Histones direct site-specific CRISPR spacer acquisition in model archaeon

Abstract

CRISPR–Cas systems provide heritable immunity against viruses and other mobile genetic elements by incorporating fragments of invader DNA into the host CRISPR array as spacers. Integration of new spacers is localized to the 5′ end of the array, and in certain Gram-negative Bacteria this polarized localization is accomplished by the integration host factor. For most other Bacteria and Archaea, the mechanism for 5′ end localization is unknown. Here we show that archaeal histones play a key role in directing integration of CRISPR spacers. In Pyrococcus furiosus, deletion of either histone A or B impairs integration. In vitro, purified histones are sufficient to direct integration to the 5′ end of the CRISPR array. Archaeal histone tetramers and bacterial integration host factor induce similar U-turn bends in bound DNA. These findings indicate a co-evolution of CRISPR arrays with chromosomal DNA binding proteins and a widespread role for binding and bending of DNA to facilitate accurate spacer integration.

Prokaryotic CRISPR–Cas (Clustered Regularly Interspaced Short Palindromic Repeats and CRISPR-associated protein) systems provide heritable, adaptive immunity against viruses, plasmids and other parasitic mobile genetic elements (MGEs) by maintaining a record of previously encountered MGEs in a repository called the CRISPR array¹. The array is a series of short (typically 30–40 base pairs (bp)) direct repeats separated by similarly sized fragments derived from invading MGE genomes, termed spacers. The organization of this array gave rise to the descriptive name for these systems. Adjacent to the CRISPR array is a regulatory DNA region called the leader which encodes promoter elements used for CRISPR RNA expression as well as sequence elements important for directing spacer integration^2–5. The CRISPR array is transcribed and processed into mature CRISPR RNAs⁶ which guide CRISPR ribonucleoprotein effector complexes to recognize and degrade nucleic acid of invading MGEs in a sequence-specific manner^7–10.

Uniquely among prokaryotic MGE defence systems, CRISPR–Cas immunity is adaptive: the array can be continuously updated through addition of new spacers against newly encountered MGEs¹¹. DNA fragments are sampled and integrated into CRISPR arrays by integrase complexes composed of highly conserved Cas1 and Cas2 proteins^7,12,13, together with accessory proteins and factors that vary among the different CRISPR–Cas subtypes and host organisms^14–20. The mechanism of spacer integration involves two sequential transesterification reactions whereby the 3′ OH groups of the spacer DNA make specific nucleophilic attacks at the 5′ borders of the repeat, followed by resolution of the integration site and DNA repair^21,22 (Extended Data Fig. 1). The end result is the expansion of the array by one new spacer-repeat unit. A common feature of diverse CRISPR systems in both bacterial and archaeal organisms is that new spacers are integrated into the CRISPR array in a polarized manner with a strong preference for the leader-adjacent repeat rather than downstream repeats within the array^11,23.

Work with different CRISPR systems has revealed that polarized spacer integration proceeds by one of two different modes in vivo. In the first mode, certain Cas1–Cas2 integration complexes show intrinsic specificity for the leader-proximal repeat: the complex directly recognizes sequence elements in the leader-proximal repeat as well as leader sequences located immediately adjacent to that repeat^12,24,25. This intrinsic directionality is exemplified by the type II-A CRISPR–Cas systems of Streptococcus thermophilus, Streptococcus pyogenes and Enterococcus faecalis where purified Cas1–Cas2 complexes specifically integrate new spacers at the first repeat in vitro without additional factors^12,24,25.

The second, and perhaps dominant mode of polarized integration, is factor mediated rather than sequence intrinsic. This is typified by types I-E and I-F CRISPR–Cas systems of Escherichia coli and Pseudomonas aeruginosa, where the polarity of integration is guided by a host-encoded chromosomal DNA binding protein, integration host factor (IHF)^26–30. In vivo, IHF is required for any detectable new spacer acquisition^26,27. In vitro, purified Cas1–Cas2 integration complexes are capable of spacer integration at CRISPR repeat sequences, but integration occurs at all repeats in an array with no preference for the leader end. When purified IHF is added to the in vitro integration reaction, spacer acquisition occurs almost exclusively at the leader-proximal repeat, showing its importance in bona fide CRISPR spacer acquisition^21,27–30. IHF binds at one or more conserved sites within the leader and induces a sharp horseshoe-like bend in the leader DNA that brings an upstream leader motif into close proximity to the repeat-bound Cas1–Cas2 complex, strongly promoting spacer integration at the leader-proximal repeat^{26–28,30,31}.

Genes encoding IHF are restricted to mostly Proteobacteria²⁸, so it remains unclear how factor-mediated polarized integration would occur in other Bacteria and in Archaea. In the crenarchaeon Saccharolobus solfataricus, in vitro reconstituted Cas1–Cas2 complexes catalyse spacer integration into purified plasmid DNA but without leader or repeat specificity. When cell lysate is added to the reaction, spacer integration is directed to the leader-adjacent repeat, but the lysate component responsible for this effect remains unknown³².

CRISPR spacer uptake in the model hyperthermophilic archaeon Pyrococcus furiosus is relatively well characterized, and as with most other species, spacer addition occurs preferentially at the leader-adjacent repeat^14,33,34. The P. furiosus genome harbours seven separate CRISPR arrays that each contain the same repeat sequences and highly conserved leaders³⁵. Spacer integration at all seven arrays is catalysed by a single Cas1–Cas2 integrase complex^14,36,37. Previous work suggests that P. furiosus utilizes a factor-mediated, polarized spacer integration mechanism as reconstituted Cas1–Cas2 integrase from this organism adds spacers to CRISPR repeats without specificity for the first repeat³⁷ in vitro while new spacers are only observed being added at the first repeat in vivo^14,33,34. These data suggest an unidentified factor is required to promote leader-proximal spacer integration in P. furiosus. Here we present evidence that archaeal histones are responsible for guiding directional spacer integration into CRISPR arrays in P. furiosus.

Results

MN-seq reveals histone occupancy at CRISPR leaders

With the aim of identifying the hypothesized factor(s) responsible for directing leader-proximal integration in P. furiosus CRISPR arrays³⁷, we conducted a genome-wide analysis of DNA binding proteins and then looked for evidence of DNA binding at CRISPR loci. For this, we carried out micrococcal nuclease (MNase) digestion of chromatin from P. furiosus COM1 strain followed by high-throughput sequencing (MN-seq) of protected (protein-bound) DNA fragments (Fig. 1). Upon digestion, we obtained a ladder-like pattern of protected DNA fragments, with particularly strong bands for 60, 90 and 120 bp sized products (Fig. 1a). To visualize patterns of MNase-protected DNA across the genome, the protected DNA fragments were excised and sequenced, and resulting reads were aligned to the reference P. furiosus genome (Fig. 1b–d). We generated colour-coded sequencing tracks to distinguish the four main fragment sizes of 30, 60, 90 and 120 bp. In general, most of the genome had some level of coverage, with more pronounced peaks interspersed throughout (Fig. 1b and Extended Data Fig. 2). There were few reads in the 30 bp size range, consistent with the lower intensity of that band in gel images (Fig. 1a).

