Cas9-mediated genome editing in the methanogenic archaeon Methanosarcina acetivorans

To determine whether the appropriate components for genome editing from S. pyogenes are functional in M. acetivorans, we constructed a Methanosarcina/Escherichia coli shuttle vector that expresses the S. pyogenes Cas9 ORF from a tetracycline inducible Methanosarcina promoter (Fig. 1A). We also constructed a derivative of this plasmid that employs a methanol-inducible promoter to express an sgRNA that targets Cas9 to ssuC, a gene required for uptake of the methanogenesis inhibitor bromoethane sulfonic acid (BES) (18) (Fig. 1 A and B). M. acetivorans was readily transformed with the Cas9-only plasmid pDN206 (78,900 ± 9,940 Pur^R transformants); however, pDN208, which contains the ssuC-targeting sgRNA in addition to Cas9, produced only 4 ± 3 Pur^R transformants. This difference in plating efficiency of more than four orders-of-magnitude strongly suggests that Cas9 is not toxic by itself, but that the Cas9–sgRNA complex from S. pyogenes is capable of generating a lethal DSB in M. acetivorans. Similar results were obtained with and without the inducers methanol and tetracycline.

Next, we determined the ability of the native HDR machinery in M. acetivorans to repair the lethal DSB generated by the sgRNA–Cas9 complex. To this end, repair templates of varying size were added to the ssuC-targeting vector. These repair templates generate a 34-bp deletion/frameshift mutation within ssuC that removes the targeting site while simultaneously introducing a diagnostic NotI restriction endonuclease site (Fig. 1B). Addition of repair templates with 1-kb homology arms to the plasmids relieved the lethal effect of targeting Cas9 to ssuC, generating nearly 20,000 Pur^R transformants per 2 μg DNA. A similar plasmid with 0.5-kbp homology arms generated roughly half as many transformants. Significantly, the 10³-fold higher transformation efficiency for pDN211 relative to pDN208 indicated that 99.9% of the Pur^R transformants are likely to be mutants (i.e., only 1 of every 1,000 Pur^R transformants would still contain the WT locus). To validate this hypothesis, 20 of these transformants were genotyped by a performing a NotI digest of a PCR amplicon containing the edited ssuC locus: all tested positive for the introduced mutation (Fig. 1C). Furthermore, as expected for null mutations in the ssu locus, all 20 transformants were resistant to 0.4 mM BES, a concentration lethal to the parent strain. Genome editing was also observed when plasmids were integrated into the chromosome using a ΦC31 integrase system (17).

The initial gene-edited strains produced in these experiments retain the targeting machinery; thus, we constructed plasmid derivatives that include a counter selectable marker (hpt) to facilitate curing of gene-editing vector. This marker confers sensitivity to the purine analog 8-aza-2,6-diaminopurine (8ADP) in strains that lack the native hpt gene (19). To validate the plasmid curing system, which has not previously been attempted in Methanosarcina, we selected 8ADP^R clones from three independent Pur^R transformants constructed using the counter selectable vectors. All 8ADP^R isolates analyzed were Pur^S and also contain the frameshift mutation at the ssuC locus (Fig. S2A). PCR-based screening with plasmid-specific primers showed that the vector was indeed cured from these strains (Fig. S2B). These proof-of-principle experiments show that a Cas9-mediated genome editing technique can be used to effectively introduce unmarked mutations in M. acetivorans.

To determine the optimal expression levels for the genome-editing machinery, we varied the transcription of Cas9 by selecting transformants on media with increasing concentrations of tetracycline, and of the sgRNA by plating on media with either methanol (induced) or trimethylamine (TMA; repressed) as growth substrates. Surprisingly, no significant difference in genome-editing efficiency was observed (Fig. 2A). In fact, the basal level of transcription provided by the two promoters in the absence of the inducers was sufficient for effective editing (Fig. 2A). A control vector identical to pDN211 but lacking the sgRNA (pDN207) was used to estimate the efficiency of genome editing. The efficiency of genome editing was measured as the ratio of mutant recovery (i.e., plating efficiency of pDN211) relative to the plating efficiency of the control vector and was estimated on media with either methanol or TMA as growth substrates. Significantly, genome editing in these experiments was particularly efficient, with edited strains being obtained at frequencies of approximately 20–25% relative to the control (i.e., one in four cells that receive the plasmid undergo gene conversion) (Fig. 2B).

To examine the maximum size of deletions that can be reliably generated by a single sgRNA, we tested repair templates with 1-kb homology arms placed at varying distance from the sgRNA-directed DSB (Fig. 2C). The transformation efficiency remained steady for templates that are ≤250 bp away from each end of the DSB, but declined precipitously when the distance increased beyond this point (Fig. 2C). Thus, a single sgRNA can be reliably used to delete up to 0.5 kbp of the chromosome, although larger deletions (up to 1 kbp) can be produced at the expense of efficiency.

