Expanding the flexibility of base editing for high-throughput genetic screens in bacteria

Sandra Gawlitt; Scott P Collins; Yanying Yu; Samuel A Blackman; Lars Barquist; Chase L Beisel

doi:10.1093/nar/gkae174

. 2024 Mar 18;52(7):4079–4097. doi: 10.1093/nar/gkae174

Expanding the flexibility of base editing for high-throughput genetic screens in bacteria

Sandra Gawlitt ^1,², Scott P Collins ^2,², Yanying Yu ³, Samuel A Blackman ⁴, Lars Barquist ^5,^6,⁷, Chase L Beisel ^8,^9,^✉

PMCID: PMC11039988 PMID: 38499498

Abstract

Genome-wide screens have become powerful tools for elucidating genotype-to-phenotype relationships in bacteria. Of the varying techniques to achieve knockout and knockdown, CRISPR base editors are emerging as promising options. However, the limited number of available, efficient target sites hampers their use for high-throughput screening. Here, we make multiple advances to enable flexible base editing as part of high-throughput genetic screening in bacteria. We first co-opt the Streptococcus canis Cas9 that exhibits more flexible protospacer-adjacent motif recognition than the traditional Streptococcus pyogenes Cas9. We then expand beyond introducing premature stop codons by mutating start codons. Next, we derive guide design rules by applying machine learning to an essentiality screen conducted in Escherichia coli. Finally, we rescue poorly edited sites by combining base editing with Cas9-induced cleavage of unedited cells, thereby enriching for intended edits. The efficiency of this dual system was validated through a conditional essentiality screen based on growth in minimal media. Overall, expanding the scope of genome-wide knockout screens with base editors could further facilitate the investigation of new gene functions and interactions in bacteria.

Graphical Abstract

Introduction

Genome-wide screens in bacteria have proven invaluable for the investigation of gene functions and interaction networks in a high-throughput manner (1). In recent years, many functional genomics studies have relied on coupling transposon-based gene perturbations with next-generation sequencing (NGS) (2). Transposon-insertion sequencing has facilitated the investigation of genes involved in a broad range of processes ranging from antibiotic resistance (3,4), virulence (5,6), metabolism and adaptation to different environmental conditions (7,8) to essential genes (3,4,9). Still, this approach is not programmable, often requiring massive libraries and being hard to adapt to query specific sets of genes.

The advent of CRISPR technologies enabled programmable, targeted and multiplexed gene perturbations as part of high-throughput screening. As one common type of perturbation, gene knockouts can be achieved with Cas9-based or Cas12a-based editing (10–12), where lethal double-strand breaks are used to enrich cells that underwent homologous recombination with a DNA repair template. The repair template encodes not only the homology arms and the desired edit but also a modification that disrupts the target site [e.g. mutation to the protospacer adjacent motif (PAM)], thus preventing target cleavage (10). Cas9-assisted editing has already been successfully used for large-scale saturation mutagenesis libraries in Escherichia coli (13,14). However, the removal of the large percentage of unedited cells reduces the total number of viable colonies and therefore the number of library members that can be subjected to a high-throughput screen (15). This problem is exacerbated in bacteria with poor recombineering or transformation efficiencies.

A separate type of CRISPR-based perturbation used for high-throughput screening is gene knockdown. This approach, also called CRISPR interference (CRISPRi), typically utilizes a catalytically dead Cas9 (dCas9) and a single-guide RNA (sgRNA) to silence target genes complementary to the guide and flanked by a PAM (16,17). The lack of a repair template and no inherent loss of viability simplify library generation and screening. Furthermore, the extent of gene silencing via CRISPRi can be tuned, enabling the investigation of essential genes (18–21). However, variable extents of silencing due to a number of factors can confound the ensuing analysis (22). These limitations have underscored the need for additional means to perform high-throughput genetic screens in bacteria.

One promising alternative for genome-wide knockout screens in bacteria is base editing. This CRISPR-based editing approach enables the irreversible conversion of one base into another in a programmable manner without inducing DNA double-strand breaks (23,24). Broadly, base editing relies on dCas9 or a nicking Cas9 fused to a deaminase enzyme. Once the base editor localizes to the target DNA, the deaminase recognizes the displaced strand and catalyzes deamination of target bases within an editing window. Cytosine deaminase base editors convert cytosine into uracil, resulting in a U·G wobble base. The host repair machinery converts the U·G into a U·A, which finally results in a T·A base pair following DNA repair or replication (23). The use of a uracil DNA glycosylase inhibitor (UGI) domain prevents the U from being converted back to the original C (25,26). By employing a Cas9 nickase, the unedited strand is nicked, initiating DNA repair using the uracil-containing strand as a template (23). So far, base editors have been developed to catalyze base transitions (C to T and A to G) (23) or base transversions (e.g. C to A and C to G) (27,28) as well as combinations thereof by combining cytidine and adenine base editor functions (29). Those specific point mutations can be used for targeted gene knockouts through the introduction of a premature stop codon. First proof-of-principle demonstrations for C-to-T base edits in bacteria have been demonstrated in E. coli (30,31) and Brucella melitensis (31). Since then, base editing has been utilized, for example, to investigate possible drug targets or antibiotic resistance genes in Staphylococcus aureus or Klebsiella pneumoniae, respectively (32,33), or for strain engineering of industrially relevant bacteria such as Corynebacterium glutamicum (34) and Pseudomonas putida (35).

Because base editing is programmable, functions without introducing DNA double-strand breaks and does not require a repair template, it has strong potential for high-throughput screens and is already widely applied in eukaryotes to investigate new gene functions (36–38). However, high-throughput screens with base editors remain underutilized in bacteria, with only one study to date describing a genome-wide screen in C. glutamicum using a cytosine base editor to reveal genes involved in toxin resistance (39). One prevailing reason is the limited number of target sites available to disrupt a target gene. Generating a gene disruption requires finding a PAM recognized by Cas9 appropriately spaced from one of the four codons that can be converted into a stop codon (40,41) such that the associated C falls within the editing window. As base editing efficiencies can vary widely (42), not all putative sites will lead to efficient disruptions. The prior base editing study in C. glutamicum began addressing this challenge by testing multiple PAM-flexible Cas9 nucleases for base editing, although only Streptococcus pyogenes Cas9 (SpCas9; NGG PAM) and SpCas9-VQR (NGA PAM) showed sufficient editing activity for a screen but had to be combined to achieve sufficient genome-wide coverage (39). Therefore, there is an outstanding need to increase the flexibility of base editing for large-scale knockout screens in bacteria (39).

In this work, we developed and applied a flexible and efficient base editing system for high-throughput genetic screens in bacteria. Using E. coli as a model, we generated a cytosine base editor derived from the Streptococcus canis Cas9 (ScCas9) exhibiting flexible PAM recognition. The resulting editor, ScBE3, was used to not only introduce premature stop codons but also mutate start codons as the basis of high-throughput screens. We identified ScBE3 target preferences and developed an enhanced screening approach by employing a two-step strategy involving editing followed by eradication of the unedited cells. Finally, we demonstrated this approach in a conditional essentiality screen that confirmed previously published genes necessary for growth in minimal media. With this enhanced and flexible screening approach, the benefits of base editing can be further exploited for genome-wide screens in bacteria and will contribute to a better understanding of their diverse genetics and physiology.

Materials and methods

Strains, plasmids and growth conditions

All strains, plasmids and primers are listed in Supplementary Tables S1–S3. Escherichia coli cells were grown in lysogeny broth (LB) (5 g/l NaCl, 5 g/l yeast extract, 10 g/l tryptone) at 37°C with shaking at 250 rpm. To maintain plasmids, the antibiotics ampicillin, chloramphenicol and/or kanamycin were added at 100, 34 and 50 μg/ml, respectively, as necessary. For screening in minimal media, M9 minimal medium (1× M9 salts, 1 mM thiamine hydrochloride, 0.4% glucose, 0.2% casamino acids, 2 mM MgSO₄, 0.1 mM CaCl₂) and MOPS minimal medium [prepared as described in (43) with 0.4% glucose] were prepared freshly prior to the screen.

Golden Gate assembly reactions

BsaI or BsmBI assembly reactions included ∼20 fmol of each plasmid or amplicon assembly fragment in equimolar amounts, in addition to the following: 1 μl of T4 DNA Ligase buffer (NEB, Cat. B0202S), 0.5 μl of BsaI or BsaI HFv2 (NEB, Cat. R3733S) or 0.5 μl BsmBI (NEB, Cat. R0580S), 0.5 μl of T7 DNA Ligase (NEB, Cat. M0318S) and nuclease-free water up to a final reaction volume of 10 μl. Reactions were incubated in thermocyclers with 20–35 cycles of [(i) 5 min of digestion at 37°C for BsaI and 2 min of digestion at 42°C for BsmBI followed by (ii) 5 min of ligation at 16°C] followed by a 30-min final digestion at 37°C or 42°C and a 10-min heat inactivation at 80°C.

Base editing in E. coli with ScBE3

The rAPOBEC1 deaminase was fused to the N-terminus of ScCas9 D10A by a 16-aa (amino acid) linker. This construct, SPC914, was co-transformed by electroporation into E. coli MG1655 alongside a constitutively expressed sgRNA targeted to produce a stop codon or disrupt a start codon initially in lacZ. The targets were flanked by the minimal NNG PAM required by ScCas9. These transformants were plated on LB with glucose and the appropriate antibiotics. Colonies were picked and inoculated into 2 ml of LB medium shaking at 37°C and 250 rpm overnight with the appropriate antibiotics. The next day, cultures were diluted 1:500 into LB supplemented with antibiotics and 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) and 0.2% l-arabinose for induction of the base editor, and cultured for 8 h. An aliquot of these cultures was plated for analysis, and another aliquot was diluted 1:500 for 16 h of further induction and culturing before plating. Base editing frequency was measured by the fraction of colonies having a white or blue color on X-gal indicator plates and/or Sanger sequencing.

Measuring GFP repression by flow cytometry

To demonstrate that the base editing system does not cause transcriptional repression in the absence of editing as part of Figure 1G, E. coli MG1655 was co-transformed with plasmids encoding (i) a GFP reporter, ScBE3 and a targeting sgRNA that induces installation of a premature stop codon; (ii) a GFP reporter, ScBE3 and an sgRNA targeting a locus without a cytidine in the appropriate editing window; (iii) a GFP reporter, ScBE3 and an sgRNA with a C-to-T mismatch, simulating prior base editing to investigate the scenario of retargeting with catalytically dead ScCas9 (ScdCas9); and (iv) a GFP reporter, dCas9 and a gfp-targeting sgRNA. A no-vector control was included in the flow cytometry experiments (v). Overnight cultures of cells harboring the above-mentioned plasmids were diluted to OD₆₀₀ = 0.01 in LB medium with the corresponding antibiotics and 1 mM IPTG and 0.2% l-arabinose to induce expression of ScBE3 and ScdCas9, respectively, and incubated for 16 h at 37°C while shaking at 250 rpm. The next morning, cultures were diluted 1:50 in fresh LB medium supplemented with the corresponding antibiotics and inducers and incubated at 37°C, while shaking at 250 rpm until reaching OD₆₀₀ ∼ 0.4. A control without inducers was also included.

For the flow cytometry measurements, cultures were diluted 1:20 in 1× phosphate-buffered saline and analyzed on an Accuri C6 flow cytometer with C6 sampler plate loader (Becton Dickinson) equipped with CFlow plate sampler, a 488 nm laser and a 530 ± 15 nm bandpass filter. Briefly, forward scatter (cutoff of 12 000) and side scatter (cutoff of 600) were used to eliminate noncellular events. The mean value within FL1-H of at least 30 000 events within a gate set for E. coli was used for data analysis. Three individual colonies were picked and inoculated to be analyzed on the flow cytometer in the same run, and their standard deviation was indicated as the error bar.

To compare gene repression efficiencies between ScdCas9 and SpdCas9 as part of Supplementary Figure S2, E. coli MG1655 was co-transformed with the ScdCas9 or SpdCas9, a GFP reporter and an sgRNA targeting the degfp sequence flanked by an NGG PAM (sgRNAs c and d, Supplementary Table S4). The cultivation procedure with the induction of ScdCas9 or SpdCas9 and the subsequent flow cytometric measurement was performed similarly to the experiment in Figure 1G.

GFP repression assay

GFP fold repression was calculated using the mean GFP fluorescence values. The value for the nontargeting guide was divided by that for the targeting guide to be tested, after subtracting the autofluorescence value of the no-vector control from both.

β-Galactosidase activity assay

To quantify the β-galactosidase activity, E. coli MG1655 was co-transformed with the ScBE3 plasmid and an sgRNA plasmid targeted to the lacZ gene, which would either mutate the start codon or introduce a premature stop codon. As a control, a nontargeting sgRNA was included. Base editing was induced according to the procedure described in the ‘Materials and methods’ section. After the base editing step, 100 μl of the 1:1000 diluted overnight culture was first plated on LB plates supplemented with ampicillin and chloramphenicol and incubated overnight at 37°C. The next day, colonies were screened for the intended edit by colony polymerase chain reaction (cPCR) and subsequent Sanger sequencing. Three single colonies, with the verified edit, were used to inoculate 5 ml LB medium supplemented with the appropriate antibiotics and 1 mM IPTG.

For the β-galactosidase assay, the overnight cultures were diluted to OD₆₀₀ = 0.1 in fresh LB medium containing the appropriate antibiotics and 1 mM IPTG and incubated at 37°C, while shaking at 250 rpm until reaching OD₆₀₀ of ∼0.5. Two milliliters of each culture was collected by centrifugation at 4°C, the supernatant was discarded and the pellet was stored on ice until all samples were collected. Next, the cell pellets were resuspended in cold Z-buffer (60 mM Na₂HPO₄·7H₂O, 38.7 mM NaH₂PO₄·H₂O, 10 mM KCl, 1 mM MgSO₄) with freshly added β-mercaptoethanol (34.5 mM), so that the cell suspension was 1 OD₆₀₀/ml. One hundred fifty microliters of chloroform and 100 μl of 0.1% sodium dodecyl sulfate was added to each tube, the tube was vortexed for 15 s and put on ice. Dilutions (1:2, 1:10) were prepared from each sample by mixing the specific volumes of the sample (upper phase) with Z-buffer. Two hundred microliters of each sample and dilution was loaded onto a 96-well plate. The spectrophotometer was set to measure at OD₄₂₀ and OD₆₀₀ every 2 min for the duration of 45 min and preheated to 28°C. Lastly, 50 μl ortho-nitrophenyl-β-galactoside (4 mg/ml in Z-buffer without β-mercaptoethanol) was added to each well using a multichannel pipette and the measurement was started immediately. The assay was performed in triplicates, using single colonies to start the overnight cultures. The β-galactosidase activity was calculated as described in the Miller assay protocol (44). The original equation was modified to incorporate the slope (ΔABS₄₂₀/Δt [min]) obtained from linear regression analyses, resulting in 1000 × {(ΔABS₄₂₀/Δt [min])/(V [ml] × ABS₆₀₀)}.

Two-step base editing and cell killing

Base editing was performed as previously, except that base editing was induced for a single time period of 16 h, back-diluted 1:500 into LB medium with antibiotics, 1 mM IPTG and 0.2% l-arabinose, shaking at 250 rpm. The culture was then back-diluted 1:100 into LB medium with antibiotics only. This was followed by a dilution of 1:100 into LB medium with antibiotics and 125 ng/ml anhydrotetracycline (aTc) for induction of ScCas9++ and cell killing. The resulting cultures were either plated for analysis or back-diluted into minimal media for selection of nonessential genes.

