High-throughput continuous evolution of compact Cas9 variants targeting single-nucleotide-pyrimidine PAMs

Despite the availability of Cas9 variants with varied protospacer-adjacent motif (PAM) compatibilities, some genomic loci—especially those with pyrimidine-rich PAM sequences—remain inaccessible by high-activity Cas9 proteins. Moreover, broadening PAM sequence compatibility through engineering can increase off-target activity. With directed evolution, we generated four Cas9 variants that together enable targeting of most pyrimidine-rich PAM sequences in the human genome. Using phage-assisted noncontinuous evolution and eVOLVER-supported phage-assisted continuous evolution, we evolved Nme2Cas9, a compact Cas9 variant, into variants that recognize single-nucleotide pyrimidine-PAM sequences. We developed a general selection strategy that requires functional editing with fully specified target protospacers and PAMs. We applied this selection to evolve high-activity variants eNme2-T.1, eNme2-T.2, eNme2-C and eNme2-C.NR. Variants eNme2-T.1 and eNme2-T.2 offer access to N4TN PAM sequences with comparable editing efficiencies as existing variants, while eNme2-C and eNme2-C.NR offer less restrictive PAM requirements, comparable or higher activity in a variety of human cell types and lower off-target activity at N4CN PAM sequences.

Supplementary Figure 5. Flow rate schedule and titers for ePACE1. SP containing wildtype, full-length Nme2ABE8e were first diversified in E.coli host cells containing pJC175e 2 and MP6 2 , isolated, then seeded into ePACE1 (eight chemostats, one lagoon each targeting each of the eight N3YTN PAMs). Flow rate stringency for each PAM is shown in the plots, as are resulting titers (measured by qPCR). If lagoons were reseeded with starting phage, the timepoint is highlighted in a green circle. The N3TTA lagoon failed prematurely due to a pump failure in the ePACE setup. LOD=limit of detection of qPCR titering, as set by the titer corresponding to the Cq for which the qPCR primers alone had been observed to amplify.   Figure 7. Flow rate schedule and titers for ePACE2. SP previously isolated from ePACE1 lagoons evolved on N3TTC and N3CTC PAMs were pooled and reseeded into ePACE2 (eight chemostats, two lagoons each targeting each of the eight N3YTN PAMs). Flow rate stringency for each PAM is shown in the plots, as are resulting titers (measured by qPCR). LOD=limit of detection of qPCR titering, as set by the titer corresponding to the Cq for which the qPCR primers alone had been observed to amplify.     number 26 49 53 71 102 124 156 163 2  47 53 68 94 119 123 154 186 323 340 361 396 409 424 431 441  wild-type R  A  H  M  R  Y  V  A  A  E  K  V  A  D  T  E  E  L  D  E  T  E  S  I  Y  CTC-L1.E2-1 16  20  20  20  20  20  20  20  20  20  20  20  20  17  20  20  20  20  20  20  20  20  20  20  20  20  18  20  20  20  20  20  20  20  20  20  20  20  20  19  20  20  20  20  20  20  20  20  20  20  20  chemostats of each of six N3WTD PAMs, where W=A or T and D=A,G, or T; 4 replicates). (a) Passage stringency schedule and resulting titers (measured by qPCR) for replicates 1 and 2 (top) or replicates 3 and 4 (bottom). Passages were done after 16-24 hr for all passages. For some passages, some conditions were passaged uniquely to others or in a different host cell line, and these changes are listed in the Notes column. Grey coloring represents titers that were not measured or the PAM had not yet been included. All N3ACD PAMs were unable to support phage propagation, which retroactively was discovered to be attributable to an AP design error (see Supplementary Note 6). LOD=limit of detection of qPCR titering, as set by the titer corresponding to the Cq for which the qPCR primers alone had been observed to amplify. (b) PANCE conditions used for N1.  each of the three N3TCD PAMs). N1 replicates 1 and 2 were pooled into "Lagoon 1" lagoons, N1 replicates 3 and 4 were pooled into "Lagoon 2" lagoons, and all N1 replicates were pooled into any "Lagoon 3" lagoons. The N3ACD PAMs all washed out, which retroactively was discovered to be attributable to an AP design error (see Supplementary Note 6). Flow rate stringency for each PAM is shown in the plots, as are resulting titers (measured by qPCR). LOD=limit of detection of qPCR titering, as set by the titer corresponding to the Cq for which the qPCR primers alone had been observed to amplify.

