Background

CRISPRs (clustered regularly interspaced short palindromic repeats) have been detected in approximately 85% of archaeal and 50% of bacterial genomes [1]. They are considered a means of prokaryotic adaptive immunity against bacteriophages [2], which are major determinants of prokaryotic abundance, diversity and community structure [3]. CRISPRs consist of conserved repeats separated by variable spacers, the latter representing incorporated fragments of viral or plasmid DNA that specify immunity upon subsequent encounters [4]. Many CRISPRs are associated with Cas gene arrays, which can be classified into three major types and ten subtypes [5, 6] and are considered essential for CRISPR function. The activity of CRISPR-Cas systems proceeds in three stages: the acquisition of protospacer sequences from foreign genetic elements and their integration into the CRISPR array, constitutive transcription of the array, and target interference through transcribed crRNA [2]. In response, phages have developed mechanisms to evade CRISPR action [7–9], suggesting a co-evolutionary arms race between bacteria and phages.

Comparative genomics has given insight into CRISPRs from Actinobacteria[10, 11], Firmicutes[12, 13], Cyanobacteria[14, 15], enterobacteria [16], and Archaea[17]. In addition, mathematical modeling has presented important concepts about CRISPR dynamics during phage-bacteria interactions [18, 19]. Most of what is known about CRISPRs has been derived from pathogenic or industrially relevant bacteria such as Salmonella[20] and Streptococcus[12]. In the case of environmental bacteria, it has been shown that CRISPRs are widespread in Cyanobacteria except for the major marine lineages Prochlorococcus and Synechococcus[15]. In freshwater Cyanobacteria, CRISPRs were used to illustrate specific host-cyanophage interactions [14]. Furthermore, CRISPRs have been linked to host-phage co-evolution, community structuring and biogeographic patterns in microbial mats [21], acidophilic biofilms [22], and hot spring microbiota [23].

CRISPRs also control genetic exchange [24, 25] and intraspecies recombination [26], hence mediating evolutionary processes [27]. They may also regulate gene expression via crRNAs [28] and ‘self-targeting spacers’ that match elements in the host genome [29]. CRISPR activity has also been linked to DNA repair [30] and can affect various bacterial phenotypes including biofilm formation [31], swarming motility [32], and pathogenicity [33]. Despite the insights afforded by these studies, the distribution, diversity and functional roles of CRISPR-Cas systems among closely related environmental bacteria remain largely unknown.

In the present study, we analyzed CRISPR-Cas and prophage content in 75 Salinispora strains from seven global collection sites. This actinomycete genus has a pan-tropical distribution in marine sediments [34, 35] and is comprised of three closely related species; the cosmopolitan S. arenicola and the regionally confined sister taxa S. pacifica and S. tropica[36, 37]. The species have been further divided into 16S rRNA phylotypes (i.e. single nucleotide variants), with the highest diversity in S. pacifica and the lowest in S. tropica[35]. The genus is recognized for the production of diverse secondary metabolites [38], with the associated biosynthetic pathways showing evidence of extensive horizontal gene transfer [39, 40].

The diversity and distribution of CRISPR-Cas systems in Salinispora spp. was investigated to (i) assess the role of CRISPRs in phage defense, (ii) characterize past interactions with foreign genetic elements, (iii) elucidate linkages between CRISPR features and Salinispora microdiversity, and (iv) identify biogeographic signatures in CRISPR and prophage content. The detected diversity of CRISPR-Cas systems, including spacers that match foreign genetic elements, supports a role in host immunity. Spacer arrays illustrated recurring encounters with related phages as well as geographically confined MGEs. These findings suggest the presence of complex CRISPR-mediated interactions between Salinispora spp. and foreign genetic elements that may influence the ecology and evolution of this broadly distributed marine actinomycete genus.