Library design for genome-wide screening in E. coli

The reference genome and annotation of E. coli K12 MG1655 (NCBI: NC_000913.3) were used for sgRNA library design. We first located potential target codons—the start codons ATG, TGG, CAA, CAG and CGA—in the first half of each coding region and searched for NNG PAM given the defined activity window from positions 5 to 8 from the 5′ end of sgRNA. sgRNAs (with flanking primer sequences) containing homopolymer stretches consisting of more than four consecutive nucleotides or BsmBI restriction sites were removed. For genes with >15 sgRNAs, a maximum of 15 sgRNAs per gene were included. Consequently, excess sgRNAs were filtered sequentially based on the number of potential off-targets and GC content until no more than 15 sgRNAs per gene were included. In the first four steps, gRNAs with off-targets harboring zero to three mismatches were removed. In the last step, sgRNAs with GC content outside of the range of 30–85 were removed. If all sgRNAs were removed in one step, the first 15 sgRNAs based on the target position in the coding sequence from the previous step were kept. The minimum number of guides per gene was 1, and the average number of guides per gene was 9. This procedure resulted in a library with 37 762 sgRNAs, including 37 362 guides covering 4084 genes and 400 randomized nontargeting guides. Potential off-targets were evaluated using SeqMap (version 1.0.12) (45).

Library cloning and verification for genome-wide screening in E. coli

The library of guides was assembled by Golden Gate BsmBI assembly in two 25 μl reactions having 150 nM of insert library and 50 nM of backbone plasmid. Assembly reactions were performed with 50 cycles of digestion at 42°C for 3 min and ligation at 16°C for 5 min, followed by a final digestion at 42°C for 60 min. After assembly, the cloning reactions were ethanol precipitated and transformed into E. coli by 12 separate electroporation reactions. Electroporated cells were recovered for 1 h in 50 ml of LB medium. A small aliquot was dilution plated on LB agar with ampicillin to confirm >50-fold library coverage. The remaining liquid culture was supplemented with ampicillin to select for library transformants. This 50 ml culture was grown until the early stationary phase, shaking at 37°C and 250 rpm, before maxi-prep DNA purification. The cloned insert library was amplified and sequenced by Illumina NGS on an iSeq 100, confirming that 98.7% of inserts were present at the target 50-fold threshold.

Genome-wide screening in E. coli

We then transformed the base editor ScBE3 (SPC914) [or the evoAPOBEC1-ScnCas9-UGI (SPC1217)] into E. coli MG1655, followed by electroporation of 1 μg of sgRNA plasmid library DNA to produce the library with base editor. To screen for gene essentiality, induction of base editing and selection of nonessential genes occurred in tandem. All culturing occurred shaking at 37°C and 250 rpm, in LB medium with antibiotics, 1 mM IPTG and 0.2% l-arabinose. The dilution and culturing series was as follows: 1:100 with 4 h of culturing, 1:100 with 4 h of culturing and 1:500 with 16 h of culturing. Cultures were sampled and plasmids isolated for sequencing. Sequencing libraries were prepared by amplifying the sgRNA sequence with primers adding the Illumina sequence adaptors. Five forward primers were used, with the terminal end staggered by one base pair so as to add to sequencing library complexity. The resulting amplicons were then further amplified to add Illumina H5XX and H7XX indices for sample identification (Supplementary Table S3). Sequencing was performed on an Illumina NextSeq system, 400M 75-bp single-end reads. The experiment was performed in duplicate starting from two independent transformations of MG1655 with the plasmid library.

Library design for screening conditionally essential genes

The library used for screening E. coli genes essential for growth in M9 and MOPS minimal media contains plasmids encoding fixed pairs of sgRNAs that guide ScBE3 to its target locus to perform C-to-T edits and, in case of unsuccessful editing, guides ScCas9++ to cleave the unedited DNA. Note that the target C within the ScBE3 editing window is part of the intact ScCas9++ PAM (NNG) encoded on the opposite strand and the guide sequences for ScCas9++ were extracted based on the editing base of ScBE3 guides, so that only cells with edited DNA will survive.

The dual-sgRNA library targets in total 298 E. coli genes, with 119 genes essential for growth in M9 and MOPS minimal media and 179 nonessential genes. For the essential genes, all sgRNAs following the design rules in the essentiality screen were included, ignoring the maximum limit. The nonessential genes were selected from genes with median log₂ fold change (logFC) higher than −0.5 between the last (after 24 h of culturing) and initial time points in the essentiality screen to minimize their differential abundance and maximize the number of sgRNAs. Similar to the previous library, to reduce the risk of off-targeting, for genes with >15 possible guides, those that had potential off-target binding sites with up to three mismatches were discarded. Nontargeting guides were ranked on the logFC values in each time point in the essentiality screen. The minimal sum and variance of the rankings were used to select from the 400 nontargeting guides. The number of guides per gene ranged between 1 and 43 with an average of 13 guides per gene. The resulting library comprises 3963 guide pairs used by ScBE3 and ScCas9++ with 224 guides targeting essential genes for growth in M9 minimal medium, 111 guides targeting essential genes for growth in MOPS minimal medium, 1159 guides targeting essential genes for growth in M9 and MOPS minimal media, 2369 guides targeting nonessential genes and 100 nontargeting guides. To avoid complications with repetitive elements during library amplification, the library cloning was performed in two steps. The first step included integration of the amplified library fragment into the pSG255 backbone, and the second step included integration of a second PCR fragment into the plasmid obtained after the first cloning step. The library oligo pool, synthesized by Twist Bioscience, encodes the following elements in this order: a 20-nt ScBE3 guide sequence, a BsaI dropout sequence and a 20-nt ScCas9++ guide sequence. Additionally, flanking universal primer binding sites for library amplification and BsmBI restriction sites for subsequent digestion of the fragment and ligation into the backbone were added. The backbone pSG255 encodes for a GFP dropout sequence flanked by BsmBI restriction sites for the integration of the amplified library as well as a PBAD promoter controlling expression of the base editing sgRNA and an sgRNA scaffold for the killing guide. The PCR fragment for the second cloning step was amplified from SPC1995. It encodes the sgRNA scaffold for the base editing guide, a PTet promoter controlling expression of the killing sgRNA and BsaI restriction sites for digestion and ligation into the plasmid obtained from the first cloning step.

Library cloning and verification for screening conditionally essential genes

The library was amplified with the KAPA HiFi HotStart Library Amplification Kit for Illumina^® platforms (Roche) using 5 ng DNA for 13 cycles following the manufacturer’s instructions (T_a = 68°C; 20 s denaturation, 15 s annealing, 20 s extension). Supplementary Table S3 contains the specific oligonucleotides used for the library amplification.

For the first cloning step, we set up two separate Golden Gate reactions by mixing 400 fmol of the amplified library, 20 fmol of backbone pSG255 (both ethanol precipitated beforehand), 1 μl of T4 DNA ligase (400 units), 1 μl of BsmBI-v2 (10 units) and 2 μl of 10× T4 ligation buffer, and adding water to reach a total volume of 20 μl. A thermocycler was used to perform 35 cycles of digestion (42°C for 2 min) and ligation (16°C for 5 min), followed by a final digestion (60°C for 10 min) and a heat inactivation step (80°C for 10 min). The reactions were pooled, ethanol precipitated and 1 μg was transformed into 100 μl of fresh electrocompetent NEB stable cells (New England Biolabs). The transformation was conducted with two separate batches of electrocompetent cells and pooled after the recovery step to ensure that enough transformants are obtained. After recovery in 900 μl of SOC at 37°C for 1 h with shaking at 250 rpm, different dilutions of the recovered cells were plated on LB agar supplemented with ampicillin and incubated for 16 h to check the number and color of the resulting colonies (ensuring an ∼2400× coverage). The rest of the recovered culture was added to 100 ml of LB medium supplemented with ampicillin and incubated for 16 h at 37°C with shaking at 250 rpm. Cells were harvested by centrifugation and subjected to plasmid (pSG256) extraction. Sanger sequencing was used to validate the correct assembly of the plasmid.

For the second cloning step, a PCR fragment was amplified from SPC1995 using primers listed in Supplementary Table S3. The new PCR fragment and the library encoding plasmid pSG256, obtained from the first cloning step, were digested with BsaI-HFv2 for 2 h at 37°C, followed by purification from agarose gels (NucleoSpin Gel and PCR Clean‐up Kit, Macherey-Nagel) and ethanol precipitation. Three separate Golden Gate reactions were set up by mixing 400 fmol digested PCR fragment, 20 fmol digested pSG256, 1 μl of T4 DNA ligase (400 units), 1 μl of BsaIHF-v2 (20 units) and 2 μl of 10× T4 ligation buffer, and adding water to reach a total volume of 20 μl. A thermocycler was used to perform 35 cycles of digestion (37°C for 5 min) and ligation (16°C for 10 min), followed by a final digestion step (60°C for 10 min) and a heat inactivation step (80°C for 10 min). The reactions were pooled, ethanol precipitated and 2 μg was transformed into 100 μl of fresh electrocompetent NEB stable cells (New England Biolabs). The transformation was conducted with two separate batches of electrocompetent cells and pooled after the recovery step to ensure that enough transformants are obtained. After recovery in 900 μl of SOC at 37°C for 1 h with shaking at 250 rpm, different dilutions of the recovered cells were plated on LB supplemented with ampicillin and incubated for 16 h to check the number of the resulting colonies (ensuring an ∼100× coverage). The rest of the recovered culture was added to 100 ml of LB medium supplemented with ampicillin and incubated for 16 h at 37°C with shaking at 250 rpm. Cells were harvested by centrifugation and subjected to plasmid (pSG257) extraction. Sanger sequencing was used to validate the final library plasmid DNA.

Screening of conditionally essential E. coli genes

Escherichia coli MG1655 was initially transformed with the ScBE3 and ScCas9++ encoding plasmids (2.0 kV, 200 Ω and 25 μF). The resulting strain SG645 was then transformed with the dual-sgRNA library by electroporation and recovered in 900 μl of SOC for 1 h at 37°C with shaking at 250 rpm. For representation of the initial guide distribution in the library, 500 μl of the recovered cell suspension was sampled and the cell pellet was stored at −20°C for later DNA extraction by miniprep (NucleoSpin Plasmid, Macherey-Nagel). Different dilutions of the recovered cell suspension were plated on LB agar containing the appropriate antibiotics and incubated for 16 h to determine the library coverage (∼11 000× and ∼7500× coverage for replicates 1 and 2, respectively). Then, the recovered culture was diluted to OD₆₀₀ = 0.01 in LB medium with appropriate antibiotics and incubated at 37°C with shaking for 16 h. The next day, 5 ml of the culture was sampled for subsequent library DNA extraction and several 1 ml cryo-cultures were prepared [70% bacterial culture mixed with 30% of 50% (v/v) glycerol] and stored at −80°C for potential repetitions of the screening procedure.

To initiate the base editing step, we diluted the overnight cultures grown in LB to OD₆₀₀ = 0.01 in fresh LB medium containing appropriate antibiotics, 1 mM IPTG and 2% (w/v) l-arabinose to induce the expression of ScBE3 and the matching base editing sgRNA. The culture was incubated for 16 h at 37°C with shaking at 250 rpm. The next day, 5 ml of the culture was sampled for subsequent library DNA extraction, representing the guide distribution after the base editing step. In a recovery step, the culture was diluted to OD₆₀₀ = 0.1 in LB containing the appropriate antibiotics and grown until OD₆₀₀ = 0.5–0.6 at 37°C with shaking at 250 rpm. For initiating the killing step, the previous culture was diluted to OD₆₀₀ = 0.1 in LB with appropriate antibiotics and 125 ng/ml aTc and grown until reaching stationary phase (OD₆₀₀ ∼ 3). Five milliliters of the culture was sampled for library DNA extraction, representing the guide distribution after the killing step.

For the screen of conditionally essential E. coli genes in M9 and MOPS minimal media, the previous culture was diluted to OD₆₀₀ = 0.01 in 100 ml of M9 or MOPS minimal medium, with a prior washing step to eliminate traces from the previously used LB medium, and incubated for 18 and 20 h, respectively, at 37°C with shaking at 250 rpm. Lastly, 5 ml for each growth condition was sampled for library DNA extraction. The experiment was performed in duplicate starting from two independent transformations of MG1655 with the plasmid library.

Library sequencing

The sequencing library was generated using the KAPA HiFi HotStart Library Amplification Kit for Illumina^® platforms (Roche) and the primers listed in Supplementary Table S3. With the first PCR, the sequencing primer binding sites were added, and with the second PCR, the unique indices and flow cell-binding sequence were added. The amplicons of the first and second PCR reactions were purified using solid-phase reversible immobilization beads (AMPure XP, Beckman Coulter) following the manufacturer’s instructions to remove excess primers and possible primer dimers. The sequencing library samples, with the required DNA concentrations (100 pg to 200 ng in a total volume of 10 μl), were submitted to the HZI NGS sequencing facility (Braunschweig, Germany) for paired-end 2 × 150-bp NGS with 4M reads per sample on a NovaSeq 6000 sequencer.

Sequencing data processing and machine learning models

Sequence reads with a perfect match were mapped to the sgRNA library using an in-house Python script. sgRNAs were first filtered by 1 count per million in a minimum of two samples in the essentiality screen and 5 counts per million in a minimum of four samples in the screen based on the dual system. Read counts for each sgRNA were normalized by factors calculated from nontargeting guides using the trimmed mean of M-values method in edgeR (version 3.28.0) (46) with default settings. RUVs in RUVSeq (version 1.20.0) (47) was used in the analysis of the essentiality screen to remove batch effects between the two replicate experiments with k equal to 1 and nontargeting guides as control. Differential abundance (logFC) of sgRNAs between time points was calculated using the quasi-likelihood F-test after fitting a generalized linear model using glmQLFTest and glmQLFit functions in edgeR with design matrix incorporating the factors of unwanted variation from RUVSeq.

The machine learning model was developed with 556 sequence features using auto-sklearn (version 0.14.6) (48), similarly to that described previously (49). Sequence features include the one-hot encoded 20-nt target and 3-nt PAM, and dinucleotide features of the 30-mer extended sequences from 4 nt upstream of sgRNA to 3 nt downstream of the PAM. For the single nucleotide features, there are four possibilities for each position within the 23-nt sequence. For the dinucleotide features, there are 16 possibilities for each of the 29 dinucleotide positions within the extended 30-nt sequence, resulting in a total of 556 features (23 × 4 + 29 × 16). Two thousand eight hundred thirteen sgRNAs targeting 307 essential genes were included for the essentiality screen, while 1494 sgRNAs targeting 119 genes essential in either M9 or MOPS minimal medium were included for the conditional essentiality screen. Essential gene lists in the LB Lennox medium or M9 minimal medium were obtained from EcoCyc (50). For MOPS minimal medium, essential genes were defined by OD₆₀₀ < 0.08 after culturing for 48 h (51). The logFC values between the the last and initial time points for ScCas9 nickase-derived base editor (ScBE3) were used as training targets in the essentiality screen, while the logFC values between post-killing and post-screening samples were used in the conditional essentiality screen. For auto-sklearn, the AutoSklearnRegressor function was used and all possible estimators were included. The following parameters were used: ‘time_left_for_this_task’ = 3600, ‘per_run_time_limit’ = 360, ‘resampling_strategy’ = ‘cv’, ensemble_kwargs = {‘ensemble_size’: 1}, ‘resampling_strategy_arguments’ = {‘fold’: 5}, ‘metric’ = autosklearn.metrics.r2, include = {‘feature_preprocessor’ : [‘no_preprocessing’]}. Feature types are categorical. The selected histogram-based gradient boosting model was saved and used with scikit-learn (version 0.24.2) (52) for downstream analysis. The training and test sets were split guide-wise based on unique sgRNA sequences. Ten-fold cross-validation was used to evaluate model performance. For each iteration, the Spearman correlation between measured depletion values and predicted values for all test samples was calculated. The histogram-based gradient boosting model was interpreted using the ‘shap_values’ function in TreeSHAP (version 0.39.0) (53) and all samples.

