconstruccions
Phylogenetic reconstructions were based on multiple concatenated genes from different sequences of each strain data sets (Table 8). Multiple analyses were performed in order to find topology changes due to the effect of the gene composition of the alignments, and to evaluate tree topology stabilities, as previously recommended (Ludwig & Klenk, 2001). Bootstrap values were performed by using the PHYML program with a total of 100 replicates.
As mentioned before, phylogenetic analyses were performed by including and excluding indels, as well as by using different data sets: one including the 16S rRNA gene in the concatenate (7,995 nucleotide alignment, Fig. 13) or disregarding it (6,513 nucleotide alignment, Fig. 14). In the first case, it was remarkable that Ebro strains (Mediterranean) were affiliated with the Peruvian and Atlantic strains (Fig.
13).
Similar results were also observed when removing the 16S rRNA gene from the analysis, and were even less related to the rest of the Mediterranean strains. Besides, Atlantic strains did not show a clear phylogenetic trend (Fig. 14). In any case, most of the trees gave congruent topologies, independently of the use of PHYML or ARB programs, or the used algorithms (ML, maximum likelihood and NJ, neighbour joining). Only ML showed different topologies when including the indels in the analysis. However, none of the tree topologies obtained showed a clear geographic trend. Despite the relatively low bootstrap value for the position of strain E7, the remaining topology was robust with bootstrap values, always above 80%. Contrarily to the same reconstruction, where the SSU rRNA gene was included, bootstrap values were lower. However, despite a lower robustness of the tree topology, there was no doubt about the common affiliations between the Peruvian and Atlantic strains, and between C9 and E3 strains (Fig.
13).
In general, the trees agreed with regard to their topologies, since only M8 acquired a stable position when including the 16S rRNA gene sequence in the analysis. Altogether, both reconstructions did not show a clear geographical segregation of the selected strains, in contrast to observations made with other extremophiles (Whitaker et al., 2003). Strains from Alicante (P13 and P18) affiliated together with those of Majorca (M8 and M31). However, the Mediterranean strains E3 and E7 affiliated together with those from the Atlantic (C9 and C14) and Peru (PR1 and PR3). Neither previous studies with fingerprinting techniques (Peña et al., 2005), nor those with an MLSA of gene stretches of nearly 8,000 homologous positions (Fig 13), were informative enough to resolve biogeographical segregation.
Results and Discussion: Chapter 1
-‐73-‐
This could be an indication that the process of genetic divergence may be still at an early stage, not rendering clearly resolvable trends. However, given that the size of the S. ruber genome is about 3,000 open reading frames (Mongodin et al., 2005), and despite the fact that the genes were selected from a set of putative phylogenetic markers (Soria-‐Carrasco et al., 2007), those may not be adequate for the understanding of a subtle geographical segregation.
In the future, intraspecific whole-‐genome comparisons with S. ruber strains might indicate which genes could be useful for understanding allopatric differentiation based on genetic drift.
Figure 13: Phylogenetic reconstruction based on a PHYML algorithm corresponding to 8 housekeeping genes including SSU rRNA gene. Strains from different geographical areas are marked with their respective colors. The bar indicates 1% of sequence divergence.
Figure 14: Phylogenetic reconstruction based on a PHYML corresponding to the 7 housekeeping genes. Strains of different geographical areas are marked with their respective colors. The bar indicates 1% sequence divergence.
Results and Discussion: Chapter 1
-‐74-‐
1.3.Metabolomic comparisons of strains
As stated above, standard genotyping techniques may not always help in clearly resolving intraspecific diversity, so there is a need to apply new approaches to understand allopatric segregation of members of the same species (Ramette & Tiedje, 2007). For this reason, a nontargeted metabolite profiling approach, using high-‐field ICR-‐FT/MS, was evaluated in the chemical extracts of our S. ruber strain collection.
1.3.1.Metabolome composition analyses
The first experiment was performed with 28 isolates of S. ruber coming from seven geographical areas in the world (Table 3), which were divided into three regions: Mediterranean (10 strains), Atlantic (13 strains) and Peruvian (5 strains). All of them were grown simultaneously with the same medium and under identical environmental conditions to avoid culture-‐dependent differences. Each cell suspension was processed for the metabolite extraction resulting in three cellular fractions: extracellular (E), cellular soluble (CS) and cellular insoluble (CI) fractions, which were grouped and analyzed by ICR-‐FT/ MS (see Materials and Methods).
The complete set of metabolomes rendered a total of 247,255 m/z, from which 11,880 were attributed to known elementary compositions containing the elements C, H, O, N and S. Single-‐peak occurrence was reduced from 11,880 (verified by isotopic assignments of elementary composition) to a total of 8,873 metabolites at a m/z lower than 550 a.m.u. (highest probable assignments), which were used for statistical analysis (Table 9).
