Catalytically active Cas9 mediates transcriptional interference to facilitate bacterial virulence

Summary

In addition to defense against foreign DNA, the CRISPR-Cas9 system of Francisella novicida represses expression of an endogenous immunostimulatory lipoprotein. We investigated the specificity and molecular mechanism of this regulation, demonstrating that Cas9 has a highly specific regulon of four genes which must be repressed for bacterial virulence. Regulation occurs through a PAM-dependent interaction of Cas9 with its endogenous DNA targets, dependent on a non-canonical small RNA (scaRNA) and tracrRNA. The limited complementarity between scaRNA and the endogenous DNA targets precludes cleavage, highlighting the evolution of scaRNA to repress transcription without lethally targeting the chromosome. We show that scaRNA can be reprogrammed to repress other genes, and with engineered, extended complementarity to an exogenous target, the repurposed scaRNA:tracrRNA-FnoCas9 machinery can also direct DNA cleavage. Natural Cas9 transcriptional interference likely represents a broad paradigm of regulatory functionality, which is potentially critical to the physiology of numerous Cas9-encoding pathogenic and commensal organisms.

Graphical Abstract

eTOC blurb

An eTOC blurb should also be included that is no longer than 50 words describing the context and significance of the findings for the broader journal readership. When writing this paragraph, please target it to non-specialists by highlighting the major conceptual point of the paper in plain language, without extensive experimental detail. The blurb must be written in the third person and refer to “First Author et al.”

Introduction

CRISPR-Cas systems are prokaryotic adaptive immune systems that restrict infection by potentially harmful foreign genetic elements (Barrangou et al., 2007; Marraffini and Sontheimer, 2008). CRISPR-Cas9 systems use the single effector protein Cas9 for target recognition and cleavage (Gasiunas et al., 2012; Jinek et al., 2012). Cas9 forms a complex with a duplex of small RNAs, one being a crRNA that is transcribed and processed from a genomic CRISPR (clustered regularly interspaced short palindromic repeats) array (Deltcheva et al., 2011). The other is tracrRNA, a small RNA transcribed from the CRISPR-Cas9 locus that contains an inverted sequence with complementarity to the repeat sequence conserved in each crRNA derived from a CRISPR array (Deltcheva et al., 2011). Another portion of the crRNA, the spacer sequence, is often complementary to an exogenous DNA target (Bolotin et al., 2005; Mojica et al., 2005; Pourcel et al., 2005). Upon infection with a nucleic acid threat, the crRNA spacer binds to the complementary sequence on the incoming DNA (the protospacer), leading Cas9 to cleave the DNA target, resulting in a double strand break (Gasiunas et al., 2012; Jinek et al., 2012).

Before the crRNA spacer can interact with the complementary sequence on the DNA target, the Cas9 complex must first recognize a short nucleotide sequence on the opposite strand and adjacent to the protospacer, called a protospacer adjacent motif (PAM) (Anders et al., 2014; Jinek et al., 2012; Jinek et al., 2014; Nishimasu et al., 2014; Sapranauskas et al., 2011). This stage of target recognition is necessary to prevent cleavage of the genomic CRISPR array from which the crRNAs are transcribed (Gasiunas et al., 2012; Jinek et al., 2012). There are no PAM sequences next to the crRNA spacers in the genome (Marraffini and Sontheimer, 2010; Mojica et al., 2009).

Interestingly, CRISPR-Cas systems can have additional roles in bacterial physiology that extend beyond defense against foreign nucleic acids (Louwen et al., 2014; Ratner et al., 2015; Westra et al., 2014). Of particular interest are type II CRISPR-Cas systems, which include CRISPR-Cas9, because they are uniquely abundant in pathogenic and commensal organisms (Chylinski et al., 2013; Fonfara et al., 2014; Sampson et al., 2013). Mutants lacking cas9 in Streptococcus agalactiae, Campylobacter jejuni, Neisseria meningitidis, and Francisella novicida are impaired in virulence processes such as attachment and invasion of host cells and intracellular survival (Louwen et al., 2013; Ma et al., 2018; Sampson et al., 2013). These defects correspond to reduced pathogenicity of cas9 mutants in S. agalactiae, C. jejuni, and F. novicida during infection (Louwen et al., 2013; Ma et al., 2018; Sampson et al., 2013).

In F. novicida, which can cause human infections, the attenuation of the cas9 mutant is due in part to F. novicida Cas9 (FnoCas9) regulation of the expression of an endogenous mRNA encoding a bacterial lipoprotein (BLP), FTN_1103 (1103) (Jones et al., 2012). Regulation of 1103 by FnoCas9 is controlled by tracrRNA and scaRNA, which is a distinct small RNA transcribed from an independent promoter near the CRISPR locus (Chylinski et al., 2013; Postic et al., 2010; Sampson et al., 2013). Interestingly, FnoCas9, tracrRNA, and scaRNA together enable robust repression of 1103 transcript levels, which occurs in the presence and absence of crRNAs (Sampson et al., 2013). Since BLPs are ligands for mammalian innate immune proteins, repression of 1103 helps facilitate the evasion of these sensors by F. novicida (Jones et al., 2012; Sampson et al., 2014; Sampson et al., 2013). However, reduction of 1103 levels alone is not sufficient to completely restore the virulence of a cas9 mutant, suggesting that additional factors are involved (Jones et al., 2012; Sampson et al., 2013).

To comprehensively characterize the endogenous regulatory role of FnoCas9, we performed a genome-wide expression analysis which revealed that FnoCas9 has a specific regulon of just two transcripts encoding four genes, including 1103. Regulation is PAM-dependent and uses catalytically active FnoCas9 and two RNAs, scaRNA and tracrRNA, that likely form an RNA duplex (scaRNA:tracrRNA) that is distinct from the duplex used for targeting foreign DNA (crRNA:tracrRNA). scaRNA is complementary to the template strand of the 5’ UTRs of the two transcripts, thus targeting FnoCas9 to specific sites on the endogenous genomic DNA to repress transcript levels. Repression of all four genes contributes to the virulence of F. novicida. These findings show for the first time that a cleavage-competent Cas9 complex can exist in two distinct states in the bacterium to mediate two different functions: binding to endogenous DNA as a transcriptional repressor and cleaving foreign DNA to prevent infection. We further demonstrate that the scaRNA can be reprogrammed to guide FnoCas9 to repress other genes in F. novicida, highlighting the potential utility of this system in the control of gene expression. Taken together, these findings likely represent a broad paradigm in the way CRISPR-Cas9 systems mediate non-canonical functions that are distinct from DNA cleavage and contribute to bacterial physiology.

Results

FnoCas9 has a highly specific regulon

To identify regulatory targets of FnoCas9, we performed a genome-wide analysis of mRNA levels in a cas9 deletion mutant compared to wild-type (WT) F. novicida (excluding small RNAs). Only 4 out of 1,782 genes (0.22%) were significantly up-regulated in the cas9 mutant by more than 2-fold, including FTN_1103 (1103) encoding a bacterial lipoprotein (BLP), which we previously demonstrated to be regulated by FnoCas9 (Figure 1A, S1A, C-F) (Jones et al., 2012; Sampson et al., 2013). Similar experiments in tracrRNA and scaRNA deletion strains revealed the identical regulon of only four genes (1103; FTN_1104, “1104”; FTN_1102, “1102”; FTN_1101, “1101”) (Figure 1A, S1A). These results, validated by Northern blots of 1104-1101 in Δcas9, ΔtracrRNA, ΔscaRNA, ΔcrRNA, and the complemented strains, indicated that similar to 1103, repression of 1104, 1102, and 1101 transcripts was dependent on FnoCas9, tracrRNA, and scaRNA (Figure 1C-F, S1A). The upregulation of the 1104-1101 transcripts upon deletion of FnoCas9, scaRNA, or tracrRNA did not occur in the crRNA mutant (Figure 1C-F, S1A). Interestingly, the four genes are encoded at the same genomic region (Figure 1B, S1A). Northern blots (Figure 1C-F) and PCR amplification over the gene junctions between 1104-1101 (Figure S1B) revealed that they are encoded on two separate transcripts, one containing the operon 1104-1102 and another with 1101 (Figure 1B-F). These data highlight that the FnoCas9 machinery has a regulon that is highly specific to a region of the genome.