Growth experiments

To evaluate the growth of E. coli overexpressing ScBE3 as part of Supplementary Figure S1, E. coli MG1655 was co-transformed with the ScBE3 plasmid and a nontargeting sgRNA plasmid (Supplementary Table S2). The wild-type (WT) MG1655 strain served as a positive growth control. Four individual colonies of each strain were used to inoculate 5 ml LB medium supplemented with the appropriate antibiotics and incubated for 16 h (overnight) at 37°C, while shaking at 250 rpm. The next morning, cultures were diluted to OD₆₀₀ = 0.02 in 200 μl of fresh LB medium on a 96-well plate supplemented with the appropriate antibiotics and with or without 1 mM IPTG or 0.2% l-arabinose to induce ScBE3 expression. The plate was incubated in the BioTek Synergy H1 plate reader at 37°C with shaking, reading the OD₆₀₀ every 5 min for 24 h.

To evaluate the growth of E. coli mutants as part of Supplementary Figure S9, E. coli MG1655 was co-transformed with the SpCas9++ plasmid, the ScBE3 plasmid and a dual-sgRNA plasmid encoding for the base editing and killing guide (Supplementary Table S2). The resulting transformants were recovered in SOC medium for 1 h at 37°C, while shaking at 250 rpm and subsequently plated on LB plates supplemented with ampicillin, kanamycin and chloramphenicol. Three individual single clones per strain were used to inoculate 5 ml LB medium supplemented with the appropriate antibiotics and grown for 8 h at 37°C, while shaking at 250 rpm. To induce the base editing, the cultures were diluted to OD₆₀₀ = 0.01 in fresh LB medium supplemented with the appropriate antibiotics, 1 mM IPTG and 0.2% l-arabinose and incubated for 16 h (overnight) at 37°C, while shaking at 250 rpm. After the base editing step, 100 μl of the 1:1000 diluted overnight culture was plated on LB plates supplemented with the appropriate antibiotics and incubated for 16 h (overnight) at 37°C. The next day, colonies were screened by cPCR and Sanger sequencing for the intended edits (mutated start codon or premature stop codon). Verified clones were selected to inoculate 5 ml of LB with appropriate antibiotics and incubated for 16 h (overnight) at 37°C, while shaking at 250 rpm. The next morning, 2 × 1 ml of the culture was collected by centrifugation at 5000 × g for 2 min. The pellets were washed twice in M9 or MOPS minimal medium, respectively. Finally, the pellets were resuspended in either M9 or MOPS minimal medium and used to inoculate 200 μl of each minimal medium on a 96-well plate to a final OD₆₀₀ = 0.01. The plate was incubated in the BioTek Synergy H1 plate reader at 37°C with shaking, reading the OD₆₀₀ every 5 min for 24 h.

Evaluation of the editing efficiency and bystander edits with ScBE3

Escherichia coli MG1655 was co-transformed with the SpCas9++ plasmid, the ScBE3 plasmid and a dual-sgRNA plasmid targeting a nonessential gene from the target pool of the conditionally essential genes (Supplementary Table S10). The base editing procedure was performed as described earlier, except that cultures were induced a single time for 16 h and directly plated after the base editing step without performing the killing step. One hundred microliters of the 1:1000 dilution was plated on LB plates with the appropriate antibiotics and incubated at 37°C overnight. The next day, colonies were screened by PCR amplifying the target region and subsequent Sanger sequencing using the oligos from Supplementary Table S3.

Investigating dual usage of the sgRNAs in the dual system

Escherichia coli MG1655 was co-transformed with the SpCas9++ plasmid, the ScBE3 plasmid and a dual-sgRNA plasmid. One dual-sgRNA construct encoded an editing guide and a killing guide targeting the lacZ target #4 and is fully complementary to the target. The second dual-sgRNA construct encoded an editing guide that coded for a C-to-T mismatch, simulating prior editing, and a nontargeting killing guide (Supplementary Table S4). The base editing procedure was performed as previously described, with base editing induced for 16 h. The next morning, 50 μl of a 1:10 000 dilution of the cultures was plated on LB plates supplemented with the appropriate antibiotics and X-gal to determine the number of colony forming units and the number of white colonies compared to blue colonies, with white colonies reflecting the edited cells (disruption of lacZ by installation of a premature stop codon). The culture was then diluted 1:1000 in fresh LB medium with the appropriate antibiotics and ±0.2% (w/v) glucose, with prior washing of the cells to remove traces of the inducers from the base editing step. After a 6-h recovery step at 37°C and shaking at 250 rpm, the cultures were diluted to an OD₆₀₀ = 0.1 in LB medium supplemented with the appropriate antibiotics, ±0.2% (w/v) glucose and 125 ng/ml aTc to initiate counterselection by ScCas9++. For the counterselection, cultures were incubated at 37°C with shaking at 250 rpm until an OD₆₀₀ ∼ 1.5 was reached. Fifty microliters of the 1:1000 and 1:10 000 dilutions were plated on LB plates supplemented with the corresponding antibiotics and X-gal to determine the number of colony forming units and the number of white colonies compared to blue colonies (Supplementary Figure S7). Three individual colonies per strain served as biological replicates.

Statistical analyses

Statistical analyses of CFU (colony forming unit) fold changes and editing efficiencies were performed using a Welch’s t-test assuming unequal variances. To compare CFUs, CFU/ml values were log transformed prior to the statistical analysis. For the GFP repression assay, a one-sample t-test was performed to compare the mean value of the sample with a hypothetical mean of zero. P > 0.05 is shown as ns, P < 0.05 is shown as *, P < 0.01 is shown as **, P < 0.001 is shown as *** and P < 0.0001 is shown as ^****.

Results

An ScCas9 cytidine base editor enables flexible introduction of premature stop codons

Given the challenges of implementing base editors for high-throughput screens in bacteria, we sought to find ways to make base editing more flexible. An important starting point is relaxing PAM requirements for the base editor, which would increase the number of targetable sites. We therefore replaced SpCas9 (canonical PAM of 5′-NGG-3′) with the more flexible ScCas9 (canonical PAM of 5′-NNG-3′) (54). The resulting base editor (ScBE3), comprising the ScCas9 nickase fused to the APOBEC cytidine deaminase and a UGI domain (Figure 1A), converts a cytosine within the editing window of the displaced strand into a uracil, which is subsequently converted into a thymine through DNA replication or repair (Figure 1B). Gene disruption is normally achieved by targeting cytidines within codons that become premature stop codons when converted to thymine (Figure 1C).

To determine the efficiency of gene disruption using different PAMs with ScBE3, we targeted different locations in the E. coli lacZ gene encoding β-galactosidase to introduce premature stop codons. An introduced premature stop codon would lead to white rather than blue colonies on LB plates supplemented with the β-galactosidase substrate X-gal, after inducing cells with the lacZ inducer IPTG (Figure 1D). We specifically targeted sites flanked by 5′-NNG-3′ as well as the noncanonical 5′-NAA-3′ PAM (Supplementary Table S4). Correct edits were confirmed by Sanger sequencing of three individual white colonies per tested guide. ScBE3 yielded robust lacZ disruption for all tested guides having the canonical 5′-NNG-3′ PAM (95% editing efficiency after 24 h of editor induction). The 5′-NAG-3′ PAM yielded the most rapid editing, reaching 100% lacZ disruption after 8 h of induction for all four guides tested. Additionally, we observed gene disruption for the noncanonical NAA PAM (∼80% editing after 24 h of editor induction) (Figure 1E). In all cases, increasing the number of cellular generations through an additional subinoculation into inducing medium resulted in an increased fraction of colonies with a disrupted lacZ gene (Figure 1E). Overexpressing ScBE3 did not lead to a noticeable growth defect (Supplementary Figure S1), even though such a defect was previously associated with overexpressing SpdCas9 (55). In sum, ScBE3 can generate premature stop codons through 5′-NNG-3′ and 5′-NAA-3′ PAMs, expanding the flexibility of targeting for gene knockout screens.

To further characterize editing with ScBE3, we analyzed the editing efficiency for five nonessential E. coli genes (msrQ, ygNN, yihQ, ypjA, purK) (Supplementary Table S7). We observed varying editing efficiencies (ranging from 0% to 100%) depending on the guide and target gene. In some cases, we also observed mixed colonies presumably due to editing occurring after the single cell began to expand to a colony (Supplementary Figure S2). We also evaluated the extent of bystander editing when multiple cytosines fall within the editing window (56). For two evaluated guides targeting msrQ, we observed bystander edits (Supplementary Figure S3). These bystander edits coincided with the intended edit in some cases, which would still yield gene disruption. However, the presence of a bystander edit alone could block targeting, potentially reducing the overall frequency of a desired edit.

Targeting with ScBE3 does not cause notable repression of transcription

ScBE3 binds target DNA and nicks the targeted strand as part of the editing process, which could lead to transcriptional repression or disruption separate from the impact of the intended edit. To our knowledge, such an effect has not been investigated with base editors. To determine the extent to which these factors contribute to gene disruption, we evaluated expression of a plasmid-expressed deGFP reporter in E. coli when individually targeted by five sgRNAs and ScBE3. We also included ScdCas9 as a repression control. One guide (a) targeted the template strand with a cytidine in the ScBE3 target window, which produces a premature stop codon. Three guides (b–d) were designed to target the nontemplate strand, without a cytidine in the editing window of ScBE3. We additionally included a version of sgRNAs b and c encoding a C-to-T mismatch (b-MM and c-MM) that simulates prior base editing to test the ability of ScdCas9 to retarget the already edited DNA (Figure 1F). Targeting the nontemplate strand of an ORF with an SpdCas9 typically blocks transcription elongation of the RNA polymerase (16), which may also be induced by ScBE3. A final guide served as a nontargeting control.

Using flow cytometry analysis, we determined the extent of deGFP repression in comparison to a nontargeting control under induced (1mM IPTG and 0.2% l-arabinose) and uninduced conditions. When inducing expression of ScBE3 or ScdCas9, we found that ScdCas9 yielded modest degfp repression in comparison to a nontarget control (fold repression = 1.4–1.7 for sgRNAs a–d) (Figure 1G). This result was surprising, as SpdCas9 has been shown to lead to considerable gene repression, albeit depending on the target site (57). We therefore performed an additional comparison between ScdCas9 and SpdCas9 using sgRNAs c and d targeting sites flanked by an NGG PAM recognized by both nucleases (Supplementary Figure S4A). We observed substantially higher gene repression with SpdCas9 (fold repression = 12.4 for sgRNA c and 18.8 for sgRNA d) compared to ScdCas9 (fold repression = 0.93 for sgRNA c and 1.14 for sgRNA d) (Supplementary Figure S4B), suggesting that ScdCas9 is less efficient in gene repression compared to SpdCas9. The presence of an intended edit, which was achieved with sgRNAs encoding a C-to-T mismatch, led to negligible to modest gene repression by ScdCas9 (fold repression = 2.5 for sgRNA b-MM and 0.89 for sgRNA c-MM) (Figure 1G). Targeting ScBE3 to a region lacking an editable cytosine, which should yield nicking, also resulted in only modest gene repression (fold repression = 1.3 for all sgRNAs) (Figure 1G). As expected, targeting with ScBE3 and a guide for the site with a cytosine in the editing window (sgRNA a) led to considerably higher degfp repression compared to sgRNAs lacking any editable cytosine or when using ScdCas9 (fold repression = 28) (Figure 1G). Under uninduced conditions, ScBE3 combined with the sgRNA that introduces a stop codon (sgRNA a) still yielded notable but reduced degfp repression (fold repression = 28 for induced and 12.2 for uninduced). All other combinations of either ScBE3 or ScdCas9 with sgRNAs b and c, b-MM and c-MM yielded only modest gene repression compared to a nontargeting control (fold repression of 1–1.4) (Figure 1G). In sum, leaky expression of ScdCas9 and ScBE3 can contribute to editing, although DNA binding and/or nicking with ScBE3 for the WT or edited target should not lead to strong gene repression.

Mutating the start codon serves as an alternative strategy for gene disruption

Beyond adopting a Cas nuclease with more flexible PAM recognition, we explored a separate means of increasing the number of targetable sites for disruption-based screens: mutating start codons. Gene disruption by start codon mutations has been demonstrated successfully in human cells and rabbit embryos (58) and more recently in the bacterium C. glutamicum (39), although it remains to be explored in E. coli. By editing the C opposite of G in the standard ATG start codon, the base editor would convert the codon into ATC that initiates translation weakly at best (0.007–3% of ATG) (59) (Figure 2A). Furthermore, the same approach could be applied to the alternative start codons GTG and TTG, which appear in 14% and 3% of all genes in E. coli, respectively (59).

Figure 2. — Mutating the start codon serves as an alternative strategy for gene disruption with base editors in bacteria. (A) Nucleobase context of ScBE3 target cytosines that lead to a mutated start codon. (B) Quantification of the β-galactosidase activity based on Miller units (see the ‘Materials and methods’ section) for base editing guides that disrupt start codons and introduce a premature stop codon, respectively. The bars and vertical lines represent the mean and standard deviation from three individual biological replicates, respectively. ****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05; ns: P > 0.05. (C) Target sites and edited bases within the *E. coli lacZ* gene. Sanger traces of the WT sequence are depicted for comparison. The guides are colored in blue and labeled as "target," while the PAM sequence flanking the spacer site is colored in yellow and labeled as "PAM."

To compare the impact of introducing a premature stop codon and mutating the start codon, we designed two different ScBE3 sgRNAs targeting lacZ. The lacZ gene initiates translation at an ATG, although it possesses another ATG as the third codon that could potentially also initiate translation (60). Fortunately, one designed guide contained both associated cytidines within the editing window, allowing us to edit both cytidines simultaneously. A second guide was designed to target the most upstream codon in lacZ that could be converted into a stop codon. After co-transforming E. coli with plasmids encoding ScBE3 and an sgRNA, we induced base editing and subsequently performed a β-galactosidase activity assay with colonies carrying the intended edits. No expression could be detected for either strategy (Figure 2B), indicating efficient editing and loss of β-galactosidase expression. The introduced base edits were confirmed by Sanger sequencing of the respective locus (Figure 2C). Mutating the start codon can thus efficiently disrupt gene expression and increase the number of possible target sites as part of base editing screens.

A genome-wide essentiality screen in E. coli reveals target preferences of ScBE3

Utilizing a Cas9 with more flexible PAM recognition and incorporating start codons as potential targets expand the targeting options for base editing screens. At the same time, gaining insights into which sites yield more efficient editing and gene disruption could further support high-throughput screening. The possibility exists as well that a bad-seed effect observed with CRISPRi (22,61) may also apply to base editors. Therefore, we performed a genome-wide essentiality screen in E. coli. Using the ScCas9-derived cytosine base editors, we could mutate the start codon or introduce premature stop codons in 4084 genes, or 96.3% of all genes in E. coli. In contrast, only 3237 genes, or 76.3% of all genes in E. coli, could be targeted through the same approach with an SpCas9-derived cytosine base editor. We designed a library of 37 362 guides mediating either the introduction of a premature stop codon or disruption of the start codon. A maximum of 15 guides per gene were selected, while an additional 400 nontargeting guides were included as normalization controls (Supplementary Tables S8 and S9). To reduce the risk of off-targeting, we discarded guides predicted to bind off-targets with up to three mismatches, albeit only for genes with >15 possible guides to ensure minimal coverage of each gene (see the ‘Materials and methods’ section).

