1. Structural Biology and Molecular Biophysics

Casposase structure and the mechanistic link between DNA transposition and spacer acquisition by CRISPR-Cas

  1. Alison B Hickman
  2. Shweta Kailasan
  3. Pavol Genzor
  4. Astrid D Haase
  5. Fred Dyda  Is a corresponding author
  1. National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, United States
https://doi.org/10.7554/eLife.50004
Share this article
Cite this article
  1. Alison B Hickman
  2. Shweta Kailasan
  3. Pavol Genzor
  4. Astrid D Haase
  5. Fred Dyda
(2020)
Casposase structure and the mechanistic link between DNA transposition and spacer acquisition by CRISPR-Cas
eLife 9:e50004.
https://doi.org/10.7554/eLife.50004
  2. Download BibTeX
  3. Download .RIS

Abstract

Key to CRISPR-Cas adaptive immunity is maintaining an ongoing record of invading nucleic acids, a process carried out by the Cas1-Cas2 complex that integrates short segments of foreign genetic material (spacers) into the CRISPR locus. It is hypothesized that Cas1 evolved from casposases, a novel class of transposases. We show here that the Methanosarcina mazei casposase can integrate varied forms of the casposon end in vitro, and recapitulates several properties of CRISPR-Cas integrases including site-specificity. The X-ray structure of the casposase bound to DNA representing the product of integration reveals a tetramer with target DNA bound snugly between two dimers in which single-stranded casposon end binding resembles that of spacer 3'-overhangs. The differences between transposase and CRISPR-Cas integrase are largely architectural, and it appears that evolutionary change involved changes in protein-protein interactions to favor Cas2 binding over tetramerization; this in turn led to preferred integration of single spacers over two transposon ends.

Introduction

CRISPR-Cas systems, encoded in the genomes of many bacterial and archaeal species, function as defense systems against invading foreign nucleic acids such as bacteriophages and plasmids (Barrangou et al., 2007; reviewed in Mohanraju et al., 2016). They provide acquired immunity in which the first step in thwarting an attack involves extracting short stretches of the foreign nucleic acid and archiving these DNA segments (or ‘spacers’) at the CRISPR locus where they alternate with host-specific repeats; this part of the process is known as adaptation (reviewed in Marraffini, 2015; Sternberg et al., 2016; Amitai and Sorek, 2016; Jackson et al., 2017). The CRISPR locus is then transcribed as a long CRISPR RNA (pre-crRNA) that is digested into shorter pieces (crRNAs). These ‘guide’ crRNAs are incorporated into a ribonucleoprotein interference complex that can recognize and cleave its complementary target sequence, thus destroying the invading nucleic acid.

Many different types of CRISPR-Cas systems have been discovered, with varying cohorts of associated genes and genomic architectures (Makarova et al., 2018). Common to all known autonomous CRISPR-Cas systems is the Cas1 protein (Makarova et al., 2013; Makarova et al., 2018; Wright et al., 2019) which acts in concert with its binding partner Cas2 to recognize, process, and integrate protospacers into the CRISPR-Cas locus (Yosef et al., 2012; Nuñez et al., 2014). Notable exceptions are the type V-C and V-D systems which lack a cas2 gene (Shmakov et al., 2015; Burstein et al., 2017). Phylogenetic analysis of Cas1 proteins suggests that they may have originated from casposons, a novel class of putative DNA transposons in which a Cas1 homolog (the ‘casposase’) functions as the transposase (Figure 1A; Krupovic et al., 2014; Koonin and Makarova, 2017). If so, the evolution of CRISPR-Cas systems would parallel that of the adaptive immune system in higher eukaryotes, which evolved from an ancient DNA transposon (Kapitonov and Jurka, 2005; Carmona and Schatz, 2017).

Figure 1
Download asset Open asset
Methanosarcina mazei casposon organization and initial biochemical characterization.

(A) (left) Schematic of a representative casposon from Methanosarcina mazei (the gene organization of casposon MetMaz1FA1A3-C2 is shown; Krupovic et al., 2016) adjacent to host tRNA-Leu gene (red arrow). The last 4 bp of the tRNA-Leu gene overlap with 4 bp of the 14 bp target site duplication (TSD; green arrows) that flanks the casposon. The casposon Terminal Inverted Repeats (TIR) are shown as orange arrows. M. mazei casposon genes are as annotated by Krupovic et al. (2016), and are in gray with the exception of the putative Cas1-type casposase (in blue). Genes corresponding to a B-family type polymerase (‘polB’) and an HNH nuclease (‘HNH’) are marked. In the sequence below, a conserved 5'-CGCA motif at the 5'-end of the TSD is underlined. The sequence of the 31 bp TIR from strain S-6 and others, located at each casposon end, is shown in orange. (right) Relative to CRISPR Cas1, family two casposases have an additional C-terminal 60–70 residues predicted to form a helix-turn-helix (HTH) domain (Krupovic et al., 2014). Amino acid numbering is that of the M. mazei casposase. (B) In vitro integration assay of the M. mazei TIR into a target plasmid. The schematic at left shows the formation of relaxed and linear products corresponding to single-end (SE) and double-end (DE) integration of an oligonucleotide into a plasmid. The oligonucleotide is shown schematically in orange where the red dot indicates the 3'-OH nucleophile. In the agarose gels shown at the right, for the ‘no end’ controls, the plasmid was incubated for 60 min in the presence of protein and metal ion, but without the addition of the 31-mer TIR. For each reaction, the time points are: 0, 1, 5, and 60 min. Target: pUC19 (lanes 1–5); +14 bp TSD sequence (lanes 6–10); +TSD+15 bp more of the tRNA-Leu gene (lanes 11–15); +TSD+30 bp more of the tRNA-Leu gene (lanes 16–20).

How Cas1-Cas2 complexes mediate spacer acquisition has been well-established through a series of genetic, biochemical, and structural studies (Marraffini, 2015; Sternberg et al., 2016; Amitai and Sorek, 2016; Jackson et al., 2017). The active integrase is an elongated (Cas1)2(Cas2)2(Cas1)2 heterohexamer in which two catalytic Cas1 monomers carry out metal-dependent transesterification to integrate protospacers into the target repeat sequence (Yosef et al., 2012; Nuñez et al., 2014; Nuñez et al., 2015a; Nuñez et al., 2015b; Wang et al., 2015; Rollie et al., 2015; Xiao et al., 2017; Wright et al., 2017). Efficient protospacer integration requires 3'-single stranded overhangs on each spacer end, and integration of the two 3'-ends into opposing strands of the target leads to the duplication of the CRISPR repeat.

Much less is known about casposases. Although the casposase from the archaeon Aciduliprofundum boonei is an active DNA integrase in vitro (Hickman and Dyda, 2015; Béguin et al., 2016; Béguin et al., 2019), in vivo transposition activity has yet to be demonstrated and how casposon ends competent for integration might be generated remains a mystery. Casposons contain varying collections of other genes (as shown schematically in Figure 1A), but only the casposase and a family B polymerase appear common to all (Krupovic et al., 2017). It is possible that casposases possess an endonuclease activity that would generate the necessary free 3’-ends, but this has not yet been demonstrated (Hickman and Dyda, 2015). Alternatively, a cellular protein might be responsible for generating free casposon ends, by analogy to those proteins such as RecBCD or AddAB implicated in the generation of protospacers (McGinn and Marraffini, 2019). It has also been suggested that the casposon-encoded family B polymerase might be responsible for synthesizing linear copies of the casposon that could then be integrated (Krupovic et al., 2017).

Given the great abundance of DNA transposons and DNA transposase genes in nature - and in particular those with a catalytic core based on the RNase H-like fold (Aziz et al., 2010) - it is intriguing that a relatively sparsely distributed transposase was selected over the course of evolution to become the core of the CRISPR-Cas spacer acquisition system (Krupovic et al., 2014). Another puzzling aspect of the relationship is that, in most cases, DNA transposition involves strict specific recognition of transposon ends that are typically integrated into random locations in the genome whereas spacer acquisition involves the integration of many different fragments of invading DNA at a specific CRISPR locus just adjacent to a leader sequence.

To understand how Cas1 of CRISPR-Cas systems may have evolved from a transposase, we have characterized the biochemical properties of a casposase from Methanosarcina mazei. We have also determined its three-dimensional structure bound to branched DNA that represents the product of casposon end insertion into a specific sequence at the 3'-end of a host tRNA-Leu gene. We show that the casposase is tetrameric when bound to its target site and can site-specifically integrate a range of substrates including casposon ends with 3'-overhangs and even short single-stranded DNA ends as well as ssRNA. Our results suggest an evolutionary process in which a transposase that on its own recapitulates several canonical biochemical properties of CRISPR Cas1-Cas2 complexes acquired the ability to bind Cas2. This acquired interaction with a Cas2 dimer would have fundamentally changed the architecture of the integration complex, turning its properties away from transposon integration and towards spacer integration.

Results

The Methanosarcina mazei casposase is a site-specific integrase

The casposase from A. boonei, a species of archaea from the deep-sea hydrothermal vent euryarchaeota two lineage, is an active integrase that generates 14–15 bp target site duplications (TSDs) in vitro (Hickman and Dyda, 2015; Béguin et al., 2016; Béguin et al., 2019). It can integrate into random DNA but demonstrates a preference for site-specific integration into the very 3'-end of a host tRNA gene (Béguin et al., 2016; Béguin et al., 2019). Recently it was reported that a large family of closely related casposons are present in multiple strains of Methanosarsina mazei (Krupovic et al., 2016), an archaeal species that lives in semiaquatic and usually temperate habitats. Different strains of M. mazei have casposons in different genomic locations, strongly suggesting that these casposons are either currently active or have recently been active. One commonly observed genomic location for casposons in this species is within tRNA-Leu genes. Analysis of integration sites suggested that there is a short consensus sequence, 5'-CGCA, within the TSDs observed in vivo (Krupovic et al., 2016).

To determine if M. mazei casposase genes encode currently active integrases and whether they exhibit an insertion site preference, we expressed and purified a representative casposase (see Figure 6A), and tested its in vitro activity using an integration assay (Figure 1B). We initially used a 31mer dsDNA oligonucleotide (‘dsLE31’) which represents one of the casposon Left End (LE) sequences found in M. mazei strains; the particular sequence used (Figure 1A) is one of the shortest Terminal Inverted Repeats (TIRs) and is identical in multiple casposon copies in M. mazei (Krupovic et al., 2016). We used pUC19 and several variants of it as the target (Figure 1B). As M. mazei casposons have been frequently observed within tRNA-Leu genes (Krupovic et al., 2016), we elected to investigate one such possible target site. We first introduced 14 bp into pUC19 (‘+TSD’) corresponding to an observed genomic TSD comprised of the final 4 bp of the tRNA-Leu gene and 10 additional bp. The particular TSD sequence chosen (found in strain 1 .F.A.1B.3 and others, shown in green in Figure 1A) conforms to the in vivo consensus sequence in which a conserved 5'-CGCA motif is followed by an AT-rich sequence (Krupovic et al., 2016). We also generated pUC19 derivatives that contain either 15 or 30 more bp of the upstream tRNA-Leu gene (designated ‘+TSD+15 bp’ and ‘+TSD+30 bp’, respectively). Under the assay conditions used, the reaction time course shows that pUC19 itself is a poor target (lanes 1–5). Although there was little or no product formed when only the TSD (lanes 6–10) or TSD+15 bp (lanes 11–15) was introduced, relaxed products consistent with single-ended integration (SE) and linear products consistent with double-ended (DE) integration were readily observed when the TSD+30 bp of the tRNA-Leu gene were included (lanes 16–20). There was no further increase in activity when longer regions of the tRNA-Leu gene were included in the target plasmid (data not shown). These results suggested that integration by the M. mazei casposase is target-specific, and that the target site is contained within the sequence corresponding to TSD+30 bp. This is similar to the reported in vitro targeted integration of the A. boonei casposon into the 3' end of a host tRNA-Pro gene, although in that case, the data suggested that the target of this casposase is contained within a distinct TSD+15 bp region (Béguin et al., 2019).

To rule out the possibility that the casposon end oligonucleotide we used could be integrated using either of its terminal 3'-OH groups, we tested substrate derivatives with different phosphorylated 3' ends that would prevent their use as nucleophiles (Figure 2A). In this assay, we used a shorter casposon end oligonucleotide (dsLE17) which was also an effective substrate in the production of relaxed and linear products when incubated with the pUC19+TSD+30 bp target plasmid (lanes 1–4). It has previously been shown for the A. boonei casposase that ends considerably shorter than the entire TIR are competent for in vitro integration (Hickman and Dyda, 2015; Béguin et al., 2019). As shown in Figure 2A, relaxed and linear products were still formed when the remote 3'-OH group of dsLE17 was blocked by phosphorylation (lanes 5–7); conversely, when the nucleophilic 3'-OH group at the casposon end was blocked, there was no detectable product formation (lanes 8–10). In the context of the blocked casposon end, we also extended the top strand with 7-nt of ssDNA, and again no activity was detected (lanes 11–13). These results indicate that the casposase differentiates between the available 3’-OH groups of supplied oligonucleotides, and uses only the 3’-OH group at the casposon end as the nucleophile for integration.

Figure 2 with 1 supplement see all
Download asset Open asset
In vitro integration activity properties of the M. mazei casposase.

(A) Effect of blocking free 3'-OH ends on product formation. (left) Sequences of TIR oligonucleotides used to assay integration into pUC19+TSD+30 bp. The 3'-end C at the casposon end is in bold. (right) For each reaction shown, time points are: 0, 1, and 60 min. (B) Integration into the specific target plasmid (pUC19+TSD+30 bp, top) compared to integration into pUC19 (bottom). Integration was assayed using a variety of substrate forms as shown (all sequences can be found in Table S1). For each reaction, time points are: 0, 0.5, 1, 5, and 60 min. Reactions were performed in triplicate, and product quantitation is in Figure 2—figure supplement 1. (C) Integration of single-stranded casposon ends. For each reaction, time points are: 0, 1, and 60 min.

As the precise nature of the casposon ends produced and used for integration is not known, we tested a variety of oligonucleotide substrates to determine if the M. mazei casposase demonstrates integration preferences similar to those of Cas1-Cas2 integrases (Figure 2A,B). Time courses were carried out in triplicate to provide quantitative information on the relative effectiveness of the substrates (Figure 2—figure supplement 1). Random oligonucleotides were included as controls of substrate specificity. One substrate we tested was TR17 (a 17mer casposon end which is recessed by five 5’-nt, ‘Top strand Recessed’, that is, there is a 5-nt 3’-overhang). This choice was motivated by the apparent preference of Cas1-Cas2 of some CRISPR-Cas systems to integrate spacers with short 3'-overhangs (Nuñez et al., 2015a; Wang et al., 2015; Xiao et al., 2017). We also included the single-stranded casposon left end bottom strand, ssLE17, as it has been shown by Béguin et al. (2019) that single-stranded casposon ends are effective substrates for the A. boonei casposase.

As shown in Figure 2B and Figure 2—figure supplement 1, with the exception of the random single-stranded 17mer (‘ssran17’) control, all oligonucleotide substrates tested produced relaxed and linear products consistent with single-ended and double-ended integrations when the specific target plasmid pUC19+TSD+30 bp was used, but not with pUC19. The level of production of the linear plasmid product (i.e., consistent with double-ended integration) was the highest for TR17; dsLE17 and ssLE17 produced similar intermediate levels, and ranTR17 and dsran17 were the least effective substrates.

