Characterization of the self-targeting Type IV CRISPR interference system in Pseudomonas oleovorans

Bacterial Type IV CRISPR-Cas systems are thought to rely on multi-subunit ribonucleoprotein complexes to interfere with mobile genetic elements, but the substrate requirements and potential DNA nuclease activities for many systems within this type are uncharacterized. Here we show that the native Pseudomonas oleovorans Type IV-A CRISPR-Cas system targets DNA in a PAM-dependent manner and elicits interference without showing DNA nuclease activity. We found that the first crRNA of P. oleovorans contains a perfect match in the host gene coding for the Type IV pilus biogenesis protein PilN. Deletion of the native Type IV CRISPR array resulted in upregulation of pilN operon transcription in the absence of genome cleavage, indicating that Type IV-A CRISPR-Cas systems can function in host gene regulation. These systems resemble CRISPR interference (CRISPRi) methodology but represent a natural CRISPRi-like system that is found in many Pseudomonas and Klebsiella species and allows for gene silencing using engineered crRNAs. Type IV CRISPR systems that interfere with the transmission of mobile genetic elements are poorly understood. Here the authors show that a Pseudomonas Type IV-A system targets DNA in a PAM-dependent manner without DNA nuclease activity.

Bacterial Type IV CRISPR-Cas systems are thought to rely on multi-subunit ribonucleoprotein complexes to interfere with mobile genetic elements, but the substrate requirements and potential DNA nuclease activities for many systems within this type are uncharacterized. Here we show that the native Pseudomonas oleovorans Type IV-A CRISPR-Cas system targets DNA in a PAM-dependent manner and elicits interference without showing DNA nuclease activity. We found that the first crRNA of P. oleovorans contains a perfect match in the host gene coding for the Type IV pilus biogenesis protein PilN. Deletion of the native Type IV CRISPR array resulted in upregulation of pilN operon transcription in the absence of genome cleavage, indicating that Type IV-A CRISPR-Cas systems can function in host gene regulation. These systems resemble CRISPR interference (CRISPRi) methodology but represent a natural CRISPRi-like system that is found in many Pseudomonas and Klebsiella species and allows for gene silencing using engineered crRNAs.
Clustered regularly interspaced short palindromic repeats (CRISPR) and CRISPR-associated (Cas) proteins elicit adaptive immunity in many bacterial species 1,2 . Foreign DNA fragments, termed protospacers, with short protospacer-adjacent motifs (PAM) are captured by Cas1-Cas2 adaptation complexes and inserted into an extending CRISPR locus [3][4][5] . Transcription and processing of the CRISPR array results in mature CRISPR RNAs (crRNAs) 2,6,7 that guide CRISPR ribonucleoprotein complexes (crRNPs) towards complementary nucleic acid target sequences 3,[8][9][10] . Two classes and six types of CRISPR-Cas systems have been classified 11 and Type IV CRISPR-Cas remains the only type without description of its endogenous activity. Subtype IV-A CRISPR-Cas was first discovered in the genome of Acidithiobacillus ferrooxidans 12 and later shown to contain the signature proteins Csf1 and the crRNP backbone subunits Csf2 (ref. 13 ). Type IV-A systems are usually present on large plasmids (>200 kb) 14 and lack adaptation modules and apparent DNA nucleases such as Cas3 or Cas10 (refs. 13,15 ). Identification of plasmid-borne protospacers targeting conjugative elements suggested that Type IV systems are mainly involved in inter-plasmid competition 13 . The characterization of recombinant Cas proteins of the Type IV-A system of Aromatoleum aromaticum 16 indicated the presence of crRNPs that resemble Type I effectors 16,17 . Heterologous anti-plasmid interference activity was reported for a Type IV-A system of Pseudomonas aeruginosa and shown to require DinG activity 18,19 . However, it is not known how its crRNPs can combat plasmids without an apparent target DNA nuclease. To pinpoint the native activity of a Type IV-A CRISPR-Cas system, we identified and characterized a complete cas gene module (csf1 (cas8-like), csf2 (cas7-family), csf3 (cas5-family), csf5 (cas6-family) and csf4 (dinG)) and a neighbouring CRISPR array on a megaplasmid of Pseudomonas oleovorans. Article https://doi.org/10.1038/s41564-022-01229-2 P. oleovorans DSM 1045 were sequenced using Illumina RNA-seq methodology and sequencing reads were found to map to all three CRISPR arrays ( Fig. 1a and Extended Data Fig. 1). Mature crRNAs were identified to contain 8-nt-long 5′-terminal repeat tags (Type IV: 5′-GUGAGCGG-3′, Type I-E: 5′-AUGAACCG-3′, Type I-F: 5′-CUCAGAAA-3′) that indicate processing of the CRISPR array transcript at the base of hairpin structures
Article https://doi.org/10.1038/s41564-022-01229-2 of the individual repeats. Each CRISPR-Cas system was found to contain a Cas6-family crRNA endonuclease that enables subtype-specific crRNA maturation. Abundance of individual crRNAs was highly variable for different CRISPR arrays, but the large numbers of mature crRNA sequences suggests that all three native systems are active. The Type IV-A CRISPR locus contains 20 spacers, generating 20 crRNAs with different targeting potential. BLAST search analyses 20 identified targets for the spacers of the Type IV-A CRISPR array in transposon and plasmid elements. Alignment of the genomic context of these target protospacers unveiled a consensus 5′-AAG-3′ PAM motif at the 5′ end of the non-targeting strand (Fig. 1c). In addition, we identified two targets of Type I-E crRNAs and both protospacers also exhibited a 5′-AAG-3′ PAM (Supplementary Table 1). Therefore, it is plausible that the Type I-E adaptation module defines the PAM requirements for both systems and compensates for the absence of cas1-cas2 genes in the Type IV-A CRISPR-Cas system.