1.3.2.Statistical analyses and proposed models
The statistical model used for data processing was the partial least squares for discriminant analysis (PLS-‐DA), which revealed statistically significant differences between the different geographical areas (P<0.05) in the three analyzed fractions (Table 9). Models were calculated independently for the three cellular fractions (extracellular, cellular soluble and cellular insoluble), and the cellular insoluble fraction (CI) was chosen as the most descriptive of the model (Fig. 15).
In addition, the Pareto scaling of the intensity values with a logarithmic transformation of the data was chosen to consider all masses equally, including those with medium-‐ and low-‐intensity values (van den Berg et al., 2006). PLS-‐DA using four significant components, R2Y (cum) was equal to 0.98 and Q2 (cum) was equal to 0.45, both with values indicating high predictive power.
Results and Discussion: Chapter 1
-‐75-‐
Although most of the analyses were performed using the whole metabolome, the common metabolome in all strains (also called core metabolome) was formed by about 2,550 unique masses distinctly expressed in the different analyzed groups. Moreover, the discriminative metabolome (i.e. not common masses to all extracts) consisted in 6,323 unique masses (Table 9). The number of discriminative masses considered for statistical analysis and for geographical discrimination, all of them analyzed in positive mode.
Results and Discussion: Chapter 1
-‐76-‐
The score scatter plot and the loading plots revealed statistically significant differences (P<0.05) between the different samples. The score scatter plot (Fig. 15a) presented a view of how well the different geographical origins (classes) were separated on the basis of their metabolomic composition (X variables). Thus, metabolome comparisons, focusing on geographically discriminative data, yielded clear-‐cut allopatric differences. In the same way, nontargeted analysis revealed unique features for each group of isolates. Fig. 15b shows the loading plot in where the different masses characteristic to each geographical area were differently coloured (red for Atlantic, green for Mediterranean and blue for Peruvian strains). The discriminative masses (variables m/z) for each origin of isolation were chosen according to their correlation coefficient value. Those having the highest coefficients were considered to be relevant (that is, variables (m/z) with a correlation value higher than |0.002|).
Figure 15: PLS-‐DA models of all cellular insoluble fractions analyzed with electrospray-‐ positive mode ICR-‐FT/MS.
(a) Score plot model shows the differentiation based on the geographical origin of the analyzed samples. (b) Loading plot represents the masses of known elementary composition (C, H, O, N, S and m/z <550) used for discriminating analysis, and correlating to the geographical origins. The masses with a high correlation with geographical origin are highlighted with their corresponding colour depending on the area, while the non-‐
discriminating masses are represented in yellow.
A
B
Results and Discussion: Chapter 1
-‐77-‐
Those masses associated with the highest correlation coefficients were represented in the van Krevelen projection (H/C versus O/C on the basis of their elementary composition values; Fig. 16)(Wu et al., 2004). In this regard, the most relevant markers were C, H and O molecules, with only a few metabolites containing sulfur or nitrogen (Fig. 16a). The comparisons of these molecules with the total metabolic spaces showed that the discriminative metabolites may be structurally aliphatic and depleted in oxygen (Fig. 16b). Thus, those components generally associated to cell membranes, such as fatty acids and terpenoids, may be responsible for the geographic differentiation.
Figure 16: Representations of all discriminating m/z values from metabolome analysis. (a) Independent of their geographical origin but coloured only as a function of their attributed elementary composition (CHO, CHON, CHOS or CHONS) and visualized in a van Krevelen diagram (H/C versus O/C). Most of the discriminative metabolites contain only C, H and O and these are compared within a van Krevelen type of diagram to the CHONS-‐containing metabolites of general metabolome databases (www.metabolome.jp, www.genome.jp/kegg/) shown in gray in the figure. Note that the triangular region corresponds to peptides (CHON and CHONS); (b) All discriminating CHO metabolites in a van Krevelen diagram coloured as a function of their origin and compared to the metabolites of general metabolome databases (www.metabolome.jp, www.genome.jp/kegg/) shown in gray in the figure.
A
B
Results and Discussion: Chapter 1
-‐78-‐
Besides, between the two principal components, a relationship for geographical distance between the isolation sites could be found. It seemed that for the second principal component, the Atlantic strains showed intermediate differences with those of the Mediterranean and Peruvian. When specifically directing the recognition of discriminative metabolites among geographically distinct metabolomes, a set of conspicuous compounds were unambiguously assigned to a sulfonolipid family (Table 10). The compounds forming this family have been observed to be major components of the cell envelope of Cytophaga (Godchaux III & Leadbetter, 1984), a member of the same phylum that S. ruber (Antón et al., 2002). These compounds, which could account for 10% of total cellular lipids, have been proposed as signatures for S. ruber identification (Corcelli et al., 2004). In fact, one of them (C35H67NO8S, m/z=660.4505), whose higher intensity was observed in the Peruvian strains, has been reported to be characteristic of S. ruber (Table 10) (Corcelli et al., 2004). The ICR-‐FT/MS approach, with a mass precision lower than 600 p.p.b., revealed that S. ruber may contain at least nine additional sulfonolipids analogous to C35H67NO8S with a mass range 644–688. These compounds differ from the originally described sulfonolipid in their elementary composition, with variations in their side chain length, insaturation or hydroxylation degree, with variations in CH2, H2 and O, respectively (Table 10). All these compounds were found in all of the analyzed samples (in negative mode) and with identical intensity ratios between isolates from the same location, except for m/z 676.4454 (C35H68NO9S) and m/z 688.4455 (C36H68NO9S), which seemed to be exclusive of the Atlantic strains (Table 10).
