Bacterial Genotoxin Accelerates Transient Infection–Driven Murine Colon Tumorigenesis

Bacterial infections have emerged as important environmental factors contributing to the global cancer burden (1). Several bacteria highly associated with human colon tumors, including enterotoxigenic Bacteroides fragilis, Fusobacterium nucleatum, and Escherichia coli strains that harbor the polyketide synthase genomic island, were recently shown to accelerate colon tumor development (2–4). These findings underscore the crucial roles of persistent infection and chronic inflammation in bacterially driven colon tumorigenesis. As major etiologic agents for diarrheal illnesses, the human pathogens enteropathogenic E. coli (EPEC) and enterohemorrhagic E. coli (EHEC), along with their murine equivalent Citrobacter rodentium (CR), elicit transient and noninvasive infections in the host via forming attaching and effacing (A/E) lesions to the intestinal epithelium (5–8). Although tumorigenic potential was indicated by early evidence (9–11), the impact of noninvasive infection by A/E pathogens on bacterially promoted colon tumorigenesis and the underlying mechanisms have remained obscure.

To explore whether A/E pathogens possess intrinsic genotoxic capability, we infected germ-free (GF) C57BL/6J mice by oral gavage with vehicle control, wild-type (WT) CR, or ΔescN CR, an isogenic mutant with failure to inject the type III secretion system (T3SS) virulence proteins into host cells due to a defective T3SS (12). Comparable CR loads were detected in the stools from WT and ΔescN CR–infected GF C57BL/6J mice, peaking 2 days postinfection (dpi) and maintaining the same colonization levels thereafter (Fig. 1A). Prominent immunofluorescence staining of phosphorylated histone H2AX (γH2AX), a sensitive biochemical marker for DNA double-strand breaks (13), on colon tissue sections derived from the infected animals revealed that WT CR infection, in comparison to vehicle control, elicited substantial DNA damage within the colonic crypts (Fig. 1B). In contrast, no detectable γH2AX staining indicative of DNA damage was observed in ΔescN CR–infected colons (Fig. 1B), despite comparable levels of overall pathogen burden in WT and ΔescN CR–infected GF animals (ref. 14; Fig. 1A). These results indicate that in the absence of a gut microbiome, CR per se prompts T3SS-dependent genotoxic damage to colonic tissue during infection in vivo, in line with the critical role of T3SS in CR infection–induced colonic crypt hyperplasia (15). We further conducted in vitro CR infection in SW480 cells to ascertain the intrinsic genotoxicity of CR. Indeed, infection with WT CR, compared with vehicle control, augmented the accumulation of γH2AX in SW480 cells, peaking at 1 hour postinfection, whereas the augmented γH2AX formation was attenuated in the cells infected with ΔescN CR (Fig. 1C and D). These results suggest that CR, via the T3SS, directly provokes genotoxic stress in the infected hosts.

The T3SS in A/E pathogens has been well documented to inject a myriad of virulence proteins into the host cells, where they interrupt multiple signaling pathways (5, 16, 17). Recent studies using stable isotope labeling with amino acids in cell culture–based mass spectrometry have identified an ever-expanding repertoire of the secreted T3SS virulence proteins in A/E pathogens (18, 19). To identify the T3SS-dependent virulence protein (effector) that confers the intrinsic DNA-damaging capability during CR infection (Fig. 1B), we ranked the T3SS-secreted proteins based on their putative functions, abundance, and conservation among A/E pathogens (Supplementary Table S1). The top 11 candidate proteins were subjected to in vitro DNA digestion assays using λ DNA as a substrate and evaluated for their potential direct DNA-damaging capabilities. Although most candidate T3SS effectors, including FkpA, ManX, NleA, and others, failed to initiate detectable DNA lesions, recombinant UshA led to more than 50% cleavage of λ DNA (Fig. 1E and F). This result demonstrates that UshA harbors intrinsic DNA-damaging capability in vitro. Using the isogenic ΔushA CR strain (with chromosomal deletion of UshA) and UshA-specific antibody (Supplementary Fig. S1A), we ascertained that UshA was secreted into the culture supernatant by CR, in an EscN/T3SS–dependent manner, when cultured in DMEM that was previously reported to induce T3SS secretion (ref. 14; Supplementary Fig. S1B). Moreover, using the type III secretion and translocation assays (20), we revealed that the N-terminus of UshA, when fused with the fluorescence-based reporter TEM1 β-lactamase, was secreted and translocated in an EscN/T3SS–dependent manner into host cells during infection (Supplementary Fig. S1C and S1D). Similar results were observed with two known T3SS effectors, NleJ and EspZ (19), but not HSPs DnaK and HtpG (Supplementary Fig. S1C and S1D), hence indicating that UshA harbors an N-terminal translocation signal as previously reported in a handful of T3SS effectors (20). Furthermore, subcellular fractionation assays showed that UshA was detected in the cytoplasm and the nucleus of SW480 cells infected with WT, but not ΔescN, CR (Supplementary Fig. S1E), further supporting the T3SS-mediated UshA injection into host cells during CR infection. Together, these results suggest that the T3SS-secreted UshA of CR could be a novel genotoxin effector.

Consistent with our in vitro screening assays (Fig. 1E and F), CR-UshA recombinant protein digested λ DNA in a dose- and time-dependent manner (Fig. 2A–D). This apparent DNA-damaging capability, expanding from the previously reported 5′-nucleotidase/UDP–sugar hydrolase (21) and NAD degradation activities (22), indicates that UshA could be a versatile bacterial enzyme. Alignment of amino acid sequences revealed that UshA is conserved among A/E pathogens as well as nonpathogenic E. coli K-12 (Supplementary Fig. S2A); as expected, recombinant EPEC, EHEC, and E. coli K-12–UshA proteins displayed comparable capacities in damaging DNA to their CR equivalent (Fig. 2E and F; Supplementary Figs. S2B–S2I and S3A–S3D). Of note, a catalytic histidine–aspartic acid (His–Asp) dyad (Fig. 2G), which was proposed to be critical for the 5′-nucleotidase activity of E. coli K-12–UshA (23), is identical among A/E pathogens and E. coli K-12 (Fig. 2H). To assess the potential impact of the His–Asp dyad on DNA-damaging activities of UshA, we generated a dual-mutant UshA, termed UshA^DM, with corresponding histidine–aspartic acid substituted with alanine–alanine (Fig. 2H). In striking contrast to WT UshA, CR-UshA^DM failed to digest λ DNA across a wide range of final concentrations (Fig. 2I and J; Supplementary Fig. S3A and S3B). Similarly, the DNA-damaging capacities of EPEC, EHEC, or E. coli K-12–UshA were markedly attenuated when the histidine and aspartic acid were mutated (Supplementary Fig. S3C–S3F), thus supporting the importance of a catalytic His–Asp dyad for the DNA-damaging activity of UshA in vitro. Besides linear double-stranded λ DNA, our in vitro DNA digestion assays revealed that CR-UshA cleaved circular plasmid DNA (Fig. 2K and L; Supplementary Fig. S3G and S3H), single-stranded DNA (Supplementary Fig. S3I and S3J), and extracted chromatin DNA (Fig. 2M and N; Supplementary Fig. S3K–S3O); likewise, EPEC, EHEC, and E. coli K-12–UshA displayed comparable activities toward these substrates (Supplementary Fig. S3C–S3J and S3M–S3O). In contrast, the DNA-digestive activities of the corresponding UshA^DM proteins were almost abolished, no matter which form of DNA substrate was used (Fig. 2I–N; Supplementary Figs. S3C–S3J and S3M–S3O). These results suggest the pivotal role of the catalytic His–Asp dyad in conferring the DNA-damaging enzymatic activity of UshA in the in vitro DNA digestion assays.

