C2566 [pTYB1-bsaXIRM] and C2566 [pTYB1-bsaXIS] cells were cultured in 1 L flasks of LB + ampicillin and grown at 30°C until OD600 reached 1.3. After sitting at room temperature for 30 min, cultures were induced with 0.4 mM IPTG and grown overnight at 18°C. The cells were harvested by centrifugation and stored at –80°C. The frozen cell pellets from 3 L of BsaXI RM culture and 6 L of BsaXI S culture were all thawed (preliminary experiment indicated that the expression level of RM subunit is higher than the S subunit), resuspended in chitin column buffer (500 mM NaCl, 20 mM Tri-HCl, pH 8.5, 1 mM EDTA, and 0.1% Triton X-100 or Tween 20), lysed through 2 passes at 30 kpsi in a Dyhydromatics HL60 cell disruptor, and centrifuged to remove cell debris. The crude supernatants were loaded through two separate 40-ml chitin resin (NEB S6651) columns by gravity, washed with 300 ml of chitin column buffer, and then cleaved over 3 days with 50 mM DTT chitin cleavage/elution buffer. All chitin eluates were pooled together in 450 ml, diluted to 300-mM NaCl with addition of 300-mL zero salt column buffer and flowed through 20-ml diethylaminoethanol (DEAE) resin to remove nucleic acids. The DEAE flow-through and wash were again diluted to 150-mM NaCl with zero salt column buffer, applied to a Heparin TSK column (11 ml) and eluted with increasing NaCl gradient. The pooled active fractions were dialyzed overnight against 100-mM NaCl, 20-mM Tris-HCl, pH 8, 1 mM DTT, 0.1 mM EDTA, and 5% glycerol, passed through a Source 15S column (21 ml), then applied to 22 ml Source 15Q and eluted with increasing NaCl gradient (0.1 to 1 M). The pooled active fractions were dialyzed overnight against 50% glycerol (NEB Diluent B) and applied to a Superdex75 SEC column (1,787 ml). The pooled active fractions were concentrated by a second run over Heparin TSK. The final yield was ∼10.0 mg of pure BsaXI REase ([RM]2 + S). One BsaXI unit (U) is defined as the amount of protein required for complete digestion of 1-μg phage λ DNA in CutSmart buffer at 37°C for 1 h.

A 2 L of IPTG-induced cells containing RM-intein-CBD or 6 L of IPTG-induced cells of S-intein-CBD were resuspended in 60 ml and 180 ml of chitin column buffer and the suspensions sonicated to lyse the cells. After the removal of the cell debris by the centrifugation at 15,000 rpm at 4°C for 30 min, clarified cell lysates were loaded onto 20-ml chitin columns by gravity flow (1 column for RM-intein-CBD, 2 columns for S-intein-CBD). The CBD-tagged enzymes in flow-through were reloaded 3 times and the columns were washed with 10 column volumes of chitin buffer (∼200 ml). A 20 ml of cleavage buffer (chitin column buffer + 50 mM DTT) was added to the top of chitin column. Approximately 2 ml of cleavage buffer passed through the column and the flow-through was discarded. The DTT-catalyzed intein cleavage reaction continued at 4°C for 2-3 days. The eluted proteins from chitin columns were diluted in a low salt buffer to reach 0.1-M NaCl concentration, which was subsequently loaded onto 5-ml Hi-Trap Heparin column (GE Healthcare). After the extensive washing with low salt, the RM or S protein was eluted with a salt gradient (0.1–1-M NaCl). Eluted peak fractions were analyzed by SDS-PAGE, concentrated in protein concentrators and protein was resuspended in enzyme storage buffer (0.2 M NaCl, 20 mM Tris-HCl, pH 7.5, 10 mM DTT, 0.5 mM EDTA, 50% sterile glycerol) to be stored at –20°C.

