Introduction

The diversity observed in the molecular architecture of the outermost surface layer of prokaryotic microorganisms is an example of the evolutionary adaptation of an organism to specific ecological and environmental conditions. The S-layer, a proteinaceous envelope that is present in many species belonging to the prokaryotic domains of Bacteria and Archaea, is regarded as the most ancient biological membrane that has remained through the evolution of microbes (Claus et al. 2005; Messner et al. 2013; Zhu et al. 2016). S-layers are monomolecular arrays of protein or glycoprotein subunits that self-assemble to form a two-dimensional lattice that completely covers the organism during all stages of growth (Sára and Sleytr 2000). Despite the abundancy of S-layer proteins (SLPs) in many prokaryotic cells, their functions have been explained only in few cases and many of them are still hypothetical. These lattices could function as protective coats, molecular sieves or ion traps, as structures involved in surface recognition, cell adhesion, inhibition of pathogens or as virulence factors (Sleytr et al. 2014; Gerbino et al. 2015b; Prado Acosta et al. 2016).

SLPs have been shown to possess exceptional physicochemical properties which make them a unique organizational structure with high potential for application in different areas of bionanotechnology such as generation of generally recognized as safe (GRAS) vehicles for the administration of antigens and other molecules of biomedical importance, and development of reactive solid-phase as biocatalysts, diagnostic devices or biosensors (Ilk et al. 2011; Sleytr et al. 2014). Moreover, they are also very useful as models to learn more about the strategies for stabilization, self-organization and functional evolution of proteins (Claus et al. 2005).

The presence of S-layers has been described in many bacterial species, including some of the genus Lactobacillus (Hynönen and Palva 2013). Lactobacilli are typically GRAS and have been isolated from different natural environments, including plants, human gastrointestinal and genital tracts, and food products. Years ago, our group demonstrated that both aggregative and non-aggregative strains of Lactobacillus kefiri, one of the most abundant lactobacilli present in the probiotic fermented milk named kefir (Garrote et al. 2005), carry a glycosylated S-layer (Garrote et al. 2004, Mobili et al. 2009b). These SLPs mediate the inhibition of Salmonella enterica invasion to Caco-2 cells (Golowczyc et al. 2007), the antagonism of Clostridium difficile toxins (Carasi et al. 2012), the interaction of L. kefiri with yeasts (Golowczyc et al. 2009), they also participate in the adhesion of L. kefiri to the gastrointestinal mucus (Carasi et al. 2014a), and protect bacterial cells against the deleterious effect of Pb2+ ions (Gerbino et al. 2015a). Despite all these interesting beneficial properties, up to now no information about the amino acid sequence of these L. kefiri SLPs has been reported.

In this work, we used both proteomic and genomic approaches to gain knowledge about the sequences of the genes encoding SLPs expressed by both aggregative and non-aggregative cultured strains of potentially probiotic L. kefiri. Differences in amino acid sequences lead to changes in secondary and conformational structures which are responsible, at least in part, for distinct surface properties of whole bacterial cells. Moreover, the knowledge of the sequence of these proteins is essential for the design and development of new bionanotechnological tools taking advantage of the extraordinary structural features of the SLPs.

Materials and methods

Bacterial strains and growth conditions

Twenty L. kefiri strains isolated from kefir grains (Garrote et al. 2001; Hamet et al. 2013), L. kefiri JCM 5818 obtained from the Japanese Collection of Microorganisms (Reiken, Japan) and L. kefiri ATCC 8007 were used in this study. Lactobacilli were cultured in deMan-Rogosa-Sharpe (MRS)-broth (DIFCO, Detroit, USA) at 37 °C for 48 h in aerobic conditions. Frozen stock cultures were stored at −80 °C in skim milk until use. The most relevant characteristics of these microorganisms and their SLPs are shown in Table 1.