As illustrated by genomic comparisons, the most striking genetic elements distinguishing A/E pathogens from nonpathogenic E. coli K-12 concentrate on the locus of enterocyte effacement (LEE) pathogenicity island, which is well conserved among CR, EPEC, and EHEC but absent in E. coli K-12 (Supplementary Fig. S4A). Given that an array of essential structural components, regulators, and chaperones of T3SS are encoded by the LEE pathogenicity island (Supplementary Fig. S4B; ref. 17) and that UshA is secreted and translocated in a T3SS/EscN–dependent manner by A/E pathogens, but not E. coli K-12 (Supplementary Figs. S1B–S1E, S4C and S4D), we speculated that UshA is a novel T3SS-dependent genotoxin, which, upon translocation, could elicit DNA damage in host cells. In mammalian cells, DNA damage responses (DDR) consisting of sophisticated signaling pathways and posttranslational modifications are rapidly activated upon detection of damaged DNA lesions (24). Indeed, the levels of γH2AX and the phosphorylated check kinase 1 (Chk1) and Chk2, two known effector kinases in DDR (25), were substantially elevated in SW480 cells during WT CR infection in comparison to vehicle control (Supplementary Fig. S5A and S5B). The WT CR infection–elicited DNA damage signaling cascade in SW480 cells was substantially attenuated when infected with ΔushA CR (Fig. 3A and B); in contrast, infection with ΔushA::UshA CR, where UshA was chromosomally complemented in the ΔushA strain, yielded a comparable UshA level as in WT CR (Supplementary Fig. S5C) and restored the phosphorylation levels of H2AX, Chk1, and Chk2 during CR infection (Fig. 3A and B). Moreover, complementing a catalytic-dead UshA^DM mutant in ΔushA CR, which led to almost identical UshA expression as in WT or ΔushA::UshA CR (Supplementary Fig. S5C), failed to trigger profound phosphorylation signaling cascade in SW480 cells (Supplementary Fig. S5D and S5E), underscoring the importance of the catalytic His–Asp dyad in UshA-conferred DNA damage in infected cells. Of note, WT, ΔushA, ΔushA::UshA, and ΔushA::UshA^DM CR strains proliferated at almost identical rates in Luria–Bertani (LB) and DMEM (Supplementary Fig. S5F), and they attached to SW480 cells to comparable extents (Supplementary Fig. S5G). Similarly, genetic manipulation of UshA did not affect the growth of EPEC in LB and DMEM and the attachment of EPEC strains to SW480 cells during infections (Supplementary Fig. S6A–S6C); however, WT EPEC triggered more profound elevation in the levels of γH2AX and phosphorylated Chk1 and Chk2 in SW480 cells, in a UshA-dependent and catalytic dyad–dependent manner (Fig. 3C; Supplementary Fig. S6D–S6F). These results suggest that although dispensable for bacterial growth and attachment to infected cells, UshA executes a critical function in initiating genotoxic stress and eliciting DDR signaling cascade in cultured cells during A/E pathogen infections.

The UshA-dependent DDR signal cascade in CR- and EPEC-infected cells (Fig. 3A–C) led us to further examine the subcellular localization of UshA during infections. Indeed, UshA was visualized by immunofluorescence staining in the cytoplasm and the nucleus of SW480 cells infected with WT, ΔushA::UshA, or ΔushA::UshA^DM CR as well as EPEC strains. Conversely, there was no detectable staining of UshA in the cells infected with vehicle control, ΔushA CR, or ΔushA EPEC (Fig. 3D; Supplementary Fig. S6G). Of note, the majority of nuclear UshA colocalized with γH2AX in the nucleus, as expected in SW480 cells responding to genotoxic stresses caused by CR and EPEC infections (Fig. 3D–F; Supplementary Fig. S6G). Interestingly, the mean fluorescence intensities (MFI) of γH2AX positively correlated with those of WT UshA, but not catalytically dead UshA^DM, in the nuclei of infected cells (Supplementary Fig. S6H). In line with the dose-dependent DNA damage caused by recombinant CR, EPEC, and EHEC UshA proteins (Fig. 2E), the positive correlation of nuclear γH2AX and WT, but not catalytic mutant, UshA MFIs in CR- and EPEC-infected SW480 cells supports that CR- or EPEC-injected endogenous UshA with intact catalytic activity is required to produce genotoxic DNA damage in host cells during A/E pathogen infections.

To ascertain that the nuclear localization of UshA causes genotoxic DNA strand breaks in the host cells during CR and EPEC infections, we performed single-cell gel electrophoresis–based alkaline comet assays, a sensitive method for direct detection of DNA strand breaks (26). Indeed, the DNA damage severity, as measured by comet tail moment, was substantially increased in SW480 cells infected with WT CR or EPEC compared with vehicle control–infected cells (Fig. 3G–I; Supplementary Fig. S6I). In contrast, comet tail moments of SW480 cells infected with ΔushA CR or ΔushA EPEC were attenuated to the levels comparable to those of vehicle control–infected cells. Of note, infections with ΔushA::UshA CR or EPEC restored the comet tail moments of SW480 cells to the levels of WT CR-infected ones, whereas ΔushA::UshA^DM strains failed to do so (Fig. 3G–I; Supplementary Fig. S6I). Thus, consistent with the elicited cellular DDR signaling (Fig. 3A–C) and UshA nuclear translocation (Fig. 3D–F), these results demonstrate that UshA, in particular its catalytic His–Asp dyad, is crucial for CR- and EPEC-induced DNA strand breaks in SW480 cells during infections.

