A total of 122 sequenced genomes available from GenBank (Supplementary Table S1) were used for the computational analysis. They belong to 15 different species of coagulase negative Staphylococcus that have been detected in humans, among other sources (Becker et al., 2014): S. auricularis (n = 1), S. capitis (n = 7), S. epidermidis (n = 34), S. haemolyticus (n = 32), S. hominis (n = 6), S. lugdunensis (n = 9), S. massiliensis (n = 1), S. pettenkoferi (n = 1), S. saprophyticus (n = 8), S. schleiferi (n = 5), S. sciuri (n = 1), S. simulans (n = 3), S. succinus (n = 1), S. warneri (n = 10), and S. xylosus (n = 3).

All the 122 CoNS genomes were analyzed with the CRISPRFinder tool (Grissa et al., 2007) for the search of possible CRISPR structures. Then, the loci of those genomes that were positive for CRISPR candidates were collected and further annotated with Artemis (Rutherford et al., 2000) for the search of cas genes, and with ISFinder (Siguier et al., 2012) for the search of insertion sequences surrounding the system.

The DRs of every CRISPR found were aligned using Clustal Omega (Sievers and Higgins, 2014); then a phylogenetic inference was done using Mr. Bayes 3.2 (Ronquist et al., 2012). The alignment was also used to build sequence logos with Weblogo 3.0 (Crooks et al., 2004). Secondary structure of each DR was predicted by RNAFold (Gruber et al., 2008). Additionally, the origin of every spacer sequence was investigated by nucleotide BLAST against the NCBI’s database.

The sequences of the cas1 gene from all the CRISPR-positive genomes were aligned by Clustal Omega and the best conserved regions were analyzed to design the oligonucleotide pairs cas1IIF (5′-AAT ATA GAG GGC CAA GCG GC-3′)/cas1IIR (5′-CGC ATG CAG CAA GTT AAT CAG C-3′) to detect a 324 bp fragment of the cas1 gene from type II CRISPR systems, and cas1IIIF (5′-TGT TAC TGC GAA GGA AAA TAG C-3′)/cas1IIIR (5′-CGT CCA CGT TTA AAT TGT TTG CC-3′), to detect a 471 bp cas1 gene of type III CRISPR systems. Both cas1IIF/R and cas1IIIF/R primer pairs were used in a multiplex polymerase chain reaction (PCR) reaction containing 2 U of the GoTaq G2 DNA Polymerase (Promega, United States) in the enzyme buffer, containing 0.2 mM of each dNTP, 0.2 mM of each primer and 0.1 mM of the primer pair 16S rRNAF/R (Chen et al., 2013) as an endogenous control. Screening was performed in 59 clinical strains of Staphylococcus haemolyticus previously analyzed by our group and shown to be phenotypic and genetically diverse (Barros et al., 2012; Rossi et al., 2016b). Amplification was performed in a T100 Thermal Cycler (BioRad, United States) with the following cycle parameters: initial denaturation at 95°C for 3 min, followed by 30 cycles (95°C for 30 s, 50°C for 1 min, and 72°C for 1 min) and 72°C for 3 min. The reaction products were analyzed after electrophoresis in 1.0% agarose gel. Staphylococcus epidermidis RP62A (Gill et al., 2005) was used as a positive control for type III-CRISPR cas1 and S. epidermidis 10 L (Rossi et al., 2017); as a positive control for type II-CRISPR cas1.

