MainBacteriophages (phages) are the most abundant and genetically diverse entities on Earth, driving a genetic arms race with their bacterial hosts that continually alters microbial life, shaping both human health and the biosphere1,2. This phage–host arms race continually generates new protein functions encoded by uncharacterized genes that constitute a large source of genes of unknown function in the biosphere3,4. To expedite characterization of the vastly underexplored genetic content of phage genomes, genome-wide experimental approaches are needed.Genome-scale CRISPRi (clustered regularly interspaced short palindromic repeats interference) methods are a common starting point for probing gene functions in diverse organisms by programmably blocking transcription using a nuclease-deactivated Cas9 or Cas12 (dCas9/dCas12)5,6,7,8,9,10,11. Recently, a dCas12-based DNA-targeting CRISPRi method was used to laboriously map essential genes for two model temperate phages one gene at a time12, but the arrayed assay format is cumbersome to scale. Furthermore, studies on nuclease-active Cas9 and Cas12 systems suggest limitations with DNA-targeting CRISPR systems when extended to lytic phages with distinct lifestyles, genomic content and genome modifications13,14,15,16,17,18,19,20,21. However, phage transcripts appear generally targetable and vulnerable during infection21. We posited that the RNA-guided RNA-binding protein dRfxCas13d (HEPN-deactivated Ruminococcus flavefaciens Cas13d, dCas13d)22 could be applied as a universal tool for targeted inhibition of phage protein expression, including for RNA phages and nucleus-forming phages where DNA-binding tools are completely ineffective23.Here we present CRISPR interference through antisense RNA targeting (CRISPRi-ART) as a robust method for suppressing protein expression. By targeting dCas13d to phage transcript-encoded ribosome-binding sites (RBS), we could achieve targeted gene expression knockdown in diverse phages. Through pooled CRISPRi-ART libraries, we implemented transcriptome-wide CRISPRi-ART screens against diverse coliphages at unprecedented scale. We identified many previously unknown phage genes critical for infection, establishing a platform for high-throughput discovery and prioritizing genes for future study.ResultsTargeting dCas13d to bacterial RBSs represses protein expressionTo determine the principles governing translational repression by dCas13d binding to target messenger RNA (mRNA) sequences (Fig. 1a)22, we systematically identified the regions within mRNA transcripts that are most susceptible to dCas13d-mediated translational repression. The PAM-less nature of dCas13d24 enabled use of a pooled, single-nucleotide-resolution CRISPR RNA (crRNA) library (Supplementary Fig. 1) under the crystal violet (CV)-inducible pJEx promoter, tiled across 18 E. coli transcripts, most of which encode at least one essential gene (Fig. 1b). This 29,473-crRNA library was transformed into cells expressing dCas13d under the aTc-inducible pTet promoter, and 15 cell doublings after induction, samples were Illumina sequenced to quantify changes in crRNA abundance between the initial and final timepoints. This competitive growth assay revealed a major dCas13d-dependent fitness defect for crRNAs binding near (within ~70 nucleotides (nt)) the ribosome-binding site (RBS) located near the start codon of the targeted essential genes (Supplementary Data 6), often producing fitness defects greater than 100-fold (Methods and Supplementary Fig. 2) (Fig. 1c,d and Supplementary Fig. 3). Notably, targeting the RBS region of known non-essential genes did not impair growth (Supplementary Fig. 4). Having identified RBS susceptibility to dCas13d targeting, we next aimed to determine whether CRISPRi-ART could inhibit phage infection through RBS targeting.Fig. 1: Design rules for dCas13d targeting by single-nucleotide-resolution profiling of E. coli essential genes.a, Overview of CRISPRi-ART. dRfxCas13d binding near the RBS reduces protein expression through inhibition of translation initiation by the 16S ribosomal subunit. b, A CRISPRi-ART (top) crRNA library tiled at single-nucleotide resolution against E. coli transcripts encoding essential genes (bottom) is used to identify regions susceptible to translational knockdown. Here, dCas13d expression is driven by the pTet