Figure 1. FnoCas9, scaRNA, and tracrRNA regulate transcript levels in a specific genomic region of F. novicida.

A) Differential transcript expression in the indicated F. novicida deletion mutants (Δcas9, ΔtracrRNA, and ΔscaRNA) relative to expression in wild-type (WT). Table represents all genes with a log fold change (log2FC) > 1 and an adjusted p-value (padj) <0.05 in each strain compared to WT. The names used for these genes throughout the paper are 1104, 1103, 1102, and 1101. The genome of F. novicida has recently been re-annotated in NCBI and the new locus tags are indicated. B) Schematic of the chromosomal locus encoding the four FnoCas9-regulated genes (1104, 1103, 1102, and 1101) and the two mRNA products of the locus. C-F) Northern blots for (C) 1104, (D) 1103, (E) 1102, (F) 1101, in wild-type (WT) and deletion mutants for each component of the F. novicida CRISPR-Cas9 system (Δcas9, ΔtracrRNA, ΔscaRNA, ΔcrRNA) and their respective complementation strains.

FnoCas9 represses transcript levels by targeting the 5’ UTR of target genes

We next investigated whether the two repressed transcripts contained regions that could serve as targeting sites by the scaRNA:tracrRNA-FnoCas9 machinery. We identified a 17 bp region in both the 1104 and 1101 5’ UTRs with 100% sequence identity as a potential site of repression (Figure 2A, S2A). To test whether the genomic regions encoding the 5’ UTRs can confer FnoCas9-dependent repression, we constructed a series of chromosomal promoter fusions driving expression of a non-native sequence gfp*, replacing the 1104-1101 locus in the genome, in strain backgrounds with Cas9 (Cas9+) and without Cas9 (Cas9−) (Figure 2B, S3A). gfp* was placed directly downstream of the 1104 promoter and the 1104 5’ UTR. Expression of gfp* was greatly enhanced in the cas9 mutant as compared to WT, consistent with a role for FnoCas9 in mediating repression (Figure 2C). No such repression of gfp* was observed in the WT strain from a similar reporter construct lacking the 1104 5’ UTR (Figure 2C). These data indicate that the genomic region encoding the 1104 5’ UTR can direct a non-native transcript to be under scaRNA:tracrRNA-FnoCas9-dependent repression.

Figure 2. FnoCas9 targets sequences coding for 5’ untranslated regions (UTRs) leading to transcriptional interference.

A) Alignment of 1104 (teal) and 1101 (red) 5’ UTRs with the location of the transcriptional start site (TSS) and start codon (ATG) highlighted. Brackets indicate the number of nucleotides in each UTR flanking the aligned sequence. B) Schematic indicating the design of the fusion constructs used to interrogate the roles of the promoters and 5’ UTRs in transcriptional repression by FnoCas9. C-D) Relative gfp* transcript level measured by quantitative real time PCR (qRT-PCR) from constructs with either the native promoter (“1104” in C, and “1101” in D), p146 synthetic promoter, or lac promoter, and with or without the native 5’ UTR (1104 in C, and 1101 in D), in a wild-type (Cas9+) or cas9 mutant (Cas9−) background. (n=3, error bars represent s.e.). **p≤0.005; ***p≤0.001.

To test whether the 1104 5’ UTR could promote FnoCas9-dependent repression downstream of non-native promoters, we generated chromosomal reporter constructs in which the 1104 promoter was replaced with either p146, a synthetic promoter with constitutive expression in F. novicida, or the broad host range T5 promoter from E. coli containing a lac operator (Bryksin and Matsumura, 2010; McWhinnie and Nano, 2014). Irrespective of the promoter used, the 1104 5’ UTR conferred FnoCas9-dependent repression of gfp* (Figure 2C). Furthermore, the genomic region encoding the 1101 5’ UTR was also capable of conferring FnoCas9-dependent repression of gfp* when located downstream of the native 1101, synthetic, or lac promoters (Figure 2D). Taken together, these data demonstrate that the 1104 or 1101 5’ UTRs can direct a transcript to be under FnoCas9 regulatory control independent of the promoter used.

scaRNA has complementarity to the 1104 and 1101 5’ UTRs

To determine how the FnoCas9 machinery might target the 5’ UTRs of 1104 and 1101, we bioinformatically searched for predicted interactions between the scaRNA or tracrRNA and these regions. We precisely defined the sequences of the scaRNA and tracrRNA using a small RNAseq analysis of WT F. novicida (Figure 3A, S2E-F) (Chylinski et al., 2013). The 56 base pair scaRNA contains the degenerated repeat predicted previously, however, the tail of the scaRNA extends to the 3’ of the repeat (Sampson et al., 2013). With these new data, we identified a sequence of 11 bases of perfect complementarity between the template strand of the 1104 5’ UTR and the tail of the scaRNA (which was included in a larger region of 15 bases of complementarity over 19 bases in the scaRNA tail; Figure 3A). This site overlapped with the 17 bp stretch of homology between the 1104 and 1101 5’ UTRs (Figure S2A), with the template strand of the 1101 5’ UTR also having 11 bases of perfect complementarity to the scaRNA tail (and 12 bases of complementarity over a region of 15 bases; Figure 3B). This suggests that scaRNA may direct FnoCas9 to the DNA encoding the 1104 and 1101 5’ UTRs since the sequence of the template strand is not encoded on the mRNA.

Figure 3. A PAM motif is required for FnoCas9 transcriptional interference.

A-B) Schematic of predicted interactions between scaRNA (orange), tracrRNA (blue), and A) the 1104 5’ UTR or B) the 1101 5’ UTR. The underlined region represents the identical sequence conserved between the 1104 and 1101 UTRs. The transcriptional start site (TSS) and PAM are shown on the coding strand of the UTRs. C-D) Relative gfp* transcript levels were measured by qRT-PCR. (C) Expression of gfp* is driven by the 1104 promoter and 1104 5’ UTRs containing either a WT TGG PAM sequence or a TAA mutation to the PAM, in strains WT (Cas9+) or Δcas9 (Cas9−) strains (n=3, error bars represent s.e., **p≤0.005). (D) Expression of gfp* is driven by the 1101 promoter and 5’ UTRs containing either a WT TGG PAM sequence or a TAA mutation to the PAM, in strains WT (Cas9+) or Δcas9 (Cas9−) strains (n=3, error bars represent s.e., *p≤0.05). E) Electromobility shift assays (EMSA) with FnoCas9 (150 nM) and scaRNA:tracrRNA duplex (300 nM), as indicated. A DNA target (1 nM) containing 11 bases of complementarity to scaRNA adjacent to a WT PAM (TGG, 9029/9030; DNA sequences in Table S2) or a mutant PAM (TAA, 9041/9042) was used.

FnoCas9 uses a PAM to interact with target 5’ UTR DNA

The Cas9 complex requires recognition of a PAM sequence in the dsDNA target before the crRNA spacer can interact with its complementary sequence in the DNA. We identified an NGG PAM sequence on the coding strand adjacent to the predicted scaRNA binding site on the template strand of both 5’ UTRs (Figure 3A-B). To test if the PAM was required for FnoCas9-dependent repression, we mutated the PAM of the 1104 5’ UTR from TGG to TAA in the gfp* reporter strain with the 1104 promoter. FnoCas9 lost the ability to repress gfp* expression upon mutation of the PAM, consistent with the possibility that FnoCas9 binds the 1104 5’ UTR DNA (Figure 3C). Similarly, we constructed the TAA PAM mutation in the 1101 promoter and 5’ UTR reporter strain and observed a loss of FnoCas9-dependent repression (Figure 3D).