We were intrigued by the capacity of the casposase to integrate ssDNA, and we repeated the assay with casposon-specific ssDNA oligonucleotides of varying length (Figure 2C). Remarkably, casposon ends as short as six nt could be specifically integrated, a result that has also recently been demonstrated for the A. boonei casposase (Béguin et al., 2019). Taken together, the results indicate that the M. mazei casposase has a wide tolerance for the type and sequence of oligonucleotides it can integrate, but with an apparent preference for a partially single-stranded casposon end. The putative evolution from transposition to spacer integration must therefore have involved acquiring the ability to discriminate among possible substrates and to integrate only short, recessed spacers of relatively specific length.

In order to quantify precise target site-specificity, we performed next-generation sequencing (NGS) of the linear plasmid products generated when pUC19+TSD+30 bp was used as the integration target (Figure 3A–C). Assays were carried out using either the 31mer casposon TIR (dsLE31) or a 29mer casposon end recessed by 5-nt on the top strand (TR29). As NGS has not been reported for casposases to date, we varied the sample preparation method in several ways. For four samples, linear reaction products resulting from the independent incubation with dsLE31 with pUC19+TSD+30 bp were isolated from agarose gels and subjected to either 17 cycles (duplicate samples S1A1, S2A1) or 30 cycles (duplicate samples S3A2, S4A2) of PCR prior to the preparation of sequencing libraries. We also carried out a large-scale reaction using TR29 so that enough linear product could be obtained directly to avoid the need for PCR amplification prior to library preparation (sample S5D). This allowed us to assess the relative integration frequency for the two possible orientations of the casposon end, as our PCR strategy would not detect products resulting from the use of the remote 3'-OH as the nucleophile.

Figure 3 with 1 supplement see all
Download asset Open asset
Characterization of target-site specificity.

(A) Double-end integration products of five independent reactions (sample S1-S5) were purified and analyzed by next generation sequencing (NGS). Sample 1–4 were PCR-amplified before sample preparation for NGS. Sample five was directly subjected to NGS sample prep. The five independent reactions show the same dominant target site (TS1). The substrate sequence 3’ of the novel insertion identifies the 14 bp target-site. Reads were collapsed by unique sequences, clipped 3’ of the spacer sequence, and aligned to the target plasmid (pUC19+TSD+30 bp). Mean read frequencies (+ /- standard error) for insertions on the positive (blue) and negative (red) strand (A) and a representative sequence logo (C) are depicted. The negative strand logo has been flipped to visualize the sequence complementarity. The CGCA motifs are highlighted. (B) Sample S5D showed additional minor target sites (TS2 and TS3) with similar target motifs (C). (D) Target variants for assay of TR17 integration into a 40-mer target. Nucleotides in red indicate where targ40 sequences were mutated. (E) As shown schematically at the right, integration into the target top strand generates a 22-mer and 35-mer product pair; integration into the bottom strand generates a 53-mer (and a 4-mer too short to be detected). For each reaction, time points are: 0, 0.25, 1, 2, 5, and 60 min. The first lane in each gel contains standard size markers for reference. (F) Integration of a 20mer casposon end (‘TR’ on the right) into targ40 as a function of the number of recessed nucleotides on the top strand. The experiment is shown schematically at the right where the recessed strand is shown as the dashed orange line. Reactions were performed in triplicate, and quantitation of product formation (as measured as band intensity) for both the bottom strand (blue) and top strands (orange) is shown in (G). Plotted points represent the mean, and the error bars correspond to the standard deviation. ssRNA can also be integrated (Figure 3—figure supplement 1).

The results reveal that integration occurred overwhelmingly and with precision into the specific target site of pUC19+TSD+30 bp (designated target site one or TS1), with sequences from all five samples revealing overlapping positive and negative strand insertions consistent with the 14 bp offset corresponding to the size of casposon TSDs (Figure 3A–C). For the four samples generated using PCR amplification, no other integration sites were detected (Figure 3B). For sample S5D, 99.4% of the detected integration events were at TS1, and only two other sites in the pUC19 backbone were detected with reads corresponding to integration into both strands of the target (TS2,TS3; Figure 3B,C). Alignment of the three sites (Figure 3B, inset) indicated that TS2 and TS3 contain a conserved 5'-CGC motif at one site of integration. Notably, all three observed target sites have a C nucleotide at the −2 position flanking the integration site but no other sequence conservation. We detected no sequences that would have been generated if the opposite 3'-end of the casposon end oligonucleotides were used as the nucleophile. Collectively, these results suggest that integration by the M. mazei casposase is remarkably sequence-specific, and that the most important target sequence features directing integration are clustered around the 5'-CGCA motif detected in vivo (Krupovic et al., 2016).

To explore aspects of target recognition by casposases in more detail, we employed an integration assay using short target oligonucleotides (Figure 3D) containing the region we established as sufficient for site-specific targeting. Integration into a 40mer target (‘targ40’) oligonucleotide by TR17, the most active substrate in the experiments in Figure 2B, was readily detectable (Figure 3E) and yielded products during the time course corresponding to the expected size for site-specific integration. The conserved 5'-CGCA target motif was critical as its mutation resulted in almost complete loss of both top and bottom strand integration (‘mutCGCA’). We also established that there was no detectable integration into a random 40mer (‘ran40’). Mutation of four bp just 5' of the CGCA motif (‘mutL’) severely reduced integration as did mutating the last five bp of the TSD (‘mutR’), although to a lesser extent. As the plasmid assay had revealed that integration is dependent upon some portion of the upstream 15–30 bp tRNA sequence, we mutated the region corresponding to bp 12–17, and this abolished activity (‘mut12-17’). We were also curious how symmetrizing the TSD (i.e., placing the 5’-CGCA motif on both strands at the target site) might impact activity and we found it was as effective a target as the original TSD (‘symTSD’).

We took advantage of this assay to more finely dissect the stimulatory effect of recessing the top strand of the casposon end (Figure 3F). Starting with a blunt-ended dsLE20 substrate, we progressively deleted single nucleotides from the top strand to thereby generate a series of substrates with progressively longer 3’-single-stranded overhangs. The assay was performed in triplicate and the product bands for integration into the top and bottom strands separately quantitated (Figure 3G). Under these conditions, there was little detectable product formation with the blunt-ended substrate and only slight stimulation with a 1-nt 3'-single-stranded overhang. Maximal activity for bottom strand integration was obtained once the 3'-overhang was 2-nt and was thereafter essentially unchanged as the length of the overhang was increased to 7-nt. Integration into the top strand as a function of 3'-overhang length lagged behind that of bottom strand integration, and reached its maximal level once the 3'-overhang was 4-nt. Although it is not clear if the difference between the top and bottom strand is significant or a consequence of the asymmetry of the target, these results suggest that only a few unpaired nucleotides are required at the casposon end for optimal integration activity. It is possible that the reduced activity we observe with blunt-ended substrates reflects a need for transient fraying at the end for binding and integration.

Lastly, encouraged by the tolerance of the M. mazei casposase towards integration substrates and by the report of direct integration of ssRNA by a type III reverse transcriptase-Cas1 fusion protein (Silas et al., 2016), we also tested the RNA version of ssLE17. As shown in Figure 3—figure supplement 1, ssRNA also can be integrated into the target oligonucleotide targ40 although far less efficiently.

M. mazei casposase is tetrameric when bound to target and its casposon ends

We next investigated the in vitro DNA binding properties of casposases using interferometric scattering mass spectrometry (iSCAMS, Young et al., 2018) and analytical size exclusion chromatography (SEC). At the low protein concentrations used for iSCAMS (25–50 nM), the M. mazei casposase had a measured MW of 95 kDa (Figure 4), consistent with a dimer and in keeping with previously characterized CRISPR Cas1 proteins and casposases (Wiedenheft et al., 2009; Hickman and Dyda, 2015). In the presence of a 17-nt ssDNA casposon end (ssLE17), the MW shifted to 109 kDa, suggesting binding while the protein remained dimeric. Due the uncertainty of this measurement, we could not distinguish between one or two casposon ends bound per dimer (104 kDa vs. 109 kDa). In the presence of a branched oligonucleotide substrate comprised of two single-stranded 8-nt casposon ends linked to a symmetrized target sequence mimicking the integration product (minus any overhangs that might be present in vivo), the resulting protein-DNA complex had a measured MW of 219 kDa, consistent with a tetramer bound to one integration product molecule.

Figure 4 with 2 supplements see all
Download asset Open asset
Characterization of casposase-DNA complexes.

Characterization by interferometric scattering mass spectrometry (iSCAMS) analysis. Casposase alone (left; MWcalcmonomer = 49.4 kDa, dimer = 98.8 kDa), casposase bound to ssLE17 (center; MWcalc dimer+ssLE17 = 104 kDa, dimer+2xssLE17 = 109.1 kDa), and casposase bound to integration complex used for crystallographic studies (right, MWcalcDNA = 13.3 kDa; tetramer+DNA = 210.9 kDa). Each experiment was performed once for each indicated binding substrate, and all of the data is included in the mass histograms. The reported error corresponds to the fitting error to a Gaussian distribution model. Characterization by size-exclusion chromatography (SEC) is shown in Figure 4—figure supplements 1 and 2.

We used SEC to resolve the uncertainty in the protein-DNA binding stoichiometry with casposon ends. When the casposase was titrated with either TR17 or ssLE17, the casposase bound with a stoichiometry of one end per dimer (Figure 4—figure supplement 1). During the course of these experiments, it became clear that the elution position of the casposase varied as a function of concentration (Figure 4—figure supplement 2) whereas its complexes with either TR17 or with the integration product studied by iSCAMS eluted at positions expected for a tetramer over a range of protein concentrations. Thus, it appears that the casposase exhibits a concentration-dependent equilibrium between dimeric and tetrameric forms, and that this equilibrium is affected upon binding to DNA substrates.

In an effort to further characterize the casposase and its interactions with its substrates, we attempted to crystallize the casposase bound to DNA oligonucleotides representing different stages of the integration reaction. Many casposase-DNA complexes yielded poorly-diffracting crystals including the casposase bound to recessed dsDNA oligonucleotides and ssDNA oligonucleotides of varying length representing the casposon ends. This was also the case with casposase bound to mimics of the post-integration step using both symmetric and asymmetric integration target sites and casposon ends of varying length and dsDNA-ssDNA composition. We were eventually able to obtain diffracting crystals of the casposase bound to a symmetric 14mer target site into which two single-stranded 8-nt casposon ends were integrated (the integration product shown in Figure 4), and the structure of the complex was solved to 3.1 Å using experimental phases obtained from a selenomethionine derivative (Table 1; Figure 5; Figure 5—figure supplement 1).

Figure 5 with 4 supplements see all
Download asset Open asset
Structure of M. mazei casposase bound to an integration product mimic.

(A) Two single-stranded casposon ends (5'-TATATCCC; shown in orange) are integrated into a 14 bp dsDNA target (5'-CGCACTTAAGCGTG; shown in green). This color code is maintained throughout. In the tetramer, catalytic subunits (B and D) are in dark blue and noncatalytic subunits in light blue (A and C). Views of the DNA alone are shown to the right of each view in the same orientation. Active site residues are shown as sticks (pink). (B) Schematic showing hydrogen bond contacts between protein and DNA using the same color code as (A). Red circles marks the phosphates at which integration occurred. Green circles correspond to bound calcium ions. (C) Detailed view of the active site.

Table 1
X-ray structure determination detail.

Data set		E1E2E3Refinement
Space groupP43212P43212P43212P43212
Wavelength (Å)0.979300.979160.968020.97918
Unit cell parameters
a = b, c (Å)106.91, 424.61106.98, 424.24106.83, 423.30106.99, 423.36
α = β = γ (°)90, 90, 9090, 90, 9090, 90, 9090, 90, 90
Resolution range (Å)30.0–3.230.0–3.230.0–3.230.0–3.1
No. of unique reflections77309774647707645362a
Redundancy7.087.137.1914.52
I/σI (3.29–3.2 Å)12.16 (1.28)12.45 (1.54)12.26 (1.93)13.3 (1.95)
BWilson (Å2)100.793.8890.13104.36
Completeness (%)99.899.899.899.8
Rmerge (%)b18.218.014.711.4
CC1/2 (last resolution shell)49.260.877.081.9
Phasing power (iso/ano)0.0/1.0450.214/0.9950.877/0.812
Figure of merit at 3.2 Å0.38
Structure Refinement
Resolution (Å)3.1
Rwork/Rfree c0.198/0.256
RMSD
Bond lengths (Å)0.01
Bond angles (°)1.17
Average B factor (Å2)142.33
No. atoms
Protein11442
DNA846
Ligands (Ca2+)6
Waters36
No. of reflections45362
Ramachandran plot (%)
Favored86.6
Allowed13.1
Outliers0.2

  1. Numbers in parentheses represent the value for the highest-resolution shell. RMSD, root-mean-square deviation.

    a Friedel's law true.

  2. b Rmerge = Σ|Ii − <I>|/ΣIi, where Ii is the intensity of measured reflection and <I> is the mean intensity of all symmetry-related reflections.

    c Rfree = ΣT||Fcalc| – |Fobs||/ΣFobs, where T is a test dataset of about 5% of the total unique reflections randomly chosen and set aside prior to refinement.

The pseudo two-fold symmetric complex is a clamp-like tetramer in which two casposase dimers hold the double-stranded target site between them, with the 5' termini of the single-stranded casposon ends emerging through a narrow slot in the bottom of the clamp (Figure 5A,B). Each casposase monomer exhibits the fold expected of a Cas1 homolog (Wiedenheft et al., 2009; Figure 5—figure supplements 2 and 3), with a canonical eight-stranded N-terminal domain and a globular catalytic domain that brings conserved active site residues (E162, H232, E247) into close proximity (Figure 5C). When compared to its most closely related structurally characterized Cas1 homolog, that from Archaeoglobus fulgidus (Kim et al., 2013) with which it shares 32% sequence identity, the N-terminal domains have an RMSD of 1.68 Å over 72 shared Cα positions and the catalytic domains are even more similar, with an RMSD of 1.28 Å over 208 shared Cα positions.

The casposase dimers that are the opposite sides of the clamp are each formed by the juxtaposition of two N-terminal β-strand domains into two extended eight-stranded β-sheets (Figure 5—figure supplement 2, top), as has been observed for CRISPR Cas1 dimers. In the integration product complex, each dimer is functionally and structurally asymmetric with a ‘catalytic’ monomer whose active site contains the scissile phosphate group that links transposon end DNA and the target site, and a ‘noncatalytic’ monomer in which the active site is located on the periphery of the assembly facing away from the DNA. This functional asymmetry at the dimer level recapitulates the observed asymmetry of Cas1 dimers in structurally characterized CRISPR Cas1-Cas2 integration complexes bound to DNA (Nuñez et al., 2015b; Wang et al., 2015; Xiao et al., 2017; Wright et al., 2017).

In the complex, there is interpretable density for all four monomers until amino acid 341, but thereafter the C-terminus is not visible in two of the monomers and, in the other two, there is interpretable density only until residues 354 and 359. We confirmed that the C-terminal residues are still present in the crystallized complex and the protein had not been proteolytically degraded (Figure 5—figure supplement 4). Close examination of solvent channels and difference electron density maps revealed density consistent with isolated α-helices that are likely regions of the C-terminal HTH domain but we are unable to assign them to a specific register. Thus it appears that when the casposase is bound to this integration product mimic, the C-terminal residues remain disordered within the crystal and do not stably engage with the rest of the protein or DNA. In an attempt to provide insight into the role of the C-terminus, we expressed and purified a C-terminally truncated version of the casposase consisting of residues 1–341 (ΔHTH, Figure 6A), and assayed its activity against the range of substrates previously evaluated for the full-length protein. As shown in Figure 6B, there was no evidence of integration activity with any of our integration substrates, using either the specific pUC19+TSD+30 bp target or on pUC19 alone. Thus, the casposase N-terminal and catalytic domains alone are not sufficient for integration, and the HTH domain must play a crucial role.

