Bacterial infection is linked to colorectal carcinogenesis, but the species that contribute to a protumorigenic ecology are ill-defined. Here we report evidence that α-hemolysin–positive (hly+) type I Escherichia coli (E. coli) drives adenomagenesis and colorectal cancer in human females but not males. We classified E. coli into four types using a novel typing method to monitor fimH mutation patterns of fecal isolates from adenoma patients (n= 59), colorectal cancer patients (n= 83), and healthy subjects (n= 85). hly+ type I E. coli was found to be relatively more prevalent in stools from females with adenoma and colorectal cancer, correlating with poor survival in colorectal cancer patients. In mechanistic studies in female mice, we found that hly+ type 1 E. coli activated expression of the glucose transporter GLUT1 and repressed expression of the tumor suppressor BIM. hly-encoded alpha hemolysin partially accounted for these effects by elevating the levels of HIF1α. Notably, colon tumorigenesis in mice could be promoted by feeding hly+ type I E. coli to female but not male subjects. Collectively, our findings point to hemolytic type I E. coli as a candidate causative factor of colorectal cancer in human females, with additional potential as a biomarker of disease susceptibility. Cancer Res; 76(10); 2891–900. ©2016 AACR.

Colorectal cancer is the third most common cancer and has become a major public health problem worldwide. Carcinogenesis of colorectal cancer has been linked to Escherichia coli (E. coli). E. coli is a commensal of the normal gut microflora of humans and other mammals, and is one of the most common causes of infections by Gram-negative bacilli (1). Enteropathogenic E. coli was reported to downregulate expression of key DNA mismatch repair proteins MSH2 and MLH1 (2), induce cancer cell detachment via cytoskeleton rearrangement, and reduce cellular apoptosis of the detached cancer cells (3). However, no in vivo evidence has been provided for the role of enteropathogenic E. coli for colorectal cancer. Another type of E. coli that belongs to B2 phylogenetic group carries a pks island and produces cyclomodulins has also been linked to colorectal cancer. pks+E. coli has been reported to cause DNA double-strand breaks and activate the DNA damage checkpoint pathway, leading to cell-cycle arrest and cell death (4). Recent studies showed that pks+E. coli was present in a significantly high percentage of inflammatory bowel disease and colorectal cancer patients and promoted colorectal cancer development in mice without affecting inflammation (5).

Although the previous evidence indicates the association between E. coli and colorectal cancer and provides some clues on the underlying mechanisms, many questions remain unanswered. Screening and comparison of a large number of clinical isolates from patients with colorectal cancer, adenoma, and healthy controls are necessary to give more information. Efforts are also desired to identify previously unidentified colorectal cancer–associated E. coli and to uncover their oncogenic mechanisms. Moreover, it is unknown whether gender-specific differences exist for colorectal cancer development and if bacteria have any role in this process. Here we identify hemolytic type I E. coli as a type of bacteria that is significantly associated with adenoma and colorectal cancer in female patients, by analyzing a large number of clinical E. coli isolates. Hemolytic type I E. coli activates the expression of the glucose transporter GLUT1 and reduces the expression of the tumor suppressor BIM at least in part by acting on hypoxia-induced α-subunit (HIF1α), and the hly-encoded alpha hemolysin is required for the regulation. We finally verify the tumorigenic capacity of hemolytic type I E. coli in two mouse models and demonstrate the requirement of alpha hemolysin of type I E. coli for colonic tumorigenesis.

fimH sequencing, clustering, and typing

Given the higher rate of point mutation in the mannose-binding lectin domain than in the pilin domain (6), a gene fragment (from −145 to +687 relative to the start codon) of fimH covering the lectin domain-encoding region was PCR amplified using primers fimH-F and fimH-R and then sequenced. For each stool sample, 4–6 E. coli isolates were randomized selected for the fimH sequencing analysis. As a control, fimH of E. coli K-12 MG1655 (a laboratory reference strain) was included for sequencing. Low-quality sequencing data were removed (i.e. sequencing quality score < 30 and error probability > 0.1%; or reading rate < 50%) and the qualified fimH sequences of fecal E. coli isolates were aligned to that of the reference strain MG1655 using BWA-MEM (7). A binary matrix was constructed on the basis of SNP information of the fimH sequences and cluster analysis was performed using the R pheatmap package. In addition to clustering, NMF package in R (8) was used to determine how many types that fimH should be grouped into. Factorization rank r, which defines the number of basis effects used to approximate the target matrix, was decided on by trying different values. 100 NMF runs were performed for a range of factorization rank r from 2 to 10. To choose the optimal rank r, we employed approaches including cophenetic correlation efficient and an approach based on the variation of the residual sum of squares (RSS) between the target matrix and its estimate. For the cophenetic correlation efficient approach, we chose the smallest value of r for which this coefficient started decreasing. For the RSS approach, we chose the value of r for which the plot of the RSS showed an infection point. We employed WebLogo (9, 10) to show the SNP composition of different types of fimH.

Genetic engineering of E. coli isolates

Gene deletion from the chromosome of E. coli isolates was achieved by using the λ-Red recombination and verified by colony PCR (11).

Primer design for detection of virulence genes

For detection of type I fimH, primers CRC-F and CRC-R3 were designed. Annealing temperature was adjusted to 62°C, at which the PCR had 100% sensitivity and 90% specificity. For detection of the pks island, primers published previously (4) were used in this study. For detection of hly and other virulence genes, conserved sequence regions were first identified across different E. coli strains by evaluation of multiple sequence alignments. Then, primers were designed to be specific to the conserved regions. All the primers are listed in Supplementary Table S4.

PCR quantification of hly+ and pks+ type I E. coli in stools

We determined percentage of hly+ or pks+E. coli relative to total E. coli by absolute quantification real-time PCR. Genomic DNA was isolated from stools using QIAamp Fast DNA Stool Mini Kit (Qiagen), and was isolated from colonic mucosa using TRIzol reagent (Life Technologies). Total E. coli were quantified by determining lacZ+ bacteria, given that most E. coli strains are positive with lacZ (12). As one E. coli cell harbors one copy of the lacZ gene, we determined the number of E. coli by quantifying the lacZ gene. This is also the case for hly+ or pks+type I E. coli. We performed absolute quantification real-time PCR for lacZ, hly, pks, and type I fimH, described in Supplementary Materials and Methods.