Results and discussion

CRISPR content in 75 Salinisporastrains

Genome sequences from 75 Salinispora strains derived from seven global collection sites were analyzed for CRISPR-Cas content (Additional file 1). In total, 335 CRISPR loci were detected, with an average of 4.4 per strain (Table 1) but considerable among strain variability, ranging between 1 and 12 (Additional file 1). Unlike many genera for which multiple genome sequences are available, all 75 Salinispora strains harbored CRISPRs, suggesting they are an ecologically relevant feature of this genus. Salinispora CRISPR content exceeded the average reported for mesophilic bacteria [1] and marine bacterial metagenomes [41] and accounted for up to 0.3% of some genomes, which is approximately a third of the reported ‘prokaryotic maximum’ [2]. CRISPRs were concentrated in genomic islands, which represent the major regions of gene acquisition in Salinispora spp. [40]. Prior evidence of extensive horizontal gene transfer in Salinispora spp. [39, 40], coupled with the ubiquity of CRISPR-Cas systems detected in the present study, suggests an ongoing dynamic between CRISPR-mediated immunity and the acquisition of foreign genetic material.

Table 1 Summary of CRISPR-Cas and prophage content in Salinispora spp.
Full size table

The 335 CRISPR loci contained 5737 spacers, of which 68% were observed only once across all genomes. Extensive differences in spacer content were detected among strains isolated at the same time from the same site (Additional file 1), suggesting that spatiotemporal encounters with mobile genetic elements (MGEs) may be highly variable. On average, S. arenicola and S. tropica contained more spacers per strain than S. pacifica, however, the numbers varied greatly among strains within each species (Table 1).

Diversity and evolution of Cas array subtypes

The 75 Salinispora strains contained 146 Cas arrays (Table 1), all of which can be classified as type I based on the inclusion of a cas3 gene [6]. Cas arrays could be further grouped into six subtypes (Figure 1), of which five occurred in all three species and one (I-U_Sa) was only observed in S. arenicola. In total, 60% of the CRISPRs were associated with Cas arrays (Table 1), with up to five different array subtypes in some strains (Additional file 1). Three of these subtypes (I-E, I-C, I-B) have been previously characterized [6], with the most common (I-E) occurring in 49 strains. Almost two-thirds of the I-E arrays were associated with paired loci, i.e., two CRISPRs (one with inverted repeat sequences) flanking internalized cas genes, as often observed in Archaea[42]. Eleven strains contained two I-E or I-C arrays (Additional file 1). BLAST analysis of the associated cas3 genes indicated that the two arrays in a given strain were acquired as independent events from different sources based on sequence similarities to homologs in different actinomycetes (Verrucosispora vs. Streptomyces spp. for I-E arrays and Frankia vs. Stackebrandtia spp. for I-C arrays). In all three species, the GC content of I-C arrays was lower than the overall genomic GC content. This was especially apparent in S. pacifica (64.2% vs. 69.8% GC), suggesting that I-C arrays have been acquired from distantly related taxa. To the best of our knowledge, three of the Cas array subtypes detected (herein designated as I-U_csb3, I-U_csx17 and I-U_Sa) have not previously been described despite containing known cas genes (csb1, csb2, csb3, csx17). These subtypes were designated as I-U based on convention [6]. However, the Integrated Microbial Genomes (IMG) database [43] revealed that a variety of bacteria from different phylogenetic groups possess equivalent arrays, indicating these are not unique to Salinispora spp.

Figure 1
figure 1

cas3 phylogeny and CRISPR repeat diversity. Condensed maximum likelihood phylogeny of cas3 nucleotide sequences reveals clades corresponding to Cas array subtype. The two major clades delineate known (I-E, I-C, I-B) and previously undescribed (I-U) subtypes. The order of genes for each subtype is displayed on the right. Gene annotations in parentheses designate hypothetical proteins with low identity to those indicated. In the I-B arrays, cas8b was replaced by a larger gene related to cst1. The total number of each array subtype among the 75 genomes is shown in the condensed nodes. Six arrays were missing several genes and therefore excluded from the analysis. Nodal support values (● above 80%, ▲ 100%) were obtained by 1000 bootstrap replicates (see Additional file 2 for the full tree including strain names and bootstrap values). Consensus repeat sequences in the associated CRISPR loci (indicated in gray) were specific to each array subtype and mostly showed considerable conservation.

