nature biotechnology 3 Article https://doi.org/10.1038/s41587-023-01759-y Engineered phage with antibacterial CRISPR-Cas selectively reduce £. coli burden in mice Received: 31 May 2022 Yilmaz Emre Gencay©1,6, Dziuginta Jasinskyte©1,6, Camille Robert1,6, Szabolcs Semsey1,6, Virginia Martinez©1,6, Anders Ostergaard Petersen1,6, Accepted- 22 March 2023 Katja Brunner1, Ana de Santiago Torio1, Alex Salazar1, Iszabela Cristiana Turcu1, Published online: 04 May 2023 Melissa Kviesgaard Eriksen1, Lev Koval1, Adam Takos©1, Ricardo Pascal1, —- Thea Staffeidt Schou1, Lone Bayer1, Tina Bryde1, Katja Chandelle Johansen1, j Check for updates_ Emilie Glad Bäk©1, Frenk Smrekar2, Timothy B. Doyle3, Michael J.Satlin4, Aurelie Gram1, Joana Carvalho1, Lene Jessen1, Björn Hallström1, Jonas Hink1, Birgitte Damholt1, Alice Troy1, Mette Grove1, Jasper Clube1, Christian Grondahl1, Jakob Krause Haaber1, Eric van der Helm©1, Milan Zdravkovic1 & Morten Otto Alexander Sommer© 1,5K Antibiotic treatments have detrimental effects on the microbiome and lead to antibiotic resistance. To develop a phage therapy against a diverse range of clinically relevant Escherichia coli, we screened a library of 162 wild-type (WT) phages, identifying eight phages with broad coverage off. coli, complementary binding to bacterial surface receptors, and the capability to stably carry inserted cargo. Selected phages were engineered with tail fibers and CRISPR-Cas machinery to specifically target E. coli. We show that engineered phages target bacteria in biofilms, reduce the emergence of phage-tolerant f. coli and out-compete their ancestral WT phages in coculture experiments. A combination of the four most complementary bacteriophages, called SNIPR001, is well tolerated in both mouse models and minipigs and reduces E. coli load in the mouse gut better than its constituent components separately. SNIPR001 is in clinical development to selectively kill E. coli, which may cause fatal infections in hematological cancer patients. Chemotherapeutic regimens used to treat hematological malignancies cause bone marrow suppression and gastrointestinal mucositis with associated increased intestinal permeability14. Translocation of gut bacteria, including Escherichia coli, from the gastrointestinal tract is a frequent cause of bloodstream infections5. The mortality related to bloodstream infections caused by enteric bacteria such as E. coli is 15-20%6; to decrease the chance of infection, antibiotics may be given before treatment in people at risk of low numbers of neutrophils in the blood7. Fluoroquinolones are used off-label in the United States, based on the results of two randomized trials demonstrating a decrease in bacterial infections in immunocompromised patients after use79. Fluoroquinolones have sideeffects, and their use in oncology patients 'SNIPR BIOME ApS, Copenhagen, Denmark. 2JAFRAL, Ljubljana, Slovenia. 3JMI Laboratories, North Liberty, IA, USA. "Division of Infectious Diseases, Weill Cornell Medicine, New York City, NY, USA. 5Novo Nordisk Foundation Center for Biosustainability, DTU Biosustain, Kongens Lyngby, Denmark. These authors contributed equally: Yilmaz Emre Gencay, Dziuginta Jasinskyte, Camille Robert, Szabolcs Semsey, Virginia Martinez, Anders 0stergaard Petersen. Sle-mail: ms@sniprbiome.com Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y cas3 and cascade genes CRISPR array (2-3 spacers) CAP leads ^fe $ AAA Efficacy screen Tail fiber engineering (1460, bl475 and bl813 where al5.2 outperformed WTal5 in the kinetic assays and subjected them to lawn kill assays. Indeed, al5.2 substantially led to a reduced number of survivors in comparison to WT al5 albeit with different levels per tested strain (Fig. 3b-d). Ten random purified colonies from WT al5 challenged group of each tested strain were as well tested for EoP with WT al5 and CAP al5.2. In accordance, results demonstrate a clear benefit of the tail fiber engineered txl5.2 over LPS-dependent WT al5, as al5 survivors mostly retained sensitivity to CAP al5.2 despite being resistant WT al5 (Fig. 3b-d, insets). CRISPR-Cas arming of phages to target E. coli To CRISPR-Cas arm the selected lytic phages and generate a library of CAPs, the type IE CRISPR-Cas system of E. coif''' was engineered (Supplementary Fig. 1) to target phylogenetically diversef. coli strains. A CRISPR-guided vector (CGV-EcCas) was generated, containing the cas3 gene (ygcB) and a downstream cascade gene complex encoded by casA (ygcL, cas8e), casB (ygcK, casll), casC (ygcj, casT), casD (ygcl, casS) and casEiygcH, cas6), and a CRISPR array targeting thef. coli genome (Fig. 1). To evaluate the killing efficiency of the CRISPR-Cas system, the CGV-EcCas was conjugated to E. coli strain b52, showing an average reduction of 3.5 log10 CFU ml"1, compared to theempty vector (Supplementary Fig. 3). Asexpected, no effect was observed after conjugating the CGV-EcCas to a nontargetf. coli strain (Supplementary Fig. 3). The killing efficiency of CGV-EcCas was further assessed on the abbreviated panel of 82E.coli strains. Conjugative delivery of the empty vector was accomplished in 75% of the isolates (Fig. 4a). For all strains where the CGV-EcCas was delivered, bacterial counts were reduced below the limit of detection (LOD, 200 CFU ml"1) corresponding to a reduction of 1-6 log10, highlighting the potent CRISPR-Cas-mediated killing (Fig. 4a). We aimed to engineer our CRISPR-Cas systems to be functional under restricted bacterial growth conditions, which have been Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y Fig. 31 Tail fiber engineering, a, EoP results of LPS-dependent WT al5, Tsx-dependent WT al7 and engineered CAP al5.2 that consolidates both WT phages' receptors. Presented titers (PFU mr1) were obtained from independent biological triplicates as dots, with averages illustrated as bars, b-d, Lawn kill assay results off. coffare shown as boxplots, whiskers indicate maximum and minimum values, box bounds indicate 25th and 75th percentile, with center line indicating the median; b!460 (b), b!475 (c), bl813 (d) with phages WT cd5 and CAP cd5.2. observed in the gut or in biofilms40. We tested two relevant promoters (Preaj41 and P6oM42) for their performance, both in planktonic cells grown in standard growth conditions (lysogeny broth (LB), 37 °C) and in biofilms, grown on peg lids in 96-well plates. Significant killing, measured as reduction of metabolic activity, was observed in E. coll biofilms when the CRISPR-Cas system was expressed from ?bolA compared to PrelB (Fig. 4b). As promoter ?bolA showed the best overall performance in the different conditions, it was chosen for transcription of the CRISPR-Cas system in the CAPs. The eight selected WT phages were CRISPR-Cas-armed to generate 15 CAP variations (Extended Data Table la). In addition to promoter PbM, the CRISPR-Cas systems were engineered to express from a synthetic constitutively expressed E. coll promoter (PJ23ioo) to further strengthen the CRISPR-Cas expression (Supplementary Fig. 1). CRISPR arrays were designed to target multiple virulence (spacers 1,2 and 3) or essential genes (spacers 4 and 5; Extended Data Table lb), as targeting multiple regions has been shown to prevent resistanceevolution43. To confirm the CRISPR-Casactivity in the CAPs, we measured the cas3 transcripts in samples obtained at 5,15 and 30 min following a synchronized infection with the equal multiplicity of infection (MOI) of CAP al5.2 in comparison to WTal5usingRT-qPCR and observed increasing levels of cas3 RNAonly upon CAP al5.2 infection (Supplementary Fig. 4). Next, we extended this assay to all four CAPs (ocl5.2, oc20.4, oc48.4 and oc51.5) and demonstrated increasing levels of cas3 transcripts highlighting that the CAPs expressed the CRISPR-Cas system during infection of a target strain (Fig. 4c-f). To demonstrate the competitive superiority of the CAPs, we performed competition experiments in which CAPs (a20.4 and al5.2) and their WT ancestral phages were cocultured with E. coll strain b230, serving as a target for both competing phages. Approximate initial ratios of 1 CAP to 9 WT phages were cocultured and passaged four times on fresh target cells in liquid cultures. After each passage, the relative abundance of CAP and WT phage particles was evaluated. Both CAPs outcompeted their WTs within four rounds; CAP a20.4 reached 68% after four rounds and CAP al5.2 reached 86% after two rounds (Fig. 4g-h),demonstratingan improved fitness compared to the WT phages. Selection and characterization of the optimal CAP cocktail The activity of the 15 CAPs was tested against the E. coll panel (n = 429) using the growth kinetics assay (Supplementary Fig. 5). The individual CAPs showed activity toward 4.1-29.4% of the strains tested. l-2 log10) lower than that of the parental strain. To maximize our coverage, we sought to rationally combine CAPs with a broad and complementary spectrum of activity. Thus, we made subsets of CAP cocktails based on our in silico predictions using individual per-formancesand tested their combinatorial in vitro performance. These results showed good compliance with our predictions (Supplementary Fig. 6). The initial 15 CAPs could be classified into four clusters based on their host-range profiles (Supplementary Fig. 5). We then excluded the seven lowest-ranking CAPs based on their redundant host-range in our cocktail predictions (Supplementary Fig. 7). Thus, eight CAPs (al5.2, al5.4, al7.2, a20.4, a46.4, a48.4, a51.5 and a51.6) were chosen for further assessments. First, all eight CAPs were individually orally dosed to mice (n = 3) and their normalized recovery (Supplementary Fig. 8) showed that all CAPs could be retrieved from fecal matter. Next, in vitro stability was assessed at accelerated conditions (40 °C, n = 3). Based on these results, two CAPs (al5.4, al7.2) were deselected as their titer dropped to below 1% of the starting material (Supplementary Fig. 9). The resulting six CAPs were individually tested (n = 6) in a mouse efficacy model (Supplementary Fig. 10), these results were combined with the predicted host range of the simulated cocktails (Supplementary Fig. 11; n = 15) and verification of complementing use of surface receptors for infection, resulting in the selection of CAPs al5.2, a20.4, a48.4 and a51.5 as the optimal CAPs for SNIPR001. The ancestors of CAPs ocl5.2, oc20.4, oc48.4 and oc51.5 are classified under the Tevenvlrlnae subfamily. Specifically, al5, a48 and a51 share 96.4%, 96.6% and 96.1% sequence similarity to E. coll phage T2, respectively, whereas a20's closest relative is E. coll phage RB69 (96.8%; Supplementary Fig. 12). In silico analyses of the genomes of SNIPR001 showed that the CAPs encode no known transposase or integrase genes, indicating