Mouse models for colorectal tumorigenesis

Six to 8-week-old female BALB/c mice were used in this study. Studies were approved by the Animal Experimentation Ethics Committee. In some experiments, mice were not treated with any carcinogen. In some experiments, mice were pretreated with azoxymethane (AOM), a colonic genotoxic carcinogen. In this study, AOM treatment was used to increase the basal level of colorectal tumorigenesis but not for direct induction of colorectal tumor formation, so that oncogenic effects of E. coli could be detected. On day 0, mice were intraperitoneally injected with AOM (10 mg/kg), according to a published protocol (13). On day 5, mice were treated with neomycin (1 g/L), vancomycin (0.5 g/L), ampicillin (1 g/L), and metronidazole (1 g/L) in drinking water to eliminate commensal flora (14) and facilitate the colonization of fed bacteria. Dextran sulfate sodium (DSS) is typically used together with AOM to induce colorectal cancer in mice (13). However, the oncogenic action of the combination of AOM and DSS are strong and may overshadow the oncogenic effects of fed bacteria. To avoid this, DSS was not used in our experiments. After the antibiotic treatment, mice were fed by oral gavage with PBS or E. coli (1 × 108 bacteria per mouse) every other day. During the experiments, mice were weighed weekly. At the end of the experiments, mice were anesthetized with sodium pentobarbital intraperitoneally (60 mg/kg) and the large intestine was excised. The large intestine was cut open longitudinally along the main axis, and a piece (approximately 2 mm in length) of tissue free of tumors was immediately stored in RNAlater (Qiagen). In some experiments, the large intestine was rolled up and fixed in 10% buffered formalin for at least 24 hours. Paraffin-embedded sections of the Swiss roll of the large intestine were then made by routine procedures. We did not rely on macroscopic evaluation of colonic tumor growth, as it did not discriminate colonic tumors from normal lymphoid aggregates. Instead, we quantified colonic tumor growth microscopically. Each colonic Swiss roll was cut into nine cross sections, which were prepared at an equal distance apart. Distance between two adjacent sections was calculated so that the entire colonic Swiss roll was sectioned. The sections of each Swiss roll were subjected to hematoxylin and eosin staining and then microscopically examined for colonic tumors arising from the epithelial tissues. Colonic inflammation in the mouse model was evaluated histomorphologically and scored using a previously published criterion (15).

E. coli is classified into four types based on fimH mutation patterns

The fimH gene of E. coli is prone to mutation and the mutation patterns help predict pathogenic potential (6). To see whether E. coli with certain fimH mutation patterns is associated with colonic adenoma and colorectal cancer, we isolated fecal E. coli from 59 adenoma patients, 83 colorectal cancer patients, and 85 healthy subjects. fimH was successfully sequenced from 711 E. coli isolates (Supplementary Table S1), and displayed single nucleotide polymorphisms (SNP) at numerous sites. Cluster analysis of the SNPs (Supplementary Fig. S1A) and nonnegative matrix factorization (Supplementary Fig. S2A and S2B) showed that fimH could be classified into four types. Among the four types, detection frequency of type I E. coli was similar among the three male groups, but was significantly higher in the female colorectal cancer patients than in the other female groups (P < 0.001; Supplementary Fig. S1B and S1C). These data suggested gender-dependent specificity of type I E. coli to colorectal cancer. It is noteworthy that the analyses based on isolate sequencing just roughly estimated detection frequency of each type of E. coli, as only a few isolates were sequenced for each stool sample. Frequency of type I E. coli in the female patients with adenoma was later found to be underestimated by the sequencing-based method, as revealed by quantitative PCR analyses of DNA isolated from stools containing a large number of E. coli. Therefore, E. coli detection rates were later precisely determined by quantitative PCR in this study.

One-third of type I E. coli produce hly-encoded alpha hemolysin

We then examined the interaction between colon epithelial cells and E. coli isolates including 32 type I isolates and 63 isolates belonging to other types. A human commensal intestinal E. coli reference strain MG1655, which was found not to promote colorectal carcinogenesis (16), was included as an additional control. After 3 hours of bacterium cell coculture at a multiplicity of infection (MOI) of 100, 10 of 32 type I E. coli killed most NCM460 cells (a normal human colon mucosal epithelial cell line), whereas other isolates did not (Fig. 1A). The type I killers at a MOI of 100 started to kill cells 1 hour after the coculture (Fig. 1B). The cell death caused by the type I killers was also observed with human colon cancer cell lines HCT116 and SW480 (Fig. 1C). The hly-encoded alpha hemolysin was reported to kill host cells (17), leading us to speculate that the cell death caused by the type I killers was due to alpha hemolysin. As speculated, all type I killers induced hemolysis but none of type I non-killers and other E. coli isolates did so (Fig. 1C). To further verify the role of hly for the type I E. coli-mediated cytotoxicity, we randomly selected a hly+ type I E. coli isolate, which was isolated from the stool of a female colorectal cancer patient, for mutagenesis analysis. This fecal isolate was hereafter referred to as J47. We deleted the entire hly operon from J47, generating a hly-null mutant named J47Δhly. Deleting hly abolished hemolysis and cytotoxicity, and transformation of the hly mutant with a hly-overexpressing plasmid (18) restored them (Fig. 1D). Thus, the hly-encoded alpha hemolysin is responsible for the cytotoxic effects of type I killers on colon epithelial cells.

Figure 1.

Cytotoxicity of type I E. coli killers. A, viability percentage (%) of NCM460 after 3 hours of coculture with type I E. coli (32 isolates) and other types of E. coli (64 isolates). B, time-course viability percentage of NCM460 cocultured with an isolate of the type I E. coli, three randomly selected isolates of other types of E. coli, and the reference strain MG1655. Type I, n = 6; other types, n = 18; MG1655, n = 6. C, cytotoxicity of type I E. coli was correlated with hemolysis. Three cell lines, including NCM460, HCT116, and SW480, were tested. In the box are the 32 isolates belonging to type I E. coli. D, effects of deleting the hly operon and restoring the hly expression on cytotoxicity and hemolysis of type I E. coli. J47, a wild-type isolate of hly+ type I E. coli; Δhly, an isogenic mutant deleted for the hly operon; Δhly/pUC, the hly-null mutant carrying an empty control vector pUC; Δhly/phly, the hly-null mutant carrying a vector pUC-hly. n = 6; error bar, SEM. ns, statistically not significant; ***, P < 0.001.

Figure 1.