promoter under aTc-inducible control, with its crRNA driven by the pJEx promoter under crystal violet-inducible control. E. coli growth is inhibited if dCas13d targets an essential gene’s RBS, while growth is unimpeded if the targeted protein is dispensable. c, The measured log2(fold-change) of guide abundances targeting a representative transcript encoding an essential gene is plotted. CRISPRi-ART-dependent fitness effects are presented in blue and crRNA-only controls in light grey (n = 3 independent replicates). d, The observed log2(fold-change) of crRNA abundances targeting the RBS region of 9 essential genes is compiled into a single plot, where points represent the centre of the crRNA spacer. Highlighted is the 100 bp surrounding the start codon (see Supplementary Fig. 3 showing a wider targeting range). Average of values at each nucleotide position are plotted across the region, along with a 95% confidence interval.Source dataFull size imageCRISPRi-ART enables targeted disruption of phage infectionTo assess whether CRISPRi-ART can discriminate phage gene essentiality through RBS targeting (Fig. 2a)21, we applied phage T4 to E. coli expressing inducible dCas13d and constitutive crRNA (Methods). We hypothesized that targeting non-essential phage genes would result in no loss of infectivity, while targeting essential phage genes would lead to a reduction in infectivity (Fig. 2b). We found that crRNAs targeting essential T4 genes25 consistently reduced the efficiency of plaquing (EOP) by 102–104-fold compared with no inhibition for crRNA targeting a non-essential gene (Fig. 2c and Supplementary Fig. 5). Inhibition occurred only upon dCas13d induction, suggesting that protein knockdown is not due to leaky dCas13d expression (Supplementary Fig. 6). To compare the performance of CRISPRi-ART to previously established double stranded (ds)DNA-targeting CRISPRi tools, we assessed the efficiency of dLbCas12a- and dSpyCas9-mediated phage inhibition. Targeting of essential T4 genes with dsDNA-targeting dLbCas12a (Fig. 2d and Supplementary Fig. 7) or dSpyCas9 (Fig. 2e and Supplementary Fig. 8) resulted in minimal anti-phage activity.Fig. 2: Translational repression provides a simple means to probe phage gene essentiality.a, Phage-encoded genome protection strategies. Phage genomes can be constituted by ssRNA+ (green), ssDNA+ (blue), dsRNA (green/purple) or dsDNA (blue/red) molecules (left), heavily modified (centre) or compartmentalized (right) with example phages tested in this study. In all cases, phage mRNA (green) is accessible to Cas13 targeting. Phage genomes not drawn to scale. b, Overview of CRISPRi-ART-mediated phage defence. dCas13d expression is driven by the pTet promoter under aTc-inducible control with its crRNA constitutively expressed. Phage infection is inhibited if dCas13d targets an essential phage transcript’s RBS, while infection is productive if the targeted protein is dispensable. Plaque images shown are cartoon illustrations representative of collected data across Figs. 3 and 4. c, EOP assays for CRISPRi-ART-mediated phage defence when targeting phage T4 genes. d, EOP assays for DNA-targeting dCas12a targeting phage T4 genes. e, EOP assays for DNA-targeting dCas9 targeting phage T4 genes. Grey bars: a negative control crRNA; dark red bars, a known T4-essential gene targeting crRNA; dark blue bars: a known T4-non-essential gene targeting crRNA. All EOP values represent the average of 3 biological replicates at 20 nM aTc dCas13d or dSpyCas9 induction or 5 nM aTc dLbCas12a induction. EOP data are presented as mean ± s.d. Minus symbols denote a consistent, ≥4-fold plaque size reduction phenotype if plaques were observed.Source dataFull size imagePolar effects, where gene repression yields additional repression of downstream genes in the operon, are known sources of false positive assignments of phage gene fitness using CRISPRi12. To explore these effects in CRISPRi-ART, we first targeted essential gene O in the well-characterized Lamba PR transcript and observed a large reduction in EOP. Next, we complemented O and observed full recovery of EOP (Supplementary Fig. 9), indicating that O knockdown does not prohibit expression of downstream essential gene Q. This result contrasts with the recent application of dLbCas12a-based DNA-targeting CRISPRi to the same Lambda transcript, which