The LEE pathogenicity island-absent nonpathogenic E. coli K-12 is known to form neither T3SS-conferred intimate attachment nor effacement lesions to the cultured cells (27). Consistently, infection with E. coli K-12 barely elevated γH2AX levels in SW480 cells (Supplementary Fig. S7A), and the comet tail moments of vehicle control–infected and E. coli K-12–infected cells were largely comparable (Supplementary Fig. S7B and S7C). These results demonstrate that E. coli K-12 failed to elicit genotoxic damage in host cells, despite that recombinant E. coli K-12 UshA shares the in vitro DNA-digesting capacity of recombinant CR- and EPEC-UshA (Supplementary Fig. S3C and S3D). In contrast, CR and EPEC possessing functional T3SS elicited robust DDR in host cells during infections (Fig. 3A–C and G–I; Supplementary Fig. S7A–S7C). Hence, these results indicate that T3SS-conferred injection and translocation into host cells could be a prerequisite for UshA to execute its genotoxin function during infections.

We further infected GF C57BL/6J mice with vehicle control, WT, ΔushA, ΔushA::UshA, and ΔushA::UshA^DM CR strains to assess the impact of UshA on CR-elicited genotoxic DNA damage in animals. As expected, there were no differences in bacterial colonization and proliferation among these CR strains, with peak CR loads from 2 dpi (Supplementary Fig. S8A). Immunofluorescence staining on the colon tissue sections derived at 7 dpi revealed that WT and ΔushA::UshA CR, compared with vehicle control, caused substantial genotoxic damage in the infected mice, as illustrated by the increased levels of γH2AX foci formation, whereas the γH2AX staining was diminished on the colon sections infected by ΔushA CR or ΔushA::UshA^DM CR (Supplementary Fig. S8B). Moreover, prominent γH2AX foci were exhibited on the colon tissue sections derived from specific pathogen–free (SPF) C57BL/6J mice infected with WT and ΔushA::UshA CR strains, whereas in contrast, infections with vehicle control, ΔushA, or ΔushA::UshA^DM CR failed to do so (Fig. 4A). Despite comparable levels of colonic colonization of these CR strains in GF and SPF C57BL/6J mice, ΔushA and ΔushA::UshA^DM CR, in contrast to WT and ΔushA::UshA strains, caused less profound genotoxic damage in the infected animals (Fig. 4A; Supplementary Fig. S8B). These results suggest that UshA, regardless of the microbiome, is crucial for CR to execute its intrinsic genotoxic capability, and the UshA catalytic activity serves as a critical determinant for the genotoxicity during CR infection in mice.

Substantial evidence implicates that accumulated DNA damage facilitates tumor initiation and progression (24). To assess whether UshA-conferred genotoxic lesions in the colon during CR infection could lead to transformative changes, we used the ex vivo three-dimensional organoids culture system. In these experiments, colon organoids were derived from colonic crypts of SPF C57BL/6J mice infected with vehicle control, WT, ΔushA, ΔushA::UshA, and ΔushA::UshA^DM CR strains at 14 dpi. Indeed, the colonic organoids derived from WT CR–infected colons grew significantly faster and larger than those from vehicle control–infected ones (Fig. 4B–D), indicating that CR augments the stemness and tumorigenic potential of colon cells in the infected C57BL/6J mice (28). In contrast, the growth rates and sizes of the organoids derived from ΔushA CR–infected colons were almost identical to the vehicle controls. Notably, only the organoids derived from ΔushA::UshA CR–infected colons exhibited similar growth rates and sizes to those from WT CR–infected colons, whereas ΔushA::UshA^DM CR infection did not accelerate the colonic organoid growth (Fig. 4B–D). These results suggest that UshA, with intact catalytic activity, confers genotoxic stress and substantially elevates colonic stemness during CR infection in C57BL/6J mice.

To assess the impact of UshA-conferred genotoxic damage during infection on colonic tumorigenesis, we orally inoculated vehicle control, WT, ΔushA, ΔushA::UshA, and ΔushA::UshA^DM CR strains in genetically susceptible Apc^MinΔ716/+ mice. These mice express a mutant Apc tumor suppressor and spontaneously develop adenomas in the colon; they have been widely used to evaluate environmental factors contributing to colon tumorigenesis (refs. 3, 29; Fig. 4E). The colonization, proliferation, and clearance of all these CR strains in the inoculated Apc^MinΔ716/+ mice, as reflected by the recovery of live CR from fecal samples, were largely comparable (Fig. 4F). Indeed, the bacterial loads of these strains peaked at 4 dpi, declined gradually thereafter, and were no longer detectable after 26 dpi (Fig. 4F), a time course similar to the bioluminescent CR (Supplementary Fig. S9A). To ascertain that CR did not cause a persistent, low-level infection in the colon, we monitored the colonic localization of CR in infected Apc^MinΔ716/+ mice using bioluminescent CR and sensitive PCR-based assays. Indeed, the majority of CR was detectable in the distal colon at 15 dpi, as illustrated by bioluminescence; however, no bioluminescence was detectable in any part of the colon at 35 dpi (Supplementary Fig. S9B). Moreover, the CR espA and espF gene-specific PCR products that were amplified using DNA isolated from the stool of CR-infected Apc^MinΔ716/+ mice revealed similar trends (Supplementary Fig. S9C). Together, these data suggest that after clearance, no trace amount of CR was detectable in the feces and the colon of infected Apc^MinΔ716/+ mice, which is consistent with the reported self-limiting and noninvasive infection of CR in C57BL/6J and many other mouse strains (5).

