Parallel evolutionary trajectories rewire enteropathogenic Escherichia coli adhesion to restore host attachment Authors: Noam Yedidi, Klil Cohen-Yaron, Miriam Ravins, Michal Bejerano-Sagie, Sigal Ben-Yehuda, Ilan Rosenshine Published: 2026 ABSTRACT AbstractEnteropathogenic Escherichia coli (EPEC) causes disease in children, presenting as chronic diarrhea that can impair physical and cognitive development. The attachment of typical EPEC (tEPEC) to the gut epithelium via bundle-forming pili (BFP) is a key factor in its virulence. Yet, infections by atypical EPEC (aEPEC), which lack BFP, have become increasingly common. To investigate how aEPEC recover host-attachment in the absence of BFP, we performed experimental evolution using a non-adherent E. coli, constructed to mimic the ancestor of aEPEC, and selected adherent progeny. Highly adherent variants evolved through phase-variable activation of type I fimbriae (T1F), followed by two alternative trajectories: bacterial filamentation, which increases T1F avidity, or point mutations in the T1F adhesin FimH that enhance ligand affinity. Extending our analysis to the genomes of 327 aEPEC strains isolated from infected patients revealed that similar FimH mutations are common. We further demonstrated experimentally that these naturally occurring variants often increase epithelial-attachment. Our findings implicate T1F in aEPEC pathogenesis and suggest it may be clinically relevant for anti-adhesion therapy. More broadly, these results indicate that impaired host-attachment can be rapidly compensated by upregulating and optimizing an alternative adhesin, and that combining experimental evolution with comparative genomics can reveal evolutionary trajectories occurring in nature.Keywords: Typical enteropathogenic E. coli, atypical enteropathogenic E. coli, EPEC, type I fimbriae, FimH, bacterial filamentation, host attachment Introduction Enteropathogenic Escherichia coli (EPEC) is a cause of infantile chronic diarrhea that can progress to life-threatening disease. 1-3 EPEC pathogenesis develops in two stages: (i) initial attachment to the gut epithelium, using specific adhesins, followed by (ii) intimate attachment through remodeling of the gut surface, achieved via Type III Secretion System (T3SS). 4 The T3SS is a molecular syringe, allowing the host-attached pathogen to inject dozens of proteins called effectors into the infected host cell. 5 These effectors manipulate host processes to benefit the pathogen. 6 A notable pathology caused by the injected effectors is called attaching and effacing (A/E) lesions, characterized by EPEC intimate attachment to the apical surface of the intestinal epithelium and the effacement of microvilli, 7 forming stable actin structures termed “pedestals”. 8 Prior to T3SS deployment, EPEC establishes initial contact with the host surface, typically via the Bundle Forming Pili (BFP). The bfp operon is encoded on a conjugative plasmid called pEAF (EPEC Adherence Factor plasmid). 9-11 BFP expression leads to aggregation of EPEC cells to form globular microcolonies, 12 , 13 and enhances initial host-attachment, thereby promoting infectivity. 14 EPEC strains possessing BFP are called typical EPEC (tEPEC). However, in recent years, atypical EPEC (aEPEC) strains, lacking pEAF and its bfp operon, have emerged in the clinic. 15-18 Infection by aEPEC usually results in less severe but more chronic symptoms, 19 causing disease also in adults. 17 Unlike tEPEC, which is a problem mainly in low- and middle-income countries, aEPEC is commonly found in patients from industrialized societies. 16 The global clinical burden of aEPEC is substantial, and may exceed that of tEPEC. 20-24 Several attachment factors shared across E. coli species may compensate for the absence of BFP and promote aEPEC attachment. 25-32 To test this hypothesis, we performed in vitro experimental evolution, selecting variants that acquired host-attachment through the emergence of alternative pathways. Our results highlight the phase-variable type I fimbriae (T1F; fim), a chaperone-usher system, capable of binding mannose moieties on the host-surface, as a strong BFP alternative. 