PAM-dependent Type IV-A CRISPR-Cas activity
We investigated P. oleovorans Type IV-A CRISPR-Cas activity using a plasmid expressing superfolder green fluorescent protein (sfGFP) and different versions of a second pUCP18 plasmid coding for crRNAs that target selected gfp protospacers (Supplementary Table 2). The GFP signal was quantified using flow cytometry and revealed GFP signal reduction for protospacer targets in the gfp coding and non-coding strand in the presence of a 5′-AAG-3′ PAM (Fig. 1d). A protospacer target in the gfp coding strand with 5′-CGG-3′ PAM and a non-targeting crRNA did not result in significant GFP signal reduction, suggesting PAM-dependent CRISPR interference. To follow plasmid curing in this system, P. oleovorans was transformed with target and crRNA-production plasmids and cultivated for 12 h without antibiotic selection for the target plasmid (Supplementary Table 2). The cells were then transferred into a medium with antibiotics to select for the presence of both plasmids. PAM-dependent plasmid curing was observed, indicating that the native Type IV-A crRNP can interfere with plasmid replication (Fig. 1e).
To analyse this targeting mechanism in greater detail, we designed a heterologous E. coli BL21-AI system to produce recombinant P. oleovorans Type IV-A crRNPs. The genes csf1-5 and a minimal CRISPR array consisting of a single spacer-repeat-spacer unit were provided on two plasmids (Supplementary Table 2). Protospacer-containing target plasmids (pCDF-Duet1) were then transformed into E. coli using electroporation and the efficiency of transformation (EoT) was calculated. We observed reduced transformation efficiency only for plasmids carrying a protospacer with a 5′-AAG-3′ PAM, indicating that the recombinant Type IV-A system facilitates PAM-dependent interference (Fig. 1g). Base pairing of PAM sequences and crRNA repeat is used to identify self-targets in Type III CRISPR-Cas systems and to prevent DNA cleavage 18,21,22 . In our system, the -1 position of the 5′-AAG-3′ PAM can base pair with the Type IV-A crRNA repeat. To obtain insights into the PAM specificity in this interaction, we mutated the -2 and -3 PAM nucleotides in the target plasmid and followed changes in the EoT for the different constructs (Supplementary Table 2). We showed that the presence of 5′-GTG-3′ and 5′-AAN-3′ PAM sequences results in significant EoT reduction (Fig. 1g). However, many other PAM sequences without complementarity to the 5′-crRNA repeat tag prevented anti-plasmid activity, suggesting the presence of a PAM selection mechanism that resembles Type I crRNP-mediated protospacer recognition. Larger EoT error bars correlate with highly variable sizes of obtained colonies and might be an indication of competition between plasmid replication and interference in these assays.
Next, we analysed Cas protein mutants in the recombinant system. The putative helicase DinG is found in all Type IV-A CRISPR-Cas systems, but its function is not known 23 . A multiple sequence alignment of several Type IV-A associated DinG enzymes revealed a conserved walker-A motif indicative of ATP-binding (Extended Data Fig. 2). EoT assays showed that a null mutation in the walker-A motif of DinG (K136A) abolished interference, suggesting that ATP-dependent helicase activity of DinG is crucial for CRISPR interference (Fig. 1f,h and Supplementary Table 2). These results agree with the requirement for DinG activity in EoT assays using recombinant P. aeruginosa crRNPs 19 . The function of the signature protein Csf1 is elusive, but it was suggested that this protein might fulfil the crRNP large subunit's role of PAM recognition and R-loop stabilization 24,25 . A conserved cysteine-rich motif of Csf1 was identified and could be part of a zinc-finger structure involved in target recognition 26,27 . EoT assays highlight the functional importance of this motif, as single mutations in each of the cysteine residues abolished CRISPR interference (Fig. 1h).
As some Type IV CRISPR arrays also contain spacers that match viral targets, we evaluated their phage-targeting potential. Therefore, plasmids coding for a crRNA that targets lambda phage geneE next to a 5′-AAG-3′ or a 5′-CGG-3′ PAM were transferred into E. coli producing recombinant Type IV-A crRNPs. Cells were infected with a virulent lambda phage (λvir) and plaque formation was quantified. We observed PAM-dependent immunity against lambda phages (Fig. 1f). The deletion of the dinG gene or the introduction of a null mutation (K136A) in DinG abolished plaque reduction, indicating that DinG is required for this activity.