In addition, the metabolomic approach allowed the targeted search for special metabolic traits considered to be relevant in the organisms’ phenotype. Previous biochemical studies on S. ruber-‐type strain M31 revealed the presence of an active, hitherto unreported, rhodopsin type of membrane proton translocation system, the xanthorhodopsin, responsible for the putative phototrophy of S. ruber (Balashov et al., 2005). In the same way, the genome sequence of the same organism revealed the coding region of one halorhodopsin (Peña et al., 2005) and two sensory rhodopsin homologous genes (Mongodin et al., 2005).
Searching for an indication of the presence of retinal, the chromophore that bound to rhodopsins, an experimental positive mass 285.22125 (theoretical 285.22129) was found in all samples. However, the m/z value was only discriminative for the Mediterranean strains.
Results and Discussion: Chapter 1
Table 10: Proposal elemental composition of masses assigned to sulfonolipids with their structural variations from C35H67NO8S, originally described by Corcelli et al. (2004) as C35H66NO8S. All these compounds were found in analyzed samples in negative mode, where n indicates the number of analyzed strains, and ND, not detected.
1.3.3.Discriminative analysis of Mediterranean strains
An independent experiment was performed to evaluate the metabolomic composition of four replicates from five Mediterranean strains: P13 and P18 from Alicante, M8 and M31 from Majorca, and IL3 from
Results and Discussion: Chapter 1
elaborated to the discriminative analysis of the Mediterranean strains. The soluble cellular fraction was chosen as representative in the OPLS-‐DA model, which rendered equivalent but clearer results than PLS-‐
DA, showing a clear separation between the three groups of strains (Fig. 17). Therefore, these differences observed between different strains may be attributed to strain-‐specific metabolisms rather than sample-‐to sample variations. pathways (confirmed by KEGG and Japanese metabolome database).
In contrast to previous results (Peña et al., 2005), when searching for discriminative phenotypes at a more reduced geographical scale using the ICR-‐FT/ MS approach, a phenotypic segregation in individual locations was observed (Fig. 17a). The main discriminative metabolomics profile features were different from those giving resolution at a larger geographical scale.
A
B
Results and Discussion: Chapter 1
-‐81-‐
In such cases, geographical differences were associated to strain-‐specific compositions of N-‐containing molecules (Fig. 17b). The confrontation of their exact masses with the KEGG and the Japanese metabolome database (www.metabolome.jp) indicated that the discriminative molecules were involved primarily in the core metabolism (that is, carbohydrate, amino acid, and fatty acid biosynthesis and metabolism).
1.4.Conclusions
These findings revealed that intraspecific metabolic diversity of S. ruber can be readily detected by the ICR-‐FT/MS approach and that such diversity can be associated to different geographical patterns at different metabolic levels, which could not be revealed by standard genetic methods previously used to assess biogeography of prokaryotes (Ramette & Tiedje, 2007; Whitaker et al., 2003).
MLSA approach, based on different gene data sets, did not resolve putative genetic-‐geographic patterns, as the genetic divergence may be too subtle for the given selection of genes. However, one must take into account that, despite the fact that large sets of concatenated genes tend to reflect the organismal phylogeny (Soria-‐Carrasco et al., 2007), perhaps only full genome sequences may reflect geographical isolation in the strain collection of S. ruber. This could result in accordance with the taxa segregation that correlates with the average nucleotide or amino-‐acid identity of shared genes (Konstantinidis &
Tiedje, 2005).However, the still sparse database of full genomes, makes the metabolomic approach a fast and less expensive alternative for revealing prokaryotic biogeography, with the added value of being discriminative at different levels at the geographical scale.
It seems clear that different studied regions led to the isolation of strains sharing common metabolic traits, such as, the distinct production of sulfonolipid derivates. However, differences were generally related to quantitative composition yields, rather than qualitative production of distinct compounds.
Nevertheless, at the molecular level, these facts could be also attributed to transcriptional or posttranscriptional regulations rather than composition changes in genes at the genomic level. In addition, the metabolic differences correlated with the geographical areas, influenced perhaps by environmental conditions such as climate and distance, since Peruvian and Mediterranean strains were found as the most different.
Results and Discussion: Chapter 1
-‐82-‐
The major forces for these differences may be related to their distinct response to the environmental
The major forces for these differences may be related to their distinct response to the environmental