33 , 34 However, T1F expression alone was insufficient to enhance attachment. This required additional adaptation: either bacterial cell elongation, which increased avidity, or point mutations in the T1F adhesin FimH that increased mannose-affinity. Importantly, we show that FimH mutants with enhanced host-attachment are common among aEPEC isolated from diarrheal patients. These results suggest that evolutionary trajectories similar to those observed in the laboratory setting have arisen repeatedly in nature, highlighting T1F-antagonism as a potential therapeutic strategy. Materials and methods Bacterial strain construction Bacteria were grown at 37°C in Luria-Bertani (LB) broth supplemented, when needed, with relevant antibiotics. Bacterial mutants were constructed using λ-Red system, 35 antibiotic-resistance cassettes were excised using pFT-A plasmid. 36 Scarless deletions were achieved with the tet-sacB cassette. 37 Bacterial strains are listed in Table S1, and their construction is described in Supplementary methods. Plasmids were created by Gibson assembly as described (Table S2). Oligonucleotides are listed in Table S3. Genomic-library was prepared as described. 38 Adherence-based enrichment experiment HeLa monolayers were infected, washed, and the host-cell-associated bacteria were regrown. After four cycles, bacteria were plated and individual colonies were isolated. See Supplementary methods for more details. Bacterial attachment measurements HeLa, Caco-2, or HT-29 cells lines were infected with indicated bacterial strains, then processed for microscopy, flow cytometry, or to determine host-associated colony-forming units (CFU). For fluorescence microscopy, bacteria, actin, tight-junctions, and nuclei, were visualized using the antibodies and stains listed in Supplementary methods and Table S4. For CFU assays, infected monolayers were washed to remove unattached bacteria, lysed, and the samples were serially diluted and plated. For flow cytometry, mScarlet-expressing bacteria associated with host cells were quantified. T1F-phase-status was determined by colony-PCR using fim-ON- and fim-OFF-specific primers, and the T1F structures were visualized by immunostaining. Measuring effector delivery using split-nanoluciferase T3SS-dependent effector delivery was measured using split-nanoluciferase. 39 , 40 HeLa cells stably expressing LgBit were infected with strains expressing EspF-HiBit, and luminescence was monitored over time as described in Supplementary methods. FimH variants analysis Published aEPEC genomes 41 were analyzed for fimH variation using the custom browser-based tools described in Supplementary methods. FimH variants were reconstructed and tested for host-attachment by microscopy and CFU. Further analysis is described in the figure captions, Supplementary methods, and extended data table. Results Atypical EPEC rapidly acquires host-attachment traits The emergence of aEPEC strains, lacking BFP, may result from the acquisition of alternative mechanisms for initial attachment. Because most E. coli strains encode multiple potential adhesins, we hypothesized that tEPEC strains that lose BFP, or commensal E. coli strains that acquire T3SS through horizontal gene transfer, can rapidly evolve into effective aEPEC by upregulating the expression of one of their pre-existing adhesins. 42 , 43 To experimentally test this hypothesis, we developed an in vitro evolution approach to select for variants that had evolved alternative mechanisms for initial attachment (Figure 1A). To mimic an ancestral strain with the potential to evolve into aEPEC, we used the well-characterized tEPEC strain E2348/69, 44 and eliminated its known attachment mechanisms, including BFP, T3SS, and flagella. 45 , 46 We verified that this parental strain exhibited only negligible attachment to HeLa cells (Figure S1). To accelerate evolution, we generated a plasmid-borne genomic-library of EPEC containing random 3–6 kb inserts (Figure S2), and introduced it into the parental strain, thereby increasing the copy number of random genomic fragments and enhancing the expression of potential attachment factors or their regulators. Finally, HeLa cells were infected with the parental strain containing the plasmid-library, washed to remove unattached bacteria, and the recovered attached bacteria were used for the next round of selection (Figure 1A). Remarkably, massive host-attachment was obtained after the fourth enrichment cycle, and 44 individual colonies derived from the attached bacteria were selected for further analysis. Enrichment of mutants that gain host-attachment. (A). Schematic of the enrichment procedure: the parental strain (RP7768) containing the genomic library was used for infection of HeLa cells. After 3 h HeLa cells were washed, and suspended in LB for overnight regrowth of the attached bacteria. This culture was used for the next infection cycle. At the fourth iteration, attached bacteria were recovered and spread on LB plates to obtain isolated