Self-targeting activity of recombinant Type IV-A crRNPs
It remains unclear whether the observed activities against phages and plasmids require DNA degradation. An earlier report showed that Type I-C CRISPR-Cas activity can be evaluated using lacZ blue-white screening 28 . In this system, white colonies correlate with large genomic deletions around the lacZ gene due to the activity of the DNA nuclease Cas3 (ref. 28 ). We used this Type I-C system activity as a control and designed a comparable set-up to monitor genomic lacZ targeting of our Type IV-A system. Plasmids coding for a crRNA with a spacer base pairing with protospacers in either the coding or non-coding strand of lacZ were transferred into E. coli cells producing Type IV crRNPs (Supplementary Table 2). The presence of a non-targeting crRNA resulted in all blue colonies. For both Type I-C and Type IV CRISPR-Cas systems, the presence of crRNAs with lacZ protospacer targets resulted in comparable amounts of white colonies ( Fig. 2a and Extended Data Fig. 3). Protospacers in both strands and adjacent to the lacZ gene region generated white colonies, providing further support for a DNA targeting mechanism (Extended Data Fig. 4). A DinG null mutant (K136A) did not induce white colony formation (Extended Data Fig. 3a). Next, we PCR-amplified the lacZ gene of selected blue and white colonies. All blue colonies contained intact lacZ. Strikingly, the PCR amplification of 14 white colonies indicated no bands for the Type I-C system and wild-type bands for the Type IV-A system (Fig. 2b,c). We sequenced these PCR amplicons and obtained wild-type sequences, confirming that Type IV-A-mediated CRISPR interference of lacZ did not involve DNA degradation or mutation. In agreement, individual blue and white colonies were picked and shown to be able to revert their phenotype when grown on new plates (Extended Data Fig. 3b).