Fig. 1 |. MNase protection assay to characterize P. furiosus DNA binding proteins.

a, Gel image of DNA fragments generated through MNase digestion of P. furiosus chromatin. Coloured boxes highlight major bands of protected DNA. Digestions were repeated three times; similar gel band patterns were observed each time. b–d, DNA bands highlighted in a were sequenced by high throughput sequencing and mapped to P. furiosus genome; genome browser tracks show the distribution of aligned reads across representative CRISPR loci. Overlapping bars are colour-coded according to the size of the DNA fragment (red, 30 ± 5 bp; golden yellow, 60 ± 5 bp; green, 90 ± 5 bp; blue, 120 ± 5 bp). The y axis indicates the cumulative depth of DNA sequencing (DNAseq) read coverage. Six replicates were sequenced and yielded similar patterns; a representative replicate is shown here. b, A wide, approximately 10 kb, view of the genome space surrounding CRISPR7 (leader in pale yellow, repeat array in black; purple star indicates position of spacer integration). c, A closer view of full CRISPR7 array. d, Insets that show the leader and first three repeats for CRISPR7, CRISPR2, CRISPR4 and CRISPR6.

Interestingly, we observed relatively high protection peaks in the leaders of the CRISPR arrays (Fig. 1b–d). In each of the seven CRISPR arrays, three 60 bp leader peaks were evident (Fig. 1b–d). The three peaks were not equally pronounced in each array but were consistently observed in different growth conditions and replicates (Extended Data Fig. 3). We named these peaks LR, L1 and L2. The LR peak partially overlapped the leader–repeat junction and site of spacer integration. The LR peak was bounded on the left side by the promoter for CRISPR array expression and on the right side by the first CRISPR repeat (Extended Data Fig. 4). The L1 peak was located ~100 bp upstream from the leader–repeat junction and was generally the site with the greatest level of protection as detected by MN-seq. The L2 peak, located further upstream, was the least pronounced at most CRISPR loci (Fig. 1 and Extended Data Figs. 3 and 4).

We noted that the leader peaks almost exclusively comprised 60 bp reads (Fig. 1b–d, golden yellow). In contrast, upstream and downstream from the arrays, and generally elsewhere in the genome, yellow peaks, corresponding to 60 bp aligned reads, were usually much lower (Fig. 1b and Extended Data Fig. 2). We evaluated genome-wide 60 bp read distributions by examining read depth in 60 bp sliding windows across the genome and found that the tallest leader peaks were in the 99.9th percentile, indicating they are outliers (Extended Data Fig. 3). We also noted that most of the other relatively tall peaks (outside of the leaders) comprised 90 bp reads (green), and many prominent 60 bp read peaks also had some 90 bp read overlap (Fig. 1b and Extended Data Fig. 2). In contrast, for the LR and L1 peaks, 60 bp read depth was about 15 times higher than 90 bp read depth, on average, a bias that was relatively rare elsewhere in the genome (Extended Data Fig. 3). Other than the peaks of 60 bp MN-seq coverage in the leaders, CRISPR arrays did not show unusually dense coverage (Fig. 1b and Extended Data Fig. 2). These interesting features prompted us to consider that MNase footprints in CRISPR leaders may reflect unique functions for DNA-binding proteins in this region.

Like most archaea^38–40, P. furiosus encodes histone proteins that bind and wrap DNA, thus protecting the DNA from MNase digestion^41,42 (Fig. 2a). P. furiosus encodes two similar but distinct histone isoforms (A and B) predicted to form both homodimers and heterodimers in solution^41,43,44. These histones are expected to bind and wrap differing lengths of DNA, depending on the number of histone subunits in a complex: 30 bp for a dimer, 60 bp for a tetramer (or dimer of dimers), 90 bp for a hexamer (trimer of dimers) and so on (Fig. 2a), which would result in the observed 30 bp laddering pattern of protected DNA when chromatin is treated with MNase (Fig. 1a). Histone tetramers (which would protect 60 bp of DNA like that observed at the CRISPR leaders (Fig. 2c)) are predicted to induce sharp, horseshoe-like or U-turn bends in the DNA^38,41 (Fig. 2a).

Fig. 2 |. DNA binding by histones.

a, Crystal structure of DNA-bound histones from Methanothermus fervidus (PDB 5T5K)⁴¹. b, Globally, MNase-protected DNA fragments in P. furiosus had a non-random distribution of AT and GC dinucleotides along their lengths, with the 10 bp periodicity observed for histones in Eukaryotes and other Aarchaea (the y-axis is the relative frequency with 1.0 being 100%) of the specific dinucleotide. c, Global AT and GC dinucleotide periodicity trends are also observed over the 60 bp protected peaks in CRISPR leaders. Three replicates of each sample were sequenced; similar patterns were observed for all.

Histone binding patterns have been studied in a wide range of organisms, leading to the identification of sequence features that promote or repel stable histone binding along stretches of DNA^45,46. We examined the global characteristics of 60 bp MNase-protected reads in our data set and observed the hallmark phased helical repeats of AT and GC dinucleotides expected for histone binding^45,46. Focusing on CRISPR leaders, the sequences underlying LR, L1 and L2 were consistent with this pattern of alternating AT and GC dinucleotides (Fig. 2c). Together, these presumed 60 bp histone tetramer binding footprints in CRISPR leaders appeared to be a unique and interesting feature, suggesting the possibility of a functional role of histone binding in polarized spacer integration.