Full size image

cas3 is the signature gene of type I arrays [6]. A cas3 phylogeny revealed clades that corresponded to Cas array subtype as opposed to taxonomic relationships (Figure 1). The finding of cas3 sequence similarities across species boundaries supports the concept that Cas arrays evolve independent of their hosts [20, 44]. Furthermore, sequences within the array subtypes reveal evidence of recombination, as different Salinispora species shared virtually identical cas3 genes. The same patterns were observed with cas1 genes and corresponding protein sequences (Additional file 2), the most common phylogenetic marker for CRISPR-Cas systems. The delineation of the Cas array subtypes was supported by the repeat sequences, which frequently shared subtype-specific conservation (Figure 1) and averaged between 29 nt (subtypes I-E and I-B) and 37 nt (subtypes I-C and I-U).

Cas-associated CRISPRs contained significantly more spacers than Cas-devoid loci (p < 1 × 10-10), as might be expected given that cas genes are required for spacer integration [2]. Furthermore, subtypes I-E, I-C and I-B contained significantly more spacers (p < 0.0001) than the three I-U subtypes. Considering the latter, I-U_Sa and I-U_csx17 lack cas1 and are thus potentially unable to incorporate additional spacers, as cas1 is involved in spacer integration [2].

CRISPRs illustrate interactions with foreign genetic elements

We assessed defensive functions of Salinispora CRISPRs by analyzing for perfect matches between Salinispora spacers and mobile genetic elements (MGEs). These included 97 prophages that were identified in the 75 genomes (Additional file 3) as well as MGEs deposited in the Aclame database [45] (the latter referred to as ‘known MGEs’). On average, 11% of spacers matched Salinispora prophages (Table 1). Prophage-devoid strains had a higher fraction of matching spacers than prophage-harboring strains (p < 0.05), supporting a functional role of CRISPRs in phage immunity. In addition, 1.1% of spacers matched known MGEs, which was comparable to observations for marine bacterial metagenomes [41] and oral pathogens [26]. Some spacers matched homologous elements from different viral genomes, suggesting they may target multiple phage strains. CRISPRTarget [46] revealed that MGEs matched by Salinispora spacers are associated with various protospacer-associated motifs (PAMs), short sequences important for protospacer acquisition [2]. This suggests that Salinispora spp. can detect different PAMs and integrate a large diversity of spacers. When including lower-quality matches (100% identity over at least 18 nt) the majority (77%) of spacers matched plasmids, suggesting that a major role for Salinispora CRISPRs is to defend against plasmid integration. As no information about the plasmid content of the strains investigated is currently available, we focused on the role of CRISPRs in phage defense, while realizing this may not present a complete picture of CRISPR functionality in Salinispora spp.

CRISPRs indicate differing defensive capacities among Salinisporaspecies

S. arenicola had four-fold more spacers matching Salinispora prophages and twelve-fold more spacers matching known MGEs than S. tropica. This corresponded to the fact that only two-third of S. arenicola but all S. tropica strains harbored prophages, with 1.3 vs. 2.1 prophages per genome, respectively (Table 1). A substantial number of S. arenicola spacers that matched Salinispora prophages were located in the I-U_Sa Cas arrays, which are specific to S. arenicola. This additional array and spacer diversity may provide superior defensive capacities for S. arenicola, which potentially contributes to its broader geographic distribution [35]. S. pacifica had an intermediate fraction of spacers matching Salinispora prophages and known MGEs, with 1.1 prophages per genome (Table 1). There was a significantly lower frequency of prophages among phylotype C and F strains (p < 0.01). While these phylotypes also contained significantly more spacers (p < 0.01), the fraction of those spacers matching Salinispora prophages and known MGEs was similar to other phylotypes. The differing phage sensitivities between S. pacifica phylotypes are thus independent from or only partially related to CRISPRs.

In contrast, the total numbers of CRISPR loci or spacers were uncorrelated with prophage content in all three species (R2 < 0.01). For instance, strains CNS-051 and CNS-205 contained 11 and 8 CRISPRs with 119 and 140 spacers, respectively. Despite these similarities, these strains harbored 0 and 5 prophages, respectively (Additional file 1). The number and diversity of Cas arrays were also uncorrelated with prophage content (R2 < 0.001). For instance, S. pacifica strain DSM-45549 contained four Cas array subtypes and three prophages while the Cas-devoid S. pacifica strain CNS-103 only contained one prophage (Additional file 1). Thus, the number of CRISPR loci as well as the diversity of associated Cas arrays appear to be affected by factors other than phage exposure.

