Presence and Variety of CRISPR-Cas Immune Systems in Bifidobacterium Bacteria

The genomes used were previously described by Milani et al. [19] (Table 1). CRISPR analyses were performed using CRISPRfinder followed by manual curation to identify and assemble repeat-spacer arrays [32]. BLAST [33] was used to identify Cas proteins and confirm subtype designation [13]. SnapGene (GSL Biotech) was used to annotate cas genes and generate the operon layout. The Cas1 alignment was performed using the MUSCLE alignment algorithm; the consensus tree was constructed in MEGA version 6 [34] using the UPGMA Jukes-Cantor model calculating bootstrap values with 500 samples. Protospacers were identified using BLASTn against publically available metagenomic datasets and a locally created bifidobacterium database. A “strong” protospacer match was considered when two examined sequences exhibited more than 90% identity over 90% of the spacer sequence length. Additional analyses were performed to detect similarity between CRISPR spacers and prophage sequences located within the bifidobacterium genomes used in this study. Once potential protospacers were identified, the flanking regions were aligned by hand and analyzed by WebLogo to determine the protospacer adjacent motif (PAM) [35]. The tracrRNA sequences for Bifidobacterium bombi, Bifidobacterium merycicum, and B. bifidum were predicted manually using BLAST to identify anti-repeat region in the respective genomes per the modules previously defined by Briner et al. [36]. NUPACK was used to generate RNA folding prediction images [37].

Table 1 | Occurrence of CRISPR-Cas systems in bifidobacteria | Each strain used in this analysis was examined for CRISPR-Cas systems. “System Type” refers to the type and subtype of the predicted system. The sequence, repeat length, and number of repeats found in a single repeat-spacer array are given for each system. When repeat-spacer arrays were detected but no cas genes were annotated, the system type was designated “Undetermined.” “None” indicates that repeat-spacer arrays and cas genes were not found in the genome. Cas indicated the presence or absence of cas genes, with a “Y” indicating the gene is present and an “N” indicating the gene is absent. Type designation was confirmed by the presence of cas3, cas9, and cas10 for Type I, II, and III respectively.

Among the 48 genomes analyzed, we observed a high rate of occurrence of CRISPR-Cas systems in the genus Bifidobacterium (77%) (Table 1), compared to the estimated prevalence in bacteria (45%, 1176/2612, as of the August 2014 update) [32]. In the 48 genomes analyzed, representing 43 species, we identified 26 Type I systems, and six Type II systems, while Type III systems appeared to be absent, according to the cas gene content and CRISPR repeat length and sequence. Interestingly, four strains contained multiple distinct CRISPR-Cas systems, including B. dentium which contained three separate systems. Each Bifidobacterium species containing a CRISPR-Cas system had a unique CRISPR repeat and cas1 gene, suggesting that these systems are phylogenetically different from one another. Strains that contained repeat-spacer arrays but did not contain cas genes were categorized as “Undetermined”; given the repetitive nature of CRISPR loci and the draft genome status of many genomes, there are possibly additional loci that could not be identified or classified. Bifidobacterium moukalabense was also defined as “undetermined” because it only contained the repeat-spacer array and cas1 and cas2 genes; no other cas genes were found within this locus. The high rate of occurrence and occasional presence of multiple systems in a single genome suggests that bifidobacteria possess active CRISPR immune systems that are important in the defense against potentially damaging foreign DNA.

A phylogenetic analysis was performed to investigate the level of conservation and divergence between the identified bifidobacterial CRISPR immune systems using the universal Cas1 protein. The Cas1 tree in Fig 1 revealed five major clusters that correspond to different types and subtypes of CRISPR systems, namely a branch for Type I-E, I-C, I-U, Type II-A and II-C Cas1 proteins. Interestingly, the Type I-E branch appears to diverge further into two distinct sub-clusters that correspond to similar locus architectures but distinct cas genetic sequences (Fig 2). B. moukalabense was most likely a I-C system before loss of the other cas genes because it grouped most closely with the I-C branch. Finally, Type I-U systems appear to be the most distantly related Cas1s, diverging very early from the other Cas1 subtypes, including other Type I systems.