Histones promote adaptation in vivo

Based on the prediction that histone complexes were specifically bound to CRISPR loci, we tested whether histones were important for spacer integration in vivo (Fig. 3). Specifically, we transformed P. furiosus strains with a recently identified conjugative plasmid, called pT33.3, that was found to greatly stimulate spacer uptake into CRISPR arrays⁴⁷. We used our established PCR-based approach (Fig. 3a) to test the ability of P. furiosus strains to acquire new spacers into four separate CRISPR arrays in a strain containing both histones A and B (the P. furiosus COM1 parental strain, referred to here as wild type) as well as in strains lacking either histone A or histone B (Fig. 3b). We were unable to generate a strain lacking both histones A and B, presumably because at least one histone is required for cell viability⁴⁸. Both the ∆histone A and ∆histone B strains were viable, showed typical growth rates and showed largely unchanged gene expression profiles by RNA sequencing (RNA-seq) compared to wild type (Extended Data Fig. 5a–c). As a negative control, we also examined spacer uptake in a strain wherein Cas1 and Cas2 are deleted and adaptation does not occur. Relative to wild type, deletion of genes for either histone A or B led to a decrease in both the percentage of cells taking up spacers and the number of integrated spacers per array for each of the four arrays examined (Fig. 3b,c). Ectopic spacer integrations at downstream repeats were not observed irrespective of histone availability (Fig. 3d). It appeared that loss of histone A had a greater impact on spacer uptake than did loss of histone B (Fig. 3b,c). To determine whether this effect could be due to differing levels of total histone abundance in these strains, we used western blotting to estimate the concentration of histone proteins from crude cell lysates. Blots showed that total histone expression was not reduced for the ∆histone A strain (Extended Data Fig. 5d), implying that total histone abundance is not responsible for the greater impact histone A deletion had on adaptation. Western blots also showed that the three strains had similar chromatin organization, as larger multimeric complexes were visible for all (Extended Data Fig. 5d). Together, these results show that histones A and B promote spacer uptake in vivo.

Fig. 3 |. Histone gene deletions alter spacer integration at the leader-adjacent repeat in vivo.

a, Schematic diagram of a PCR-based, in vivo adaptation assay to detect integration of new spacers at the leader-adjacent repeat. The relative position of PCR primers is labelled by arrows. b, Gel image of PCR products from the assay shown in a. Strains used here contained only the type I-B effector complex. Where indicated, strains were transformed with the conjugative plasmid pT33.3 to enhance adaptation. *, unexpanded array;], expanded arrays. The assay was performed with primers for all seven CRISPR arrays to evaluate the consistency of the phenotype (CR7, CR4, CR5 and CR6 are shown here). c, Intensity of all PCR bands, as in b, was quantified using ImageJ. The proportion of expanded to unexpanded products was determined, along with the average number of new spacers in an array, as calculated based on relative band intensity. ∆, spacer integration genes deleted (∆cas1, ∆cas2, ∆cas4-1, ∆cas4-2). d, Primers that amplify downstream from the leader-adjacent repeat (spacer 1–3 or 3–6, as labelled) were used to detect ectopic (that is, internal to the array) spacer integrations. Assays were conducted either five times (CR5 and CR7) or twice (CR4 and CR6) with consistent results; representative gels and quantifications are shown here.

Histones guide proper spacer integration in vitro

We turned to in vitro studies to further characterize the role of histone proteins in specifying spacer integration (Fig. 4a). We had previously reconstituted CRISPR spacer integration in vitro by incubating purified P. furiosus Cas1–Cas2 integrase complexes with a synthetic spacer and a DNA substrate, a plasmid (called pCR7-short) encoding a CRISPR array containing a leader and three repeat-spacer units³⁷ (Fig. 4a). Sites of spacer integration into pCR7-short are detected via PCR using a primer complementary to the newly acquired spacer and a primer downstream of the CRISPR array (Fig. 4a). Previously, we reported that Cas1–Cas2 integrate spacers at each of the three repeats without major preference (ref. 37 and Fig. 4a). Here we added purified histones from either P. furiosus or from Thermococcus kodakarensis, a related species of Thermococcales with two histone genes that are well studied and highly similar to those of P. furiosus (Extended Data Fig. 6). We compared the spacer integration results for histone-free plasmid DNA versus the same plasmids following incubation with histones A, B or both combined (that is using chromatin CRISPR DNA substrates; Fig. 4a). As expected from previous results³⁷, in the absence of histones, Cas1–Cas2 complexes catalysed integration at each of the three repeats with no apparent preference (Fig. 4b). In contrast, addition of purified histones (A alone, B alone or a mixture of A and B) led to a concentration-dependent shift in spacer integration to the leader-proximal repeat (that is, repeat 1) (Fig. 4b). We observed the same effect when either P. furiosus (Fig. 4b,c) or the T. kodakarensis histones proteins were used (Fig. 4d–g). PCR products sometimes appeared as doublets, which we believe is due to variable electrophoretic mobility of the same PCR product in the ethidium bromide-stained agarose gels (Fig. 4d). Quantification of relative PCR band intensities showed that histones A and B from both archaeal species increased preference for the leader-proximal repeat 1 from ~30% (histone-free conditions) to between 60% and 80% (at the highest histone concentration tested) (Fig. 4c,e,g). These results provide evidence that both histone A and B proteins per se promote integration of spacers by the Cas1–Cas2 complex at the leader-adjacent repeat, mirroring the integration site used in vivo.

Fig. 4 |. In vitro evaluation of spacer integration in the presence of archaeal histones.

a, Diagram of experimental set-up; spacers are integrated into pCR7-short by the Cas1–Cas2 adaptation complex in the presence or absence of histones A and/or B (from either P. furiosus or T. kodakarensis). PCR was used to assess the positions and relative efficiencies of in vitro integration events into pCR7-short; primers spanned from the inserted spacer to a downstream plasmid position (PCR 1) or from the leader to the inserted spacer (PCR 2). b,c, Gel images show products from PCR 1 for assay done with P. furiosus (b) and T. kodakarensis (c) histones. Pfu, P. furiosus; Tko, T. kodakarensis. The expected sizes for PCR products resulting from integrations at repeat 1, repeat 2 and repeat 3 are marked with R1, R2 and R3, respectively (the black asterisk indicates a non-specific spacer integration event). Red asterisks mark the band for integration at repeat 1, which is the natural, preferred point of integration in vivo. Histograms below each gel image show relative band intensities, quantified using ImageJ. d, Gel image shows products from PCR 2 for integration assay done with T. kodakarensis histones. Histograms below the gel image show band intensity quantifications, as above. Assays and gel imaging were done in duplicate (Pfu) or triplicate (Tko) with similar results; representative gels and quantifications are shown here.

Several additional experiments further confirmed a specific role for histones in guiding or confining spacer integration to the leader-adjacent repeat in vitro. First, when the plasmid-encoded CRISPR array was expanded from 3 spacer-repeat units (pCR7-short) to 11 spacer-repeat units (pCR7-long; Fig. 5a), we observed the same phenomenon: integrations occurred at all repeats in the absence of histones but were mostly at the first repeat in the presence of histones (Fig. 5b, lanes 2 versus 3 and lanes 4 versus 5; Fig. 5c). The effect was observed with both supercoiled plasmid DNA or linearized plasmid DNA (Fig. 5b, lanes 6 versus 7 and lanes 8 versus 9; Fig. 5d) implying that histones do not simply act by inducing changes to DNA supercoiling. We also found that the effect was not unique to the CR7 array. Histones increased the preference for repeat 1 in all of the three additional arrays that we tested in vitro (plasmids bearing CR5, CR6 and CR8; Extended Data Fig. 7). To confirm that purified P. furiosus histones did indeed bind the plasmid-encoded leader, we conducted an MNase protection assay on pCR7-long mixed with purified histones A and B, followed by sequencing. L1 and LR protection peaks were observed in the same relative positions as seen in vivo (Extended Data Fig. 8).