Cytotoxicity of type I E. coli killers. A, viability percentage (%) of NCM460 after 3 hours of coculture with type I E. coli (32 isolates) and other types of E. coli (64 isolates). B, time-course viability percentage of NCM460 cocultured with an isolate of the type I E. coli, three randomly selected isolates of other types of E. coli, and the reference strain MG1655. Type I, n = 6; other types, n = 18; MG1655, n = 6. C, cytotoxicity of type I E. coli was correlated with hemolysis. Three cell lines, including NCM460, HCT116, and SW480, were tested. In the box are the 32 isolates belonging to type I E. coli. D, effects of deleting the hly operon and restoring the hly expression on cytotoxicity and hemolysis of type I E. coli. J47, a wild-type isolate of hly+ type I E. coli; Δhly, an isogenic mutant deleted for the hly operon; Δhly/pUC, the hly-null mutant carrying an empty control vector pUC; Δhly/phly, the hly-null mutant carrying a vector pUC-hly. n = 6; error bar, SEM. ns, statistically not significant; ***, P < 0.001.

Close modal

hly and pks coexist in some hly+ type I E. coli

We then detected by PCR the presence of a set of well-known E. coli virulence genes in 97 E. coli isolates including hly+ type I (16 isolates), hly type I (26 isolates), other types of E. coli isolates (38 isolates), and the reference strain MG1655. We also determined the phylogenetic type and assayed the ability of each isolate to cause hemolysis (Supplementary Table S2). Among the genes tested, only the hly operon and the pks island were present in all hly+ type I isolates but absent from hly type I and other types of E. coli isolates. The hly operon was accompanied with the pks island [a pathogenic island encoding giant modular nonribosomal peptide and polyketide synthases (4)] in all the 16 hly+ type I isolates, but 3 of 19 pks+ type I E. coli isolates did not carry hly. Phylogenetic typing showed that 73.8% (31/42) of the type I E. coli isolates belonged to type B2 and the other 26.2% belonged to type D. Among them, all the hly+ type I E. coli and pks+ type I E. coli belonged to type B2. In contrast, most of other types of E. coli belonged to type A, B1, or D. The data based on the 97 E. coli isolates might not provide precise and full information on the association between pks and hly. We therefore further determined the association by PCR analysis of DNA isolated from at least 50 mg of stools that contained a large number of E. coli. We found that stool samples could be positive with both pks+ type I E. coli and hly+ type I E. coli, but could also be positive with only one of them (Fig. 2A). This indicates that pks and hly do not always coexist. pks+E. coli has previously been demonstrated to promote colorectal cancer development (4, 5), which promoted us to ask whether hly is a colorectal cancer–associated factor in addition to pks and if they play different roles in colorectal cancer development.

Figure 2.

Detection rates of hly+ and pks+ type I E. coli in different subject groups. A, partial overlap of hly+ and pks+ type I E. coli in stools of different subject groups. B, detection rates of pks+ type I E. coli, hly+ type I E. coli, and hly type I E. coli in stools of male subject groups. C, detection rates of pks+ type I E. coli, hly+ type I E. coli, and hly type I E. coli in stools of female subject groups. H, healthy subjects; AD, adenoma patients; CRC, colorectal cancer patients; ns, statistically not significant; ***, P < 0.001.

Figure 2.

Detection rates of hly+ and pks+ type I E. coli in different subject groups. A, partial overlap of hly+ and pks+ type I E. coli in stools of different subject groups. B, detection rates of pks+ type I E. coli, hly+ type I E. coli, and hly type I E. coli in stools of male subject groups. C, detection rates of pks+ type I E. coli, hly+ type I E. coli, and hly type I E. coli in stools of female subject groups. H, healthy subjects; AD, adenoma patients; CRC, colorectal cancer patients; ns, statistically not significant; ***, P < 0.001.

Close modal

hly+ but not pks+ type I E. coli is associated with both adenoma and colorectal cancer in females

Quantitative PCR analyses of stool DNA showed that detection rates of neither pks+ nor hly+ type I E. coli were significantly different among male subject groups (Fig. 2B). When female subjects were analyzed, detection rate of pks+ type I E. coli was higher in the colorectal cancer patients than in the healthy subjects, but did not differ between the healthy subjects and adenoma patients. In contrast, hly+ type I E. coli was detected in 36.1% (13/36) of the female colorectal cancer patients, 39.3% (11/28) of the female adenoma patients, but in only 6.7% (4/60) of the female healthy subjects (Fig. 2C). Thus, unlike pks+ type I E. coli that is linked with colorectal cancer but not with adenoma in females, hly+ type I E. coli is associated with both colorectal cancer and adenoma in females. This association is independent of colorectal cancer stage (P = 0.882) or age (P = 0.445), as revealed by backward elimination analysis. Smoking is a risk factor for many cancers (19), which led us to ask whether colonization of hly+ type I E. coli in the female patients is associated with smoking. None of the female colorectal cancer patients with smoking information were smokers or ex-smokers (Supplementary Table S1), ruling out this possibility. In contrast, the three female groups displayed similar detection rates of hly type I E. coli, so did the three male groups (Fig. 2C). This indicates that hly type I E. coli is not associated with colorectal cancer or adenoma in either males or females.

hly+ but not pks+ type I E. coli correlates with poor survival in female colorectal cancer patients

Survival analyses showed that the presence of hly+ type I E. coli did not affect survival in male colorectal cancer patients (Fig. 3A) but was associated with shorter survival in female colorectal cancer patients (Kaplan–Meier, P = 0.026; Fig. 3B). Survival of the female colorectal cancer patients was not affected by age (P = 0.836) or colorectal cancer stage (P = 0.516) at the point of sample collection, as revealed by stepwise regression analysis through backward elimination. Thus, hly+ type I E. coli influenced survival independently of these host factors. When colorectal cancer patients carrying hly type I E. coli (i.e. carrying type I E. coli but negative with hly) were compared with other colorectal cancer patients, no difference was observed in survival in either male or female patients (Fig. 3C and D). Thus, hly type I E. coli is not associated with survival in colorectal cancer patients. Unlike hly+ type I E. coli that specifically correlated with shorter survival in female colorectal cancer patients, pks+ type I E. coli had no effects on survival in either male or female colorectal cancer patients (Fig. 3E and F). Collectively, hly+ type I E. coli but not pks+ type I E. coli correlates with poor survival in female colorectal cancer patients.

Figure 3.

Association between hly+ type I E. coli and survival of colorectal cancer patients. Survival of male (A) or female (B) colorectal cancer patients carrying hly+ type I E. coli and those free of this type of E. coli. Survival of male (C) or female (D) colorectal cancer patients carrying hly type I E. coli and those free of this type of E. coli. Survival of male (E) or female (F) colorectal cancer patients carrying pks+ type I E. coli and those free of this type of E. coli.

Figure 3.