The synthetic genes (gene blocks) encoding S variants were purchased from IDT, which carry 24-27 bp overlapping vector sequences at the 5′ and 3′ ends. They were assembled into pTYB1 using Gibson assembly kit and the assembled DNA was transferred into T7 expression strain C2566 (T7 Express, NEB) by transformation. To construct 6xHis-tagged S expression clone, A PCR fragment containing bsaXIS gene was cloned into pET21b (NdeI and XhoI cut, Novagen) to achieve C-terminal 6xHis tag of the targe protein. The pET21-bsaXIS plasmid was transferred into T7 Express (C2566) for expression. The BsaXI S (6xHis) tagged protein was purified from a nickel agarose beads column (10 ml, NEB) by gravity flow using a protocol supplied by NEB.

WT: TRD1-CR1-TRD2-CR2 (54.98 kDa)

TRD1 and TRD2: The aa residues are shown in blue and green, respectively. The CR1 and CR2, underlined aa residues. The boundary of CR1 (42-aa) and CR2 (42-aa) are predicted by structured-guided Phyre2 aa sequence alignment (Kelley et al., 2015). The PIPP residues shown in red indicates the boundary between TRD2 and CR2 (see Supplementary Figure 1 for the sequence alignment). The boundary between TRD1 and CR1 is less clear-cut as CR1 may include 1–2 additional aa residues depending on the secondary structure prediction software; similarly, the boundary between CR1 and TRD2 has a 5-aa sequence (HFNIN) that is not part of either domain (see Supplementary Figure 1).

CspC0110I TRD homolog (partial) was found in GenBank by BlastP search with 59% aa sequence identity to BsaXI TRD2 (see below for GenBank accession number). It was annotated as a Type I HsdS partial sequence in Cyanothece sp. CCY0110.

Hypothetical Type I restriction enzyme EcoEI specificity protein (S protein, partial)

WP_008279016, 273 aa, linear, BCT 05-JUN-2013.

hypothetical Type I restriction enzyme EcoEI specificity protein (S protein), partial [Cyanothece sp. CCY0110].

WP_008279016.1, GI:495554437. ORGANISM: Cyanothece sp. CCY0110 (Bacteria; Cyanobacteria; Oscillatoriophycideae;

Chroococcales; Cyanothece).

Protein: 1.273.

Product: hypothetical Type I restriction enzyme, EcoEI specificity protein (S protein).

Note: EcoKI restriction–modification system protein HsdS; Provisional.

Chimeric S subunit TRD1-CR1-[CstTRD]-CR2

The BsaXI TRD2 and Cst TRD2 share 28% aa sequence identity. The Cst S subunit is annotated as a Type I

specificity protein in GenBank (accession number WP_005530828).

LOCUS: WP_005530828 371 aa

DEFINITION: type I restriction modification DNA specificity protein

SOURCE: Corynebacterium striatum

ORGANISM classification: Bacteria; Actinobacteria; Actinobacteridae; Actinomycetales;

Corynebacterineae; Corynebacteriaceae; Corynebacterium.