Self-targeting activity of the native system
We aimed to evaluate the biological relevance of such Type IV-A self-targeting mechanism. The native Type IV-A CRISPR array in P. oleovorans contains with spacer 1 a perfect match for the host chromosome gene pilN, which encodes for a Type IV pilus assembly protein. Therefore, crRNA1 should be self-targeting, which is further supported by the presence of a correct 5′-AAG-3′ PAM sequence next to the pilN target (Fig. 1c). To visualize the scanning of self-targets in the P. oleovorans genome, we fused the mNeonGreen fluorescent protein to Cas6 in both a wild-type (WT) and a Type IV CRISPR array deletion (ΔCRISPR) strain, allowing us to quantify changes in the crRNP dynamics at the single-molecule level. Diffusion patterns of Cas6-mNeonGreen revealed Article https://doi.org/10.1038/s41564-022-01229-2 two molecule populations mainly distributed in the bacterial nucleoid ( Fig. 3a and Extended Data Fig. 5a,b). The distribution of molecules over the nucleoid and the fact that most of the confined molecules (dwell events) in the WT strain are located there, suggest the existence of a self-target region (that is, pilN) (Fig. 3b,c). Similarly, among the mobile molecules, there is an increase of molecules with low diffusion rate in the WT strain (60.2%) compared with 24.8% of molecules in the ΔCRISPR strain, which again might be biologically associated with a longer scanning time of the crRNPs attached to the fluorophore in the WT strain (Fig. 3d). SMTracker 2.0 software 29 was used to estimate the number of crRNPs and we observed an average of 26 crRNPs in a single P. oleovorans cell (Extended Data Fig. 5c,d). Therefore, the presence of a substantial number of Type IV crRNPs in the cells correlates well with the observed RNA-seq data of mature crRNAs (Fig. 1a).
Type IV-A self-targeting at the pilN gene should influence pilN transcript levels. We used RNA-seq to compare the transcriptomes of wild-type P. oleovorans cells and the ΔCRISPR strain. The effective absence of sequence reads mapping to the CRISPR locus verified its successful deletion in the ΔCRISPR strain. We calculated the changes in transcript abundance and observed the strongest increase for transcripts of pilN and neighbouring genes in the ΔCRISPR strain (Fig. 3e,g). Quantitative PCR with reverse transcription (RT-qPCR) verified significantly increased pilN transcript levels in this strain (Fig. 3f). These results suggest that the native pilN operon is downregulated by crRNA1 of the P. oleovorans Type IV-A system. It is possible that the self-targeting of pilN is a result of neutral evolution that is tolerated in the cell as the target is not degraded. In agreement, we did not observe a clear growth advantage for the ΔCRISPR strain. Alternatively, P. oleovorans might even benefit from a reduction of type IV pilus formation in an environment where its presence is not essential and the bacterium might save energy. Type IV pili are involved in the uptake of DNA (natural transformation and bacteriophage infection) and P. aeruginosa type IV pili were shown to bind DNA 30 . Thus, it is intriguing to speculate that a CRISPR-Cas system that evolved to target foreign DNA might benefit from downregulating a cellular component that is involved in the uptake of such foreign genetic material.
Finally, we introduced synthetic crRNAs into wild-type P. oleovorans cells to analyse the efficiency of gene silencing applications based on endogenous Type IV-A CRISPR-Cas activity. A first target was the gene trpE designed to create a tryptophan auxotroph strain (Fig. 4a). Successful trpE silencing was evidenced by nearly four orders of magnitude lower colony numbers after crRNA-trpE induction in minimal medium without tryptophan (Extended Data Fig. 6). As a second target, a crRNA against gene hmgA was shown to stimulate production of the pigment pyomelanin (Fig. 4b). These phenotypes observed upon production of synthetic crRNAs indicate that they can compete with endogenous crRNAs for incorporation into wild-type Type IV-A effectors and enable sequence-specific inactivation of targeted pathways.