Figure 6
Download asset Open asset
Details of the casposase structure.

(A) SDS-PAGE analysis of proteins expressed and purified in this work. (B) Integration activity of the C-terminally truncated casposase, ΔHTH. There is no evidence for integration activity with either a target plasmid containing the casposon specific target site (top) or pUC19 alone (bottom). (C) Interactions at the casposase tetramerization interface. Shown are the hydrogen bond contacts between monomers A (noncatalytic) and D (catalytic), which are recapitulated between monomers B and C (not shown).

In the tetrameric complex with the integration product DNA, we observe two dimer-to-dimer interactions (Figure 6C) formed by the convergence of helix α1 (residues 59–67) and a loop (residues 39–44 between β4 and β5) from each noncatalytic monomer with two loops (residues 182–199 and 275–278) from the opposing catalytic monomer. Each set of interactions buries a surface area of ~1200 Å in the tetramerization interface and together form the base of the clamp in which the dsDNA target is enclosed.

dsDNA target recognition

The M. mazei casposase generates 14 bp TSDs (Figure 3A–C), in sharp contrast to CRISPR-Cas systems where the CRISPR repeats that are the TSDs (generated after spacer integration and subsequent repair by DNA replication) are generally 25–50 bp. In the case of CRISPR-Cas integrases, both Cas1 and Cas2 participate in binding the CRISPR repeat (Xiao et al., 2017; Wright et al., 2017). In the casposase complex structure, the total casposase-dsDNA interface buries ~4700 Å2 of solvent-accessible surface but only two specific contacts mediate the interaction with the target DNA: between R191 and G22 (i.e., the base opposite the first C of the 5'-CGCA motif) and between R192 and G20 (the base opposite the second C of the 5'-CGCA motif) (Figure 7A,C). The remaining interactions are non-specific, including contacts to the phosphate backbone from the main chain amides of T309 and E233 and from the side chains of N69 and H323. There is also a divalent ion-mediated interaction from a calcium ion contributed by the crystallization buffer with two phosphate groups and the carboxylate of D46. The bound target in the integration product complex is almost perfectly straight B-form DNA, unlike the severely bent targets observed in other DNA transpososomes and retroviral intasomes (for example, Montaño et al., 2012; Maertens et al., 2010).

Figure 7 with 1 supplement see all
Download asset Open asset
Binding of the casposon target site and casposon ends.

(A) Hydrogen-bond contacts between the casposase and target DNA (casposon ends omitted for clarity). Bound Ca2+ ions are shown in green. (B) Comparison of binding of ssDNA casposon ends (left) and the ssDNA 3'-overhangs of bound spacers in the E. coli type I-E (center; 5vvk) and E. faecalis type II-A (right; 5xvp) integrases. Structures were aligned based on the catalytic monomers. Active site residues are shown in pink, and the red arrow indicates the 3'-OH group where the casposon end is joined to target. (C) Detailed view of the interactions of R191 and R192. (D) Detailed view of the interactions involving C7 and C8. (E) Detailed view of the interactions involving C6. (F) Integration activity of point mutants using pUC19+TSD+30 bp as the target plasmid and TR17 as the casposon end mimic. For each protein, time points are: 0, 1, and 60 min. Comparison of integration activity of purified casposase mutants with varying casposon end oligonucleotides is shown in Figure 7—figure supplement 1.

To confirm the catalytic importance of R191 and R192, we expressed and purified a double mutant (R191A/R192A; Figure 6A) and tested its activity in the plasmid integration assay. When the optimal oligonucleotide substrate, TR17, was used, there was little or no relaxed or linear products observed (Figure 7F). We also tested a range of different casposon end substrates, and again there was little evidence for any activity (Figure 7—figure supplement 1), supporting the notion that these two arginines play important roles during integration.

ssDNA casposon end binding resembles that of spacer 3' overhangs

A rich set of protein-DNA contacts line the path of each single-stranded casposon end from the catalytic site to the protein surface (Figure 7B). Although the only base-specific interaction is between the first C of the casposon DNA (C8) and the side chain of K238, several residues interact with the phosphate backbone (H232, R250, Y206, R280, and R287) and there are arrays of stacking interactions. Most notably, the two first Cs of the casposon end are stacked on each other and both sandwiched between Y243 and M194; Y243 in turn is stacked on the side chain of N239 (Figure 7D). The third C of the casposon end (C6) is stacked between Y206 and R280 (Figure 7E). Finally, the fifth base of the end (A4) is stacked on W77 (Figure 7E). Through these extensive interactions, each single-stranded casposon end buries ~1700 Å total solvent-accessible surface area. We also note that the phosphate groups of C6 and C7 of each casposon end are located within 5 Å of the phosphate backbone of the target DNA, an unusually close approach stabilized in part by the binding of a Ca2+ ion, which in turn is enabled by the syn conformation of C7.

To test the relative importance of the observed interactions in integration activity, we expressed and purified several mutants expected to disrupt sets of interactions including the double-mutants M194A/Y243A, Y206A/R280A, and the single mutant W77A (Figure 6A). Each of these mutants had severely reduced integration activity when assayed using TR17, most notably for the production of the linear product (Figure 7F). When other casposon end substrates were tested (Figure 7—figure supplement 1), there were varying levels of residual activity for the production of the relaxed form of the pUC19+TSD+30 bp plasmid (consistent with SE joining), but none of the mutant proteins generated a linear product consistent with double-ended integration. Interestingly, all of the mutants showed a higher activity level with a casposon-specific ssDNA end (ssLE17) than with non-specific DNA (ssran17).

The path followed by ssDNA from the casposase active sites is reminiscent of those of the 3' overhangs of CRISPR spacers (Wang et al., 2015; Nuñez et al., 2015b; Xiao et al., 2017) although the number and type of interactions differ (Figure 7B). Consistent with the importance of ssDNA overhangs at the end of spacers, several CRISPR Cas1 proteins have been reported to bind ssDNA, and although ssDNA can be disintegrated by Cas1 (Rollie et al., 2015), stable half-site integration of single-stranded protospacers has not been detected (Nuñez et al., 2015b; Fagerlund et al., 2017; Rollie et al., 2018) although an exception has been reported (Silas et al., 2016). These collective observations suggest that the putative evolution from transposition to spacer integration may have involved the loss of the coupling between ssDNA binding and subsequent integration.

Converting a casposase into a CRISPR-Cas integrase

To understand the structural relationship between the casposase integration product complex and CRISPR Cas1-Cas2 assemblies bound to DNA representing a spacer integrated into a CRISPR repeat, we first compared their aligned dimer structures. As can be seen in Figure 8A, the mode of target binding by the casposase resembles that of E. faecalis and E. coli Cas1 bound to CRISPR repeats, and the binding mode of the ssDNA casposon ends resembles that of 3'-overhangs of CRISPR spacers (Wang et al., 2015; Nuñez et al., 2015b; Xiao et al., 2017; Wright et al., 2017). However, when the casposase tetramer is considered (Figure 8B), it is clear that one of the casposase monomers occupies the same architectural space as one of the Cas2 subunits in CRISPR Cas1-Cas2 complexes (Figure 8B, center), and it would not be possible to retain the ability to tetramerize in the manner observed and to simultaneously bind Cas2.

Figure 8
Download asset Open asset
Features of the casposase-DNA complex.

(A) Comparison of a DNA-bound casposon dimer (left) to full-site bound CRISPR Cas1-Cas2 integrases of E. faecalis type II-A CRISPR Cas1-Cas2 (PDB ID 5xvp, center) and E. coli type I-E (PDB ID 5vvk, right). Aligned dimers are boxed in gray. Casposon ends and spacers are shown in orange, target in green, and leader DNA in red. The final visible amino acid in the electron density on each casposase C-terminal end is indicated. (B) Superposition of the structures (center; casposase monomer D aligned to E. coli Cas1c) shows that the location of the casposase monomer B (in blue) overlaps that of Cas2.

At the level of the casposase dimer, however, there does not appear to be an obvious structural impediment to interacting with a Cas2 dimer (Figure 9A). Although the specifics of Cas1-Cas2 interactions are not conserved across CRISPR-Cas systems (Wang et al., 2015; Nuñez et al., 2015b; Xiao et al., 2017), when the relevant interfaces were compared, it appears that mutation of only a few surface residues of the casposase might be sufficient to convert it into one with both a complementary shape and the necessary residues to bind Cas2. For example, the interaction between E. faecalis Cas1 and Cas2 is mediated in part by a β-sheet formed between β1 of Cas1 and a β-strand at the C-terminus of Cas2, and β1 of the casposase appears similarly poised (Figure 9A).

Figure 9 with 1 supplement see all
Download asset Open asset
Proposed transformation from transposase to spacer integrase.

(A) Comparison of the surfaces of casposase dimer and Cas1-Cas2 of E. faecalis (5xvp). The structures were aligned based on the catalytic monomers of each. Also shown is a close-up of the interaction between the E. faecalis Cas1 N-terminal domain (green) and Cas2 (gold); the aligned casposon N-terminal domain is overlaid in blue. In the inset, the view is based on the alignment between noncatalytic domains. (B) Proposed evolutionary transformation from casposase to CRISPR-Cas integrase in which Cas2 becomes a preferred binding partner of a casposase dimer, in turn leading to the preferential integration of a spacer over the original casposon. See Figure 9—video 1.

It is not clear what changes in DNA recognition might need to accompany the architectural change from tetramer to Cas2-binder. Comparison of target binding sites in both integrase systems shows that the overall mode of binding is very similar, although the M. mazei casposase has base-specific contacts involving R191 and R192 that presumably would have to be lost since Cas1-Cas2 structures have shown little evidence for sequence-specific contacts with the CRISPR repeats (Wright et al., 2017; Xiao et al., 2017). Another obvious difference is the lack of the casposase HTH domain in Cas1-Cas2, which we have shown is crucial for casposon integration. Until we have a better understanding of the role of the HTH, missing in the structure here, it remains unclear how Cas1-Cas2 compensated for its loss.

Discussion

Evolutionary studies on CRISPR-Cas systems have suggested that some Cas proteins of effector complexes as well as Cas1 of the spacer integrase have arisen from mobile genetic elements (Krupovic et al., 2014; Koonin and Makarova, 2017; Faure et al., 2019). The deep branching of casposons within the Cas1 phylogeny (Makarova et al., 2013; Makarova et al., 2018) and the clear mechanistic similarity between DNA transposition and CRISPR spacer acquisition makes the idea of an evolutionary relationship highly attractive. Like transposases and integrases with RNase H-like catalytic domains, the CRISPR-Cas integrase carries out transesterification to insert one piece of DNA into another, generates uniform-sized TSDs upon insertion, and produces a head-to-tail array of integrated DNA segments through tandem insertion as often seen with insertion sequences and transposons (Siguier et al., 2015). Yet, the putative evolutionary ancestor of the CRISPR-Cas integration system is not from one of the prevalent DNA transposon superfamilies but, rather, from the relatively sparsely distributed casposons (Krupovic et al., 2014). The reason for this may lie in the unusual capacity of the M. mazei casposase to integrate a variety of substrate types, including nonspecific and random DNA, into a strongly preferred target site, the important characteristic if a CRISPR array is to serve as an effective library of a range of previously encountered foreign DNA. Thus, on the hypothetical evolutionary pathway towards spacer integration, casposases appear well-suited to act as the starting point for a system to integrate a range of largely unrelated DNA sequences.

Our biochemical results described here indicate that the M. mazei casposase already possesses several observed properties of Cas1-Cas2 integrases (Marraffini, 2015; Sternberg et al., 2016; Amitai and Sorek, 2016; Jackson et al., 2017), as has also been recently reported for the A. boonei casposase (Béguin et al., 2019). For example, the crucial identity of a few bp on each side of the top strand casposon integration site (Figure 3E, mutCGCA and mutL) corresponds to the conserved feature of CRISPR-Cas integration for specificity to the leader-repeat border. Specific integration of a transposon avoids any possible deleterious effects of random integration on the host cell genome, whereas for CRISPR-Cas, this also ensures that the latest encountered cellular invader is effectively neutralized as it generates the most robust immune response (McGinn and Marraffini, 2016). Furthermore, the requirement for a sequence motif upstream of the casposon integration site (Figures 1B and 3E mut12-17) appears to correspond to the observation that in some CRISPR-Cas systems, sequences upstream within the leader are important for integration (Alkhnbashi et al., 2016). This requirement is understood for the E. coli Type I-E system in which IHF binds within the last 40 bp of the leader and is needed to bend and deform the target site so that integration can occur (Nuñez et al., 2016; Wright et al., 2017). It is possible that the role of the casposase HTH domain is to recognize a specific sequence just upstream of the integration site and thereby direct integration; this would be consistent with our observation that it is crucial for activity (Figure 6B). We have expressed and purified the HTH domain alone, but have been unable to confirm an interaction with DNA (data not shown). As these shared properties are already features of casposon transposition, they are not a consequence of the evolution of one system into the other.

Based on the biochemistry and structure presented here, we suggest an evolutionary pathway in which the crucial change was architectural (Figure 9B, Figure 9—video 1). In this scenario, the ability of the casposase to tetramerize on DNA upon target binding was lost when it acquired a preference for binding to a Cas2 molecule rather than to itself. This seems a reasonable possibility as the total protein surface buried by the casposase by tetramerization (~2400 Å2) is substantially smaller than the total Cas1-Cas2 buried surface in the heterohexamer of either E. coli (~7800 Å2; Nuñez et al., 2014) or of E. faecalis (~4000 Å2; Xiao et al., 2017). As shown by the structure here, if an intervening Cas2 dimer were introduced as a bridge between the two noncatalytic subunits of the casposase tetramer, the active sites of the catalytic monomers would not only remain oriented towards each other but Cas2 would provide an additional DNA binding surface that could preferentially direct a single, short piece of DNA between two active sites.

In this model, the CRISPR repeat sequence would be therefore derived from the original target site preferred by the casposase, and any original specificity for the casposon terminal inverted repeats would devolve to a residual spacer sequence preference. It is also possible that one evolutionary path could conserve the ability of the M. mazei casposase to integrate RNA as well as DNA (Silas et al., 2016). We note that the RAG1 protein of the V(D)J recombinase, also derived from a DNA transposon, conserves properties of its original transposase in that specific signal sequences are recognized and cleaved out from its flanking coding sequences (Schatz and Swanson, 2011) while the random integration activity has been severely reduced to avoid genomic damage. Thus, in both cases, it is the logic of the particular system (i.e., biochemical properties of its transposase) that has been evolutionarily conserved.