Fig. 5 |. In vitro spacer integration into different DNA substrates in the presence of P. furiosus histones.

a, Diagrammatic representation of the PCR used to assess in vitro integration into pCR7-long. Primer 1 binds the spacer (S₀), while primer 2 binds either spacer 5 (S₅) or spacer 10 (S₁₀); in this diagram spacer 10 products are outlined. b, Gel image shows products from the PCR outlined in a; pCR7-long was either supercoiled or linearized by inverse PCR. The S₀–S₅ label indicates PCR products when the spacer 5 primer was used; S₀–S₁₀ label indicates PCR products when spacer 10 primer was used. Red asterisks highlight the repeat 1 band. Assay and gel imaging was done in duplicate; a representative gel with quantification is shown here. c,d, Intensity of all PCR bands in b was quantified using ImageJ, and the proportion of integration events at all repeats was determined for both supercoiled (c) and linear (d) plasmid DNA.

To further address the specificity of the histone-mediated effects on shaping the site of spacer integration, in vitro spacer integration assays were carried out with two other abundant DNA binding proteins found in P. furiosus, TrmBL2⁴⁹ and Alba^38,50. Titration of either protein failed to reproduce the changes in integration observed for histones A and B, although higher concentrations of TrmBL2 appeared to suppress integration at all repeats (Extended Data Fig. 9; purified proteins shown in Extended Data Fig. 6). Finally, we verified the PCR-based results using an established sequencing-based approach that does not rely on array-specific primers³⁷. Without histones, numerous integration events were detected at each CRISPR 5′ repeat boundary (Extended Data Fig. 10). However, with histones there was a clear shift in the number of integrations at the leader-adjacent repeat, and that preference scaled with histone concentration (Extended Data Fig. 10). In contrast, addition of the Alba protein did not alter the sites of integration across repeat junctions at any concentration (Extended Data Fig. 10). Taken together, the in vitro studies provide compelling evidence that histones play a specific role in directing spacer integrations to the leader-adjacent repeat.

Discussion

Prokaryotic CRISPR arrays serve as memory banks of past infection that are used to immunize an organism against potentially deleterious viruses and other MGEs. This immunity is passed on to future generations, making CRISPR the only example of adaptive immunity found in prokaryotes. The polarized acquisition of new spacers to the leader-adjacent CRISPR repeat is ubiquitous throughout diverse CRISPR–Cas systems of Bacteria and Archaea⁵¹. To date, only one bacterial host factor has been shown to drive leader-proximal spacer integration—the proteobacterial IHF protein^26–30. Beyond this example, the identity of host factors underlying spacer integration polarity remains unknown for the majority of CRISPR-containing species. Here, in vivo and in vitro findings reveal that histones play an important role in CRISPR immunity by serving as a key host factor that directs site-specific CRISPR spacer acquisition in the model archaeon, P. furiosus.

While the molecular details of how the tested P. furiosus and T. kodakarensis histones act to promote spacer integration at the leader-adjacent repeat are not yet understood, we propose a general model that mirrors that of IHF (Fig. 6). Both IHF and histones are abundant host chromosomal binding proteins that wrap DNA and, in doing so, induce sharp bends and topological changes. In the case of IHF-mediated integration, one or two IHF dimers bind conserved IHF binding motifs present in the CRISPR leader²⁸ and bend the DNA such that upstream leader motifs contact the Cas1–Cas2 spacer integration complex (Fig. 6). In the E. coli system, this IHF binding has been proposed to expose the leader–repeat phosphate for nucleophilic attack by the 3′-OH of the incoming spacer present within the Cas1–Cas2 integrase complex^25,27. In our model, histone tetramers bind to one, two or even three 60 bp conserved leader-encoded motifs and are predicted to bend the DNA in a manner similar to IHF (Fig. 6).

Fig. 6 |. Hypothetical model for involvement of DNA binding proteins in spacer integration in bacterial and archaeal systems.

a,c, Cartoon diagrams, based on recent structural and functional data, showing a general model for how histones (a) and IHF (c) bind CRISPR leaders and guide integration to the leader-adjacent repeat. b, Crystal structure of histone B from M. fervidus PDB 5T5K ref. 65. d, Crystal structure of IHF from E. coli PDB 1OWG ref. 68. e, Components of the integration reaction: DNA structure with positions of LR and L1 histone binding sites (yellow) relative to promoter elements (BRE and TATA in purple and red, respectively), as well as CRISPR repeat (blue) and spacer integration site at leader–repeat junction; histone tetramer (dimers labelled in orange and green) bound to 60 bp of DNA; Cas1–Cas2 integration complex with bound pre-spacer (purple). f, Binding of histone tetramers to L1 (left), or LR (right) induces a strong bend in CRISPR leader sequence. g, Binding of histone tetramers to both L1 and LR with Cas1–Cas2 bound at the leader–repeat spacer integration site.

There are several possibilities for how DNA binding and bending by histones could promote integration at the leader-proximal repeat in P. furiosus. Leader bending may allow a secondary upstream motif to contact Cas1–Cas2 complex, as in the E. coli IHF system^26,28,30. Alternatively, the Cas1–Cas2 complex may directly contact the histone tetramer, and some aspect of this physical interaction could aid integration. A third possibility is that the bending or torsion induced by the histones may promote integration by making the leader–repeat phosphate more exposed for nucleophilic attack by the spacer DNA. We expect that the site of integration at the leader–repeat junction would be readily accessible on the outer surface of the LR positioned histone tetramer (Fig. 6). Additional experiments and structural data will be necessary to understand if one or more of these possible mechanisms underlie directional spacer integration in P. furiosus.

Another interesting parallel between IHF and histones in our study is that both are required for detectable integration in their respective systems in vivo, but for these same systems in vitro, Cas1 and Cas2 efficiently carry out integration at array repeats without leader–end specificity. This discrepancy between in vivo and in vitro effects suggests that there may be other factors or cellular features in vivo that suppress integration at downstream repeats, even in the absence of leader-binding proteins like IHF and histones. Overall, we find it remarkable that CRISPR spacer acquisition appears to rely on similar modes of DNA binding and bending across the great evolutionary distance between Bacteria and Archaea (Fig. 6).