that the phages are not temperate, and thus not predicted to be capable of inserting their DNA in bacterial cells. In addition, we observed no antimicrobial resistance markers or virulence genes in the phage genomes (Supplementary Table2). We investigated whether SNIPR001 CAPs cause generalized transduction and found no evidence of transduction with the LOD of 2 x 10~7 for frequency of transduction (Supplementary Table 3). Developing a drug product from individual CAPs Manufacturing a stable drug product comprised of four engineered phage particles requires establishinga phage and bacterial host collection, creating a Bacterial Master Cell Bank and a Master Phage Seed and turning the four resulting individual drug substances into a final Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y a 106 105 ■ TD1 104 ■ ZD U- O r 103 ■ 102 ■ 101 I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I Strain o 60 % 106 104 102 7= 10° 108 106 104 102 10° C (X15.2 ^ 0.05, two-sided Student's r-test, Nature Biotechnology 4848485353532348232348482353232323234848482348 Article https://doi.org/10.1038/s41587-023-01759-y JMI panel n = 382 100% - 8^ 80% - 60% - • 40% - O 20% -0% - SNIPR panel n = 429 • OO 1 80% 60% 40% 20% 0% 100% 92.2% 91.7% AMR type Bacteremia Plaquing result I Lysis zone I Negative I Plaques Fig. 51 In vitro validation of SNIPR001 on clinical E. coli strains, a, An unrooted phylogenetictree of theJMI strains displaying a clinical panel of382 E. coli strains encompassing nine phylogroups and 118 MLSTs. Plaquingdata reflects a single plaquing repl icate. One strain, E. coli b4038, with a long branch (indicated by a break) has been truncated to 37% of the original length. Phylogeneticdistance scale indicated below the phylogenetic tree, as computed by MASH, b, A spot assay was used toanalyze the efficacy of SNIPR001 against the clinical panel of382 E. coli strains (fromJMI Laboratories) isolated from bloodstream infections and the internal 429 E. coli strain panel. The spot assay was conducted as two independent experiments, with bars indicating average cumulative panel fraction and dots indicating the results of each duplicate relative to prior means, c, Coverage of SNIPR001 does not depend on antibiotic-resistant phenotypes; consequently, 100% 93% I I I I I 4 3 2 1 0 Redundancy SNIPR001 targets > 90% off. coli strains that are carbapenem resistant, ESBL-producing or MDR, and 89% of fluoroquinolone-resistant E. coli strains. Numbers indicated in each green or gray bar indicate the number of bacteria susceptible or resistant to SNIPR001, respectively, for each resistance category generated from a screening of382 strains and subset to the number of strains with a given resistance, d, Amidpoint-rooted phylogenetic tree of the 72 fluoroquinolone-resistant E. coli strains isolated from fecal samples of hematological cancer patients. A total of 67 of the 72 strains are susceptible to at least one of the four CAPS in SNIPR001. Plaquing results are generated by the conservative consensus between two runs of plaquing, that is displaying the outcome with lower plaquing efficiency, e, Redundancy distribution showing82% of the fluoroquinolone-resistant E. coli strains (n = 72) from d are targeted by at least two different CAPs. FDR corrected with Holm's method) of the SNIPR001 cocktail or any of the SNIPR001 CAPs on non-f. coli strains, while the growth off. coli was significantly inhibited (P < 0.05, two-sided Student's f-test, FDR corrected with Holm's method). Thus, SNIPR001 is not expected to impact the gut microbiome beyond the target E. coli. SNIPR001 in vitro host-range in clinical target population To understand the potential effect in strains relevant to hematological cancer patients, the coverage of SNIPR001 was tested against our internal E. coli panel (429 strains) and a set of382 clinical E. coli strains (JMI Laboratories). TheseJMI strains originated from patients with bloodstream infections hospitalized in hemato-oncology units across four different regions from 2018-2020 (Asia-Pacific 54 isolates, Europe 161 isolates, Latin America 26 isolates and North America 141 isolates; Supplementary Fig. 16). The genotypic distribution off. coli strains in the patient population was determined using whole genome sequencing and was found to be diverse, representing nine phylogroups and 118 multilocus sequence types (MLSTs; Fig. 5a and Supplementary Fig. 17). We recorded phage infectivity against the E. coli panel using a spotting assays. Visible single plaques were differentiated from lysis zones in cases where single plaques could not be verified. All spotting assays were run in duplicates. We observed overall coverages of 90.4 ± 1.6% of SNIPR001 in the 382JMIE. coli panel, and of 95.6 ± 0.3% of SNIPR001 on the internal E. coli panel (429 strains). Furthermore, we observed plaques in 53.1 ± 7.7% and lysis zones in 37.3 ± 6.1% of the JMI panel strains, and similarly, plaques in 60.1 ± 6.6% and lysis zones in 35.4 ± 6.3% of the internal panel strains (Fig. 5b). SNIPR001 showed 100% coverage in the B2 phylogroup, representing 53% of the JMI panel. This phylogroup is correlated with multidrug resistance and virulence. Additionally, we observed that SNIPR001 covered 91.7% (n = 55) of strains classified as multidrug resistant (MDR), 100% (ft = 5) of carbapenem-resistant strains, 92.2% (ft = 95) of extended-spectrum [^lactamases producing strains and 88.9% (ft = 176) of strains that are resistant to fluoroquinolones, such as ciprofloxacin and levofloxacin (Fig. 5c). Finally, we validated SNIPR001 on a clinical panel (n = 72) of fluoroquinolone-resistant E. co/istrains that were isolated from either a fecal sample or a perianal swab from hematological cancer patients. This population represents the expected clinical target patient population being pursued (SNIPR001 has been designated fast-trackstatus by the FDA). A subset of these strains gave rise to bloodstream infection (Fig. 5d). 82% of thef. coli strains (n = 72) were susceptible to at least two or more of the CAPs in SNIPR001, and 93% of the strains were susceptible to the whole SNIPR001 cocktail (Fig. 5e). These data demonstrate Nature Biotechnology 8 Article https://doi.org/10.1038/s41587-023-01759-y Day 2 8 h after treatment (1 dose) Day 3 24 h after treatment (3 doses) Day 4 48 h after treatment (6 doses) # ^ r ^° ^° # ^ r ^° ^° sgi x- <$■ ^ S- f7> _N 1010 109 10s 107 106 105 10" ,03 102 101 10° Day 1 Day 2 before treatment Day 2 8 h after treatment (1 dose) NS Day 3 24 h after start of treatment (3 doses) NS ft •! 10io 109 - 10s 107 106 105 10" 103 LOD1Q2± 101 10° Day 1 before dose ***** Day 2 before dose Day 2 8 h after first dose (1 dose) Day 3 24 h after first dose (3 doses) Day 4 48 h after first dose (6 doses) it* ?^vVv° #^VVv& #^VVv& ^VV^° #^VVv& >/of /o°^Oe Ä#Oe /V/O8 .^^cf <9 Fig. 61 SNIPROOlin vivo evaluation in mice and minipigs. a, CAP recovery in minipigs feces after a single p.o. dose of 2 x 1012 PFU of SNIPR001 (n = 8, green) or vehicle (n = 6, gray) over 1 week with daily sampling. Trend lines indicate average recovered phage in PFU per gram feces, dots indicate individual measurement points. LODof 33 PFU g'feces indicated by the dotted line, b, CAP recovery in minipig feces after a single p.o. dose of 2 x 1012 PFU of a single CAP (n = 8 minipigs received either al5.2, ot20.4 or o51.5; n = 7minipigs received a48.4) over 1 week with daily sampling. Trend lines indicate average recovery, while points indicate individual measurements. Recovery was measured in PFU per gram feces. LOD of 33 PFU g'feces (dotted line), c, CAP recovery in mouse feces 8 h, 24 h and 48 h after the start of treatment with three times daily administration of varying doses of SNIPR001 (n = 10 for low, medium and high, green), vehicle (n = 10, gray), or gentamicin (n = 4, gray). Recovery is measured in PFU g1 feces, LOD of 371 PFU g1 feces (dotted line), d, E. colibn recovery in mouse feces indicates increased SNIPR001 effect with increased dose; color legend and group sizes are the same as in c. *P< 0.05, **P< 0.01, ***P< 0.001; statistical analyses were performed using two-sided Kruskal-Wallis tests for comparison of all SNIPR001-treated groups, two-sided Mann-Whitney Utest was used for comparison of treated groups with vehicle corrected using Holm's method separately for each day. The exact Pvalues are shown in Extended Data Table 5. Recovery is measured in CFU per gram feces, with a LOD of 371CFU g'feces. Animals that have begun SNIPR001 treatment are indicated in green, others in gray, e, E. coli bl7 recovery in mouse feces 8 h and 24 h after the start of treatment with three times daily administration of CAPs al5.2, a20.4, a48.4 or o51.5 (n = 6 for each CAP) and in combination as SNIPR001 (n = 6) confirming synergy of the CAPs, as well as vehicle (n = 6), and gentamicin (n = 3). Differences in CFU per gram tested by a two-sided Mann-Whitney U test, Pvalues corrected with Holm's method. Adjusted Pvalues for comparisons of vehicle and SNIPR001 are both 0.022 for days 2 and 3. Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y the benefit of SNIPR001 compared to the individual CAPs with regards to improving the spectrum of efficacy. Tolerability and recovery ofSNIPROOlinminipigs The tolerability and gastrointestinal recovery of SNIPR001 were evaluated in female Gottingen minipigs. Blood and feces were sampled over 7 d following oral administration of 2 x io12 PFU of SNIPR001 or vehicle. No CAPs were recovered from plasma, indicating no systemic exposure, while CAPs were recovered in the feces up to 7 d after SNIPR001 administration with a peak of 2 x io7 PFU 24 h postdosing (Fig. 6a). The minipigs exhibited no clinical signs and no significant changes wereobserved in hematology or biochemistry parameters, in particular, no changes were seen in any immune cells (Supplementary Fig. 18), compared to vehicle treatment, supporting that SNIPR001 was well tolerated (Supplementary Figs. 19,22-25, and Supplementary Table 5). Similar recoveries were obtained with the individual CAPs (Fig. 6b). In conclusion, SNIPR001 appears to be well tolerated in Gottingen minipigs with gastrointestinal recovery. Efficacy in a mouse colonization model To assess the in vivo efficacy of the four selected CAPs in reducing E. coll, we adapted a mouse gut colonization model from ref. 44 for E. coll strain bl7 (Supplementary Fig. 20). Streptomycin was administered for 3 d to reduce Gram-negative bacteria from the mouse gastrointestinal tract, after which streptomycin administration was stopped and animals were inoculated once perorally with E. coll bl7 (1 x io7 CFU). This allowed stable colonization for 3-4 d. Aiming at assessing the efficacy of CAPs on established