We sought to assess the far-reaching impact of UshA on colonic tumor formation at 3 months postinfection, a time point well after the clearance of CR infection, in Apc^MinΔ716/+ mice infected with vehicle control or CR strains (Fig. 4E). As previously reported (29, 30), vehicle control–inoculated Apc^MinΔ716/+ mice spontaneously developed few adenomas in the cecum and distal colon at 12 weeks of age, as conveyed by whole-mount methylene blue staining (Fig. 4G and H); WT CR infection dramatically accelerated colonic adenoma development, as reflected by increased tumor numbers and tumor load (Fig. 4G–I). Moreover, the tumor tissues in the colon were further evaluated by histologic analysis and revealed substantially elevated β-catenin signaling (Supplementary Fig. S9D). In striking contrast, deletion of UshA in CR nearly abolished the infection-accelerated colon tumor formation (Fig. 4G–I), without affecting the bacterial colonization and clearance (Fig. 4F) in Apc^MinΔ716/+ mice. To examine the possibility that ΔushA CR-diminished promotion in colon tumor development may be caused by impaired colonization and localization of this strain in the colon, we inoculated Apc^MinΔ716/+ mice with vehicle control, WT, or ΔushA bioluminescent CR, respectively (Supplementary Fig. S10A). Indeed, WT and ΔushA bioluminescent strains colonized predominantly in the distal colon with comparable bioluminescence density at 8 dpi (Supplementary Fig. S10B), further indicating that UshA does not affect colonic colonization, proliferation, and distribution of CR in the infected animals. In addition, to assess whether UshA deletion leads to any fitness defect or competitive disadvantage during CR infection in Apc^MinΔ716/+ mice, the animals were infected with a mixture of WT and ΔushA CR strains at a 1:1 ratio (Supplementary Fig. S10C). Via a genomic DNA–targeted PCR strategy (Supplementary Fig. S10D), we confirmed that colonization of WT and ΔushA strains in this mixed CR infection in Apc^MinΔ716/+ mice was comparable to their separate inoculations (Fig. 4F; Supplementary Fig. S10E). Of note, the ratios of WT CR to ΔushA CR remained approximately 1:1 at 6 dpi and 13 dpi, respectively, as in the mixed inoculum prior to infection (Supplementary Fig. S10F). These results demonstrate that under the same infection microenvironment in Apc^MinΔ716/+ mice, ΔushA CR displayed no detectable fitness defect or competitive disadvantage in comparison with the WT strain, which is in line with the finding that UshA is dispensable for CR growth and proliferation in LB and DMEM in vitro (Supplementary Fig. S5F). Furthermore, the crucial role of UshA in colon tumorigenesis was further supported by the evidence that infection with the ΔushA::UshA strain restored the promotion of colon tumor development in Apc^MinΔ716/+ mice to the comparable levels as WT CR; conversely, inoculation of ΔushA::UshA^DM CR did not accelerate tumor development (Fig. 4G–I). Of note, infections with WT, ΔushA, ΔushA::UshA, and ΔushA::UshA^DM CR strains did not significantly affect tumor sizes (Fig. 4J). All together, these results indicate that UshA, via the catalytic His–Asp dyad, conferred DNA-damaging capability and executed an important role in initiating tumor development in the CR-infected colons, even after complete clearance of CR in the host.

We further examined the possibility that UshA-induced colonic tumor development was affected by CR-initiated immune responses during infection in Apc^MinΔ716/+ mice. Indeed, CR infection caused a robust lengthening and thickening of colonic crypts and almost complete loss of goblet cells in infected Apc^MinΔ716/+ mice compared with the vehicle controls at 11 dpi (Supplementary Fig. S11A and S11B), coinciding with the peak bacterial load in the inoculated animals (Fig. 4F). However, infections with WT and ΔushA CR led to almost identical tissue damage and inflammation in the infected Apc^MinΔ716/+ mice, as illustrated by hematoxylin and eosin staining and pathologic analysis (Supplementary Fig. S11A and S11B). This result indicates that UshA may be dispensable for CR-elicited tissue inflammation. Moreover, there were no discernable differences in the elevated expression of known proinflammatory cytokines/chemokines, including Nos2, Il17a, Cxcl1, Cxcl2, and Ccl2 in the colons infected by WT CR and the ΔushA strain, in comparison to the vehicle control (Supplementary Fig. S11C). Consistent with the self-limiting infection and comparable clearance dynamics in Apc^MinΔ716/+ mice (Fig. 4F), our immune profiling in the lamina propria lymphocytes demonstrates that WT and ΔushA CR comparably triggered colonic immune responses (Supplementary Fig. S11D–S11H). These results suggest that UshA is dispensable for CR-elicited acute immune response and tissue inflammation in Apc^MinΔ716/+ mice, thus ruling out the possibility that UshA deletion attenuates colon tumor development via dampening the infection-associated immune responses in the host.

We characterized the mutational profiles in colon tumors derived from Apc^MinΔ716/+ mice infected with vehicle control, WT CR, and ΔushA CR by whole-exome sequencing (WES) to assess whether UshA-conferred DNA damage results in any distinct mutational signatures observed in human colon cancers. To specifically identify somatic alternations in colon tumor development, the reference DNAs from the matched liver tissue of Apc^MinΔ716/+ mice were obtained for WES and used for germline variant filtering (31). Single-base substitutions (SBS) are the most identified somatic mutations in colon tumors derived from Apc^MinΔ716/+ mice infected with vehicle control (552 SBS in 10 tumors), wild-type CR (733 SBS in 13 tumors), and ΔushA CR (801 SBS in 13 tumors), and their distinct SBS mutational catalogs were profiled by SBS 96-trinucleotide analysis (ref. 32; Fig. 5A). The SBS mutational signatures currently collected in the Catalogue of Somatic Mutations in Cancer (COSMIC) database (https://cancer.sanger.ac.uk/cosmic) were fit to the observed SBS mutational catalogs, and our analyses revealed that SBS12, SBS15, SBS17a, SBS23, SBS26, SBS28, SBS37, and SBS89 were the top eight mutational signatures that contribute to the mutational catalogs of colon tumors derived from the infected Apc^MinΔ716/+ mice (Fig. 5B). Among them, SBS15, SBS17a, SBS26, SBS28, and SBS37 were previously reported to contribute to human colorectal cancers (Fig. 5C); SBS89 was identified as a mutational signature of colorectal epithelial cells prior to neoplastic alterations (33). Moreover, SBS89, which is featured with a high percentage of T>G mutation at GTG trinucleotides (the mutated base is underlined) unique among all SBS mutational signatures (Fig. 5D), contributed significantly to the SBS mutational catalogs observed in colon tumors derived from vehicle control–, wild-type CR–, and ΔushA CR–infected Apc^MinΔ716/+ mice (Fig. 5B and E). These results indicate that the colon tumor mutational spectra in Apc^MinΔ716/+ mice are highly related to human colorectal cancers. Intriguingly, the relative contributions of SBS26, which is characterized by high proportions of T>C mutations (Fig. 5F), were substantially augmented in the mutational catalogs in the colon tumors from WT CR–infected Apc^MinΔ716/+ mice compared with those from vehicle control–infected animals, whereas such elevated contributions were attenuated in the tumors from ΔushA CR–infected animals (Fig. 5B and G). Of note, despite that SBS26 and SBS15 were both linked with defective DNA mismatch repair (34), the UshA-dependent augments in the relative contributions of SBS26, but not SBS15, to the tumor mutational spectra indicate that the SBS26-specific defect in the DNA mismatch repair pathway may confer the accelerated colon tumorigenesis in WT CR–infected Apc^MinΔ716/+ mice (Fig. 4H and I).