Table 1 Characteristics of L. kefiri strains and their S-layer proteins
Full size table

Surface protein (S-layer) extraction

S-layer protein extraction from bacterial cells was performed using 5 M LiCl, as previously described (Carasi et al. 2012). Briefly, 100 ml of MRS culture of L. kefiri was harvested at stationary phase, collected by centrifugation (10,000×g at 10 °C for 10 min), washed three times with phosphate buffered saline (PBS, KH2PO4 0.144 g/l, NaCl 9 g/l, Na2HPO4 0.795 g/l, pH 7.2) and bacteria were resuspended in 15 ml of 5 M LiCl (J.T. Baker, Mallinckrodt Baker S.A., Edo de Mexico, Mexico) giving a bacterial suspension of OD550 = 25. The mixture was incubated in a shaker at 300×g at 4 °C for 60 min. Then, it was centrifuged (16,000×g at 10 °C for 30 min) and the supernatant was filtered through a membrane filter with 0.45 µm pore diameter (Millipore, USA). The supernatant was then dialyzed against PBS with 0.05% (v/v) Tween 20 for 24 h at room temperature, following by three additional buffer changes of 5 l each, using a cellulose membrane (SpectraPor membrane tube, MWCO 6000–8000, Spectrum Medical Industries, Rancho Dominguez, CA, USA). SLPs extracts were visualized by sodium dodecylsulphate–polyacrylamide gel electrophoresis (SDS–PAGE) in 12% separating and 4% stacking gels using the discontinuous buffer system according to Laemmli (1970). Gels were migrated on a BioRad Mini-Protean II (BioRad Laboratories, Richmond, CA, USA), with LMW Marker kit (GE Healthcare, Sweden) as molecular weight reference and were revealed using Colloidal Blue Staining.

Mass spectrometry

In-gel protein digestion

Gel pieces were destained in 3 × 100 µl washes of 25 mM NH4HCO3, 5% acetonitrile (ACN) (pH 8,5), followed by reduction with 100 µl 10 mM DTT in 25 mM NH4HCO3 at room temperature for 30 min and then alkylation with 25 mM iodoacetamide in 25 mM NH4HCO3, at room temperature for 30 min. Gel pieces were dessicated with 100 µl of ACN 100% for 10 min at room temperature and rehydrated with trypsin (Promega Trypsin Gold, TPCK treated) in 25 mM NH4HCO3, at an approximate trypsin to protein ratio of 1:20. The enzymatic reaction was carried out at 37º C for 3 h and peptides were extracted from the gel pieces with 100 µl of 0.2% TFA. The eluted peptides were dried with a Speed-Vac™ and then suspended in 5 µl of 50% ACN, 0.1% TFA. All assays were carried out in MultiScreen solvinert filter plates (Millipore) with a MultiScreen™ Vacuum Manifold 96-well system (MILLIPORE, Billerica, MA).

MALDI–TOF–MS analysis

Peptide mass fingerprint (PMF) analysis was used to determine protein identities. PMF and MS/MS fragmentation spectra were acquired with an UltrafleXtreme MALDI–TOF mass spectrometer (Bruker Daltonics, Bremen, Germany) at the SePBioEs Proteomics Service (UAB, Barcelona, Spain). Samples were spotted on a ground steel target plate (Bruker Daltonics) mixing 0.5 µl of each sample with 0.5 µl of freshly prepared matrix solution of a-cyano-4-hydroxycinnamic acid (Bruker Daltonics) at 10 mg/ml in a 30% ACN and 0.1% TFA aqueous solution. An external calibration was performed using a standard peptide mixture (Bruker Daltonics). Peptide masses were acquired within a range of ca. m/z 800–4000. Protein identification was carried out with Mascot search engine (Matrix Science Inc., Boston, MA) using the following parameters: two missed cleavages, 100 ppm tolerance, cysteine carbamidomethylation and methionine oxidation were set as variable modifications. Searches were performed using the NCBInr database restricted to Firmicutes.

The tandem mass spectra of selected peptides were obtained and de novo sequencing was performed manually using the table of amino acid masses.

Primer design, PCR conditions, sequencing and restriction profiles

The presence of SLP encoding genes was analyzed by specific PCR using purified chromosomal DNA as a template. Based on Lactobacillus buchneri CD034 genome, primers located outside (F1, F2, R1) and inside (TREG-f, TREG-r, GWIY-f) the ORF were constructed for PCR sequencing strategy. Their sequences are shown in Table 2, while their localization in the L. buchneri CD034 genome is shown in Online Resource 1. The PCR was performed as follows: one step at 95 °C for 5 min; 35 steps 94 °C for 30 s, 54 °C for 30 s and 72 °C for 3 min; and a final step at 72 °C for 10 min.