colonization, treatment was started 2 d after inoculation and the study was terminated on day 4after inoculation, as the colonization starts to drop. To ensure maximum exposure to CAPs, mice were treated with three daily doses, administered 8 h apart, for a total of six doses over 2 d. Mice were treated by oral gavage with a high, medium or low dose (2 x 10" PFU, 2 x 10" PFU and 1 x IO7 PFU, respectively) of SNIPR001, vehicle (negative control) or gentamicin (positive control). CAP recovery in the feces ranged from 3 x io7 PFU g_1 in the low dose to 1 x io10 PFU g"1 in the high dose, confirming successful Gl passage (Fig. 6c). These levelsof CAPs wereassociatedwitha significant (P< 0.05, two-sided Mann-Whitney U test, FDR corrected) dose-dependent reduction in the target E. coll population compared to vehicle treated mice, after 24 h of treatment (day 3). At the high dose, SNIPR001 led to a 4 log10 CFU g"1 reduction (Fig. 6d). Despite an increased variability in bacterial recovery on day 4, possibly due to clearance of the colonizing strain as illustrated in the vehicle group, similar reductions were observed after 2 d of treatment (day 4). While the medium dose did not reach statistical significance (P< 0.05, two-sided Mann-Whitney U test), there was nevertheless a numerical reduction in comparison to the vehicle group. Subsequently, the efficacies of the individual CAPs were compared to the SNIPR001 cocktail in this model. In this experiment, a greater reduction in the colonization of the target strain was observed with SNIPR001 compared to any single CAP (which showed a numerical, but not statistically significant reduction) highlighting a benefit in efficacy from the combination (Fig. 6e). We also assayed the resistance profile of randomly sampled surviving bacteria and found no isolates that were resistant to the SNIPR001 cocktail. We did identify one isolate from one animal which was resistant to three of the four phages of the cocktail (Supplementary Fig. 21). Overall, these data demonstrate the ability of SNIPR001 to decrease the target E. coll'm the Gl tract of colonized mice. Discussion Here we describe the development of SNIPR001 designed to target gut E. coll that frequently translocate in the bloodstream to cause bloodstream infections in patients with hematological cancers who are neutropenic. While fluoroquinolones are being used off-label, these patients continue to have high morbidity and mortality. The useof traditional antibiotics has led to significant bacterial resistance development, and the number of deaths attributable to bacterial antimicrobial resistance in 2019 has been estimated to be 1.27 million, with E. coll being the leading pathogen45. In this study, we describe the development of SNIPR001, a combination of engineered phages with the potential to address challenges related to antibiotic resistance in immunocompromised patients. SNIPR001 combines state-of-the-art phage screening, with phage tail fiber engineering and CRISPR Cas arming. Traditionally, phage therapy has been used experimentally with limited characterization and often applied in a highly individualized way because of the often narrow host range of individual phages46. Buildingon recent advances in phage engineering that have enabled the manipulation of virulent phages47andtheabilitytoengineertailfibers26andCRISPR-Casarrnthe phages, we enhanced the potency of the phages comprising SNIPR001 toenableitto target a broader range of clinically relevantf. co/i, including strains that are resistant to current therapies. To deliver a development candidate ready for clinical testing, we established a traceable manufacturing process resulting in stable CAP substances, and final confirmation of theefficacy of SNIPROOlon largeand clinically relevant strain panels supports theclinical potential of the SNIPR001 cocktail. The observed 4 log10 reduction of E. coll in our in vivo model is a clear improvement over the previous studies35,48. SNIPR001 is an orthogonal antimicrobial approach as it has shown activity in MDRstrains. In addition, there isemergingevidence that maintaining a normal microbiome is important for upholding immunological tonus and potentially benefiting the outcome of oncology treatments49, and this has also been recognized in the most recent guidance on prophylactic management of patients at risk of febrile neutropenia7. In this context, in vitro studies with SNIPR001 have shown specificity toward E. coll with no off-target effects toward any of the tested non-f. coll strains, thereby having a less detrimental effect on the microbiome. In the future, individualized combinations of narrow-spectrum antibiotics such asSNIPROOl may be used first-line rather than use in addition to broad-spectrum antibiotics such as fluoroquinolones. As with any nonclinical study, the translatability of the in vitro and preclinical findings into humans requires investigation, in particular for MDR strains. Although we did not observe structural resistance toward SNIPR001 in mice, resistance development, and the synergy that a combination of CAPs provides, are challenging to study in vivo with a complex drug product like SNIPR001. Furthermore, part of the activity spectrum of SNIPR001 is driven by lysis zone formation and not plaquing, and it is to be investigated how this phenotype translates into clinical efficacy. Therefore, a clinical study to evaluate the ability of SNIPR001 to ascertain safety and its ability to reduce E. coll in the gut without perturbing the overall gut microbiome is currently ongoing in the United States (NCT05277350). SNIPR001 exemplifies a potentially significant therapeutic advance in the field of antimicrobials for high-risk patient populations and can serve as a blueprint for narrow-spectrum therapies for other life-threatening antimicrobial-resistant pathogens in high-risk patient populations. Online content Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availabilityareavailableat https://doi.org/10.1038/s41587-023-01759-y. References 1. Pulte, D., Jansen, L. & Brenner, H. Changes in long term survival after diagnosis with common hematologic malignancies in the early 21st century. Blood Cancer J. 10, 56 (2020). Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y 2. Riley, M. et a I. Epidemiology and burden of mucosal barrier injury laboratory-confirmed bloodstream infections in bone marrow transplant and hematology-oncology units. Am. J. Infect. Control 45, S38 (2017). 3. Dandoy, C. E. et al. Incidence, risk factors, and outcomes of patients who develop mucosal barrier injury-laboratory confirmed bloodstream infections in the first 100 days after allogeneic hematopoietic stem cell transplant. JAMA Netw. Open 3, e1918668 (2020). 4. Velden, W. J. F. M., Herbers, A. H. E., Netea, M. G. & Blijlevens, N. M. A. Mucosal barrier injury, fever and infection in neutropenic patients with cancer: introducing the paradigm febrile mucositis. Br. J. Haematol. 167,441-452 (2014). 5. Girmenia, C. et al. Incidence, risk factors and outcome of pre-engraftment gram-negative bacteremia after allogeneic and autologous hematopoietic stem cell transplantation: an italian prospective multicenter survey. Clin. Infect. Dis. 65, 1884-1896 (2017). 6. Klastersky, J. et al. Bacteraemia in febrile neutropenic cancer patients. Int. J. Antimicrob. Agents 30, 51-59 (2007). 7. Taplitz, R. A. et al. Antimicrobial prophylaxis for adult patients with cancer-related immunosuppression: ASCO and IDSA clinical practice guideline update. J. Clin. Oncol. 36, JCO.18.00374 (2018). 8. Cullen, M. et al. Antibacterial prophylaxis after chemotherapy for solid tumors and lymphomas. N. Engl. J. Med. 353,988-998 (2005). 9. Bucaneve, G. et al. Levofloxacin to prevent bacterial infection in patients with cancer and neutropenia. N. Engl. J. Med. 353, 977-987 (20 05). 10. See, I., Freifeld, A. G. & Magill, S. S. Causative organisms and associated antimicrobial resistance in healthcare-associated, central line-associated bloodstream infections from oncology settings, 2009-2012. Clin. Infect. Dis. 62,1203-1209 (2016). 11. Trecarichi, E. M. et al. Current epidemiology and antimicrobial resistance data for bacterial bloodstream infections in patients with hematologic malignancies: an Italian multicentre prospective survey. Clin. Microbiol. Infect. 21,337-343 (2015). 12. Marin, M.etal. Factors influencing mortality in neutropenic patients with haematologic malignancies or solid tumours with bloodstream infection. Clin. Microbiol. Infect. 21, 583-590 (2015). 13. Averbuch, D. et al. Antimicrobial resistance in gram-negative rods causing bacteremia in hematopoietic stem cell transplant recipients: intercontinental prospective study of the infectious diseases working party of the European bone marrow transplantation group. Clin. Infect. Dis. 65,1819-1828 (2017). 14. Miedzybrodzki, R. et al. Bacteriophages (Springer, 2018) https://doi.org/10.1007/978-3-319-40598-8_31-1 15. Miedzybrodzki, R. et al. Bacteriophages (Springer, 2021) https://doi.org/10.1007/978-3-319-41986-2_31 16. Schooley, R. T. et al. Development and use of personalized bacteriophage-based therapeutic cocktails to treat a patient with a disseminated resistant acinetobacter baumannii infection. Antimicrob. Agents Chemother. 61, e00954-17 (2017). 17. Dedrick, R. M. et al. Engineered bacteriophages for treatment of a patient with a disseminated drug resistant Mycobacterium abscessus. Nat. Med. 25,730-733 (2019). 18. Jennes, S. et al. Use of bacteriophages in the treatment of colistin-only-sensitive Pseudomonas aeruginosa septicaemia in a patient with acute kidney injury—a case report. Crit. Care 21,129 (2017). 19. Jault, P. et al. Efficacy and tolerability of a cocktail of bacteriophages to treat burn wounds infected by Pseudomonas aeruginosa (PhagoBurn): a randomised, controlled, double-blind phase 1/2 trial. Lancet Infect. Dis. 19,35-45 (2018). 20. Sarker, S. A. et al. Oral phage therapy of acute bacterial diarrhea with two coliphage preparations: a randomized trial in children from Bangladesh. eBiomedicine 4,124-137 (2016). Nature Biotechnology 21. Leitner, L. et al. Intravesical bacteriophages for treating urinary tract infections in patients undergoing transurethral resection of the prostate: a randomised, placebo-controlled, double-blind clinical trial. Lancet Infect. Dis. 21,427-436 (2020). 22. Sabino, J, Hirten, R. P & Colombel, J. Alimentary Pharmacology & Therapeutics, Vol. 51 (Wiley Online Library, 2020). 23. Görski, A., Borysowski, J. & Miedzybrodzki, R. Phage therapy: towards a successful clinical trial. Antibiotics 9, 827 (2020). 24. Federici, S. et al. Targeted suppression of human IBD-associated gut microbiota commensals by phage consortia for treatment of intestinal inflammation. Cell 185,2879-2898 (2022). 