The BsaXI restriction–modification (R–M) system was originally identified from random shotgun sequencing of B. stearothermophilus 25B (Geobacillus sp. 25B) gDNA (data are not shown). A part of the sequences was verified by the primer walking using Sanger sequencing. The R–M system contains two genes, the bsaXIRM gene encoding the restriction–modification subunit (RM fusion, 911 aa, predicted molecular mass 107.04 kDa) and the bsaXIS gene coding for the specificity subunit (S, 476 aa, molecular mass 54.97 kDa). The RM and S genes are probably regulated in a single transcription unit with RM gene preceding the S gene. The DNA sequence has been deposited in GenBank (accession number OM373208). The endonuclease domain in the N-terminal region of RM subunit contains two potential PD-D/ExK catalytic sites (D-X16-DxK or D-X17-ExK). It is predicted that the second motif may form the catalytic site since it is analogous to the catalytic residues of BcgI: PE-X12-ExK (Kong, 1998). The exact catalytic residues remain to be confirmed experimentally. The methylase domain in the C-terminal region contains typical N6mA (6 mA) methyltransferase (MTase) conserved sequence motifs. The methylase activity is predicted to modify the adenine (A) in the target site 5′ ↓N9 AC N5 CTCC N10↓ 3′, and the adenine in the bottom strand opposite to the T base. The MTase activity of the purified BsaXI RM-S enzyme complex is low and only provided partial modification in vitro (see below). The sequence alignment with the S subunits of other Type I and Type IIB systems by Phyre2 indicated that the S subunit contains protein subdomains TRD1-CR1-TRD2-CR2 in which CR1 and CR2 are predicted to form long α-helix coiled-coil and interact with the RM subunits (see below for more detailed analysis of TRDs, Supplementary Figure 1). The BsaXI RM and S gene products are nearly identical to two putative R–M systems listed in REBASE: The 99–100% sequence identity to Gth3921I and Gka8005I (RM fusion, WP_052369193; S subunit, WP_052369191; Geobacillus kaustophilus NBRC 102445 DNA sequence contig, NZ_BBJV01000045), suggesting that the BsaXI R–M system (or isoschizomers) may have been evolved in other Geobacillus strains via horizontal gene transfer. The homologs of BsaXI R–M systems are widely distributed in other microbial genomes: More than 60 RM fusion homologs, with 44–100% aa sequence identity, are found in GenBank in a recent BlastP search (data are not shown).

The co-expression of BsaXI RM and S genes in pUC19 appeared to be unstable (data are not shown). The reason for this might be insufficient methylation. Therefore, the two genes were expressed separately in fusion with intein and chitin binding domain (CBD) in pTYB1. The fusion to intein and CBD may also contribute to the reduced toxicity of RM expression. The cell extracts containing RM and S subunits could be mixed in test tube and the RM/S subunits in the complex was co-purified. Figure 1A shows the co-purified BsaXI after chromatography through six columns (Chitin, DEAE, Heparin Sepharose, Source 15 S, Source 15Q, and gel filtration). Based on the protein molecular mass and band intensity (Figure 1A and Supplementary Figure 2), it was estimated that enzyme stoichiometry of RM:S is 2:1 [band intensity (total pixels ratio) = 2:1.2 as measured by gel imaging software, Bio Rad], and the active enzyme complex can be shown as trimeric complex [RM]2 S similar to the first characterized Type IIB enzyme BcgI (Kong, 1998). However, we cannot rule out the possibility of forming dimer of trimer ([RM]2 S)2 in the presence of cognate DNA. Two BsaXI subunits could be purified separately by chromatography through chitin and heparin columns (Figure 1B). To increase the expression level for BsaXI S subunit, we also constructed an expression clone in pET21b with 6xHis tag. Then BsaXI S (6xHis) was purified by chromatography through a nickel agarose column (Supplementary Figure 2). Figure 2A shows the restriction activity of co-purified BsaXI endonuclease. At high enzyme concentrations, the activity was inhibited either due to protein aggregation or to a lack of available target sites as “overcrowded” enzyme non-specific binding. The BsaXI activity can also be reconstituted by mixing the partially purified RM and S (6xHis) subunits to achieve partial digestion (Figure 2B). Increasing BsaXI RM subunit concentration slightly improved the cleavage reaction, but still showing partial digestion (Supplementary Figure 3). The activity of in vitro reconstituted BsaXI activity by mixing the two subunits together in digestion of λ-DNA is lower than the co-purified enzyme complex, probably due to slow subunit unit association, non-optimal subunit ratio or non-productive complex formation.

It has been observed that TRDs in a Type I RM system can be shuffled with circular permutation (Loenen et al., 2014). Due to the modular organization of TRD1 and TRD2, we constructed and purified the rearranged BsaXI TRDs in circular permutated form TRD2-CR2-TRD1-CR1. The S variant is active in cleavage of M13 dsDNA in complex with the RM subunits. Figure 5 shows the run-off sequencing of two BsaXI sites 5′ GGAG N5 GT 3′ in M13 dsDNA, which confirmed the recognition sequence and slightly shifted cleavage distance at N12-13 and N10-13 over the longer staggered cuts. The shorter staggered cuts were not affected at N7 and N9. It is not clear why TRD2-CR2 variant could not be expressed in E. coli, but S variants CR1-TRD2-CR2 and TRD2-CR2-TRD1-CR1 could be expressed and purified.

A BsaXI TRD2 homolog CspC0110I TRD (partial sequence) was found in GenBank by a BlastP search. The two TRDs share 59% aa sequence identity (see Materials and Methods section for the aa sequence). It was annotated as a Type I HsdS partial sequence in Cyanothece sp. CCY0110 which is similar to EcoEI HsdS. We constructed and purified the chimeric protein consisting of TRD1-CR1-[CspC0110T TRD]-CR2. The restriction activity was reconstituted when the chimeric S subunit is combined with the BsaXI RM subunits and the cleavage pattern is identical to BsaXI (data are not shown).

Another S variant (Chimeric S subunit: TRD1-CR1-[CstTRD]-CR2) (28% aa sequence identity between Cst TRD and BsaXI TRD2) could not be expressed in E. coli. We concluded from this experiment that highly homologous TRD (59% sequence identity) can be used to replace BsaXI TRD2 and generate active enzyme with BsaXI RM subunit. TRDs with low sequence identity to BsaXI TRD2 may be problematic possibly due to incompatible TRD with BsaXI CR1 and CR2.

The results of all BsaXI TRD rearrangement and mutagenesis are summarized in Table 1.

The BsaXI is predicted to carry both endonuclease and methylase activity. To detect BsaXI methylase activity of the RM/S complex, phage λ-DNA was first incubated with BsaXI in the presence of methyl donor SAM without Mg²⁺ cations, and in the presence of 1 mM EDTA (restriction activity is inhibited by the absence of divalent cations). After 2 h methylation, MgCl₂ (10 mM) was added to the reaction, and 2 U of BsaXI was added to chase the cleavage reaction. After DNA methylation reaction, the DNA was partially resistant to BsaXI digestion (data not shown). The substrate DNA was only partially modified, in agreement with the in vivo expression result that the co-expression of RM and S genes in E. coli host is toxic and unstable on the same plasmid due to insufficient methylation. It is not clear whether there is another MTase that modified the same sequence from the native host. We have not extensively analyzed the methylase activities of the newly engineered TRDs (TRD1-CR1, TRD1-CR1-TRD1-CR2, and CR1-TRD2-CR2) in complex with the WT RM subunits.

Only two homologs of BsaXI standalone TRD1 are found in GenBank (WP_052564242 = 177 aa and KAA0244900 = 175 aa). Phyre2 search predicted that both TRD1 homologs lack the CR region (data are not shown). Therefore, either sequencing errors resulted in premature termination (missing CR region) or a separate peptide for the CR region or the TRD homologs may function in the absence of CR sequence. In the genome of Candidatus Brocadia sinica, two ORFs encoding TRD2 (240 aa) and RM (928 aa) are also present, suggesting they might encode an R–M system (CbrSI). It was noted that Cbr TRD1 and RM putative proteins share high sequence identity to the counterparts in BsaXI (39.3% and 55.9% aa sequence identity). However, Cbr TRD2 only shared 23.3% aa sequence identity to BsaXI TRD2. Most BsaXI TRD1 homologs are present as fusion to other TRDs in the S subunits of putative Type I and IIB restriction systems.

BlastP search also identified more than 36 standalone BsaXI TRD2 homologs (196 to 248 aa long). Some homologs are shown in Supplementary Figure 5. Most of the BsaXI TRD2 homologs are fused to BsaXI TRD1 homologs or fused to other TRDs with low sequence identity, suggesting BsaXI TRD2 homologs may have partnered with other TRDs to create new specificities.