Table 2 Primers and their nucleotide sequences
Full size table

Sequencing reactions were performed using a ABI PRISM® BigDyeTM Terminator Cycle Sequencing Kit (Applied Biosystems, USA). The fluorescent-labeled fragments were purified following a BigDye® XTerminator™ purification protocol. The samples were resuspended in distilled water and subjected to electrophoresis in an ABI 3730xl sequencer (Applied Biosystems, USA).

The amplicons obtained using the primers F1 and R1 (or amplicon F2/R1 in the case of L. kefiri CIDCA 5818) were cleaved simultaneously with EcoRI and NcoI (New England BioLabs, UK) according to the manufacturer’s instructions. The digested DNA fragments were separated on a 1.5% (w/v) agarose gel.

In silico analysis of nucleotide and amino acid sequences

The SLP gene sequences obtained in this study, as well as those from the databases, were analyzed by ClustalOmega (Goujon et al. 2010; Sievers et al. 2011; McWilliam et al. 2013).

The amino acid sequence was deduced from the nucleotide sequence using the ExPASy Translate Tool (http://web.expasy.org/translate/) and in silico tryptic digestion was performed using ExPASy Peptide Mass (http://web.expasy.org/peptide_mass/), both softwares available on ExPASy portal (Artimo et al. 2012). The obtained profiles were compared manually with the experimental spectra previously obtained for each SLP. The theoretical physicochemical properties (pI, Mw, %Asp-Glu, %Arg, Lys) of the amino acid sequences were predicted using the software ProtParam (http://web.expasy.org/protparam/) (Gasteiger et al. 2005) and SignalP 4.0 (Petersen et al. 2011) was used to identify the signal peptide. Secondary structure prediction of SLPs was computed with PSIPRED protein structure prediction server (http://bioinf.cs.ucl.ac.uk/psipred/).

Nucleotide sequence accession numbers

The sequence data of complete genes have been deposited in the EMBL nucleotide sequence database under accession numbers LT601591, LT601592, LT601593, LT601594, LT601595, LT601596, LT601597, LT601598, LT601599 and LT601600.

Results and discussion

Mass spectrometry analysis of SLPs from L. kefiri strains

Due to the importance of L. kefiri as a potentially probiotic microorganism as well as a component of kefir microbiota (Garrote et al. 2005; Golowczyc et al. 2007; Carasi et al. 2014a, b, 2015), some studies about the biochemical and functional properties of SLPs belonging to L. kefiri strains have been conducted by our group in recent years (Mobili et al. 2009a, b; Golowczyc et al. 2007, 2009; Carasi et al. 2012, 2014a; Gerbino et al. 2015a). However, this is the first work reporting the amino acid sequences of SLPs from both aggregative and non-aggregative L. kefiri strains.

Lactobacillus SLPs are among the smallest so far described, with molecular masses ranging from 25 to 71 kDa (Hynönen and Palva 2013; Wasko et al. 2014), which agrees with the results presented in this work as well as to previous reports for the SLPs isolated from L. kefiri strains (Table 1). The identity of SLPs extracted from L. kefiri and digested by trypsinization was confirmed by PMF analysis, finding homology with L. buchneri CD034 and Lactobacillus parafarraginis F0439 SLPs. The amino acid sequence of several peptides was confirmed by MS/MS fragmentation and subsequent de novo sequencing. Interestingly, some of these fragments also matched with peptides present in SLPs of other phylogenetically related species, such as Lactobacillus otakiensis and Lactobacillus kisonensis (Endo and Okada 2007; Watanabe et al. 2009). Representative PMFs of the SLPs from L. kefiri strains are shown in Fig. 1. In concordance with previous results (Mobili et al. 2009b), SLPs from aggregating strains showed very similar spectral patterns (Fig. 1, spectra A–D), on the other hand more heterogeneity was observed in the various spectra corresponding to SLPs from non-aggregative strains (Fig. 1, spectra E–S). The analysis of all tryptic peptides detected in the PMF experiments (data not shown), led to identification of two peptides that were present in all samples. The presence of the tryptic peptides m/z 1103.6 and m/z 1200.6 was observed in the mass spectra from the SLPs of all L. kefiri strains studied, regardless of their aggregative ability. Moreover, these fragments were also present in the SLPs of L. buchneri CD034 and L. parafarraginis F0439. Both peptides were further analyzed by MS/MS fragmentation and de novo sequencing, and their sequences could be validated. The 1103.6 Da peptide corresponded to TREGDLWVK sequence, and the 1200.6 Da peptide corresponded to TYRGWIYGGK sequence.