In the evolutionary model presented here, when Cas2 was acquired as a preferred binding partner, it seems likely that Cas1 active sites would be forced farther apart in CRISPR integrases than in casposases, thereby expanding the length of the target site duplication. As has been pointed out (Wright et al., 2019), this could have been important to allow the sampling of longer pieces of DNA which would produce more specific crRNA for incorporation into effector complexes at a later step in the immunity process. However, it is not yet clear why - or even if - the two integration pathways are necessarily mutually exclusive, if appropriate substrates are supplied. Indeed, the presence in M. mazei strains of ‘solo-TIRs’ of specific length (either 19 or 31 bp; Krupovic et al., 2016) might be evidence of the capacity of a casposase to integrate short spacer-like sequences or an early step along the path to the development of a CRISPR array (Jackson et al., 2017). It is an intriguing possibility that some casposases may already possess the ability to integrate short CRISPR-like spacers, and nothing in the casposase structure appears to rule it out except perhaps the tight quarters of the compact tetramer and the lack of an extended dsDNA binding surface such as that provided by Cas2.

It has recently been reported that a type V-C Cas1 protein is active in vitro for spacer integration in the absence of Cas2, and has been proposed to be an early ancestor of the Cas1-Cas2 spacer acquisition system (Wright et al., 2019). Like the casposase here, the V-C Cas1 forms a tetrameric assembly upon binding to oligonucleotides that represent spacer integration products but differs in three key features. The first is its exquisite preference for spacers that are recessed by 0–2 nt on one strand: spacers that are recessed by three nt or longer are not functional for spacer integration whereas the M. mazei casposase is capable of integrating blunt-ended substrates, substrates with varying numbers of single-stranded nucleotides at its 3'-end, as well as completely single-stranded casposon ends of varying length. The second difference from casposases is the length of the target site as the 25 bp V-C target site, although shorter than average for CRISPR-Cas systems, is substantially larger than the ~14 bp average for casposases (Krupovic et al., 2014). Finally, the V-C Cas1 is less sequence-specific in its integration site preference relative to other characterized Cas1-Cas2 integrases (Wright et al., 2019), in marked contrast to the stringent sequence-specificity we observed here for the M. mazei casposase (Figure 3B). It will be extremely interesting to understand the structural basis of the transformation from casposase to a functional tetrameric spacer integrase.

The structure here provides insight into how a casposase recognizes single-stranded casposon ends and its target site sequence, but the target DNA used does not contain all of the elements required for site-specific integration. Determining the role of sequences just upstream of the top strand integration site may provide insight into the ancestral role of what became the critical CRISPR leader-repeat border. Another intriguing question relates to the role of the casposase C-terminal domain. It is tempting to speculate that it might play a role similar to that of IHF in the Type I-E system in bending target (Nuñez et al., 2016), perhaps through recognition of a specific sequence upstream of the integration site, but the answer awaits further investigation. Our results explain the capability of casposases to use ssDNA as an integration substrate but do not shed light on the nature of the in vivo substrate. As it is not known if an integratable casposon is generated through an excision mechanism or through the process of replication, where ssDNA fits into the puzzle remains to be established. The only other transposons known to site-specifically integrate ssDNA are those of the IS200/IS605 family (Barabas et al., 2008). However, their integration is fundamentally different as both the 3' and 5' ends of the elements are joined to a single-stranded target. It will be interesting to determine if the effectiveness of ssDNA casposon ends for integration might be linked more closely to the mechanism of casposon mobility, as originally proposed (Krupovic et al., 2014). Certainly, the fellow-traveling polB gene in casposons is suggestive that ssDNA and replication may be central to casposon mobility.

Materials and methods

Key resources table
Reagent type
(species) or resource		DesignationSource or referenceIdentifiersAdditional information
Gene (Methanosarsina mazei)casposaseWP_011035139.1
Strain (E. coli)Top10Invitrogen/ThermoFisherC404003
Strain (E. coli)B834(DE3)MilliporeSigma69041–3
Commercial kitSelenomethionine Medium CompleteMolecular DimensionsMD12-500
Recombinant DNA reagentpUC19InvitrogenP/N 54357
Chemical compound, drugPEG 8000Hampton ResearchHR2-535
Chemical compound, drugCHES, pH 9.0Hampton ResearchHR2-256-05
Commercial kitNextera XT DNA sample prepIllumina#FC1311024
Software, algorithmRsubread packagehttp://bioconductor.org/packages/RsubreadV1.34.7DOI: 10.18129/B9.bioc.Rsubread
Software, algorithmGviz packagehttps://bioconductor.org/packages/Gviz/V1.26.5
Software, algorithmggseqlogohttps://www.rdocumentation.org/packages/ ggseqlogo/versions/0.1V0.1
Software, algorithmBiostringshttps://bioconductor.org/packages/release/bioc/html/Biostrings.htmlV2.50.2, RRID:SCR_016949
Software, algorithmRsamtoolsV1.34.1
Software, algorithmShortReadV1.40.0
Software, algorithmGenomicAlignmentsV1.18.1
Software, algorithmJalviewwww.jalview.orgV1.0, RRID:SCR_006459
Software, algorithmXDSxds.mpimf-heidelberg.mpg.deRRID:SCR_015652
Software, algorithmXPREPhttps://www.bruker.comBruker
Software, algorithmSHELXDhttp://shelx.uni-ac.gwdg.de/Usón and Sheldrick, 1999; RRID:SCR_014220
Software, algorithmSHARP
https://www.globalphasing.com > sharp		Bricogne et al., 2003
Software, algorithmCCP4https://www.ccp4.ac.ukWinn et al., 2011; RRID:SCR_007255
Software, algorithmDMhttps://www.ccp4.ac.ukCowtan, 1994; RRID:SCR_007255
Software, algorithmOhttp://xray.bmc.uu.se/alwyn/TAJ/Home.htmlJones and Kjeldgaard, 1997
Software, algorithmCNS1.3http://cns-online.org/v1.3/Brunger, 2007
Software, algorithmBUSTER 2.10.3
https://www.globalphasing.com > buster		Bricogne et al., 2017
RRID:SCR_015653
Software, algorithmPhenix 1.10www.phenix-online.orgAdams et al., 2010
RRID:SCR_014224
Software, algorithmPyMOL v1.7https://pymol.orgSchrodinger, IncRRID:SCR_000305

Cloning, protein expression, and purification

Request a detailed protocol

The gene encoding the casposase from Methanosarsina mazei strain 3.F.A.1A.1 (WP_011035139.1) was cloned into a modified pBAD vector as previously described for the A. boonei casposase (Hickman and Dyda, 2015). For optimal expression and purification, the single point mutation C184S was introduced and 17 additional amino acids were added between the N-terminal TEV protease cleavage site and M1 to allow complete removal of the thioredoxin (Trx) tag. Purification was essentially as previously described (Hickman and Dyda, 2015). For selenomethionine-substituted protein, the coding sequence corresponding to the Trx-casposase fusion protein was subcloned into pET-29b, and transformed into E. coli strain B834(DE3). Expression was performed using the M9 minimal media kit and selenomethionine from Molecular Dimensions Inc (Maumee, OH) according to the manufacturer's instructions. Full selenomethionine substitution was confirmed by LC/MS (Biochemistry Core Facility, NHLBI). Purification and crystallization were identical to those of the unsubstituted protein. All purification steps were carried out at 4°C.

A C-terminally truncated casposase, residues 1-341 (ΔHTH), was generated by the deletion of residues 342-406 from the Trx-casposase fusion protein by PCR mutagenesis. Expression and purification was identical to that of the full-length version. Single and double point mutations were introduced by PCR, and sequences of the full coding frame confirmed. Purification was essentially as the wild-type with the exception of the Y206A/R280A double mutant, which yielded only ~10% of the level of soluble protein when compared to wild-type and was not completely cleaved by TEV, despite prolonged incubation. Due to the low level of recovered protein, it was assayed without performing the usual final size exclusion chromatography step.

Enzymatic assays

In vitro DNA integration into pUC19 and modified pUC19 target plasmids

Request a detailed protocol

Oligonucleotides (IDT) were resuspended in 10 mM Tris and, as needed, mixed in the appropriate molar ratio, heated to 95°C for 15 minutes, and allowed to slowly cool to room temperature. Modified pUC19 target plasmids were constructed using the QuikChange method by introducing target sequences between bp 575 and 576. For assays, purified protein (75 nM) and oligonucleotide (200 nM) were mixed in buffer consisting of 25 mM Tris pH 7.5, 150 mM KCl, 0.05 mg/ml BSA, and 5 mM MnCl2 and incubated at room temperature for the times indicated. For each set of reactions, a "no end" control was performed in which the oligonucleotide was omitted. Reactions were initiated by the addition of 150 ng target plasmid and stopped at time points indicated by adding EDTA to a final concentration of 25 mM followed by proteinase K digestion for 30 min at 37◦C (NEB; 40 U per reaction tube). Glycogen was then added, the reaction products ethanol-precipitated, and the tubes air-dried. Products were resuspended in 1X DNA loading buffer, run on a 1.5% agarose gel in 1X TAE at 100V, and visualized using ethidium bromide. Bands were quantified where indicated using ImageJ.

In vitro DNA integration into oligonucleotide targets

Request a detailed protocol

Protein (1.4 μM) and casposon end oligonucleotides (1.2 μM) were mixed in buffer consisting of 25 mM Tris pH 7.5, 150 mM KCl, 0.05 mg/ml BSA, and 5 mM MnCl2. Reactions at room temperature were initiated by the addition of the target oligonucleotide, and were stopped at the time points indicated by mixing 1:1 with formamide gel buffer. Reactions were run on 15% acrylamide TBE-urea gels (Novex) at 250V in 1X TBE and visualized using SYBR Gold (Invitrogen). Bands were quantified as indicated using the AlphaView band analysis software of an Alpha Innotech imager.

Library preparation and sequencing

Request a detailed protocol

To prepare four of the DNA samples for sequencing, the in vitro integration reaction was carried out in quadruplicate for 60 min using dsLE31 and pUC19+TSD+30bp as described above. The resulting samples were run on 1.5% agarose and, for each, the band corresponding to the linear product of integration was isolated using the QIAquick Gel Extraction Kit (Qiagen). The resulting DNA was then used as the template for four independent PCR reactions using a LE31 primer to amplify the product. The reaction was carried out for either 17 or 30 cycles. Amplified products corresponding to the size of linearized pUC19 were again isolated from agarose prior to library preparation. For the fifth sample, the integration reaction was scaled up four-fold and carried out in triplicate for 2 hours using TR29 as the integration substrate. Bands corresponding to the linear integration product were isolated from agarose, subjected to one round of extension using Pfu Ultra II Fusion DNA polymerase (Agilent) for 30 sec at 72°C, and purified again using the QIAquick PCR purification kit (Qiagen). The five sequencing libraries were prepared from the supplied DNA using Nextera XT DNA sample prep kit (Ilumina, #FC1311024) according to the manufacturer’s protocols. The libraries were then sequenced in a stranded 75 bases paired-end run on the MiSeq instrument and analyzed.

NGS computational analysis

Request a detailed protocol

The read 1 fastq files for each sample were processed, aligned and analyzed in R (V3.5.1) using custom scripts. Specifically, the reads were first collapsed by sequence, abundance of individual sequences was recorded, and results were stored in fasta format. Each fasta file was filtered, extracting only the unique sequences containing casposon end sequence (5`-GAGTTACCTATATCCC) for further analysis. These sequences were then trimmed, removing the casposon end fragment and retaining only downstream regions for aligned to the plasmid using Rsubread package (V1.34.7). Generated alignment files were used to visualize sequence alignments and coverages using Gviz package (V1.26.5). To generate sequence logos, the aligned sequences were first extracted from alignment files by strand, and each sequence was then duplicated based on the recorded abundance information obtained from the fastq files. The sequences were then aligned at the 5`end and alignment was used to calculate nucleotide frequency matrix extending from positions +1 up to +30 nucleotides downstream of integration of casposon end sequence. The logos were then visualized using ggseqlogo (V0.1) and negative strand logos flipped using Adobe Illustrator to reflect the complimentary base pairing. The alignment, conservation and upstream sequences of different target sites were performed and depicted using Jalview (V1.0) software. Additional packages used in the analysis include Biostrings (V2.50.2), Rsamtools (V1.34.1), ShortRead (V1.40.0), GenomicAlignments (V1.18.1). The scripts are available upon request. The fastq files have been deposited to GEO under the accession number: GSE139037.

Assay of DNA binding by size exclusion chromatography

Request a detailed protocol

After samples were mixed at various protein:DNA ratios and dialyzed overnight into 35 mM CHES pH 9.0, 0.5 M KCl, 105 (w/v) glycerol, 0.2 mM TCEP, protein multimerization and DNA binding was assessed. Samples were injected onto a Superdex 200 Increase 3.2/300 column equilibrated in dialysis buffer and eluted at 0.05 ml/min at 4°C. A standard curve was generated using solutions at 1 mg/ml in dialysis buffer of blue dextran, thyroglobulin (669 kDa), apoferritin (443 kDa), β-amylase (200 kDa), alcohol dehydrogenase (150 kDa), bovine serum albumin (66 kDa), and carbonic anhydrase (29 kDa).

Interferometric scattering mass spectrometry (iSCAMS)

Request a detailed protocol

Samples at ~1 mg/ml (~25 μM) were diluted to 25-50 nM in buffer containing 30 mM CHES pH 9.0, 2 mM EDTA, and either 1 M or 0.5 M KCl immediately prior to measurement using a G10 RefeynOne mass photometer. For each experiment, thousands of protein molecules were counted and their contrasts measured. Contrast values were converted to molar masses using a standard curve prepared using BSA (monomer+dimer), alcohol dehydrogenase (monomer+dimer), ovalbumin, and thyroglobulin in PBS buffer. Distributions were fitted using a Gaussian distribution model.

Complex formation for crystallography

Request a detailed protocol

Oligonucleotides were combined in several molar excess with purified casposase and dialyzed overnight into Complex Buffer 1 (30 mM CHES pH 9.0, 0.5 M KCl, 10% w/v glycerol, 0.2 mM TCEP, 0.2 mM EDTA). This was followed by size exclusion chromatography in the same buffer on a 16/60 Superdex 200 column. Fractions containing the complex were subsequently combined and concentrated at room temperature in ThermoScientific 30K CO units. Glycerol was removed by extensive dialysis against Complex Buffer 1 without glycerol. Complexes were stored at 10°C to prevent cold-induced precipitation.

Crystallization and structure determination

Request a detailed protocol

Crystallization was performed in hanging drop format at room temperature. Briefly, protein-DNA complexes were mixed with equal volume of reservoir solution containing 0.15-0.185 M calcium acetate, 84 mM sodium acetate pH 4.4-3.8, 7.5-9% PEG 8000, and 0.5 M KCl. Crystals grew to final size within 48h, and addition of 0.5 M NaCl to the reservoir encouraged further growth. Crystals were cryoprotected by transfer to mother liquor and quick serial transfer in increasing concentrations of ethylene glycol to 15-17.5% (v/v) final. Data was collected at the Southeast Regional Collective Access Team (SERCAT) beamline ID22 of the Advanced Photon Source.

Diffraction data collected at three energies around the Se K edge (Table 1) were integrated and scaled with XDS. Zero-dose (Diederichs et al., 2003) corrected data were used to calculate Fa coefficients (absolute value of the energy independent part of the Se structure factors) in XPREP (Bruker) and 27 Se positions were found with SHELXD (Usón and Sheldrick, 1999). The positional and occupancy parameters of the Se substructure were optimized with SHARP (Bricogne et al., 2003) and additional 5 Se sites were added. Solomon solvent-flattened electron density map was used to generate masks for non-crystallographic symmetry (NCS) averaging using CCP4 (Winn et al., 2011). The asymmetric unit contained a tetramer of the casposase composed of two dimers related by a proper two-fold axis. However, the monomers forming the dimers were not related by a proper rotational symmetry. Therefore, one NCS averaging mask was generated for a catalytic domain and another for one of the dimers. Multi-domain non-crystallographic symmetry averaging, further solvent flattening, histogram matching, and slow phase extension were carried using DM (Cowtan, 1994).

