Cas1–Cas2 physically and functionally interacts with DnaK to modulate CRISPR Adaptation

Abstract Prokaryotic Cas1–Cas2 protein complexes generate adaptive immunity to mobile genetic elements (MGEs), by capture and integration of MGE DNA in to CRISPR sites. De novo immunity relies on naive adaptation—Cas1–Cas2 targeting of MGE DNA without the aid of pre-existing immunity ‘interference’ complexes—by mechanisms that are not clear. Using E. coli we show that the chaperone DnaK inhibits DNA binding and integration by Cas1–Cas2, and inhibits naive adaptation in cells that results from chromosomal self-targeting. Inhibition of naive adaptation was reversed by deleting DnaK from cells, by mutation of the DnaK substrate binding domain, and by expression of an MGE (phage λ) protein. We also imaged fluorescently labelled Cas1 in living cells, observing that Cas1 foci depend on active DNA replication, and are much increased in frequency in cells lacking DnaK. We discuss a model in which DnaK provides a mechanism for restraining naive adaptation from DNA self-targeting, until DnaK is triggered to release Cas1–Cas2 to target MGE DNA.


INTRODUCTION
Prokaryotes utilize specialised chromosomal sites called CRISPRs (Clustered Regularly Interspaced Short Palindromic Repeats) and Cas (CRISPR-associated) proteins to provide adapti v e immunity against mobile genetic elements (MGEs). Immunity is generated by CRISPR 'adaptation' ( 1 , 2 ), which depends on the Cas1-Cas2 protein complex to capture fragments of MGE DNA (or RNA) and integrate them into a CRISPR as 'spacers' (3)(4)(5)(6). Cas1-Cas2 captures DNA fragments that are defined by length and end sequences called Protospacer Adjacent Motifs (PAMs) (5)(6)(7)(8)(9). (Cas1) 4 -(Cas2) 2 complexes bind pre-spacer DNA in two Cas1 acti v e sites held either side of the Cas2 dimer, defining the distance between acti v e sites and ther efor e DNA fr agment length. Integr ation of ca ptured DN A, without its PAM sequence, as a spacer along with synthesis of one new repeat per spacer, establishes immunity that is deli v ered by CRISPR 'interfer ence' r eactions. Interfer ence depends on transcription of CRISPR to RN A, w hich is cleaved within the r epeat r egions into crRNAs r epr esenting a single spacer, and which are bound into interfer ence complex es (Cascade-Cas3 in E. coli ) ( 10 ). These survey DNA for PAMs, 'locking' into an R-loop where a PAM is at MGE sequence complementary to the crRNA, triggering nuclease destruction of the MGE (11)(12)(13)(14). PAMs ther efor e provide functional coupling of adaptation with interference, for effecti v e immune responses.
Interaction of Cas1-Cas2 with interference nucleases temporall y and spatiall y targets ada ptation to MGE DN A, providing new DNA fragments for capture (15)(16)(17). This is 'primed' adaptation ( 17 , 18 ), which ther efor e r elies on pre-existing CRISPR immunity that has already generated spacer-crRNAs. But if there is no pre-existing immunity 'nai v e' adaptation by Cas1-Cas2 generates immunity de nov o, by tar geting MGE DN A independentl y from interference nucleases by mechanisms that are unclear. When Cas1-Cas2 is ov er-e xpressed in cells ectopically (e.g. from an inducible plasmid) in the absence of interference complexes it readily deri v es ne w spacers from the host chromosome. This is in accord with the PAM pr efer ence of Cas1-Cas2 (ATG in E. coli ) for sequences abundant across host chromosomes and MGEs, but does not provide for targeting of MGE DNA ( 19 ). Host proteins that assist DNA capture by Cas1-Cas2, including RecBCD helicase (20)(21)(22), and RecJ, DnaQ and Cas4 nucleases (23)(24)(25), do not appear to contribute to Cas1-Cas2 distinguishing MGE DNA as a target ( 20 , 26 ). We report multiple lines of evidence indicating that in E. coli the widely conserved 'hub' chaperone DnaK (Hsp70) ( 27 , 28 ) regulates nai v e adaptation by restraining Cas1-Cas2. This protects the host chromosome from targeting by Cas1-Cas2. We show that inhibition of adaptation can be re v ersed by mutation of DnaK and by expression of MGE pr otein. This may pr ovide DNA target selection to MGEs, when Cas1-Cas2 is released from DnaK that is recruited by MGE proteins.