Our results could indicate a conserved role for histones in archaeal CRISPR immunity as histones are widespread in archaeal lineages^38–40. Notable exceptions are the Crenarchaeota and the Diaforarchaea clades⁵², which lack histones. Intriguingly, histones are not present in the crenarchaeon S. solfataricus, where factor-mediated spacer integration is also known to exist³², so polarized spacer integration in crenarchaeon CRISPR systems may be ensured by yet another co-opted host chromosomal DNA binding protein.

Previous work aimed at understanding the determinants for histone binding to DNA showed that underlying sequences are the primary predictor for archaeal histone occupancy^45,46. Important DNA features that recruit archaeal histones include helically phased AT and GC dinucleotides and certain GC-rich stretches (and the absence of certain AT-rich DNA stretches^45,46). The regions of P. furiosus CRISPR leaders with peaks of MNase-Seq read coverage contain these DNA features and are bordered by histone-repelling AT-rich sequences that comprise elements involved in transcription initiation (the BRE/TATA elements) and the CRISPR repeat (Figs. 1 and 2 and Extended Data Fig. 4). Histone binding at these same sites was also observed in vitro at the leader of a plasmid-encoded CRISPR locus (Extended Data Fig. 8), confirming that histone binding is largely mediated by sequence features, rather than by other protein(s) or DNA topology. Such a sequence requirement suggests that, as with IHF sites in Proteobacteria, Archaeal CRISPR leaders co-evolved with host-associated chromosomal DNA binding proteins.

Like many archaea, P. furiosus and T. kodakarensis encode two highly similar histone isoforms, referred to as A and B (Extended Data Fig. 6). Consistent with previous findings with T. kodakarensis⁴⁸, we found that either P. furiosus A or B histone could be deleted but a double histone knockout (A + B) was not possible, indicating that at least one histone is necessary for viability. It is not clear why two (or more) related histones commonly occur in archaea⁴⁴. Our in vivo data indicated that loss of histone A disrupted spacer uptake more than loss of histone B, potentially indicating a specialized role for histone A in integration (Fig. 3). In contrast, our in vitro work showed that both histone A and B (from either P. furiosus or T. kodakarensis) promoted spacer integration to the first repeat with approximately equal efficiencies (Fig. 4). Together, the results show that both histone A and B, which likely bind DNA with similar properties⁴⁵, are competent in conferring integration specificity to the Cas1–Cas2 integrase complex but leave open the possibility that additional layers of regulation make histone A more effective at directing new spacer uptake in vivo.

It is now clear that factor-mediated mechanisms (histones in Archaea and IHF in Bacteria) ensure that new spacer integration is localized to the leader-adjacent end of the repeat array. However, it is still unclear what evolutionary driving forces promote this localization. Although studies have shown that spacers throughout an array are competent for interference, spacers in the leader-adjacent position are often more highly expressed and give rise to more potent protection than leader-distal spacers^5,51,53. In addition, the numerous repeats in a CRISPR array may make central spacers less stable (subject to loss through recombination). Finally, the combinatorial requirement of leader–repeat sequence elements together with leader-bound protein factors may ensure specificity toward CRISPR arrays, thereby minimizing ectopic integration of foreign DNA fragments at repeat-like sequences elsewhere in the genome.

The fact that new spacers are consistently added to host CRISPR arrays has been leveraged into powerful technologies that harness CRISPR arrays as ‘molecular recording devices’^54–58. These technologies exploit the capacity of diverse Cas1–Cas2 integrases to continuously capture and directionally store DNA or RNA information over time in CRISPR loci of living cells. Ideally, sequencing and analysis of spacer location within the molecular recording arrays will allow researchers to derive time-ordered histories of biological processes, including transcription events and temporal gene expression pathways. CRISPR-based recorders may be optimized and/or their capacities expanded by considering the contribution of the protein factors like histones in new spacer uptake.

Methods

MNase DNA digestion and high-throughput sequencing analysis

P. furiosus COM1 strain⁵⁹ was grown in 250 ml cultures in defined media to either exponential or stationary phase, pelleted and flash frozen. Pellets were re-suspended in MNase buffer (50 mM Tris–HCl pH 8.0, 100 mM NaCl and 1 mM CaCl₂) then lysed mechanically by freeze/thawing with liquid nitrogen while grinding with a mortar and pestle. Lysate was clarified by centrifugation to remove cellular debris. RNA was removed by digestion with RNase A (Qiagen) at a final concentration of 4 mg ml⁻¹ and incubation at 37 °C for 1 h. About 500 units of MNase (ThermoFisher Scientific) was then added to the lysate, and 100 µl fractions were taken out at various timepoints. MNase reactions were stopped by the addition of phenol/chloroform/isoamyl alcohol (25:24:1) followed by vigorous mixing and centrifugation at 4 °C for 5 min at 21,000 x g. The aqueous layer was transferred into a 1.5 ml tube with 200 µl of 1 M Tris–HCl pH 8.0. DNA was ethanol precipitated, washed and re-suspended in 10 mM Tris–HCl pH 8.0.

MNase-digested DNA was visualized by ethidium bromide staining after separation on a 4% Nusieve agarose gel. Bands corresponding to digested products 30, 60, 90 and 120 bp in length were excised from the gel and purified using the Zymo Gel DNA Recovery Kit (Zymo Research). Purified DNA was then prepared for Illumina platform sequencing using the NEBNext Ultra II DNA library prep kit (New England Biolabs). As MNase-digested DNA was already fragmented, the fragmentation steps of the library prep protocol were excluded. Completed libraries were analysed for size distribution and DNA concentration using TapeStation (Agilent) and Qubit (ThermoFisher Scientific) instruments, respectively, then pooled and sequenced on a NextSeq 550 (Illumina) using a 2 × 75 paired end protocol and high output kit.

Following sequencing, reads were demultiplexed by index, adapter trimmed and aligned to a the P. furiosus reference genome and pT33.3 plasmid by bowtie2 version 2.5.0 (ref. 60). To visualize patterns of MNase-protected DNA, alignment outputs were processed using samtools version 1.9 and bedtools version 2.29.0 to generate genome coverage profiles (coverage BED files) for each sample^61,62, and then custom genome browser tracks were made with publicly available tools provided and described by the University of California Santa Cruz (kent-tools, https://genome.ucsc.edu). To examine genome-wide trends in read depth (as in Extended Data Fig. 3), custom python scripts were written to take genome coverage depth BED files as input and determine the depth of 60 bp and 90 bp read coverage in sliding 60 bp windows across the genome (1 bp increments). This generated a set of densities which was then binned, tallied and analysed to determine (1) the distribution of 60 bp read densities, (2) the ratio of 60 bp to 90 bp read coverage in each sliding window and (3) the distribution of 60 bp to 90 bp read coverage ratios. These ratios were output as matrix files which could be viewed in Microsoft Excel version 16.69.1.