The resulting experimental electron density map was of excellent quality at 3.2 Å with clearly visible side chain densities at most places and clear density for the bound DNA. The molecular model was built using O (Jones and Kjeldgaard, 1997) and refined with several cycles of cartesian molecular dynamics, energy minimization and ADP refinement using CNS1.3 (Brunger, 2007) while the final refinement was carried out with BUSTER 2.10.3 (Bricogne et al., 2017). Composite simulated annealed omit map based on the final refined model for model verification was computed using Phenix 1.10 (Adams et al., 2010). The current model contains all residues from M1 to I341 in all four protein chains and to L354 for the B chain and V359 in the C chain; however, due to the lack of interpretable side chain densities, the last few residues in the B and C chains were modeled as serine. There was no visible connected electron density for the C-terminal domain; however, some disconnected but clearly helical densities were observed. These were modeled as polyserines.

Interfaces were analyzed using Protein interfaces, surfaces and assemblies' service PISA (Krissinel and Henrick, 2007) at the European bioinformatics Institute (http://www.ebi.ac.uk/pdbe/prot_int/pistart.html). All structural figures were generated using PyMOL (Schrodinger, Inc).

Data availability

Diffraction data for the casposase-DNA complex have been deposited under PBD ID 6OPM. The GSE number for the NGS data is GSE139037.

The following data sets were generated
    1. Hickman AB
    2. Kailasan S
    3. Genzor P
    4. Haase AD
    5. Dyda F
    (2019) NCBI Gene Expression Omnibus
    ID GSE139037. Casposase structure and the mechanistic link between DNA transposition and spacer acquisition by CRISPR-Cas.
    https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE139037

References

  10. Software
    1. Bricogne G
    2. Blanc E
    3. Brandl M
    4. Flensburg C
    5. Keller P
    6. Paciorek W
    7. Roversi P
    8. Sharff A
    9. Smart OS
    10. Vonrhein C
    11. Womack TO
    (2017)
    BUSTER, version 2.10.3
    Global Phasing Ltd, Cambridge.
    1. Cowtan KD
    (1994)
    'dm': An automated procedure for phase improvement by density modification
    Joint CCP4 and ESF-EACBM Newsletter on Protein Crystallography 31:34–38.

Decision letter

  1. Blake Wiedenheft
    Reviewing Editor; Montana State University, United States
  2. John Kuriyan
    Senior Editor; University of California, Berkeley, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

Cas1 and Cas2 are characteristic of the CRISPR-Cas adaptive immune systems found in most archaeal and many bacterial genomes. It is hypothesized that Cas1 evolved from a rare class of transposases, called casposases. Here, Hickman et al. express and purify a casposase from Methanosarsina mazei. They demonstrate that the MmCasposase integrates a wide variety of RNA and DNA substrates into a specific location. Integration requires substrates with a free 3'-OH and a target that contains specific sequence motifs. To understand the mechanism of integration the authors crystallize the casposase bound to DNA. The structure reveals a tetrameric architecture in which two casposase dimers sandwich the target DNA. The authors compare this structure to available crystal structures of Cas1 and Cas1-2 integration complexes. This analysis shows that the casposase tetramer is mutually exclusive with Cas2 binding and suggests how architectural changes in the Cas1-Cas2 complex may have led to mechanistic differences between the two types of integrases.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Casposase structure and the mechanistic link between DNA transposition and spacer acquisition by CRISPR-Cas" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife, in its present form.

In this manuscript, Hickman et al. present the first crystal structure of a transposon-encoded casposase bound to an integration product mimic. The structure reveals a tetrameric architecture in which two casposase dimers sandwich the target DNA. The authors compare this structure to available crystal structures of Cas1 and Cas1-2 complexes. This analysis shows that the casposase tetramer is mutually exclusive with Cas2 binding and suggests how architectural changes in the Cas1-Cas2 complex may have led to mechanistic differences between the two types of integrases.

The reviewers all agree that the structure presented in this manuscript is potentially important. However, the structural data are not supported by biochemical experiments that advance our understanding of mechanism.

Some of the specific points raised during the review and in the discussion between reviewers and the editors are noted below, and the individual reviews are appended. We concluded that it could take substantial time to address all of these issues in a satisfactory manner. At this point, the authors may simply wish to submit this high quality paper elsewhere, without delay. If, however, these points can be addressed satisfactorily then the paper could be resubmitted to eLife for further consideration, should the authors wish to do so.

1) Clarification on the DNA substrates used in the biochemical assay, and an explanation for how these substrates relate to the biology are required. Additional controls and additional time points are necessary. Important controls are missing from the results presented in Figure 1, most notably negative controls in which the plasmid is incubated with casposase in the presence of divalent metal ions but no transposon end substrates. Previous Cas1 studies have observed considerable nicking activity (or possibly nicking/cleavage from contaminating proteins), so it is crucial that the baseline activity of the casposase protein prep on plasmid DNA is evaluated. Furthermore, for the data presented in panels C and D, no lanes exist showing the starting plasmid substrate at time 0 (i.e. in the absence of protein and transposon end substrate); since any given plasmid prep will have variable ratios of supercoiled to open-circular plasmid, it is critical that the reader have this comparison – for the exact substrate used in the experiments – to fully evaluate the biochemical activity of casposase and discriminate this from the starting status of the substrate. Labeling the bands "target," "SE," and "DE," is misleading, since a plasmid will naturally purify from a bacterial culture with some fraction in the open circular form, which will migrate the same as a SE product but is certainly not a single-end integration product.

2) Provide more experimental support and/or justification for the specific choice of the 8-nt overhang integration product. Why was 8 selected as the overhang length? Was a larger crystallization screen performed, and this was the only substrate geometry that resulted in diffracting crystals? Is there any further evidence that this is a physiologically relevant ssDNA end sequence? For at least one reviewer, it was unclear what the SEC data provided that was not show by the iSCAMS analysis. All reviewers commented on misplaced labels and inappropriate units (time rather than volume) in the SEC. There was some discussion about moving the SEC to the supplement but regardless of where it is discussed, the interpretation of these data must be clarified.

3) The authors gloss over the interesting finding that the HTH domain is disordered in the crystals and fail to follow-up with further experiments to investigate the role of this domain on biochemical integration activity, and whether the domain may be important in the context of more physiologically relevant substrates.

4) A major take-away from this manuscript would be that this casposase integrates substrates primarily at a specific target site rather than randomly, as has been previously touched upon in the literature. However, apparent integration events in pUC19 alone and farther away from the target site have been detected in this and previous studies. No sequencing data for these events are shown, and the authors fail to perform additional analysis on integration patterns. The sequences should be shown and discuss. It has become routine in spacer acquisition studies with Cas1 to deep sequence half-site integration products in order to gain deeper insights into specificity. Similar NGS experiments are warranted here to enhance their biochemical results. While NGS experiments are not required for publication in eLife, the authors should perform bioinforamtics analyses to determine whether related casposons are always present in genomes adjacent to similar target sites. Please mention, in the Discussion, the recent observation that the Type V-C Cas1 is much less sequence-specific in its integration site preference.

5) When describing the structure, the authors note several protein-DNA contacts, but it is unclear whether any of these contacts contribute to the specificity for the TSD target or LE substrate. From the data shown in Figure 1C, it appears that the casposase is somewhat specific for the LE sequence, especially for shorter single-stranded DNAs. Can this specificity be explained by the structure? Do R191 and R192 contribute to the specificity of the casposase for the TSD? The authors speculate about the importance of stacking interactions for ssDNA substrate binding (subsection “ssDNA casposon end binding resembles that of spacer 3' overhangs”). Did the authors perform any mutagenesis experiments to establish the importance of these contacts for target and substrate specificity? Experiments supporting structure-based hypotheses proposed by the authors are required.

6) The order in which substrates are shown on the left gel in Figure 1C is confusing. If possible, it would be helpful if this gel could be re-run to place casposon-specific and random substrates of similar type (e.g. TR17 vs ranTR17) in adjacent lanes on the gel. Is a dsran17 substrate (i.e. a fully double-stranded random substrate of 17 bp) also integrated? In addition, ssran12 and ssLE11 could be omitted from this gel, as these substrates are also tested in the rightmost gel in the panel.

7) In paragraph two of the Discussion section, the authors suggest a similarity in the lack of sequence specificity on the bottom strand of the casposase TSD with a lack of sequence specificity for spacer-side integration by Cas1-Cas2. However, it is important to note that Cas1-Cas2 does display specificity for the entire direct repeat sequence, likely through sequence-specific deformations of the repeat that occur during full-site integration (as proposed by Wright et al., 2017). In that study, it was shown that mutations in the repeat affect spacer-side integration, but do not have an effect on leader-side integration. Thus, the conclusion stated by the authors that bottom strand integration is sequence-independent and instead based on a "molecular ruler" is not consistent with current models for spacer-side integration.

Reviewer #1:

Cas1 and Cas2 are hallmarks of the CRISPR-Cas adaptive immune systems found in most archaeal and many bacterial genomes. Koonin and colleagues have previously identified homology between Cas1 and casposon family proteins (doi: 10.1186/1741-7007-12-36) and predicted that CRISPR repeats evolved from the target site duplications (doi: 10.1016/j.mib.2017.04.004). These predictions have been largely validated with in vitro biochemical studies (doi.org/10.1093/nar/gkv1180, doi.org/10.1093/nar/gkw821, doi.org/10.1093/nar/gkz447). Here, Hickman et al., extend the previous work by defining essential features of the target site and testing a wide-variety of integration substrates. This on its own would not warrant publication in eLife. However, they use this qualitative biochem to identify substrates that result in higher order assemblies and they determine the structure of the tetrameric complex bound to DNA. Overall the experiments appear technically sound, and the paper delivers the first structure of a casposase. These mechanistic insights and evolutionary connection of this system to CRISPR-Cas immune systems are important and warrant publication in eLife. However, the text and figure are a challenge to interpret (esp Figure 1) and considerable revision is required to make this work accessible to a broad readership.

The authors purify a casposases from Methanosarcina mazei and determine sequence requirements for integration. The authors describe a sequencing method to determine the location of integration. 7 of 8 integration occurs at the TSD sequence with concomitant generation of a 14 bp TSD. "The other sequenced product arose from integration into a site ~110 bp away from the specific sequence and was also accompanied by a 14 bp TSD." Please include more information. Was there sequence similarity between these sites? If not, what does this say about the mechanism of recognition and how does this relate what the structures teaches us about recognition?

Figure 1A. It is not clear what the authors are intending to teach the reader in panel A. Based on the figure legend title ("Casposase structure and function") the reader is expecting to see a structure, rather than a schematic of the genomic context. Most of the arrows in the schematic are different colors, which have no explanations, the sequence inserted below the schematic is too small to decipher, and the blue box inserted to the right is not labeled or sufficiently explained. To easily understand this figure, readers will need to be familiar with the papers by Koonin et al. I think this is an unfair expectation of readers.

Seems like it would be worth showing a gel or at least referencing the gel in Figure 5—figure supplement 4, when the authors first claim to have "expressed and purified a representative casposase".

Figure 1B. I was unable to understand how the TIR (introduced in font size ~2 in panel B) is different form the TSD, which is introduced in panel A. I understand that it's a tandem inverted repeat but where is the sequence, how long, how is it related to the biology (I'm lost)..? Why is the first gel white on black and the second is black on white? Are these different imaging methods of the same assay? If the 5'-CGCA motif is conserved, then why do the authors mutate the first 6 rather than the first 4 bps of the TSD sequence?

Figure 1C. What are the red dots one end of the orange oligonucleotides?

Figure 2A. In my opinion, seeing a gel of the purified protein is more important to me than seeing a series of SEC runs with increasing concentration of the DNA substrates. If the authors choose to show these in the main text, then a vertical line extending from the protein alone peak down though the other runs will help illustrate the shift. The authors disclose that "the complex did not produce the expected shift on SEC for a tetramer", but the peak does not look symmetrical to me and the left edge appears to be almost vertical. This would suggest that the complex may be eluting in the void, but the x-axis is in time rather than volume, so it is hard to know what this mean for a16/60 Superdex 200 column. Try a Sup 6 column and include a regression line fit with standers. It appears that the labels for the x-axis are misplaced for the last column ("integration products"). In my opinion this entire figure could go in the supplemental material and Figure 1 should be refined and split into two figures.

The text for Table 1 overlaps the table itself making it a little hard to interpret. I suggest adding a sentence to the text that states how the structure was determined (e.g., MR or experimental phases).

Figure 3. Active site subunits "shown in pink sticks" are nearly impossible to see. I could not find an explanation for the green spheres in panel B.

Figure 4. The authors state that "Each casposase monomer exhibits the fold expected of a Cas1 homolog" but none of the figures do a good job of showing this and I could not find an RMSD for a define set of equivalently positioned atoms.

Discussion paragraph five: delete word "is". “It is an intriguing possibility is that some casposases may already possess the ability to integrate short CRISPR-like spacers,…..”

Reviewer #2:

In their manuscript, Hickman and colleagues present the first crystal structure of a transposon-encoded casposase bound to an integration product mimic, revealing a tetrameric architecture in which two Cas1 dimers sandwich the target DNA destined to be duplicated during downstream molecular steps within the pathway. Biochemical experiments are presented to argue that single-stranded transposon ends may be the natural substrate utilized during integration, and that specific sequence motifs in both the transposon end and the target site affect biochemical integration activity. By comparing their casposase-DNA structure to available crystal structures of Cas1 homologs in the context of the Cas1-Cas2 complexes that mediate spacer acquisition during adaptive immunity in CRISPR-Cas systems, the authors argue a plausible evolutionary scenario by which gaining Cas2 interactions could have directed casposases towards an integration mechanism that is more akin to spacer acquisition than transposition.

The crystal structure is a notable advance that broadens our understanding of the Cas1/casposase protein family, and is undoubtedly a substantive contribution to the field. However, in this reviewer's opinion, the biochemical experiments that accompany this finding fall short of providing compelling evidence for a specific integration mechanism, and instead leave more to be desired in terms of what overhang geometries are optimal, how integration sites are chosen, and how casposons are excised. In addition, the authors gloss over the interesting finding that the HTH domain is disordered in the crystals and fail to follow-up with further experiments to investigate the role of this domain on biochemical integration activity, and whether the domain may be important in the context of more physiologically relevant substrates. I believe additional experiments are merited for this study to go beyond reporting a novel structure, to helping yield new mechanistic insights for the process of casposon transposition.

1) Some important controls are missing from the results presented in Figure 1, most notably negative controls in which the plasmid is incubated with casposase in the presence of divalent metal ions but no transposon end substrates. Previous Cas1 studies have observed considerable nicking activity (or possibly nicking/cleavage from contaminating proteins), so it is crucial that the baseline activity of the casposase protein prep on plasmid DNA is evaluated. Furthermore, for the data presented in panels C and D, no lanes exist showing the starting plasmid substrate at time 0 (i.e. in the absence of protein and transposon end substrate); since any given plasmid prep will have variable ratios of supercoiled to open-circular plasmid, it is critical that the reader have this comparison – for the exact substrate used in the experiments – to fully evaluate the biochemical activity of casposase and discriminate this from the starting status of the substrate. I would even argue that naming the bands "target," "SE," and "DE," is misleading, since a plasmid will naturally purify from a bacterial culture with some fraction in the open circular form, which will migrate the same as a SE product but is certainly not a single-end integration product.