Our results demonstrate that UshA is a novel T3SS-dependent genotoxin, which confers the intrinsic genotoxicity of A/E pathogens and accelerates the bacterially induced colon tumorigenesis in mice. UshA is metabolically dispensable for bacterial growth and proliferation in culture; deletion of UshA does not affect bacterial colonic colonization, proliferation, distribution, and fitness during CR infection in animals. Notably, the secretion and translocation of UshA into host cells during infection are conferred by the T3SS encoded in the LEE pathogenicity island, which is well conserved among A/E pathogens. The catalytic His–Asp dyad–conferred DNA-damaging activity of UshA is indispensable for in vitro DNA digestions and activation of the DNA-damaging cascade in in vitro CR infection in cultured cells, in vivo CR infection in GF and SPF mice, and CR-promoted colon tumor formation in Apc^MinΔ716/+ mice. These results support the crucial role of UshA in eliciting genotoxic stress, cell stemness, and tumorigenesis in the infected animals. Although it has been extensively investigated how T3SS-translocated effectors interfere with and subvert inflammatory response and immune defensive signaling pathways in host cells during CR infection, the long-standing impacts of these crucial host–pathogen interactions on host tissue/cells have remained elusive. Indeed, the infected colon tissue/cells have sustained DNA damage/genomic instability caused by the T3SS-translocated UshA while rapidly responding to a substantial number of other effectors injected along with UshA via the T3SS during infection. Hence, underneath the frontline of host–pathogen interactions concentrating on the immune signal cascades, UshA, with its DNA cleavage activity, induces DNA damage in the infected hosts. The substantially elevated burden in DNA damage/genomic instability, conferred by UshA during CR infection, accelerates mutations in the infected colonic cells/tissue. These mutations, resulting from DNA double-strand breaks, are comparable by nature to those induced by other insults such as ionizing irradiation. Such accumulated mutations are crucial and sufficient for initiating the stemness and transformative changes even after the complete clearance of CR infection. As such, transient infection of CR, using the T3SS-dependent genotoxin effector UshA, elicits DNA damage and promotes far-reaching impact on accelerating colonic tumor formation.

CR-induced colitis in mice and EPEC/EHEC-caused diarrheal symptoms in humans are generally believed to be transient disorders that are ameliorated upon pathogen clearance (35, 36). However, our results demonstrate that acute and noninvasive infections with A/E pathogens could have far-reaching impact, particularly on the development of colon cancer, a tumor known to have a lead time of typically at least a decade from initiation to diagnosis. Each year, more than 200 million episodes of EPEC/EHEC infection–elicited diarrheal disease are reported worldwide (37), among which the populations in sub-Saharan Africa have had higher exposure to A/E pathogen infections (38). Notably, colon cancer incidence rates in the countries in sub-Saharan Africa have increased steadily in the past decades, as reported in an increasing number of epidemiologic studies (39, 40). Indeed, colon cancer is the fifth most common malignancy in sub-Saharan Africa (41), frequently occurring at an early age with distinctive histologic characteristics (40). A disproportionately high number of patients with colon cancer in these countries are young (19%–38% are individuals younger than 40 years compared with 3%–7% in developed countries; refs. 39, 40), and most early-onset colon cancers are sporadic adenocarcinomas (40); the left-sided colon cancers are predominant in sub-Saharan Africa cases, whereas the right-sided lesions are more prevalent in the developed world (39). These distinctive characteristics indicate that the etiologic mechanisms of colon cancers in sub-Saharan Africa could differ from those in the developed world. Of note, our results here demonstrate that acute and noninvasive infection with A/E pathogen CR harbors a far-reaching impact, in a genotoxin UshA–dependent manner, on the development of colon cancer in mice. It is indicative that the prevalent EPEC/EHEC infections in sub-Saharan Africa may be one of the previously neglected important environmental factors that contribute to the rising incidence of colon cancers in this region. Given the global prevalence of EPEC and EHEC infections (42) and the anticipated surge in global colon cancer (43), approaches to interrupt the UshA-conferred intrinsic genotoxicity in A/E pathogens could be relevant to the development of more robust strategies for prevention of colon cancer globally.

All animal experiments were performed according to protocol number MO19H269, approved by the Johns Hopkins University's Animal Care and Use Committee and in direct accordance with the NIH guidelines for housing and care of laboratory animals. Apc^MinΔ716/+ mice (expressing a mutant gene encoding an adenomatous polyposis coli protein truncated at amino acid 716) and wild-type C57BL/6J mice, fed autoclaved food and water ad libitum, were maintained in a specific pathogen-free or germ-free facility, as previously described (29, 30).

SW480 (CCL-228) and HeLa (CCL2) cells were obtained from ATCC and cultured in DMEM containing 10% FCS, 2 mol/L glutamine, 100 U/mL penicillin, and 100 U/mL streptomycin. The following antibodies were used: anti-DnaK (8E2/2) from Thermo Fisher Scientific; anti-γH2AX (JBW301) from Millipore; anti-H2AX (A300–082A) from Bethyl Laboratories; anti-PARP1 (#9542), anti–Caspase-3 (#9662), anti-pChk1 (D12H3), and anti–Histone H3 (D1H2) from Cell Signaling Technology; anti-Chk1 (G-4) and anti-Chk2 (A-12) from Santa Cruz Biotechnology; anti-pChk2 (NB100–92592) from Novus Biologicals; and anti–β-actin (AC-15) from Sigma-Aldrich. Polyclonal antibodies specific for CR and UshA were generated using heat-killed CR or recombinant CR UshA protein as immunogens, respectively, following a standard immunization protocol. 4′,6-Diamidino-2-phenylindole (DAPI), dithiothreitol (DTT), and ethylenediaminetetraacetic acid (EDTA) were obtained from Sigma-Aldrich.

WT CR (DBS 100 strain), EPEC (E2348/69 strain), EHEC (EDL933 strain), E. coli K-12 (MG1655 strain), bioluminescent CR (ICC180 strain), and genetically manipulated bacterial strains used are summarized in Supplementary Table S2. Chromosomal deletion or supplement of indicated genes was conducted with a “scar”-free in-frame deletion method based on the suicide vector pRE112 via sacB-based allelic exchange (44). CR, EPEC, EHEC, and E. coli K-12 strains were grown from single colonies on LB plates in LB broth at 37°C overnight with shaking. For growth curve measurement, the overnight bacterial culture was diluted as indicated to grow in LB or DMEM at 37°C with shaking, and the culture was taken at indicated time periods to measure absorbance at optical density 600 (OD₆₀₀) on a POLARStar Omega Plate Reader (BMG Labtech).