Discussion
In conclusion, we showed that the Type IV-A CRISPR-Cas system of P. oleovorans targets DNA in a PAM-dependent manner and can be used to downregulate gene expression in the absence of DNA nuclease activity. The mechanism of this system resembles CRISPRi methodology that uses CRISPR-Cas proteins with inactivated catalytic sites (for example, dead Cas9 (dCas9)) or missing nucleases (Cascade without    Cas3) to stably bind targets and block the transcription machinery [31][32][33] . The heterologous production of P. oleovorans Type IV crRNPs in E. coli yielded white colonies in the presence of all cas genes and a crRNA spacer against the chromosomal lacZ gene (Extended Data Fig. 3a). However, a distinct fraction of cells remained blue, and we also noted that Type IV CRISPR-Cas activity sometimes generated heterogenous colonies with distinct blue and white patterns (Extended Data Fig. 3b). CRISPRi activity would be expected to regulate lacZ expression in uniform fashion across the population and it remains to be established why some cells escape lacZ targeting. Possible explanations include plasmid copy number variations for the cas gene and crRNA expression plasmids, or the presence of feedback mechanisms.
To exclude undesired effects of heterologous crRNP production, we induced gene regulation for the native P. oleovorans Type IV CRISPR-Cas system. Here, the addition of synthetic crRNAs resulted in Type IV-A CRISPR-mediated silencing of target genes in the host chromosome. Therefore, we propose that endogenous Type IV-A CRISPR-Cas systems (for example, those found in many pathogenic Pseudomonas and Klebsiella strains) can be utilized for convenient regulation of host gene expression by providing engineered crRNA sequences.  [34][35][36] ). Data analysis, coverage plots and scatterplots were generated using the R packages ggplot2 3.3.6 and DEseq2 v1. 36.0 (refs. 37,38 ). Mapped reads were visualized and inspected using IGV 2.13.2 (ref. 39 ).

Conjugation of P. oleovorans
Genetic constructs for genome editions in the P. oleovorans strains were delivered via conjugation following an adapted protocol from ref. 40   Article https://doi.org/10.1038/s41564-022-01229-2 maintenance of the E. coli strain during the first step of conjugation. The absence of DAP allowed the elimination of E. coli and visualization of only P. oleovorans transformed cells. For the first event of conjugation, 1 ml and 0.5 ml of a fresh overnight culture of P. oleovorans and E. coli WM3064, respectively, were collected separately by centrifugation at 8,000 × g for 3 min at room temperature. Cell pellets were washed twice with LB supplemented with 0.3 mM DAP (LB-DAP) and finally resuspended in a total volume of 100 µl of LB-DAP. The whole suspension was pipetted as a single drop onto an LB-DAP agar plate. Plates were incubated for 5-7 h at 37 °C. After incubation, cells were collected by adding 2 ml of plain LB medium to the plate and scraping the agar with an inoculation loop. Cells were washed two times with 1 ml of plain LB medium to remove traces of DAP. Finally, cells were resuspended in 1 ml of LB medium, and serial dilutions of 10 −1 and 10 −2 were plated onto agar plates supplemented with 50 µg ml −1 kanamycin or 30 µg ml −1 gentamicin. The P. oleovorans strain before conjugation was considered as a negative control. Plates were incubated at 37 °C for 36 h.

Generation of P. oleovorans gene knock-in and knock-out strains
Insertions and deletions of specific elements in the P. oleovorans genome were carried out following the adapted protocols for endonucleases-mediated recombination 40,41 . The suicide vector pEMG was used for delivery of genome-specific regions flanking the place of the desired insertion or deletion of parts. The first construct was designed by fusing mNeonGreen fluorescent protein (mNeonGreen FP) to the C terminal of the native Cas6 protein in the Type IV-A CRISPR-Cas of P. oleovorans. In this construct, mNeonGreen gene flanked by ~500 bp of homologous region upstream and downstream of the insertion site of interest (that is, between csf5 and csf1) was cloned into the pEMG suicide vector between BamHI and EcoRI restriction sites. For doing multiple insertions or deletions in the same strain, the helper plasmid pSEVA6213S was used for digestion of the suicide vector pEMG. A second construct was designed to further knockout elements of the native system including the CRISPR array. In this case, two fragments of ~500 bp of homologous regions upstream and downstream of the CRISPR array were cloned into the pEMG vector between BamHI and EcoRI restriction sites. A list of plasmids used in this work is found in Supplementary Table 4.

Transformation efficiency assays
E. coli BL21-AI cells were transformed with plasmids enabling production of recombinant Type IV-A crRNPs. A pETDuet-1 plasmid contained all five Type IV-A cas genes. Individual mutants (ΔDinG, DinG K136A, ΔCsf1, Csf1 C30A, Csf1 C33A, Csf1 C66A or Csf1 C69A) were created via Quikchange mutagenesis. A second plasmid, pRSFDuet-1 carried a minimal CRISPR array with crRNA1 from P. oleovorans. E. coli cells producing Type IV-A crRNP variants were then transformed with a target pACYCDuet-1 vector carrying a perfectly complementary protospacer1 to crRNA1 with a 5′-AAG-3′ PAM. A pCDFDuet-1 vector with a proto-spacer1 and a 5′-CGG-3′ PAM or a non-base-pairing protospacer served as controls. Transformation efficiency was calculated with the formula: Transformation efficiency = c.f.u. (sample)/c.f.u. (non-matching spacer control). To identify functional PAM elements, protospacer1 sequences were synthesized with different 3 nt PAM combinations and cloned into vector pCDFDuet-1. These vectors were transformed into E. coli BL21-AI cells containing recombinant wild-type Type IV-A crRNPs and the transformation efficiency was recorded as described above.

FACS measurements
P. oleovorans cells harboured a pHERD30T vector expressing sfGFP and a pUCP18 vector with a minimal CRISPR array. Different constructs contained spacers targeting the coding strand of gfp with a 5′-AAG-3′ PAM or a 5′-CGG-3′ PAM, the non-coding strand of gfp with a 5′-AAG-3′ PAM, or a non-targeting crRNA (Supplementary Table 2). Individual colonies were cultivated in LB medium overnight at 37 °C. The cultures were washed twice and diluted 100 times with phosphate-buffered saline solution. Fluorescence intensity measurements were conducted using a BD Fortessa flow cytometer and GFP was excited by the 488 nm laser line. For each sample, 10,000 events were recorded and the ungated average fluorescence intensity of each measurement was recorded. Data were analysed with BD FACSDiva 8.0.1 Bacteriophage plaque assay E. coli BL21-AI cells producing all Type IV-A Cas proteins were transformed with a pCDFDuet-1 plasmid carrying a minimal CRISPR array, with a spacer targeting the coding or non-coding strand of geneE of the virulent λvir variant of phage lambda. A pCDFDuet-1 vector with a minimal CRISPR array carrying a random spacer sequence was transformed in control experiments. Individual colonies were inoculated in LB medium for overnight incubation and 100 µl of the overnight culture was pre-incubated with 10 µl of lambda phage (titre 1.2 × 10 7 plaque-forming units per ml). After incubation for 10 min, cultures were mixed with 3 ml of selective 0.7% soft LB agar supplemented with 2 mM MgCl 2 . The mixture was transferred onto a plain LB agar plate containing 1 mM IPTG, spectinomycin (100 µg ml −1 ) and ampicillin (100 µg ml −1 ). Cells and phages were co-incubated at 30 °C for 10 h and plaques were counted.

Gene silencing assays
(1) Genomic lacZ targets in E. coli: E. coli BL21-AI cells producing all Type IV-A Cas proteins expressed in pETDuet-1 were transformed with a pCDFDuet-1 vector containing a minimal CRISPR array with a spacer targeting the coding or non-coding strand of lacZ (Supplementary  Table 2) or a pCDFDuet-1 vector carrying a random sequence. As control, a pCas3cRh vector encoding a complete Type I-C CRISPR-Cas system and a crRNA targeting lacZ were used as previously detailed 28 . After transformation, cells were transferred onto LB agar plates containing 0.005% X-gal, 0.2% arabinose and 1 mM IPTG. After overnight incubation at 37 °C, images of plates were captured and analysed with OpenCFU 3.8-BETA 42 . For recognition of white colonies, the colour filter was set to a hue angle of 0-80.
(2) Genomic targets in P. oleovorans: pHERD30T plasmids with a Type IV-A repeat-spacer-repeat construct under the control of an araBAD promoter were conjugated into wild-type P. oleovorans cells as described above. The crRNA spacer sequences are provided in Supplementary Table 2. Cells producing a crRNA targeting trpE (anthranilate synthase component I) were grown overnight in M9 minimal medium (Na 2 HPO 4 x 2H 2 O, 5.6 g l −1 ; KH 2 PO 4 , 3 g l −1 ; NaCl, 0.5 g l −1 ; NH 4 Cl, 1 g l −1 ; 2 mM MgSO 4 ; 0.1 mM CaCl 2 ; 0.4% (w/v) glucose) supplemented with 0.2% arabinose for crRNA induction, and absorbance at OD 600 was monitored to analyse tryptophan auxotrophy. For complementation, 0.6 mM tryptophan was added to the medium. For the plated assay, cells were grown overnight in LB, washed with M9 minimal medium or tryptophan-complemented M9 medium. The cell suspensions were adjusted to the same OD 600 and plated onto M9 (+tryptophan) agar plates additionally containing 0.2% arabinose and selective antibiotic. After incubation at 37 °C for 2-3 d, the c.f.u. ml −1 was determined from the colony number of countable dilutions. Cells producing a crRNA targeting hmgA (coding for homogentisate 1,2-dioxygenase) were grown in LB medium for 48 h, and absorbance at OD 400 of the supernatant was monitored to quantify pyomelanin accumulation in the medium.

RT-qPCR of wild-type and ΔCRISPR array P. oleovorans strains
Primer efficiency test (PETest) was performed for the primers designed for pilN and the housekeeping gene recA from P. oleovorans using a 3-log dilution series of complementary DNA from WT P. oleovorans. RT-qPCR was performed on a CFX Connect real-time PCR detection system (Bio-Rad) using iQ SYBR Green Supermix (Bio-Rad). Amplification conditions were as follows: denaturation at 95 °C for 3 min Article https://doi.org/10.1038/s41564-022-01229-2 followed by 40 cycles of 95 °C for 30 s, and 60 °C for 30 s followed by an additional final extension of 55 °C for 5 s and 95 °C/0.5 °C. Three technical replicates were used per dilution, and five serial concentrations were considered for constructing the linear equation and finding the regression coefficients (R 2 ). Adequate primers with a PETest of 90-110% and R 2 > 0.7 were selected (Supplementary Table 5). The reaction mix consisted of 1X iQ SYBR Green Supermix (Bio-Rad), 1 µl 1:50 cDNA, 0.5 µM of each primer (0.5 µl of primer mix) and adjustment with nuclease-free water to a final volume of 10 µl. Three technical replicates were used for each biological replicate (that is, four different colonies per treatment) following the RT-qPCR protocol described previously. Gene expression analysis was performed by normalizing the internal control as the cycle threshold (Ct) values of the housekeeping gene recA and calculating the relative transcript level of the pilN gene using the 2 −ΔΔCT method 43 .

Single-molecule microscopy of mNeonGreen-Cas6 tagged Type IV-A crRNPs
To investigate the intracellular dynamics of the Type IV-A crRNP complex in its native system and in a system lacking the CRISPR array, mNeonGreen FP was fused with the Type IV-A Cas6 protein, which is a stable part of the Type IV-A crRNP complex 16 . A wild-type P. oleovorans expressing mNeonGreen in pHERD30T plasmid was used as a control (Supplementary Table 4). P. oleovorans cultures were grown until OD 600 ~0.3 and prepared for single-molecule microscopy following sample preparation, in an Olympus IX71 microscope 44 with a ×100 objective (UAPON 100×OTIRF; numerical aperture, 1.49; oil immersion). In general, this microscopy technique relies on strong excitation of the fluorophores, in this case mNeonGreen, followed by rapid bleaching, which allows tracking of a few unbleached molecules. Image acquisition was performed continuously during laser excitation with the electron-multiplying CCD camera iXon Ultra (Andor Technology). For each movie, 2,500 frames were taken at an acquisition time of 20 ms. Further processing was done with Oufti 1.2 (ref. 45 ) to set the cell meshes. Track generation was performed with a minimum track length of five steps U-track 2.2.1 (ref. 46 ). Bleaching curves were analysed in ImageJ 2.0 to verify single-molecule observations. Analytical evaluation was carried out using the SMTracker 2.0 software 29 . To determine the presence or absence of signals corresponding to single molecules and the number of Cas6-mNeonGreen molecules per cell, the photon count of single bleaching steps was quantified (that is, single-fluorescent-protein bleaching) towards the end of the acquisitions, and the total fluorescence intensity at the beginning of the acquisition was divided by the fluorescence intensity of single fluorophores, relative to cell size. SMTracker 2.0 was used to automatically determine background signals in individual cells and to subtract these from the specific point spread functions from single molecules 29 . Estimation of the Cas6 protein copy number was based on the single-molecule-tracking pipeline 29 .

Statistics and reproducibility
RNA-seq, RT-qPCR and Type IV-A CRISPR-Cas activity assays (EoT assays, plasmid curing and phage assay) were performed in triplicate (n = 3 biologically independent samples; n = 4 for RT-qPCR); all attempts at replication were successful. No sample size calculation was performed and no data were excluded during the analyses. The experiments of this study compare different bacterial cells with defined plasmid sequences and genotypes. A selection bias should not affect the results. Statistical analyses and P values for all experiments were obtained using an unpaired two-sided t-test (and F-test for the RT-qPCR experiment).

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data are available in the manuscript or the Extended Data files. Illumina sequence data generated in this study have been deposited in the NCBI Sequence Read Archive database under project ID PRJEB48544. Raw data from single-molecule microscopy analyses are provided at https://doi.org/10.6084/m9.figshare.20359071. Source data are provided with this paper. Article https://doi.org/10.1038/s41564-022-01229-2 Extended Data Fig. 6 | Gene silencing assay for trpE in P. oleovorans plated on minimal medium agar. The control strain is carrying an empty pHERD30T plasmid instead of pHERD30T coding for the crRNA targeting trpE. Experiments were performed in triplicates (n = 3 biologically independent samples) and four dilutions (10 −2 -10 −5 ) were plated, respectively. Statistical analysis was performed using an unpaired t-test. Data are presented as mean values +/−SD with a p = 0.0069 (**).

March 2021
Field-specific reporting Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences
Behavioural & social sciences Ecological, evolutionary & environmental sciences For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative. The axis labels state the marker and fluorochrome used (e.g. CD4-FITC).
The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysis of identical markers).
All plots are contour plots with outliers or pseudocolor plots.
A numerical value for number of cells or percentage (with statistics) is provided.

Methodology
Sample preparation Instrument Software the single molecule microscopy experiments are available at https://doi.org/10.6084/m9.figshare.20359071. Additional extended data and source data are provided.
Experiments were performed in triplicate (n=3), based on similar studies in the field which generated reproducible results. No statistical methods were used to determine the sample size.
No data were excluded.
Experiments (RNA-seq, Type IV-A CRISPR-Cas activity assays) were performed in triplicate, except qRT-PCR which was performed with 4 biologically independent colonies. All attempts at replication were successful.
The experiments of this study compare different bacterial cells with defined plasmid sequences and genotypes. A selection bias does not affect the results in these studies. Therefore, randomization is not needed in this experimental design.
Investigators were not blinded as blinding is not applicable in these studies.