To perform the screen, E. coli harboring the ScBE3 plasmid was transformed with the pooled guide library plasmids, with a theoretical library coverage of 92-fold. Following induction, the induced culture was back-diluted multiple times over 24 h to increase the number of cell doublings to better differentiate different strengths of growth defects (62). The initial transformed library as well as the extracted library after 4, 8 and 24 h post-induction were analyzed by NGS (Figure 3A). The sgRNA guide sequence served as a barcode to quantify abundance and then measure depletion of any library member due to a growth defect induced by the sgRNA.

Figure 3. — A genome-wide base editor essentiality screen in *E. coli* reveals target preferences of ScBE3. (A) Schematic of the genome-wide gene essentiality screen in *E. coli* using ScBE3. (B) Comparison of guide depletion between essential and nonessential gene targets at different time points of ScBE3 expression. The logFC of sgRNAs targeting essential or nonessential genes or including a nontargeting control were compared based on the difference between time points 0 and 4, 8 or 24 h. Boxplots: Centerline represents the median, the boxes indicate first and third quartiles, respectively, while the whiskers are 1.5 times the interquartile range above or below the edges of the boxes. The logFC values are based on the number of guides for each group (395, 34 196 and 2809 for NT, nonessential and essential target genes, respectively). (C) SHAP (SHapley Additive exPlanations) values for the top 10 features affecting ScBE3 target preferences.

The screen resulted in modest depletion of sgRNAs targeting essential genes, with depletion increasing with induction time (median logFC −1.19 after 24 h of base editor induction) (Figure 3B and Supplementary Figure S5A). There was additionally no evidence of a prevailing bad-seed effect (22), as the distribution of sgRNAs targeting nonessential genes showed no depletion even after 24 h (median logFC 0.22) (Figure 3B). We additionally tested a second base editor variant, evoAPOBEC1-ScnCas9-UGI (abbreviated as evoBE), engineered to overcome the limitations of the WT APOBEC1 deaminase to edit GC contexts (63); however, we observed even less guide depletion compared to the ScBE3 variant (Supplementary Figure S6A and B). Therefore, all the following experiments are based on ScBE3.

To identify determinants of efficient editing and gene disruption, we applied automated machine learning to guides targeting essential genes to develop a model for the prediction of efficient gene disruption (see the ‘Materials and methods’ section). Targeting preferences were then elucidated using SHAP values (53), which assigns an importance to each feature in a model based on the magnitude of the value and whether it is positive (positive effect) or negative (negative effect) (64). For one, we observed that mutating the start codon of essential genes showed higher depletions (median logFC −3.34 for ATG) in contrast to introducing premature stop codons (median logFC −1.67 for TGG; median logFC −1.04 for CAG; median logFC −0.95 for CAA; median logFC −0.74 for CGA) (Supplementary Figure S5B). We further found that edited cytosines within a TC motif yielded the largest guide depletion (median logFC −3.23) while cytosines within a GC motif yielded the smallest guide depletion (median logFC −0.33) (Supplementary Figure S5C), in line with known preferences for base editors with an rAPOBEC1 domain (23). We further found that essential genes were depleted more efficiently when the target cytosine fell between positions 5 and 7 of the target (median logFC −1.31 for position 5; median logFC −1.65 for position 6; median logFC −1.61 for position 7), compared to position 8 of the target (median logFC −0.48) (Supplementary Figure S5D). Finally, we ranked the PAM motifs flanking the guide target sequences according to the highest and lowest sgRNA depletion (median logFC −4.72 for 5′-CAG-3′ versus median logFC −0.26 for 5′-TTG-3′), showing that 5′-NAG-3′ PAMs resulted in the highest depletion of essential genes (Supplementary Figure S5E). These findings provide a set of design rules for efficient gene disruption in E. coli using ScBE3 (Figure 3C).

Combining base editing with Cas9-induced cell killing increases the apparent editing efficiency

We suspected that the modest depletion of guides targeting essential genes in the screen stemmed from editing occurring slowly in relation to cell division, allowing unedited cells harboring the targeting guide to dominate. We therefore explored one option to increase the fraction of edited cells: using Cas9 to remove unedited cells (65). Making use of the broad PAM range of the ScCas9 (54), we developed a dual system that combines ScBE3 base editing with subsequent ScCas9-mediated cleavage of the same locus in case of an unsuccessful editing event (Figure 4A). For DNA cleavage, we selected the ScCas9++ variant that exhibits tight DNA binding and cleavage across a broader set of NNG PAMs (66). In the first step, ScBE3 and the base editing guide are induced with the addition of IPTG and l-arabinose. In the second step, IPTG and L-arabinose are removed through washing, while ScCas9++ and its corresponding guide are induced with the addition of aTc. While both nucleases can use the same guide, the use of orthogonal inducers helps ensure that cells are undergoing either base editing or counterselection. Furthermore, the guides were designed such that the C modified by the base editor is positioned on the opposite strand of the ScCas9++ target site, complementary to the guanosine within the 5′-NNG-3′ PAM, recognized by ScCas9 (Figure 4B). As a result, unedited cells can be eliminated from the population (65), leading to higher apparent editing rates and a stronger depletion of guides with essential targets as part of a genetic screen. This approach parallels one used to enrich cells edited with an adenine base editor (67), with the prior study using inducible expression of the base editor followed by a Cas9 with the same guide.

Figure 4. — A dual system that combines base editing with Cas9-induced cell killing helps rescue less efficient target sites. (A) Schematic of the constructs for the dual system facilitating inducible base editing and Cas9 counterselection. (B) Schematic of the experimental setup that combines base editing with subsequent Cas9-induced cleavage of the unedited DNA. In the first step, the ScBE3 and base editing guide (top strand shown in dark blue) are expressed leading to the introduction of a stop codon by converting cytosines to thymines. In the next step, ScCas9++ and the killing guide (bottom strand shown in light blue) are expressed, which mediates DNA cleavage in case of the presence of an intact 5′-NGG-3′ PAM, reflecting unsuccessful base editing on the opposite DNA strand. Consequently, the unedited cell fraction is eradicated from the population (crossed out cells). (C) Analysis of the total number of colonies and proportion of white colonies representing edited cells after applying the dual system. CFU/ml (top graph) and percentage of white colonies (bottom graph) after induction of ScBE3 alone or utilizing the dual system combining ScBE3 with ScCas9++ and plating on X-gal plates. The *lacZ* gene was targeted at different loci utilizing sgRNAs #1–6 (or in case of the dual system with a base editing and killing sgRNA pairs #1–6) (Supplementary Table S4). The dots represent three individual biological replicates and the horizontal bar represents the mean (top graph). The bars and vertical lines represent the mean and standard deviation from three individual biological replicates, respectively (bottom graph). ****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05; ns: P > 0.05.

To validate the efficiency of this dual system, we disrupted the E. coli lacZ gene and performed a blue–white screen to quantify the frequency of editing and survival. After co-transforming E. coli with the ScBE3, ScCas9++ and dual-sgRNA plasmids, we sequentially induced expression of ScBE3 and the guide mediating the cytosine conversion followed by ScCas9++ and the guide responsible for DNA cleavage. We tested guides exhibiting a range of editing efficiencies (Supplementary Table S4), where low-performance guides would be expected to yield higher editing frequencies at the cost of fewer colonies. In line with this expectation, guides exhibiting poor editing efficiencies (i.e. 0% for sgRNA #3, 9.8% for sgRNA #5) were associated with the largest CFU reductions following the killing step (i.e. 40-fold for sgRNA #3, 566-fold for sgRNA #5), whereas guides exhibiting high editing efficiencies (i.e. 78% for sgRNA #1, 83% for sgRNA #4) were associated with the smallest CFU reductions following the killing step (i.e. 2.1-fold for sgRNA #4, 9.8-fold for sgRNA #6). In all cases, the reduction in CFUs coincided with an increase in the fraction of white cells, in line with enrichment of cells harboring the intended edit.

One concern with our dual-targeting setup is that leaky expression of ScCas9++ or the editing sgRNA could induce cell killing independent of editing, whether during the base editing step (i.e. leaky expression of ScCas9++) or the killing step (i.e. leaky expression of the editing sgRNA). We therefore examined the CFUs and the percentage of white colonies for the targeting dual-sgRNA system versus one encoding an editing guide with a mismatch simulating editing (to gauge potential crosstalk that kills edited cells) and a nontargeting killing guide (Supplementary Figure S7A). Despite the introduced mismatch, the editing guide still resulted in editing at levels that were not significantly different from the original targeting guide (Supplementary Figure S7B). For the base editing step, there was no significant difference in CFUs between the two systems (mean CFU/ml = 6.7E5 and 1.0E6, P = 0.9), arguing against use of the editing guide by uninduced ScCas9++. For the killing step, the targeting system resulted in a 78-fold reduction in CFUs compared to the control system, arguing against a major contribution due to crosstalk. However, some CFU reduction may still result from ScCas9++ using the uninduced editing sgRNA with the mismatch. Therefore, we included glucose in the killing step to further suppress the pBad promoter controlling expression of the editing sgRNA. Doing so enhanced the fold reduction in CFUs with targeting but not significantly (78-fold to 289-fold, P = 0.18). We also noticed that CFUs increased significantly by adding glucose (targeting dual guide: mean CFU/ml = 7.3E4 to 3.9E5, P = 0.048; MM-NT dual guide: mean CFU/ml = 1.2E6 to 9.8E7, P = 0.013), although this increase may be attributed to higher growth rates and stationary phase density on this rich carbon source. Therefore, crosstalk was not an apparent issue for the two-sgRNA system.

Applying the dual system to a high-throughput screen that decouples editing and phenotyping screening

With the dual system, we first enrich the population of edited cells before applying selective pressure. In contrast, in the original gene essentiality screen, editing and screening occurred simultaneously, which complicated the analysis of the resulting guide abundances and resulted in generally low guide depletion due to only moderate base editing efficiencies. Therefore, decoupling editing and screening could prevent unedited cells from masking the actual screening effect and thus make a base editing screen more efficient and reliable.

We next wanted to apply the dual system to a setup where we could decouple editing from the screening itself. Therefore, we focused on screening for genes that are essential only for growth in minimal media. Accordingly, we designed a library of 3863 guides targeting 298 genes, with 119 genes essential in M9 or MOPS minimal medium. Guides were designed to either facilitate the introduction of a premature stop codon or mutate the start codon. From the total number of guides, 224 were targeted to genes essential in M9 minimal medium, 111 were targeted to genes essential in MOPS minimal medium and 1159 were targeted to genes that are necessary for growth in both conditions. Additionally, we included 2369 guides targeting nonessential genes (derived from the previous gene essentiality screen) and 100 nontargeting guides (Supplementary Tables S10 and S11). To maintain all possible guides within this targeted library, we did not discard any due to off-targeting concerns.

Prior to the screen, we co-transformed E. coli with the ScBE3, ScCas9++ and the dual-sgRNA library, induced base editing followed by a recovery phase and finally initiated removal of unedited cells by inducing ScCas9++ with its corresponding guide (Figure 5A and Supplementary Table S5). The library obtained from each step was subjected to NGS to quantify guide abundances. As expected, no substantial depletion of targeting guides was observed in the library extracted after transformation, growth in nutrient-rich LB medium and induction of base editing (Supplementary Table S6). A relatively high depletion of guides was observed when comparing the library after the base editing step and the library after the killing step for guides targeting conditionally essential genes (median logFC = −2.12) as well as nonessential genes (median logFC = −1.89). The high depletion values suggest the removal of unedited cells by ScCas9++, although genome targeting or editing may have introduced some growth defect.

Figure 5. — Screening conditionally essential *E. coli* genes with the dual system confirms previously published hits and results in an updated version of ScBE3 target preferences. (A) Schematic of the base editor screen with a dual system combining base editing with subsequent Cas9-induced DNA cleavage of the unedited cell fraction. (B) Comparison of the guide depletion in M9 and MOPS minimal media between conditionally essential and nonessential gene targets. LogFC of sgRNAs targeting conditionally essential genes or nonessential genes based on the difference in guide abundance from the library after the killing step and after the screen in minimal media. A nontargeting control is included in the experiment. (C) Rank order of the 30 genes with the highest guide depletion. The sorting is based on M9 (left) and MOPS (right) minimal media. Genes in blue (*nadB*, *panZ*, *pdxB*) are essential only in M9 minimal medium (69), genes in green (*panD*) are essential only in MOPS minimal medium (51,70) and all other genes are essential in both media. (D) Comparing the guide depletion in M9 and MOPS minimal media between different dinucleotide contexts harboring the target cytosine. LogFC of sgRNAs targeting different dinucleotide motifs (AC, CC, GC, TC) within conditionally essential genes. (E) Comparing guide depletion in M9 and MOPS minimal media between different positions of the target cytosine within the ScBE3 editing window. LogFC of sgRNAs targeting essential genes with the target cytosine at position 5, 6, 7 or 8 within the guide sequence. (F) Comparing the guide depletion in M9 and MOPS minimal media when utilizing different PAMs. LogFC of sgRNAs targeting different 5′-NNG-3′ PAMs. (G) SHAP values for the top 10 features affecting ScBE3 target preferences. For panels (B) and (D)–(F), boxplots: Centerline represents the median, the boxes indicate first and third quartiles, respectively, while the whiskers are 1.5 times the interquartile range above or below the edges of the boxes. The logFC values are based on the number of guides (numbers within the boxes) for each group.

A screen decoupled from editing reveals E. coli genes conditionally essential in minimal media

Following the base editing procedure, we screened for conditionally essential E. coli genes in minimal media. For this purpose, we cultured our mutant library in M9 or MOPS minimal medium, selected for those genotypes in which the respective conditionally essential genes are intact and analyzed the guide abundances by NGS. Regarding the M9 minimal medium, we detected efficient depletion of guides targeting genes necessary for growth (median logFC = −2.79 and −1.51 for guides targeting conditionally essential and nonessential genes, respectively) (Supplementary Table S6). Analyzing the gene hits from the screen in M9 minimal medium showed that 29 of the 30 genes with the most strongly depleted guides [median logFC values ranging between −7.66 (carA) and −4.26 (pdxH)] were previously identified as essential for growth in M9 (68,69) (Figure 5C). However, we also found several examples of guides targeting genes previously defined as nonessential showing higher depletion values in this screen (e.g. median logFC = −2.04 for yfcG and −3.47 for gatZ).

In MOPS minimal medium, guides targeting conditionally essential genes were depleted as well, but less severely (median logFC = −1.79 and −1.06 for guides targeting conditionally essential and nonessential genes, respectively) (Supplementary Table S6). The weaker depletion in MOPS minimal medium (Figure 5B and Supplementary Figure S8A) was likely due to slower growth indicated by the lower optical density of the culture compared to the culture in M9 minimal medium (Supplementary Table S6). However, analyzing the gene hits from screening in MOPS medium revealed that 27 of the 30 genes with the most strongly depleted guides were previously identified as essential for growth in MOPS minimal medium (51,68,70). Also in this screening condition, several guides targeting genes assumed to be nonessential for growth in MOPS minimal medium showed relatively high depletion values [median logFC ranging between −2.01 (ybjT) and −2.94 (dsdA) for MOPS minimal medium].