2) A major question is left unaddressed (and mostly unmentioned) in the manuscript, namely, in what form the casposon is excised and integrated. I understand this is still unclear in the literature, that previous attempts to monitor excision by this group were unsuccessful (i.e. Hickman and Dyda, 2015), and that there is as yet no in vivo transposition data for casposons. Nevertheless, the reader should certainly be given more context on this topic, in the Introduction/Discussion, and in the Results section to better motivate the range of donor substrates tested. I would also like to see the authors test a larger panel of "Top Strand Recessed" substrates, and to more quantitatively compare integration activities amongst these substrates, and between these substrates and blunt dsDNA substrates. Similar experiments in the context of spacer acquisition by Cas1 complexes (e.g. Wright et al., 2019) have revealed the likely substrates that are used biologically, based on a clear optimum in terms of integration product formation kinetics. The authors have an opportunity to make similar discoveries with their casposase, but the use of just two timepoints, the lack of any product quantification, and the relatively small set of substrates tested, limits the potential of their biochemical assays.

3) Related to #2, I would prefer more experimental support and/or justification for the specific choice of the 8-nt overhang integration product. Why was 8 selected as the overhang length? Was a larger crystallization screen performed, and this was the only substrate geometry that resulted in diffracting crystals? Is there any further evidence that this is a physiologically relevant ssDNA end sequence? What happens if the casposon end substrates are shorter, or partially double-stranded? For example, it seems that the iSCAMS analysis could be scaled up to test a wider range of model integration products, to investigate whether tetramer formation is more or less favored with certain substrates, which could help suggest what may or may not be more 'relevant.'

4) The authors point out that capsposases have an additional C-terminal HTH domain that is absent in Cas1, and that this domain could not be modeled into the experimental electron density. In light of these interesting observations, the authors should test the biochemical activity casposase mutants in which this domain is truncated; does these mutants show behave similarly? Differently? Although it would be interesting to further investigate whether this domain becomes ordered with different DNA substrates, I understand that pursuing additional structures may be too much to ask. However, certainly some biochemical investigations are warranted here. I would also caution the authors against stating that the HTH domain "is not in a folded state" – although I am not a structural biologist, the absence of ordered density in a X-ray crystal structure is not equivalent to a domain being unfolded.

5) A major take-away from this manuscript would be that this casposase integrates substrates primarily at a specific target site rather than randomly, as has been previously touched upon in the literature. However, apparent integration events in pUC19 alone and farther away from the target site have been detected in this and previous studies. No sequencing data for these events are shown, and the authors failed to perform additional analysis on integration patterns. It has not become routine in spacer acquisition studies with Cas1 to deep sequence half-site integration products in order to gain deeper insights into specificity, and the authors should perform similar NGS experiments here to enhance their biochemical results. These experiments could provide much more data on the target site requirements, but looking for conservation (or lack of conservation) in the motifs that are preferred adjacent to integration sites other than at the 3' end of tRNA-Leu.

Reviewer #3:

In this study, Hickman et al. describe the activity and structure of a casposase, the hypothesized evolutionary ancestor of the Cas1 integrase found in CRISPR-Cas adaptive immune systems. The authors demonstrate in vitro DNA integration by the casposase from Methanosarcina mazei, and perform biochemical experiments to establish the target sequence specificity and substrate preferences for this casposase. They also report, for the first time, a structure of the casposase bound to an integration product mimic. The casposase forms a tetramer composed of two casposase dimers sandwiching the DNA. The structure provides insight into the mechanism of substrate and target recognition by the casposase. In addition, comparison of the casposase and Cas1-Cas2 complex structures show that the casposase tetramer is mutually exclusive with Cas2 binding, and suggests how architectural changes in the Cas1-Cas2 complex may have led to mechanistic differences between the two types of integrases. The structure of the casposase is an important first step in understanding casposon function and the evolution of CRISPR-Cas adaptation mechanisms, and the authors propose several interesting hypotheses based on the structure. However, these hypotheses are largely speculative and could greatly benefit from experimental evidence. In particular, the structural data is not supported by any biochemical experiments to establish a structure-function relationship. Thus, the structure is largely descriptive, although the authors speculate on the importance of a number of observed interactions within the structure. Several examples are provided in comment 1 below, in addition to few other concerns that should also be addressed by the authors.

1) When describing the structure, the authors note several protein DNA contacts, but it is unclear whether any of these contacts contribute to the specificity for the TSD target or LE substrate. From the data shown in Figure 1C, it appears that the casposase is somewhat specific for the LE sequence, especially for shorter single-stranded DNAs. Can this specificity be explained by the structure? Do R191 and R192 contribute to the specificity of the casposase for the TSD? The authors speculate about the importance of stacking interactions for ssDNA substrate binding (subsection “ssDNA casposon end binding resembles that of spacer 3' overhangs”). Did the authors perform any mutagenesis experiments to establish the importance of these contacts for target and substrate specificity? In general, the study would be greatly improved by experiments supporting the structure-based hypotheses proposed by the authors when describing the structure.

2) The order in which substrates are shown on the left gel in Figure 1C is confusing. If possible, it would be helpful if this gel could be re-run to place casposon-specific and random substrates of similar type (e.g. TR17 vs ranTR17) in adjacent lanes on the gel. Is a dsran17 substrate (i.e. a fully double-stranded random substrate of 17 bp) also integrated? In addition, ssran12 and ssLE11 could be omitted from this gel, as these substrates are also tested in the rightmost gel in the panel.

3) In paragraph two of the Discussion section, the authors suggest a similarity in the lack of sequence specificity on the bottom strand of the casposase TSD with a lack of sequence specificity for spacer-side integration by Cas1-Cas2. However, it is important to note that Cas1-Cas2 does display specificity for the entire direct repeat sequence, likely through sequence-specific deformations of the repeat that occur during full-site integration (as proposed by Wright et al., 2017). In that study, it was shown that mutations in the repeat affect spacer-side integration, but do not have an effect on leader-side integration. Thus, the conclusion stated by the authors that bottom strand integration is sequence-independent and instead based on a "molecular ruler" is not consistent with current models for spacer-side integration.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Casposase structure and the mechanistic link between DNA transposition and spacer acquisition by CRISPR-Cas" for further consideration by eLife. Your revised article has been evaluated by John Kuriyan (Senior Editor) and a Reviewing Editor.

The reviewers agree that the manuscript has been dramatically improved and there is unanimous consensus that the paper should be published in eLife. The reviewers offer a few additional suggestions, as outlined below:

Summary:

In this manuscript, Hickman et al. express and purify a casposase from Methanosarsina mazei. They demonstrate that the MmCasposase integrates a wide variety of RNA and DNA substrates into a specific location. Integration requires substrates with a free 3'-OH and a target that contains specific sequence motifs. To understand the mechanism of integration the authors attempt to crystallize the casposase bound to DNA oligonucleotides representing different stages of the integration reaction. Diffracting crystals of the casposase bound to a substrate that mimics the post integration step were used to determine a 3.1 Å resolution structure. The structure reveals a tetrameric architecture in which two casposase dimers sandwich the target DNA. The authors highlight several interactions with target and the substrate, and mutants are used to test the importance of these interactions. The authors compare this structure to available crystal structures of Cas1 and Cas1-2 complexes. This analysis shows that the casposase tetramer is mutually exclusive with Cas2 binding and suggests how architectural changes in the Cas1-Cas2 complex may have led to mechanistic differences between the two types of integrases. Collectively, the biochemistry and structure presented support a speculative discussion that connects architectural changes in the casposase that would accommodate Cas2 binding and the transition to a Cas1-Cas2 complex that is necessary for spacer integration. While clarity of presentation could be improved in places, the work is rigorously performed, and the conclusions are of interest to a broad readership.

Organization of the presentation:

The initial characterization of casposon substrate and target specificities is much improved by the new experiments. However, the presentation is a bit disjointed, as it jumps back and forth between characterization of substrate and target requirements. For example, Figures 1C, 3A-B and 3D all investigate casposon end requirements, while Figures 2 and 3C are on target specificity. Indeed, in one case this requires that the reader read ahead to understand the experiment explained in Figure 2 ("for the rationale, see Figure 3A and associated discussion"). This section could flow better if the authors combine Figure 1C, 3A-B, and 3D into a new Figure, which could be Figure 2. This would also allow the authors to show a panel describing all substrates used in these three panels, as it is currently confusing that TR17 and ssLE17 are shown in Figure 1C but not mentioned until much later in the text. Figures 2 and 3C could then be combined to form a new Figure 3. With these changes, the authors could first define casposon end requirements, and then describe the characterization of the target sequence.

https://doi.org/10.7554/eLife.50004.sa1

Author response

[Editors’ note: what follows is the authors’ response to the first round of review]

In this manuscript, Hickman et al. present the first crystal structure of a transposon-encoded casposase bound to an integration product mimic. The structure reveals a tetrameric architecture in which two casposase dimers sandwich the target DNA. The authors compare this structure to available crystal structures of Cas1 and Cas1-2 complexes. This analysis shows that the casposase tetramer is mutually exclusive with Cas2 binding and suggests how architectural changes in the Cas1-Cas2 complex may have led to mechanistic differences between the two types of integrases.

The reviewers all agree that the structure presented in this manuscript is potentially important. However, the structural data are not supported by biochemical experiments that advance our understanding of mechanism.

Some of the specific points raised during the review and in the discussion between reviewers and the editors are noted below, and the individual reviews are appended. We concluded that it could take substantial time to address all of these issues in a satisfactory manner. At this point, the authors may simply wish to submit this high quality paper elsewhere, without delay. If, however, these points can be addressed satisfactorily then the paper could be resubmitted to eLife for further consideration, should the authors wish to do so.

1) Clarification on the DNA substrates used in the biochemical assay, and an explanation for how these substrates relate to the biology are required.

We have repeated all of the biochemical assays and incorporated more controls and time points (see below). In doing so, we took the opportunity to perform the biochemical assay first discussed in the text using an authentic M. mazei casposon 31mer TIR (Figure 1B). In the text, we now reference the work on CRISPR Cas1-Cas2 proteins to explain our exploration of substrates with 3'-OH overhangs. The use of single-stranded substrates was the outcome of performing control reactions (or what we thought would be control reactions), and their activity was unexpected. Since then, Krupovic et al. (2019) have reported that a single-stranded casposon end is an effective substrate for the A. boonei casposase, and we have cited their work. Unfortunately, as there are many unknowns regarding the experimental biology for casposons, we can only speculate as to how these particular substrates relate to their biology.

Additional controls and additional time points are necessary. Important controls are missing from the results presented in Figure 1, most notably negative controls in which the plasmid is incubated with casposase in the presence of divalent metal ions but no transposon end substrates. Previous Cas1 studies have observed considerable nicking activity (or possibly nicking/cleavage from contaminating proteins), so it is crucial that the baseline activity of the casposase protein prep on plasmid DNA is evaluated. Furthermore, for the data presented in panels C and D, no lanes exist showing the starting plasmid substrate at time 0 (i.e. in the absence of protein and transposon end substrate); since any given plasmid prep will have variable ratios of supercoiled to open-circular plasmid, it is critical that the reader have this comparison – for the exact substrate used in the experiments – to fully evaluate the biochemical activity of casposase and discriminate this from the starting status of the substrate. Labeling the bands "target," "SE," and "DE," is misleading, since a plasmid will naturally purify from a bacterial culture with some fraction in the open circular form, which will migrate the same as a SE product but is certainly not a single-end integration product.

We have repeated the experiments presented in the original manuscript (including those described in Figure 1) and included additional controls and time points. In our revised manuscript, for all experiments involving plasmid substrates, a negative control is shown in which plasmid is incubated with the casposase and metal ion in the absence of a transposon end substrate. We have re-labeled the product band positions as suggested by the reviewers, and these are now indicated as "relaxed", and "linear". The target plasmid is now labelled "sc". We have retained the SE and DE terminology in the schematic in Figure 1B, as these are the products expected by the model. As a reader can now judge, there is not much nonspecific nicking with our purified casposase under the assay conditions used.

2) Provide more experimental support and/or justification for the specific choice of the 8-nt overhang integration product. Why was 8 selected as the overhang length? Was a larger crystallization screen performed, and this was the only substrate geometry that resulted in diffracting crystals? Is there any further evidence that this is a physiologically relevant ssDNA end sequence?

We apologize, as in our original submission we failed to mention that we tried a wide range of DNA oligonucleotides for structural studies over the course of many months, and that – to date – the only substrate that has yielded crystals that diffract X-rays well enough to allow us to solve the structure is the 8-nt overhang integration product. We believe that such exploration (riddled with many failures) is typical in experimental structural biology. As shown in Figure 3B, an 8-nt ssDNA casposon sequence produces products in an integration assay that are specific for the target site (i.e., they are not seen with pUC19 alone). Thus, it has the specificity expected for a biologically relevant end sequence but we cannot say more than this in the absence of an in vivo assay for casposon mobility.

For at least one reviewer, it was unclear what the SEC data provided that was not show by the iSCAMS analysis. All reviewers commented on misplaced labels and inappropriate units (time rather than volume) in the SEC. There was some discussion about moving the SEC to the supplement but regardless of where it is discussed, the interpretation of these data must be clarified.

We agree with the reviewer that the SEC data are perhaps not as important in light of the iSCAMS analysis; therefore we have chosen to present it (and some new data) now as supplemental material (Figure 4—figure supplement 1 and 2). The supplemental figures now show the chromatograms with corrected labels and volume units. We have been able to clarify the apparent disconnect between the iSCAMS and SEC results with further SEC experiments in which we performed a dilution series with the protein alone and complexes with TR17 and the crystallized integration product. These revealed that the elution properties of the casposase alone and its complexes change as a function of protein concentration. We conclude that the lower concentration used for iSCAMS more accurately reflects the property of the casposase in vivo. These results are now discussed in the revised text.

3) The authors gloss over the interesting finding that the HTH domain is disordered in the crystals and fail to follow-up with further experiments to investigate the role of this domain on biochemical integration activity, and whether the domain may be important in the context of more physiologically relevant substrates.

We used the structure as a guide to produce a soluble version of the casposase missing the C-terminal domain (residues 1-341, "ΔHTH") – solubility being a property that had frustratingly eluded us in the absence of structural information. We have therefore been able to follow up with integration experiments using a range of substrates, and these results are now presented in Figure 6 and associated text. Indeed, the C-terminal domain is crucial, although we do not know if it is recognizing the oligonucleotide end or the target. To date, we have not been able to confirm an interaction with either.

4) A major take-away from this manuscript would be that this casposase integrates substrates primarily at a specific target site rather than randomly, as has been previously touched upon in the literature. However, apparent integration events in pUC19 alone and farther away from the target site have been detected in this and previous studies. No sequencing data for these events are shown, and the authors fail to perform additional analysis on integration patterns. The sequences should be shown and discuss. It has become routine in spacer acquisition studies with Cas1 to deep sequence half-site integration products in order to gain deeper insights into specificity. Similar NGS experiments are warranted here to enhance their biochemical results. While NGS experiments are not required for publication in eLife, the authors should perform bioinforamtics analyses to determine whether related casposons are always present in genomes adjacent to similar target sites. Please mention, in the Discussion, the recent observation that the Type V-C Cas1 is much less sequence-specific in its integration site preference.