Association between hly+ type I E. coli and survival of colorectal cancer patients. Survival of male (A) or female (B) colorectal cancer patients carrying hly+ type I E. coli and those free of this type of E. coli. Survival of male (C) or female (D) colorectal cancer patients carrying hly type I E. coli and those free of this type of E. coli. Survival of male (E) or female (F) colorectal cancer patients carrying pks+ type I E. coli and those free of this type of E. coli.

Close modal

hly+ type I E. coli colonizes colonic mucosa, which is dependent on FimH

Colorectal cancer–causing/promoting bacteria have to be capable of colonizing the colon. As all the four types of E. coli produced the fimH-encoded pilus (FimH) that binds to D-mannose structures on the surface of host cells (20), we speculate that the FimH might facilitate the intestinal colonization of hly+ type I E. coli. We examined this possibility in female BALB/c mice. Specifically, mice were pretreated with antibiotics to eliminate native bacteria and to facilitate colonization of fed bacteria. Mice were then daily fed by oral gavage with J47, its isogenic mutant J47Δhly or J47ΔfimH for 3 days. E. coli feeding was then stopped and the fed bacteria in stools were quantified. Compared with J47 and J47Δhly, colonization levels of J47ΔfimH were lower and dropped more quickly after E. coli feeding was stopped. In contrast, colonization levels of J47 and J47Δhly were comparable (Supplementary Fig. S3A). In another experiment, antibiotic-pretreated mice were fed by oral gavage with J47, J47Δhly, or J47ΔfimH every other day for 4 weeks and then left untreated. Colonization of the fed bacteria in the rectum mucosa was tested at 10 weeks. J47 accounted for 2.86% (SEM, 1.33%) of total mucosal E. coli and J47Δhly accounted for 0.36% (SEM, 0.21%). There was no significant difference between J47 and J47Δhly in colonization in the colonic mucosa (P = 0.095). In contrast, J47ΔfimH was barely detected in the colonic mucosa (Supplementary Fig. S3B). These in vivo data indicate that FimH is required for colonization of the intestinal tract by hly+ type I E. coli, making it possible for the bacteria to interact with the host cells.

hly+ type I E. coli activates multiple cancer pathways in vitro

We next performed a Cancer Pathway Finder PCR array to evaluate the tumorigenic potential of hly+ type I E. coli. The NCM460 cell line was co-cultured with J47, J47Δhly or PBS for 3 hours, followed by PCR array analysis. MOI was reduced to 25 so that most cells were not killed by bacteria after the coincubation. The reference strain MG1655, which did not promote colorectal carcinogenesis (16), was employed as an additional control so that gene regulation caused by irrelevant E. coli could be ruled out (Fig. 4A). Compared with PBS, MG1655 enhanced expression of only 10 oncogenes, whereas J47 upregulated expression of 17 oncogenes involved in angiogenesis, antiapoptosis, cell cycle, glucose transportation, epithelial-to-mesenchymal transition, metabolism, and response to hypoxia (Supplementary Table S3), suggesting that hly+ type I E. coli activates multiple carcinogenic pathways in vitro. J47 also increased expression of 8 genes responsible for apoptosis, cell senescence, and DNA repair, and MG1655 elevated expression of 6 such genes, showing that induction of these genes is a common feature of both colorectal cancer–associated and -irrelevant E. coli. Thus, compared with the irrelevant strain MG1655, J47 has a carcinogenic potential. Deleting hly abolished regulation of 10 oncogenes that were induced by J47 (10/17 oncogenes, 58.8%), suggesting that the hly-encoded alpha hemolysin is a key player in the in vitro carcinogenesis induction mediated by hly+ type I E. coli.

Figure 4.

Regulation of cancer pathways by hly+ type I E. coli. A, PCR array gene expression clustergram of NCM460 treated with PBS or E. coli strains. B, Western blot analysis of expression of colorectal cancer-associated genes that were regulated by hly+ type I E. coli J47 via the hly-encoded alpha hemolysin. C, correlation between expression of GLUT1, BIM, and HIF1α in the rectum mucosa of mice fed for 10 weeks with J47 or J47Δhly. *, P < 0.05; ***, P < 0.001.

Figure 4.

Regulation of cancer pathways by hly+ type I E. coli. A, PCR array gene expression clustergram of NCM460 treated with PBS or E. coli strains. B, Western blot analysis of expression of colorectal cancer-associated genes that were regulated by hly+ type I E. coli J47 via the hly-encoded alpha hemolysin. C, correlation between expression of GLUT1, BIM, and HIF1α in the rectum mucosa of mice fed for 10 weeks with J47 or J47Δhly. *, P < 0.05; ***, P < 0.001.

Close modal

hly+ type I E. coli activates GLUT1 and represses BIM through HIF1α in vivo

We next focused on those that were upregulated by J47 relative to J47Δhly (>2-fold change). These included three oncogenes CCL2, FOXC2, and SLC2A1 and one tumor suppressor gene BCL2L11. BCL2L11 encodes a proapoptotic protein Bcl-2–like protein 11 (BIM) involved in apoptosis of colorectal cancer cells (21), whereas CCL2 (22, 23), FOXC2 (24, 25), and SLC2A1 (encoding GLUT1; refs. 26, 27) are overexpressed in human colonic tumors or indicative of poor prognosis. The above in vitro data might not reflect in vivo expression regulation. We therefore tested if the hly-encoded alpha hemolysin was required for activation of their expression in vivo. These four genes were evaluated for their expression in the colonic tissues of mice fed for 10 weeks with J47 or J47Δhly. No difference was detected between the two groups of mice in terms of expression of FOXC2 (P = 0.177) or CCL2 (P = 0.368). In contrast to the in vitro data, feeding J47 resulted in a reduction in the expression of BIM, which is a tumor suppressor, as compared with its isogenic mutant J47Δhly (P = 0.022). In agreement with the in vitro PCR array data, J47 significantly increased the in vivo expression of GLUT1 than J47Δhly (P = 0.018; Fig. 4B). GLUT1 is known to be activated by hypoxia-induced factor-1 (HIF1), a transcription factor consisting of a constitutively expressed β-subunit and HIF1α (28, 29). This led us to examine whether the hly-encoded alpha hemolysin somehow upregulated HIF1α. Western blot analysis showed that J47 significantly increased the levels of HIF1α, compared with J47Δhly (P < 0.001; Fig. 4B). Moreover, the GLUT1 levels of the colonic mucosa were linearly correlated with the HIF1α levels (P = 0.001; Pearson correlation coefficient, r = 0.876; Fig. 4C), indicating that the J47-mediated upregulation of GLUT1 is dependent on increased HIF1α levels. BIM is known to be repressed by HIF1 (30, 31). In agreement with this, our Western blot analysis of protein expression in the colonic mucosa showed that the BIM levels were negatively correlated with the HIF1α levels (P = 0.048; Pearson correlation coefficient, r = −0.637). An inverse correlation was also observed between GLUT1 and BIM (P = 0.028; Pearson correlation coefficient, r = −0.688; Fig. 4C). Taken together, these in vivo data show that the hly-encoded alpha hemolysin is involved in the induction of GLUT1 expression and repression of BIM at least partially by acting on HIF1α. The inconsistency of in vitro and in vivo BIM regulation by hly+ type I E. coli could be due to the fact that the in vitro cell–E. coli interaction is greatly different from the complex interplay between host cells and E. coli in the colon.