25. Kauffman, K. M. et al. Resolving the structure of phage-bacteria interactions in the context of natural diversity. Nat. Commun. 13, 372 (2022). 26. Yehl, K. et al. Engineering phage host-range and suppressing bacterial resistance through phage tail fiber mutagenesis. Cell 179,459-469(2019). 27. Westra, E. R. et al. CRISPR immunity relies on the consecutive binding and degradation of negatively supercoiled invader DNA by cascade and Cas3. Mol. Cell. 46, 595-605 (2012). 28. Sinkunas, T. et al. Cas3 is a single-stranded DNA nuclease and ATP-dependent helicase in the CRISPR/Cas immune system. EMBOJ. 30,1335-1342 (2011). 29. Bikard, D. etal. Development of sequence-specific antimicrobials based on programmable CRISPR-Cas nucleases. Nat. Biotechnol. 32,1146-1150(2014). 30. Citorik, R. J., Mimee, M. & Lu, T. K. Sequence-specific antimicrobials using efficiently delivered RNA-guided nucleases. Nat. Biotechnol. 32,1141-1145 (2014). 31. Neil, K. et al. High-efficiency delivery of CRISPR-Cas9 by engineered probiotics enables precise microbiome editing. Mol. Syst. Biol. 17, e10335 (2021). 32. Edgar, R. & Qimron, U. The Escherichia coli CRISPR system protects from X lysogenization, lysogens, and prophage induction. J. Bacteriol. 192, 6291-6294 (2010). 33. Hsu, B. B. etal. In situ reprogramming of gut bacteria by oral delivery. Nat. Commun. 11, 5030 (2020). 34. Yosef, I., Manor, M., Kiro, R. & Qimron, U. Temperate and lytic bacteriophages programmed to sensitize and kill antibiotic-resistant bacteria. Proc. Natl Acad. Sei. USA 112,7267-7272 (2015). 35. Seile, K. et al. In vivo targeting of Clostridioides difficile using phage-delivered CRISPR-Cas3 antimicrobials. mBio 11, e00019-20 (2020). 36. Ochman, H. & Seiander, R. K. Standard reference strains of Escherichia coli from natural populations. J. Bacteriol. 157, 690-693(1984). 37. Oechslin, F. Resistance development to bacteriophages occurring during bacteriophage therapy. Viruses 10, 351 (2018). 38. Dunne, M. etal. Salmonella phage S16 tail fiber adhesin features a rare polyglycine rich domain for host recognition. Structure 26, 1573-1582 (2018). 39. Brouns, S. J. J. et al. Small CRISPR RNAs guide antiviral defense in prokaryotes. Science 321, 960-964 (2008). 40. Knight, D. & Girling, K. Gut flora in health and disease. Lancet 361, 1831 (2003). 41. Christensen, S. K., Mikkelsen, M., Pedersen, K. & Gerdes, K. RelE, a global inhibitor of translation, is activated during nutritional stress. Proc. Natl Acad. Sei. USA 98,14328-14333 (2001). 42. Vieira, H. L. A., Freire, P. & Arraiano, C. M. Effect of Escherichia coli morphogene bolA on biofilms. Appl. Environ. Microbiol. 70, 5682-5684 (2004). 43. Uribe, R. V. et al. Bacterial resistance to CRISPR-Cas antimicrobials. Sei. Rep. 11,17267 (2021). Article 44. Galtier, M. et a I. Bacteriophages to reduce gut carriage of antibiotic resistant uropathogens with low impact on microbiota composition. Environ. Microbiol. 18, 2237-2245 (2016). 45. Antimicrobial Resistance Collaborators. Global burden of bacterial antimicrobial resistance in 2019: a systematic analysis. Lancet 399, 629-655 (2022). 46. Hyman, P. & Abedon, S. T. Bacteriophage host range and bacterial resistance. Adv. Appl. Microbiol. 70, 217-248 (2010). 47. Lemay, M.-L, Renaud, A., Rousseau, G. & Moineau, S. Targeted genome editing of virulent phages using CRISPR-Cas9. Bio. Protoc. 8, e2674 (2018). 48. Lam, K. N. et al. Phage-delivered CRISPR-Cas9 for strain-specific depletion and genomic deletions in the gut microbiome. Cell Rep. 37,109930-109930 (2021). 49. Lee, K. A. et al. The gut microbiome: what the oncologist ought to know. Br. J. Cancer 125,1197-1209 (2021). https://doi.org/10.1038/s41587-023-01759-y Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/. © The Author(s) 2023 Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y Methods Phage collection and isolation procedures The starting point for the phage screening was a collection of 162 lytic WTphages, 82 were isolated in-house from commercial cocktails and environmental sources, 71 phages were obtained from a phage bank (LyseNTech, Korea) and two phages from ATCC, one phage was donated by the University of Copenhagen and six wereobtained from Kirikkale University, Turkey50 (Supplementary Table 1). Phage isolation was carried out by usingf. coli strain panels (seef. coli panels and isolation procedures). In brief, 100 pi of overnight cultures of each E. coli strain were mixed with 100 pi of each phage cocktail or wastewater sample. Following 6 min incubation at room temperature (in this period infection should occur), 3 ml of prewarmed top agar containing Ca2+ were added to the E. coW/phage or wastewater mixtures and poured immediately on an LB plate. Alternatively, tenfold dilutions of each cocktail were spotted on lawns prepared with isolation strains. After drying, plates were incubated at 37 °C overnight. Plaques were picked from each plate and resuspended in 500 pi of SM buffer, vor-texed and stored at 4 °C. Tenfold dilutions were spotted on the isolation strain which the plaque was originally picked from. To increase the likelihood of obtaining plaques corresponding to single phages, the procedure was repeated at least three times. Lysates were prepared from single plaques picked at the previous round of propagation, DNA was extracted and their genomes were sequenced. E. coli panels and isolation procedures Three E.coli panels, one internal SNIPRBiome panel and two clinically relevant panels were included in this study. The internal panel consists of429 phylogenetically diverse E. coli strains, isolated from the blood of patients with bloodstream infections and urinary tract infections, from feces of humans with no known disease, animals and theenvironment. The strains cover seven different phylogroups (A, Bl, B2, C, D, E and F), 114 MLST groups, serotypes (K- and O-type), antibiotic resistance profiles and different geographical locations of isolation. The JMI panel comprises 382 strains E. coli clinical collection obtained from JMI Laboratories. These strains were isolated from patients with bloodstream infections hospitalized in hematology and oncology units across four different regions (Asia-Pacific 54 isolates, Europe 161 isolates, Latin America 26 isolates and North America 141 isolates), sourced through the SENTRY Antimicrobial Surveillance Program (2018-2020), which is composed of a network of more than 150 medical centers in more than 28 countries worldwide (https://www. jmilabs.com/sentry-surveillance-program). Finally, the panel comprising 72 fluoroquinolone-resistant E. coli strains is isolated from either fecal samples or perianal swabs of hematological cancer patients hospitalized for hematopoietic cell transplantation51,52. E. coli strains were cultivated at 37 °C in LB at 250 rpm in liquid media or on agar plates containing 1.5% (wt/vol) agar. When necessary, cultures were supplemented with ampicillin (100 pg ml"1), kanamycin (50 pg ml^gentamicin (15 pg mL1) oramikacin (50 pg mL1). All media for the growth of conjugation donor E. coWJKE201 (ref. 53) and its derivatives were supplemented with 1,6-diaminopimelic acid (80 pg mL1) to complement their auxotrophy. Both E. coli strain b52, which was used to produce al5.2, a48.4 and a51.5, and E. coli strain b2479, which was selected to produce a20.4, belong to phylogroup A. Strain E. coli bl7 was used as colonizing strain in the in vivo efficacy models as the strain is susceptible to all SNIPR001 CAPs and is part of the SNIPR Biome strain bank. Phage screening by growth kinetics In vitro susceptibility of the internal E. coli panel (n = 429) to the 162 WT phages was evaluated using a growth kinetics assay. The assay measures themetabolic activity of a bacteria bytrackingthereduction of a tetrazolium dye to a purple compound that aggregates during bacterial growth. Thecolorimetric reading was recorded every 15 min over a 24-h period by using the OmniLog (Biolog)-adapted from ref. 54. The inhibitory area under the curve (iAUC) was calculated from the kinetic curves over the course of the experiment and was defined as the ratio between the normalized AUC of the phage-treated bacterial growth curveand the bacteria-only control. Susceptibility was defined at iAUC values >0.2. Prescreening, including 48 phages, was carried out at MOI10, after which 114 phages were screened at MOI1. Calculation of bacterial growth inhibition using iAUC The growth inhibitory effect of SNIPR001 was determined usinggrowth kinetic curves constructed using the OmniLog apparatus. To limit technical variability in measurement between timepoints, a cubic smoothing spline function was applied to the data in Scala using the 'umontreal.ssj.functionfit' package. To identify appropriatep and weight variables, every combination of p and weight 0.1 and 0.5 was applied in 0.1 increments (that is, 0.1,0.2,... 0.5). The spline with the lowest mean absolute error was chosen for area under the curve (AUC) calculation. The initial cumulative amount of fluorescent dye at the initial timepoint varies slightly from well to well, leading to artificial inflation of the AUC of certain wells. Using the best smoothed square spline, the mean signal for the first 1.5 h,beforeany measurable growth, was removed from all growth curves to approximate a zero-growth signal intercept. The total iAUC was calculated as the sum of the Riemann midpoint sums for each timepoint along the smoothed square spline. Lastly, wecalculated the iAUC as iAUC = 1 - AUCSamp|e/AUCCorltr0|, where AUCSampie is the AUC of the spline created by a given bacteria and SNIPR001, while AUCContr0| refers to the AUC of the spline created with a given bacteria without a given phage or CAP, or a combination of those. Thus, iAUC values usually lie between 0 and 1, where 0 indicates no growth inhibition and 1 indicatescomplete growth inhibition. Some biological and technical noise does result in iAUC values outside these bounds on occasion but is considered negligible. Host range was calculated as the fraction of a panel that had an iAUC < 0.2 for each repeat. Reported standard deviations were calculated as the deviance in the number of strains with an iAUC < 0.2, and then normalized to the size of the panel, by dividing the s.d. with thesizeofthe panel. Combination complementarity prediction Phage and CAP complementarity were evaluated in silico under the assumption of complementarity-if at least one CAP in a combination of phages can strongly inhibit a given bacterial strain, the combination of CAPs is assumed to strongly inhibit that bacterial strain. In in vitro studies, the total host range was estimated by calculating the fraction of a panel that was inhibited by one or more of the