The expression of cas1 was evaluated in triplicate by RT-PCR in Staphylococcus haemolyticus clinical strains positive for the gene. Overnight cultures grown on Brain Heart Infusion (BHI-BD, France) agar plates were used to prepare bacterial suspensions using a 0.5 McFarland turbidity standard. The bacterial suspensions were then diluted in 1:100 fresh BHI and incubated at 37°C under agitation (150 rpm) for 4 h (exponential growth), 16 h (stationary phase) and 48 h for the formation of biofilms in the walls of the glass flasks used for cultivation. Cells were then harvested and ressuspended in RNAlater (Thermo Fisher Scientific, United States) and kept at -20°C until RNA extraction. RNA extraction was performed with the RNAeasy mini kit (Qiagen, Germany) according to the manufacturer’s instructions, with cells being previously disrupted with Lysing Matrix Tubes B (MP Biomedicals, United States) in a Mini-beadbeater-24 (BioSpec, United States) in three rounds of 40 s. RNA quantity and quality were evaluated by spectrophotometry and gel electrophoresis, respectively. Then, RNA was treated with RQ1 RNAse-Free DNAse (Promega, United States), following the manufacturer’s instructions. Complete removal of DNA was evaluated by PCR with primers for the rpoB gene, rpoB-2643F and rpoB-3241R, with amplification parameters as suggested by the author of the primers (Drancourt and Raoult, 2002). cDNA first strand synthesis was performed with the DNAse-treated RNA, using the ImProm-II^TM Reverse Transcription System (Promega, United States). Then, a PCR reaction was conducted with the primer pair cas1IIIF/cas1IIIR, 1 U of GoTaq G2 DNA Polymerase (Promega, United States) in the enzyme buffer, containing 0.2 mM of each dNTP and 0.2 mM of each primer. Amplification was done as described above. The reaction products were analyzed after electrophoresis in 1.0% agarose gel. As negative controls, we used RNA that was subjected to the same reactions of cDNA synthesis, but without the addition of the Reverse Transcriptase.

A total of 122 public available genomes, belonging to 15 strains of coagulase negative staphylococci species that have already been described in the human microbiota (among other sources), were analyzed for the presence of CRISPR-like sequences by the CRISPRfinder software. Among them, 32 strains were suspicious of containing CRISPR/Cas systems (Supplementary Table S1). However, many of these sequences were too small (comprising three DRs and two spacers), and most of them are not adjacent to cas genes. Therefore, these repetitive sequences are not included in real CRISPR/Cas systems. Overall, only 18 (15%) strains, belonging to eight species (S. capitis, S. epidermidis, S. massiliensis, S. schleiferi, S. haemolyticus, S. lugdunensis, S. simulans, and S. warneri) contain CRISPR/Cas systems, characterized by 3 (S. lugdunensis M23590) to 37 (S. schleiferi 1360-13) DRs and an adjacent cluster of cas genes (Figure 1 and Supplementary Figure S1). The proportion is even lower in CoNS that act as important human pathogens: among the S. epidermidis strains, only 9% (3/34) of them presented characteristic CRISPR/Cas systems, and among S. haemolyticus; only 3% (1/32).

These systems are all complete and are either of type II, containing the cas1, cas2, and cas9 genes; or of type III, which also contains the universal cas1-2 genes, in addition to cas6, cas10, and csm2-6. Curiously, some systems (the type III ones from S. capitis CR03 and S. schleiferi TSCC54) contain repetitive sequences both up and downstream the cas-gene cluster, although one of them seems to be more active than the other, given the difference in the number of spacers presented. In addition, two strains contain two independent CRISPR/Cas systems in their genome: while the S. simulans FDAARGOS_124 strain contains two type II systems, the S. schleiferi TSCC54 strain contains one type II system (like all other strains from this species), but harbors an additional system, of type III.

The presence of a CRISPR/Cas system in S. schleiferi TSCC54 different from the one that is common for the species (Figure 1) lead us to investigate whether they could have had different origins. For that purpose, we first analyzed the DRs of each CRISPR found and made a Bayesian inference of the phylogeny of those sequences, which are usually conserved among species. Most DR formed groups in a species-dependent manner, in addition to defining two clades: one of type II CRISPR system DR, and other from type III (Figure 2A).

All the DR from type III systems have very conserved regions containing the motif CCCC, separated by eight nucleotides from a GGGG stretch. Both sequences are likely to interact with one another, to form the typical hairpin structures that are both important for the initial processing of the CRISPR transcript and the interaction of the CRISPR small RNA (crRNA, involved in the interference process) with the Cas proteins. Although the same motifs are not observed in the DR from type II systems, which are more varied, all the DR are also capable of forming the typical secondary structures (Figure 2B and Supplementary Figure S2).