The cDNAs encoding the indicated bacterial proteins were cloned by PCR from genomic DNA of CR, EPEC, EHEC, or E. coli K-12 into a pGEX-6P-1 plasmid (GE Healthcare), using the primers with detailed sequences in Supplementary Table S3. The glutathione S-transferase fusion proteins in E. coli (BL21) were induced and purified as previously described (45).

The DNA digestion assays were amended from previously described assays (46). In brief, the reaction contained a standard assay buffer (10 mmol/L Tris-HCl, 2.5 mmol/L MgCl₂, 0.5 mmol/L CaCl₂, pH 7.6), indicated concentrations of recombinant proteins, and variant DNA substrates including λ DNA (New England Biolabs), plasmids pRE112UA and pGEX-6P-1, single-stranded DNAs (with detailed sequences in Supplementary Table S3), and isolated chromatin DNA. The DNA digestion reactions were incubated for an indicated time at 37°C and quenched by heating at 80°C for 10 minutes. When chromatin DNA was used as a substrate, the digestion reactions were subjected to protein digestion by Proteinase K at 37°C for 30 minutes followed by heat inactivation at 65°C for 10 minutes. The DNA was separated by 2% agarose gel electrophoresis, stained with ethidium bromide, and photographed using a FluorChem E System (Protein Simple). The densities of DNA bands were quantified using NIH ImageJ software version 10.2.

CR infection in mice was conducted as described previously (47). In brief, SPF C57BL/6J mice (6–8 weeks) or Apc^MinΔ716/+ mice (4 weeks) were fasted for 8 hours before being orally inoculated with 200 μL PBS containing 2 × 10⁹ colony-forming units (CFU) of indicated CR strains or PBS vehicle control. GF C57BL/6J mice (6–8 weeks) were directly infected by oral inoculation of 2 × 10⁹ CFUs of indicated CR strains or vehicle control. For fecal CR burden analysis, stool was collected from live animals at various time periods postinoculation. The stool and tissue were homogenized, diluted in sterile PBS, and plated on MacConkey agar plates (VWR), followed by overnight incubation at 37°C. CFUs were enumerated the following day and normalized to the stool weight. To validate CR clearance from infected mice, total DNA was extracted from mouse stools using the Quick-DNA Fecal/Soil Microbe Kit (Zymo Research) and subjected to PCR using primers specific for CR espA and espF genes with sequences detailed in Supplementary Table S3. WT and ΔushA CR mixed infection in Apc^MinΔ716/+ mice and fecal burden measurements were conducted as described above. To validate the identities of WT and ΔushA CR strains, random colonies on MacConkey agar plates were picked for colony-PCR using the ushA-specific primers with sequences detailed in Supplementary Table S3.

Infection by CR, EPEC, and E. coli K-12 in SW480 cells was performed as previously described (47). Briefly, the indicated bacterial strain was washed with ice-cold PBS and resuspended in prewarmed DMEM containing 10% heat-inactivated FBS. Bacterial concentration was measured by absorbance at OD₆₀₀, followed by a serial dilution and seeding on a MacConkey agar plate to confirm the administered CFU. SW480 cells were infected by the indicated bacterial strains at a multiplicity of infection of 100 for indicated time periods, followed by whole-cell lysis for immunoblotting or fixation for immunofluorescence staining, as previously described (47). To measure CR and EPEC attachment during infections, SW480 cells were washed twice with DMEM, harvested, and plated directly on MacConkey agar plates for overnight incubation at 37°C. The CFU of indicated bacterial strains were enumerated the following day.

Histology, IHC, and immunofluorescence staining of colon tissue sections were performed as previously described (30). In brief, after euthanizing mice, the colons were removed under aseptic conditions and washed once with ice-cold PBS, the colon was fixed in 10% buffered formalin for 24 hours and embedded in paraffin, and 5-μm sections were cut and processed for hematoxylin and eosin staining. Histopathology scores were determined in a blinded fashion using the following criteria as previously described (48). For immunostaining, the colon tissue sections were subjected to antigen retrieval in citrate buffer, pH 6.0 (Life Technologies), in a Pressurized 2100 Retriever (Electron Microscopy Science). After blocking with appropriate sera and incubating with appropriate antibodies, sections were washed and incubated with either horseradish peroxidase–conjugated second antibodies, followed by DAB staining (Vector Laboratories), or fluorescence dye–conjugated second antibodies and 1 μg/mL DAPI (Sigma-Aldrich). Immunofluorescence-stained sections were washed and mounted under a coverslip using Fluoro-gel with Tris Buffer (Electron Microscopy Sciences) and examined on a Leica DMi8 fluorescence microscope (Leica Microsystems).

For cellular immunofluorescence staining, cells were fixed with 4% paraformaldehyde in PBS and then mounted onto slides. After permeabilization with 0.05% Triton X-100 in PBS and blocking with appropriate sera, the fixed cells were incubated with appropriate antibodies, counterstained with DAPI, and examined on a Leica SP8 confocal fluorescence microscope (Leica Microsystems). The images were obtained with matched exposure settings and in a blinded fashion.

Cells or colon tissues were harvested and lysed on ice by 0.4 mL lysis buffer containing 50 mmol/L Tris-HCl (pH 8.0), 150 mmol/L NaCl, 1% NP-40 and 0.5% sodium deoxycholate, and 1× complete protease inhibitor cocktail (Roche Applied Science) for 30 minutes. The lysates were centrifuged at 10,000 × g at 4°C for 10 minutes and separated by SDS-PAGE under reduced and denaturing conditions. The resolved protein bands were transferred onto nitrocellulose membranes and probed by the Super Signaling system (Thermo Scientific) according to the manufacturer's instructions and imaged using a FluorChem E System (Protein Simple), as previously described (30). Immunoblot of bacterial secreted proteins was performed as previously described (49). In brief, bacteria were pelleted by centrifugation from 20 mL CR culture in DMEM at 37°C (OD₆₀₀ 0.5–0.7), and supernatants were collected and filtered through a 0.22-μm syringe filter (Celltreat). Bacterial secreted proteins in supernatants were precipitated by trichloroacetic acid (Sigma-Aldrich) at a final concentration of 10% (volume for volume; v/v) for 1 hour on ice, followed by centrifugation at 16,000 × g at 4°C for 10 minutes. The resolved proteins were subjected to SDS-PAGE separation and immunoblot.