To investigate whether the unexpectedly high guide depletion for those genes previously described as nonessential was screen dependent or caused by biologically relevant growth defects, we performed gene disruption and growth experiments in pure cultures for some representative examples (msrQ, yggN and ypjA) and included a positive and a negative control (purK and yihQ, respectively) for comparison (Supplementary Table S7). Growth curves of the selected examples demonstrated no obvious growth defects caused by gene disruption (Supplementary Figure S9), suggesting that the relatively high guide depletion (logFC < −2) in the screen was affected by either competition between the co-cultured strains or the overall lenient threshold to categorize guides as differentially depleted from the reference. Although some gene hits with guide depletion values just around the threshold of logFC < −2 could not be validated as conditionally essential (Supplementary Figure S9), the screen reliably revealed previously reported conditionally essential genes required for growth in M9 and MOPS minimal media. In sum, the NGS results demonstrated the selective depletion of guides targeting conditionally essential genes using the dual ScBE3–ScCas9++ system (Figure 5B).

While we elucidated targeting rules for ScBE3 with the essentiality screen, we asked how these rules changed with the introduction of the counterselection step as well as separating editing from screening. To accomplish this, we compared the guide library extracted after the ScCas9++ killing step and after the screen in minimal media. Similar to the gene essentiality screen in E. coli, a machine learning model was developed to predict features contributing to efficient ScBE3 editing and gene disruption (see the ‘Materials and methods’ section).

The machine learning model for the screen of conditionally essential genes was trained with guides targeting genes essential in both minimal media, using the average logFC and including the same sequence features as the initial gene essentiality screen. Despite employing the same sequence features as those used in the initial screen, the machine learning model identified some additional ScBE3 sequence preferences specific to the dual system. We confirmed the TC dinucleotide preference and low editing efficiency in GC contexts known for rAPOBEC1 domains (23) (median logFC −3.37/−3.02 for TC in M9/MOPS minimal medium versus logFC −2.66/−2.05 for GC in M9/MOPS minimal medium) (Figure 5D) and determined the preferred base editing window of the target C to be in positions 5–7 versus position 8 (median logFC −2.9/−2.17 for position 5 in M9/MOPS minimal medium; median logFC −2.69/−1.99 for position 6 in M9/MOPS minimal medium; median logFC −2.54/−2.28 for position 7 in M9/MOPS minimal medium; median logFC −2.54/−1.89 for position 8 in M9/MOPS minimal medium) (Figure 5E). Furthermore, we again found that 5′-NAG-3′ PAMs yielded the highest editing efficiencies, with the highest guide depletion with a 5′-GAG-3′ PAM (median logFC −4.78/−4.08 for M9/MOPS minimal medium) and the lowest guide depletion for a 5′-TTG-3′ PAM (median logFC −1.68/−1.22 for M9/MOPS minimal medium) (Figure 5F). Interestingly, in contrast to the first screen in which higher guide depletion was achieved by mutating start codons than by introducing premature stop codons, no such preference could be detected for the conditional essentiality screen (Supplementary Figure S8B). Overall, the screen of conditionally essential E. coli genes confirmed most of the target preferences determined through the genome-wide screen of essential E. coli genes but differed in some sequence features, likely due to the additional counterselection step involved in the dual system (Figure 5G).

Discussion

Base editing is a promising alternative to currently used technologies for genome-wide knockout and knockdown screens. However, base editing is not yet widely applied for such screens in bacteria, with only one study describing a genome-wide base editing screen in C. glutamicum (39). This disconnect can be attributed in part to the limited number of available target sites, which restricts how many genes can be reliably screened. Therefore, improved flexibility of base editing would greatly contribute to enabling depletion-based high-throughput screens and advance the field of functional genomics in bacteria. We applied three approaches to improve targeting flexibility: adopting a Cas nuclease with flexible PAM recognition (i.e. ScCas9) (54), mutating start codons on top of introducing premature stop codons and adding a counterselection step following base editing to rescue less efficient target sites. An added benefit of using ScCas9 in particular was poor gene repression through DNA binding (Figure 1G), limiting this effect as a confounding variable for high-throughput screening. We also found that ScBE3 could generate bystander edits (Supplementary Figure S3) similar to other base editors (56). Assessing their impact on high-throughput depletion screens such as those performed in this work could introduce additional rules for guide design.

Apart from the options presented in this work to increase the number of possible target sites, we can take advantage of the large collection of previously characterized nucleases with divergent PAM profiles (71–73) or use truncated or extended sgRNAs that shift the base editing window (74) to find suitable target sites. In addition, other studies performed orthogonal expression of different base editor variants (39) to cover the majority of all possible target sites. Overall, the measures demonstrated in this work allow for more flexible base editing, which could enable a more widespread application of this technology for high-throughput screens specifically in bacteria.

To improve upon the efficiency of base editing as part of the first high-throughput screen, we introduced a dual system that includes a counterselection step after base editing. A conceptually similar approach was described previously (67), where Cas9 counterselection following base editing enriches for edited cells and consequently increases editing efficiencies. Apart from not applying the approach in a library format, an important distinction is that the prior study used the same sgRNA for the base editor and nuclease, thereby relying on a mismatch at a less sensitive location in the target. In contrast, we employed a second sgRNA in which modifying the target cytosine disrupts the PAM, the most sensitive feature of target recognition by a DNA-targeting Cas nuclease (Figure 4B). Regardless of the approach, enriching the fraction of edited cells in the population should yield greater changes in guide abundance in a phenotypic screen. The trade-off is that sgRNAs associated with poor editing will be greatly depleted in the population, potentially requiring deeper sequencing to quantify changes in the abundance of these sgRNAs as part of a screen. Although this approach limits screening to genes that are nonessential under the editing conditions (e.g. growth in rich LB medium), we recommend this strategy for base editing screens.

Some base editing guide design or library design tools are already publicly available (42,75–77). However, depending on the bacterial species, base editing system or specific experimental setups, the predictability of the designed guides may vary drastically. By training our machine learning model on ScBE3 guides targeting (conditionally) essential E. coli genes and, for the second screen, incorporating the effects of counterselection and decoupling of editing and screening, we integrate crucial features that affect the editing efficiency. We also interpret the model using SHAP values (53), which identify preferred ScBE3 sequence features (e.g. PAM preferences, optimal target cytosine position within the editing window, preferred dinucleotide contexts of the target cytosine). Incorporating those features becomes especially important when guides are used where individual guide testing does not occur prior to use. From the initial screen, we identified important ScBE3 target preferences, including a known preference of the rAPOBEC1 domain for TC motifs, optimal editing in a window between positions 5–7 of the guide, and the highest editing efficiency for NAG PAMs recognized by ScCas9, which were in line with results from previous studies (29,55). We again applied automated machine learning to the data obtained from the screen using the dual system, as the additional counterselection step as well as decoupling editing and screening could impact the model’s predictability. Most of the top features for the screen based on the dual system were similar to those determined by the model in the first screen. However, there were also minor deviations. For example, the feature with the greatest influence on the editing efficiency in the initial screen—an A in the second PAM position—was completely absent from the model for the second screen. Such differences emphasize the importance of specifically fitting the model to the selected experimental setup to predict optimal guide features or target preferences. These findings could also provide useful guidance for future library designs for ScCas9-derived base editors.

The increased flexibility of base editing could promote its broader application in bacteria for gene disruptions and other specific point mutations. For gene disruption, our base editing approach could be extended to bacteria with lower transformation efficiencies (e.g. Clostridia, Streptomyces) (78,79), where Cas9 editing does not yield surviving cells or transposon mutagenesis libraries have insufficient coverage of the genome due to a limited number of transformants. Beyond gene disruption, base editors can also be used for other targeted point mutations, such as altering specific amino acid residues (80,81) or screening regulatory elements (82). Apart from the study of gene essentiality, base editing screens offer a variety of potential applications in bacteria, ranging from the optimization of industrially relevant metabolic pathways (35,83) to the identification of therapeutically relevant proteins (84) or virulence genes (85,86). In conclusion, with our dual system and other improvements, we present an enhanced and flexible base editing screening approach that could unlock new possibilities for leveraging the base editing technology for genome-wide screens in bacteria. This advancement could contribute to a deeper understanding of the genetics and physiology of diverse bacteria, enabling us to obtain new insights and unravel complex mechanisms.

Supplementary Material

gkae174_Supplemental_Files

gkae174_supplemental_files.zip^{(7.1MB, zip)}

Acknowledgements

Author contributions: S.G., S.P.C. and C.L.B. devised the study. S.G., S.P.C., Y.Y. and S.A.B. performed experiments. S.G. and C.L.B. wrote the original draft. S.G., S.P.C., Y.Y., L.B. and C.L.B. reviewed and edited the manuscript. S.G. and Y.Y. generated the figures. L.B. and C.L.B. acquired funding and supervised the project.

Contributor Information

Sandra Gawlitt, Helmholtz Institute for RNA-based Infection Research (HIRI), Helmholtz Centre for Infection Research (HZI), 97080 Würzburg, Germany.

Scott P Collins, Department of Chemical and Biomolecular Engineering, North Carolina State University, Raleigh, NC 27695, USA.

Yanying Yu, Helmholtz Institute for RNA-based Infection Research (HIRI), Helmholtz Centre for Infection Research (HZI), 97080 Würzburg, Germany.

Samuel A Blackman, Department of Chemical and Biomolecular Engineering, North Carolina State University, Raleigh, NC 27695, USA.

Lars Barquist, Helmholtz Institute for RNA-based Infection Research (HIRI), Helmholtz Centre for Infection Research (HZI), 97080 Würzburg, Germany; Department of Biology, University of Toronto Mississauga, Mississauga, Ontario L5L 1C6, Canada; Medical Faculty, University of Würzburg, 97080 Würzburg, Germany.

Chase L Beisel, Helmholtz Institute for RNA-based Infection Research (HIRI), Helmholtz Centre for Infection Research (HZI), 97080 Würzburg, Germany; Medical Faculty, University of Würzburg, 97080 Würzburg, Germany.

Data availability

The NGS data were deposited in NCBI’s Gene Expression Omnibus (GEO) (87) and are accessible through GEO Series accession number GSE225335. Source code was deposited on Zenodo (DOI: 10.5281/zenodo.8186157).

Supplementary data

Supplementary Data are available at NAR Online.

Funding

European Research Council [865973 to C.L.B.]; National Institutes of Health [1R35GM119561 to C.L.B.]; Bavarian State Ministry for Science and Art (to L.B.). Funding for open access charge: European Research Council.

Conflict of interest statement. C.L.B. is a co-founder and member of the Scientific Advisory Board for Locus Biosciences as well as a member of the Scientific Advisory Board for Benson Hill. The other authors have no conflicts of interest to declare.