To determine whether scaRNA mediates FnoCas9 DNA binding as suggested by the previous data, and to further investigate whether a PAM is required, we conducted electrophoretic mobility shift assays (EMSAs) with scaRNA:tracrRNA, FnoCas9, and DNA oligonucleotide targets containing the 11 bases of identity to the scaRNA tail shared between the 1104 and 1101 5’ UTRs, and either a WT TGG PAM or a mutated TAA PAM (Figure 3E). FnoCas9 binding to DNA was dependent on scaRNA:tracrRNA as well as the WT PAM, since there was no binding observed to DNA encoding the mutated PAM (Figure 3E). These data suggest that scaRNA interacts with DNA in a PAM-dependent manner.

To determine whether FnoCas9 interacts with the 1104 5’ UTR DNA in F. novicida, we crosslinked intact cells expressing either a FLAG epitope-tagged WT Cas9 or a FLAG-tagged point mutant (Cas9:R59A-FLAG) that is unable to interact with any of the FnoCas9-associated RNAs due to a mutation in the RNA binding domain. We next immunoprecipitated Cas9-FLAG from bacterial lysates, fragmented and isolated crosslinked DNA, and amplified the 1104 5’ UTR region by qPCR. 1104 was enriched in the WT Cas9-FLAG pulldown relative to the Cas9:R59A-FLAG mutant, demonstrating that FnoCas9 does indeed interact with DNA (Figure S2B). Furthermore, this interaction was dependent on the scaRNA since Cas9-FLAG in a scaRNA deletion strain behaved similarly to Cas9:R59A-FLAG in the WT (Figure S2B). 1101 followed the same trend of enrichment in the WT Cas9-FLAG pulldown compared to the Cas9:R59A-FLAG and ΔscaRNA Cas9-FLAG strains (Figure S2C). A control gene, FTN_0544, that is not regulated by FnoCas9 (Figure 7B-C) was not enriched in the Cas9-FLAG and Cas9:R59A-FLAG pulldowns compared to WT (Figure S2D). The overall reduced rate of 1101 DNA enrichment compared to 1104 is likely the result of the fewer bases of complementarity between scaRNA and the 1101 5’ UTR compared to the 1104 5’ UTR, which may reduce the affinity of the pulldown as well as the level of transcriptional repression (Figure S2B-C). Collectively, these data indicate that scaRNA mediates the interaction of FnoCas9 with 1104 and 1101 DNA, representing the first natural example of Cas9-mediated transcriptional interference.

Figure 7. scaRNA can be reprogrammed to repress new targets.

A) Schematic of the scaRNA target site when reprogrammed to interact with the intergenic region between two F. novicida genes, 0544 and 0545, that are transcribed in opposite directions (TSS indicated with arrows). 0544 and 0545 ORFs are 98 bp apart in the genome and were targeted upstream of the two TSSs. B-D) qRT-PCR for transcript levels of (B) 0544, (C) 0545, and (D) 1104 in WT, Δcas9, reprogrammed scaRNA (WT with the scaRNA tail reprogrammed to target 0544-0545), and Δcas9+reprogrammed scaRNA (Δcas9 with the scaRNA tail reprogrammed to target 0544-0545) (n=6, error bars represent s.e., *p≤0.05; **p≤0.005; ***p≤0.001). E) Percent survival of WT, reprogrammed scaRNA, Δ0544, and Δcas9+reprogrammed scaRNA strains 6 hours after polymyxin treatment (100 μg/mL) relative to untreated strains (n=3, error bars represent s.e., *p≤0.05).

Extent of complementarity to scaRNA modulates transcriptional interference

We next evaluated which factors control the level of transcriptional interference exhibited by FnoCas9. The importance of complementarity between the scaRNA and 1104 5’ UTR was tested by measuring gfp* transcript level from chromosomal fusion constructs with either 0, 8, 11, or 15 bases of complementarity to scaRNA on the template strand, followed by a PAM, as well as from the equivalent plasmid-based fusion constructs (Figure S3A). Compared to the strain without any complementarity between scaRNA and the 1104 5’ UTR, the highest level of gfp* repression was observed in the strain with 15 bp of complementarity to scaRNA, with lesser repression in the strain with 11 bp of complementarity (Figure 4A, S4). Repression of gfp* was alleviated in the strain with only 8 bp of scaRNA complementarity (Figure 4B, S4). These data indicate that 11 to 15 bp of identity between the scaRNA and its target is sufficient for repression, with 15 bp providing the strongest effect (Figure 4A, S4). We mutated the PAM in the construct with 15 bp complementarity to scaRNA, and found that this mutation restored gfp* levels to that of a construct with 0 bp complementarity to scaRNA (Figure S3B).

Figure 4. FnoCas9 transcriptional interference is controlled by degree of scaRNA complementarity and target proximity to the TSS.

A) Relative gfp* transcript levels were measured by qRT-PCR from constructs containing sequences with different lengths of complementarity to the scaRNA tail (0, 8, 11, 15 bp) between a synthetic promoter and gfp* sequence in the chromosome (n=3, error bars represent s.e., ***p≤0.001). B-F) Electromobility shift assays (EMSA) with a FnoCas9:scaRNA/tracrRNA complex (1:2 molar ratio FnoCas9:RNA duplex) and DNA target (1 nM) containing different extents of complementarity to scaRNA: (B) 0 bp (9025/9026; DNA sequences in Table S2), (C) 8 bp (9027/9028), (D) 11 bp (9029/9030), (E) 15 bp (9031/9032), (F) 20 bp (9033/9034). G) Plasmid inhibition assay of WT F. novicida and Δcas9 with plasmids containing PAM-adjacent target sequences with 0, 11, 15, and 20 bases of complementarity to scaRNA between a synthetic promoter and gfp*. Results are presented as percent transformation into WT relative to Δcas9 (n=3, error bars represent s.e., **p≤0.005). H) Relative gfp* transcript levels were measured by qRT-PCR in strains with varying numbers of additional bases (0, 5, 10, 20 bp) placed between the TSS of a synthetic promoter and a sequence with 11 bp of complementarity to the scaRNA, followed by gfp*. A strain with gfp* placed downstream of the synthetic promoter and 0 bp of complementarity to scaRNA was used as a control (n=3, error bars represent s.e., **p≤0.005; ***p≤0.001). I-K) EMSAs with a FnoCas9:scaRNA/tracrRNA complex (1:2 molar ratio FnoCas9:RNA duplex) and DNA targets (1 nM) with 5, 10, or 20 bp between the TSS and the 11 bp region of complementarity to scaRNA: (I) 5 bp from TSS (9035/9036; DNA sequences in Table S2), (J) 10 bp from TSS (9037/9038), and (K) 20 bp from TSS (9039/9040).

The degree of transcriptional interference correlated with the affinity of FnoCas9 binding. EMSAs revealed that the affinity of the scaRNA:tracrRNA-FnoCas9 complex for the DNA increased with the amount of scaRNA complementarity (8, 11, 15 bp) (Figure 4C-E). The FnoCas9 complex did not bind to DNA with no scaRNA complementarity or in the absence of the scaRNA:tracrRNA duplex (Figure 4B).

To determine if either strand of the DNA could be targeted by FnoCas9 for repression, the reverse complement of the construct with 15 bp scaRNA complementarity was placed between the promoter and gfp*, resulting in a fusion construct with the PAM and 15 bp of complementarity to scaRNA on the coding strand (Figure S3A). FnoCas9 repressed transcription equally when targeted to the coding and template strands of the DNA (Figure S3C-D).