Type I systems, including the I-E, I-C, and I-U subtypes, were the most commonly encountered bifidobacterial CRISPR-Cas system, occurring at a rate of 54% (26/48), consistent with this system being also most prevalent in bacteria [13]. Ten systems were categorized as Type I-C based on a similar gene content and layout as B. dentium (Fig 2). Visible on the Cas1 tree (Fig 1), 12 species were split between two distinct Type I-E locus architectures: one cluster included B. animalis subsp. animalis and the other cluster included Bifidobacterium angulatum, with the main differentiating feature being the sequence of the cas genes, including the signature cas3 and the universal cas1 gene (Fig 2). Type I-U systems are very unique in that they contain a cas4/cas1 fusion gene. Three of the Type I-U systems matched this typical gene layout, as seen in Bifidobacterium thermophilum, whereas B. animalis subsp. lactis appears to have undergone a unique genomic rearrangement event of its Type I-U CRISPR-Cas system causing a reversed orientation of all relevant genes, thereby constituting a unique layout.

Type II systems, which are fairly rare in nature only occurring in 5% of bacteria [38], occur at a much higher frequency in the genus Bifidobacterium; Type II systems were found in six of the 48 genomes analyzed (12.5%), with two Type II-A systems and four Type II-C. The II-A cas9’s canonical arrangement and content are longer than the II-C cas9s; the II-A B. merycicum cas9 is approximately 900 nucleotides longer than the Type II-C cas9 in B. bombi. The predicted tracrRNA is located between cas9 and cas1 on the minus strand in B. merycicum, while the tracrRNA in B. bombi, a II-C system, is upstream of the 5’ end of cas9 on the positive strand.

The size of the repeat-spacer arrays found in bifidobacteria is highly diverse, and ranging from 2 to 171 spacers (Fig 3, Table 1). Type I-E systems average 80 spacers per locus with some loci encoding as many as 171 unique genetic immunization events. Type I-C systems are, on average, smaller than Type I-E, averaging 50 spacers per locus, but can reach as many as 155 spacers. Type I-U systems average 19 spacers per locus with a range of 6 to 52 spacers. The identified bifidobacterial Type II-C systems contain, on average, the smallest repeat-spacer arrays (14 spacers). Because each unique spacer is acquired from invasive foreign genetic elements like phages and plasmids, we hypothesize that Type I-E systems are the most active loci as evidenced by their large size and highest rate of occurrence in the genomes. Such large arrays found in bifidobacteria also suggest these bacteria live in an environment rich in exogenic DNA requiring the need for an active defense system. Furthermore, such large repeat-spacer arrays could provide a locus with high genotyping potential. The hypervariable spacer sequences can be used to infer strain divergence and ecological relationships between different strains [7, 39].

By investigating the foreign DNA source each spacer targets, insights can be gained as to the functional role of CRISPR-Cas systems. A protospacer is the DNA sequence homolog in a phage, plasmid, or chromosomal DNA that matches the spacer sequence stored in a CRISPR locus. Several matches to publically available metagenomic data and matches to prophage sequences in a local Bifidobacterium database (S1 Table), reveal that spacers in different genomes have homology to predicted (pro)phage genomes indicating bifidobacteria have acquired immunity against such (pro)phages (Fig 4).

The availability of 48 genome sequences of recognized bifidobacterial taxa [23, 29], allowed the identification of bifidobacterial prophage protospacers matching the spacer sequences of CRISPR loci identified in this study. Furthermore, as shown in S1 Table, investigation of flanking regions displaying a DNA similarity with the spacer sequences greater than 91% (e-value less that 0.0001), allowed us to precisely map their homologous region within the prophage regions identified in these bifidobacterial genomes.