References

1. Przybyla L., Gilbert L.A. A new era in functional genomics screens. Nat. Rev. Genet. 2022; 23:89–103. [DOI] [PubMed] [Google Scholar]
2. Cain A.K., Barquist L., Goodman A.L., Paulsen I.T., Parkhill J., van Opijnen T. A decade of advances in transposon-insertion sequencing. Nat. Rev. Genet. 2020; 21:526–540. [DOI] [PMC free article] [PubMed] [Google Scholar]
3. Coe K.A., Lee W., Stone M.C., Komazin-Meredith G., Meredith T.C., Grad Y.H., Walker S. Multi-strain Tn-Seq reveals common daptomycin resistance determinants in Staphylococcus aureus. PLoS Pathog. 2019; 15:e1007862. [DOI] [PMC free article] [PubMed] [Google Scholar]
4. Jana B., Cain A.K., Doerrler W.T., Boinett C.J., Fookes M.C., Parkhill J., Guardabassi L. The secondary resistome of multidrug-resistant Klebsiella pneumoniae. Sci. Rep. 2017; 7:42483. [DOI] [PMC free article] [PubMed] [Google Scholar]
5. Shames S.R., Liu L., Havey J.C., Schofield W.B., Goodman A.L., Roy C.R. Multiple effector virulence phenotypes revealed through high-throughput analysis of targeted mutant libraries. Proc. Natl Acad. Sci. U.S.A. 2017; 114:E10446–E10454. [DOI] [PMC free article] [PubMed] [Google Scholar]
6. Warr A.R., Hubbard T.P., Munera D., Blondel C.J., Abel Zur Wiesch P., Abel S., Wang X., Davis B.M., Waldor M.K. Transposon-insertion sequencing screens unveil requirements for EHEC growth and intestinal colonization. PLoS Pathog. 2019; 15:e1007652. [DOI] [PMC free article] [PubMed] [Google Scholar]
7. Liu H., Shiver A.L., Price M.N., Carlson H.K., Trotter V.V., Chen Y., Escalante V., Ray J., Hern K.E., Petzold C.J. et al. Functional genetics of human gut commensal Bacteroides thetaiotaomicron reveals metabolic requirements for growth across environments. Cell Rep. 2021; 34:108789. [DOI] [PMC free article] [PubMed] [Google Scholar]
8. Mandal R.K., Kwon Y.M. Global screening of Salmonella enterica serovar Typhimurium genes for desiccation survival. Front. Microbiol. 2017; 8:1723. [DOI] [PMC free article] [PubMed] [Google Scholar]
9. Goodall E.C.A., Robinson A., Johnston I.G., Jabbari S., Turner K.A., Cunningham A.F., Lund P.A., Cole J.A., Henderson I.R. The essential genome of Escherichia coli K-12. mBio. 2018; 9:e02096-17. [DOI] [PMC free article] [PubMed] [Google Scholar]
10. Jiang W., Bikard D., Cox D., Zhang F., Marraffini L.A. RNA-guided editing of bacterial genomes using CRISPR–Cas systems. Nat. Biotechnol. 2013; 31:233–239. [DOI] [PMC free article] [PubMed] [Google Scholar]
11. Jiang Y., Qian F., Yang J., Liu Y., Dong F., Xu C., Sun B., Chen B., Xu X., Li Y. et al. CRISPR–Cpf1 assisted genome editing of Corynebacterium glutamicum. Nat. Commun. 2017; 8:15179. [DOI] [PMC free article] [PubMed] [Google Scholar]
12. Vento J.M., Crook N., Beisel C.L. Barriers to genome editing with CRISPR in bacteria. J. Ind. Microbiol. Biotechnol. 2019; 46:1327–1341. [DOI] [PMC free article] [PubMed] [Google Scholar]
13. Liu R., Liang L., Freed E.F., Choudhury A., Eckert C.A., Gill R.T. Engineering regulatory networks for complex phenotypes in E. coli. Nat. Commun. 2020; 11:4050. [DOI] [PMC free article] [PubMed] [Google Scholar]
14. Garst A.D., Bassalo M.C., Pines G., Lynch S.A., Halweg-Edwards A.L., Liu R., Liang L., Wang Z., Zeitoun R., Alexander W.G. et al. Genome-wide mapping of mutations at single-nucleotide resolution for protein, metabolic and genome engineering. Nat. Biotechnol. 2016; 35:48–55. [DOI] [PubMed] [Google Scholar]
15. Cui L., Bikard D. Consequences of Cas9 cleavage in the chromosome of Escherichia coli. Nucleic Acids Res. 2016; 44:4243–4251. [DOI] [PMC free article] [PubMed] [Google Scholar]
16. Qi L.S., Larson M.H., Gilbert L.A., Doudna J.A., Weissman J.S., Arkin A.P., Wendell A Repurposing CRISPR as an RNA-guided platform for sequence-specific control of gene expression. Cell. 2013; 152:1173–1183. [DOI] [PMC free article] [PubMed] [Google Scholar]
17. Bikard D., Jiang W., Samai P., Hochschild A., Zhang F., Marraffini L.A. Programmable repression and activation of bacterial gene expression using an engineered CRISPR–Cas system. Nucleic Acids Res. 2013; 41:7429–7437. [DOI] [PMC free article] [PubMed] [Google Scholar]
18. Li X.-T., Jun Y., Erickstad M.J., Brown S.D., Parks A., Court D.L., Jun S. tCRISPRi: tunable and reversible, one-step control of gene expression. Sci. Rep. 2016; 6:39076. [DOI] [PMC free article] [PubMed] [Google Scholar]
19. Fontana J., Dong C., Ham J.Y., Zalatan J.G., Carothers J.M. Regulated expression of sgRNAs tunes CRISPRi in E. coli. Biotechnol. J. 2018; 13:e1800069. [DOI] [PubMed] [Google Scholar]
20. Peters J.M., Colavin A., Shi H., Czarny T.L., Larson M.H., Wong S., Hawkins J.S., Lu C.H.S., Koo B.-M., Marta E. et al. A comprehensive, CRISPR-based functional analysis of essential genes in bacteria. Cell. 2016; 165:1493–1506. [DOI] [PMC free article] [PubMed] [Google Scholar]
21. Lee H.H., Ostrov N., Wong B.G., Gold M.A., Khalil A.S., Church G.M. Functional genomics of the rapidly replicating bacterium Vibrio natriegens by CRISPRi. Nat. Microbiol. 2019; 4:1105–1113. [DOI] [PubMed] [Google Scholar]
22. Cui L., Vigouroux A., Rousset F., Varet H., Khanna V., Bikard D. A CRISPRi screen in E. coli reveals sequence-specific toxicity of dCas9. Nat. Commun. 2018; 9:1912. [DOI] [PMC free article] [PubMed] [Google Scholar]
23. Komor A.C., Kim Y.B., Packer M.S., Zuris J.A., Liu D.R. Programmable editing of a target base in genomic DNA without double-stranded DNA cleavage. Nature. 2016; 533:420–424. [DOI] [PMC free article] [PubMed] [Google Scholar]
24. Nishida K., Arazoe T., Yachie N., Banno S., Kakimoto M., Tabata M., Mochizuki M., Miyabe A., Araki M., Hara K.Y. et al. Targeted nucleotide editing using hybrid prokaryotic and vertebrate adaptive immune systems. Science. 2016; 353:aaf8729. [DOI] [PubMed] [Google Scholar]
25. Mol C.D., Arvai A.S., Sanderson R.J., Slupphaug G., Kavli B., Krokan H.E., Mosbaugh D.W., Tainer J.A. Crystal structure of human uracil-DNA glycosylase in complex with a protein inhibitor: protein mimicry of DNA. Cell. 1995; 82:701–708. [DOI] [PubMed] [Google Scholar]
26. Tang W., Liu D.R. Rewritable multi-event analog recording in bacterial and mammalian cells. Science. 2018; 360:eaap8992. [DOI] [PMC free article] [PubMed] [Google Scholar]
27. Zhao D., Li J., Li S., Xin X., Hu M., Price M.A., Rosser S.J., Bi C., Zhang X. Glycosylase base editors enable C-to-A and C-to-G base changes. Nat. Biotechnol. 2020; 39:35–40. [DOI] [PubMed] [Google Scholar]
28. Tong H., Liu N., Wei Y., Zhou Y., Li Y., Wu D., Jin M., Cui S., Li H., Li G. et al. Programmable deaminase-free base editors for G-to-Y conversion by engineered glycosylase. Natl Sci. Rev. 2023; 10:nwad143. [DOI] [PMC free article] [PubMed] [Google Scholar]
29. Grünewald J., Zhou R., Lareau C.A., Garcia S.P., Iyer S., Miller B.R., Langner L.M., Hsu J.Y., Aryee M.J., Joung J.K. A dual-deaminase CRISPR base editor enables concurrent adenine and cytosine editing. Nat. Biotechnol. 2020; 38:861–864. [DOI] [PMC free article] [PubMed] [Google Scholar]
30. Banno S., Nishida K., Arazoe T., Mitsunobu H., Kondo A. Deaminase-mediated multiplex genome editing in Escherichia coli. Nat. Microbiol. 2018; 3:423–429. [DOI] [PubMed] [Google Scholar]
31. Zheng K., Wang Y., Li N., Jiang F.-F., Wu C.-X., Liu F., Chen H.-C., Liu Z.-F. Highly efficient base editing in bacteria using a Cas9–cytidine deaminase fusion. Commun. Biol. 2018; 1:32. [DOI] [PMC free article] [PubMed] [Google Scholar]
32. Gu T., Zhao S., Pi Y., Chen W., Chen C., Liu Q., Li M., Han D., Ji Q. Highly efficient base editing in using an engineered CRISPR RNA-guided cytidine deaminase. Chem. Sci. 2018; 9:3248–3253. [DOI] [PMC free article] [PubMed] [Google Scholar]
33. Wang Y., Wang S., Chen W., Song L., Zhang Y., Shen Z., Yu F., Li M., Ji Q. CRISPR–Cas9 and CRISPR-assisted cytidine deaminase enable precise and efficient genome editing in Klebsiella pneumoniae. Appl. Environ. Microbiol. 2018; 84:e01834-18. [DOI] [PMC free article] [PubMed] [Google Scholar]
34. Wang Y., Liu Y., Liu J., Guo Y., Fan L., Ni X., Zheng X., Wang M., Zheng P., Sun J. et al. MACBETH: multiplex automated Corynebacterium glutamicum base editing method. Metab. Eng. 2018; 47:200–210. [DOI] [PubMed] [Google Scholar]
35. Volke D.C., Martino R.A., Kozaeva E., Smania A.M., Nikel P.I. Modular (de)construction of complex bacterial phenotypes by CRISPR/nCas9-assisted, multiplex cytidine base-editing. Nat. Commun. 2022; 13:3026. [DOI] [PMC free article] [PubMed] [Google Scholar]
36. Xu P., Liu Z., Liu Y., Ma H., Xu Y., Bao Y., Zhu S., Cao Z., Wu Z., Zhou Z. et al. Genome-wide interrogation of gene functions through base editor screens empowered by barcoded sgRNAs. Nat. Biotechnol. 2021; 39:1403–1413. [DOI] [PubMed] [Google Scholar]
37. Hanna R.E., Hegde M., Fagre C.R., DeWeirdt P.C., Sangree A.K., Szegletes Z., Griffith A., Feeley M.N., Sanson K.R., Baidi Y. et al. Massively parallel assessment of human variants with base editor screens. Cell. 2021; 184:1064–1080. [DOI] [PubMed] [Google Scholar]
38. Kweon J., Jang A.-H., Shin H.R., See J.-E., Lee W., Lee J.W., Chang S., Kim K., Kim Y. A CRISPR-based base-editing screen for the functional assessment of BRCA1 variants. Oncogene. 2020; 39:30–35. [DOI] [PMC free article] [PubMed] [Google Scholar]
39. Liu Y., Wang R., Liu J., Lu H., Li H., Wang Y., Ni X., Li J., Guo Y., Ma H. et al. Base editor enables rational genome-scale functional screening for enhanced industrial phenotypes in Corynebacterium glutamicum. Sci. Adv. 2022; 8:eabq2157. [DOI] [PMC free article] [PubMed] [Google Scholar]
40. Billon P., Bryant E.E., Joseph S.A., Nambiar T.S., Hayward S.B., Rothstein R., Ciccia A. CRISPR-mediated base editing enables efficient disruption of eukaryotic genes through induction of STOP codons. Mol. Cell. 2017; 67:1068–1079. [DOI] [PMC free article] [PubMed] [Google Scholar]
41. Kuscu C., Parlak M., Tufan T., Yang J., Szlachta K., Wei X., Mammadov R., Adli M. CRISPR-STOP: gene silencing through base-editing-induced nonsense mutations. Nat. Methods. 2017; 14:710–712. [DOI] [PubMed] [Google Scholar]
42. Marquart K.F., Allam A., Janjuha S., Sintsova A., Villiger L., Frey N., Krauthammer M., Schwank G. Predicting base editing outcomes with an attention-based deep learning algorithm trained on high-throughput target library screens. Nat. Commun. 2021; 12:5114. [DOI] [PMC free article] [PubMed] [Google Scholar]
43. Neidhardt F.C., Bloch P.L., Smith D.F. Culture medium for enterobacteria. J. Bacteriol. 1974; 119:736–747. [DOI] [PMC free article] [PubMed] [Google Scholar]
44. Miller J.H. Assay of β-galactosidase. Experiments in Molecular Genetics. 1972; Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press. [Google Scholar]
45. Jiang H., Wong W.H. SeqMap: mapping massive amount of oligonucleotides to the genome. Bioinformatics. 2008; 24:2395–2396. [DOI] [PMC free article] [PubMed] [Google Scholar]
46. Robinson M.D., McCarthy D.J., Smyth G.K. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics. 2009; 26:139–140. [DOI] [PMC free article] [PubMed] [Google Scholar]
47. Risso D., Ngai J., Speed T.P., Dudoit S. Normalization of RNA-seq data using factor analysis of control genes or samples. Nat. Biotechnol. 2014; 32:896–902. [DOI] [PMC free article] [PubMed] [Google Scholar]
48. Feurer M., Klein A., Eggensperger K., Springenberg J., Blum M., Hutter F.. Cortes C., Lawrence N.D., Lee D.D., Sugiyama M., Garnett R. Efficient and robust automated machine learning. Advances in Neural Information Processing Systems 28. 2015; Red Hook, NY: Curran Associates, Inc; 2962–2970. [Google Scholar]
49. Yu Y., Gawlitt S., de Andrade E Sousa L.B., Merdivan E., Piraud M., Beisel C.L., Barquist L Improved prediction of bacterial CRISPRi guide efficiency from depletion screens through mixed-effect machine learning and data integration. Genome Biol. 2024; 25:13. [DOI] [PMC free article] [PubMed] [Google Scholar]
50. Keseler I.M., Gama-Castro S., Mackie A., Billington R., Bonavides-Martínez C., Caspi R., Kothari A., Krummenacker M., Midford P.E., Muñiz-Rascado L. et al. The EcoCyc Database in 2021. Front Microbiol. 2021; 12:711077. [DOI] [PMC free article] [PubMed] [Google Scholar]
51. Baba T., Ara T., Hasegawa M., Takai Y., Okumura Y., Baba M., Datsenko K.A., Tomita M., Wanner B.L., Mori H. Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Mol. Syst. Biol. 2006; 2:2006.0008. [DOI] [PMC free article] [PubMed] [Google Scholar]
52. Pedregosa F., Varoquaux G., Gramfort A., Michel V., Thirion B., Grisel O., Blondel M., Prettenhofer P., Weiss R., Dubourg V. et al. Scikit-learn: machine learning in Python. J. Mach. Learn. Res. 2011; 12:2825–2830. [Google Scholar]
53. Lundberg S.M., Erion G., Chen H., DeGrave A., Prutkin J.M., Nair B., Katz R., Himmelfarb J., Bansal N., Lee S.-I. From local explanations to global understanding with explainable AI for trees. Nat. Mach. Intell. 2020; 2:56–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
54. Chatterjee P., Jakimo N., Jacobson J.M. Minimal PAM specificity of a highly similar SpCas9 ortholog. Sci. Adv. 2018; 4:eaau0766. [DOI] [PMC free article] [PubMed] [Google Scholar]
55. Cho S., Choe D., Lee E., Kim S.C., Palsson B., Cho B.-K. High-level dCas9 expression induces abnormal cell morphology in Escherichia coli. ACS Synth. Biol. 2018; 7:1085–1094. [DOI] [PubMed] [Google Scholar]
56. Rees H.A., Liu D.R. Base editing: precision chemistry on the genome and transcriptome of living cells. Nat. Rev. Genet. 2018; 19:770–788. [DOI] [PMC free article] [PubMed] [Google Scholar]
57. Qi L.S., Larson M.H., Gilbert L.A., Doudna J.A., Weissman J.S., Arkin A.P., Lim W.A. Repurposing CRISPR as an RNA-guided platform for sequence-specific control of gene expression. Cell. 2021; 184:844. [DOI] [PubMed] [Google Scholar]
58. Chen S., Xie W., Liu Z., Shan H., Chen M., Song Y., Yu H., Lai L., Li Z. CRISPR start-loss: a novel and practical alternative for gene silencing through base-editing-induced start codon mutations. Mol. Ther. Nucleic Acids. 2020; 21:1062–1073. [DOI] [PMC free article] [PubMed] [Google Scholar]
59. Hecht A., Glasgow J., Jaschke P.R., Bawazer L.A., Munson M.S., Cochran J.R., Endy D., Salit M. Measurements of translation initiation from all 64 codons in E. coli. Nucleic Acids Res. 2017; 45:3615–3626. [DOI] [PMC free article] [PubMed] [Google Scholar]
60. Munson L.M., Stormo G.D., Niece R.L., Reznikoff W.S. lacZ translation initiation mutations. J. Mol. Biol. 1984; 177:663–683. [DOI] [PubMed] [Google Scholar]
61. Depardieu F., Bikard D. Gene silencing with CRISPRi in bacteria and optimization of dCas9 expression levels. Methods. 2020; 172:61–75. [DOI] [PubMed] [Google Scholar]
62. Wang T., Guan C., Guo J., Liu B., Wu Y., Xie Z., Zhang C., Xing X.-H. Pooled CRISPR interference screening enables genome-scale functional genomics study in bacteria with superior performance. Nat. Commun. 2018; 9:2475. [DOI] [PMC free article] [PubMed] [Google Scholar]
63. Thuronyi B.W., Koblan L.W., Levy J.M., Yeh W.-H., Zheng C., Newby G.A., Wilson C., Bhaumik M., Shubina-Oleinik O., Holt J.R. et al. Continuous evolution of base editors with expanded target compatibility and improved activity. Nat. Biotechnol. 2019; 37:1070–1079. [DOI] [PMC free article] [PubMed] [Google Scholar]
64. Lundberg S., Lee S.-I. A unified approach to interpreting model predictions. Adv. Neural Inf. Process. Syst. 2017; 30:4765–4774. [Google Scholar]
65. Gomaa A.A., Klumpe H.E., Luo M.L., Selle K., Barrangou R., Beisel C.L. Programmable removal of bacterial strains by use of genome-targeting CRISPR–Cas systems. mBio. 2014; 5:e00928-13. [DOI] [PMC free article] [PubMed] [Google Scholar]
66. Chatterjee P., Jakimo N., Lee J., Amrani N., Rodríguez T., Koseki S.R.T., Tysinger E., Qing R., Hao S., Sontheimer E.J. et al. An engineered ScCas9 with broad PAM range and high specificity and activity. Nat. Biotechnol. 2020; 38:1154–1158. [DOI] [PubMed] [Google Scholar]
67. Xin X., Li J., Zhao D., Li S., Xie Q., Li Z., Fan F., Bi C., Zhang X. Double-check base editing for efficient A to G conversions. ACS Synth. Biol. 2019; 8:2629–2634. [DOI] [PubMed] [Google Scholar]
68. Joyce A.R., Reed J.L., White A., Edwards R., Osterman A., Baba T., Mori H., Lesely S.A., Palsson B.Ø., Agarwalla S. Experimental and computational assessment of conditionally essential genes in Escherichia coli. J. Bacteriol. 2006; 188:8259–8271. [DOI] [PMC free article] [PubMed] [Google Scholar]
69. Patrick W.M., Quandt E.M., Swartzlander D.B., Matsumura I. Multicopy suppression underpins metabolic evolvability. Mol. Biol. Evol. 2007; 24:2716–2722. [DOI] [PMC free article] [PubMed] [Google Scholar]
70. Feist A.M., Henry C.S., Reed J.L., Krummenacker M., Joyce A.R., Karp P.D., Broadbelt L.J., Hatzimanikatis V., Palsson B.Ø. A genome-scale metabolic reconstruction for Escherichia coli K-12 MG1655 that accounts for 1260 ORFs and thermodynamic information. Mol. Syst. Biol. 2007; 3:121. [DOI] [PMC free article] [PubMed] [Google Scholar]
71. Walton R.T., Christie K.A., Whittaker M.N., Kleinstiver B.P. Unconstrained genome targeting with near-PAMless engineered CRISPR–Cas9 variants. Science. 2020; 368:290–296. [DOI] [PMC free article] [PubMed] [Google Scholar]
72. Nishimasu H., Shi X., Ishiguro S., Gao L., Hirano S., Okazaki S., Noda T., Abudayyeh O.O., Gootenberg J.S., Mori H. et al. Engineered CRISPR–Cas9 nuclease with expanded targeting space. Science. 2018; 361:1259–1262. [DOI] [PMC free article] [PubMed] [Google Scholar]
73. Kleinstiver B.P., Sousa A.A., Walton R.T., Tak Y.E., Hsu J.Y., Clement K., Welch M.M., Horng J.E., Malagon-Lopez J., Scarfò I. et al. Engineered CRISPR–Cas12a variants with increased activities and improved targeting ranges for gene, epigenetic and base editing. Nat. Biotechnol. 2019; 37:276–282. [DOI] [PMC free article] [PubMed] [Google Scholar]
74. Wang Y., Liu Y., Li J., Yang Y., Ni X., Cheng H., Huang T., Guo Y., Ma H., Zheng P. et al. Expanding targeting scope, editing window, and base transition capability of base editing in Corynebacterium glutamicum. Biotechnol. Bioeng. 2019; 116:3016–3029. [DOI] [PubMed] [Google Scholar]
75. Arbab M., Shen M.W., Mok B., Wilson C., Matuszek Ż., Cassa C.A., Liu D.R. Determinants of base editing outcomes from target library analysis and machine learning. Cell. 2020; 182:463–480. [DOI] [PMC free article] [PubMed] [Google Scholar]
76. Song M., Kim H.K., Lee S., Kim Y., Seo S.-Y., Park J., Choi J.W., Jang H., Shin J.H., Min S. et al. Sequence-specific prediction of the efficiencies of adenine and cytosine base editors. Nat. Biotechnol. 2020; 38:1037–1043. [DOI] [PubMed] [Google Scholar]
77. Dandage R., Després P.C., Yachie N., Landry C.R. A computational workflow for designing libraries of guide RNAs for CRISPR-mediated base editing. Genetics. 2019; 212:377–385. [DOI] [PMC free article] [PubMed] [Google Scholar]
78. Li Q., Seys F.M., Minton N.P., Yang J., Jiang Y., Jiang W., Yang S. CRISPR–Cas9 nickase-assisted base editing in the solvent producer Clostridium beijerinckii. Biotechnol. Bioeng. 2019; 116:1475–1483. [DOI] [PubMed] [Google Scholar]
79. Tong Y., Whitford C.M., Robertsen H.L., Blin K., Jørgensen T.S., Klitgaard A.K., Gren T., Jiang X., Weber T., Lee S.Y. Highly efficient DSB-free base editing for Streptomycetes with CRISPR-BEST. Proc. Natl Acad. Sci. U.S.A. 2019; 116:20366–20375. [DOI] [PMC free article] [PubMed] [Google Scholar]
80. Hao W., Cui W., Cheng Z., Han L., Suo F., Liu Z., Zhou L., Zhou Z. Development of a base editor for protein evolution via in situ mutation in vivo. Nucleic Acids Res. 2021; 49:9594–9605. [DOI] [PMC free article] [PubMed] [Google Scholar]
81. Després P.C., Dubé A.K., Seki M., Yachie N., Landry C.R. Perturbing proteomes at single residue resolution using base editing. Nat. Commun. 2020; 11:1871. [DOI] [PMC free article] [PubMed] [Google Scholar]
82. Cheng L., Li Y., Qi Q., Xu P., Feng R., Palmer L., Chen J., Wu R., Yee T., Zhang J. et al. Single-nucleotide-level mapping of DNA regulatory elements that control fetal hemoglobin expression. Nat. Genet. 2021; 53:869–880. [DOI] [PMC free article] [PubMed] [Google Scholar]
83. Li J., Liu Y., Wang Y., Yu P., Zheng P., Wang M. Optimization of base editing in Corynebacterium glutamicum. Chin. J. Biotechnol. 2020; 36:143–151. [DOI] [PubMed] [Google Scholar]
84. Ding X.-Y., Li S.-S., Geng Y.-M., Yan M.-Y., Li G.-B., Zhang G.-L., Sun Y.-C. Programmable base editing in using an engineered CRISPR RNA-guided cytidine deaminase. Front. Genome Ed. 2021; 3:734436. [DOI] [PMC free article] [PubMed] [Google Scholar]
85. Ipoutcha T., Rideau F., Gourgues G., Arfi Y., Lartigue C., Blanchard A., Sirand-Pugnet P. Genome editing of veterinary relevant mycoplasmas using a CRISPR–Cas base editor system. Appl. Environ. Microbiol. 2022; 88:e0099622. [DOI] [PMC free article] [PubMed] [Google Scholar]
86. Rodrigues S.D., Karimi M., Impens L., Van Lerberge E., Coussens G., Aesaert S., Rombaut D., Holtappels D., Ibrahim H.M.M., Van Montagu M. et al. Efficient CRISPR-mediated base editing in Agrobacterium spp. Proc. Natl Acad. Sci. U.S.A. 2021; 118:e2013338118. [DOI] [PMC free article] [PubMed] [Google Scholar]
87. Edgar R., Domrachev M., Lash A.E. Gene Expression Omnibus: NCBI gene expression and hybridization array data repository. Nucleic Acids Res. 2002; 30:207–210. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