Conjugative transfer of pT33.3 plasmid

Conjugative transfer of pT33.3 plasmid to P. furiosus strains were carried out as described in ref. 47. Briefly, Thermococcus donor strains were cultured in complex media at 85 °C, and the Pyrococcus recipient strains were cultured in defined selective media at 95 °C, all grown overnight for ~16 h. About 5 ml of each fresh, overnight culture was centrifuged at ~3,800 × g for 20 min. The supernatant was discarded and all remaining liquid media removed. Using a wide-bore pipette tip, cells were re-suspended in 50 µl of 1× Pfu Base Salts⁶³. Recipient cells were mixed with the appropriate donor in a 1.5 ml tube, for a total of 100 µl of cell re-suspension. The entire cell mixture was spotted in the centre of selective media plates. Plates were grown anaerobically for three nights at 95 °C. The spot of growth was scraped using the flat side of a micropipette tip into 5 ml of defined selective liquid media and grown overnight at 95 °C.

P. furiosus strain construction

Strains with histone A or B gene deletions were generated by homologous recombination with PCR products⁶⁴.

Growth of P. furiosus cultures retaining both or one histone-encoding loci

Strains were inoculated 1:100 from overnight cultures and growth monitored by changes in optical density at 600 nm at 95 °C. Anaerobic media was prepared and reduced as previously described⁶⁵. Culture density was recorded hourly for ~32 h. Average culture densities with standard error of the mean are reported for at least three biological replicates of each strain. Un-inoculated cultures were also monitored to provide a baseline as small changes in media absorbance occur during extended incubation at 95 °C.

Analyses of histone protein levels in total cell lysate from P. furiosus cultures retaining both or one histone-encoding loci by western blotting

Biomass from anaerobic 95 °C cultures was collected via centrifugation at mid-exponential and early-stationary phase, respectively, re-suspended and lysed in 25 mM Tris–HCl pH 8.0, 500 mM NaCl, 10% glycerol, 2% SDS, and total protein concentrations of each sample were quantified using a Qubit Protein Assay (Invitrogen). About 3 µg of total protein from each strain collected from either exponential or stationary phase cells was boiled and resolved through 4–20% Criterion TGX Stain-Free Protein gels alongside 500 ng of purified, recombinant Pyrococcus furiosus histone proteins A and B (HPfA and/or HPfB). Despite boiling samples in 2% SDS before loading the gels, the hyperthermophilic histones from P. furiosus retain dimer, tetramer and larger multimeric complexes even when resolved in SDS–PAGE, resulting in a ladder of histone complexes revealed by western blotting. Gels were transferred, probed with polyclonal antibodies raised against T. kodakarensis HTkA (that recognize HTkA, HTkB, HPfA and HPfB) and developed as previously described⁶⁶. Six independent biological replicates (culturing, biomass preparations and western analyses of total histone levels from each strain) were done with similar results.

RNA-seq analysis of wild type and histone deletion strains

For each of the three strains (wild type, ∆histone A, ∆histone B), five separate 250 ml volumes of liquid culture grown to stationary phase were centrifuged to pellet cells. Cell pellets were frozen and later re-suspended in 5 ml TRI Reagent RT (Molecular Research Center), incubated at room temperature for 5 min and then placed on ice. About 250 µl of 4-bromoanisole (BAN; Molecular Research Center) was added, and samples were mixed by inversion and then centrifuged at 12,000 × g for 15 min at 4 °C. The resulting 3 ml of aqueous layer was removed, added to 4.5 ml of 100% isopropyl alcohol and moved to −20 °C for at least 15 min before being centrifuged again for 30 min. Pelleted material was washed with 75% ethanol, re-suspended in 450 µl nuclease-free water (ThermoFisher) + 50 µl DNase I Reaction Buffer (NEB) + 4 µl DNase I (NEB) and incubated at 37 °C for 30 min. About 500 µl of 4 °C acid-phenol:chloroform IAA (ThermoFisher) was added, and the sample was mixed by inversion and centrifuged at 16,000 × g for 10 min at 4 °C. About 450 µl of aqueous layer was removed and added to 45 µl 3 M sodium acetate + 750 µl 100% isopropyl alcohol and moved to −20 °C for at least 15 min before being centrifuged at 16,000 × g again for 30 min. Pelleted material was washed with 75% ethanol and re-suspended in 40 µl nuclease-free water, and isolated RNA was quantified using a NanoDrop spectrophotometer (ThermoFisher).

The NEBNext rRNA Depletion Kit (NEB; E6310) was utilized to deplete P. furiosus rRNAs before making libraries for RNA sequencing. The ‘rRNA Depletion Solution’ provided in the kit was replaced with a mixture of 82 custom oligonucleotides (IDT) with complementary sequences to P. furiosus rRNA sequences and containing an equimolar concentration of 1 µM for each oligo. The protocol in the manual for the kit was followed using an input amount of 840 ng P. furiosus RNA for each sample with the only exception being that after the enzymatic steps the RNA was purified by using a Monarch RNA Cleanup Kit (NEB; T2030) rather than magnetic beads. rRNA-depleted RNA was eluted in 7 µl, and then 5 µl of that volume was used to make libraries via using a NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (NEB; E7760) and the protocol section in the manual for intact rRNA-depleted RNA. The libraries were amplified via 10 cycles of PCR with NEBNext Multiplex Oligos for Illumina (NEB; E6440), quantified by analysing on a 2100 Bioanalyzer (Agilent), mixed and paired-end sequenced 2 × 150 bp on an Illumina NovaSeq 6000 instrument. Sequencing was conducted on two separate runs. For the first run, sequencing resulted in between 109,066 and 140,590 reads (average 124,032) for each library. For the second run, sequencing resulted in between 105,874 and 44,868,432 reads (average 29,067,816) for each library. For both sequencing runs, the average insert size was between 142 bp and 159 bp.

Following sequencing, reads were demultiplexed by index, adapter trimmed and aligned to the P. furiosus COM1 reference genome by bowtie2 in local mode, with default settings⁶⁰. To look for differences in transcript abundances between wild type and ∆histone A or ∆histone B strains, we took the aligned RNA-seq data and determined per-gene read counts using bedtools intersect version 2.29.0⁶² and the available gene annotations for our reference genome. A custom python script was used to parse the output and generate a read count matrix which was then imported into DESeq2 (ref. 67). Samples were evaluated for differences in gene expression in two separate comparisons (wild type versus ∆histone A, wild type versus ∆histone B).