colonies. (B). The emerging host-attached bacteria. Monolayers of Caco-2 cells were infected for 3.5 h with representative filamentous (F-1, NY10270) and coating (C-1, NY10273) clones or the parental strain. The infected cells were fixed and stained with DAPI and anti-ZO-1 antibody (magenta), and bacteria are marked by GFP expression. Scale Bar 10 µm. (C). Attachment to HT-29 and Caco-2 cells. Monolayers of HT-29 and Caco-2 cells were infected for 3.5 h with strains mentioned in (B) and wild-type EPEC (IR1). After infection, cells were washed to discard unattached bacteria. Serial dilutions of the recovered attached bacteria were spotted on LB plates. The dilution factor is indicated on the top of the panels. See Figure S3 and Supplementary methods for more details. To identify the candidate attachment factor(s), we sequenced the genomic inserts carried by the library-plasmids. Intriguingly, none of them contained a known or predicted attachment factor. Yet, this analysis showed that several colonies contained identical inserts, indicating that the 44 colonies likely originated from eight different clones (Table 1). We next examined host-attachment by the obtained clones, revealing two distinct patterns: four clones displayed elongated bacterial cells, sporadically attached to HeLa (filamentous phenotype), whereas four additional clones showed extensive diffused adherence to the host surface (coating phenotype) (Figures S4A and S9A). These two attachment patterns were also observed following infection of polarized Caco-2 cells, whereas the parental strain exhibited only negligible attachment (Figure 1B). We further quantified bacterial attachment to monolayers of Caco-2 and HT-29 cell lines by determination of host-associated CFUs, revealing that the filamentous clones exhibited an approximately 10-fold increase in attachment efficiency compared with the parental strain, whereas the coating clones showed more than a 100-fold increase (Figures 1C and S3). Notably, the attachment efficiency of the filamentous clones is likely underestimated because each bacterial chain, which may consist of multiple cells, is expected to give rise to a single CFU. Table 1.Analysis of the enrichment experiment.PhenotypeFilamentationCoatingCloneF-1F-2F-3F-4C-1C-2C-3C-4Colonies232111349Insert size (bp)34732299216424504336–32632181Genes in library insert rssA-rssB-galU (dauA)-prs-(ispE) (bcsC)-(pdeK) (aas)-lplT-(tas) (recB)-(recD) – (agaV)-agaW-(agaA) (astC)-xthA-(ydjX) Plasmid dependent+–Fim-ON9/98/99/98/97/97/98/98/9FimHWild-typeV56AY64DY64NG73R44 colonies were picked after the enrichment experiment. 27 displayed bacterial filamentation. 17 were coating HeLa cells. Sequencing of the inserts revealed that these colonies were expanded from 8 different clones. Insert sizes and the genes in these inserts are mentioned. Truncated genes are in (). Filamentation is plasmid-dependent, causative loci are in bold. All clones reside primarily in the T1F-ON-phase according to PCR. Coating clones acquired non-synonymous mutations in fimH, according to WGS. Analysis of the enrichment experiment. 44 colonies were picked after the enrichment experiment. 27 displayed bacterial filamentation. 17 were coating HeLa cells. Sequencing of the inserts revealed that these colonies were expanded from 8 different clones. Insert sizes and the genes in these inserts are mentioned. Truncated genes are in (). Filamentation is plasmid-dependent, causative loci are in bold. All clones reside primarily in the T1F-ON-phase according to PCR. Coating clones acquired non-synonymous mutations in fimH, according to WGS. Attachment of filamentous bacteria requires T1F expression To investigate the filamentous adhesion pattern, we isolated the plasmids carried by clones displaying this phenotype. Reintroduction of these plasmids into the plasmid-free parental strain restored both the filamentous morphology and host-attachment (Figures 2A and S4B). To identify the genes responsible for filamentation, the four plasmids were subjected to deletion analysis, mutagenesis, and expression validation by RT-qPCR. These experiments revealed that overexpression of either galU, prs, the HAMP-GGDEF domains of pdeK, or the 5′ ~200 bp of lplT ORF, were each sufficient to induce filamentation (Figure S5). Filamentation of bacteria expressing T1F promotes host-attachment. (A). HeLa cells were infected with clone F-3 (NY10280) or parental strain transformed with the plasmid extracted from this clone (parental/pF-3). At 3 h post infection cells were fixed and stained with phalloidin rhodamine (red, actin) and