gkae174_Supplemental_Files

gkae174_supplemental_files.zip^{(7.1MB, zip)}

Data Availability Statement

[B1] 1. Przybyla L., Gilbert L.A. A new era in functional genomics screens. Nat. Rev. Genet. 2022; 23:89–103. [DOI] [PubMed] [Google Scholar]

[B2] 2. Cain A.K., Barquist L., Goodman A.L., Paulsen I.T., Parkhill J., van Opijnen T. A decade of advances in transposon-insertion sequencing. Nat. Rev. Genet. 2020; 21:526–540. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B3] 3. Coe K.A., Lee W., Stone M.C., Komazin-Meredith G., Meredith T.C., Grad Y.H., Walker S. Multi-strain Tn-Seq reveals common daptomycin resistance determinants in Staphylococcus aureus. PLoS Pathog. 2019; 15:e1007862. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B4] 4. Jana B., Cain A.K., Doerrler W.T., Boinett C.J., Fookes M.C., Parkhill J., Guardabassi L. The secondary resistome of multidrug-resistant Klebsiella pneumoniae. Sci. Rep. 2017; 7:42483. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B5] 5. Shames S.R., Liu L., Havey J.C., Schofield W.B., Goodman A.L., Roy C.R. Multiple effector virulence phenotypes revealed through high-throughput analysis of targeted mutant libraries. Proc. Natl Acad. Sci. U.S.A. 2017; 114:E10446–E10454. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B6] 6. Warr A.R., Hubbard T.P., Munera D., Blondel C.J., Abel Zur Wiesch P., Abel S., Wang X., Davis B.M., Waldor M.K. Transposon-insertion sequencing screens unveil requirements for EHEC growth and intestinal colonization. PLoS Pathog. 2019; 15:e1007652. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B7] 7. Liu H., Shiver A.L., Price M.N., Carlson H.K., Trotter V.V., Chen Y., Escalante V., Ray J., Hern K.E., Petzold C.J. et al. Functional genetics of human gut commensal Bacteroides thetaiotaomicron reveals metabolic requirements for growth across environments. Cell Rep. 2021; 34:108789. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B8] 8. Mandal R.K., Kwon Y.M. Global screening of Salmonella enterica serovar Typhimurium genes for desiccation survival. Front. Microbiol. 2017; 8:1723. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B9] 9. Goodall E.C.A., Robinson A., Johnston I.G., Jabbari S., Turner K.A., Cunningham A.F., Lund P.A., Cole J.A., Henderson I.R. The essential genome of Escherichia coli K-12. mBio. 2018; 9:e02096-17. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B10] 10. Jiang W., Bikard D., Cox D., Zhang F., Marraffini L.A. RNA-guided editing of bacterial genomes using CRISPR–Cas systems. Nat. Biotechnol. 2013; 31:233–239. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B11] 11. Jiang Y., Qian F., Yang J., Liu Y., Dong F., Xu C., Sun B., Chen B., Xu X., Li Y. et al. CRISPR–Cpf1 assisted genome editing of Corynebacterium glutamicum. Nat. Commun. 2017; 8:15179. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B12] 12. Vento J.M., Crook N., Beisel C.L. Barriers to genome editing with CRISPR in bacteria. J. Ind. Microbiol. Biotechnol. 2019; 46:1327–1341. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B13] 13. Liu R., Liang L., Freed E.F., Choudhury A., Eckert C.A., Gill R.T. Engineering regulatory networks for complex phenotypes in E. coli. Nat. Commun. 2020; 11:4050. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B14] 14. Garst A.D., Bassalo M.C., Pines G., Lynch S.A., Halweg-Edwards A.L., Liu R., Liang L., Wang Z., Zeitoun R., Alexander W.G. et al. Genome-wide mapping of mutations at single-nucleotide resolution for protein, metabolic and genome engineering. Nat. Biotechnol. 2016; 35:48–55. [DOI] [PubMed] [Google Scholar]

[B15] 15. Cui L., Bikard D. Consequences of Cas9 cleavage in the chromosome of Escherichia coli. Nucleic Acids Res. 2016; 44:4243–4251. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B16] 16. Qi L.S., Larson M.H., Gilbert L.A., Doudna J.A., Weissman J.S., Arkin A.P., Wendell A Repurposing CRISPR as an RNA-guided platform for sequence-specific control of gene expression. Cell. 2013; 152:1173–1183. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B17] 17. Bikard D., Jiang W., Samai P., Hochschild A., Zhang F., Marraffini L.A. Programmable repression and activation of bacterial gene expression using an engineered CRISPR–Cas system. Nucleic Acids Res. 2013; 41:7429–7437. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B18] 18. Li X.-T., Jun Y., Erickstad M.J., Brown S.D., Parks A., Court D.L., Jun S. tCRISPRi: tunable and reversible, one-step control of gene expression. Sci. Rep. 2016; 6:39076. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B19] 19. Fontana J., Dong C., Ham J.Y., Zalatan J.G., Carothers J.M. Regulated expression of sgRNAs tunes CRISPRi in E. coli. Biotechnol. J. 2018; 13:e1800069. [DOI] [PubMed] [Google Scholar]

[B20] 20. Peters J.M., Colavin A., Shi H., Czarny T.L., Larson M.H., Wong S., Hawkins J.S., Lu C.H.S., Koo B.-M., Marta E. et al. A comprehensive, CRISPR-based functional analysis of essential genes in bacteria. Cell. 2016; 165:1493–1506. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B21] 21. Lee H.H., Ostrov N., Wong B.G., Gold M.A., Khalil A.S., Church G.M. Functional genomics of the rapidly replicating bacterium Vibrio natriegens by CRISPRi. Nat. Microbiol. 2019; 4:1105–1113. [DOI] [PubMed] [Google Scholar]

[B22] 22. Cui L., Vigouroux A., Rousset F., Varet H., Khanna V., Bikard D. A CRISPRi screen in E. coli reveals sequence-specific toxicity of dCas9. Nat. Commun. 2018; 9:1912. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B23] 23. Komor A.C., Kim Y.B., Packer M.S., Zuris J.A., Liu D.R. Programmable editing of a target base in genomic DNA without double-stranded DNA cleavage. Nature. 2016; 533:420–424. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B24] 24. Nishida K., Arazoe T., Yachie N., Banno S., Kakimoto M., Tabata M., Mochizuki M., Miyabe A., Araki M., Hara K.Y. et al. Targeted nucleotide editing using hybrid prokaryotic and vertebrate adaptive immune systems. Science. 2016; 353:aaf8729. [DOI] [PubMed] [Google Scholar]

[B25] 25. Mol C.D., Arvai A.S., Sanderson R.J., Slupphaug G., Kavli B., Krokan H.E., Mosbaugh D.W., Tainer J.A. Crystal structure of human uracil-DNA glycosylase in complex with a protein inhibitor: protein mimicry of DNA. Cell. 1995; 82:701–708. [DOI] [PubMed] [Google Scholar]

[B26] 26. Tang W., Liu D.R. Rewritable multi-event analog recording in bacterial and mammalian cells. Science. 2018; 360:eaap8992. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B27] 27. Zhao D., Li J., Li S., Xin X., Hu M., Price M.A., Rosser S.J., Bi C., Zhang X. Glycosylase base editors enable C-to-A and C-to-G base changes. Nat. Biotechnol. 2020; 39:35–40. [DOI] [PubMed] [Google Scholar]

[B28] 28. Tong H., Liu N., Wei Y., Zhou Y., Li Y., Wu D., Jin M., Cui S., Li H., Li G. et al. Programmable deaminase-free base editors for G-to-Y conversion by engineered glycosylase. Natl Sci. Rev. 2023; 10:nwad143. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B29] 29. Grünewald J., Zhou R., Lareau C.A., Garcia S.P., Iyer S., Miller B.R., Langner L.M., Hsu J.Y., Aryee M.J., Joung J.K. A dual-deaminase CRISPR base editor enables concurrent adenine and cytosine editing. Nat. Biotechnol. 2020; 38:861–864. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B30] 30. Banno S., Nishida K., Arazoe T., Mitsunobu H., Kondo A. Deaminase-mediated multiplex genome editing in Escherichia coli. Nat. Microbiol. 2018; 3:423–429. [DOI] [PubMed] [Google Scholar]

[B31] 31. Zheng K., Wang Y., Li N., Jiang F.-F., Wu C.-X., Liu F., Chen H.-C., Liu Z.-F. Highly efficient base editing in bacteria using a Cas9–cytidine deaminase fusion. Commun. Biol. 2018; 1:32. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B32] 32. Gu T., Zhao S., Pi Y., Chen W., Chen C., Liu Q., Li M., Han D., Ji Q. Highly efficient base editing in using an engineered CRISPR RNA-guided cytidine deaminase. Chem. Sci. 2018; 9:3248–3253. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B33] 33. Wang Y., Wang S., Chen W., Song L., Zhang Y., Shen Z., Yu F., Li M., Ji Q. CRISPR–Cas9 and CRISPR-assisted cytidine deaminase enable precise and efficient genome editing in Klebsiella pneumoniae. Appl. Environ. Microbiol. 2018; 84:e01834-18. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B34] 34. Wang Y., Liu Y., Liu J., Guo Y., Fan L., Ni X., Zheng X., Wang M., Zheng P., Sun J. et al. MACBETH: multiplex automated Corynebacterium glutamicum base editing method. Metab. Eng. 2018; 47:200–210. [DOI] [PubMed] [Google Scholar]

[B35] 35. Volke D.C., Martino R.A., Kozaeva E., Smania A.M., Nikel P.I. Modular (de)construction of complex bacterial phenotypes by CRISPR/nCas9-assisted, multiplex cytidine base-editing. Nat. Commun. 2022; 13:3026. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B36] 36. Xu P., Liu Z., Liu Y., Ma H., Xu Y., Bao Y., Zhu S., Cao Z., Wu Z., Zhou Z. et al. Genome-wide interrogation of gene functions through base editor screens empowered by barcoded sgRNAs. Nat. Biotechnol. 2021; 39:1403–1413. [DOI] [PubMed] [Google Scholar]