We agree with the reviewer's suggestion that the need for an NGS experiment was justified. We recruited the expertise of our colleagues, Dr. Pavel and Dr. Haase, who helped us to perform the NGS analysis on the linear product of targeted integration. They are now included as co-authors, and the new data are presented in Figure 2 and discussed in the text. As suggested, we have added a sentence regarding Type V-C Cas1 to the Discussion as follows: “Finally, the V-C Cas1 is less sequence-specific in its integration site preference relative to other characterized Cas1-Cas2 integrases (Wright et al., 2019), in marked contrast to the stringent sequence-specificity we observed here for the M. mazei casposase (Figure 2).”

5) When describing the structure, the authors note several protein-DNA contacts, but it is unclear whether any of these contacts contribute to the specificity for the TSD target or LE substrate. From the data shown in Figure 1C, it appears that the casposase is somewhat specific for the LE sequence, especially for shorter single-stranded DNAs. Can this specificity be explained by the structure? Do R191 and R192 contribute to the specificity of the casposase for the TSD? The authors speculate about the importance of stacking interactions for ssDNA substrate binding (subsection “ssDNA casposon end binding resembles that of spacer 3' overhangs”). Did the authors perform any mutagenesis experiments to establish the importance of these contacts for target and substrate specificity? Experiments supporting structure-based hypotheses proposed by the authors are required.

As suggested, in order to explain the specificity for short ssDNAs, we generated several mutants to test the roles of the residues observed to form interactions with the ssDNA. These include three mutant versions of the casposase that the structure predicted should disrupt the stacking interactions, as well as the double-mutant R191A/R192A. We have compared their activity on the TSD target vs. pUC19 to evaluate target specificity, and have tested their integration activity using a range of substrates to probe the role of these residues in substrate specificity (Figure 7F and Figure 7—figure supplement 1). These data are now discussed in the text.

6) The order in which substrates are shown on the left gel in Figure 1C is confusing. If possible, it would be helpful if this gel could be re-run to place casposon-specific and random substrates of similar type (e.g. TR17 vs ranTR17) in adjacent lanes on the gel. Is a dsran17 substrate (i.e. a fully double-stranded random substrate of 17 bp) also integrated? In addition, ssran12 and ssLE11 could be omitted from this gel, as these substrates are also tested in the rightmost gel in the panel.

We apologize for the haphazard organization of the gels presented in our original submission. As suggested, we have repeated the experiment to include a dsran17 substrate and the order of the lanes has been changed. These data are now in Figure 3A.

7) In paragraph two of the Discussion section, the authors suggest a similarity in the lack of sequence specificity on the bottom strand of the casposase TSD with a lack of sequence specificity for spacer-side integration by Cas1-Cas2. However, it is important to note that Cas1-Cas2 does display specificity for the entire direct repeat sequence, likely through sequence-specific deformations of the repeat that occur during full-site integration (as proposed by Wright et al., 2017). In that study, it was shown that mutations in the repeat affect spacer-side integration, but do not have an effect on leader-side integration. Thus, the conclusion stated by the authors that bottom strand integration is sequence-independent and instead based on a "molecular ruler" is not consistent with current models for spacer-side integration.

Upon repeating our initial experiments with time courses, it has become clear that our original statement has to be revised. As shown in Figure 3C, when the activity of the "mutR" substrate is compared to that of "symTSD", two substrates that differ in sequence at the site of "bottom strand" integration, there is clearly a difference. We have removed the entire section from the revised text.

Reviewer #1:

Cas1 and Cas2 are hallmarks of the CRISPR-Cas adaptive immune systems found in most archaeal and many bacterial genomes. Koonin and colleagues have previously identified homology between Cas1 and casposon family proteins (doi: 10.1186/1741-7007-12-36) and predicted that CRISPR repeats evolved from the target site duplications (doi: 10.1016/j.mib.2017.04.004). These predictions have been largely validated with in vitro biochemical studies (doi.org/10.1093/nar/gkv1180, doi.org/10.1093/nar/gkw821, doi.org/10.1093/nar/gkz447). Here, Hickman et al., extend the previous work by defining essential features of the target site and testing a wide-variety of integration substrates. This on its own would not warrant publication in eLife. However, they use this qualitative biochem to identify substrates that result in higher order assemblies and they determine the structure of the tetrameric complex bound to DNA. Overall the experiments appear technically sound, and the paper delivers the first structure of a casposase. These mechanistic insights and evolutionary connection of this system to CRISPR-Cas immune systems are important and warrant publication in eLife. However, the text and figure are a challenge to interpret (esp Figure 1) and considerable revision is required to make this work accessible to a broad readership.

The authors purify a casposases from Methanosarcina mazei and determine sequence requirements for integration. The authors describe a sequencing method to determine the location of integration. 7 of 8 integration occurs at the TSD sequence with concomitant generation of a 14 bp TSD. "The other sequenced product arose from integration into a site ~110 bp away from the specific sequence and was also accompanied by a 14 bp TSD." Please include more information. Was there sequence similarity between these sites? If not, what does this say about the mechanism of recognition and how does this relate what the structures teaches us about recognition?

We have now performed next-generation sequencing on the linear product of the plasmid integration reaction, and included these results in the revised manuscript. As shown in Figure 2C, there is indeed sequence similarity between the favored integration site and the very few other detected sites of integration.

Figure 1A. It is not clear what the authors are intending to teach the reader in panel A. Based on the figure legend title ("Casposase structure and function") the reader is expecting to see a structure, rather than a schematic of the genomic context. Most of the arrows in the schematic are different colors, which have no explanations, the sequence inserted below the schematic is too small to decipher, and the blue box inserted to the right is not labeled or sufficiently explained. To easily understand this figure, readers will need to be familiar with the papers by Koonin et al. I think this is an unfair expectation of readers.

In light of these comments, we have substantially revised Figure 1A, and have included more labeling, increased the font size for the TSD sequence, and have included the sequence of the 31mer TIR. We have also added titles to the two figure sections. The figure legend title has also been changed to: "Methanosarcina mazei casposon organization and initial biochemical characterization."

Seems like it would be worth showing a gel or at least referencing the gel in Figure 5—figure supplement 4, when the authors first claim to have "expressed and purified a representative casposase".

In the revised manuscript, in addition to the gel in the supplemental figure showing dissolved crystals and the protein control (Figure 5—figure supplement 4), we now show a gel for all of the purified proteins used in this study in Figure 6A.

Figure 1B. I was unable to understand how the TIR (introduced in font size ~2 in panel B) is different form the TSD, which is introduced in panel A. I understand that it's a tandem inverted repeat but where is the sequence, how long, how is it related to the biology (I'm lost)..?

We have now included a better schematic including the entire relevant genomic sequence in Figure 1A.

Why is the first gel white on black and the second is black on white?

We thank the reviewer for pointing this out and, in fact, there was no good reason. This has been corrected in the revised figures.

Are these different imaging methods of the same assay?

The gels have now been standardized to black-on-white throughout the manuscript.

If the 5'-CGCA motif is conserved, then why do the authors mutate the first 6 rather than the first 4 bps of the TSD sequence?

The experiment shown in the original submission was performed before we recognized that the first 4 bp are important, not 6. We have therefore removed this experiment panel from Figure 1B; instead, we have tested the effect of mutating only the first 4 bp of the TSD sequence in the experiment shown in Figure 3C.

Figure 1C. What are the red dots one end of the orange oligonucleotides?

The red dot indicates the 3'-OH that is the nucleophile for the integration reaction, and this is now stated in the figure legend for the schematic in Figure 1B. We have replaced the original schematic involving red dots with the explicit sequences of the oligonucleotides used (Figure 3C).

Figure 2A. In my opinion, seeing a gel of the purified protein is more important to me than seeing a series of SEC runs with increasing concentration of the DNA substrates.

We agree with the reviewer, and an SDS-PAGE gel is now shown in Figure 6A (in addition to the control lane in Figure 5—figure supplement 4), and the SEC data are now in supplementary material (Figure 4—figure supplement 1 and 2).

If the authors choose to show these in the main text, then a vertical line extending from the protein alone peak down though the other runs will help illustrate the shift.

We appreciate the suggestion, and vertical lines are now incorporated.

The authors disclose that "the complex did not produce the expected shift on SEC for a tetramer", but the peak does not look symmetrical to me and the left edge appears to be almost vertical. This would suggest that the complex may be eluting in the void, but the x-axis is in time rather than volume, so it is hard to know what this mean for a16/60 Superdex 200 column.

To resolve these issues, we performed further SEC experiments over a range of protein concentrations and generated a standard curve for the Superdex 200 Increase 3.2/300 column to allow us to estimate MWs. Although the peak shape is indeed asymmetric, at three different concentrations the integration product complex eluted with a peak position consistent with a tetramer bound to one DNA substrate.

Try a Sup 6 column and include a regression line fit with standers.

We believe we have resolved the basis for our original confusion (we had not recognized the effect of protein concentration on elution position), so have not repeated the experiment on a Superose 6 column. The regression line fit is now shown in Figure 4—figure supplement 2.

It appears that the labels for the x-axis are misplaced for the last column ("integration products").

Thank you for pointing this out, and the figure has been corrected.

In my opinion this entire figure could go in the supplemental material and Figure 1 should be refined and split into two figures.

We agree, and this has been done.

The text for Table 1 overlaps the table itself making it a little hard to interpret. I suggest adding a sentence to the text that states how the structure was determined (e.g., MR or experimental phases).

As suggested, we have included the following when we first introduce the structure: “…was solved to 3.1 Å using experimental phases obtained from a selenomethionine derivative.” The text for the table has been corrected.

Figure 3. Active site subunits "shown in pink sticks" are nearly impossible to see. I could not find an explanation for the green spheres in panel B.

The active site is now shown in a close-up view (Figure 5C), and an explanation for the green spheres (Ca2+ ions from the crystallization buffer) has been included in the figure legend to Figure 5B.

Figure 4. The authors state that "Each casposase monomer exhibits the fold expected of a Cas1 homolog" but none of the figures do a good job of showing this and I could not find an RMSD for a define set of equivalently positioned atoms.

We have added a figure (Figure 5—figure supplement 2) that shows the side-by-side structures of the casposase dimer and that of Cas1 from A. fulgidus, as well as those of the aligned N-termini and catalytic domains. The following sentence has now been included in the text: "When compared to its most closely related structurally characterized Cas1 homolog, that from Archaeoglobus fulgidus (Kim et al., 2013) with which it shares 32% sequence identity, the N-terminal domains have an RMSD of 1.68 Å over 72 shared Cα positions and the catalytic domains are even more similar, with an RMSD of 1.28 Å over 208 shared Cα positions."

Discussion paragraph five: delete word "is". “It is an intriguing possibility is that some casposases may already possess the ability to integrate short CRISPR-like spacers,…..”

Done – thank you.

Reviewer #2:

In their manuscript, Hickman and colleagues present the first crystal structure of a transposon-encoded casposase bound to an integration product mimic, revealing a tetrameric architecture in which two Cas1 dimers sandwich the target DNA destined to be duplicated during downstream molecular steps within the pathway. Biochemical experiments are presented to argue that single-stranded transposon ends may be the natural substrate utilized during integration, and that specific sequence motifs in both the transposon end and the target site affect biochemical integration activity. By comparing their casposase-DNA structure to available crystal structures of Cas1 homologs in the context of the Cas1-Cas2 complexes that mediate spacer acquisition during adaptive immunity in CRISPR-Cas systems, the authors argue a plausible evolutionary scenario by which gaining Cas2 interactions could have directed casposases towards an integration mechanism that is more akin to spacer acquisition than transposition.

The crystal structure is a notable advance that broadens our understanding of the Cas1/casposase protein family, and is undoubtedly a substantive contribution to the field. However, in this reviewer's opinion, the biochemical experiments that accompany this finding fall short of providing compelling evidence for a specific integration mechanism, and instead leave more to be desired in terms of what overhang geometries are optimal, how integration sites are chosen, and how casposons are excised. In addition, the authors gloss over the interesting finding that the HTH domain is disordered in the crystals and fail to follow-up with further experiments to investigate the role of this domain on biochemical integration activity, and whether the domain may be important in the context of more physiologically relevant substrates. I believe additional experiments are merited for this study to go beyond reporting a novel structure, to helping yield new mechanistic insights for the process of casposon transposition.

1) Some important controls are missing from the results presented in Figure 1, most notably negative controls in which the plasmid is incubated with casposase in the presence of divalent metal ions but no transposon end substrates. Previous Cas1 studies have observed considerable nicking activity (or possibly nicking/cleavage from contaminating proteins), so it is crucial that the baseline activity of the casposase protein prep on plasmid DNA is evaluated. Furthermore, for the data presented in panels C and D, no lanes exist showing the starting plasmid substrate at time 0 (i.e. in the absence of protein and transposon end substrate); since any given plasmid prep will have variable ratios of supercoiled to open-circular plasmid, it is critical that the reader have this comparison – for the exact substrate used in the experiments – to fully evaluate the biochemical activity of casposase and discriminate this from the starting status of the substrate. I would even argue that naming the bands "target," "SE," and "DE," is misleading, since a plasmid will naturally purify from a bacterial culture with some fraction in the open circular form, which will migrate the same as a SE product but is certainly not a single-end integration product.

These issues have been addressed by repeating the experiments in Figure 1, as discussed above.

2) A major question is left unaddressed (and mostly unmentioned) in the manuscript, namely, in what form the casposon is excised and integrated. I understand this is still unclear in the literature, that previous attempts to monitor excision by this group were unsuccessful (i.e. Hickman and Dyda, 2015), and that there is as yet no in vivo transposition data for casposons. Nevertheless, the reader should certainly be given more context on this topic, in the Introduction/Discussion, and in the Results section to better motivate the range of donor substrates tested.

We agree with the reviewer that the excision of casposons is an open question at this point. However, one possible consequence of the model for self-synthesizing transposons is that excision may not in fact be occurring as the element might be synthesized at the donor site without the need for excision. However, this is pure speculation at this point. We have included a few sentences in the revised Introduction in which we discuss how free casposon ends might be generated.

I would also like to see the authors test a larger panel of "Top Strand Recessed" substrates, and to more quantitatively compare integration activities amongst these substrates, and between these substrates and blunt dsDNA substrates. Similar experiments in the context of spacer acquisition by Cas1 complexes (e.g. Wright et al., 2019) have revealed the likely substrates that are used biologically, based on a clear optimum in terms of integration product formation kinetics. The authors have an opportunity to make similar discoveries with their casposase, but the use of just two timepoints, the lack of any product quantification, and the relatively small set of substrates tested, limits the potential of their biochemical assays.

As suggested, we have performed the plasmid integration assay with a larger panel of "Top Strand Recessed" substrates, ranging from blunt-ended to 7-nt recessed (Figure 3D). We have also performed the assay to compare blunt dsDNA, recessed, and ssDNA (Figure 3A). We also repeated the assay in triplicate in order to provide a quantitative comparison of their integration activities. These results are plotted in Figure 3—figure supplement 1.

3) Related to #2, I would prefer more experimental support and/or justification for the specific choice of the 8-nt overhang integration product. Why was 8 selected as the overhang length? Was a larger crystallization screen performed, and this was the only substrate geometry that resulted in diffracting crystals? Is there any further evidence that this is a physiologically relevant ssDNA end sequence? What happens if the casposon end substrates are shorter, or partially double-stranded?