History of Salinisporainteractions with a common prophage

We focused on a common prophage that is related to the Streptomyces SV1 phage and was detected in 24 Salinispora strains from all three species (Figure 2A, Additional file 3). Six percent of Salinispora spacers matched SV1-related sequences, suggesting that this phage represents a major challenge to the genus. Strains without an integrated SV1 prophage had a larger fraction of spacers matching SV1 in Cas-associated loci (89%) compared to those with an integrated SV1 prophage (75%), supporting a specific targeting of this phage. The history of encounters with SV1-related phages was determined for six strains per species (three harboring and three lacking SV1) by analyzing the location of matching spacers within spacer arrays according to the concept that ancestral spacers are commonly located at the ‘trailer’ end and more recent spacers at the ‘leader’ end of a spacer array [4]. Matching spacers, the majority with unique sequences, were detected throughout the spacer arrays (Figure 2B) suggesting recurring encounters with SV1-related phages over time. Given that the SV1 phage represents a vector for genetic exchange [47], it is interesting to speculate that it may represent a source of beneficial genetic material in addition to a survival challenge.

Figure 2
figure 2

Streptomyces SV1-related prophage content in Salinispora spp. A) Comparison of genes from six intact, SV1-related prophages in Salinispora genomes with the SV1 type phage (gray; with GenBank Gene IDs in parentheses) by color-coding percent sequence similarities over a 12400 bp region within SV1. B) History of SV1 encounters in S. tropica CNR-699 as reflected in multiple spacer matches in a single array. Incomplete matches are indicated by asterisks, perfect matches by encircled asterisks with sequences shown above.

Full size image

Linkages of CRISPR-Cas features with microdiversity

Salinispora microdiversity on the subspecies level has been defined based on 16S rRNA phylotypes (Additional file 1) and a multilocus phylogeny (Additional file 4). We detected several correlations between CRISPR-Cas features and microdiversity. For instance, one well-supported S. pacifica lineage contained the only strains (CNT-796 and CNT-851) with a modified I-C array lacking cas1/cas2, suggesting these genes have been lost in this lineage. Another S. pacifica lineage (containing strains CNQ-768 and CNS-103) was unique in being entirely devoid of cas genes. Also, certain clades were characterized by chromosomal relocations of CRISPR-Cas systems, as seen with I-E arrays in S. pacifica (strains CNT-796 and CNT-851) and S. tropica (strains CNS-197 and CNR-699).

The most distinct linkages were observed among the four S. arenicola phylotype B strains, which contained significantly more CRISPRs and spacers than strains from S. arenicola phylotypes A and ST (p < 0.05). Many spacers were unique to phylotype B, underlining that spacer composition can reflect population structure and evolutionary relationships [48, 49]. CRISPR characteristics not only distinguished phylotype B from other phylotypes, but also the two subclades within phylotype B (strains CNH-941 and CNP-193 vs. CNH-964 and CNP-105; Additional file 4). For instance, a paired CRISPR locus and flanking genes were inverted in one of the subclades (Additional file 5). Furthermore, there were subclade-specific differences in spacer content. While multiple spacers were shared by all phylotype B strains, which is consistent with observations among other closely related bacteria [46], spacer array alignments revealed three sets of spacers that were specific to one of the subclades (Figure 3). This probably illustrates subclade-specific deletions or insertions of whole spacer groups [49]. Sixty-five percent of the group 1 spacers in CNH-941 and CNP-193 matched plasmids from Alphaproteobacteria, while the group 2 spacers in CNH-964 and CNP-105 equally matched phages and largely gammaproteobacterial plasmids. This may coincide with differing defensive capacities or varying modes of interaction with MGEs between the two subclades. While prophage content appeared independent of these observations (Additional file 2) MGEs are also involved in diversification [50, 51], niche adaptation [52], and microdiversity [53]. It is hence interesting to speculate that these differences may influence the evolutionary or ecological divergence within S. arenicola phylotype B.

Figure 3
figure 3

Subclade-specific spacer content in S. arenicola phylotype B. Alignment of spacers from a homologous CRISPR locus present in all four S. arenicola phylotype B strains. The phylogenetic relationships among the four strains are depicted on the right. Spacers are indicated by rectangles arranged from the oldest (trailer end; right) to the most recent spacers (leader end; left). Vertically aligned spacers are identical and separated by subclade-specific groups of spacers (1, 2, 3), which were likely deleted from the respective other lineage or selectively acquired.