Fig. 1
figure 1figure 1

Peptide mass fingerprints (PMF) of S-layer proteins (SLP) from representative aggregative (A–D) (SLP CIDCA 8345, 8347, 8348* and 83115) and non-aggregative (E–S) (SLP CIDCA 8310, 8314, 8315, 8317, 8332, 8335, 8343, 8381, 8385, 83111, 83113*, 83116, ATCC 8007 and JCM 5818*) strains of L. kefiri. *Mass spectra of these SLPs were also reported by Mobili et al. in a previous work (2009b)

Full size image

In general, the deduced amino acid sequences of mature Lactobacillus SLPs vary considerably, thus a similarity can only be found among related species (Antikainen et al. 2002; Avall-Jääskeläinen and Palva 2005; Hagen et al. 2005). Indeed, the remarkable similarities between SLPs from L. acidophilus-related bacteria have led to the proposal of employing their LC–MS/MS analysis for typing strains within this group (Podlesny et al. 2011). Similarly, an exceptionally high conservation was reported for the SLPs of two Lactobacillus hilgardii strains (Dohm et al. 2011) as well as for the SLPs of L. acidophilus NCFM and ATCC 4356 (Hynönen and Palva 2013), although the strains were clearly distinguishable. In the case of L. kefiri strains, although the presence of SLPs in L. kefiri strains was described more than eleven years ago (Garrote et al. 2004), and their heterogeneity had been characterized by using specific monoclonal antibodies and MALDI–TOF analysis in our laboratory (Mobili et al. 2009b), we were not able to obtain suitable tools to start to get reliable information about the amino acid sequence of these proteins until the complete genome of L. buchneri CD034 was published (Heinl et al. 2012).

L. kefiri SLP-encoding gene amplification

Taking into account the data obtained by mass spectrometry analysis, along with the availability of both the complete genomic sequence of L. buchneri CD034 and sequence read archives (SRA) available from other related species (L. parafarraginis, L. kisonensis, L. otakiensis, L. parabuchneri, L. parakefiri) (Endo and Okada 2007; Watanabe et al. 2009; Oki et al. 2012), the alignment of the SLP-encoding gene sequences from those lactobacilli was performed using the bioinformatics software platform Geneious (Biomatters Limited). Based on the alignment, the regions showing the highest homology were selected and the set of SLP-specific primers referred to as TREG-r (within the peptide of m/z 1103.6) and GWIY-f (within the peptide of m/z 1200.6) were designed, using the L. buchneri CD034 genetic code. Using these primers, an amplicon of 340-bp was obtained from genomic DNA of all L. kefiri strains. The fragments were then sequenced, and employing the software available on the ExPASy portal (http://web.expasy.org/translate), nucleotide sequences were translated into peptides. The five aggregative strains (group I) showed no differences in the amino acid sequence of this 106-mer fragment among them, whereas the non-aggregative strains could be separated into four different groups based on the similarities observed in this fragment: group II (strains CIDCA 8314, 8315, 8326, 83111, ATCC 8007), III (strains CIDCA 8310, 8317, 8319, 8332, 8343, 8344, 8381, 8385, 83110, 83116), IV (strains 8335, 83113) and V (strain JCM 5818).

Regarding these results, a total of ten strains belonging to different groups were selected and sequencing of the complete SLP-encoding genes was performed. The partial DNA sequences obtained with the combinations of F1/R1, GWIY-f/TREG-r, GWIY-f/F1 and TREG-r/R1 were assembled to obtain the complete S-layer gene sequences from L. kefiri CIDCA 8310, 8314, 8321, 8335, 8343, 8348, 83111, 83113 and 83115. For amplification of DNA from L. kefiri JCM 5818, the combinations F1/R1 and GWIY-f/F1 were replaced by F2/R1 and GWIY-f/F2 respectively, because of the absence of F1 region in this strain. All the S-layer gene sequences were deposited in the EMBL database with the corresponding accession numbers given in the “Materials and methods” section.