Strains, plasmids and media
Esc heric hia coli strains are described in the Supplementary  Table S7. We generated an E. coli dnaK strain by recombineering ( 29 ) to insert kanamycin resistance followed by P1 vir transduction into BW25113 ( 30 ). The dnaK phenotype was confirmed via plaque formation and temperature sensitivity tests. Cells were grown at 37 • C in LB broth (10 g / l bacto-tryptone, 5 g / l yeast extract, 10 g / l NaCl) and on LB agar plates (supplemented with 15 g of agar / L for solid media) unless otherwise stated. Antibiotics were added to LB plates at final concentr ations: tetr acycline 10 g / ml, ampicillin at 100 g / ml, and chloramphenicol at 34 g / ml. Plasmids are detailed in Supplementary Table S8. Briefly, pBad-HisA (Invitro gen) was used for expression of Cas1-Cas2 under control of arabinose inducible ar aB AD promoter as in previous studies ( 30 ).The plasmid pACY-Cduet (Novagen) was used for expression of DnaK and other proteins, each under control of the IPTG inducible T7 promoter.

BioID2 identification of Cas1-Cas2 interactor proteins
E. coli strain EB377 (Table S7) was transformed with pCas1 BioID2 -Cas2 or pBioID2 and grown on ampicillin agar. Individual colonies were used to inoculate LB medium supplemented with ampicillin and 0.2 (w / v) % L-arabinose and grown for 18 h to provide starter cultures. These starter cultur es wer e used to inoculate LB supplemented with 0.2% (w / v) L-arabinose. Cells were grown for 60 min prior to harvesting at 4000 × g for 5 min. Biomass was washed three times with 10 ml of 1 × PBS and then resuspended in 1 ml of Lysis Buffer (50 mM Tris pH 7.5, 150 mM NaCl, 0.4% SDS, 1% Nonidet P40, 1.5 mM MgCl 2 ) for lysis by sonication and clarification by centrifugation at 16 000 × g for 15 min. Lysates were incubated with Pierce ™ High-Capacity Streptavidin Agarose Beads (Thermo Scientific ™) overnight at 4 • C with gentle agitation. Beads were washed three times in 1 × PBS to remove unbound protein, and sent to the Cambridge Centre for Proteomics for analysis using a 120min LC-MS / MS run, with the resulting raw data available accompanying this work (Table S1). The NSAF value for each protein was determined as the number of spectral counts (SpC) identifying a protein, divided by the protein's length (L), divided by the sum of SpC / L for all protein in the experiment ( 31 ).

In-vivo co-expression and pull down of a DnaK-Cas1 complex
Strain EB377 was co-transformed with p His DnaK and pCas1 Strep -Cas2, or respecti v e empty v ector controls, and grown on agar with antibiotic selection. Overnight cultures containing antibiotics were inoculated with single colonies and grown for 18 h to provide starter cultures for growth to OD 600 of 0.6 prior when cells were treated with 0.2% (w / v) L-arabinose and IPTG to 1 mM to induce Cas1, Cas2 and DnaK protein expression. After 3 h cells were harvested and resuspended in 1 ml of Pull-Down Buffer (20 mM Mops pH7, 200 mM NaCl) supplemented with phen ylmethylsulf on yl fluoride (PMSF) (0.1 M final concentration). Cells were lysed by sonication and clarified by centrifuga tion a t 16 000 × g for 30 min, and lysa te incubated for 60 min at 4 • C with gentle agitation following addition of 50 l of Iminodiacetic acid Sepharose ® (Merck) pre-charged with NiCl 2 . Samples washed 3 × 1 ml of Pull-Down Buffer with 50 mM imidazole, were heat treated in 1 x SDS Dena tura tion Buf fer (50 mM Tris pH 6.8, 2% SDS, 10% Glycerol, 0.1 M DTT, 6 M urea, 0.5 M imidazole, bromophenol blue). Sample separation used a 12.5% SDS gel and transferred onto Amersham ™ Hybond ™ P 0.2 PVDF membrane (Cytivia ™). Standard western blotting methods w ere follow ed with membrane blocked in Blocking Buffer (3% milk po w der in 1 × TBS-Tween). Primary antibodies: Mouse 6 ×-His Tag Monoclonal Antibody (HIS.H8), Biotin (Invitrogen ™) and Mouse Anti-Strep-tag II mAb Monoclonal Antibody (MBL ® ), and secondary antibodies: Goat Anti-Biotin HRP-linked, and Goat anti-Mouse IgG (H + L) Secondary Antibody HRP, all added at 1:2000 dilution in blocking buffer. The membrane was treated with ECL Western Blotting Substrate (Promega ™) and imaged using a LAS-3000 mini (FUJIFILM ™).