Alpha hemolysin confers the ability of hly+ type I E. coli to promote colorectal tumorigenesis in wild-type mice

Elevated glucose transport and GLUT1 expression are required for the oncogenic transformation of mammalian cells, and are associated with poor survival of cancer patients (26, 27, 29). BIM repression is regarded as an important step during colonic tumorigenesis (32). As the above data showed that hly+ type I E. coli activates GLUT1 expression and reduces BIM expression in a manner dependent on alpha hemolysin, we asked whether this type of E. coli could induce colonic tumorigenesis in vivo and if alpha hemolysin is required for this process. We employed two mouse models for testing these possibilities. In the first model, AOM-pretreated female BALB/c mice were fed by oral gavage with J47, J47Δhly, or PBS alone every other day for 10 weeks. Feeding J47 reduced body weight growth of mice (Fig. 5A), but had no significant effects on colon length or inflammation scores (Supplementary Fig. S4A–S4C; ref. 15). There were polyp-like outgrowths in the colons of all the groups, and most of them were not real tumors but lymphoid aggregates that are normally present in the colon (Supplementary Fig. S4D). To correctly quantify colonic tumors, each colon was made into a Swiss roll, which was then cut into nine cross sections. Microscopic evaluation of the Swiss rolls revealed that mice fed with J47 developed significantly more colonic tumors than mice fed with PBS (P = 0.013), demonstrating that hly+ type I E. coli promotes colonic tumorigenesis. Comparison of the J47 group and the J47Δhly group showed that deleting hly reduced the ability of type I E. coli to induce tumor growth, but the effects were marginally significant in the AOM-pretreated mice (P = 0.059; Fig. 5B and C). We further examined the tumorigenic potential of J47 in mice without AOM pretreatment. At 22 weeks, 4 of 12 mice (33.3%) fed with J47 developed colonic tumors, whereas mice fed with J47Δhly did not develop any tumors (P = 0.039; Fig. 6A and B). We finally tested whether hly+ type I E. coli had any effects on colonic tumorigenesis in male mice. AOM-pretreated male BALB/c mice were fed by oral gavage with J47, J47Δhly, or PBS alone every other day for 10 weeks, and then examined for colonic tumor formation as described above. Although feeding J47 resulted in colonic tumor formation in some male mice (1 of 11 male mice; 9.1%), the tumorigenic effects of J47 were not significant in the male mice as compared with either J47Δhly (P = 0.478) or PBS (P = 0.519; Fig. 7). Thus, in contrast to what was observed in female mice in which hemolytic E. coli significantly increased colonic tumor formation as compared with PBS buffer (P = 0.013), the effects of hemolytic E. coli on colonic tumorigenesis were not significant in male mice.

Figure 5.

Effects of hly+ type I E. coli on colorectal tumorigenesis in female mice pretreated with AOM. A, reduction of body weight of mice as a result of oral gavage with J47. B, number of colonic tumors determined by microscopic examination of nine cross sections of each colonic Swiss roll stained with hematoxylin and eosin. C, representative cross-section picture of colonic Swiss roll of mice fed with J47, J47Δhly, or PBS. Magnified are three representative colonic tumors (red arrows) detected in the colon of mice fed with J47. Error bar, SEM; *, P < 0.05.

Figure 5.

Effects of hly+ type I E. coli on colorectal tumorigenesis in female mice pretreated with AOM. A, reduction of body weight of mice as a result of oral gavage with J47. B, number of colonic tumors determined by microscopic examination of nine cross sections of each colonic Swiss roll stained with hematoxylin and eosin. C, representative cross-section picture of colonic Swiss roll of mice fed with J47, J47Δhly, or PBS. Magnified are three representative colonic tumors (red arrows) detected in the colon of mice fed with J47. Error bar, SEM; *, P < 0.05.

Close modal
Figure 6.

Effects of hly+ type I E. coli on colorectal tumorigenesis in female mice without AOM pretreatment. A, number of colonic tumors developed in mice, determined by microscopic examination of nine cross sections of each colonic Swiss roll stained with hematoxylin and eosin. Error bar, SEM. B, representative cross-section picture of colonic Swiss roll of mice fed with J47 or J47Δhly. Magnified is a representative colonic tumor detected in the colon of mice fed with J47. *, P < 0.05.

Figure 6.

Effects of hly+ type I E. coli on colorectal tumorigenesis in female mice without AOM pretreatment. A, number of colonic tumors developed in mice, determined by microscopic examination of nine cross sections of each colonic Swiss roll stained with hematoxylin and eosin. Error bar, SEM. B, representative cross-section picture of colonic Swiss roll of mice fed with J47 or J47Δhly. Magnified is a representative colonic tumor detected in the colon of mice fed with J47. *, P < 0.05.

Close modal
Figure 7.

Effects of hly+ type I E. coli on colorectal tumorigenesis in male mice pretreated with AOM. A, number of colonic tumors developed in mice, determined by microscopic examination of nine cross sections of each colonic Swiss roll stained with hematoxylin and eosin. Error bar, SEM. B, representative cross-section picture of colonic Swiss roll of mice fed with J47 or J47Δhly. Magnified is a representative colonic tumor detected in the colon of mice fed with J47. ns, P > 0.05.

Figure 7.

Effects of hly+ type I E. coli on colorectal tumorigenesis in male mice pretreated with AOM. A, number of colonic tumors developed in mice, determined by microscopic examination of nine cross sections of each colonic Swiss roll stained with hematoxylin and eosin. Error bar, SEM. B, representative cross-section picture of colonic Swiss roll of mice fed with J47 or J47Δhly. Magnified is a representative colonic tumor detected in the colon of mice fed with J47. ns, P > 0.05.