Since November 2015, two Genome Assemblies and Annotation reports for L. kefiri JCM 5818 are available in the NCBI site (Sun et al. 2015). Three different loci encoding hypothetical SLPs can be found. One of them shows 100% nucleotide identity with the sequence reported in this work for L. kefiri JCM 5818 (accession AYYV01000002.1; locus KRM54156), which confirms our results, meanwhile the other two show 72% (accession AYYV01000037.1; locus KRM52454) and 60% (accession AYYV01000004.1; locus KRM53976) nucleotide identity, respectively. These findings suggest that at least three SLP-encoding genes are present in the genome of the reference strain JCM 5818. Locus KRM52454 encodes for a polypeptide of 540 amino acids and locus KRM53976 encodes for a polypeptide of 565 amino acids whereas the protein described in this work shows a total length of 578 amino acids. The presence of more than one SLP-encoding gene has been described in different lactobacilli species, including Lactobacillus brevis ATCC 14869 (Jakava-Viljanen et al. 2002); L. acidophilus ATCC 4356 (Boot et al. 1995; Palomino et al. 2016) and L. acidophilus NCFM (Goh et al. 2009). Moreover, it was reported that the expression level of different SLP’s genes could change depending on environmental conditions such as high salt concentration or gastrointestinal tract milieu (Ramiah et al. 2009; Palomino et al. 2016).

Although the primers located inside the ORF (TREG-f, GWIY-f) could generate amplification products of the three mentioned putative genes, it is important to note that the primers used for sequencing, which are located outside the ORF (F2 and R1), only match to the flanking region of the locus KRM54156 of L. kefiri JCM 5818. This finding allows us to assume that the amplicons obtained correspond to a gene located in that region of the genome. The presence of other SLP encoding genes in the L. kefiri strains of our collection as well an analysis of differential expression patterns will be subjects of future studies.

Besides F1/R1 amplicons (or F2/R1 amplicon in the case of L. kefiri JCM 5818) of 2500–3000 bp-length were obtained from genomic DNA of these L. kefiri strains. The digestion of these amplicons with a mix of EcoRI and NcoI, resulted in different fragmentation patterns that match with the distinct groups described above (Fig. 2).

Fig. 2
figure 2

Fragmentation patterns of amplicons F1/R1 (1 L. kefiri CIDCA 8310; 2 L. kefiri CIDCA 8343; 3 L. kefiri CIDCA 8314; 4 L. kefiri CIDCA 83111; 5 L. kefiri CIDCA 8335; 6 L. kefiri CIDCA 83113; 7 L. kefiri CIDCA 8321; 8 L. kefiri CIDCA 8348; 9 L. kefiri CIDCA 83115) and F2/R1 (10 L. kefiri JCM 5818) with a mix of EcoRI and NcoI. Lane M molecular weight marker

Full size image

Amino acid sequences from L. kefiri SLPs

The SLP encoding genes of the selected L. kefiri strains were translated into the corresponding protein sequences, and then confirmed by comparison of each experimental PMF spectrum with the theoretical tryptic digestion obtained with the PeptideMass tool (http://web.expasy.org/peptide_mass/). Even though it is not possible to discard the presence of other contaminant proteins in the gel band used to obtain the PMF, more than 79% of the total intensity of the experimental spectra is covered by the amino acid sequences obtained for each SLP (Online Resource 2). Additionally, due to the presence of sugar residues, there are some peaks corresponding to glycopeptides that cannot be assigned in these analyses. The deduced amino acid sequence alignment of ten SLPs is shown in Fig. 3.

Fig. 3
figure 3figure 3

Alignment of ten SLPs sequences from L. kefiri strains. The sequence corresponding to GWIY-f/TREG-r fragments (blue box) and the O-glycosylation site (red box) are marked. Asterisks indicates positions which have a single, fully conserved residue, colon indicates conservation between groups of strongly similar properties—scoring >0.5 in the Gonnet PAM 250 matrix, period indicates conservation between groups of weakly similar properties—scoring ≤0.5 in the Gonnet PAM 250 matrix

Full size image

The total length of the mature proteins varies from 492 to 576 amino acids, with the SLPs from L. kefiri CIDCA 8335 and 83113 the shortest and the SLP from L. kefiri CIDCA 8310 the longest polypeptides respectively. As expected, cysteine is absent in all L. kefiri SLPs analyzed here, and the percentages of hydrophobic and hydroxylated amino acids varies from 34.9 to 38.3% and from 24.6 to 29.2%, respectively among strains (Table 3). The amount of positively charged amino acid residues ranged from 9.5 to 10.5% and is always higher than the amount of negatively charged residues (5.7–7.3%), thus leading to a calculated isoelectric point between 9.37 and 9.60 (Table 3). All these results are similar to those reported for other lactobacillus SLPs (Avall-Jääskeläinen and Palva 2005; Wasko et al. 2014), and there are only slight differences between aggregative and non-aggregative strains.

Table 3 Amino acid composition of mature SLPs of L. kefiri strains
Full size table

All proteins start with a predicted 31-mer leader peptide for membrane translocation showing the same amino acid sequence in all the tested strains, except for the residue 27, where a threonine present in the non-aggregative strains is changed by a serine in the aggregative strains (Fig. 3). Additionally, the sequence of all the signal peptides includes the A–X–A motif that precedes the cleavage site for type I signal peptidases commonly found in Gram-positive bacteria (van Roosmalen et al. 2004). The similarities observed between L. kefiri SLPs and those from L. buchneri, L. parafarraginis, L. farraginis, L. otakiensis, and L. kisonensis among others, correlate with the phylogenetical relationship that exists among these species of lactobacilli (Sun et al. 2015). Interestingly, protein BLAST analyses showed that there is almost a 100% identity of their predicted leader peptides with those from strains of phylogenetically related species such as L. parakefiri, L. buchneri, L. parabuchneri, L. parafarraginis, L. farraginis, L. sunkii, L. kisonensis and L. otakiensis. Additionally, the threonine to serine change at residue 27 was also observed in the predicted leader peptides of SLPs from the strains L. buchneri CD034 and NRRL B-30929, respectively.

Nevertheless, all L. kefiri SLPs described here share some specific characteristics with the SLPs from other lactobacilli, such as the absence of cysteines and the highest ratio of positively/negatively charged residues among all the bacterial SLPs (Avall-Jääskeläinen and Palva 2005; Hynönen and Palva 2013; Wasko et al. 2014). These properties result in the absence of intra- or inter-catenary disulfide bonds and also a high predicted isoelectric point comparable to the SLPs from other lactobacilli. Although the SLPs of lactobacilli are mainly reported as non-glycosylated proteins (Hynönen and Palva 2013), the occurrence of glycosylation as a post-translational modification was previously reported for the SLPs of twelve strains of L. kefiri by our group (Mobili et al. 2009b) and in the present work it has also been demonstrated for eleven new strains. In this sense, all L. kefiri SLPs display at least one glycosylation site located in the N-terminal region of the proteins, which has also been described in SLPs of L. buchneri 41021/251 (Möschl et al. 1993) and L. buchneri CD034 and NRRLB-30929 (Anzengruber et al. 2014). This agrees with other similarities observed at primary sequence level between SLPs belonging to L. kefiri and L. buchneri species.

The molecular weight of mature SLPs did not match with that observed in SDS-PAGE for the gel band that was analysed (differences around 10 kDa) (Tables 1, 3). This apparent discrepancy could be explained considering the additional contribution of sugar moieties to the protein molecular weight and/or other structural characteristics that could influence the electrophoretic mobility of SLPs, even in denaturing conditions.

Most SLPs display two structural regions, i.e. the region involved in the attachment to the cell envelope and the region involved in S-layer assembly. These regions have been characterized at least for seven SLPs from of different lactobacilli strains belonging to L. acidophilus (Smit et al. 2001), Lactobacillus crispatus (Antikainen et al. 2002; Chen et al. 2009; Hu et al. 2011; Sun et al. 2013), L. brevis (Avall-Jääskeläinen et al. 2008) and L. hilgardii species (Dohm et al. 2011). The C-terminal region of the SLPs of L. acidophilus and L. crispatus, and the N-terminal region of the SLPs of L. brevis and L. hilgardii are the most conserved part of the protein and are responsible for anchoring to the cell envelope. On the other hand, the most variable part of the protein seems to be involved in the self-assembly of the SLPs monomers on the bacterial surface (Smit et al. 2001; Antikainen et al. 2002; Åvall-Jääskeläinen et al. 2008; Dohm et al. 2011; Hu et al. 2011; Sun et al. 2013). These findings, together with the absence of investigations carried out in lactobacilli species more phylogenetically related to L. kefiri, highlight the need for additional studies using truncated proteins to characterize these regions in the SLPs of aggregative and non-aggregative L. kefiri strains.

Table 4 Secondary structure prediction of mature SLPs of L. kefiri strains
Full size table

A prediction of secondary structure of the L. kefiri SLPs was performed using the Psipred software (Table 4). There are no major differences among the strains, except for SLPs from the strains L. kefiri CIDCA 8335 and CIDCA 83113 which showed a higher percentage of α-helix and a lower percentage of random coil than the other SLPs. Our results are similar to those obtained for the SLP of L. buchneri CD034 using the same software. On the other hand, no differences in β-sheet contents were observed between aggregative and non-aggregative strains, in disagreement with the results previously reported by Mobili et al. (2009a, b) using FT-IR spectroscopy. This discrepancy could be explained considering that glycan residues (that were not included in our predictive analysis) could influence the secondary structure of the whole protein.

Our results show that, unlike the relatively high intra-strain homology observed at amino acid sequence level in the N-terminal region of L. kefiri SLPs, the C-terminal part of the proteins shows the most evident differences among strains. Different groups can be distinguished regarding the complete amino acid sequences of the mature SLPs of ten representative strains (Fig. 3). Some similarities found in the SLPs from non-aggregative strains allowed us to group the studied strains as follows: L. kefiri CIDCA 8314 and 83111; L. kefiri CIDCA 8310 and 8343; L. kefiri CIDCA 8335 and 83113; and L. kefiri JCM 5818, corresponding to the previously mentioned groups II, III, IV and V respectively.

It should be noted that the analysis of the similarities observed in the PMF spectra (Fig. 1), allowed us to extend this classification to all L. kefiri SLPs studied here, and the results were consistent with those coming from the sequencing analyses. Interestingly, the SLPs of the three aggregative strains (L. kefiri CIDCA 8321, 8348 and 83115) display a 100% of sequence identity, even though each strain was isolated from a different natural source (kefir grains AGK2, AGK4 and AGK1, respectively) (Table 1). This result strongly suggests that some common structural characteristics of these proteins are responsible, at least in part, for the aggregative ability of the whole bacteria. However, it cannot be discarded that sugar residues bound to the polypeptide skeleton could also contribute to this surface characteristic. Moreover, other non-covalently bound exoproteome components, named as SLAPs in other lactobacilli (Johnson et al. 2015), might contribute to aggregative and non-aggregative phenotype of each strain. Further studies will be necessary to elucidate the specific SLPs domains involved in this phenomenon.

Comparing N-terminal versus C-terminal regions of L. kefiri SLPs, some differences were found regarding their amino acid composition (Table 5) and the sequence homology among strains (Fig. 3). Clearly, in the N-terminal region of the proteins, the percentage of positively charged amino acids is nearly 2.5-fold higher than the percentage of negatively charged residues, which correlates with their high predicted pI values. Moreover, the N-terminal region is relatively conserved among strains, showing a low intra-species variability in this portion of the SLPs. On the other hand, the C-terminal region behaves quite differently. Positively and negatively charged amino acids are almost equally distributed along this region and they appear in very similar percentages in all the SLPs analyzed. Interestingly, the C-terminal region of the SLPs harbors the major differences in the amino acid sequence among all the strains studied here.

Table 5 Amino acid composition of N-terminal and C-terminal regions of mature SLPs of L. kefiri strains
Full size table

The knowledge of the amino acid sequence of the SLPs of different L. kefiri strains provides relevant data not only for a better understanding of the mechanisms involved in the functionality of these proteins with exceptional physicochemical properties, but also to contribute to the development of products of biotechnological interest from safe and potentially probiotic lactic acid bacteria.