Full size image

Biogeographic patterns in CRISPR and prophage content

The strains analyzed in this study originate from seven global collection sites and were derived from independent sediment samples. While sampling efforts were not uniform across locations and may have affected the biogeographic patterns observed, it is interesting to note that 40% of the spacers observed in more than one strain were restricted to specific locations and/or biomes, the latter describing major oceanic regions distinguished by oceanographic factors such as nutrient concentrations and primary productivity [54]. Location-specific spacers provide evidence of exposure to local virus populations [41, 55], with the majority of localized spacers occurring in strains from the Sea of Cortez (Figure 4A). This is a highly productive sea [56] enclosed by a distinct geographical barrier and the only site classified as a Coastal biome [54]. While these results are preliminary, it is intriguing to speculate that spacer sequences can be used to trace location-specific interactions with distinct MGE pools, as also observed in other ecosystems [23, 57, 58]. A more nuanced biogeographic pattern was the detection of identical spacers with location-specific nucleotide substitutions, as found in strains from Hawaii, Fiji and Palau (Figure 4B). This may illustrate the presence of widespread MGEs that maintain location-specific genetic variants. Furthermore, SV1-related prophages could be resolved into geographically confined lineages (Figure 4C), supporting the concept that Salinispora strains are exposed to location-specific MGEs.

Figure 4
figure 4

Biogeographic patterns among Salinispora spacers and prophages. A) Distribution of spacers unique to certain locations (SC = Sea of Cortez, HW = Hawaii, FJ = Fiji, PY = Palmyra, BH = Bahamas, PA = Palau) and biomes; B) Location-specific spacer variants reflected in single nucleotide polymorphisms (marked in red) within a conserved 41 nt spacer in strains from Hawaii (3 strains with 3 variants), Fiji (4 strains with 1 variant) and Palau (1 strains with 1 variant); C) Maximum likelihood phylogeny (1000 bootstrap replicates) of SV1-related prophages in genomes from geographically distant Salinispora strains, showing location-specific SV1 lineages.

Full size image

Self-targeting spacers

Several studies have reported the occurrence of ‘self-targeting spacers’ that match regions within the host genome [29]. While self-targeting spacers can be deleterious and strongly selected against [29, 59], they have also been suggested to function as regulatory elements [20, 60, 61] or to affect genome content [62]. Interestingly, a third of the 75 Salinispora strains harbored such spacers, with perfect matches to e.g. a cytochrome P450 within a terpenoid biosynthetic pathway [40] and two DNA-modifying genes (Table 2). However, experimental evidence would be required to determine potential regulatory roles. In addition, several self-targeting spacers matched resident prophages, suggesting that CRISPR interference may be ineffective in some cases. Alternatively, self-targeting may be prevented by selective self vs. non-self mechanisms, such as variations in spacer flanking sequences [63].

Table 2 Chromosomal matches of select self-targeting spacers
Full size table

Conclusions

This study describes a comprehensive survey of CRISPR-Cas systems among a large collection of strains from a cosmopolitan marine actinomycete genus. The finding of ubiquitous and diverse CRISPR-Cas systems suggests that Salinispora maintains a robust mechanism to mediate interactions with MGEs, which may be of ecological and evolutionary relevance in virally rich marine sediments [3]. Future surveys of CRISPR-Cas systems will provide additional opportunities to assess the evolutionary history of MGE exposure, the effectiveness of these systems as mechanisms of adaptive defense, and how CRISPRs may be linked to the ecology and evolution of Salinispora.

Methods

Genome sequences and CRISPR-Cas classification

The 75 Salinispora genome sequences (Additional file 1) were downloaded from the IMG database (https://img.jgi.doe.gov). CRISPRs were predicted using CRISPRFinder [64] on pseudochromosomes generated from the genome sequences (i.e. contigs assembled using a closed reference genome) [39] and unmapped contigs. Only CRISPRs classified as ‘confirmed’ were considered. Predicted CRISPRs were manually checked and adjacent loci combined if separated by Ns and having the same repeat sequences. Annotated cas genes were verified by determining similarities to known cas genes using BLAST [65] and UniProt [66]. The naming of cas genes and their classification into Cas array subtypes was done following [6]. The IMG database was searched for equivalent Cas arrays in other sequenced bacterial genomes. CRISPRmap was used to classify repeats into motifs, families, and superclasses based on similarities to known repeat sequences [67]. Repeat consensus sequences were obtained using WebLogo [68].