In vivo spacer integration assay

After the conjugation of the pT33.3 plasmid into P. furiosus strains, a population of cells (~20 colonies) was added to 5 ml of defined, selective, liquid media and allowed to grow at 95 °C overnight (~16 h). About 1 ml of overnight culture was then used for genomic DNA extraction using the Quick-DNA miniprep kit (Zymo Research). About 10 ng of genomic DNA was used in each PCR reaction. To detect addition of a new leader-adjacent spacer, PCR primers were designed to amplify between the leader and the first known spacer in the array. PCR products were then run on a 2.5% agarose gel to separate unexpanded bands (no integration of new spacers between the leader and the first known spacer) from expanded bands (integration of a new leader-adjacent spacer-repeat unit), and the DNA was detected by ethidium bromide staining.

Recombinant protein expression and purification

P. furiosus Cas1, Cas2, Alba, Trmbl2 and histone A and B genes were PCR amplified from genomic DNA, cloned into pET21d and transformed into E. coli BL21-RIPL expression strains. The C-terminal 6× histidine tag proteins (Cas1, Cas2, Alba, Trmbl2) were expressed and purified by heat precipitation and nickel chromatography as described in ref. 37. Histone A and B genes (from P. furiosus and T. kodakarensis) were expressed as native (untagged) proteins and purified by heat precipitation and heparin chromatography as described in ref. 45.

In vitro integration assay

A full CRISPR leader and an array consisting of either three repeats and two spacers (pCRISPR-short) or 11 repeats and 10 spacers (pCRISPR-long) were added to Blunt II TOPO vectors via blunt end ligation. These CRISPR-containing plasmids were used in in vitro integration assays to monitor the position of new spacer integration events. A final concentration of 1 µM Cas1, 1 µM Cas2, 1 mM DTT, 10 mM MgCl₂ and 50 nM of spacer DNA (with 5 nt overhangs) was incubated in reaction buffer (20 mM Tris, 100 mM KCl, 5% glycerol, pH 7.5) for 1 h at 4 °C. Histone concentrations were selected to range from fully saturated (that is, one histone dimer for every 30 bp of DNA in the reaction) to half saturation, to one-tenth saturation, which was 3.4 µM, 1.7 µM and 0.34 µM, respectively. A final concentration of 50 nM pCRISPR and varying concentrations as indicated of histones, Alba or TrmBL2 (0–5.1 µM) were added and incubated at 70 °C for 1 h. After incubation, the Cas1–Cas2 mixture was added, and the reaction was incubated for 1 h at 70 °C. The integration reactions were then stopped by addition of 25 mM EDTA, and proteins were degraded using 10 µg proteinase K (Life Technologies). DNA was purified from each reaction with the DNA Clean and Concentrator Kit (Zymo Research) and eluted in a volume of 15 µl of TE buffer. Integration reaction products were then assayed by PCR: 1 µl of the reaction product was added to a PCR with primers designed to amplify from the CRISPR leader to the incoming spacer, or from spacer 5 or 10 on the CRISPR array to the incoming spacer. PCR products were then run on a 4% Nusieve agarose gel at 120 V for 45–50 min for detection by ethidium bromide staining.

ImageJ and high-throughput sequencing analysis of in vitro integration products

The average intensity of each band (corresponding to an integration event at a repeat) in each lane was acquired using ImageJ version 1.53 (NIH). The normalized intensity of those bands was then determined: for a given band, the average intensity of that band was divided by the total band intensities observed for the whole lane multiplied by 100. For high-throughput sequencing, in vitro integration products were treated, prepared for sequencing and analysed as previously described³⁷. Briefly, the in vitro spacer integration assay was performed and DNA was isolated as described above. Excess un-integrated pre-spacer was removed using Agencourt AMPure XP beads (Beckman Coulter). Illumina adapter sequence with an N10 random primer was annealed to the plasmid DNA and extended (thermocycler conditions: 98 °C for 30 s, 25 °C for 30 s, 35 °C for 30 s, 45 °C for 30 s, and 72 °C for 5 min). Excess Illumina adapter was then removed using AMPure beads. A PCR was done to specifically amplify plasmid DNA that contained an integrated pre-spacer; the forward primer targeted the pre-spacer, while reverse primers targeted the Illumina adapter introduced with the random anneal and extension step. Illumina barcodes and additional adapter sequences were added with a final PCR, and the resulting library was separated on a 1% agarose gel to select for DNA in a 400 bp to 700 bp size range. DNA was isolated using the Zymo Gel DNA Recovery Kit (Zymo Research) and sequenced on an Illumina MiSeq in a 100 by 50 paired end run.

Biological materials used in this study are available upon request.

Reporting summary

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

Extended Data

Extended Data Fig. 1 |. Steps of integration.

Cartoon diagram shows the steps by which a spacer is integrated into the CRISPR array at the leader-adjacent repeat.

Extended Data Fig. 2 |. Distributions of aligned DNA sequences from MNase protection assay.

MNase protected DNA (Fig. 1a) was sequenced by HTS, aligned, and genome browser tracks were generated to examine the genome-wide distribution of aligned DNA fragments. Overlapping subtracks are color-coded according to the size of the DNA fragment (red: 30 bp +/− 5; golden yellow: 60 bp +/− 5; green: 90 bp +/− 5; blue 120 bp +/− 5). Positions of CRISPR loci are indicated (leader in pale yellow, repeat array in black). Approximately 10 kb of genome space upstream and downstream from each CRISPR locus is shown. Six replicates (three stationary growth, three exponential) were sequenced and yielded similar patterns; a representative replicate is shown here.

Extended Data Fig. 3 |. Positions of MNase protected DNA fragments and promoter elements within leaders of the seven CRISPR loci in P. furiosus.

Genome browser tracks were examined to determine the distribution of aligned MNase-protected DNA fragments within CRISPR leaders. Overlapping subtracks are color-coded according to the size of the DNA fragment (red: 30 bp +/− 5; golden yellow: 60 bp +/− 5; green: 90 bp +/− 5; blue 120 bp +/− 5). The x-axis indicates the cumulative depth of read coverage. Six replicates (three stationary growth (St), three exponential (Ex)) are shown. We noted that the 60 bp read coverage (peak height) over L1 and LR was unusually high compared to 60 bp coverage elsewhere in the genome. To quantify this, we analyzed sliding 60 bp windows across the genome, determined the density of 60 bp reads in those sliding windows, then binned and tallied the results to determine the distribution of 60 bp read densities. The values found under the label “60mer peak height” indicate the percentile of the L1 or LR peaks for each replicate; for example, the LR peak in CR1 exponential replicate 1 is higher than 99.92% of all 60 bp windows in the genome. We also noted that L1 and LR peaks usually had more 60mer coverage than 90mer coverage, even though 90mer reads were more abundant elsewhere in the genome. We used the sliding windows to quantify the 60mer to 90mer read density ratios across the genome. The values found under the label “60/90 peak ratio” indicate how biased towards 60mer read covereager the L1 or LR peak is; for example, the LR peak in CR1 exponential replicate 1 is more biased towards 60mer read coverage than 97.68% of all 60 bp windows in the genome.