anti-EPEC (green). Scale bar 10 µm. (B). Visualization of T1F by immunostaining. Fluorescence microscopy was performed using anti-T1F antiserum (α-T1F). The F-2 clone (NY10274) is compared to the fim-ON-locked (NY11899) and fim-OFF-locked (NY11900) mutants. Scale bar 2 µm. (C). T1F is required for attachment of the filamentous bacteria. EPEC mutant lacking BFP, flagella, T3SS and T1F genes (NY10872, parental ∆fim) was transformed with a plasmid for inducible expression of T1F operon (pNY11212, pFIM), and supplemented also with filamentation-inducing plasmid from clone F-3 (pF-3, NY11276), or the corresponding empty vector (EV, NY11273), together with mScarlet-expressing plasmid. HeLa cells were infected with cultures as indicated, and expression of T1F was induced (100 ng/ml aTc), or not (No aTc). At 3 h post infection cells were fixed and visualized. Scale bar 10 µm. See Supplementary methods for more details. We noticed, however, that occasionally, bacteria derived from isolated colonies of the filamentous clones failed to attach despite their filamentous appearance. This apparent bimodality is characteristic of phase-variable traits, such as that of type I fimbriae (T1F), driven by inversion of the DNA segment containing the fim promoter. 33 To test the role of T1F in host-attachment by filamentous bacteria, we used PCR to examine the ON/OFF-phase in isolated colonies of the parental strain and of the filamentous clones. While the parental strain resided mostly in OFF-state, in all the filamentous clones the ON-state was dominant (Figure S6A). Consistently, immunostaining with anti-T1F serum indicated the presence of fimbriae spanning the entire length of the bacterial filaments (Figures 2B and S6B). To further examine whether T1F is required for filament attachment, we deleted the fim locus, fimBEAICDFGH, from the parental strain genome (parental-Δfim), and complemented this deletion with a plasmid expressing the T1F structural genes, fimAICDFGH, under an inducible promoter (pFIM). 47 Host-attachment was observed only upon inducing T1F expression. When the four filamentation-inducing plasmids were introduced into this parental-Δfim/pFIM strain, the attachment levels were further boosted (Figures 2C and S7). To quantify bacterial attachment, we used flow cytometry, measuring the load of fluorescently-tagged bacteria per HeLa cell. We found that all four filamentous strains showed increased attachment capacity upon fim induction (Figure S8B). These results suggest that the increased T1F avidity conferred by filamentation enhances bacterial attachment. To test this hypothesis, we induced filamentation by several independent interventions, including disruption of FtsZ ring positioning in a ΔminCD mutant, overexpression of the small RNA OxyS, which indirectly inhibits FtsZ activity, and treatment with sublethal concentrations of ampicillin, which inhibit septum formation (Figure S8A). 48-51 We found that all three perturbations induced filamentation accompanied by enhanced host-attachment, with ampicillin exhibiting a dose-dependent effect (Figures S7 and S8C). These findings suggest that enhanced attachment results from the combination of two events: (i) acquisition of a filamentous morphology, and (ii) switching T1F to the ON-phase, generating bacterial particles with high T1F avidity. We did not further investigate how overexpression of galU, prs, pdeK, or lplT induces filamentation because our data indicate that the specific mechanism is not critical for the enhanced attachment phenotype. The coating phenotype is caused by affinity-gains through mutations in FimH We next investigated the clones displaying the coating phenotype. Analysis of their plasmids showed that they either lacking an insert (Table 1), or that the inserts are irrelevant to attachment (Figures 3A and S9), suggesting that spontaneous mutations might be responsible for the phenotype. We thus subjected the genomes of the four independent clones to whole-genome sequencing (WGS). Strikingly, all four genomes were identical to each other and to the parental strain, except for single-nucleotide changes which converged into the fimH gene of T1F (Figure 3B). PCR analysis, as well as immunostaining, showed that all four clones were predominantly residing in the T1F-ON-phase (Figure S10). The coating phenotype is caused by expression of mutated T1F. (A). HeLa cells were infected with clone C-3 (NY10269), or parental strain transformed with the plasmid extracted from this clone (parental/pC-3). At 3 h post infection cells were fixed and stained with