members of a given CAP or phage combination. In OmniLog screenings, a strain was considered inhibited if the iAUC of phage was above 0.2 compared to control. When usingplaquingresults,a strain wasconsidered inhibited ifaplaqueorlysiszone was observed. In in vivo studies, theeffect of CAP combinations was considered complementary,and the efficacy of individual CAPs was assessed as the log10-transformed difference in CFU per gram between a vehicle and a given CAP. The predicted effect of a combination was thus evaluated as the sum of these log10 reductions for each member of a combination. In silico marginal host-range calculation To get an overview of theability of a CAP to participate in an efficacious CAP combination, weevaluate the marginal host ranges for each CAP. The marginal host range is a measure of the gained host range when a given CAP is incorporated in a combination. This is calculated as the difference in host range between a combination with and without a given CAP of interest. By calculating the marginal host ranges of each combination for each CAP, we can compare the different CAPs with regard to their utility in adding host ranges. However, the composition Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y of the CAP panel can lead to unfair scoring-the addition of a CAP to a combination, where one of the composing CAPs has a very similar inhibitory profile, would have an unfairly low marginal host range. Similarly, if a CAP is added to a combination of CAPs that all have very similar inhibitory profiles, the marginal utility gain would be unfairly high. If the set of CAPs being screened does not equally represent different types of inhibitory profiles, some CAPs will have misleading marginal host-range distributions. To avoid this issue, we do not generate combinations of CAPs that contain multiple CAPs that originate from the same WT phage. To identify CAPs whose marginal host range tended to be good, we used the mode to differentiate the CAPs. The mode of the distribution for each phage was used to calculate the overall utility of phage using the densityO function in R v. 4.1.0. Engineering phages with a CRISPR-Cas system Phages were CRISPR-Cas armed by using homologous recombination. We inserted the payload in the region between the pin (encoding the inhibitor of host Lon protease) and vs.7 (encoding a conserved hypothetical protein) gene. Recombination was carried out in bacterial cells during phage propagation. Cells carried a plasmid that served as a recombination template. Recombination template plasmids carried the sequences that were aimed to be inserted into the phage genome between -200 bp and 700 bp flanking sequences that were homologous to the phage sequences at the insertion site. For each phage, we inserted the type IE CRISPR-Cas system endogenous to E. coll (Genbank CP032679.1), that is, the cas3 gene iygcB) and the downstream genes encoding the cascade complex, casA (ygcL),casB (ygcK), casC (ygcj), casD (ygcl) and casElygcH), as well as a CRISPR array targeting selected E. coll sequences. For all CAPs selected, the cas genes originating from E. coll are identical. Insertion of the CRISPR-Cas system resulted in the deletion of-7 kbp deletion of phage DNAin the pin - vs.7. The sequences of the resulting CAPs were verified by NGS (BaseClear). Transduction of CGVs in biofilms E. coll b52 cells were grown in 96-well plates, and biofilms were allowed to develop on peg lids. Each well contained 180 ul M9 medium (Sigma-Aldrich, M6030) supplemented with 20 mM glucose, 2 mM MgS04,0.1 mM CaCI2,0.1% Amicase (Sigma-Aldrich) and 0.1% mannitol. Wells were inoculated with 1 ul of overnight b52 culture. The peg lid was inserted, and the microtiter plate was incubated statically for 24 h at 37 °C. Next, the peg lid was transferred to a new plate with fresh media without washing, and the plate was incubated for an additional 24 h. After incubation, a new plate was prepared with 100 ul media and 100 ul of CGV transducing particles (-108 particles) in each well (three replicates). Biofilms grown on the pegs were rinsed three times in sterile H20 (200 ul) before transferring them on the new plate. The plate was incubated statically for 5 h at 37 °C. To assay the metabolic activity of cells in the biofilms, lids were rinsed three times in sterile H20 (200 ul) before placing them in a plate with 20 ul Alamarblue stain (Thermo Fisher Scientific) and 180 ul media in each well. Plates were incubated for 1.5 h at 37 °C and moved to a microplate reader (Synergy HI, Biotek). Fluorescence (excitation, 560 nm; emission, 590 nm) and absorbance (600 nm) were recorded for each well. The metabolicactivities of the biofilms treated with CGVs carrying one of the promoters (Preffl or PbM) were reported relative to the metabolic activities of biofilms treated with a CGV not carrying a promoter transcribing the cas genes. Plasmid and strain construction To construct CGV-EcCas, cas3 and cascade genes from E. coll were amplified and cloned into a ColEl-type plasmid, pZE21 (ref. 55), containing kanamycin, gentamycin and amikacin resistance markers, andoriTRP4. DNA fragments encoding a 3-spacer array targeting genes in E. coll were synthesized as gBlock fragments (IDT) flanked by Aarl restriction enzymes (gB149, gB150, gB152 and gB153; Supplementary Table 6). Similarly, constitutive promoter J23100 (ttgacggctagctcag tcctaggtacagtgctagc) was synthesized as a gBlock fragment (IDT) (gB-d2; Supplementary Table 6) to drive the expression of the CRISPR array. The array contains nucleotides from the genome of E. coll per target locus separated by direct repeats. The protospacer adjacent motif is located adjacent to the selected target sequences in the genome off. coll. cas3 and cascade genes from E. coll were amplified with primers containing Aarl restriction sites (TH556 and TH558; Supplementary Table 6). Similarly, pMO constitutive promoter to drive the expression of the cas genes (ggattaacaatataagctgaccttcaagtattgaat) was amplified with primers TH402 and TH403 (Supplementary Table 6). To combinecas3and cascade genes with the CRISPRarray, all plasmids were digested with Bsal and ligated with T4 DNA ligase. Finally, to generate CGV-EcCas, the CRISPR-Cas system was moved into conju-gative plasmid pZE21 by InFusion HD cloning using primers TH712 to TH715 (Supplementary Table 6). Transformation assays Overnight cultures were diluted (1:100) in fresh LB medium and grown to mid-exponential phase (OD600 ~ 0.6). Subsequently, cells were prepared for electroporation and concentrated 50-fold in ice-cold MilliQ water. Cells were then electroporated with appropriate plasmids, allowed to recover for 1 h at 37 °C in super optimal broth, and plated on LB plates supplemented with antibiotics. Conjugation assays Conjugation experiments assessing the transfer and killingefficiency of CGV-EcCas were established using E. co/i JKE201 as the donor and E. coll clinical isolates as recipients (including target and nontarget and E. coll strains as controls). Plasmids were conjugated into E. coll recipients by liquid mating. Briefly, overnight cultures were diluted (1:100) in fresh LB medium, grown to OD600 ~ 0.4, washed, and suspended in fresh LB to OD600 = 0.25.125 ul of donor and 25 ul of recipient cell suspensions were mixed for 5:1 mating in a 96-well microplate and incubated for 16 h at 37 °C. The conjugation efficiency was determined by platinga dilution seriesof conjugation reactions onto LB agar supplemented with antibiotics (to selectforthetransconjugants). The specific killingefficiency was quantified by plating 90 ul of the conjugation reactions on selective plates. The CGV-EcCas plasmid encodes kanamycin, gentamycin and amikacin resistance to enable selection for transconjugants. Viability was calculated by counting CFUs on the plates, and data were recorded as viable cell concentration (CFU mL1). Synchronized CAP infection and cas3 expression assay An overnight culture of the test strain in LB was 100-fold diluted and incubated to stationary phase in LB at 37 °C with shaking, and 10-ml aliquots were subsequently separated into 50-ml falcon tubes. Each aliquot was then seeded with 50 ul of high-titer lysateof the individual CAPs, and incubation was continued under the same conditions. Additionally, a mock 10 ml LB volume for each CAPs was also seeded with 50 ul of CAP lysates and used for 0 min phage enumeration. At 5 min, 15 min and 30 min postseeding, aliquots were collected for total RNA extraction and phage enumeration. Phage enumeration aliquots were syringe filtered (0.2 um, Sartorius AG) and subjected to an EoP assay. For total RNA extraction, 1 ml aliquots of individual cultures were centrifuged at 13.3kg using a table-top centrifuge for 15 s, and supernatants were discarded. Then, pellets were immediately resuspended in cold RNA Later (Thermo Fisher Scientific, AM7020) and stored at -20 °C until extraction. Total RNA was extracted using a GeneElute Total RNA kit (Sigma-Aldrich) following the manufacturer's protocol for extraction of RNA from bacteria. After the first elution, Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y 1 pi of Dnase I (1U pi"1) was added and incubated overnight at 37 °C. The reaction was terminated by incubation at 70 °C for 15 min. The RNA was re-purified on a GeneElute column and eluted in 35 ul of kit elution buffer. Total RNA concentration was estimated on a NanoDrop instrument (Thermo Fisher Scientific, One/OneC), and 0.5-2 ug of RNA was added to a cDNA synthesis reaction containing SuperScriptlll RT enzyme (Thermo Fisher Scientific) and random decamers to prime synthesis in a 20-ul reaction volume. ThecDNA reaction was diluted to 100 ul in water. RT-PCR was conducted in triplicate using 5 ul of cDNA as template, 10 pi of Power SYBR Green PCR Master mix (Thermo Fisher Scientific) and 0.2 pM of each PCR primer. PCRs were performed on an AB QuantStudio5 system (Applied Biosystems) using the standard two-step thermocycling protocol for Power SYBR Green PCR Master Mix with 60 °C annealing/extension. Theforward and reverse primers for gapA (reference gene) were 5'-cgctaacttcgacaaatatgctggc-3', and 5'-aggacgggatgatgttctgggaa-3', and for cas3 were 5'caagtatgctaccaa cggctaaag-3' and 5'- ccaatcaaaatcaacgtcgagtga-3'. Single PCR products were confirmed for these primer pairs by melting curve analysis. Relative levels of transcripts were estimated using tenfold dilutions of purified PCR products as standards, and values were expressed as the ratio of cas3 to gapA transcripts. Phage competition assay Lysates of the two phages were mixed at 9:1 (WT:CAP) ratio and the phage mixtures were added to 10 ml 2xYT medium containing 10 mM CaCI2 and 20 mM MgCI2, and 100 pi overnight of E. coli strain