We also attempted to investigate the ability of scaRNA to bind and repress a construct to which it had 20 bases of complementarity. However, we observed a significantly reduced number of transformants while attempting to make such a strain, and the transformants had mutations in the scaRNA target (Figure S3E). We reasoned that this might be due to lethal targeting of recombinants by the scaRNA:tracrRNA-FnoCas9 complex, and set out to test this. First, we validated that the scaRNA:tracrRNA-FnoCas9 complex was indeed able to bind a 20 bp target in vitro (Figure 4F). We next used a transformation inhibition model to directly test whether modulating the amount of complementarity between scaRNA and artificial exogenous targets (both a linear allelic exchange fragment and a plasmid) harboring the 5’ UTR region with the PAM was sufficient to block transformation. As expected, transformation of constructs containing 8, 11, and 15 bases of complementarity to scaRNA into WT and cas9 mutant strains resulted in an equal number of transformants. However, when the region of complementarity between the target plasmid and the scaRNA tail was artificially extended to 20 bases, we observed a significantly reduced number of plasmid transformants in the WT but not the cas9 mutant strain (Figures 4G). We then constructed mutants with 20 bp of complementarity to scaRNA between a promoter and gfp* in a dCas9 strain, in which FnoCas9 has point mutations in the active sites used for DNA cleavage. The mutations present in dCas9 did not alter the ability of FnoCas9 to repress transcription of 1104 (Figure S3F). dCas9 was able to repress transcription of gfp* in strains with 20 bases of complementarity to scaRNA, and repression was more efficient than the native repression from the scaRNA interaction with the 1104 5’ UTR (Figure S3G). Finally, we constructed a dCas9 strain with 20 bases of complementarity to a crRNA. dCas9 repressed transcription using 20 bases of crRNA complementarity at the same level as WT FnoCas9 repressed transcription from the 1104 5’ UTR using scaRNA. Further, there was less repression from both of these constructs than by scaRNA from a 20 bp target with dCas9 (Figure S3G). These results suggest that scaRNA is capable of guiding the FnoCas9 complex to cleave a DNA target only when the extent of complementarity is sufficient, and that repression via binding occurs independently of the cleavage active sites. The viability of F. novicida indicates that the native complementarity between scaRNA and the 1104 and 1101 5’ UTRs is not sufficient to cause cleavage of the genomic DNA.

Proximity of the scaRNA binding site to the TSS is required for transcriptional interference

We next investigated the importance of proximity between the scaRNA target in the 5’ UTR and the transcriptional start site (TSS). To do this, we measured gfp* transcript levels from fusion constructs with either 0, 5, 10, or 20 bases between the TSS and the 1104 5’ UTR region with complementarity to scaRNA (Figure S3A). We observed that constructs with 0, 5, and 10 bases between the TSS and the scaRNA complementarity region effectively repressed gfp* (Figure 4H). However, the construct with 20 bp between the TSS and the scaRNA complementarity region exhibited significantly reduced repression (Figure 4H). To test if this was the result of altered binding, we conducted in vitro DNA binding assays with constructs that have 0, 5, 10, or 20 bases between the TSS and an 11 bp scaRNA target region. All of the constructs had comparable binding affinities to FnoCas9 in complex with a scaRNA:tracrRNA duplex, suggesting that the difference in transcriptional inhibition observed when the FnoCas9 binding site is moved further from the TSS is not the result of decreased binding affinity (Figure 4D,I-K). These data highlight that the scaRNA complementarity region must be in close proximity to the TSS for effective FnoCas9-dependent transcriptional interference to occur. Together, these results indicate that binding affinity through complementarity to scaRNA can modulate the level of transcriptional interference nearby a TSS, until the number of bases alters Cas9 function from DNA binding to cleavage.

Cleavage-capable FnoCas9 binds competing RNAs to form two distinct complexes with different functions

We demonstrated that the presence of crRNA does not contribute to the repression of 1104-1101 by the scaRNA:tracrRNA-FnoCas9, and that scaRNA with artificially extended complementarity to a target could inhibit transformation with a target-containing plasmid (Figure 1C-F, S3E). To test whether DNA targeting by the CRISPR array was independent of scaRNA, we transformed WT F. novicida and deletion mutants of crRNA, scaRNA, tracrRNA, and cas9 with a plasmid containing a target of an endogenous crRNA (with and without a PAM), and evaluated the ability of each strain to restrict transformation relative to WT. All strains were transformed with the plasmid lacking a PAM at the same frequency. However, cas9, tracrRNA, and crRNA mutants were unable to restrict transformation with a target plasmid containing a PAM, while the scaRNA mutant inhibited transformation similarly to WT (Figure S5A). Together, these results led us to hypothesize that FnoCas9 forms two distinct cleavage-capable complexes in F. novicida, one with a scaRNA:tracrRNA duplex, and another with a crRNA:tracrRNA duplex (Figure S5B). To test this, we measured the interaction between purified FnoCas9 and preformed complexes with either scaRNA:tracrRNA or crRNA:tracrRNA. This analysis revealed that FnoCas9 can indeed interact with both small RNA duplexes (Figures S5C-D), and raised the question of how these distinct small RNAs may affect each other.

We performed Northern blot analysis for tracrRNA, crRNA, and scaRNA from a panel of mutant and complemented strains. As expected, the abundance of each small RNA was dependent on the presence of FnoCas9 since each was undetected in a cas9 mutant but restored in the complemented strain. In addition, analysis of the tracrRNA mutant and complemented strains indicated that the presence of crRNA and scaRNA was dependent on tracrRNA. tracrRNA processing was retained upon deletion of the crRNAs, suggesting that scaRNA can guide the processing of tracrRNA, likely through a similar duplex interaction with the tracrRNA anti-repeat as crRNA. Interestingly, deletion of scaRNA did not significantly alter the levels of crRNA, while deletion of crRNA led to an increase in scaRNA levels that was recovered to WT levels in a crRNA complement (Figures 5A-C). To investigate the impact of the increased abundance of scaRNA in a crRNA mutant, we measured the level of 1104-1101 in a crRNA mutant compared to WT by qRT-PCR. In a crRNA mutant, 1104-1101 were expressed at lower levels than in WT F. novicida, consistent with enhanced transcriptional repression (Figure 5E-H, 1C-F). Complementation of the crRNA mutant restored wild-type levels of crRNA expression and repression of 1104-1101 (Figure 5D-H). Taken together, these results suggest that there is competition for FnoCas9 between the scaRNA:tracrRNA and crRNA:tracrRNA complexes.

Figure 5. FnoCas9 forms complexes with two different RNA duplexes.

(A-C) Northern blots for (A) tracrRNA, (B) crRNA, and (C) scaRNA in wild-type (WT), mutants for each component of the FnoCas9 complexes (Δcas9, ΔtracrRNA, ΔcrRNA and ΔscaRNA), and the complementation strains. D-H) qRT-PCR for transcript levels of (D) crRNA, (E) 1104, (F) 1103, (G) 1102, and (H) 1101 in WT, ΔcrRNA, and ΔcrRNA + crRNA complemented strains ((D-G) n=7-9, (H) n=4-6, error bars represent s.e., *p≤0.05; **p≤0.005; ***p≤0.001).

Repression of each gene in the FnoCas9 regulon contributes to virulence

It was unclear whether repression of each gene in the FnoCas9 regulon contributes to F. novicida virulence. To test this, we infected mice subcutaneously with cas9 double mutants also lacking one of the repressed genes. This revealed that 1104 expression plays a major role in the attenuation of the cas9 mutant, similar to 1103 (Figure S6B). Deletion of 1104 from a cas9 mutant led to a greater than 4-log enhancement in the levels of bacteria recovered after infection as compared to the cas9 mutant (Figure S6B). 1102 and 1101 played more minor roles, but still contributed to attenuation of F. novicida when expressed, reducing virulence by 1-2 logs (Figure S6C-E). To test if virulence could be completely restored to the cas9 mutant by deletion of the entire 1104-1101 locus, we infected mice with a Δcas9Δ1104-1101 strain and evaluated the bacterial burden in the spleen 48 hours post-infection. We found that unlike the Δcas9Δ1103 mutant, virulence was restored to WT levels by deletion of the four genes in the FnoCas9 regulon from a Δcas9 strain (Figure 6). These data demonstrate that F. novicida uses Cas9 to specifically repress four genes whose expression would otherwise lead to attenuation of bacterial virulence.

Figure 6. Deletion of 1104-1101 restores virulence of a cas9 mutant.