phalloidin rhodamine (red) and anti-EPEC (green). Scale bar 10 µm. (B). FimH mutations found in the four coating clones. FimH protein contains a signal peptide (black), lectin-domain (orange) and pilin-domain (green). The mutations found in the coating clones (i.e., NY10273, C-1; NY10272, C-2; NY10269, C-3; NY10271, C-4) and their locations in the structure model (PDB: 4XOB, visualized using ChimeraX) are indicated. (C). Mutated FimH is required for coating attachment. HeLa cells were infected with EPEC mutant lacking BFP, flagella, T3SS and T1F genes (Parental ∆fim, NY10872), which was transformed with a plasmid expressing the T1F operon with mutated fimH (pNY11666, pFIMV56A) together with mScarlet-expressing plasmid. T1F expression was induced, or not, with aTc and at 3 h post infection cells were fixed and visualized. Scale bar 10 µm. (D). Quantification of the coating phenotype. Parental ∆fim strain described in C was supplemented with plasmids expressing wild-type fim operon (NY11273, pFIMwt), or fim with the four fimH mutations as indicated (pNY11666, pFIMV56A; NY11665, pFIMY64D; NY11680, pFIMY64N and NY11681, pFIMG73R) as well as a combined phenotype coating & filamentation (NY11667, pFIMY64N + pF-2). Cultures were used for infection of HeLa cells followed by flow cytometry analysis as described in Supplementary methods. A representative set is shown out of 3 biological repeats (extended data table). The FimH subunit is located at the tip of T1F, and serves as the adhesin, recognizing exposed mannose displayed on the host-surface. It consists of pilin-domain that connects it to the fimbriae and lectin-domain that contains the mannose-binding site. Association between these two domains results in reduced mannose-affinity, while their dissociation shifts FimH into higher-affinity state. 52 , 53 The mutations we identified: V56A, Y64D, Y64N, and G73R (PDB: 4XOB), cluster into the alpha-switch, a small region within the lectin-domain (Figure 3B), involved in switching to the high-affinity configuration (PDB: 4XOC). 54 , 55 These results suggest that the coating phenotype arose from two events (i) switching the T1F to ON-phase and (ii) acquiring a point mutation in fimH. To verify the role of the mutations, we transformed the parental-Δfim strain with pFIM plasmids containing each of the identified fimH mutations. Indeed, expression of these four fim-mutated alleles induced the coating phenotype (Figures 3C and S11). Using flow cytometry to quantify host-attachment further demonstrated the increased adherence associated with the mutated FimH. Finally, bacterial filamentation combined with expression of high-affinity fimH allele resulted in additional increase in attachment (Figure 3D). We concluded that upon selection for improved host-attachment, two phenotypes evolved, both required switching the T1F to the ON-phase and either increasing avidity by bacterial filamentation or increasing affinity to the host-surface by gaining point mutations in fimH (Table 1). The coating and filamentation phenotypes enhance T3SS effector delivery We predicted that enhanced attachment mediated by T1F would enhance T3SS-mediated effector injection. To test this premise, we used an EPEC E2348/69 derivative cured of the BFP-expressing plasmid, 56 and genetically locked in the fim-ON-state. This strain was transformed with one of the filamentation plasmids or mutated to contain the fimH Y64D variant. For both genotypes, we noted enhanced attachment to Caco-2 or HeLa cells, associated with the formation of actin-pedestals, a readout for T3SS activity (Figures 4A and S12A). To quantify the T3SS activity, we employed a split-nanoluciferase system. 40 HeLa cells, stably expressing a large portion of nanoluciferase (LgBit), were infected with pathogens expressing the native EspF T3SS effector, 57-59 translationally fused with the complementing portion of nanoluciferase (HiBit). 