b230, serving as a target for both competing phages. After 2 h incubation in a 37 °C shaking incubator, the cultures were centrifuged and 1 pi of the supernatant was added to a new b230 culture. The same steps were repeated twice. The ratio of phages was assessed by PCR with three primers, resulting in two specific products, one for the WT phage and one for the CAP (al5/15.2-5'-ttcattgcgtatttgtagatgaagctc-3', 5'cttttcagactt atcttgcgtttcttaagaagttctacaagttct-3', 5'-gtacgactgattgatcccaccagc-3'; oc20/20.4-5'-atggcttttattgctaccgggt-3', 5'aaatctagagcggttcagt actcaaggaaatcatcccagaaactc-3', 5'-tgctatctttggctccactgtgat-3'). PCR products were separated on a 1% agarose gel and DNA bands were stained by SYBRsafe and visualized and quantified by the ChemiDoc XRS + System (model 1708265, Bio-Rad). The background-corrected intensity of the band corresponding to the WT phage was divided by the intensity of the band corresponding to the CAP in the same lane, to obtain the ratio of the two band intensities (WT/CAP). The fraction of CAP compared to the total phage content (WT + CAP) was determined based on the calibration curve, which was made by usinga set of different mixtures of the two phages and f ittinga curve to the measured band intensity ratios (WT/CAP). The estimated error of the reported values is less than 20%. Lawn killing assay An overnight culture of the test strain in LB was adjusted to 109 CFU ml"1. Hundred pi aliquots of CFU ml"1 adjusted strain was mixed with 100 pi of 109 PFU ml"1 to achieve a multiplicity of infection of 1 of either CAP al5.2 or WT al5 in 15 ml falcon tubes, mixed with 3 ml of molten and pretempered top agar and spread on LB plates. After lawns were solidified, plates were incubated at 37 °C overnight, and the total number of surviving colonies was counted for CAP ocl5.2 or WT ocl5 groups the next day. Assays were performed as independent biological duplicates whereeach experiment comprised ten technical replicates. Statistical significance was established using both replicates using a two-sided Mann-Whitney [/test. Generalized transduction assay The transduction ability of each CAP was evaluated via the generalized transduction assay. Briefly, transducing lysates were prepared by propagating each CAP on E. coli MG1655 lamBv.Cm. This strain was modified from the WT MG1655 (700926, American Type Culture Collection) to carry a chloramphenicol selection marker. Experiments were conducted in parallel with the well-characterized lytic T4 phage (negative control), and its transducing mutant T4GT7 (ref. 56; positive control). Following this step, the WT E. coli MG1655 strain was infected at an OD600 of 0.3 with each transducing lysate at MOI of 0.5, 0.1 and 0.01, and spread on LB plates containing chloramphenicol. Next day, the number of transductant colonies was recorded for each CAP and control and different MOIs. The frequency of transduction was calculated as the number of transductants divided by the titer of the transducing lysate. Sequence analysis of CAPs Sequencesof the individual SNIPR001 CAPs wereanalyzed for the presence of antibiotic resistance, virulence genes and lysogeny associates genes (transposases and integrases) using databases (Extended Data Table 2). Furthermore, for release criteria during the CMC process (Supplementary Table 2), phage samples were analyzed using whole genome sequencing. This typically results in >1000x coverage of the whole phage genome. Assemblies are constructed by down-sampling the data to 1000* average coverage for the phage and assembling usingSKESA. To detect differences between samples and to detect non-majority mutations the raw reads were mapped back to the assembly using BWA (version 0.7.17). Phage specificity assay using liquid killing assay SNIPR001 CAPs (0(15.2, 0:20.4, 1x48.4 and oc51.5) and SNIPR001 killing specificity were evaluated via a biopotency assay against a panel of human-relevant, aerobic (n = 6) and anaerobic (n = 3) bacterial strains. An E. coli strain b2480 was included as a positive control for phage-mediated killing (Extended Data Table 3). In brief, overnight cultures were adjusted to 106 CFU ml"1 in LB broth. SNIPR001 CAPs or SNIPR001 (in which each CAP was combined in equal ratio) were added at an MOI of 1 before incubation for 4 h. Untreated bacteria were cultured in parallel as controls for bacterial growth. CFU counts were recorded at0hand4h post phage treatment, and data are represented as Alog10 CFU ml"1 by subtracting the initial inoculum (0 h) from the assay endpoint CFU per milliliter (4 h). CMC The in vitro stability of phages was assessed by following the potency of CAPs in the formulation buffer overtime at 2-8 °C and at accelerated temperature (40 °C). Polypropylene cryovials were filled with one milliliter of each phage for storage at the appropriate temperature. At each timepoint, the potency of each phage was assessed by EoP method in triplicates. T0 was measured before the initiation of storage. Spotting assay and EoP For counting of phage titers, phage lysates or the equal volume mix of SNIPR001 CAPs were serially diluted tenfold in SM buffer or PBS, respectively. Bacterial lawns were prepared by addinglOO pi or300 pi of bacterial overnight culture to 3 ml or 10 ml of 0.5% top agar (containing Ca2+ and Mg2+), which was vortexed briefly and poured onto a round or square LB plate. Five microliters of the dilution series of test phages were then spotted on lawns and left to dry at room temperature with an open lid before incubation at 37 °C overnight. The strains b52, b2479 and bl7 were used as controls of the assay and included in each round of assays. The next day, results were assessed (Extended Data Table 4). In this assay, a susceptible strain is defined as one producing plaques that are countable in PFU per milliliter as well asone without visible plaques but demonstrating impairment of bacterial growth (that is, lysis zones). Coverage defines the percentage of the total number of susceptible strains. Images of all plates were recorded. Figures illustrating EoP results first had titers log10 transformed and then standard deviances Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y and averages were calculated subsequently. The clinical panels and control strains were tested in two independent experiments. Animals and housing Mouse studies were performed with female CD-I IGS mice (approximately 6-7 weeks of age upon arrival) from Charles River. The animals were housed in groups of three to five mice per cage within a climate-controlled room (temperature, 20-23 °C; relative humidity, 30-70%) under a 12 h light/12 h dark cycle (illuminated, 07:00-19:00). Standard pelleted chow and tap water were available ad libitum. Animals were allowed an acclimatization period of at least 7 d before the start of the experimental procedures. Thirty female Gottingenminipigs (approximately 4-7 months of age upon arrival) from Ellegaard Gottingenminipigs A/S were used for tolerability and kinetic studies. Animals were allowed an acclimatization period of at least 14 d before the start of experiments. Pigs were housed in groups of two to three animals and given standard pig diet twice daily and tap water was available ad libitum. All procedures wereconducted in accordance with guidelines from the Danish Animal Experiments Inspectorate, Ministry of Environment and Food of Denmark and in accordance with the institutional license (BioAdvice, animal license 2015-15-0201-00540). Mouse gut colonization model The mouse gut colonization model was adapted from ref. 44. Briefly, pretreatment with streptomycin (5 g I"1) in the drinking water was given 3 d before inoculation with E. coli bl7 to decrease the level of native bacteria. On day 0, an inoculum of 3 x 107 CFU of E. coli bl7 was prepared from a frozen glycerol stock and administered to all mice in 0.25 ml by oral gavage. Treatment was administered three times daily for 2 d starting 2 d after inoculation. Right before each administration, the four CAPs were mixed in a 1:1:1:1 ratio to form SNIPR001 at a high, medium or low concentration resulting in dose levels of 2 x 10", 2 x 10" and 1 x 107 PFU. At the time of treatment, mice wereadministered 0.1 ml of 10% sodium bicarbonate by oral gavage followed by the oral administration of 0.3 ml of SNIPR001, saline (vehicle) or 43.5 mg kg"1 gentamicin. CAP recovery and tolerability studies Gottingen minipigs were first given a cocktail of antibiotic comprising neomycin (60 mg kg"1, orally, once daily for 4 d) and cefquinome (2 mg kg"1, intramuscular once daily for 3 d) before SNIPROOlor single CAP administration to decrease the level of Gram-negative bacteria in the GI tract and therefore limiting phage replication. Animals were then fasted overnight and lightly sedated before administration of a single CAP, or SNIPR001 cocktail, onceorally at 2 x 1012 PFU in 100 ml, following an oral administration of 50 ml of 10% sodium bicarbonate. Fecal samples werecollected daily for CAPs quantification by plaque assay. In addition, for the tolerability study, blood samples were collected for hematology and blood chemistry analysis, including C-reactive protein, and plaque assay. Animals were closely monitored following SNIPR001 administration, and their body temperature was recorded regularly. Quantification off. colibU and CAPs in feces Fecal samples were homogenized and serially diluted in SM buffer. Triplicates of 10 \i\ of each dilution were then spotted on McConkey agar plates (Sigma-Aldrich, M7408) supplemented with streptomycin (1 mg ml"1) and incubated for 12-16 h at 37 °C for E. co/ienumeration. Plaque assays were performed for enumeration of CAPs in feces samples. Briefly, homogenized samples were centrifuged at 10,000g for 10 min, and the supernatant was serially diluted. Triplicates of 10 ul of each dilution were spotted on an E. coli bl7 overlay and incubated for 12-16 hat 37 °C. To quantify the presence of in vivo resistors, three colonies from each mouse fecal sample in the medium dose group at three different time points were picked from the McConkey agar plates. Colonies were incubated for 12-16 h at 37 °C in LB broth and used to make top agar overlays on LBagar plates. Then, plates were dried for 15 min in the LAF bench. TheSNIPROOl cocktail, as well as the four individual CAPs, were spotted as a dilution series from 1 x 10s PUF ml"1 stocks. As a control, a top agar overlay of colonization strain E. coli bl7 was spotted in the same way. Plates were left to dry in the LAF bench with the lid on and subsequently incubated upside down for 12-16 h at 37 °C. Whole genome sequencing off. coli strains from JMI Total genomic DNA was extracted and purified using the KingFisher Cell and Tissue DNA kit (Thermo Fisher Scientific) in