Type III secretion and translocation assays for CR and EPEC T3SS effectors were conducted using TEM1 β-lactamase as a fluorescence-based reporter, as previously described (20). In brief, the PCR products of indicated effectors were amplified using the primers with detailed sequences in Supplementary Table S3 and cloned via NdeI/EcoRI sites into pCX340 plasmid that encodes the mature form of TEM1 under the control of the isopropyl-β-D-thiogalactopyranoside–inducible Ptrc promoter, thus generating T3SS effector–TEM1 fusion proteins. HeLa cells were washed with Hank's Balanced Salt Solution (HBSS), followed by infection with the DMEM culture of WT and ΔescN EPEC expressing either TEM1 control or indicated TEM1 fusion proteins. After infection, HeLa cells were washed twice with HBSS, covered with 1 × CCF2/AM solution in HBSS freshly prepared with the CCF2/AM loading kit (Thermo Scientific), and incubated at room temperature for 90 minutes. Fluorescence was quantified on a Synergy H1 microplate reader (BioTek Instruments) with excitation at 410 nm, and emission was detected via 450-nm (blue fluorescence) and 520-nm (green fluorescence) filters; translocation was expressed as the emission ratio at 450 nm/520 nm to normalize the TEM1 β-lactamase activity to cell loading and the number of present cells.

Subcellular fractionation was performed by differential centrifugation as previously described (50). Briefly, cells were resuspended in ice-cold buffer A [10 mmol/L HEPES pH 7.9, 10 mmol/L KCl, 1.5 mmol/L MgCl₂, 0.1 mmol/L EDTA, 0.5 mmol/L DTT, 0.4% NP-40, 0.5 mmol/L phenylmethylsulfonylfluoride (PMSF), complete protease inhibitor cocktail] at 4°C for 5 minutes. Lysates were centrifuged at 4°C, 500 × g for 3 minutes, and supernatants were collected as cytosolic fractions. Pellets were incubated in Buffer C (20 mmol/L HEPES pH 7.9, 420 mmol/L NaCl, 1.5 mmol/L MgCl₂, 25% glycerol, 0.5 mmol/L PMSF, 0.2 mmol/L EDTA, 0.5 mmol/L DTT, complete protease inhibitor cocktail) at 4°C for 10 minutes. Supernatants were collected as nuclear fractions following a centrifuge at 4°C, 13,800 × g for 10 minutes. Chromatin fractionation was carried out with modifications from a previous protocol (51). In brief, cells were washed with PBS, resuspended in ice-cold Buffer A1 [10 mmol/L HEPES pH 7.9, 10 mmol/L KCl, 1.5 mmol/L MgCl₂, 0.34 mol/L sucrose, 10% (vol/vol) glycerol, 1 mmol/L DTT, 10 mmol/L NaF, 1 mmol/L Na₂VO₃, complete protease inhibitor cocktail], and lysed on ice for 5 minutes. Lysates were centrifuged at 4°C, 1,300 × g for 5 minutes, and supernatants were collected as cytoplasmic fractions. Pellets were washed once with Buffer A1 and lysed in Buffer B (3 mmol/L EDTA, 0.2 mmol/L EGTA, 1 mmol/L DTT) on ice for 10 minutes. Following a centrifuge at 4°C, 10,000 × g for 1 minute, pellets were collected as chromatin, resuspended in 50 mmol/L Tris-HCl (pH 8.0), and subjected to DNA digestion assays.

Comet assays were carried out using a Comet Assay Kit (Trevigen) as described previously (45). Briefly, SW480 cells were infected with indicated bacterial strains for indicated periods. After two washes in PBS, 3 × 10⁵ cells were collected and combined with 1% molten LMAgarose at 37°C at a ratio of 1:10 (v/v) and immediately pipetted onto slides. Slides were immersed in prechilled lysis solution for 1 hour on ice to lyse cells, followed by alkaline unwinding of chromatin. Alkaline electrophoresis of gelled slides was performed using an Econo-Sub Horizontal System (CBS Scientific) at 24 V (0.7 V/cm) at 4°C for 30 minutes. After visualization of the DNA by SYBR Green staining, images were taken under a Leica DMi8 fluorescence microscope and analyzed by CometScore software (TriTek).

After dilution of samples in sterile H₂O, measurements of bioluminescence were taken over a 10-second period with an integration time of 1 second, using a TD-20/20 luminometer (Turner Designs), and results were expressed as relative light units. The in vivo measurement of CR burden was conducted as previously described (52). In brief, at indicated time periods after CR infection, mice were euthanized, and the colon and cecum were removed, cut longitudinally, and opened flat. Tissues were placed into the light-tight chamber of an IVIS50 Pre-clinical In Vivo Imaging System (PerkinElmer). After switching off the light source, photons emitted from bioluminescent CR on the tissue were quantified. For the anatomic localization, pseudocolor images representing light intensity (blue, least intense to red, most intense) were generated using the LIVING IMAGE software (PerkinElmer).

Organoids were established from the colonic crypts derived from C57BL/6J mice as previously described (53). Briefly, after euthanizing mice, the colons were removed under aseptic conditions, washed with ice-cold PBS, and incubated with ice-cold PBS containing 1.5 mmol/L DTT and 30 mmol/L EDTA for 20 minutes. The colonic tissue was then incubated with warm PBS containing 15 mmol/L EDTA for 6 minutes, followed by extensive shaking to release the colonic crypts. After centrifugation, the resulting crypt pellet was washed with 30× volume of organoid basal medium (DMEM/F12 supplemented with 2 mmol/L GlutaMAX, 10 mmol/L HEPES, 100 U mL⁻¹ penicillin–streptomycin, and 1× N2 and B27 supplements). The purified crypts were filtered through a 100-μm cell strainer (Celltreat) and embedded in growth factor–reduced Matrigel (Corning, Inc.), and the resultant organoids were maintained in Organoid Growth Medium (StemCell Technologies).

The visualization and quantification of colon adenomas in mice were conducted as previously described (30). Briefly, mice were sacrificed at 3 months of age. Colon tissues were excised, cleaned with cold PBS, opened longitudinally, fixed in 10% neutral buffered formalin (3.7% formaldehyde, 1.2% methanol, 6.5 g/L sodium phosphate dibasic, 4.0 g/L sodium phosphate monobasic) at 25°C overnight, and stained with 0.2% (w/v) methylene blue solution. The adenomas were quantified and sized under a dissecting scope. Average tumor size and tumor load per individual mouse were determined by averaging diameters of all tumors present and summing the diameters of all tumors presented in a given mouse.

Total RNA was isolated from colon tissues using TRIzol reagent (Life Technologies) and treated with the TURBO DNA-free Kit (Life Technologies) to remove residual genomic DNA. cDNA was synthesized using the ProtoScript First Strand cDNA Synthesis Kit (New England Biolabs) according to the manufacturer's instructions. Gene-specific products were amplified using SsoAdvanced SYBR Green Supermix (Bio-Rad Laboratories) using the primers with detailed sequences in Supplementary Table S3.