Mice were subcutaneously infected with either WT F. novicida, Δcas9, Δcas9Δ1103, or Δcas9Δ1104-1101. Spleens were homogenized and plated for enumeration of colony forming units (CFU) at 48 hr post-infection (n=5, bar represents geometric mean, **p≤0.005).

scaRNA can be reprogrammed to guide FnoCas9 to repress non-native targets

We next sought to determine whether scaRNA can be reprogrammed to repress new targets and whether promoter regions can be targeted for repression similar to the 5’ UTRs. We replaced the 1104-1101 targeting portion of the scaRNA tail with a 16 bp sequence complementary to a portion of the 98 bp intergenic region between FTN_0544 (0544; naxD, new NCBI locus tag: FTN_RS02820) and FTN_0545 (0545; flmF2, new NCBI locus tag: FTN_RS02825), located upstream of the TSS for each gene (Figure S7). These two genes are required for the modification of outer membrane lipid A that leads to resistance to the antibiotic polymyxin B; they are transcribed in opposite directions, and are not regulated by FnoCas9 (Figure 7A-C, S7A) (Kanistanon et al., 2008; Llewellyn et al., 2012). In the scaRNA reprogrammed strain, we observed a significant reduction in transcript levels of both 0544 and 0545 compared to WT (Figures 7B, C). This suggests that the FnoCas9 CRISPR-Cas9 system can be engineered to repress the expression of new targets, and that repression is independent of the strand targeted. Furthermore, this repression was dependent on FnoCas9, since a cas9 mutant harboring the reprogrammed scaRNA exhibited WT levels of 0544 and 0545 (Figures 7B, C). In addition, when the natural scaRNA was reprogrammed for 0544-0545, the strain lost the ability to repress 1104 (Figure 7D). Reprogrammed scaRNA repressed transcription of 0544 at the same efficiency in strains harboring WT Cas9 or dCas9 (Figure S7B). The repression of 0544 and 0545 in the scaRNA reprogrammed strain led to an increased susceptibility to polymyxin B of almost 100-fold, similar to a 0544 deletion strain (Figure 7E). The susceptibility was reversed by deletion of cas9 from the reprogrammed scaRNA strain (Figure 7E). These results indicate that FnoCas9 can be reprogrammed to repress expression from new targets in a scaRNA:tracrRNA-dependent manner, highlighting the scaRNA:tracrRNA-FnoCas9 machinery as a potential new tool to control gene expression and modulate bacterial physiology.

Discussion

Using an analysis of the native FnoCas9 transcriptome to elucidate the specificity of endogenous gene regulation, we located the site of interaction between the FnoCas9 complex and the DNA of the 5’ UTR of each transcript in its regulon. FnoCas9 uses scaRNA to interact with the template strand of the 5’ UTR by recognition of a PAM and a scaRNA-complementary sequence on the DNA target. By targeting the 5’ UTR DNA, FnoCas9 functions as a transcription factor, repressing gene expression. Through this interaction, FnoCas9 regulates the expression of four endogenous genes with remarkably high specificity. We determined that repression is dependent on a PAM in the 5’ UTR, and that the sensitivity of natural FnoCas9 regulation could be modulated by the length of the RNA-target interaction and proximity of the scaRNA binding site to the TSS. Further, transcriptional repression by FnoCas9 could be achieved through the targeting of either strand. We demonstrated that the extent of complementarity between scaRNA and the DNA target alters the binding affinity of the dual-RNA-FnoCas9 complex to the DNA. We also observed that the distance of the TSS from the scaRNA target region does not affect the binding affinity of the complex to the DNA. Using this knowledge of scaRNA:tracrRNA-FnoCas9 interaction with DNA, we reprogrammed scaRNA such that FnoCas9 targeted the promoters of desired genes to repress transcription, highlighting the potential use of scaRNA:tracrRNA-FnoCas9 in the control of gene expression.

Previous work from our lab suggested a different model for FnoCas9-mediated regulation of 1103 mRNA that depended on its direct interaction with tracrRNA. We therefore closely re-examined the experiments that suggested RNA targeting in Sampson et al., 2013, and have either been unable to reproduce critical experiments or identified flaws that led to misinterpretation of the previous results, as summarized below. While we cannot rule out the possibility of a low level of direct targeting of 1103 mRNA by the scaRNA:tracrRNA-FnoCas9 complex or an alternative mechanism of FnoCas9-mediated gene regulation, the data presented here clearly demonstrate that transcriptional interference is the dominant contributor. We sincerely apologize for the previous misleading data and the impact they may have had on others in the field.

We have obtained different results when trying to reproduce the RNA degradation experiment (Figure 2a) from Sampson et al 2013 and no longer observe differences in the rates of 1103 mRNA stability between the wild-type and cas9 mutant strain (Correction of Sampson et al., 2013, Nature). We do not have an explanation for this discrepancy, but think it may in part be due to the complications derived from measuring 1103 transcript stability in strains with vastly different baseline levels of this mRNA. We have also been unable to replicate the immunoprecipitation experiments in Figure 2g of Sampson et al., 2013 (Correction of Sampson et al., 2013, Nature). The inherent noise in the RNA pulldowns, which did not incorporate a cross-linking step like the DNA pulldowns reported herein, may have contributed to the previous results. Further, the pulldown of 1103 mRNA may have been caused by binding of FnoCas9 to the 1104 DNA where a low level of transcriptional read-through could have occurred, enabling the enrichment of nascent mRNA that are undergoing transcription and are attached to the DNA. However, consistent with Sampson et al., 2013, here we present robust data indicating scaRNA and tracrRNA together with FnoCas9 are required for FnoCas9-dependent repression of RNA levels (Figure 1A, 1C-F, 3E, 5A-H, 7B-D, S2B-D, S5C-D) (Sampson et al., 2013). Finally, in Figure 2h of Sampson et al, 2013, a tracrRNA mutant containing substitutions in the putative 1103 interaction site was used to suggest that 1103 was repressed via interaction with the tracrRNA. However, due to the location of the tracrRNA mutation in a stem loop that interacts with FnoCas9, we feel that this data can no longer be used to draw a conclusion as to the role of the mutated bases in the repression of 1103, and therefore we no longer have evidence to support a role of tracrRNA in direct interaction with 1103 (Correction Sampson et al., 2013, Nature). Therefore, we no longer have conclusive evidence to support a role for RNA degradation in FnoCas9-mediated regulation of 1103. Although we cannot definitively conclude that the native FnoCas9 system cannot target RNA in some conditions, the data in this manuscript clearly indicate that DNA targeting is the predominant mechanism of endogenous gene regulation.

In some systems, an artificial FnoCas9 complex has been engineered to target RNA. We showed that an FnoCas9 complex with an engineered guide RNA led to a reduction in hepatitis C virus (HCV) levels in human cells (Price et al., 2015). Since HCV is an RNA virus whose genome is only in the form of RNA and never DNA, the reduction in HCV levels by FnoCas9 were due to targeting of RNA, the result of either repression of viral genome replication, translational inhibition, or RNA degradation. In this system, we employed an FnoCas9 catalytic domain point mutant to show that FnoCas9 catalytic activity was not required for repression of HCV levels, ruling out direct RNA degradation via the RuvC and HNH motifs (Price et al., 2015). Another group has also shown that FnoCas9 can repress the levels of tobacco mosaic virus, another RNA virus that also lacks a DNA stage in its lifecycle, during infection of plants (Zhang et al., 2018). However, these mechanisms of RNA targeting are distinct from the regulation of 1104-1101 at the DNA level by the native FnoCas9 system. Whereas the native system uses scaRNA and tracrRNA to target DNA, RNA targeting by the artificial system utilizes a single-guide RNA with tracrRNA modifications. Furthermore, a PAM is not required for HCV repression, while a PAM is involved in FnoCas9 repression of endogenous gene transcription from the DNA.

The mechanism of gene regulation via transcriptional inhibition we have described here is also unique compared to other examples of repression mediated by Cas9 orthologs. NmeCas9, SauCas9 and CjeCas9 have recently been shown to degrade ssRNA in a PAM-independent manner that requires the HNH catalytic domain and perfect or near perfect complementarity with a crRNA spacer (Dugar et al., 2018; Rousseau et al., 2018; Strutt et al., 2018). ssRNA cleavage by Cas9 has been proposed to have roles in endogenous gene regulation (CjeCas9) and foreign nucleic acid defense (SauCas9) (Dugar et al., 2018; Strutt et al., 2018). However, at least in vitro, FnoCas9 is not capable of this mechanism of ssRNA targeting (Strutt et al., 2018). FnoCas9-mediated transcriptional repression of 1104-1101 through interaction with the DNA is also distinct from the engineered targeting of RNA by SpyCas9, which is mediated by Cas9 interaction with the RNA target and requires supplementation of a short PAMmer sequence (O’Connell et al., 2014).