39 Upon its injection, the EspF-HiBit complements the host-encoded LgBit, allowing monitoring of EspF injection over time (Figure 4B). Strong reduction in EspF injection was detected for EPEC lacking BFP compared to its parental tEPEC, attributed to the contribution of BFP to tEPEC efficient attachment. While injection by aEPEC was slightly improved by filamentation, the FimHY64D-mutant surpassed tEPEC injection capacity, further demonstrating T1F potential as a compensatory adhesin. Consistently, the effect of the Y64D mutation was completely abolished by D-mannose, a soluble FimH ligand (Figure 4C). Since all these strains display similar level of T3SS expression (Figure S12B), we concluded that enhanced attachment via T1F correlates with enhanced effector injection. Filamentation and coating improve the T3SS activity. (A). Attachment to Caco-2 cells and pedestal formation. Caco-2 cells were infected with aEPEC strain containing fim-ON-locked mutation (JPN15 T1F ON, NY12809), or an isogenic strain containing a filamentation-inducing plasmid expressing the prs gene (JPN15 T1F ON/p-prs, NY12811), or containing fimH Y64D mutation (JPN15 T1F ON/fimHY64D, NY12813). Bacterial GFP expression was induced (0.2 mM IPTG). At 3.5 h post infection cells were washed, fixed, and stained with phalloidin rhodamine (actin, orange) and anti-ZO-1 (magenta). Scale bar 10 µm. (B). Kinetic measurement of EspF injection. HeLa expressing LargeBit were infected with bacterial strains expressing EspF-HiBit and T1F: wild-type tEPEC (NY12401, black), JPN15 (NY12517, yellow), JPN15 + pF-2 (NY12560, blue) and JPN15 fimH Y64D (NY12554, magenta). Substrate was added with the infecting bacteria. Luminescence was measured at 30 min intervals and data was normalized as described in Supplementary methods. Error bars represent standard deviation of 3 technical repeats. One representative set is shown out of 3 biological repeats (extended data table). (C). Kinetic measurement of EspF injection as in B, with infection media containing 1% D-mannose. Atypical EPEC strains often contain FimH mutations We next sought to determine whether similar FimH mutations, increasing host-attachment, evolved also in nature. To this end, we analyzed deposited WGS data of 327 aEPEC clinical isolates, 41 and examined their fimH sequences. We found that 160 isolates in this database have a mutated fimH gene, with a total of 35 different mutations, spread throughout the FimH sequence (Figure 5A). Notably, most of the clinical FimH mutations fall in areas that influence the shift to the high-affinity configuration, including mutations in the lectin-pilin domains interface and in the alpha-switch (Figure 5B), one of which (Y64D) also emerged in our in vitro evolution experiment (Figure 3B). Additional mutations occurred in the mannose binding-site, and in other areas in the lectin-domain. Other mutations were mapped to the FimH signal-peptide and pilin-domain. The impact of these mutations on attachment remain unclear. 60 , 61 FimH mutations in clinical isolates of aEPEC can promote host-attachment. (A). Linear schematic of FimH and list of the mutations, color-coded to the respective domain in the linear presentation (A) and the structure model (B). See extended data table for per-isolate and per-variant information. (B). The FimH backbone (PDB: 4XOB, ChimeraX), with indication of the locations of the mutated residues from (A), designated by the known functional loops: mannose binding site (magenta), alpha-switch (red), interface between the domains (blue), pilin-domain (green), and lectin-domain (yellow). (C). Upper panel: plausible trajectory of the FimH mutations accumulation. The numbers of clinical isolates with each FimH variant are mentioned in gray circles. Beneath the tree, highlighted in pale pink, are the branch-corresponding mutations, which were reconstructed as singles. Colors correspond to the attachment score: FimH variants with no significant effect on attachment to HeLa (<2 fold compared to the consensus) are in gray. Variants with >2, >4, >10, >100-fold improvement in attachment appear in yellow, orange, pink, and red, respectively. Lower panel: visual presentation of scoring method. Numbers of clinical isolates with each attachment score appear in gray circles. See Supplementary methods for analysis and experimental details, and extended data table for adherence score analysis. Interestingly, the fimH genes of 108 isolates contained 2-4 amino-acid changes. In most cases, it is possible to infer the trajectories of FimH evolution based on intermediates. For example, the quadruple-mutant N70S, S78N, A27V, G160D, found in a single isolate, can be traced to ancestor variants containing triple, double or single corresponding mutations, which appeared in 2, 6, and 16 isolates, respectively, supporting the proposed trajectory (Figure 5C). FimH variants that evolved in aEPEC can enhance initial attachment We suspected that host-attachment might be a selective pressure driving FimH evolution in the clinical isolates. To evaluate the attachment activity of the FimH mutants we generated a panel of isogenic strains, each expressing different fimH allele (Table S2 “FimH-mutants collection”). These strains were also locked in the T1F ON-phase