Close modal

In this study, we reveal that hemolytic type I E. coli is a causative factor for colonic adenoma and colorectal cancer in females but is not associated with these diseases in males. Prevalence of hemolytic type I E. coli is dramatically increased in female patients with colonic adenoma and colorectal cancer compared with healthy female subjects, and correlates with poor survival in female colorectal cancer patients (P = 0.026). In contrast, prevalence of hemolytic type I E. coli is similar between male colorectal cancer patients, male adenoma patients and healthy male subjects, and has no effects on the survival in male colorectal cancer patients. Animal experiments showed that hemolytic type I E. coli induces tumorigenesis by activating the oncogenic protein GLUT1 expression and repressing the tumor suppressor BIM expression by elevating levels of HIF1α, in a manner dependent on the hly-encoded alpha hemolysin. The carcinogenic capacity of hemolytic type I E. coli and its requirement for alpha hemolysin were verified in female mice.

pks+E. coli have been reported to be associated with colorectal cancer (4, 5). In this study we found that pks+E. coli sometimes overlap with hly+ type I E. coli, which is frequently observed in male subjects and in female colorectal cancer patients but relatively rare in healthy female subjects or females with adenoma. Although pks+E. coli partially overlap with hly+ type I E. coli, their roles for colorectal cancer and adenoma are different in at least two aspects. First, hly+ type I E. coli correlates with shorter survival in female patients while pks+ type I E. coli does not; second, hly+ type I E. coli is associated with both adenoma and colorectal cancer, whereas pks+ type I E. coli is associated with colorectal cancer but not with adenoma in females. The link between hly+ type I E. coli and adenoma suggests that these bacteria pre-exist the onset of colorectal cancer and supports our in vitro and in vivo evidence that they are the cause but not the consequence of colorectal cancer. As it takes a long time for adenoma to develop into colorectal cancer, the adenoma specificity of hly+ type I E. coli suggest that these bacteria can be used as potential antecedent biomarkers for colorectal cancer.

Alpha hemolysin binds nonspecifically to host cells and triggers cellular reactions without the need for a receptor (33). We speculate that hemolytic type I E. coli elevates levels of HIF1α by indirect mechanisms. Alpha hemolysin of E. coli is a pore-forming toxin, generating small cation-permeable channels in the host cell membrane. Alpha hemolysin enables hemolytic E. coli to induce “focal leaks” both in colonic cells and tissues, within which hemolytic E. coli accumulates (34, 35). E. coli is a group of facultative anaerobic bacteria, making ATP by aerobic respiration in the presence of oxygen. Thus, the hemolytic E. coli trapped in colonic tissues would deplete oxygen, generating a hypoxic microenvironment, which is known to stabilize HIF1α and increase HIF1α activity (36). Therefore, hemolytic type I E. coli mediates activation of HIF1α and consequently induces colonic tumorigenesis at least in part by generating focal hypoxia. Although many intestinal bacteria consume oxygen, most of them are incapable of making pores in the colonic tissue and generating a hypoxic microenvironment surrounding colonic cells. Thus, the hypoxic mechanisms could be specific to these “pore-making” hemolytic bacteria, which are primarily hemolytic type I E. coli in the colon. By creating channels in the host cell membrane, hemolytic E. coli also cause ion oscillations such as elevation of intracellular calcium (33, 37), which in turn could lead to calcium-dependent HIF1α activation (38). Thus, it is speculated that hemolytic E. coli regulates HIF1α levels and activity probably through multiple pathways.

HIF1 regulates diverse oncogenic pathways (31), and accordingly hemolytic E. coli could regulate a quite number of downstream genes through HIF1. We showcased that the HIF1-activated GLUT1 and HIF1-repressed BIM were induced and repressed in vivo by hemolytic E. coli, respectively. Elevated GLUT1 expression and glucose uptake are features of oncogenic transformation of mammalian cells. GLUT1 is also associated with poor survival of cancer patients (26, 27, 29). Therefore, not only the carcinogenic capability of hemolytic E. coli but also the poor survival in female colorectal cancer patients carrying this type of E. coli could be partially attributed to the HIF1-mediated activation of GLUT1.

Despite the requirement of alpha hemolysin for tumorigenesis mediated by this type of E. coli, alpha hemolysin is also cytotoxic to colon epithelial cells. It is a paradox that the hemolysin acts as a carcinogenic factor but is meanwhile cytotoxic and apoptotic to both normal and cancerous cells. Unlike in vitro conditions, under which cultured cells are quickly killed by alpha hemolysin, in vivo environments support continuous repairment of colonic tissues so that alpha hemolysin is carcinogenic to colonic tissues but is not as potent as to inhibit tumor development. These may explain why the net outcome of hly+ type I E. coli infection is increased colorectal cancer carcinogenesis. In addition, in vitro expression regulation may sometimes fail to mimic in vivo regulation. Indeed, our data showed that the tumor suppressor BIM was induced by hly+ type I E. coli in vitro but repressed by this bacterium in vivo.

Another interesting finding is that hemolytic type I E. coli is specifically associated with adenoma and colorectal cancer in females. It is beyond the scope of this study but warrants further investigation why females are more vulnerable to colorectal cancer carcinogenesis mediated by hemolytic type I E. coli. Gender-dependent differences have been known to exist in the etiology and pathogenesis of a long list of diseases such as heart failure (39), viral infections (40), and Alzheimer disease (41). Mechanisms underlying the gender difference remain poorly understood and may involve complex processes. In the case of the gender specificity of hemolytic E. coli-mediated colonic tumorigenesis, the genetic factor is ruled out because our in vitro data showed that hemolytic E. coli had cytotoxic and tumorigenic effects on cell lines that are derived from men. It is therefore speculated that the gender specificity observed in this study might be associated with immunological, hormonal, or behavioral factors that have been linked to gender difference in disease development (40–43). Our animal experiments showed that hemolytic E. coli significantly induced colonic tumorigenesis in female mice but failed to do so in male mice, confirming the gender specificity of hemolytic E. coli-mediated colonic tumorigenesis. It is, however, noteworthy that gender difference in mice does not fully mimic that in humans, as the latter is further complicated by social factors.

No potential conflicts of interest were disclosed.

Conception and design: Y. Jin, J.J.Y. Sung, J. Yu

Development of methodology: Y. Jin, J. Yu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Jin, W. Li, S.C. Ng, M.W.Y. Chan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Jin, S. Tang, M. Chan

Writing, review, and/or revision of the manuscript: Y. Jin, S.C. Ng, J.J.Y. Sung, J. Yu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Jin, W. Li, J. Yu

Study supervision: Y. Jin, S.C. Ng, J. Yu

This project was supported by 863 Program China (2012AA02A506 to J. Yu), 973 Program China (2013CB531401 to J. Yu), National Natural Science Foundation of China (NSFC) for Young Scientists (C010301 to Y. Jin), Shenzhen Technology and Innovation Project Fund (JSGG20130412171021059 to J. Yu), Shenzhen Virtual University Park Support Scheme to CUHK Shenzhen Research Institute (J. Yu), and 973 Program China (2014CB745200 to Y. Jin).

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.

1.
Kaper
JB
,
Nataro
JP
,
Mobley
HL
. 
Pathogenic Escherichia coli
.
Nat Rev Microbiol
2004
;
2
:
123
40
.
2.
Maddocks
OD
,
Short
AJ
,
Donnenberg
MS
,
Bader
S
,
Harrison
DJ
. 
Attaching and effacing Escherichia coli downregulate DNA mismatch repair protein invitro and are associated with colorectal adenocarcinomas in humans
.
PLoS One
2009
;
4
:
e5517
.
3.
Choi
HJ
,
Kim
J
,
Do
KH
,
Park
SH
,
Moon
Y
. 
Enteropathogenic Escherichia coli-induced macrophage inhibitory cytokine 1 mediates cancer cell survival: an invitro implication of infection-linked tumor dissemination
.
Oncogene
2013
;
32
:
4960
9
.
4.
Nougayrede
JP
,
Homburg
S
,
Taieb
F
,
Boury
M
,
Brzuszkiewicz
E
,
Gottschalk
G
, et al
Escherichia coli induces DNA double-strand breaks in eukaryotic cells
.
Science
2006
;
313
:
848
51
.
5.
Arthur
JC
,
Perez-Chanona
E
,
Muhlbauer
M
,
Tomkovich
S
,
Uronis
JM
,
Fan
TJ
, et al
Intestinal inflammation targets cancer-inducing activity of the microbiota
.
Science
2012
;
338
:
120
3
.
6.
Dreux
N
,
Denizot
J
,
Martinez-Medina
M
,
Mellmann
A
,
Billig
M
,
Kisiela
D
, et al
Point mutations in FimH adhesin of Crohn's disease-associated adherent-invasive Escherichia coli enhance intestinal inflammatory response
.
PLoS Pathogens
2013
;
9
:
e1003141
.
7.
Li
H
,
Durbin
R
. 
Fast and accurate long-read alignment with Burrows-Wheeler transform
.
Bioinformatics
2010
;
26
:
589
95
.
8.
Gaujoux
R
,
Seoighe
C
. 
A flexible R package for nonnegative matrix factorization
.
BMC Bioinformatics
2010
;
11
:
367
.
9.
Schneider
TD
,
Stephens
RM
. 
Sequence logos: a new way to display consensus sequences
.
Nucleic Acids Res
1990
;
18
:
6097
100
.
10.
Crooks
GE
,
Hon
G
,
Chandonia
JM
,
Brenner
SE
. 
WebLogo: a sequence logo generator
.
Genome Res
2004
;
14
:
1188
90
.
11.
Jin
Y
,
Wu
J
,
Li
Y
,
Cai
Z
,
Huang
JD
. 
Modification of the RpoS network with a synthetic small RNA
.
Nucleic Acids Res
2013
;
41
:
8332
40
.
12.
Bej
AK
,
Steffan
RJ
,
DiCesare
J
,
Haff
L
,
Atlas
RM
. 
Detection of coliform bacteria in water by polymerase chain reaction and gene probes
.
Appl Environ Microbiol
1990
;
56
:
307
14
.
13.
Thaker
AI
,
Shaker
A
,
Rao
MS
,
Ciorba
MA
. 
Modeling colitis-associated cancer with azoxymethane (AOM) and dextran sulfate sodium (DSS)
.
J Vis Exp
2012
;
11
:
pii:4100
.
14.
Rakoff-Nahoum
S
,
Paglino
J
,
Eslami-Varzaneh
F
,
Edberg
S
,
Medzhitov
R
. 
Recognition of commensal microflora by toll-like receptors is required for intestinal homeostasis
.
Cell
2004
;
118
:
229
41
.
15.
Erben
U
,
Loddenkemper
C
,
Doerfel
K
,
Spieckermann
S
,
Haller
D
,
Heimesaat
MM
, et al
A guide to histomorphological evaluation of intestinal inflammation in mouse models
.
Int J Clin Exp Pathol
2014
;
7
:
4557
76
.
16.
Rubinstein
MR
,
Wang
X
,
Liu
W
,
Hao
Y
,
Cai
G
,
Han
YW
. 
Fusobacterium nucleatum promotes colorectal carcinogenesis by modulating E-cadherin/beta-catenin signaling via its FadA adhesin
.
Cell Host Microbe
2013
;
14
:
195
206
.
17.
Gur
C
,
Coppenhagen-Glazer
S
,
Rosenberg
S
,
Yamin
R
,
Enk
J
,
Glasner
A
, et al
Natural killer cell-mediated host defense against uropathogenic E. coli is counteracted by bacterial hemolysinA-dependent killing of NK cells
.
Cell Host Microbe
2013
;
14
:
664
74
.
18.
Mansson
LE
,
Kjall
P
,
Pellett
S
,
Nagy
G
,
Welch
RA
,
Backhed
F
, et al
Role of the lipopolysaccharide-CD14 complex for the activity of hemolysin from uropathogenic Escherichia coli
.
Infect Immun
2007
;
75
:
997
1004
.
19.
Vineis
P
,
Alavanja
M
,
Buffler
P
,
Fontham
E
,
Franceschi
S
,
Gao
YT
, et al
Tobacco and cancer: recent epidemiological evidence
.
J Natl Cancer Inst
2004
;
96
:
99
106
.
20.
Krogfelt
KA
,
Bergmans
H
,
Klemm
P
. 
Direct evidence that the FimH protein is the mannose-specific adhesin of Escherichiacoli type 1 fimbriae
.
Infect Immun
1990
;
58
:
