The round 1 plasmid library insert was constructed using a degenerate oligo essentially as described previously, but for a single step69. 15-cycle PCR reactions were pooled from 96 wells, PCR-purified, digested with AgeI/XbaI, and then gel-extracted from a band migrating at the expected length. Approximately 10 ng of pB1H2_UV2-ω-dCas9_UV5-sgRNA plasmid template was used per PCR reaction. The gel-purified insert DNA was ligated overnight at 16 °C with ~20 µg gel-purified backbone DNA that had been likewise digested with AgeI/XbaI (5:1 molar ratio of insert-to-backbone). Ligation reactions were then cleaned and concentrated by ethanol precipitation and split into 15 aliquots of electrocompetent US0 cells for electroporation. After recovering the pooled transformants for 1 h in a 1 L flask, an aliquot was withdrawn for serial dilution and plating to measure the CarbR and CarbRKanR CFUs, and carbenicillin was added to the culture before returning to the shaker. The CarbRKanR CFUs are presumed to result from self-ligation of singly-cut backbone DNA and uncut backbone plasmid; their counts were subtracted from total CarbR counts to estimate the final library size (6.5e7 members). ODs were periodically measured from the library culture by blanking against a similarly treated control culture inoculated in parallel with a proportional volume of electrocompetent cells that were electroporated without DNA. The round 2 plasmid libraries were constructed similarly, except that backbone DNA was derived from the K2re plasmid, and each insert was generated with 16 error-prone PCR reactions using standard oligos and the GeneMorph II (Agilent) kit. The pB1H2_UV2-ω-dCas9[KG]_UV5-sgRNA plasmid or wild-type equivalent was used as template DNA. BsaI-HFv2 (NEB) was used for digestion of both the insert and backbone DNA. Small-scale ligations were initially performed to determine the template concentrations needed to achieve a desired mutation rate for the PID region, according to guidelines described previously70, and ~10 ng per reaction was found to be optimal. Mutation rates were estimated by Sanger-sequencing of 10–12 random transformants from each ligation, isolated on rich media. The final library sizes for the mWt and mKG libraries were estimated at 4.7e8 and 4.8e8, respectively, with PID mutation rates of ~1.0% and ~1.2% per bp, respectively. In accordance with procedures used for the round 1 PID library, the randomized PAM library was built using degenerate oligos and a plasmid template, except that 5 or fewer electrocompetent cell aliquots were needed for electroporation, and transformants were selected on plates after overnight storage at 4 °C. The final library size (2.9e7) was estimated as described above for transformant pools except that KanR and KanRCarbR CFU were enumerated from each transformation aliquot and the net KanR CFU counts were then summed.