Phylogenetic and structural analyses of Cas arrays

cas1 and cas3 nucleotide and corresponding Cas1 and Cas3 amino acid sequences were aligned using MAFFT v7.017 (L-INS-i algorithm, 100PAM/k = 2 scoring matrix, gap open penalty 1.53, offset value 0.123) [69] and manually curated. The best substitution models (cas1: TN93 + G + I; Cas1: WAG + G + F; cas3: T92 + G; Cas3: JTT + G) were determined using MEGA5 [70]. Maximum likelihood phylogenies were computed with MEGA5 (using the best model and 100 bootstrap replicates) and RAxML (with default settings and 1000 bootstrap replicates) implemented on the CIPRES Science Portal [71], always giving the same topology. Nucleotide sequences of cas1 [KM526976-KM527070] and cas3 [KJ677987-KJ678124] have been deposited at GenBank (Additional file 6). Architectures of selected loci and flanking regions were analyzed with progressiveMauve [72]. Spacer arrangement in S. arenicola phylotype B was evaluated by aligning concatenated spacer sequences (sorted from trailer to leader end) with MAFFT [69].

Prophage prediction and sequence comparison

Prophages were predicted using PHAST [73] on both the pseudochromosomes and unmapped contigs. Predicted intact prophages classified as being related to the Streptomyces SV1 phage were compared with the sequenced SV1 type phage (GenBank accession number NC_018848) using the CGView Comparison Tool [74]. Nucleotide sequences of SV1-related prophages were aligned using Mugsy [75] and the resulting alignment file converted to Fasta using the Galaxy web server [76]. The alignment was manually curated and the best substitution model (GTR + G) determined using MEGA5 [70]. A maximum likelihood phylogeny was computed using MEGA5 with 1000 bootstrap replicates (Additional file 7).

Analysis of spacers

Spacers were extracted from genome sequences and sorted by unique (only found once across all 75 genomes) and shared (found in ≥2 genomes). Spacers were searched against different databases (Aclame MGE_0.4, PHAST_virus, PHAST_prophage_virus, CRISPRFinder spacer) with the standard BLAST parameters for short query sequences (word size 7; match/mismatch scores 1,-3; gap costs 5,2) using Geneious Pro v5.5 (available from http://geneious.com). In addition, short-query BLAST was used to determine spacers matching Salinispora prophages as well as self-targeting spacers matching non-CRISPR regions. Furthermore, short-query BLAST against Salinispora prophages was done with spacers from five representative strains from each species that were sorted by Cas-associated, Cas-devoid, associated with known Cas array subtypes (I-E, I-C, I-B), and associated with herein designated Cas array subtypes (I-U). Only perfect matches with 100% identity over the entire spacer length were considered. A separate BLAST search against Aclame was performed which also considered incomplete hits (100% sequence identity over at least 18 nt), as this may still be indicative of the targeted MGE type. The 18 nt threshold corresponds to 2/3 of the average Salinispora repeat length, which has been suggested as the minimum for a functioning spacer [1]. Also, 100% coverage hits are possibly rare since the vast majority of phage diversity is likely still unknown [3].

Statistical evaluation

The number of CRISPR loci, prophages and MGE genes per strain were normalized by genome size and gene count, respectively. Values were compared by species, location, and biome (both between and within species) as well as phylotype (only within species) using the Kruskal-Wallis one-way analysis of variance implemented in R [77] to test for significant differences. In case of a significant result (p < 0.05) the Wilcoxon rank-sum test implemented in R [77] was used to test the specific sample pairs for significant differences (p < 0.05). The fraction of spacers matching Salinispora prophages in strains with and without prophages was compared using Student’s t-test. Correlations between the number of CRISPRs/spacers/Cas arrays and prophages were calculated using least squares regression.

Availability of supporting data

All supporting data are included within the article and its additional files.