led to the misclassification of non-essential nin genes between O and Q as essential12. We conclude that CRISPRi-ART avoids such polar effects, providing a notable advantage in accurately assigning gene essentiality.CRISPRi-ART is broadly effective across E. coli phage phylogenyTo test whether CRISPRi-ART is applicable across diverse bacteriophages, we applied it to 12 coliphages including single-stranded (ss)RNA+, ssDNA+, dsDNA, chemically modified and compartmentalized genomes, as well as temperate, chronic and lytic lifestyles (Fig. 3a and Supplementary Data 8)13,16,17,18,20,25,26,27. For each phage, we designed two crRNAs (crRNA1 and crRNA4) (Supplementary Fig. 1b) targeting an essential gene encoding the major capsid protein (MCP) and measured infection productivity via EOP and plaque size (Fig. 3b). At least one crRNA per phage caused a strong reduction in EOP and plaque size reduction (Supplementary Figs. 10a–j and 11). For a few phages (T7, T5, EdH4, SUSP1 and M13), effective crRNAs caused strong plaque size reduction without a major reduction in EOP. Overall, however, RNA-targeting CRISPRi-ART was far more consistent in its ability to restrict a diverse range of phages compared with DNA-targeting CRISPRi (Supplementary Figs. 12 and 13).Fig. 3: Translational repression is broadly active against phage diversity.a, Network graph representation of E. coli phages and their relatives21. Nodes represent phage genomes connected by edges if they share similarity determined by vContact2 (ref. 91). Red and blue nodes represent E. coli and non-E. coli phages, respectively. Phages assessed here for dCas13d sensitivity are shaded in black. b, Anti-phage activity conferred by dCas13d when targeting an essential gene in 12 diverse phages (mean of 3 replicates), scored by EOP reduction (top) or plaque size (bottom). c, crRNA multiplexing facilitates more efficient knockdown than component guides, using phage T7 as an example. d, Transcriptome-wide knockdown screen in ssRNA phage MS2 using RBS-targeting guides. The best of two guides tested is shown. -con is a non-targeting crRNA control. dCas13d was induced as described in Methods.Source dataFull size imageA consistent observation for phages targeted with CRISPRi-ART was a reduction in plaque size when targeting genes expected to affect phage fitness (Fig. 3b and Supplementary Fig. 10j). To test whether repressing two essential genes in these phages would enhance infection inhibition, we employed crRNAs targeting two essential genes either individually or in combination (Fig. 3c). Although each individual crRNA reduced plaque size without EOP reduction, we observed near-complete elimination of plaque formation when using both crRNAs simultaneously. These results suggest that crRNA multiplexing can have synergistic effects.We next used CRISPRi-ART to analyse the short, 3.6 kb ssRNA genome28 of phage MS2, which replicates without DNA intermediates26 and thus evades DNA-based tools (Fig. 3b). At least one crRNA targeting each of the four known MS2 genes inhibited infection, while crRNAs targeting inside the coding sequence (CDS) but outside of the susceptible RBS region (Fig. 1d and Supplementary Fig. 1b) on either +sense or −sense RNA strands did not, ruling out direct obstruction of genome synthesis (Fig. 3d and Supplementary Fig. 11a,b). Differences in magnitude of knockdown may reflect a limitation of the crRNAs tested or differential sensitivities to CRISPRi-ART (Supplementary Fig. 11c). Together, these results demonstrate the ability to perform transcriptome-wide knockdown screens in diverse phages (Figs. 2a and 3).We also tested CRISPRi-ART in four diverse E. coli strains sensitive to phage PTXU04, extending applicability to diverse wild-type hosts. CRISPRi-ART achieved substantial reduction of PTXU04 EOP in all four ECOR strains when targeting essential phage genes relative to a non-targeting control crRNA (Supplementary Fig. 14). These results demonstrate that CRISPRi-ART can be successfully applied to genetically diverse E. coli strains.CRISPRi-ART uncovers diverse superinfection immunity suppressorsWe next used CRISPRi-ART to investigate the role of the widespread yet poorly understood genetic module rII in subverting RexAB-based superinfection immunity encoded by lambda lysogens. The RexAB system protects against superinfecting