a robotic KingFisher Flex Magnetic Particle Processor (Thermo Fisher Scientific) workstation. Total genomic DNA was used as input material for library construction. DNA libraries were prepared using the Nextera XT library construction protocol and index kit (lllumina) and sequenced on a MiSeq Sequencer (lllumina) using MiSeq Reagent Kits v3 (600 cycles). Resistance phenotype definitions The extended-spectrum piactamase-phenotype was defined for E. coli as a minimum inhibitory concentration (MIC) value >2 mg I"1 for ceftriaxone, ceftazidime and/or aztreonam (https://clsi.org/). Carbapenem-resistantFnferoftacfera/es was defined as any isolate displaying imipenem, doripenem and/or meropenem resistance with MIC > 2 mg I"1 (https://clsi.org/). Assembly of whole-genome sequencing data Raw sequencing reads were trimmed using Trimmomatic57 (version 0.39) with the settings 'LEADINGS TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:36'. Trimmed reads were assembled using SPAdes58 (version 3.14.1) with default settings. Contigs shorter than 500 bp or with a sequencing depth below two times were removed from the final assemblies. Comparative genomic methods for clinical E. coli strains MLST was performed using MLST2 (ref. 59) on the assembled genomes of the E. coli bacteria using default settings, with the MLST database downloaded on ljuly 2021, from the MLST2 repository (https://bitbucket.org/genomicepidemiology/mlst db/src/master/). Phylogroup classification was conducted using ClermonTyping60 on the assembled E. coli genomes using default settings. Distance matrices for phylogenetic tree construction were generated using MASH61 with a k-mer size of 21 and 10,000 sketches per genome. Sketches were then compared to create the MASH distance in a pairwise manner to create a distance matrix of E. coli genomes. Phage synteny analysis To generate the synteny plot, WT sequences of the four phages included in the final cocktail, plus the two closely related and well-known reference phages (RB69 AY303349.1 and T2 NC_054931.1) were annotated with RAST to extract predicted protein sequences. All protein sequences for each phage were queried again all other phage genomes using tblastn (v 2.12.0), with an F-value cutoff of 1 x 10"10. The synteny plot was then generated using a custom Python (v 3.7.10) script (see Data availability), using the drawSvg library (v 1.9.0). The plot shows the phage genomes in order of similarity and displays all tblastn hits as synteny blocks shaded by their protein identity. The proteins of the two reference phages were manually classified as belonging to each of the functional groups 'DNA metabolism', 'structure' or 'other' and colored accordingly. Data processing and visualization Figures and key statistics were generated using R version 4.1.0. For figure generation, the following packages were used: RcolorBrewer Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y v. 1.1-2, ape v. 5.5, ggsignif v. 0.6.2, ggpubr v. 0.4.0, matrixStats 0.59, reshape2 v. 1.4.4, ggimage v. 0.3.0, here v. 1.0.1, purr v. 0.3.4, ggtree62 v. 3.0.2, systemfonts v. 1.0.2, Cairo v. 1.5-12.2, cowplot v. 1.1.1, reaxxl v. 1.3.1, ggplot2 v.3.3.3, openxlsx, v. 4.2.3, patchwork v. 1.1.1, dplyr v. 1.0.7 and ggh4x v. 0.2.3. Averages and standard deviations are calculated after transforming the values to the scale shown on a given figure, for example when a log10 scale is used, the averages and standard deviations are calculated after log10 transformation. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Data availability All data and results that were generated during this study are deposited at https://github.com/sniprbiome/SNIPR001 paper. Additional data are available in the Article, Online methods and Supplementary tables. To reproduce the results, no further data is needed. Phage genome sequences are deposited at Genbank under access numbers OQ067373 - 76. The MLST database was downloaded on July 1,2021, from the MLST2 repository (https://bitbucket.org/genomicepidemiology/mlstdb/ src/master/). For annotation of the CAP sequences, the following tools anddatasets were used ResFinder 4.1 (https://cge.cbs.dtu.dk/services/ ResFinder), VirulenceFinder-2.0 (https://cge.cbs.dtu.dk/services/ VirulenceFinder/),PHASTERProphage/VirusDB (https://phaster.ca/). Code availability All code needed to produce this study is available at https://github. com/sniprbiome/SNIPROOl paper. References 50. Gencay, Y. E., Ayaz, N. D., Copuroglu, G. & Erol, I. Biocontrol of shiga toxigenic Escherichia coli 0157:H7 in Turkish raw meatball by bacteriophage. J. FoodSaf. 36,120-131 (2016). 51. Satlin, M.J. et al. Colonization with fluoroquinolone-resistant enterobacterales decreases the effectiveness of fluoroquinolone prophylaxis in hematopoietic cell transplant recipients. Clin. Infect. Dis. 73,1257-1265 (2021). 52. Satlin, M.J. et al. Colonization with levofloxacin-resistant extended-spectrum 3-lactamase-producing enterobacteriaceae and risk of bacteremia in hematopoietic stem cell transplant recipients. Clin. Infect. D/s.67,1720-1728 (2018). 53. Harms, A. et al. A bacterial toxin-antitoxin module is the origin of inter-bacterial and inter-kingdom effectors of Bartonella. PLoS Genet. 13, e1007077 (2017). 54. Henry, M. et al. Development of a high throughput assay for indirectly measuring phage growth using the OmniLogTM system. Bacteriophage 2,159-167 (2014). 55. Lutz, R. & Bujard, H. Independent and tight regulation of transcriptional units in Escherichia coli via the LacR/O, the TetR/O and AraC/11-12 regulatory elements. Nucleic Acids Res. 25, 1203-1210(1997). 56. Young, K. K., Edlin, G. J. & Wilson, G. G. Genetic analysis of bacteriophage T4 transducing bacteriophages. J. Virol. 41, 345-347(1982). 57. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for lllumina sequence data. Bioinformatics 30, 2114-2120 (2014). 58. Bankevich, A. et al. SPAdes: a new genome assembly algorithm and its applications to single-cell sequencing. J. Comput. Biol. 19, 455-477(2012). 59. Larsen, M. V. etal. Multilocus sequence typing of total-genome-sequenced bacteria. J. Clin. Microbiol. 50,1355-1361 (2012). 60. Beghain, J., Bridier-Nahmias, A., Nagard, H. L, Denamur, E. & Clermont, O. ClermonTyping: an easy-to-use and accurate in silico method for Escherichia genus strain phylotyping. Microb. Genom. 4, e000192 (2018). 61. Ondov, B. D. et al. Mash: fast genome and metagenome distance estimation using MinHash. Genome Biol. 17,132 (2016). 62. Yu, G. Using ggtree to visualize data on tree-like structures. Curr. Protoc. Bioinform. 69, e96 (2020). 63. Gibson, S. B. et al. Constructing and characterizing bacteriophage libraries for phage therapy of human infections. Front. Microbiol. 10,2537 (2019). Acknowledgements SNIPR Biome acknowledges funding from Combating Antibiotic-Resistant Bacteria Biopharmaceutical Accelerator (CARB-X). Research reported in this publication is supported by CARB-X. CARB-X's funding for this project is sponsored by the Cooperative Agreement Number IDSEP160030 from ASPR/BARDA and by an award from Wellcome Trust. The content is solely the responsibility of the authors and does not necessarily represent the official views of CARB-X or any of its funders. We would also like to thank our collaborators at SSI (Denmark) and Minerva Imaging (Denmark) for their contribution to the manuscript. We would like to acknowledge E. S0ndberg (SNIPR BIOME) who illustrated Fig. 1. Author contributions E.v.d.H., J.K.H., K.B., Y.E.G, D.J., V.M., C.R., A.S., A.S.T., S.S., AT., J.C., L.J., B.D., M.G., J.H., AT., J.C., C.G., M.Z. and M.O.A.S. conceptualized and designed the work. K.B., Y.E.G., D.J., V.M., C.R., A.S., A.S.T., S.S., AT., L.B., T.B., M.K.E., K.C.J., L.K., R.P., T.S.S., IT., A.0.P., E.G.B., B.H., A.G., M.J.S., J.K.H. and E.v.d.H. were responsible for data acquisition and analysis. All authors contributed to data interpretation and finalization. Competing interests All authors affiliated with SNIPR Biome are present or past employees of SNIPR Biome and maybe share- or warrant holders. F.S. and T.B.D. are subcontractors of SNIPR Biome. M.J.S. received research funding from Merck, Biomérieux and SNIPR Biome. M.J.S. is an unpaid consultant for SNIPR Biome and has been consulting for Shionogi and participated on a Data Safety Monitoring Board for AbbVie. Patent applications have been filed based on material described in this article. SNIPR, CRISPR-Guided Vectors and CGVare trademarks of SNIPR Biome ApS. Data not included in the publication is commercially sensitive as SNIPR Biome is in the processor securing patent protection for these aspects, which precludes their inclusion in the paper at this stage. Upon request, SNIPR Biome is willing to share all data with other parties with no competing interest. Additional information Extended data is available for this paper at https://doi.org/10.1038/s41587-023-01759-y. Supplementary information The online version contains supplementary materialavailableat https://doi.org/10.1038/s41587-023-01759-y. Correspondence and requests for materials should be addressed to Morten Otto Alexander Sommer. Peer review information Nature B/otechnoŕogythanks Antónia Sagona and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Reprints and permissions information is available at www.nature.com/reprints. Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y Extended Data Table 11 a Overview of the 15 CAPs generated from the selected WT phages resulting in four CAPs making up SNIPR001. The E. coli genes targeted by the five individual spacers and the sequence are listed below, b The E. coli genes targeted by the five individual spacers, that make up the array, and the sequence used in the CAPs Backbone Phage CAP Arrays Cas and Arrays Tail fiber engineering (original location/ added location) Select based on marginal utility (Supplementary Fig. 7) Select based on in vivo PK (Supplementary Fig. 8) Select based on accelerated stability at 40 °C (Supplementary Fig. 9) Select based on host range and in vivo efficacy (Supplementary Fig. 11) 15 a15.2 1-2-3 Separate sites a15wt/a17 / / / / a15.4 1-2-3 Separate sites a17/ a21 / / x 17 a17.2 4-5 co-transcribed NA / / x 20 a20.4 4-5 co-transcribed NA / / / / 31 a31.3 4-5 co-transcribed NA x a31.4 4-5 co-transcribed NA 33 a33.3 4-5 co-transcribed NA x a33.4 4-5 co-transcribed NA x 46 a46.3 4-5 co-transcribed NA x a46.4 4-5 co-transcribed NA / / / x 48 a48.3 4-5 co-transcribed NA x a48.4 4-5 co-transcribed NA / / / / 51 a51.4 4-5 co-transcribed NA x a51.5 4-5 co-transcribed NA / / / / a51.6 4-5 co-transcribed NA / / / x Total 15 8/15 8/8 6/8 4/6 spacer Target gene Sequence 1 bolA AGTGGGAAGGGTTGCAGGACACCGTCTTTGCC 2 rpoH CCGATGTTACCTTCCTGAATCAAATCCGCCTG 3 fimH CGAATGACCAGGCATTTACCGACCAGCCCATC 4 iptA TGATTGACGGCTACGGTAAACCGGCAACGTTC 5 murA GCTGTTAACGTACGTACCGCGCCGCATCCGGC Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y Extended Data Table 21 List of databases used for the analysis of SNIPR001 CAP sequences Database Source Analysis ResFinder 4.1 https://cge.cbs.dtu.dk/services/ResFinder Identification of acquired genes and/or chromosomal mutations mediating antimicrobial resistance Virulence Finder-2.0 https://cge.cbs.dtu.dk/services/VirulenceFinder/ Detection of virulence genes including exotoxins PHASTER Prophage/Virus DB https://phaster.ca/ Detection of potential transposases and integrases Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y Extended Data Table 31 Panel of bacterial strains (Aerobic: n=6. Anaerobic: n=3, Aerobic/Anaerobic: n=1) tested via a biopotency assay, showing Gram-type classification, growth conditions and source/I D Strain/name Gram type Growth condition Source/ID Acinetobacter bayiyi Negative Aerobic ATCC 33304 Kiebsieiia pneumonia Negative Aerobic SNIPRBiomelDb2951 Enterococcus faecaiis (Andrewes and Horder) Schleifer and Kipper-Balz Positive Aerobic ATCC 47077 Streptococcus thermophiius Orla-Jensen Positive Aerobic ATCC 19258 Baciiius coaguians Hammer Positive Aerobic ATCC 7050 Staphylococcus aureus subsp. aureus Rosenbach Positive Aerobic ATCC 12600 Eubacterium iimosum Eggerth Positive Anaerobic ATCC 8486 ßactero/des vuigatus Eggerth and Gagnon Negative Anaerobic ATCC 8482 ßactero/des thetaiotaomicron (Distaso) Castellani and Chalmers Negative Anaerobic ATCC 29148 £. co/i Negative Aerobic/Anaerobic Takara Cat. #636763 Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y Extended Data Table 41 Criteria used to evaluate results of the spot assay and define strain susceptibility following standards46'63 Spot assay categories Observation Strain susceptibility to SNIPR001 Plaques Visible plaques were counted, and PFU/mL was calculated by multiplying with volume and dilution Susceptible strain Lysis zone Impairment of bacterial growth observed as lysis zones. No plaques are visible; the highest di lution of visible zones is recorded Susceptible strain Negative Neither plaques nor lysis zones are detected Non-susceptible strain Nature Biotechnology Article https://doi.org/10.1038/s41587-023-01759-y Extended Data Table 51 Exact P-values resulting from the statistical analysis of the data shown in Fig. 6D Test used Time point Comparison group Nominal P-value FDR corrected with Holm's method for timepoint Two-sided Mann-Whitney Utest Day 2-8 hours after first dose. 1 dose total SNIPR001 High 1,08E-05 3,25E-05 SNIPR001 Medium 1/15E-02 2,30E-02 SNIPR001 Low 1,26E-02 2,30E-02 Day 3-24hours after first dose. 3 doses total SNIPR001 High 1,08E-05 3,25E-05 SNIPR001 Medium 3,25E-04 3,25E-04 SNIPR001 Low 2/17E-05 4,33E-05 Day 4-48hours after first dose. 6 doses total SNIPR001 High 4,37E-04 1,31 E-03 SNIPR001 Medium 1,23E-01 1,23E-01 SNIPR001 Low 4,87E-04 1,31 E-03 Two-sided Kruskill-Wallis test Day 2-8 hours after first dose. 1 dose total SNIPR001 low, medium, and high 3,07E-03 NA Day 3-24hours after first dose. 3 doses total SNIPR001 low, medium, and high 2,07E-03 NA Day 4-48hours after first dose. 6 doses total SNIPR001 low, medium, and high 1,04E-02 NA Nature Biotechnology nature research Reporting Summary Corresponding author(s): Morten Sommer Last updated by author(s): Mar 14, 2023 Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist. Statistics For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section. n/a □ □ □ □ □ □ □ Confirmed _>^1 The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement ^1 A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly K-pi The statistical test(s) used AND whether they are one-or two-sided ^ Only common tests should be described solely by name; describe more complex techniques in the Methods section. "J A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates {e.g. regression coefficient) ^ AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) rt-7| For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted ^ Give P values as exact values whenever suitable. "j For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings "J For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes ^1 Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated Our web collection on statistics for biologists contains articles on many of the points above. Software and code Policy information about availability of computer code Data collection Figures and key statistics were generated using R version 4.1.0. For figure generation the following packages were used: RcolorBrewer v. 1.1-2, ape v. 5.5, ggsignif v. 0.6.2, ggpubr v. 0.4.0, matrixStats 0.59, reshape2 v. 1.4.4, ggimage v. 0.3.0, here v. 1.0.1, purr v. 0.3.4, ggtree63 v. 3.0.2, systemfonts v. 1.0.2, Cairo v. 1.5-12.2, cowplot v. 1.1.1, reaxxl v. 1.3.1, and ggplot2 v.3.3.3, openxlsx, v. 4.2.3, patchwork v. 1.1.1, dplyr v. 1.0.7, and ggh4x v. 0.2.3. Averages and standard deviations are calculated after transforming the values to the scale shown on a given figure, e.g. when a loglO scale is used, the averages and standard deviations are calculated after loglO transformation. The synteny plot was then generated using a custom Python (v 3.7.10) script, using the drawSvg library (v 1.9.0). Data analysis All code needed to produce this study is available at https://github.com/sniprbiome/SNIPR001_paper. For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information. Data Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A list of figures that have associated raw data -A description of any restrictions on data availability All data and results that were generated during this study is deposited at https://github.com/sniprbiome/SNIPR001_paper. Additional data are available in the Article, Online methods and Supplementary tables. The MLST database was downloaded on July 1, 2021, from the MLST2 repository (https://bitbucket.org/ genomicepidemiology/mlst_db/src/master/). For annotation of the CAP sequences the following tools and datasets were used ResFinder 4.1 (https:// cge.cbs.dtu.dk/services/ResFinder), VirulenceFinder-2.0 (https://cge.cbs.dtu.dk/services/VirulenceFinder/), PHASTER Prophage/Virus DB (https://phaster.ca/). To reproduce the results, no further data is needed. Phage genome sequences are deposited at Genbank under access numbers OQ067373 - 76 Field-specific reporting Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection. Life sciences ] Behavioural & social sciences ] Ecological, evolutionary & environmental sciences For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summarv-flat.pdf Life sciences study design All studies must disclose on these points even when the disclosure is negative. Sample size No explicit sample size calculations were performed since the strains assayed where part of the SENTRY surveillance program. The JMI panel comprises of 382 strain E. coli clinical collection obtained from JMI Laboratories (North Liberty, IA, USA). These strains were isolated from patients with bloodstream infections hospitalized in hematology and oncology units across four different regions (Asia-Pacific 54 isolates, Europe 161 isolates, Latin America 26 isolates, and North America 141 isolates), sourced through the SENTRY Antimicrobial Surveillance Program (2018-2020), which is composed of a network of more than 150 medical centers in more than 28 countries worldwide (https:// I www.jmilabs.com/sentry-surveillance-program). Data exclusions no exclusions were performed Replication all experiments contain at least two biological replicates and for each method the number of technical replicates are stated. The SNIPR001 I assay against the 382 strains was performed in duplicate and the duplicate results are explicitly shown in Figure 5B by the two dots. Randomization [not relevant for this work as all 382 E. coli isolates were exposed to SNIPR001 and received the same treatment. Blinding not relevant for this work as the 382 E. coli isolates were exposed to SNIPR001. More specifically for counting of phage titers, phage lysates or the equal volume mix of SNIPR001 CAPs were serially diluted 10-fold in SM buffer or PBS, respectively. Bacterial lawns were prepared by adding 100 or 300 uiof bacterial overnight culture to 3 or 10 mLof 0.5% top agar (containing Ca2+and Mg2+), which was vortexed briefly and poured onto a round or square LB plate. Five u.1 of the dilution series of test phages were then spotted onto lawns and left to dry at room temperature with an open lid prior to incubation at 372C overnight. The strains b52, b2479 and bl7 were used as controls of the assay and included in each round of assays. The next day, results were assessed (Extended Data Table 4). In this assay, a susceptible strain is defined as one producing plaques that are countable in PFU/mL as well as one without visible plaques but demonstrating impairment of bacterial growth (i.e., lysis zones). Coverage defines the percentage of the total number of susceptible strains. Images of all plates were recorded. Figures illustrating efficiency of plating results first had titers loglO transformed and then standard deviances and averages were calculated subsequently. The clinical panels and control strains were tested in two independent experiments. Reporting for specific materials, systems and methods_ We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material, system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response. Materials & experimental systems n/a □ Involved in the study "J Antibodies ~2 Eukaryoticcelllines "j Palaeontology and archaeology ^1 Animals and other organisms "j Human research participants "J Clinical data ~2 Dual use research of concern Methods n/a Involved in the study □ ChlP-seq ~2 Flow cytometry "J MRI-basedne uroimaging Animals and other organisms Policy information about studies involvine animals: ARRIVE guidelines recommended for reporting animal research Laboratory animals [female CD-I® IGS mice (approximately 6-7 weeks of age upon arrival) from Charles River (Freiburg, Germany). And female Gbttingen 2 Laboratory animals (minipigs (approximately 4-7 months of age upon arrival) from Ellegaard Gottingen minipigs A/S, Denmark was used for tolerability [and kinetic studies. Wild animals did not involve wild animals Field-collected samples no field samples were collected - - Ethics oversight All procedures were conducted in accordance with guidelines from the Danish Animal Experiments Inspectorate, Ministry of Environment and Food of Denmark and in accordance with the institutional license (BioAdvice, animal license no. 2015-15-0201-00540). Note that full information on the approval of the study protocol must also be provided in the manuscript.