[B37] 37. Hanna R.E., Hegde M., Fagre C.R., DeWeirdt P.C., Sangree A.K., Szegletes Z., Griffith A., Feeley M.N., Sanson K.R., Baidi Y. et al. Massively parallel assessment of human variants with base editor screens. Cell. 2021; 184:1064–1080. [DOI] [PubMed] [Google Scholar]

[B38] 38. Kweon J., Jang A.-H., Shin H.R., See J.-E., Lee W., Lee J.W., Chang S., Kim K., Kim Y. A CRISPR-based base-editing screen for the functional assessment of BRCA1 variants. Oncogene. 2020; 39:30–35. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B39] 39. Liu Y., Wang R., Liu J., Lu H., Li H., Wang Y., Ni X., Li J., Guo Y., Ma H. et al. Base editor enables rational genome-scale functional screening for enhanced industrial phenotypes in Corynebacterium glutamicum. Sci. Adv. 2022; 8:eabq2157. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B40] 40. Billon P., Bryant E.E., Joseph S.A., Nambiar T.S., Hayward S.B., Rothstein R., Ciccia A. CRISPR-mediated base editing enables efficient disruption of eukaryotic genes through induction of STOP codons. Mol. Cell. 2017; 67:1068–1079. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B41] 41. Kuscu C., Parlak M., Tufan T., Yang J., Szlachta K., Wei X., Mammadov R., Adli M. CRISPR-STOP: gene silencing through base-editing-induced nonsense mutations. Nat. Methods. 2017; 14:710–712. [DOI] [PubMed] [Google Scholar]

[B42] 42. Marquart K.F., Allam A., Janjuha S., Sintsova A., Villiger L., Frey N., Krauthammer M., Schwank G. Predicting base editing outcomes with an attention-based deep learning algorithm trained on high-throughput target library screens. Nat. Commun. 2021; 12:5114. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B43] 43. Neidhardt F.C., Bloch P.L., Smith D.F. Culture medium for enterobacteria. J. Bacteriol. 1974; 119:736–747. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B44] 44. Miller J.H. Assay of β-galactosidase. Experiments in Molecular Genetics. 1972; Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press. [Google Scholar]

[B45] 45. Jiang H., Wong W.H. SeqMap: mapping massive amount of oligonucleotides to the genome. Bioinformatics. 2008; 24:2395–2396. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B46] 46. Robinson M.D., McCarthy D.J., Smyth G.K. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics. 2009; 26:139–140. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B47] 47. Risso D., Ngai J., Speed T.P., Dudoit S. Normalization of RNA-seq data using factor analysis of control genes or samples. Nat. Biotechnol. 2014; 32:896–902. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B48] 48. Feurer M., Klein A., Eggensperger K., Springenberg J., Blum M., Hutter F.. Cortes C., Lawrence N.D., Lee D.D., Sugiyama M., Garnett R. Efficient and robust automated machine learning. Advances in Neural Information Processing Systems 28. 2015; Red Hook, NY: Curran Associates, Inc; 2962–2970. [Google Scholar]

[B49] 49. Yu Y., Gawlitt S., de Andrade E Sousa L.B., Merdivan E., Piraud M., Beisel C.L., Barquist L Improved prediction of bacterial CRISPRi guide efficiency from depletion screens through mixed-effect machine learning and data integration. Genome Biol. 2024; 25:13. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B50] 50. Keseler I.M., Gama-Castro S., Mackie A., Billington R., Bonavides-Martínez C., Caspi R., Kothari A., Krummenacker M., Midford P.E., Muñiz-Rascado L. et al. The EcoCyc Database in 2021. Front Microbiol. 2021; 12:711077. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B51] 51. Baba T., Ara T., Hasegawa M., Takai Y., Okumura Y., Baba M., Datsenko K.A., Tomita M., Wanner B.L., Mori H. Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Mol. Syst. Biol. 2006; 2:2006.0008. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B52] 52. Pedregosa F., Varoquaux G., Gramfort A., Michel V., Thirion B., Grisel O., Blondel M., Prettenhofer P., Weiss R., Dubourg V. et al. Scikit-learn: machine learning in Python. J. Mach. Learn. Res. 2011; 12:2825–2830. [Google Scholar]

[B53] 53. Lundberg S.M., Erion G., Chen H., DeGrave A., Prutkin J.M., Nair B., Katz R., Himmelfarb J., Bansal N., Lee S.-I. From local explanations to global understanding with explainable AI for trees. Nat. Mach. Intell. 2020; 2:56–67. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B54] 54. Chatterjee P., Jakimo N., Jacobson J.M. Minimal PAM specificity of a highly similar SpCas9 ortholog. Sci. Adv. 2018; 4:eaau0766. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B55] 55. Cho S., Choe D., Lee E., Kim S.C., Palsson B., Cho B.-K. High-level dCas9 expression induces abnormal cell morphology in Escherichia coli. ACS Synth. Biol. 2018; 7:1085–1094. [DOI] [PubMed] [Google Scholar]

[B56] 56. Rees H.A., Liu D.R. Base editing: precision chemistry on the genome and transcriptome of living cells. Nat. Rev. Genet. 2018; 19:770–788. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B57] 57. Qi L.S., Larson M.H., Gilbert L.A., Doudna J.A., Weissman J.S., Arkin A.P., Lim W.A. Repurposing CRISPR as an RNA-guided platform for sequence-specific control of gene expression. Cell. 2021; 184:844. [DOI] [PubMed] [Google Scholar]

[B58] 58. Chen S., Xie W., Liu Z., Shan H., Chen M., Song Y., Yu H., Lai L., Li Z. CRISPR start-loss: a novel and practical alternative for gene silencing through base-editing-induced start codon mutations. Mol. Ther. Nucleic Acids. 2020; 21:1062–1073. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B59] 59. Hecht A., Glasgow J., Jaschke P.R., Bawazer L.A., Munson M.S., Cochran J.R., Endy D., Salit M. Measurements of translation initiation from all 64 codons in E. coli. Nucleic Acids Res. 2017; 45:3615–3626. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B60] 60. Munson L.M., Stormo G.D., Niece R.L., Reznikoff W.S. lacZ translation initiation mutations. J. Mol. Biol. 1984; 177:663–683. [DOI] [PubMed] [Google Scholar]

[B61] 61. Depardieu F., Bikard D. Gene silencing with CRISPRi in bacteria and optimization of dCas9 expression levels. Methods. 2020; 172:61–75. [DOI] [PubMed] [Google Scholar]

[B62] 62. Wang T., Guan C., Guo J., Liu B., Wu Y., Xie Z., Zhang C., Xing X.-H. Pooled CRISPR interference screening enables genome-scale functional genomics study in bacteria with superior performance. Nat. Commun. 2018; 9:2475. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B63] 63. Thuronyi B.W., Koblan L.W., Levy J.M., Yeh W.-H., Zheng C., Newby G.A., Wilson C., Bhaumik M., Shubina-Oleinik O., Holt J.R. et al. Continuous evolution of base editors with expanded target compatibility and improved activity. Nat. Biotechnol. 2019; 37:1070–1079. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B64] 64. Lundberg S., Lee S.-I. A unified approach to interpreting model predictions. Adv. Neural Inf. Process. Syst. 2017; 30:4765–4774. [Google Scholar]

[B65] 65. Gomaa A.A., Klumpe H.E., Luo M.L., Selle K., Barrangou R., Beisel C.L. Programmable removal of bacterial strains by use of genome-targeting CRISPR–Cas systems. mBio. 2014; 5:e00928-13. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B66] 66. Chatterjee P., Jakimo N., Lee J., Amrani N., Rodríguez T., Koseki S.R.T., Tysinger E., Qing R., Hao S., Sontheimer E.J. et al. An engineered ScCas9 with broad PAM range and high specificity and activity. Nat. Biotechnol. 2020; 38:1154–1158. [DOI] [PubMed] [Google Scholar]

[B67] 67. Xin X., Li J., Zhao D., Li S., Xie Q., Li Z., Fan F., Bi C., Zhang X. Double-check base editing for efficient A to G conversions. ACS Synth. Biol. 2019; 8:2629–2634. [DOI] [PubMed] [Google Scholar]

[B68] 68. Joyce A.R., Reed J.L., White A., Edwards R., Osterman A., Baba T., Mori H., Lesely S.A., Palsson B.Ø., Agarwalla S. Experimental and computational assessment of conditionally essential genes in Escherichia coli. J. Bacteriol. 2006; 188:8259–8271. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B69] 69. Patrick W.M., Quandt E.M., Swartzlander D.B., Matsumura I. Multicopy suppression underpins metabolic evolvability. Mol. Biol. Evol. 2007; 24:2716–2722. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B70] 70. Feist A.M., Henry C.S., Reed J.L., Krummenacker M., Joyce A.R., Karp P.D., Broadbelt L.J., Hatzimanikatis V., Palsson B.Ø. A genome-scale metabolic reconstruction for Escherichia coli K-12 MG1655 that accounts for 1260 ORFs and thermodynamic information. Mol. Syst. Biol. 2007; 3:121. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B71] 71. Walton R.T., Christie K.A., Whittaker M.N., Kleinstiver B.P. Unconstrained genome targeting with near-PAMless engineered CRISPR–Cas9 variants. Science. 2020; 368:290–296. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B72] 72. Nishimasu H., Shi X., Ishiguro S., Gao L., Hirano S., Okazaki S., Noda T., Abudayyeh O.O., Gootenberg J.S., Mori H. et al. Engineered CRISPR–Cas9 nuclease with expanded targeting space. Science. 2018; 361:1259–1262. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B73] 73. Kleinstiver B.P., Sousa A.A., Walton R.T., Tak Y.E., Hsu J.Y., Clement K., Welch M.M., Horng J.E., Malagon-Lopez J., Scarfò I. et al. Engineered CRISPR–Cas12a variants with increased activities and improved targeting ranges for gene, epigenetic and base editing. Nat. Biotechnol. 2019; 37:276–282. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B74] 74. Wang Y., Liu Y., Li J., Yang Y., Ni X., Cheng H., Huang T., Guo Y., Ma H., Zheng P. et al. Expanding targeting scope, editing window, and base transition capability of base editing in Corynebacterium glutamicum. Biotechnol. Bioeng. 2019; 116:3016–3029. [DOI] [PubMed] [Google Scholar]

[B75] 75. Arbab M., Shen M.W., Mok B., Wilson C., Matuszek Ż., Cassa C.A., Liu D.R. Determinants of base editing outcomes from target library analysis and machine learning. Cell. 2020; 182:463–480. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B76] 76. Song M., Kim H.K., Lee S., Kim Y., Seo S.-Y., Park J., Choi J.W., Jang H., Shin J.H., Min S. et al. Sequence-specific prediction of the efficiencies of adenine and cytosine base editors. Nat. Biotechnol. 2020; 38:1037–1043. [DOI] [PubMed] [Google Scholar]

[B77] 77. Dandage R., Després P.C., Yachie N., Landry C.R. A computational workflow for designing libraries of guide RNAs for CRISPR-mediated base editing. Genetics. 2019; 212:377–385. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B78] 78. Li Q., Seys F.M., Minton N.P., Yang J., Jiang Y., Jiang W., Yang S. CRISPR–Cas9 nickase-assisted base editing in the solvent producer Clostridium beijerinckii. Biotechnol. Bioeng. 2019; 116:1475–1483. [DOI] [PubMed] [Google Scholar]

[B79] 79. Tong Y., Whitford C.M., Robertsen H.L., Blin K., Jørgensen T.S., Klitgaard A.K., Gren T., Jiang X., Weber T., Lee S.Y. Highly efficient DSB-free base editing for Streptomycetes with CRISPR-BEST. Proc. Natl Acad. Sci. U.S.A. 2019; 116:20366–20375. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B80] 80. Hao W., Cui W., Cheng Z., Han L., Suo F., Liu Z., Zhou L., Zhou Z. Development of a base editor for protein evolution via in situ mutation in vivo. Nucleic Acids Res. 2021; 49:9594–9605. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B81] 81. Després P.C., Dubé A.K., Seki M., Yachie N., Landry C.R. Perturbing proteomes at single residue resolution using base editing. Nat. Commun. 2020; 11:1871. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B82] 82. Cheng L., Li Y., Qi Q., Xu P., Feng R., Palmer L., Chen J., Wu R., Yee T., Zhang J. et al. Single-nucleotide-level mapping of DNA regulatory elements that control fetal hemoglobin expression. Nat. Genet. 2021; 53:869–880. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B83] 83. Li J., Liu Y., Wang Y., Yu P., Zheng P., Wang M. Optimization of base editing in Corynebacterium glutamicum. Chin. J. Biotechnol. 2020; 36:143–151. [DOI] [PubMed] [Google Scholar]

[B84] 84. Ding X.-Y., Li S.-S., Geng Y.-M., Yan M.-Y., Li G.-B., Zhang G.-L., Sun Y.-C. Programmable base editing in using an engineered CRISPR RNA-guided cytidine deaminase. Front. Genome Ed. 2021; 3:734436. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B85] 85. Ipoutcha T., Rideau F., Gourgues G., Arfi Y., Lartigue C., Blanchard A., Sirand-Pugnet P. Genome editing of veterinary relevant mycoplasmas using a CRISPR–Cas base editor system. Appl. Environ. Microbiol. 2022; 88:e0099622. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B86] 86. Rodrigues S.D., Karimi M., Impens L., Van Lerberge E., Coussens G., Aesaert S., Rombaut D., Holtappels D., Ibrahim H.M.M., Van Montagu M. et al. Efficient CRISPR-mediated base editing in Agrobacterium spp. Proc. Natl Acad. Sci. U.S.A. 2021; 118:e2013338118. [DOI] [PMC free article] [PubMed] [Google Scholar]

[B87] 87. Edgar R., Domrachev M., Lash A.E. Gene Expression Omnibus: NCBI gene expression and hybridization array data repository. Nucleic Acids Res. 2002; 30:207–210. [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

Expanding the flexibility of base editing for high-throughput genetic screens in bacteria

Sandra Gawlitt

Scott P Collins

Yanying Yu

Samuel A Blackman

Lars Barquist

Chase L Beisel

Abstract

Graphical Abstract

Graphical Abstract.

Introduction

Materials and methods

Strains, plasmids and growth conditions

Golden Gate assembly reactions

Base editing in E. coli with ScBE3

Measuring GFP repression by flow cytometry

Figure 1.

GFP repression assay

β-Galactosidase activity assay

Two-step base editing and cell killing

Library design for genome-wide screening in E. coli

Library cloning and verification for genome-wide screening in E. coli

Genome-wide screening in E. coli

Library design for screening conditionally essential genes

Library cloning and verification for screening conditionally essential genes

Screening of conditionally essential E. coli genes

Library sequencing

Sequencing data processing and machine learning models

Growth experiments

Evaluation of the editing efficiency and bystander edits with ScBE3

Investigating dual usage of the sgRNAs in the dual system

Statistical analyses

Results

An ScCas9 cytidine base editor enables flexible introduction of premature stop codons

Targeting with ScBE3 does not cause notable repression of transcription

Mutating the start codon serves as an alternative strategy for gene disruption

Figure 2.

A genome-wide essentiality screen in E. coli reveals target preferences of ScBE3

Figure 3.

Combining base editing with Cas9-induced cell killing increases the apparent editing efficiency

Figure 4.

Applying the dual system to a high-throughput screen that decouples editing and phenotyping screening

Figure 5.

A screen decoupled from editing reveals E. coli genes conditionally essential in minimal media

Discussion

Supplementary Material

Acknowledgements

Contributor Information

Data availability

Supplementary data

Funding

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases