Success of Escherichia coli O25b:H4 ST131 clade C associated with a decrease in virulence

Escherichia coli of sequence type (ST) 131 resistant to fluoroquinolones and producer of CTX-M-15 is globally one of the major extraintestinal pathogenic E. coli (ExPEC). ST131 phylogenesis showed that multidrug-resistant ST131 strains belong to a clade called C, descending from an ancestral clade called B, comprising mostly antibiotic-susceptible strains. Antibiotic resistance could appear as one of the keys of the clade C global success. We hypothesized that other features of ST131 clade C could contribute to this success since other major global ExPEC clones (ST73, ST95) are mostly antibiotic-susceptible. To test this hypothesis, we measured the growth abilities, early biofilm formation and virulence-factor content of a collection of clade B and clade C strains. Moreover, using competition assays, we measured the capacity of selected representative strains of clades B and C to colonize the mouse intestine and urinary tract, and to kill mice in a septicemia model. Clade B and C strains had similar growth ability. However, clade B strains were more frequently early biofilm producers, expressed mostly faster their type 1 fimbriae and displayed more virulence factor-encoding genes than clade C strains. Clade B outcompeted clade C in the gut and/or urinary tract colonization models and in the septicemia model. These results strongly suggest that clade C strain evolution includes a loss of virulence, i.e. a process that could enhance micro-organism persistence in a given host and thus optimize transmission. This process, associated with acquired antibiotic-resistance, could ensure clade C strain survival in environments under antibiotic pressure. Importance Extraintestinal pathogen Escherichia coli (ExPEC) are virulent but mostly antibiotic-susceptible. One worrying exception is ST131, a major multidrug resistant ExPEC clone that has spread worldwide since the 2000s. To contain the emergence of this threatening clone, we need to understand what factors favored its emergence and dissemination. Here, we investigated whether multidrug-resistant ST131 had advantageous phenotypic properties beyond multidrug resistance. To this end, we competed the emergent multidrug-resistant ST131 with its antibiotic-susceptible ancestor in different conditions: biofilm production, in vivo colonization and virulence experiments. In all in vivo competitions, we found that multidrug-resistant ST131 was losing to its ancestor, suggesting a lesser virulence of multidrug-resistant ST131. It was previously described that losing virulence can increase micro-organism persistence in some populations and subsequently its level of transmissibility. Thus, a decreased level of virulence, associated with multidrug resistance, could explain the global success of ST131.

The evolutionary success of ST131 is still largely unexplained. Resistance to antibiotics, 99 notably to fluoroquinolones and ESC may explain the success of clade C strains, as these 100 resistances were shown not to impact the clade C growth fitness (11,12). However, clones 101 susceptible to antibiotics, such as ST73 and ST95, are as successful as ST131 clade C among 102 bacteremia, suggesting that factors other than antibiotic resistance can participate in the 103 success of a clone (9), 104 The aim of our study was thus to investigate the potential role of factors other than antibiotic 105 resistance, notably growth rate, biofilm formation, colonization ability and virulence, in the 106 success of ST131 clade C strains. To this purpose, we first established a collection of 39 E. 107 coli O25b:H4 ST131 comprising representatives of the ST131 clade and subclade genetic 108 diversity and tested them in a variety of in vitro phenotypic tests. We then selected three 109 strains for further in vitro and in vivo competition assays, including various mouse models.