The DR from S. schleiferi TSCC54 type III system grouped with the DR from S. capitis CR03, which raised the hypothesis of a horizontal transference between them. To investigate the location of CRISPR loci within mobile genetic elements, we first searched for insertion sequences in the vicinity of these elements. Many of the systems found in here are nearby insertion sequences (Supplementary Table S2), but an in depth analysis and annotation of genes around the CRISPR/Cas systems of S. schleiferi TSCC54 and S. capitis CR03 showed that they both are located within SCCmec elements (Figure 3A). These elements display an almost identical region (98% of identity) that is also very similar to a SCCmec previously described for the S. aureus strain 80BA02176 (Golding et al., 2012). Surprisingly, these three systems share most of their spacer sequences, evidencing a common origin (Figure 3B). Given the number and array of spacers, we hypothesize that this CRISPR/Cas element was originally from a S. capitis strain that diverged into two different strains (Figure 3C): (i) one that acquired the spacers S14 and S15, originating the S. capitis CR01 strain and (ii) another that acquired the spacers S16 and S18-S22, originating the strain that served as the donor of the CRISPR-containing SSCmec to S. aureus 80BA02176 (which incorporated the S17 spacer and lost S08-S09) and to S. schleiferi TSCC54 (which acquired S23-24 and lost S03-S04). It is also possible that other intermediate strains were involved in this process of SSCmec and CRISPR/Cas systems exchange process.

Considering that some spacers are identical between the S. capitis CR01 and S. schleiferi TSCC54 strains and that spacers were found repeatedly within the same species (among S. schleiferi strains) and even in the same CRISPR of a single strain (S. massiliensis CCUG55927), there are overall 194 unique spacer sequences. Among them, the great majority (93%) are from an unknown origin, as they do not present homology with any sequence available from GenBank. Only 13 (7%) sequences are known, of which seven (4%) are identical to a variety of plasmids found in different species belonging to the Staphylococcus genus (Supplementary Table S3). A spacer from S. haemolyticus W75 (shae20), for example, has a sequence that is identical to that of the S. aureus plasmid pSAM12-0145 (GenBank ID KU521355.1) and the S. cohnii plasmid pHK01 (GenBank ID KC820816.1). The remaining six (3%) spacers originally may have come from bacteriophages (Supplementary Table S3), some of which were isolated from other Staphylococcus species.

As well as having very distinct DR sequences, CRISPR systems from CoNS also present sequences for the cas1 gene that differ from on type to the other. A comparison between the cas1 gene sequences from all the strains analyzed in this work showed that when the sequences from both type II and type III cas1 are aligned together, they present an identity as low as 29% (data not shown). However, when the sequences are analyzed in their own separate CRISPR-type groups, this value increases to 54% for type II cas1, and to 70% for type III. These differences lead us to design primers to cas1 of each group separately, so that the amplification of that gene could be performed in a multiplex PCR reaction that allowed us not only to detect more efficiently strains harboring that CRISPR marker gene, but also permitted their differentiation according to the type of system (Figure 4). In addition, the reaction allows the identification of strains containing more than one type of CRISPR system in its genome, as in the case observed computationally for the S. schleiferi TSCC54 strain, described above.

Because S. haemolyticus is one of the most important CoNS and no CRISPRs have been described to the species so far, the multiplex PCR reaction was employed in 59 clinical strains of this species previously shown by our group to be phenotypic and genetically diverse (Rossi et al., 2016b). The cas1 gene was amplified in only 3% of the strains (2/59), namely MD03 and MD25. The proportion found is the same observed in the in silico analysis for the species. However, while in the genome available for S. haemolyticus W75 the CRISPR system belongs to the type II, in both of our clinical strains the amplicon size indicates they belong to type III (data not shown).

To get an insight on the functionality of the CRISPR systems found in the screening, we evaluated the expression of the cas1 gene in strains MD03 and MD05 during exponential growth, stationary phase and in the biofilm form. As expected for a surveillance-related gene, the expression of cas1 was observed in all situations (Figure 5).