Extended Data Fig. 4 |. Nucleotide alignment of seven CRISPR array leaders with MNase protected DNA peaks, promoter elements, and integration sites mapped.

Nucleotide sequences of seven leaders were aligned; conserved nucleotide regions are highlighted in gray. Peaks of MNase protected DNA are highlighted in yellow; the purple star and line at the leader-repeat junction show the position of typical, in vivo spacer integration. Promoter elements in the leader are labeled (BRE and TATA).

Extended Data Fig. 5 |. Growth, RNA expression, and histone abundance characteristics in wildtype, ∆histone A, and ∆histone B P. furiosus strains used in this study.

Total, ribo-depleted RNA was sequenced for wildtype, ∆histone A, and ∆histone B strains (five replicates each) and analyzed using the DESeq2 package to identify annotated transcripts with differential expression between (a) wildtype and ∆histone A and (b) wildtype and ∆histone B. Volcano plots show raw p-values against log₂-fold changes in abundance for each annotated transcript. A difference in transcript abundance was considered significant if the log₂-fold difference was equal or greater than 0.58 and the adjusted p-value was equal or less than 0.1; significantly different points are shown in blue; all other points for transcripts not meeting this threshold are colored gray. (c) Wildtype, ∆histone A, and ∆histone B strains were grown in liquid medium supplemented with pyruvate and Na₂S at 95 °C and optical density (OD_600nm) was measured every hour as a proxy for cell growth. Blank sample contained medium but no inoculum. Error bars show standard error of the mean for three biological replicates. (d) Western blots, employing polyclonal antibodies raised against HTkA (that recognize HTkA, HTkB, HPfA, and HPfB), were done on total protein from each strain harvested either from exponential (E, 10 hour time point) or stationary (S, 30 hour time point) phase cells. Recombinantly purified HPfA and HPfB were resolved on the gels to provide size references. Six biological replicates were run with similar results; a representative blot image is shown here. Although samples were boiled in 2% SDS before gel loading, histones retained dimer, tetramer, and larger multimeric complexes, giving rise to the multi-band ladder appearance visible on the gel.

Extended Data Fig. 6 |. Conservation of histones in Euryarchaeota and purification of recombinant proteins.

(a) Clustal Omega Multiple Sequence Alignment of histones from P. furiosus (Pfu A and Pfu B), Thermococcus kodakarensis (Tko A and Tko B), and Methanothermus fervidus (HMf A and HMf B). * = completely conserved residues,: = conservation between groups with strongly similar properties,. = conservation between groups with weakly similar properties. (b) SDS PAGE gels with the purified proteins used in the in vitro integration assays. Cas1 = 37.5 kDa, Cas2 = 10 kDa, P. furiosus histones A = 7.4 kDa and B = 7.3 kDa, T. kodakarensis histones A = 7.3 kDa and B = 7.1 kDa, TrmBL2 = 30.6 kDa.

Extended Data Fig. 7 |. In vitro evaluation of spacer integration into four CRISPR arrays in the presence or absence of P. furiosus histones.

(a) Gel images show representative results from PCR (carried out as in Fig. 5d) with primers targeting four CRISPR arrays: CRISPR5, CRISPR6, CRISPR7, and CRISPR8. The expected sizes for PCR products resulting from integrations at repeats 1 – 5 are marked with R1 – R5, respectively. Red asterisks identify the band corresponding to integrations at repeat 1, which is the natural, preferred point of integration in vivo. (b) Intensity of all PCR bands was quantified using ImageJ and the proportion of integration events at the five repeats was determined. Single biological replicates were evaluated with two different primer pairs producing with similar outcomes; one PCR reaction is shown and quantitated here.

Extended Data Fig. 8 |. MNase protection assay to characterize binding patterns for P. furiosus histones incubated with the pCR7-long plasmid.

(a) Gel image of DNA fragments generated by micrococcal nuclease digestion of pCR7-long plasmid incubated with purified recombinant P. furiosus histones A and B. Colored boxes highlight major bands of protected DNA. The gel image is for a digest containing 100 units of MNase; a 500 unit digestion was also done and resulting DNA bands were much fainter. (b) DNA bands highlighted in A were sequenced and genome browser tracks were generated. Overlapping subtracks are color-coded according to the size of the DNA fragment (red: 30 bp +/− 5; golden yellow: 60 bp +/− 5; green: 90 bp +/− 5; blue 120 bp +/− 5). The full CRISPR7 and part of one repeat are shown, with the purple line indicating the position where integration would typically occur in vivo. The x-axis indicates the cumulative depth of read coverage. One replicate was done for each treatment (100 and 500 units MNase).

Extended Data Fig. 9 |. In vitro evaluation of spacer integration into pCR7-long in the presence or absence of DNA-binding proteins.

(a) Diagrammatric representation of the PCR used to assess in vitro integration into pCR7-long. (b) Gel images show representative results from the PCR in (a) when recombinant purified histones, Alba, or Trmbl2 were added to the reaction at the indicated concentrations. The expected sizes for PCR products resulting from integrations at repeats 1 – 4 are marked with R1 - R4, respectively. Red asterisks identify the band corresponding to integrations at repeat 1, which is the natural, preferred point of integration in vivo. (c) Intensity of all PCR bands was quantified using ImageJ and the proportion of integration events at the five repeats was determined. Three replicates were assessed with similar outcomes, representative gels and quantifications are shown here.

Extended Data Fig. 10 |. Unbiased HTS sequencing of in vitro integration of a spacer into pCR7-long.

The in vitro integration assay with and without histones (a) or alba (b) was carried out, as before, and integration events were then sequenced using a two-step, semi-degenerate PCR protocol which targets all DNA fragments bearing the spacer. Sequencing results were used to generate genome browser tracks; the amplitude of the black peaks in these tracks indicate the total number of reads supporting an integration event at that nucleotide position. The positions of the leader (pale yellow) and repeats (black) are shown. Assay products with histones were sequenced for six biological replicates yielding very similar patterns; with alba one biological replicate was sequenced.

Data availability

Sequencing data generated for this study have been deposited in the NCBI Sequence Read Archive (SRA) under the BioProjectID PRJNA901420. Source data are provided with this paper.