It is not clear why FnoCas9 evolved to repress endogenous gene expression. One explanation for the ability of FnoCas9 to repress transcription from the DNA is as an expansion of the phage defense toolkit, to not only block replication of phage DNA, but also repress the transcription of harmful phage genes. It is hypothesized that scaRNA evolved from a degenerated CRISPR array that contains repeats with impaired complementarity to the inverted repeat of tracrRNA (Chylinski et al., 2013). We found that scaRNA is still capable of interacting with tracrRNA, likely through the repeat similar to the interaction between crRNA and tracrRNA. Transcriptional repression of 1104-1101 by FnoCas9 could be the result of genome shuffling events in the area of the CRISPR arrays, leading to the evolution of the self-targeting scaRNA through environment-specific fitness advantages of 1104-1101 repression. Alternatively, a spacer might have been acquired from the bacterial genome and degenerated during this process, avoiding self-cleavage. In a later step, a promoter may have evolved upstream of this spacer, allowing the bacterium to control its expression independently of the array.

It is particularly interesting that in spite of the degeneration of its repeat sequence, scaRNA has retained the ability to direct DNA cleavage. When F. novicida is transformed with an artificial target containing 20 bases of identity to the scaRNA, FnoCas9 restricts transformation. However, the 11 consecutive bases of perfect complementarity between scaRNA and the native 1104 and 1101 5’ UTRs is sufficient for robust transcriptional repression, which we hypothesize is due to the inability of FnoCas9 to enter a cleavage-favorable conformation with a partial scaRNA-DNA target interaction. If so, this would be similar to what has been observed with shortened crRNA spacers, which guide Cas9 to bind but not cleave a DNA target (Bikard et al., 2013; Sternberg et al., 2015). Thus, modification of the length of the targeting sequence of the guiding scaRNA:tracrRNA duplex determines whether FnoCas9 represses transcription or cleaves its DNA target. We utilized this knowledge to reprogram scaRNA to target genes involved in polymyxin resistance. This led to efficient repression of the targeted genes and greatly increased sensitivity to polymyxin.

The use of a catalytically active Cas9 for gene repression makes this system unique compared to engineered CRISPRi technologies that use catalytically inactive mutants of the protein (dCas9) in complex with an RNA guide containing 20 bases of complementarity to its target (Bikard et al., 2013; Larson et al., 2013; Qi et al., 2013). This highlights the potential for scaRNA:tracrRNA-FnoCas9 complexes to be used to control gene expression, and especially for applications that seek to multiplex DNA cleavage and transcriptional control. Furthermore, within F. novicida, FnoCas9 is able to prevent transformation and regulate gene expression simultaneously, suggesting that, at least in our experimental conditions, sufficient FnoCas9 molecules are bound to both duplexes (tracrRNA:crRNA and tracrRNA:scaRNA) to fulfill each function with minimal effect on the other. This would be analogous to the multiple spacers protecting against different phages in parallel without out-competing each other.

The limited scaRNA:target complementarity required for FnoCas9 transcriptional repression, as compared to 20 bp crRNA or guide RNA complementarity to cleavage targets, could increase the risk of off-target effects. Typically, analyses to identify off-target cleavage sites are performed with full length (20 bp) target sequences, however, we propose that for any use of Cas9, such analyses should include an examination of potential off-target transcriptional effects as well. Similarly, potential endogenous regulatory functions of native CRISPR-Cas9 systems could be identified by decreasing the stringency of self-targeting spacer identification.

FnoCas9 transcriptional repression is critical for F. novicida virulence and we found that repression of each gene in the FnoCas9 regulon has a distinct contribution to virulence in a mouse model (Figure S6B). We find that the attenuation of a cas9 mutant of F. novicida can be reversed by deletion of the 1104-1101 locus (Figure 6). 1104, 1103, and 1101 all exhibit conserved features of Gram-negative bacterial lipoproteins, while 1102 has some but not all of these features (Figure S6A).

Together, these results highlight a novel role for CRISPR-Cas9 systems in endogenous gene regulation and provide a mechanistic explanation of the role Cas9 plays in the virulence of F. novicida. Interestingly, F. novicida utilizes two distinct RNA duplexes for foreign DNA restriction and transcriptional repression, although both are capable of DNA restriction. The prevalence of these systems and the minimal base pair requirements needed for a shift between DNA cleavage and transcriptional interference suggest that a role of Cas9 as a transcriptional regulator may be a broader phenomenon in bacterial physiology than previously expected.

STAR Methods

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, David Weiss (david.weiss@emory.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

In vivo animal work

Specific-pathogen free mice were kept in filter-top cages at Yerkes National Primate Center, and provided food and water ad libitum (Sampson et al., 2013). Emory University Institutional Animal Care and Use Committee (protocol #YER-2000573-061314BN) approved all procedures (Sampson et al., 2013). Female C57BL/6J mice (Jackson) between 8 and 12 weeks were used for all experiments.

Bacterial strains and growth conditions

For information about the strains used and growth conditions, please refer to the strain list (Table S1) and the “METHOD DETAILS” for each experiment.

In vitro studies

The DH5α derivative of E. coli K12 (NEB 5-alpha) was used for plasmid isolation and cells were grown at 37°C in LB medium supplemented with 30 μg/mL kanamycin. FnoCas9 was purified from E. coli NiCo21(DE3) (NEB) expressing Cas9 from pEC657. Cultures for Cas9 purification were grown at 37°C to OD₆₀₀ of 0.7–0.8, Cas9 expression was induced with IPTG (5mM) and cultures were grown overnight at 13°C, as previously described (Fonfara et al, 2014).

METHOD DETAILS

Construction of cDNA libraries for total RNA analysis

Biological triplicates of Francisella novicida U112 WT, Δcas9, ΔtracrRNA and ΔscaRNA were grown overnight on LB plates (37°C), cultured into LB medium (37°C shaking), and grown until OD_{620 nm}= 0.1. Twenty-five mL of bacterial culture were mixed with 25 mL of 1:1 acetone/ethanol and total RNAs were extracted using TRIzol (Ambion) and treated using turbo DNAse (Ambion). RNA integrity was checked using a bioanalyzer (RIN > 8). cDNA libraries were prepared at the HZI genome analytics platform in Braunschweig, Germany as described with some modifications (Dötsch et al., 2012). Briefly, rRNAs were removed using the MICROBExpress kit (Ambion) and samples were treated with TAP (tobacco acid phosphatase). The RNAs were fragmented using sonication (Covaris) to fragments of 200 nucleotides. T4 Polynucleotide kinase (Fermentas) was used to phosphorylate 5’ ends and remove the 3’ phosphate. Successively, 3’ and 5’ adapters were added using T4 RNA ligase. Reverse transcription was performed using SuperScript II (Invitrogen) followed by 15 cycles of PCR with Phusion (New England Biolabs) and agarose gel purification. The sequencing was performed using 50 nucleotide single end reads (HiSeq2500). cDNA libraries for small RNA analysis were generated as in Chylinski et al., 2013 (Chylinski et al., 2013).

Analysis of total RNA sequencing

The raw data files were demultiplexed using the specific barcode sequences. The reads were trimmed from adapter sequences and the read quality was assessed using fastQC. The reads were mapped using STAR to the Francisella novicida U112 reference genome (NC_008601.1) (Dobin et al., 2013). We retrieved from 201525 to 518855 of uniquely mapped reads. The number of reads for each gene were counted using HTseq and the differential expression analysis was done for each deletion mutant compared to the WT using DESeq2 (Anders et al., 2015; Love et al., 2014).