The genomes of B. angulatum, B. adolescentis and Bifidobacterium minimum contain the most spacer sequences that target predicted prophages in bifidobacterial genomes (Fig 4). B. angulatum possesses 26 distinct spacers that target presumed prophage sequences in 11 different Bifidobacterium species. Conversely, the prophage sequences harbored in the genomes of Bifidobacterium stercoris and B. moukalabense taxa are potential targets of 14 and 13 CRISPR spacers, respectively. CRISPR systems that were shown to contain more than 10 spacers matching prophages in other genomes, did not contain prophages targeted by any other CRISPR systems. Conversely, genomes without CRISPR-Cas systems or that contained apparently degenerated CRISPR-Cas systems were more likely to have prophages that were targeted by other CRISPR systems. We speculate that one reason prophages may have been able to integrate into host chromosomes is the lack of an interfering CRISPR-Cas system. There is a correlation between genomes targeted by CRISPR spacers and absence of functional cas genes. However, no one type of CRISPR-Cas systems was correlated with higher numbers of spacers targeting prophages.

CRISPR-Cas systems defined as “undetermined” and degenerate rarely had spacer sequences that target prophages; only B. longum subsp. suis had four spacer sequences that showed homology to prophage sequences (Fig 4). Of the 32 characterized and putative functional CRISPR-Cas systems in the genus Bifidobacterium, 25 had at least one spacer sequence that targeted a prophage. The most frequently targeted prophage-containing species are B. longum subsp. suis, Bifidobacterium catenulatum and Bifidobacterium mongoliense which range from 11 to 12 protospacer targets in their prophage regions (Fig 4). Interestingly, Bifidobacterium actinocoloniiforme and Bifidobacterium cuniculi display some spacer matches against their own prophage sequences and none against other bifidobacterial taxa. Self-targeting spacers suggest that those CRISPR-Cas systems may be inactive to prevent Cas protein cleavage of the chromosome, as self-targeting is known to be a lethal event [40].

To fully characterize the Type II CRISPR-Cas systems in three Bifidobacterium genomes, we identified the tracrRNA sequence, predicted the structure of the crRNA:tracrRNA duplex, and investigated the PAM sequence targeted by Cas9 by identifying protospacers for the spacer sequences in these genomes (Figs 5 and 6, S1 Table). Three representative organisms were selected for further characterization: B. merycicum and B. bifidum which contain Type II-A systems and B. bombi which contains a II-C system.

The tracrRNAs predicted for the three organisms are all congruent with conserved tracrRNA: crRNA structures as previously published in the literature [36, 38], and contain a double stemmed nexus previously only seen in lactobacilli. Combining preliminary RNASeq analyses (S1 Text) and predicting RNase III processing sites as defined by Pertzev and Nicholson, the boundaries of the crRNA suggest that the 5’ processing site occurs either just below the bulge at the top of the lower stem, or after the first nucleotide of the upper stem, in the crRNA:tracrRNA modules previously defined by Briner et al., 2014 [36, 41, 42] (Figs 5 and 6, S1 Fig). For both II-A systems, B. bifidum and B. merycicum, the crRNA:tracrRNA duplexes appear to have one processing site in the middle of the upper stem and a second processing site one to two nucleotides after the bulge; these similar processing sites are likely due to similarity of crRNA and tracrRNA sequences [36, 41, 42].

Boundaries for the crRNA and tracrRNA molecules are important for characterizing how Type II CRISPR-Cas systems function, but the protospacer adjacent motif (PAM) is another element that is important for functionality; the PAM is a recognition motif used by the Cas proteins to bind foreign DNA (Fig 5) [10–11,43]. In Type II-A systems, Cas9 first binds to the PAM and then interrogates the DNA from 3’ to 5’ [44]. The PAM identified for B. merycicum, 5’-NGG-3’, is the same as the well-characterized Streptococus pyogenes PAM, and flanks the 3’ edge of the protospacer (Figs 4 and 5, S1 Table). Interestingly, the PAM identified for B. bombi, a Type II-C system, is predicted to be a single nucleotide 5’-NNG-3’ flanking the 3’ end of protospacers. Altogether, using the PAM, tracrRNA, CRISPR repeat, and novel Cas9, Type II systems can be exploited to develop new genome editing tools [17, 19].