and GFP-labeled. Attachment of these strains to HeLa cells was evaluated by measuring cell-associated GFP (Figure 5C). Half of the variants, found in 95 clinical isolates, showed increased attachment compared to the consensus FimH, while the rest of the variants, found in 65 clinical isolates, had no significant impact on host-attachment (Figure 5C). Some variants strengthened the attachment by 10–100-fold, of which the most prominent were Y64D and S62P. Similar increased attachment was evident by determination of CFU associated with HT-29 monolayers (Figure S13A). Similar growth rates (Figure S13B) and fimbriae levels (Figure S13C) reinforced the notion that the FimH mutations are the cause for improved attachment. Importantly, many of the fimH alleles that acquired multiple mutations displayed improved host-attachment compared to each of the corresponding single mutations (Figure 5C, light boxes). For example, the V30M mutation improved attachment by more than 10-fold when acquired on top of T-6N A119V G273A, but had no significant impact as a single mutation. This trend may point to a consistent pressure to improve FimH affinity, either in a single impactful substitution like Y64D, or through a step-by-step trajectory. These results further show that evolutionary trajectories such as those observed in our laboratory experiment have arisen repeatedly in nature, suggesting a possible adaptation to host infection via fimH mutations. Discussion Our study was designed to understand how a pathogen can reassign dominance from one adhesin to another. We investigated this question by utilizing a model mimicking the emergence of aEPEC strains. Using an engineered strain lacking the known attachment machineries, we selected variants that evolved to regain host-adherence. In this selection process, phase-variation served as a first step, generating a T1F-expressing subpopulation, and unlocking access to improved adherence without committing the entire population. However, fim-ON alone was insufficient. Enhanced attachment consistently required a second change in T1F avidity or affinity. On the avidity side, perturbations that cause bacterial filamentation lead to enhanced T1F-dependent attachment. These perturbations include overexpression of galU, prs, the HAMP-GGDEF domains of pdeK, the first ~200 bp of lplT and exposure to sublethal concentrations of ampicillin. 62 , 63 Notably, cell elongation is a common bacterial response to diverse stress conditions. 64-72 Moreover, even without stress, stochastic cell-to-cell variations in gene expression can generate subpopulations of transiently elongated bacteria. 73 , 74 Such cells may then serve as substrates for selection, further enhancing attachment through the acquisition of adaptive mutations. On the affinity side, we identified single amino-acid substitutions in the FimH lectin-domain (V56A, Y64D, Y64N, G73R) that increase host-attachment. These residues fall in or near a structural loop that influences the reversibility (KOFF) of mannose engagement, in line with prior biochemical evidence. 54 Strikingly, comparative analysis of 327 aEPEC clinical genomes revealed the same logic operating in nature, with half the isolates encoding FimH substitutions, including Y64D. We showed that many of these mutations impact attachment, with reports linking them to host-adherence in other E. coli pathotypes. 75-80 Moreover, many aEPEC fimH alleles contain multiple substitutions, with intermediate genotypes present across different isolates, allowing us to infer likely mutational trajectories. Often, single substitutions conferred modest gains, while combinations produced stronger effects. The key role of T1F in the virulence of uropathogenic E. coli (UPEC) is well documented. 81 This is consistent with the presentation of mannosylated glycans on the apical surface of urothelial cells. 82 , 83 In intestinal E. coli pathotypes, however, the contribution of T1F is more context dependent. For tEPEC, T1F was reported not to affect adherence, consistent with the dominant role of BFP. 84 This does not imply that FimH targets are absent from the intestine. Rather, such targets might be heterogeneous and differentially exposed across intestinal niches. 85 Alternatively, expression of wild-type fim or the observed mutants may impose fitness costs, which could result in environmental trade-offs. For instance, binding of FimH-expressing EPEC to mannosylated glycoprotein-2 (GP2) at the M-cell-luminal interface, or FimH engagement with TLR4, are primarily associated with immune sampling and inflammatory signaling rather than stable colonization. 