phages by inducing membrane depolarization and growth arrest upon detection of phage infection. Phage-encoded RIIA and RIIB proteins counteract RexAB29,30, whereas loss-of-function mutants of rIIA and rIIB render T4 susceptible to RexAB superinfection immunity (Fig. 4a)31. Nearly 7 decades after the discovery of these systems29, the specificity of this phage–host interaction remains unclear. Given the low sequence identity between diverse rIIAB and rIIAB-like genes (Fig. 4b), we wondered whether divergent rII systems counteract the Lambda Rex exclusion system or have adapted to preferentially suppress distinct Rex or other immune systems. We confirmed that CRISPRi-ART knockdown of RIIA or RIIB encoded by T4, MM02, EdH4, SUSP1 and N4 phages—spanning four distinct subfamilies and five genera—does not inhibit their infection of E. coli lacking Rex, suggesting that these genes are not broadly critical for efficient infection in the absence of Rex-encoding prophages (Fig. 4c,d and Supplementary Figs. 15–20). In contrast, CRISPRi-ART knockdown of RIIA and RIIB reduced EOP during infection of E. coli expressing Lambda RexAB, indicating that divergent rII systems ( CD-SEARCH > HHpred. Any deviations were made on the basis of annotation detail and annotation confidence.In addition to the above annotations, genes were assigned ‘class’ and ‘annotation quality’ scores. The ‘class’ annotation included the following annotations: anti-defence, chaperonin, lysis, nucleotide metabolism, replication, transcription, translation, tRNA, virion, or unknown/other. ‘Anti-defence’ refers to genes involved in subverting phage-defence systems, including restriction modification systems. ‘Chaperonin’ refers to genes involved in phage virion or protein maturation, but not a structural component of the phage virion. ‘Lysis’ refers to genes involved in lysis, regulation of lysis timing or degradation of cell wall components. ‘Nucleotide metabolism’ refers to genes responsible for nucleotide biosynthesis, degradation, modification and regulation thereof, but not directly a part of replication. ‘Replication’ refers to genes involved in phage replication liberally applied. ‘Transcription’ refers to genes that modulate transcription in either the phage or host genome. ‘Translation’ refers to genes that modulate translation, including genes that modulate RNA or tRNA stability. ‘tRNA’ refers to tRNA genes, but not genes that modify them. ‘Virion’ refers to genes that are structural components or part of the virion produced in infection. ‘Unknown/other’ refers to all other genes encoded in phage. ‘Annotation quality’ was assigned manually based on both confidence, detail and known literature of the annotation and its source content: ‘known’ for genes with known function, ‘ambiguous’ for known genes with ambiguity to substrate or role of the gene, and ‘unknown’ for genes of unknown function.In general, these were in agreement with PHROG category, with the following exceptions for visualization simplicity: (1) all predicted phage structural components were grouped into the ‘virion’ category including packaged phage proteins; (2) many genes that are critical for phage lifecycle but of unknown molecular function (for example, T5 genes A1 and A2) were grouped into ‘replication’; (3) any gene responsible for assisting folding or assembly was overridden to fall under the ‘chaperonins’ category; (4) genes responsible for anti-phage defence through nucleotide modification were labelled as ‘nucleotide metabolism’; (5) genes with overlapping category functions (for example, RNase H) were labelled with a primary annotation on the basis of literature25 and (6) predicted subgenomic mobile elements (for example, homing endonucleases) were assigned ‘unknown/other’ for simplicity. All phage annotations are listed in Supplementary Data 6.Analysis of phage CRISPRi-ART-seqFollowing CRISPRi-ART-seq processing, phage genes were interpreted for fitness. To identify FC thresholds for Fit and Semi-fit genes, for each phage–MOI condition, we identified the lowest fitness score with a K–S P value less than 0.05 on the right tail of the fitness distribution (that is, positively enriched). This value was used as an inclusive FC threshold for fitness. Thus, we determined Fit genes through the following metrics: T4 (10 MOI) (FC > 0.7, P −3, P 1.2, P −1.1, P