Naiv e adaptation assa ys and corr esponding measur ement of plasmid instability and cell viability
Nai v e adaptation assays were based on the procedure described in ( 1 ). E. coli EB377 cells transformed with plasmid vector lacking Cas1-Cas2 (pControlA), pEB628 (pCas1-Cas2) or pTK145 (pCas1R84G-Cas2) were inoculated into 5 ml of LB and aerated at 37 • C for 16 h in LB containing 0.2% (w / v) L-arabinose, and then sub-cultured ('passaged') by diluting 1:300 into fresh LB again supplemented with 0.2% (w / v) L -arabinose. Cells were harvested at identical time points and genomic DNA extracted using a Gene-JET Genomic DNA Purification Kit (Thermo Scientific ™). Spacer acquisition was monitored by PCR, utilizing 10 ng of genomic DNA and primers SW1 and SW2 (Supplementary Table S9), with products separated using a 1.25% agarose gel stained with ethidium bromide and imaged using a U:Genius3 (Syngen Biotech). For nai v e adaptation assays expressing Cas1-Cas2 alongside other proteins, strain EB377 was co-transformed with pEB628 and pACYC plasmids containing the gene of interest. Cells were passaged as described above and stopped in P2 at OD 600 0.4, with spacer acquisition monitored by PCR as described above. Analysis of acquisition was carried out from three independent replicates, with band quantification of PCRs carried out with ImageJ ( 32 ). Plasmid instability during nai v e adaptation assays was analysed by comparison of cell viability on LB and antibiotic selection plates. At the end of each 'passage' samples were taken and serially diluted in 1 × M9 Minimal Salts. 10 l of each dilution was spotted onto LBagar plates with and without selection and grown overnight prior to colony quantification.

Protein purification
Cas1 was purified with a C-terminal StrepTag ® II (Cas1 Strep ), Cas2 with an N-terminal StrepTag ® II ( Strep Cas2), and DnaK proteins with an N-terminal hexahistidine tag ( His DnaK). Individual transformants of BL21-AI cells containing the relevant plasmid were used to pr epar e fr esh overnight cultur es which wer e subsequently diluted 1:100 in 3 l of LB supplemented with selection marker. Cells were grown to an OD 600 of 0.6 and protein expression induced by addition of IPTG (1 mM) and L-arabinose (0.02% w / v). After 3 h cells were harvested and resuspended in Buffer A (20 mM Tris pH 7.5, 150 mM NaCl and 10% glycerol) supplemented with PMSF to 0.5 mM. Cas1 Strep and Strep Cas2 were individually loaded onto 5 ml Strep-Avidin ™ XT Superflow ™ High-Capacity Cartridges (IBA Life Sciences GmbH) in Buffer A before being eluted via an isocratic elution in Buffer A supplemented with 50 mM biotin. Cas1 Strep was further purified using a 1 ml HiTrap Heparin Hp (Cytivia ™) in Buffer A and eluted in a gradient of 0.15-1 M NaCl. Cas1 Strep containing fractions were loaded onto a HiLoad 16 / 600 Super de x 200 pg (Cytivia ™) equilibrated in Buffer A. Cas1 C-Strep was concentrated using a VivaSpin ® 6 10 kDa cut-off centrifugal concentrator (Sartorius) prior to storage at -80 • C. Strep Cas2 was further purified by loaded onto a 1 ml HiTrap Q XL column (Cytivia ™) and collected in the flow through before being dialysed overnight at 4 • C against 20 mM Tris pH 7.5, 150 mM NaCl and 25% glycerol prior to storage at -80 • C.
His DnaK and mutant variants were loaded onto a 5 ml HiTrap Chelating HP (Cytivia ™) and washed with Buffer B (20 mM Tris pH 7.5, 500 mM NaCl, 20 mM imidazole, 10 mM MgCl 2 and 5 mM ATP and 10% glycerol), before elution in Buffer A using a gradient of 20-500 mM imidazole. His DnaK containing fractions were dialysed overnight a t 4 • C against Buf fer C (50 mM Tris pH 7.5, 100 mM NaCl, 1 mM DTT and 10% glycerol) before loading a 5 ml Hi-Trap Q HP (Cytivia ™) in Buffer C and eluted with a gradient of 100-1000 mM NaCl with His DnaK containing fractions dialysed overnight at 4 • C in 50 mM Tris pH 7.5, 50 mM NaCl, 1 mM DTT and 20% glycerol prior to storage at -80 • C.

In-vitro spacer integration assays
Spacer integration (SpIn) assays were carried out in a reaction buffer containing 20 mM HEPES-NaOH pH 7.5, 25 mM KCl, 10 mM MgCl 2 , 1 mM DTT and 0.1 mg / ml BSA. Reactions wer e pr epar ed to a final volume of 10 l, with the relevant concentrations of proteins used in each assays indicated in the relevant figure or figure legend. Cas1 Strep , or Cas1 Strep and Strep Cas2, were left to incubate on ice for 10 min in reaction buffer prior to addition of His DnaK, followed by an additional 5 min incubation on ice. Reactions were initiated by addition of 20 nM Cy5-labelled Pre-spacer (formed from annealed TK24 and TK25, Supplementary Table S10) and 100 ng of supercoiled pCRISPR (Supplementary Table S8). Reactions were immediately transferred to 37 • C for 60 min, before being quenched by addition 1 l of Stop Buffer (0.2 mg / ml proteinase K, 2% SDS and 100 mM EDTA), and left for a further 60 min at 37 • C. 1 × DNA loading dye (2.5% Ficol 400, 3 mM Tris pH 8.0, 10 mM EDTA, 0.08% SDS, Orange G) was added to samples prior to loading onto a 1.5% agarose TAE gel and left to migrate for 60 min at 120 V before imaging on a Typhoon ™ laser scanner platform (Cytivia ™). All gel images were processed using ImageJ.

Electrophoretic mobility shift assays
Cas1 and DnaK proteins were diluted to working concentrations in 20 mM Tris pH 8.0, 100 mM NaCl and 1 mM DTT prior to addition. Cas1 was preincubated for 5 min with 20 nM Cy5 labelled DNA fork substrate and 20 mM Tris pH 8.0, 0.1 mg / ml BSA, 7% Glycerol. DnaK was added to the reaction followed by glycerol to 25% and left for a further 25 min at 37 • C. Samples were loaded onto a 5% nati v e acrylamide gel and left to migrate for 90 min before imaging on a Typhoon ™ laser scanner platform. All gel images were processed using ImageJ.

Single-image microscopy
Fresh overnight cultures of strains of interest were diluted 100-fold in fresh LB broth supplemented with Nucleic Acids Research, 2023, Vol. 51, No. 13 6917 ampicillin (50 g / ml) if r equir ed and incubated with vigorous aeration at 37 • C until A 600 reached 0.2. L -Arabinose (Sigma) was added to a final concentration of 0.1% and the culture incubated 60 min for protein expression and matura tion. If necessary, the DNA d ye Hoechst 33342 (Invitrogen ™) was added to a final concentration of 200 ng / ml, incubated for 5 min at room temperature and imaged without washes. 1 l of the culture was pipetted onto an agarose pad and air-dried. For generation of pads a 65 l (15 × 16 mm) GeneFrame (Thermo Scientific ™) was added to a conventional microscopy slide. 1% of SeaK em LE agar ose (Lonza) was added to 1 × M9 minimal medium (diluted from a 5 × stock, Sigma-Aldrich) and heated until the agarose was completely dissolved. 95 l of the solution was added into the GeneFrame chamber and the chamber sealed immediately with a conventional microscopy slide. Once set, the top slide was removed and the agarose pad air-dried for no more than 5 min at 37 • C and used immediately. Once the sample was added and air-dried the GeneFrame chamber was sealed by adding a 22 × 22 mm cover slip. Visualisation was by using a T i -U inverted microscope (Nikon) with a CFI Plan Fluor DLL 100 × objecti v e (Nikon) and an ORCA Flash 4.0 LT plus camera (Hamamatsu). Phase contrast images were taken using a pE-100 single LED wavelength source (CoolLED). For fluorescence the pE-4000 illumination system (CoolLED) was used. The relevant filters for visualisation of DAPI, eYFP and mCherry were Nikon DAPI-50LP-A, Zeiss filter set 46 (eYFP), as well as Nikon TXRED-A-Basic Filter (mCherry). Images wer e captur ed using the NIS Elements-BR softwar e V4.51 (Nikon) and exported to tiff. Postprocessing, such as cropping and rotating, was performed in Adobe Photoshop CC (V23.0.0).

Time-lapse microscopy
Fr esh overnight cultur es of strains of inter est wer e diluted 100-fold in fresh M9 with 0.4% (v / v) glycerol, supplemented with ampicillin (50 g / ml) if r equir ed. M9 minimal medium was used for time-lapse experiments to avoid the autofluorescence typical for LB broth, which requires longer exposure times and consequently much more rapid photobleaching of the fluor ophores. Glycer ol was used as carbon source because glucose would r epr ess the arabinosecontr olled pr omoter. Cultur es wer e incubated with vigorous aeration at 37 • C until OD 600 reached 0.2. L -Arabinose (Sigma) was added to a final concentration of 0.1% and the culture incubated 60 min for protein expression and ma tura tion. 1 l of the sample was pipetted onto an agarose pad and air-dried. Pads were generated as described above. Once the sample was added and air-dried the Gene-Frame chamber was sealed by adding a 22 × 22 mm cover slip. Cells were visualised using the T i -U system described above. The temperature was maintained at 37 • C using an environmental chamber (Digital Pixel). Time-lapse stacks wer e captur ed using the NIS Elements-BR softwar e V4.51 (Nikon) and either exported to a single mp4 file or individual tiff files. Postprocessing of tiff images, such as cropping and rotating, was performed in Adobe Photoshop CC (V23.0.0).

DnaK physically interacts with Cas1 and inhibits naive adaptation by Cas1-Cas2 in E. Coli
We in vestigated f or interactors of Cas1-Cas2 in replicating E. coli cells by using BioID2, a proximity-dependent biotin protein labelling technique ( 33 ) that we adapted for E. coli . Cas1 was fused at its C-terminus via a tri-peptide repeat (GGS) 8 linker to the biotin-protein ligase R40G mutant from Aquifex aeolicus , under inducible control and alongside Cas2 in the same plasmid (Figure 1 A). The resulting protein complex (Cas1 BioID2 -Cas2) catalysed nai v e adaptation in E. coli cells (Figure 1 B). Proteins entering physical proximity to induced Cas1 BioID2 -Cas2 were biotinylated in cells growing in 50 nM biotin (summarised in Figure 1 c), providing a biotinylated proteome differing from control cells expressing only the biotin-protein ligase (Supplementary Figur e S1a). Str eptavidin extraction of the biotinylated pr oteins fr om these pr oteomes, followed by peptide mass fingerprinting and normalised spectral abundance factor (NSAF) analysis of these proteins ( 31 ) Table S1), compared with the control. This suggested physical interaction of DnaK with Cas1 BioID2 -Cas2, consistent with a previous E. coli proteomics study that identified physical interaction of DnaK with YgbT protein, now called Cas1 ( 28 ). We further validated this physical interaction by co-expressing His DnaK and Cas1 Strep -Cas2 from plasmids in E. coli cells, alongside controls expr essing str ep-ta gged or (His) 6  To assess whether DnaK modulated nai v e adaptation in E. coli --which depends on Cas1-Cas2 --we measured acquisition of new DNA spacers into the chromosomal CRISPR-1 locus. Cas1-Cas2 was inducibly over-expressed from a plasmid, to overcome repression of chromosomal Cas1 ( ygbT ) by H-NS ( 1 , 6 , 21 , 30 , 34 ). In nai v e adaptation assays Cas1-Cas2 captures DNA from the chromosome ( 21 , 22 ) and integrates it as spacers, observed as expansion of the chromosomal CRISPR-1 site, summarised in Figure 1 f. Nai v e adaptation was detectab le in passages two and three of E. coli cell growth (P2 and P3) observed across optical densities compared with control cells containing plasmid lacking Cas1 ( ygbT ) and Cas2 ( ygbF ) (Figure 1 G,  lanes 5 and 6). Cells expressing Cas1 R84G -Cas2 complex, in which Cas1 cannot bind to DNA ( 30 ), gave no detectable nai v e adaption (lanes 7-9), confirming nai v e adaptation dependent specifically on plasmid Cas1-Cas2. Cell viabilities were similar in P2 for each E. coli cell type grown in ampicillin or not (Table S3), confirming that plasmid instability or cell death is not responsible for observed differences in spacer acquisition (Supplementary Figure S1c  and d). Co-expressing DnaK alongside Cas1-Cas2 inhibited nai v e adaptation in P2, compared with cells e xpressing Cas1-Cas2 alongside the empty plasmid expression vector for dnaK ( Figure 1 H and I). HtpG, a chaperone that  Table S4 shows the raw data measurements of spacer acquisition. modulates Cas3 in E. coli ( 35 ) , and SecB, which has no connection with CRISPR systems, did not inhibit nai v e adaptation ( Figure 1 H and I). We conclude that DnaK and Cas1 physically interact in E. coli cells, and that DnaK inhibits nai v e adaptation by Cas1-Cas2.

Naive adaptation is released from inhibition by mutation or deletion of DnaK, and by expression of MGE proteins
DnaK binds to and releases proteins from its C-terminal Substrate Binding Domain (SBD), triggered by allosteric modulation of its N-terminal ATP hydrolysis domains, summarised in Figure 2 A, re vie wed in ( 27 ). By introducing mutations that deactivate either the SBD (DnaK S427P or DnaK N451K ) or ATPase (DnaK E171A ) sites ( 36 ) and coexpressing the mutant proteins with Cas1-Cas2 we investigated how DnaK inhibits nai v e adaptation (Figure 2 b and  c). DnaK E171A was as effecti v e as DnaK + , but DnaK S427P or DnaK N451K released nai v e adaptation from inhibition (Figure 2 Figure S1e) and western blotting confirmed that each DnaK mutant protein was detectably expressed similarly to DnaK + during the assays (Supplementary Figure S1e).

B and C). Viability of cells expressing each mutant were similar (Supplementary
Deletion of chromosomal dnaK ( dnaK ) in cells expressing Cas1-Cas2 was ther efor e pr edicted to also de-r epr ess nai v e adaptation. E. coli dnaK was deleted from the chromosome by recombineering ( 29 ) and the resulting cells were phenotyped at the commencement of each experiment to avoid suppressor mutations that compensate for dnaK --dnaK cells should be unable to support phage infection, and show temperature sensitivity (Supplementary Figure S2a We next tested for physiological conditions that release nai v e adaptation from inhibition. DnaK undergoes dynamic binding and release cycles when chaperoning 'client' proteins, including when DnaK is 'hijacked' by MGE proteins for DNA replication and quality control for assembly of phage particles ( 27 , 28 , 37 ). Specific interactions of pr oteins fr om phage with DnaK in E. coli are essential for the lytic cy cle --e x emplified in figur e S2a --ther efor e we co-expr essed pr oteins fr om the same plasmid as DnaK, and alongside Cas1-Cas2, and assessed nai v e adaptation. Expression of V protein restored fully functioning nai v e adaption compared with control cells expressing DnaK alone alongside Cas1-Cas2 (Figure 2 F,  G and Supplementary Table S7). V protein forms phage coat / tail structures, and has not previously been shown to interact with DnaK. We also tested P protein, which is r equir ed to interact with DnaK for initiating phage replication (38)(39)(40), and observed modest de-repression of nai v e adapta tion tha t was not as ef fecti v e as V protein (Supplementary Figure S2e and f, and Supplementary  Table S7).

Cas1 foci require DNA binding and are controlled by DnaK
To observe the behaviour of Cas1-Cas2 in living cells, and whether it is influenced by DnaK, we fused Cas1 to the fluorophore eYFP. By fusing Cas1 at its C-terminus to eYFP via a (GGS) 8 linker (Cas1-LFP, Linker Fluorescent Protein), and with Cas2 inserted downstream of eYFP for co-expression from the same plasmid (Cas1-LFP + Cas2, Figure 3 A) --therefore similarly to the Cas1-BioID2 fusion --we were able to detect naive adaptation in cells, confirming that Cas1-LFP + Cas2 was functioning (Figure 3 B). Cas1-LFP produced bright foci corresponding with the space occupied by the nucleoid (Figure 3  Our imaging data is consistent with Cas1-LFP focus formation that is str ongly pr omoted by Cas1 binding to DNA. Cas1-LFP foci in dnaK cells were dramatically changed in frequency and distribution compared with wild type cells (Figure 4 ), with no foci observed in dnaK cells expressing only eYFP (Supplementary Figure S3). Cells with multiple foci increased to 90%, with 8% of dnaK cells generating greater than 10 foci (Figure 4 C). Increased frequencies of well-defined Cas1-LFP foci in dnaK cells was accompanied by additional 'classes' of foci, highlighted encircled (I) in Figure 4 A and B. The additional foci formed a larger area and showed une v en distribution of brightness, which we describe as 'haze-like' (Figure 4 B, highlighted). When haze-like foci were included in focus counting > 60% of cells had multiple foci (Figure 4 C). When we assessed Cas1 R84G -LFP foci in dnaK cells we observed substantially reduced in frequency; 85% of cells showed no foci at all, compared with < 10% of dnaK cells expressing 'wild-type' Cas1-LFP (Figure 4 C). When present, the Cas1 R84G -LFP foci pr esent wer e consistently always haze-like in morphology, localised throughout the cell (Figure 4 A and B). Thus, normal-looking foci were specifically reduced if Cas1 R84G -LFP was expressed, while only minor changes were observed for the haze-like foci (Figure 4 C). The behaviour of Cas1-LFP foci compared with foci from Cas1 R84G -LFP, and the stimulatory effects of dnaK on nai v e adaptation and focus formation, are consistent with DnaK restraining Cas1-Cas2 from DNA capture.

Cas1 foci form during active DNA replication
Nai v e adaptation in E. coli has been suggested to target acti v e DNA replication forks ( 21 ). To investigate this using Cas1-LFP foci we generated cells conditionally unable to support DNA replication, but which can be switched over to acti v e DNA r eplication, pr edicting that this may alter the presence and absence of foci. E. coli cells were synchronised using a temperature-sensiti v e allele of the replication initiation protein DnaA ( dnaA46 ) that can initiate replication at oriC at 30 corresponding to the replisome, which were absent after a 90 min incubation period at 42 • C (Figure 5 A), confirming this as effecti v e for pre v enting ne w DNA replication initiating. We then visualised Cas1-LFP by splitting an exponentially growing culture. One half was incubated at 30 • C, while the other half was shifted to 42 • C. After each culture had been incubated for 30 min, arabinose was added to induce Cas1-LFP + Cas2 expr ession. Both cultur es wer e then incubated for 60 further min at their respecti v e temperatur es befor e cell imaging. Cells that remained at 30 • C showed robust numbers of Cas1-LFP foci, but no foci were observed in cells shifted to 42 • C (Figure 5 B). This provides direct support for Cas1 targeting DNA during acti v e DNA replication.

DnaK inhibits DNA binding and integration by Cas1 in vitro
We investigated the mechanism for DnaK inhibiting nai v e adaptation by physical interaction targeting Cas1 (Figure 1 E). Nai v e adaptation in E. coli r equir es that the Cas1-Cas2 complex binds to a DNA fragment and integrates it as a new CRISPR spacer ( 8 ). On formation of Cas1-Cas2 complex from individually purified Cas1 and Cas2 proteins (Supplementary Figure S4a) we observed integration of a model 'protospacer' Cy-5 labelled DNA fragment (Suppl . Table 11)  We then assessed whether DnaK influenced Cas1 DNA binding to flayed duplex DNA in complexes sufficiently stab le to observ e in EMSAs ( 22 , 42 ). Providing Cas1 with a mixture of Cy-5 end labelled 'decoy' ssDNA and flayed duplex substrate of the same sequence, formed a stable complex with only the flayed duplex DNA (Figure 6 c i lanes 2-4) with Cas1 e v entually aggregating DNA into the gel wells (i, lane 5). DnaK did not bind to DNA in EMSAs, as expected, (Figure 6 Cii, lanes 6-10), but DnaK titration (0.25-1 M) into pre-bound Cas1 (0.5 M)-DNA complex es, r eleased flayed duplex DNA from binding by Cas1, including DNA from gel wells (Figure 6 Ciii, lanes 11-16). When we replaced DnaK with purified DnaK S427P (Supplementary Figure S4b), DnaK with the SBD mutation that was ineffecti v e at inhibiting na ïve adaptation, we observed Cas1 prebound to DNA was not removed, consistent with this DnaK mutant being unable to associate with Cas1 (Supplementary Figure S4c and d). These in vitro data are consistent with DnaK being able to restrain naive adaptation by pre v enting Cas1-DNA binding, and provides a plausib le explanation for the altered behaviour of Cas1-LFP foci in cells.

DISCUSSION
Our work identifies that chaperoning of Cas1-Cas2 by DnaK in E. coli cells regulates nai v e adaptation. The influence of DnaK on m ultiple biolo gical pr ocesses acr oss prokaryotes is reflected in its widespread conservation, and we now re v eal that DnaK physically and functionally interacts with and controls a CRISPR system. We detected it using BioID-based proteomics, which was validated by physical interaction between DnaK and Cas1 that was sufficiently robust to extract Cas1 from Cas1-Cas2 complex when they are expressed together in cells (Figure 1 ). This re v ealed the inhibitory effect of DnaK on nai v e adaptation only when we inducibly expressed DnaK from a plasmid, to 'compete' with Cas1-Cas2 that was also inducib ly e xpressed from a plasmid --in typical nai v e adaptation assa ys an y effect of DnaK is not noticed, likely because le v els of plasmid-induced Cas1-Cas2 ov erwhelm the le v els of chromosomally encoded DnaK, resulting in self DNA targeting by Cas1-Cas2 that is typical of nai v e adaptation assays. The robust Cas1-DnaK interaction was also apparent from DnaK removing Cas1 from DNA to Cy5 labeled single stranded DNA, f ork, and f or k-Cas1 comple x are indicated to the sides of panels. All gels were 5% acrylamide, imaged using a Typhoon ™ laser scanner platform (Cytivia).
which it was pre-bound in vitro. Prokaryotic CRISPR immunity systems all deploy Cas1-Cas2 to generate immunity, and DnaK is highly conserved across all prokaryotes, except in hyperthermophilic archaea, and across di v erse CRISPR types --DnaK residues Ser-427 and Asn-451 that we observed are required for inhibition of nai v e adaptation are conserved in DnaK in prokaryotes with all CRISPR systems types, I-VI. We suggest that DnaK may control CRISPR adaptation beyond the Type IE E. coli CRISPR system. Inhibition of nai v e adaptation by DnaK raises the question of how Cas1-Cas2 may be released for DN A ca pture, and if this may provide targeting of Cas1-Cas2 to MGEs. We observed that inhibition of naive adaptation by DnaK was re v ersed in the presence of phage proteins. Recruitment of DnaK to plasmid and phage MGE proteins is r equir ed to initiate their DNA replication and to control quality and chaperoning of phage head and tail packaging proteins ( 27 , 43-49 ) --the essential role of DnaK is evident from the inability of phage to infect E. coli dnaK cells (Supplementary Figure S2a). These processes are controlled by dynamic binding and release of DnaK client proteins from the substrate binding domain, governed by ATPhydrolysis triggered allosterically by other proteins, including the DnaJ co-cha perone, recentl y re vie wed ( 27 ). Physical interaction of DnaK and Cas1 may be a binding and release process to protect the chromosome from targeting by Cas1-Cas2, until an MGE client protein of DnaK triggers release of Cas1-Cas2. This may spatially and temporally control Cas1-Cas2 to target MGE DNA that is generated in abundance during invader MGE genome replication. In this model (Figure 7 ), Cas1-Cas2 would not be distinguishing MGE DNA fr om host chr omosomal DNA, but is instead restrained from self-targeting and released to target MGEs through interaction with DnaK. Our observation that Cas1 foci can be triggered by acti v e DNA replication, but are not detected when we suppress initiation of replication ( Figure 5 ), raises an intriguing possible mechanism for Cas1-Cas2 targeting to MGE DN A, w hen MGE proteins suppress host chromosomal replication in preference for triggering that of the MGE genome. Our data from several lines of investigation provide new insights about how DNA is targeted so that CRISPR immunity can be established. It newly reveals interplay between natural CRISPR systems and a host factor (DnaK) that is a central controller 'hub' for other cellular processes, including of DNA replica tion. Further investiga tion will be able to identify whether the DnaK-Cas1 interactions that we have identified are common across CRISPR systems, and similarly control adaptation.

DA T A A V AILABILITY
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD042090.

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.