Supplementary
Supplementary Figure 16. PANCE dilution schedule and titers for N2. SP containing wild-type split-dNme2Cas9, pooled ePACE1/ePACE2 split-Nme2Cas9, or pooled ePACE3 split-Nme2Cas9 were first diversified in E.coli host cells containing pJC175e 2 and MP6 2 , isolated, then seeded into PANCE2 (N2, eight chemostats of each of eight N3YTN PAMs, where Y=C or T; three replicates). (a) Passage stringency schedule and resulting titers (measured by qPCR) for replicates 1-3. Passages were done after 16-24 hr for all passages. For some passages, some conditions were passaged uniquely to others or in a different host cell line, and these changes are listed in the Notes column. Grey coloring represents titers that were not measured or the PAM had not yet been included. LOD=limit of detection of qPCR titering, as set by the titer corresponding to the Cq for which the qPCR primers alone had been observed to amplify. (b) PANCE conditions used for N2.  CTT TTA TTC TTG TTT CTA CTC CTG CTT TTA TTC TTG TTT CTA CTC CTG CTT TTA TTC TTG TTT  1  20  20 20  20 20 20  20 20  20 20 20  20 20  20 20  2  20  20 20  20 20 20  20 20  20 20 20  20 20  20 20  3 Replicate 1 dilution table  Replicate 2 dilution table  Replicate 3

dilution table
Supplementary Figure 17. Flow rate schedule and titers for ePACE5. SP from N2 replicate 3, passage 7, were combined and seeded into corresponding PAMs in ePACE5 (eight chemostats, two lagoons each of the eight N3YTN PAMs, where Y=C or T). Flow rate stringency for each PAM is shown in the plots, as are resulting titers (measured by qPCR). As most lagoons were unable to support consistent phage propagation, the timepoints used for isolating and sequencing phage (Extended Data Figure 5) are marked by a black box. LOD=limit of detection of qPCR titering, as set by the titer corresponding to the Cq for which the qPCR primers alone had been observed to amplify.   ACC-1  0  0  3  111  0  0  0  0  GCT-1  0  0  14  271  0  0  0  6  CCC-1  4  11  149  1612  0  6  6  70  CCG-1  0  1  13  168  0  0  0  12  GCA-1  0  1  25  426  0  0  0     Nme off-targets SpRY off-targets Nme off-targets SpRY off-targets Nme off-targets SpRY off-targets Nme off-targets SpRY off-targets Nme off-targets SpRY off-targets Nme off-targets SpRY off-targets target sites nominated by GUIDE-seq for eNme2-C, eNme2-C.NR, SpRY, or SpRY-HF1 nucleases (Supplementary Table 3). Off-target indel formation by Nme2Cas9, eNme2-C.NR, SpRY, or SpRY-HF1 nuclease at nominated off target sites for the sgRNAs targeting Site 3 (a), Site 4 (b), Site 5 (c), or Site 6 (d). Off-target adenine base editing by Nme2-ABE8e, eNme2-C-ABE8e, SpRY-ABE8e, or SpRY-HF1-ABE8e at nominated off-target sites for the sgRNAs targeting Site 3 (d), Site 4 (e), Site 5 (f), or Site 6 (g). Mean±SEM is shown and reflects the average activity and standard error of n=3 independent biological replicates measured at the maximally edited position within each given genomic site. On-target activity is shown at the left-most entry for each site.       TGGAGTTCAGACGTGTGCTCTTCCG  ATCTTCCCTCTGGCTTCTTTTAAGTT  TT  NR-Site4-OT3  genomic site   ACACTCTTTCCCTACACGACGCTCTT  CCGATCTNNNNAGCACAGAGATAAAA  GGACAGAA   TGGAGTTCAGACGTGTGCTCTTCCG  ATCTTCCCTCTGGCTTCTTTTAAGTT  T  NR-Site4-OT4  genomic site   ACACTCTTTCCCTACACGACGCTCTT  CCGATCTNNNNCCGGAGGTCTCTACT  TCCCA   TGGAGTTCAGACGTGTGCTCTTCCG   As IPP devices are sensitive to changes in pressure at valves and in connected media bottles, we developed an 8-channel pressure regulator that can be used to regulate these pressures through the eVOLVER framework. The device consists of sets of two proportional valves that can limit air flow from a high-pressure source and a vent at atmospheric pressure.
By connecting an electronic pressure gauge to the output of this valve configuration, it is possible to implement proportional-integral-derivative (PID) control over the valves in order to set the output pressure to any desired level between the input and atmospheric pressure. We Large pressure deviations (>0.5 psi) that can affect the performance of the devices were observed with the fixed regulator, but were successfully eliminated with our automated pressure regulator scheme. We further characterized the effects of pressure changes at various locations in the system in order to optimize performance of the IPP devices for the course of a PACE experiment (Supplementary Figure 3).