112
Technical details for each section are available in the supplemental material Text S1.

113
Bacterial strains 114 The 39 studied O25b:H4 ST131 E. coli strains comprised 18 fimH22 and 21 fimH30 strains 115 obtained between 1993 and 2012 from different geographic origins and sources (Table S1). 116 Nalidixic acid and ciprofloxacin resistance had been determined by the agar diffusion method 117 and interpreted following the 2015 EUCAST recommendations (www.eucast.org) and ESBL 118 production by the double disk synergy test (13). CFT073 and E. coli K-12 MG1655 strains 119 were used as positive and negative controls, respectively, in the septicemia mouse model. pMLST (14)]. Then, Abricate (15) was used to detect genes encoding antibiotic resistance 124 with Resfinder (16) and virulence factors with a custom virulence database composed of 125 VirulenceFinder (17), the virulence factor database (VFDB) (18), and classical genes 126 characterizing ExPEC. Virotypes were determined as previously described (19). All contigs 127 were submitted to the MicroScope Platform (20) for further gene investigation such as 128 gyrA/parC alleles and genes involved in the biofilm formation (21). When necessary, 129 presence or absence of some genes was controlled by PCR. 130 We investigated how the different subclades B and C of ST131 were represented in our   values ranging from 20 (absence of biofilm formation) to 0 (high biofilm formation), is 155 inversely proportional to the biofilm formation ability. A BFI value of 10 was chosen as the 156 biofilm production cut-off (biofilm formation: BFI≤10, no biofilm formation: BFI>10). The 157 experiments were repeated at least three times for each strain. BHI broth without any strains 158 was used as negative control, and strains S250 and 39, previously described with this method 159 as strong and never biofilm producers, respectively, as control strains (24).

160
Yeast cell agglutination assay 161 Expression of type 1 fimbriae was assessed by using the yeast cell (Saccharomyces 162 cerevisiae) agglutination assay as previously described (25), and after adaptations to highlight 163 the early expression of type 1 fimbriae, by using 10 µL of a pellet obtained after 164 centrifugation (3000 g, 10 min) of 3 mL of LB broth culture after 2 and 5 h-incubations for 165 the agglutination assay. Based on the biofilm formation and the sequence of the fim operon, 166 strains S250 and H1447 were used as positive and negative controls, respectively. To determine the relative fitness of strains in planktonic conditions, they were grown in 188 couple in LB broth during 100 generations, i.e. during ten days, as previously described (29).

189
Competitive indexes (CI) were obtained using the following formula:  Six-week-old female mice CD-1 from Charles River® (L'Arbresle, France) pre-treated with 199 streptomycin before inoculation of challenging strains were used to assess strains' relative 200 ability to colonize the mouse intestine, as previously described (28). At least five mice were Female mice OF1 of 14-16 g (4-week-old) from Charles River ® were used to assess the 207 strains' relative ability to cause sepsis, as previously described (31). After subcutaneous 208 inoculation of individual strain or mixed strains at a 1:1 ratio, time to death was monitored 209 during seven days. Five to ten mice were used for each assay. In all experiments, CFT073  Urinary tract infection mouse model 218 CBA female mice of 8-22g (8-week-old) from Janvier ® (Le Genest-Saint-Isle, France) were used 219 to assess strains' relative ability to cause an ascending unobstructed urinary tract infection, as 220 previously developed in our lab (33). Ten mice were used for each competition assay, and the 221 experiments were repeated twice. CIs were obtained as described above.  To visualize all data of our strain collection in one graph, a multiple correspondence analysis   (Table S5).

278
Genes encoding antibiotic resistance 279 As shown in Table S5, our collection comprised strains that displayed a diverse antibiotic   291 Maximum growth rate (MGR) 292 We assessed the fitness of our 39 strains by measuring MGR in planktonic conditions ( Figure   293 3). MGR did not differ significantly between clade B and clade C strains (P = 0.08), and 294 between clinical and feces isolates (P=0.07) ( Figures 3A and 3B). Within clade B strains, 295 those resistant to nalidixic acid seemed to have a lower MGR than the susceptible ones 296 13 ( Figure 3C), but this difference was not significant (P=0.2). Regarding clade C strains, no 297 significant MGR differences were observed between nalidixic acid/ciprofloxacin-resistant 298 strains and susceptible strains (P=0.9) ( Figure 3D).