Northern blot analysis

After RNA extraction (see above), RNAs were resolved on 1% agarose containing 0.74% formaldehyde and transferred by capillarity to a nylon membrane (Hybond™ N+, GE healthcare). Forty pmol of oligonucleotide probes were 5′ radiolabeled using [γ-³²P] ATP (Hartmann Analytics) and T4 polynucleotide kinase (Fermentas) and purified over Microspin™ G-25 columns (GE Healthcare) following the manufacturers’ instructions. The probes were hybridized at 55°C using Rapid-hyb buffer (GE healthcare). The radioactive signal was visualized using a Thyphoon FLA-9500 phosphorimager (GE healthcare) and the transcript sizes were determined using a 3’-end radiolabeled RiboRuler high range RNA ladder (Thermo Scientific™). 16S rRNA was used as a loading control.

Electrophoretic mobility shift assay

tracrRNA, crRNA and scaRNA were in vitro transcribed from annealed oligonucleotides or PCR fragments (Table S2) using the AmpliScribe™ T7-Flash™ Transcription Kit (Epicenter). scaRNA:tracrRNA or crRNA:tracrRNA duplexes were hybridized by heating to 95°C and cooling to room temperature in hybridization buffer (200 mM NaCl, 20 mM HEPES pH 7.5). DNA substrates were generated by annealing complementary DNA oligonucleotides (Table S2) in a similar manner. Annealed oligos were 5′ radiolabeled using [γ-³²P] ATP (Hartmann Analytics) and T4 polynucleotide kinase (Fermentas) according to the manufacturers’ instructions, and purified using Microspin™ G-50 Columns (GE Healthcare). Cas9 was pre-incubated at 37°C with two-fold molar excess of prehybridized scaRNA:tracrRNA duplex for 15 min in DNA-binding buffer (20 mM Tris HCl pH 7.4, 100 mM KCl, 5 mM CaCl₂, 1 mM DTT, 5% (w/v) glycerol, 40 ng/μl poly(dI-dC)), then 1 nM labelled DNA substrate was added. For duplex RNA EMSAs, tracrRNA was dephosphorylated using FastAP Alkaline Phosphatase (Thermo Scientific™) and then 5′ radiolabeled as described above, before hybridization with scaRNA or crRNA. The reaction was performed as described above using 100 ng/ul tRNA as competitor. Binding reactions were incubated for 1 h. The samples were loaded on a native 5% polyacrylamide gel, which was run in 0.5X TBE supplemented with 5 mM CaCl₂. The gels were exposed on autoradiography films and visualized by phosphorimaging.

Francisella novicida strain construction and growth conditions

All strains are listed in Table S1. Francisella novicida U112 derived strains were grown in Tryptic Soy Broth (VWR International) supplemented with 0.2% cysteine (BD Biosciences) at 37 °C with shaking. Strains were plated on Tryptic Soy Agar (TSA, VWR International) plates supplemented with 0.1% cysteine. Deletion and fusion construct mutants were constructed by allelic exchange (primers listed in Table S3). Promoter fusions were inserted in place of the 1104-1101 locus in the chromosome in Δ1104-1101 background strains. Chromosomal promoter and 5’ UTR fusions were made using a fragment of gfp derived from the pBav-kGFP vector, gfp* (bases 170-499). CRISPR-Cas9 system mutants Δcas9,Δ scaRNA, ΔcrRNA, ΔtracrRNA, their respective complemented strains, and Δcas9Δ1103, Cas9-FLAG, Cas9:R59A-FLAG, and ΔscaRNA Cas9-FLAG were described previously (Table S1) (Jones et al., 2012; Sampson et al., 2013; Weiss et al., 2007). Phusion high-fidelity DNA polymerase (New England Biolabs) was used to amplify homologous sequences (500-1000 bp) flanking the region of interest from genomic DNA isolated with DNeasy Blood and Tissue Kit (Qiagen). Overlapping PCR was used to construct the allelic exchange substrate by inserting a kanamycin selectable marker containing Flp recombinase target sites (FRT) between the flanking sequences (Llewellyn et al., 2011). Allelic exchange substrates were transformed into chemically competent F. novicida (Llewellyn et al., 2011). Mutants were selected on media supplemented with kanamycin sulfate (30 μg/ml, Fisher Scientific). Mutants were confirmed by PCR amplification from outside of the recombined region followed by sequencing (Genewiz) using “seq” primers (Table S3). The selection cassette was removed from the mutants using a temperature sensitive suicide vector at 30 °C, pFFlp, carrying the Flp recombinase in trans (Gallagher et al., 2008). pFFlp was selected for on TSA plates with 1% cysteine and 15 μg/ml tetracycline (Alfa Aesar). Following unmarking the strains were moved to 37 °C to remove the plasmid, as described previously (Gallagher et al., 2008).

Quantitative real-time PCR

RNA was isolated from bacterial cultures at OD600 nm of 0.8-1.0 using TRI-reagent and a Direct-zol RNA MiniPrep Kit (Zymo Research). DNA was removed with Turbo DNaseI (Ambion Biosciences). qRT-PCR was performed with biological triplicates using the primers indicated in Table S3 and Power Sybr Green RNA-to-CT one-step kit (Applied Biosystems). CT values for each gene were normalized to the Francisella novicida housekeeping gene DNA helicase II (uvrD, FTN_1594) to determine 2-ΔΔct for each condition (Sampson et al., 2013). Results are plotted as relative transcript levels of percent transcript in WT (intact Cas9) compared to the transcript level in a cas9 mutant.

Cas9-FLAG crosslinking and immunoprecipitation

DNA was crosslinked and immunoprecipitated as described previously, with the indicated modifications to optimize for F. novicida and Cas9-FLAG (Jaggavarapu and O'Brian, 2014). Cultures grown to OD₆₀₀ nm of 0.6-0.8 were crosslinked by adding formaldehyde to a final concentration of 1% in 10 mM PO₄ buffer, and shaking for 10 min at RT. Reactions were quenched using 1:10 volume of 100 mg/ml glycine and shaken for 30 min at 4°C. Cells were pelleted, washed 2X in PBS, and concentrated 20X in lysis buffer (100 mM Tris-HCl pH 8.0, 300 mM NaCl, 10 mM EDTA) with protease inhibitor cocktail (Bacterial ProteaseArrest, G-Biosciences). Samples were further lysed and DNA was fractionated to 500-3000 bp fragments by sonication (15x 10 sec pulses with 15 sec pauses). Lysates were cleared by centrifugation and Cas9-FLAG was pulled-down from the supernatant using anti-Flag M2 agarose beads (Sigma). Following elution from the anti-FLAG beads, DNA was uncrosslinked from Cas9 by adding NaCl (final concentration 0.2 M) and incubating overnight at 65°C. DNA was purified (Qiagen PCR purification kit), and used as a template for qPCR. Results were normalized to the input DNA levels and a housekeeping gene (see qRT-PCR methods).

PCR amplification from cDNA

RNA and extracted from WT and Δcas9 (see above) and converted to cDNA using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Segments were PCR amplified with Phusion high-fidelity DNA polymerase (New England Biolabs) from the cDNA of each strain and compared to amplification from WT gDNA (Table S3).

Plasmid construction

Plasmids were constructed using the primers indicated in Table S3. The broad host range vector pBAV1K-T5-GFP (pBAV) was used as the control plasmid and backbone for the plasmids in all assays (Bryksin and Matsumura, 2010). Plasmids that were used to measure repression and transformation inhibition were made by replacing the promoter and RBS driving expression of gfp in the pBav vector with the synthetic constitutive promoter (p146) followed by different amounts of complementarity to the scaRNA tail and a PAM sequence directly upstream of gfp (McWhinnie and Nano, 2014). Plasmids were constructed using a Gibson Assembly Cloning Kit (NEB) and transformed and isolated from competent E. coli (NEB 5-alpha) using Zyppy Mini, Midi and Maxi prep kits (Zymo Research). Cas9 target plasmids containing a Cas9 crRNA spacer with and without a PAM were previously described (Price et al., 2015).