Supplementary Note 2. ePACE2 recombination and cheating
During ePACE2, evolving Nme2Cas9 variants on the SP appeared to propagate well in all lagoons on targeted PAMs (each of the eight N3YTN PAMs). Phage were sampled from some lagoons, and the insert was amplified via PCR. Following agarose gel electrophoresis, we found that these SP pools appeared to lose the expected Nme2Cas9 insert, as the resulting bands no longer corresponded to the correct insert size (Supplementary Fig. 8a).
Sanger sequencing of the incorrectly sized band revealed a region of nucleotide homology between the N-terminus of the gIII construct on the AP and gVI in the phage genome ( Supplementary Fig. 8b,c). This site of homology was likely acting as a recombination site enabling some phage to incorporate the gIII-C half into the SP genome. As gIII-N is constitutively expressed in the original SAC-PACE selection, this enables phage to propagate in a selection genomic siteree manner. For subsequent evolutions, we altered the codon usage of the N-terminus of gIII within the AP, such that the nucleotide homology was no longer present (pTPH412, Supplementary Table 7). Following this alteration, recombination was no longer observed.

Supplementary Note 3. Validation of the split base editor SAC-PACE selection
To enable control over the expression of active enzyme in the SAC-PACE selection, we split Nme2ABE8e at the linker sequence between TadABE8e and Nme2Cas9. The TadABE8e Fig. 6a, 10a). TadABE8e Fig. 3c). We note, however, that the difference in activity between the variants was nuanced, as the overall trend reflects general improvements to activity on all N4CN PAMs (Extended Data Fig. 3b).

Supplementary Note 5. Reversion analysis of eNme2-C RuvC/HNH domain mutations.
Simple reversion of the RuvC-inactivating mutation D16A in eNme2-C did not yield a robust nuclease Cas9. Upon reversion of the eight mutations in the RuvC/HNH domains and their associated linker regions to their wild-type residues, the resulting variant eNme2-C.NR had robust nuclease activity across N4CN PAM sites. However, reversion of these mutations had a detrimental effect on base editing activity, as the ABE8e version of eNme2-C.NR had a 1.8 genomic siteold reduction in adenine base editing activity relative to eNme2-C-ABE8e (Extended Data Fig. 7e). These results would suggest that some or all the mutations in the RuvC/HNH domains are important for Nme2Cas9 activities necessary for base editing, but detrimental to the subsequent activation or catalytic activity of Nme2Cas9 nuclease.
To further explore this idea and to potentially find an optimal dual base editor/nuclease To analyze BE-PPA sequenced files, the demultiplexed fastq files were filtered using the seqkit package/grep function 11 to search for two flank sequences near either end of the amplicon. For ABE-PPA profiled variants, groups of PAMs were UMI-tagged, and the specific UMI tag was used in place of one of the flank sequences. Filtered files were then binned into individual fastq files per PAM using the same function. The resulting PAM-specific fastq files were analyzed using standard CRISPResso2 12 analysis.

APs.
When designing the dual PAM split SAC-PACE APs (pTPH418b, see Supplementary