299
Kinetic of early biofilm formation and expression of type 1 and curli fimbriae 300 Clade B strains were more frequently biofilm producers than clade C strains at 2 and 5 h 301 (P<0.001). We found three significantly different phenotypes of early biofilm production 302 (P<0.0001) within our 39 strains ( Figure 2, Table S5): early and persistent producers (BFI≤10 303 at 2, 3 and 5 h), delayed producers (BFI>10 at 2 and 3 h but ≤10 at 5 h) and never producers 304 (BFI>10 at 2, 3 and 5 h). As BFI values after 2 and 3 h of incubation classified a given strain 305 into the same biofilm production phenotype (Figure 2A), only BFI values obtained after 2 and 306 5 h of incubation were considered for further analyses. The 20 clade C strains were either 307 delayed producers (n=8) or never producers (n=12). There was no significant association 308 between subclades C and biofilm phenotype. Clade B strains produced biofilm more 309 frequently than clade C strains. Five (71%) of the B5 strains were early producers and two 310 (29%) were delayed producers, whereas all eight B4 strains were delayed or never producers.

312
As the ability to form early biofilm is linked to type 1 fimbriae and curli fimbriae expression 313 (36), we studied these expressions in our strains by using the yeast agglutination test and 314 analyzing colony aspect of bacteria spotted on Congo red agar, respectively. After 2 and 5 h 315 of shaking growth, we observed that the early biofilm producers (n=8) expressed their type 1 316 fimbriae more quickly than the delayed and never producers (P≤0.0001) ( To investigate the genetic factors potentially involved in differences in biofilm formation, we 326 examined the diversity in the fim operon across all strains, knowing that Fim proteins 327 participate in biofilm formation (21) and that an insertion sequence ISEc55 has been 328 described in the clade C strain fimB gene (25) that encodes a co-factor of the type 1 fimbriae 329 synthesis regulation. We found at least one non-synonymous SNP per gene in the fim operon  Figures 2B and 2C). ∆fimB variants were also submitted to yeast agglutination test and 342 Congo red assay. Both ∆fimB variants were negative for early yeast agglutination test, unlike 343 parent strains, but were positive for standard test. No changes were observed on Congo red 344 phenotype for both variants (data not shown).

345
Further genes and/or deduced proteins previously described as involved in E. coli biofilm 346 formation (21) were compared between our strains by taking S250 strain as reference (data 347 not shown). We observed identical protein sequence modifications in all B4 strains that were 348 all delayed biofilm producers: substitution K142Q in FlgD that is required for flagellar hook   (Table S5) and MGR values. Among clade B strains, we selected strain 358 S250 i.e. one of the two most ancestral strains (subclade B1) that we called "Ancestor".

359
Among clade C strains, we selected a C1 subclade strain, CES164C, harboring the clade C-   Figure 5B). However, a small competitive advantage was observed for both Ancestor and 393 Hybrid against Emergent, after 7 days in anaerobic biofilm conditions. Lack of the fimB gene 394 did not impact the relative fitness of Ancestor and Hybrid in these conditions ( Figure 5B).  Urinary tract infection mouse model 419 Ancestor displayed a better urinary tract colonization capacity than Emergent. In total, 30 420 mice were inoculated, with an overall infection rate of 87%, and no death was observed 48 h 421 after the inoculation. Bladders and kidneys collected 48 h after inoculation contained an 422 average of 9.10 7 CFU/g (range: 4.10 4 -2.10 9 CFU/g) and of 5.10 5 CFU/g (range: 6.10 1 -2.10 6 423 CFU/g), respectively, showing a satisfactory colonization rate of mice (data not shown).                       expressed in log, and CI above zero means that isolate 1 outcompetes isolate 2, and conversely.

853
An interval of ±1 log was considered as the limit of discrimination. Each dot represents a mouse

866
Competitive index (CI) is expressed in log, and CI above zero means that isolate 1 outcompetes 867 isolate 2, and conversely. An interval of ±1 log was considered as the limit of discrimination.