To explore the possibility of using multiple Cas9-mediated DSBs to create larger deletions, or to simultaneously introduce more than one mutation, we tested two alternate arrangements for the expression of multiple sgRNAs. In the first arrangement, sgRNAs were expressed individually, whereas in the second they were expressed as a single transcript separated by a 30-bp linker sequence (Fig. 3A). Plasmids with sgRNAs in either arrangement were equally efficient in generating strains with complete deletions (approximately 2 kbp) of the mtmCB1 and mtmCB2 loci, which are highly homologous genes encoding monomethylamine methyltransferase isozymes (Fig. 3B and Fig. S3). Thus, to reduce the size of mutagenic plasmids and simultaneously simplify their design, the placement of sgRNAs on a single transcript was preferred. Subsequently, we generated a plasmid containing all four sgRNAs and each of the corresponding repair templates to simultaneously delete mtmCB1 and mtmCB2 (Fig. S4). Surprisingly, transformants that simultaneously acquired both the ΔmtmCB1 and ΔmtmCB2 mutation were obtained at the same frequency as transformants that acquired only one of two mutations (Fig. 3C). Furthermore, the genomes of two random, independent isolates each for the ΔmtmCB1, ΔmtmCB2, and ΔmtmCB1ΔmtmCB2 mutants were completely sequenced and no off-target activity was detected (Table S1). This finding is especially notable given the high levels of homology between the sgRNA target sites in the two genes (Fig. S5). Hence, Cas9-mediated genome editing is remarkably precise in M. acetivorans.

To assess the efficacy of gene knockins, plasmids designed to insert either a 3.05-kbp fragment containing the mtmCB1 locus or a 2.53-kbp fragment containing the mtmCB2 locus were constructed and used to introduce WT copies of each gene into the ssuC gene of the ΔmtmCB1ΔmtmCB2 double mutant (Fig. S6). Although 20–60% fewer transformants were observed (relative to the simple 34-bp deletion mutation described above), all transformants screened contain the desired gene insertions.

Two methods have previously been used to test gene essentiality in Methanosarcina (17, 20); however, both approaches are laborious and time-consuming. We therefore sought to use the efficient repair of the DSB generated by the Cas9–sgRNA complex to assay gene essentiality in M. acetivorans. For nonessential genes (ssuC, mtmCB1, mtmCB2), 10³- to 10⁴-fold more transformants are consistently observed when a repair template to generate a deletion is provided in addition to the sgRNA–Cas9 complex. We expected that this would not be true for essential genes, because HDR-directed repair using a deletion cassette would also be lethal. To test this idea, we constructed plasmids, with and without a repair template, that target the previously established essential genes mcrA and hdrED (17, 21). In contrast to the results with the nonessential ssuC, mtmCB1 and mtmCB2 loci, where we obtained thousands of transformants in the presence of a repair template, plasmids that targeted the essential genes generated fewer than five transformants, regardless of whether a repair template is present (Table 1). Thus, the ratio of transformants obtained in the presence versus absence of a repair template can be used as a reliable and simple test for gene essentiality in M. acetivorans.

A lethal phenotype for plasmids expressing a Cas9–sgRNA complex in the absence of a repair template was uniformly observed across a wide range of sgRNAs tested in this study, suggesting that NHEJ does not occur in M. acetivorans (Table 1). This result is consistent with the absence of genes related to Ku and LigD in the completely sequenced genome (22). Nevertheless, in some circumstances HDR-independent gene editing would be very useful. Therefore, we examined whether NHEJ could be established in Methanosarcina for use in conjunction with the Cas9–sgRNA complex. For this purpose, we chose the NHEJ machinery from the closely related methanogen M. paludicola, which has previously been reconstituted in vitro (23, 24). An artificial operon encoding four M. paludicola NHEJ proteins [DNA ligase (Lig), polymerase (Pol), phosphoesterase (PE), and Ku] was synthesized and transcriptionally fused to the moderately expressed serC promoter (25) to allow transcription in M. acetivorans (Fig. S7). This cassette was then added to the Cas9 ssuC-targeting vector without a repair template. Transformation with this plasmid was approximately 100-fold less efficient than the corresponding HDR vector, but approximately 10-fold higher than with plasmids lacking the NHEJ system (Fig. 4A). Therefore, expression of the M. paludicola NHEJ machinery overcame the lethal effect of the Cas9 ssuC-targeting vector without a repair template. Molecular analysis of the ssuC locus in these transformants revealed deletions ranging from 75 bp to 2.7 kb in length, often occurring at naturally occurring regions of microhomology 6–11 bp in length (Fig. 4B). Thus, the combined Cas9/NHEJ system provides the opportunity to generate a variety of mutations surrounding a single target site. Importantly, these plasmids are much simpler to construct, requiring only addition of target-specific sgRNA. We also tested whether addition of two sgRNAs targeting DNA sequences approximately 450 bp apart in conjunction with NHEJ could be used to generate precise deletions without a repair template. Interestingly, attempts to construct ssuC deletions via this method were not successful: only a handful of colonies were obtained (6 ± 3 per 2 μg plasmid) and none had the precise deletion desired. We examined 20 transformants obtained by this method. Two contained a 1.3-kb deletion of the ssuC locus, which occurred at a region of microhomology (Fig. 4B). The remainder had WT copies of the ssuC gene and, thus, are likely to be so-called escape mutants in which the Cas9 gene or sgRNA has mutated on the targeting plasmid (26).

All chemicals were purchased from Sigma-Aldrich unless otherwise specified. M. acetivorans strains were grown in single-cell morphology (38) at 37 °C in bicarbonate-buffered high-salt (HS) liquid medium containing 125 mM methanol or 50 mM TMA hydrochloride in Balch tubes with N₂/CO₂ (80/20). Plating solid medium was conducted in an anaerobic glove chamber (Coy Laboratory Products) as described previously (25). Solid media plates were incubated in an intrachamber anaerobic incubator maintained at 37 °C with N₂/CO₂/H₂S (79.9/20/0.1) in the headspace, as described previously (39). Puromycin (CalBiochem), the purine analog 8ADP (R. I. Chemicals) and BES were added to a final concentration of 2 μg/mL, 20 μg/mL, 0.4 mM, respectively, from sterile, anaerobic stock solutions. Anaerobic, sterile stocks of tetracycline hydrochloride in deionized water were prepared fresh shortly before use and added to a final concentration as indicated. E. coli strains were grown in LB broth at 37 °C with standard antibiotic concentrations. WM4489, a DH10B derivative engineered to control copy-number of oriV-based plasmids (40), was used as the host strain for all plasmids generated in this study (Table S2). Plasmid copy number was increased by adding sterile rhamnose to a final concentration of 10 mM.

All plasmids used in this study are listed in Table S2. The plasmid pMJ0806 was obtained from Jennifer Doudna, University of California, Berkeley, CA (Addgene plasmid # 39312). The S. pyogenes (Spy) Cas9 ORF was fused to the PmcrB(tetO1) promoter in pJK027A (17), and linearized with NdeI and HindIII by the Gibson assembly method, as described previously (41). The DNA segments containing sgRNA flanked by putative mtaCB1 promoter and terminator sequences from M. acetivorans were synthesized as double-stranded DNA fragments (“gBlocks”) from Integrated DNA Technologies and used for cloning purposes per the manufacturer’s instructions. A 3.25-kbp artificial operon with the NHEJ polymerase (Mcp_2125), DNA ligase (Mcp_2126), phosphoesterase (Mcp_2127), and Ku (Mcp_0581) genes from M. paludicola SANAE fused to the Methanosarcina barkeri Fusaro serC promoter was ordered from the GeneArt gene synthesis service (Life Technologies). All synthetic DNA fragments and repair templates were introduced in the appropriate vector backbone linearized with either AscI or PmeI by the Gibson assembly method, as described previously (41). The entire pC2A plasmid was introduced in the appropriate pJK027A-derived vector (carrying the λattB site) by retrofitting with pAMG40 (carrying the λattP site) using the BP Clonase II master mix (Invitrogen) per the manufacturer’s instructions. WM4489 was transformed by electroporation at 1.8 kV using an E. coli Gene Pulser (Bio-Rad). Standard techniques were used for the isolation and manipulation of plasmid DNA. All pJK027A-derived plasmids were verified by Sanger sequencing at the Roy J. Carver Biotechnology Center, University of Illinois at Urbana–Champaign, and all pAMG40 cointegrates were verified by restriction endonuclease analysis. Primers used in this study are listed in Table S3. The plasmid sequence and annotations for pDN211 have been submitted to GenBank (accession no. KY436376).

All target sequences used in this study are listed in Table S4. Target sequences were designed using the CRISPR site finder tool in Geneious version R9 (42). The M. acetivorans chromosome and the plasmid pC2A were used to score off-target binding sites.

All M. acetivorans strains used in this study are listed in Table S5. Liposome-mediated transformation was used for M. acetivorans, as described previously(43), and 10 mL of late-exponential phase culture of M. acetivorans and 2 μg of plasmid DNA were used for each transformation.

Genomic DNA from M. acetivorans was extracted using a protocol described previously (44). DNA libraries were prepared with the Hyper Library construction kit (Kapa Biosystems) and quantified using qPCR. All libraries were sequenced on one lane of an Illumina MiSeq v2 (Illumina) at the Roy J. Carver Biotechnology Center, University of Illinois at Urbana–Champaign using a 500 cycles v2 sequencing kit (Illumina). Trimmed, paired end 250-nt reads were mapped to the M. acetivorans reference genome (NC_003552) using default parameters for breseq v0.25 (45). Trimmed genome sequencing reads have been deposited in the Sequenced Reads Archive at the National Center for Biotechnology Information under accession no. PRJNA352863.