Transformation Assays

Competent cells of F. novicida were made by concentrating cultures at an OD_{600 nm} of 0.8-1.0 10X in 4°C chemical transformation buffer (CTB) (Llewellyn et al., 2011). For the transformations, DNA was added to 100 μl competent cells, and transformations were incubated by shaking at 37°C for 20 minutes. 1 ml of recovery media (TSB+0.2% cysteine) was then added and transformations were incubated for another 2 hours (shaking, 37°C). Transformations were plated on TSA plates with kanamycin selection and incubated at 37°C overnight. For transformations with plasmid vectors, 500 ng of plasmid was used. For transformation inhibition experiments, transformants per 100 ng plasmid were enumerated and compared between strains to determine transformation efficiency. To measure Cas9 repression of plasmid gfp expression, plasmids used to transform WT and cas9 mutants. Transformants from each strain were isolated and grown in TSB+0.2% cysteine+kanamycin 30 μg/ml selection to an OD_{600 nm} of 0.8-1.0, RNA was isolated and gfp transcript level was measured (see qRT-PCR methods) and normalized to the kanamycin resistance cassette on pBAV to account for variations in plasmid copy number. For transformation assays with allelic exchange fragments, DNA of purified allelic exchange fragments were normalized by concentration and transformed. Transformation efficiency after 24 hours was measured by comparing the number of transformants into WT and a cas9 mutant.

5’ RACE

RNA was isolated from WT cultures as described above. 5’ ends of 0544 and 0545 mRNA were mapped using the reagents and protocols from the 5’/3’ RACE Kit, 2^nd Generation (Roche) unless otherwise noted (primers listed in Table S3). cDNA was synthesized using the primer “GSP1” for each gene and purified with QIAquick PCR purification kit (Qiagen). PolyA-tails were added with terminal transferase and the 5’ ends were amplified using “0544_GSP2” or “0545_GSP3”, and “oligo dT anchor primers” (Table S3 and kit) by PCR as described above. Products were isolated by agarose gel electrophoresis, extracted using a QIAquick Gel Extraction Kit (Qiagen), and 5’ ends sequenced using “0544_GSP3” or “0545_GSP2” primers (Table S3, Genewiz Sanger Sequencing).

Polymyxin susceptibility assay

Overnight cultures of WT, scaRNA_0544/0545, Δ0544 and Δcas9+scaRNA_0544/0545 strains of F. novicida were prepared as described previously (Llewellyn et al., 2012). Strains were incubated with 100 μg/ml polymyxin B (Tokyo Chemical Industries Japan) shaking at 37°C for 6 hours and then plated to enumerate CFU surviving bacteria from each condition. Results are presented as % survival of each strain treated with polymyxin relative to untreated.

Mouse infections

Mice were infected subcutaneously with ~2 × 10⁵ cfu bacteria (Weiss et al., 2007). Spleens were harvested at 48 hours post infection, homogenized in PBS, and the bacterial burden per organ was determined.

QUANTIFICATION AND STATISTICAL ANALYSIS

Prism 5 Graphpad Software was used for statistical analyses. The significance of the bacterial and culture experiments (qPCR of RNA and DNA, killing assays, transformation assays) was determined using a two-tailed student’s t-test, for data with normal distribution. Significance was determined using the Mann-Whitney test for the mouse infections, as not all data was normally distributed. Biological replicate number and error are indicated in the figure legends. Representative gel images are shown for in vitro experiments.

DATA AND SOFTWARE AVAILABILITY

The RNA sequencing data reported in this paper have been deposited in the National Center for Biotechnology Information Sequence Read Archive database, http://www.ncbi.nlm.nih.gov/sra (accession no. SRP148943). The unprocessed gel images have been published at Mendeley under http://dx.doi.org/10.17632/jtjvh9m7zk.1.

Supplementary Material

3

Table S3. RT-PCR, Cloning, and Other Primers, Related to the STAR Methods.

Key Resources Table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-Flag M2 agarose beads	Sigma-Aldrich	Product#A2220
Bacterial and Virus Strains
NEB 5-alpha competent E. coli (High Efficiency)	New England Biolabs (NEB)	Cat#C2987I
NiCo21(DE3) competent E. coli	NEB	Cat#C2529H
See Table S1 for all other strains and plasmids	N/A	N/A
Chemicals, Peptides, and Recombinant Proteins
T4 Polynucleotide kinase	Fermentas, Thermo Fisher Scientific	Cat#EK0031
Phusion® High-Fidelity DNA Polymerase	NEB	Cat#M0530
Hybond™ N+	GE healthcare	Product#RPN119B
[γ-32P] ATP	Hartmann Analytics	Product#SCP-501-150
Rapid-hyb buffer	GE healthcare	Product#RPN1635
RiboRuler high range RNA ladder	Thermo Fisher Scientific	Cat#SM1821
FastAP Alkaline Phosphatase	Thermo Fisher Scientific	Cat#EF0652
Kanamycin sulfate	Thermo Fisher Scientific	Cat#11815024
Bacterial ProteaseArrest	G-Biosciences	Cat#786-330
Polymyxin B	Tokyo Chemical Industries Japan	Product#P1923
Tetracycline	Alfa Aesar	Cat#AAJ6171414
tobacco acid phosphatase	Epicenter	Cat#T19050
Critical Commercial Assays
TURBO DNase	Ambion, Thermo Fisher Scientific	Cat#AM2239
MICROBExpress kit	Ambion, Thermo Fisher Scientific	Cat#AM1905
Microspin™ G-25 columns	GE Healthcare	Mfr#27-5325-01
SuperScript II	Invitrogen, Thermo Fisher Scientific	Cat#18064014
AmpliScribe™ T7-Flash™ Transcription Kit	Epicenter	Cat#ASF3275
DNeasy blood & tissue kit	Qiagen	Cat#69504
TRI-reagent and a Direct-zol RNA MiniPrep Kit	Zymo Research	Cat#R2051
Power Sybr Green RNA-to-CT one-step kit	Applied Biosystems, Thermo Fisher Scientific	Cat#4389986
Qiagen PCR purification kit	Qiagen	Cat#28104
High Capacity cDNA Reverse Transcription Kit	Applied Biosystems, Thermo Fisher Scientific	Cat#4368813
Gibson Assembly Cloning Kit	NEB	Cat#E2611S
Zyppy Plasmid Miniprep	Zymo Research	Cat#D4036
Zyppy Plasmid Midiprep	Zymo Research	Cat#D4025
and Maxi prep kits	Zymo Research	Cat#D4027
5’/3’ RACE Kit, 2nd Generation	Roche	Cat#03 353 621 001
QIAquick Gel Extraction Kit	Qiagen	Cat#28704
Deposited Data
RNA sequencing data	This paper; NCBI sequence read archive database	http://www.ncbi.nlm.nih.gov/sra (accession no. SRP148943)
Unprocessed image files	This paper; Mendeley Data	http://dx.doi.org/10.17632/jtjvh9m7zk.1
Experimental Models: Organisms/Strains
Mice: C57BL/6J	Jackson	JAX stock #000664
Oligonucleotides
Refer to Table S2 and S3 for all oligonucleotides used in the study	N/A	N/A
Recombinant DNA
Refer to Table S1 for recombinant plasmids and strains used in the study.	N/A	N/A
pEC657 (FnoCas9 expression plasmid for purification)	(Fonfara et al, 2014)	N/A
pFFlp	(Gallagher et al., 2008)	N/A
pBAV1K-T5-GFP	(Bryksin and Matsumura, 2010)	N/A
Software and Algorithms
HTSeq	(Anders et al, 2014)	https://htseq.readthedocs.io/en/release_0.11.1/
STAR	(Dobin et al, 2013)	https://code.google.com/archive/p/rna-star/
DESeq2	(Love et al, 2014)	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
FastQC		https://www.bioinformatics.babraham.ac.uk/projects/fastqc/

Highlights.

Highlights are 3–4 bullet points of no more than 85 characters in length, including spaces, and they summarize the core results of the paper in order to allow readers to quickly gain an understanding of the main take-home messages.

• FnoCas9 uses scaRNA to bind endogenous DNA and repress transcription

• The limited length of scaRNA:target complementarity prevents DNA cleavage

• Cleavage-competent FnoCas9 uses distinct RNAs for repression versus cleavage

• scaRNA can be reprogrammed to guide FnoCas9 to repress a new target