86-90 Under inflammatory conditions, mannosylated CEACAM6 and basement-membrane ligands may support FimH-dependent adherence to the intestinal epithelium. 91-93 Previous studies have shown that aEPEC isolates carry and express functional T1F. 94-96 Our data indicate that increased affinity of FimH mutants can compensate for the heterogenous availability of accessible mannosylated ligands, thereby permitting bacterial attachment. Importantly, similar FimH variants have been identified in Crohn's disease-associated adherent-invasive E. coli (AIEC), supporting the tuning of FimH affinity as a general strategy to promote host-attachment under limited receptor availability. 76 Increased T1F-mediated attachment translated into more efficient T3SS activity. Both filamentation and FimH mutations increased effector delivery, and the latter exceeded the injection capacity of tEPEC. We interpret this as a direct consequence of improved attachment. Consistently, D-mannose that inhibit attachment also repressed protein injection. These results suggest that in the absence of BFP, T1F can become the dominant adherence factor. If this is correct, mannoside-class FimH antagonists developed against UPEC may be repurposed against aEPEC. 97-99 Conversely, antibiotic regimens that induce cell elongation may increase early-stage adherence. Where possible, pairing or substituting such treatments with anti-adhesion strategies could mitigate this risk. From a diagnostic perspective, simple hemagglutination assays that report on FimH-mediated attachment, 94 or fimH genotyping, could point to patients likely to benefit from anti-FimH interventions. However, these possibilities remain speculative at this stage. Although T1F emerged as the central module in our system, our results do not exclude parallel solutions in other aEPEC lineages. E. coli pan-genomes encode a plethora of chaperone-usher fimbriae, autotransporters, and curli that could likewise be harnessed to gain host-attachment through tuned expression, sequence changes, and morphology-driven avidity. Altogether, the emergence of aEPEC strains highlights a general principle: when a dominant adhesin is lost, rapid compensatory evolution proceeds by switching-ON an alternative system and then tuning it through avidity or affinity gains. The similarity between our experimental evolution and the trajectories evident in clinical genomes suggests that adhesin tuning may follow predictable rules. By focusing on T1F and FimH, we outline how aEPEC regains its grip on host tissue and provide a rationale for loosening that grip with targeted anti-adhesion strategies. Supplementary Material extended data table.xlsx Supplementary material supplemental revised2 further changes.docx Acknowledgments We thank Prof. Gad Frenkel for providing antibodies, Prof. Shoshy Altuvia for providing plasmids, Dr. Alex Rouvinski for providing lentivirus tools, Dr. Mark Saper for protein visualization, and IR and SBY lab members for insightful discussions. Special thanks to Zhilin Wu for data acquisition and analysis during revisions. Funding Statement This work was supported by the European Research Council Grant 810186; Israel Science Foundation Grant 1342/20; Einstein Foundation Grant BJ-2024-836. Disclosure of potential conflicts of interest No potential conflicts of interest were disclosed. Data availability statement WGS results are deposited in NCBI under bioproject PRJNA1420383. Custom codes are archived on Zenodo under DOI: 10.5281/zenodo.21396901. Data supporting the findings of this study are available within the article, supplementary materials and extended data table. Generative AI During the preparation of this work the authors used ChatGPT 5.4 to optimize custom codes for handling a collection of WGS of clinical isolates, and to improve the grammar and style of the text. After using this tool, the authors reviewed and edited the content as needed, and take full responsibility for the content of the published article. Supplementary material Supplemental data for this article can be accessed at https://doi.org/10.1080/19490976.2026.2725431. References Associated Data Supplementary Materials extended data table.xlsx Supplementary material supplemental revised2 further changes.docx Data Availability Statement WGS results are deposited in NCBI under bioproject PRJNA1420383. Custom codes are archived on Zenodo under DOI: 10.5281/zenodo.21396901. Data supporting the findings of this study are available within the article, supplementary materials and extended data table.