As discussed above, we have tried a wide range of substrates in crystallization experiments and with the exception of the 8-nt overhang, none has yielded crystals of sufficient diffraction quality. We have added the following so the reader can understand why this particular complex was studied: "In an effort to further characterize the casposase and its interactions with its substrates, we attempted to crystallize the casposase bound to DNA oligonucleotides representing different stages of the integration reaction. Many casposase-DNA complexes yielded poorly-diffracting crystals including the casposase bound to recessed dsDNA oligonucleotides and ssDNA oligonucleotides of varying length representing the casposon ends. This was also the case with casposase bound to mimics of the post-integration step using both symmetric and asymmetric complete integration target sites and casposon ends of varying length and dsDNA-ssDNA composition. We were eventually able to obtain diffracting crystals of the casposase…"

For example, it seems that the iSCAMS analysis could be scaled up to test a wider range of model integration products, to investigate whether tetramer formation is more or less favored with certain substrates, which could help suggest what may or may not be more 'relevant.'

We agree with the reviewer, and this line of inquiry is one of our future plans.

4) The authors point out that capsposases have an additional C-terminal HTH domain that is absent in Cas1, and that this domain could not be modeled into the experimental electron density. In light of these interesting observations, the authors should test the biochemical activity casposase mutants in which this domain is truncated; does these mutants show behave similarly? Differently?

As discussed above, we have performed these experiments with a 1-341 construct (ΔHTH) and they are shown in Figure 6B. We note that we have not been able to express a range of truncated casposase mutants, as all other deletion constructs we generated prior to the structure determination were insoluble. In our hands, only Δ341 and Δ343 have been soluble.

Although it would be interesting to further investigate whether this domain becomes ordered with different DNA substrates, I understand that pursuing additional structures may be too much to ask.

Our pursuit of additional structures by X-ray crystallography is ongoing. However, in light of our experience to date with the difficulties of obtaining diffraction-quality crystals with interesting DNA constructs, we are also pursuing single particle cryo-EM experiments.

However, certainly some biochemical investigations are warranted here. I would also caution the authors against stating that the HTH domain "is not in a folded state" – although I am not a structural biologist, the absence of ordered density in a X-ray crystal structure is not equivalent to a domain being unfolded.

The reviewer is correct, and we have reworded this to indicate that "the C-terminal residues do not appear to be folded into a compact domain."

5) A major take-away from this manuscript would be that this casposase integrates substrates primarily at a specific target site rather than randomly, as has been previously touched upon in the literature. However, apparent integration events in pUC19 alone and farther away from the target site have been detected in this and previous studies. No sequencing data for these events are shown, and the authors failed to perform additional analysis on integration patterns. It has not become routine in spacer acquisition studies with Cas1 to deep sequence half-site integration products in order to gain deeper insights into specificity, and the authors should perform similar NGS experiments here to enhance their biochemical results. These experiments could provide much more data on the target site requirements, but looking for conservation (or lack of conservation) in the motifs that are preferred adjacent to integration sites other than at the 3' end of tRNA-Leu.

As mentioned above, NGS experiments have been performed and the results discussed in the text and shown in Figure 2.

Reviewer #3:

In this study, Hickman et al. describe the activity and structure of a casposase, the hypothesized evolutionary ancestor of the Cas1 integrase found in CRISPR-Cas adaptive immune systems. The authors demonstrate in vitro DNA integration by the casposase from Methanosarcina mazei, and perform biochemical experiments to establish the target sequence specificity and substrate preferences for this casposase. They also report, for the first time, a structure of the casposase bound to an integration product mimic. The casposase forms a tetramer composed of two casposase dimers sandwiching the DNA. The structure provides insight into the mechanism of substrate and target recognition by the casposase. In addition, comparison of the casposase and Cas1-Cas2 complex structures show that the casposase tetramer is mutually exclusive with Cas2 binding, and suggests how architectural changes in the Cas1-Cas2 complex may have led to mechanistic differences between the two types of integrases. The structure of the casposase is an important first step in understanding casposon function and the evolution of CRISPR-Cas adaptation mechanisms, and the authors propose several interesting hypotheses based on the structure. However, these hypotheses are largely speculative and could greatly benefit from experimental evidence. In particular, the structural data is not supported by any biochemical experiments to establish a structure-function relationship. Thus, the structure is largely descriptive, although the authors speculate on the importance of a number of observed interactions within the structure. Several examples are provided in comment 1 below, in addition to few other concerns that should also be addressed by the authors.

1) When describing the structure, the authors note several protein DNA contacts, but it is unclear whether any of these contacts contribute to the specificity for the TSD target or LE substrate. From the data shown in Figure 1C, it appears that the casposase is somewhat specific for the LE sequence, especially for shorter single-stranded DNAs. Can this specificity be explained by the structure? Do R191 and R192 contribute to the specificity of the casposase for the TSD? The authors speculate about the importance of stacking interactions for ssDNA substrate binding (subsection “ssDNA casposon end binding resembles that of spacer 3' overhangs”). Did the authors perform any mutagenesis experiments to establish the importance of these contacts for target and substrate specificity? In general, the study would be greatly improved by experiments supporting the structure-based hypotheses proposed by the authors when describing the structure.

As discussed above, we have now performed these experiments and the results are shown in Figure 7F and Figure 7—figure supplement 1.

2) The order in which substrates are shown on the left gel in Figure 1C is confusing. If possible, it would be helpful if this gel could be re-run to place casposon-specific and random substrates of similar type (e.g. TR17 vs ranTR17) in adjacent lanes on the gel. Is a dsran17 substrate (i.e. a fully double-stranded random substrate of 17 bp) also integrated? In addition, ssran12 and ssLE11 could be omitted from this gel, as these substrates are also tested in the rightmost gel in the panel.

We have rerun the assay to both expand the range of substrates tested and to place them in a logical order. The results are shown in Figure 3A.

3) In paragraph two of the Discussion section, the authors suggest a similarity in the lack of sequence specificity on the bottom strand of the casposase TSD with a lack of sequence specificity for spacer-side integration by Cas1-Cas2. However, it is important to note that Cas1-Cas2 does display specificity for the entire direct repeat sequence, likely through sequence-specific deformations of the repeat that occur during full-site integration (as proposed by Wright et al., 2017). In that study, it was shown that mutations in the repeat affect spacer-side integration, but do not have an effect on leader-side integration. Thus, the conclusion stated by the authors that bottom strand integration is sequence-independent and instead based on a "molecular ruler" is not consistent with current models for spacer-side integration.

We thank the reviewer for pointing this out. As discussed above, upon further examination of our data, we realize that our original statement was not correct. We have removed the entire section from the revised manuscript.

[Editors’ note: what follows is the authors’ response to the second round of review.]

Organization of the presentation:

The initial characterization of casposon substrate and target specificities is much improved by the new experiments. However, the presentation is a bit disjointed, as it jumps back and forth between characterization of substrate and target requirements. For example, Figures 1C, 3A-B and 3D all investigate casposon end requirements, while Figures 2 and 3C are on target specificity. Indeed, in one case this requires that the reader read ahead to understand the experiment explained in Figure 2 ("for the rationale, see Figure 3A and associated discussion"). This section could flow better if the authors combine Figure 1C, 3A-B, and 3D into a new Figure, which could be Figure 2. This would also allow the authors to show a panel describing all substrates used in these three panels, as it is currently confusing that TR17 and ssLE17 are shown in Figure 1C but not mentioned until much later in the text. Figures 2 and 3C could then be combined to form a new Figure 3. With these changes, the authors could first define casposon end requirements, and then describe the characterization of the target sequence.

We agree that most of these suggestions would help with the logic and the flow. However, a minor problem arises with Figure 3D which is certainly related to substrate specificity (as Figure 1C, 3A-B) but uses an assay that is introduced and validated only in Figure 3C (i.e., substrates are integrated into an oligo targ40 but not ran40). The solution we chose is to keep 3C and 3D together. We have therefore created a new Figure 2 with 1C, 3A-B, and combined old Figures 2 and 3C-3D into new Figure 3.

https://doi.org/10.7554/eLife.50004.sa2

Article and author information

Author details

  1. Alison B Hickman

    Laboratory of Molecular Biology, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, United States
    Contribution
    Conceptualization, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7666-0249

  2. Shweta Kailasan

    Laboratory of Molecular Biology, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, United States
    Present address
    Integrated BioTherapeutics, Inc, Rockville, United States
    Contribution
    Conceptualization, Formal analysis, Validation, Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0876-6812

  3. Pavol Genzor

    Laboratory of Cell and Molecular Biology, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, United States
    Contribution
    Conceptualization, Formal analysis, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared

  4. Astrid D Haase

    Laboratory of Cell and Molecular Biology, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, United States
    Contribution
    Conceptualization, Formal analysis, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared

  5. Fred Dyda

    Laboratory of Molecular Biology, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, United States
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology
    For correspondence
    Fred.Dyda@nih.gov
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1689-9041

Funding

National Institute of Diabetes and Digestive and Kidney Diseases (Intramural Program)

  • Alison B Hickman
  • Shweta Kailasan
  • Pavol Genzor
  • Astrid D Haase
  • Fred Dyda

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Dalibor Kosek for assistance with X-ray data collection, Duck-Yeon Lee of the Biochemical Core Facility at the NHLBI for LC/MS support, and Greg Piszczek and Di Wu of the Biophysics Core Facility at the NHLBI for help with iSCAMS measurements. Next-generation sequencing was performed by the NHLBI Sequencing and Genomics Core. This work was supported by the Intramural Program of the National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health. Data were collected at Southeast Regional Collaborative Access Team (SER-CAT) 22-ID beamline at the Advanced Photon Source, Argonne National Laboratory. SER-CAT is supported by its member institutions (www.ser-cat.org/members.html), and equipment grants (S10_RR25528 and S10_RR028976) from the National Institutes of Health. Use of the Advanced Photon Source was supported by the U. S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. W-31–109-Eng-38. Some of the computations were carried out using the High Performance Computing Systems at the NIH.

Senior Editor

  1. John Kuriyan, University of California, Berkeley, United States

Reviewing Editor

  1. Blake Wiedenheft, Montana State University, United States

Publication history

  1. Received: July 8, 2019
  2. Accepted: January 8, 2020
  3. Accepted Manuscript published: January 8, 2020 (version 1)
  4. Version of Record published: January 23, 2020 (version 2)

Copyright

This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.

Metrics

  • 2,625
    Page views
  • 457
    Downloads
  • 13
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Alison B Hickman
  2. Shweta Kailasan
  3. Pavol Genzor
  4. Astrid D Haase
  5. Fred Dyda
(2020)
Casposase structure and the mechanistic link between DNA transposition and spacer acquisition by CRISPR-Cas
eLife 9:e50004.
https://doi.org/10.7554/eLife.50004

Categories and tags

Further reading

    1. Structural Biology and Molecular Biophysics

    Robust membrane protein tweezers reveal the folding speed limit of helical membrane proteins

    Seoyoon Kim, Daehyo Lee ... Duyoung Min
    Tools and Resources Updated

    Single-molecule tweezers, such as magnetic tweezers, are powerful tools for probing nm-scale structural changes in single membrane proteins under force. However, the weak molecular tethers used for the membrane protein studies have limited the observation of long-time, repetitive molecular transitions due to force-induced bond breakage. The prolonged observation of numerous transitions is critical in reliable characterizations of structural states, kinetics, and energy barrier properties. Here, we present a robust single-molecule tweezer method that uses dibenzocyclooctyne cycloaddition and traptavidin binding, enabling the estimation of the folding ‘speed limit’ of helical membrane proteins. This method is >100 times more stable than a conventional linkage system regarding the lifetime, allowing for the survival for ~12 hr at 50 pN and ~1000 pulling cycle experiments. By using this method, we were able to observe numerous structural transitions of a designer single-chained transmembrane homodimer for 9 hr at 12 pN and reveal its folding pathway including the hidden dynamics of helix-coil transitions. We characterized the energy barrier heights and folding times for the transitions using a model-independent deconvolution method and the hidden Markov modeling analysis, respectively. The Kramers rate framework yields a considerably low-speed limit of 21 ms for a helical hairpin formation in lipid bilayers, compared to μs scale for soluble protein folding. This large discrepancy is likely due to the highly viscous nature of lipid membranes, retarding the helix-helix interactions. Our results offer a more valid guideline for relating the kinetics and free energies of membrane protein folding.

    1. Biochemistry and Chemical Biology
    2. Structural Biology and Molecular Biophysics

    The structural basis of the multi-step allosteric activation of Aurora B kinase

    Dario Segura-Peña, Oda Hovet ... Nikolina Sekulic
    Research Article Updated

    Aurora B, together with IN-box, the C-terminal part of INCENP, forms an enzymatic complex that ensures faithful cell division. The [Aurora B/IN-box] complex is activated by autophosphorylation in the Aurora B activation loop and in IN-box, but it is not clear how these phosphorylations activate the enzyme. We used a combination of experimental and computational studies to investigate the effects of phosphorylation on the molecular dynamics and structure of [Aurora B/IN-box]. In addition, we generated partially phosphorylated intermediates to analyze the contribution of each phosphorylation independently. We found that the dynamics of Aurora and IN-box are interconnected, and IN-box plays both positive and negative regulatory roles depending on the phosphorylation status of the enzyme complex. Phosphorylation in the activation loop of Aurora B occurs intramolecularly and prepares the enzyme complex for activation, but two phosphorylated sites are synergistically responsible for full enzyme activity.

    1. Biochemistry and Chemical Biology
    2. Structural Biology and Molecular Biophysics

    Calcium and bicarbonate signaling pathways have pivotal, resonating roles in matching ATP production to demand

    Maura Greiser, Mariusz Karbowski ... Liron Boyman
    Research Article

    Mitochondrial ATP production in cardiac ventricular myocytes must be continually adjusted to rapidly replenish the ATP consumed by the working heart. Two systems are known to be critical in this regulation: mitochondrial matrix Ca2+ ([Ca2+]m) and blood flow that is tuned by local ventricular myocyte metabolic signaling. However, these two regulatory systems do not fully account for the physiological range of ATP consumption observed. We report here on the identity, location, and signaling cascade of a third regulatory system -- CO2/bicarbonate. CO2 is generated in the mitochondrial matrix as a metabolic waste product of the oxidation of nutrients that powers ATP production. It is a lipid soluble gas that rapidly permeates the inner mitochondrial membrane (IMM) and produces bicarbonate (HCO3-) in a reaction accelerated by carbonic anhydrase (CA). The bicarbonate level is tracked physiologically by a bicarbonate-activated adenylyl cyclase, soluble adenylyl cyclase (sAC). Using structural Airyscan super-resolution imaging and functional measurements we find that sAC is primarily inside the mitochondria of ventricular myocytes where it generates cAMP when activated by HCO3-. Our data strongly suggest that ATP production in these mitochondria is regulated by this cAMP signaling cascade operating within the inter-membrane space (IMS) by activating local EPAC1 (Exchange Protein directly Activated by cAMP) which turns on Rap1 (Ras-related protein 1). Thus, mitochondrial ATP production is shown to be increased by bicarbonate-triggered sAC signaling through Rap1. Additional evidence is presented indicating that the cAMP signaling itself does not occur directly in the matrix. We also show that this third signaling process involving bicarbonate and sAC activates the cardiac mitochondrial ATP production machinery by working independently of, yet in conjunction with, [Ca2+]m-dependent ATP production to meet the energy needs of cellular activity in both health and disease. We propose that the bicarbonate and calcium signaling arms function in a resonant or complementary manner to match mitochondrial ATP production to the full range of energy consumption in cardiac ventricular myocytes in health and disease.