Isolation of lamina propria lymphocytes (LPL) was conducted as described previously (54). In brief, after euthanizing mice, the colons were removed under aseptic conditions, opened longitudinally, and washed in cold PBS to remove feces. The harvested colons were cut into 0.5-cm segments and incubated in RPMI 1640 medium containing 3% FBS, 5 mmol/L EDTA, 1 mmol/L DTT, and 20 mmol/L HEPES at 37°C with shaking for 25 minutes. The colons were further digested in RPMI 1640 medium containing 5 μg/mL DNase I (Roche), 100 μg/mL Liberase TL (Roche), and 20 mmol/L HEPES at 37°C for 50 minutes, followed by filtering through a 70- and 40-mm cell strainer and centrifugation on a Percoll gradient to isolate colonic LPLs.

Flow cytometry was performed on a FACSCanto II flow cytometry system (BD Biosciences), as previously described (55). Briefly, isolated LPLs were stimulated with 50 ng/mL PMA (MilliporeSigma), 1 μg/mL Ionomycin (MilliporeSigma), and 5 μg/mL Brefeldin A solution (BioLegend) at 37°C for 4 hours. Dead cells were excluded with the fixable live/dead staining (Life Technologies). Anti-CD16/32 antibody was used to block nonspecific antibody binding. For surface staining, cells were stained at 4°C for 30 minutes with anti-CD45 (Clone 30-F11, AmCyan conjugated; BioLegend), anti-CD4 (clone RM4–5, PE conjugated; BioLegend), and anti-CD3 (17A2, PE-Cy7 conjugated; BioLegend). For intracellular staining, cells were permeabilized with Fix/Perm buffer (eBioscience) for 30 minutes, washed, and stained with anti-Foxp3 (FJK-16s, PerCP-Cy5–5 conjugated; eBioscience), anti-IL17 (TC11–18H10.1, APC-Cy7 conjugated; BioLegend), and anti-IFNγ (XMG1.2, APC conjugated; BioLegend) at 4°C for 30 minutes. Data were analyzed using the FlowJo software version 10 (BD Life Sciences).

For WES, colon tumor tissue and matched liver tissue samples were harvested from tumor-laden mice, which were originally infected with vehicle control or indicated CR strains. Genomic DNA was isolated using the PureLink Genomic DNA Mini Kit (Thermo Fisher Scientific) according to the manufacturer's instructions, and Illumina DNA libraries with mouse exome capture were prepared using the SureSelect Mouse All Exon Kit (Agilent). All samples were sequenced using an Illumina NovaSeq6000 S4 System with ∼150× base coverage. Following quality control and trimming by fastp (56), the valid reads were mapped against the mouse reference genome version mm10 using the Burrows–Wheeler Aligner (57) version v0.7.5 with settings “bwa mem -M” to produce a single mapping file for each sample. Somatic mutations calling in all tumor samples were carried out using the Genome Analysis Toolkit (GATK) Mutect2 (58) version v.4.17 with tumor-only mode. Prior to that, the matched liver tissue samples were used to generate a panel of normal file containing background mutations with default parameters. High-confidence somatic mutations in tumor samples were then extracted from the generated raw somatic mutation data sets using GATK FilterMutectCalls, and SBS mutations in each sample were extracted using GATK SelectVariants individually. Then the software tool ANNOVAR (2019Oct24) was employed to perform gene-, region-, and filter-based annotations on high-confidence SBS mutations. After removing mutations located in repetitive regions of mouse genome, the mutations located within the exons of mouse genome were subjected to further analyses. For mutational signature analysis, 96 trinucleotide SBS subcategory counts of each sample were calculated using the Python package SigProfiler (34). The 96-trinucleotide SBS mutation percentages in each infection group were grouped into six classes (C>A, C>G, C>T, T>A, T>C, T>G) compared by R package MutationalPatterns (59) and plotted with the ggplots R package (https://ggplot2.tidyverse.org/). The profiles of previously defined SBS mutational signatures, excluding possible sequencing artifacts, were retrieved from the COSMIC mutational signatures database version v3.2 (https://cancer.sanger.ac.uk/cosmic/signatures) for comparative purposes.

Statistical analysis was performed using GraphPad Prism version 6.0 (GraphPad Software). SEMs were plotted in graphs. Significant differences were considered: ns (nonsignificant difference); *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 by multiple comparison tests (one-way ANOVA or Kruskal–Wallis depending on data distribution) followed by post hoc tests (Bonferroni or Dunn multiple comparisons tests), as detailed in the figure legends.

WES data have been deposited in the Sequence Read Archive at the NCBI under accession number PRJNA681077.

D. Xu reports grants from American Association of Immunologists (Career in Immunology Fellowship to D. Xu) during the conduct of the study. X. Xia reports grants from National Institutes of Health during the conduct of the study. C.L. Sears reports grants from Bloomberg Philanthropies during the conduct of the study; grants from Bristol Myers Squibb and grants from Janssen outside the submitted work. F. Wan reports grants from National Institutes of Health, grants from American Cancer Society, grants from Department of Defense, grants from Willowcroft Foundation, and grants from Graham Memorial Trust outside the submitted work. No disclosures were reported by the other authors.

Y. Liu: Data curation, software, formal analysis, investigation, writing–review and editing. K. Fu: Formal analysis, investigation. E.M. Wier: Formal analysis, investigation. Y. Lei: Formal analysis, investigation. A. Hodgson: Formal analysis, investigation. D. Xu: Formal analysis, investigation. X. Xia: Formal analysis, investigation. D. Zheng: Data curation, software, formal analysis. H. Ding: Resources, formal analysis, investigation. C.L. Sears: Resources, investigation. J. Yang: Data curation, software, formal analysis. F. Wan: Conceptualization, formal analysis, supervision, funding acquisition, writing–original draft, writing–review and editing.

We thank Drs. Gad Frankel, Anne-Marie Hansen, Alison O'Brien, Hongbing Yu, and Huimin Yu for kindly sharing reagents; Drs. Diane Griffin and Jiou Wang for use of equipment; and the Wan lab members for constructive and productive discussions. This work was supported in part by grants from the NIH (R01GM111682, R01CA244350, and R21AI137719 to F. Wan; T32CA009110 to E.M. Wier and X. Xia), American Heart Association (19PRE34380234 to Y. Liu), American Association of Immunologists (Career in Immunology Fellowship to D. Xu), Germ-Free Facility of Johns Hopkins University (to H. Ding), Bloomberg Philanthropies (to C.L. Sears), American Cancer Society (RSG-13–052–01-MPC, to F. Wan), Department of Defense (W81XWH-19–1-0479, to F. Wan), Willowcroft Foundation (1902, to F. Wan), and Graham Memorial Trust (Samuel J. Graham Award, to F. Wan).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.