1995
8
.
21.
Wiener
Z
,
Band
AM
,
Kallio
P
,
Hogstrom
J
,
Hyvonen
V
,
Kaijalainen
S
, et al
Oncogenic mutations in intestinal adenomas regulate Bim-mediated apoptosis induced by TGF-beta
.
Proc Natl Acad Sci U S A
2014
;
111
:
E2229
36
.
22.
Wolf
MJ
,
Hoos
A
,
Bauer
J
,
Boettcher
S
,
Knust
M
,
Weber
A
, et al
Endothelial CCR2 signaling induced by colon carcinoma cells enables extravasation via the JAK2-Stat5 and p38MAPK pathway
.
Cancer Cell
2012
;
22
:
91
105
.
23.
Baier
PK
,
Eggstein
S
,
Wolff-Vorbeck
G
,
Baumgartner
U
,
Hopt
UT
. 
Chemokines in human colorectal carcinoma
.
Anticancer Res
2005
;
25
:
3581
4
.
24.
Cui
YM
,
Jiao
HL
,
Ye
YP
,
Chen
CM
,
Wang
JX
,
Tang
N
, et al
FOXC2 promotes colorectal cancer metastasis by directly targeting MET
.
Oncogene
2015
;
34
:
4379
90
.
25.
Watanabe
T
,
Kobunai
T
,
Yamamoto
Y
,
Matsuda
K
,
Ishihara
S
,
Nozawa
K
, et al
Gene expression of mesenchyme forkhead 1 (FOXC2) significantly correlates with the degree of lymph node metastasis in colorectal cancer
.
Int Surg
2011
;
96
:
207
16
.
26.
Saigusa
S
,
Toiyama
Y
,
Tanaka
K
,
Okugawa
Y
,
Fujikawa
H
,
Matsushita
K
, et al
Prognostic significance of glucose transporter-1 (GLUT1) gene expression in rectal cancer after preoperative chemoradiotherapy
.
Surg Today
2012
;
42
:
460
9
.
27.
Haber
RS
,
Rathan
A
,
Weiser
KR
,
Pritsker
A
,
Itzkowitz
SH
,
Bodian
C
, et al
GLUT1 glucose transporter expression in colorectal carcinoma: a marker for poor prognosis
.
Cancer
1998
;
83
:
34
40
.
28.
Chen
C
,
Pore
N
,
Behrooz
A
,
Ismail-Beigi
F
,
Maity
A
. 
Regulation of glut1 mRNA by hypoxia-inducible factor-1. Interaction between H-ras and hypoxia
.
J Biol Chem
2001
;
276
:
9519
25
.
29.
Macheda
ML
,
Rogers
S
,
Best
JD
. 
Molecular and cellular regulation of glucose transporter (GLUT) proteins in cancer
.
J Cell Physiol
2005
;
202
:
654
62
.
30.
Whelan
KA
,
Caldwell
SA
,
Shahriari
KS
,
Jackson
SR
,
Franchetti
LD
,
Johannes
GJ
, et al
Hypoxia suppression of Bim and Bmf blocks anoikis and luminal clearing during mammary morphogenesis
.
Mol Biol Cell
2010
;
21
:
3829
37
.
31.
Benita
Y
,
Kikuchi
H
,
Smith
AD
,
Zhang
MQ
,
Chung
DC
,
Xavier
RJ
. 
An integrative genomics approach identifies Hypoxia Inducible Factor-1 (HIF-1)-target genes that form the core response to hypoxia
.
Nucleic Acids Res
2009
;
37
:
4587
602
.
32.
Greenhough
A
,
Wallam
CA
,
Hicks
DJ
,
Moorghen
M
,
Williams
AC
,
Paraskeva
C
. 
The proapoptotic BH3-only protein Bim is downregulated in a subset of colorectal cancers and is repressed by antiapoptotic COX-2/PGE(2) signalling in colorectal adenoma cells
.
Oncogene
2010
;
29
:
3398
410
.
33.
Valeva
A
,
Walev
I
,
Kemmer
H
,
Weis
S
,
Siegel
I
,
Boukhallouk
F
, et al
Binding of Escherichia coli hemolysin and activation of the target cells is not receptor-dependent
.
J Biol Chem
2005
;
280
:
36657
63
.
34.
Troeger
H
,
Richter
JF
,
Beutin
L
,
Gunzel
D
,
Dobrindt
U
,
Epple
HJ
, et al
Escherichia coli alpha-haemolysin induces focal leaks in colonic epithelium: a novel mechanism of bacterial translocation
.
Cell Microbiol
2007
;
9
:
2530
40
.
35.
Bucker
R
,
Schulz
E
,
Gunzel
D
,
Bojarski
C
,
Lee
IF
,
John
LJ
, et al
alpha-Haemolysin of Escherichia coli in IBD: a potentiator of inflammatory activity in the colon
.
Gut
2014
;
63
:
1893
901
.
36.
Lee
JW
,
Bae
SH
,
Jeong
JW
,
Kim
SH
,
Kim
KW
. 
Hypoxia-inducible factor (HIF-1)alpha: its protein stability and biological functions
.
Exp Mol Med
2004
;
36
:
1
12
.
37.
Uhlen
P
,
Laestadius
A
,
Jahnukainen
T
,
Soderblom
T
,
Backhed
F
,
Celsi
G
, et al
Alpha-haemolysin of uropathogenic E. coli induces Ca2+ oscillations in renal epithelial cells
.
Nature
2000
;
405
:
694
7
.
38.
Yuan
G
,
Nanduri
J
,
Bhasker
CR
,
Semenza
GL
,
Prabhakar
NR
. 
Ca2+/calmodulin kinase-dependent activation of hypoxia inducible factor 1 transcriptional activity in cells subjected to intermittent hypoxia
.
J Biol Chem
2005
;
280
:
4321
8
.
39.
Azad
N
,
Kathiravelu
A
,
Minoosepeher
S
,
Hebert
P
,
Fergusson
D
. 
Gender differences in the etiology of heart failure: a systematic review
.
J Geriat Card
2011
;
8
:
15
23
.
40.
Klein
SL
,
Hodgson
A
,
Robinson
DP
. 
Mechanisms of sex disparities in influenza pathogenesis
.
J Leukoc Biol
2012
;
92
:
67
73
.
41.
Webber
KM
,
Perry
G
,
Smith
MA
,
Casadesus
G
. 
The contribution of luteinizing hormone to Alzheimer disease pathogenesis
.
Clin Med Res
2007
;
5
:
177
83
.
42.
Roberts
CW
,
Walker
W
,
Alexander
J
. 
Sex-associated hormones and immunity to protozoan parasites
.
Clin Microbiol Rev
2001
;
14
:
476
88
.
43.
Barrat
F
,
Lesourd
B
,
Boulouis
HJ
,
Thibault
D
,
Vincent-Naulleau
S
,
Gjata
B
, et al
Sex and parity modulate cytokine production during murine ageing
.
Clin Exp Immunol
1997
;
109
:
562
8
.

Supplementary data