Association between antibiotic use during early life and early‐onset colorectal cancer risk overall and according to polygenic risk and FUT2 genotypes

Abstract Early‐onset colorectal cancer (EOCRC) has been increasing worldwide. Potential risk factors may have occurred in childhood or adolescence. We investigated the associations between early‐life factors and EOCRC risk, with a particular focus on long‐term or recurrent antibiotic use (LRAU) and its interaction with genetic factors. Data on the UK Biobank participants recruited between 2006 and 2010 and followed up to February 2022 were used. We used logistic regression to estimate adjusted odds ratios (ORs) and 95% confidence intervals (95% CIs) of the associations between LRAU during early life and EOCRC risk overall and by polygenic risk score (constructed by 127 CRC‐related genetic variants) and Fucosyltransferase 2 (FUT2), a gut microbiota regulatory gene. We also assessed the associations for early‐onset colorectal adenomas, as precursor lesion of CRC, to examine the effect of LRAU during early‐life and genetic factors on colorectal carcinogenesis. A total of 113 256 participants were included in the analysis, with 165 EOCRC cases and 719 EOCRA cases. LRAU was nominally associated with increased risk of early‐onset CRC (OR = 1.48, 95% CI = 1.01‐2.17, P = .046) and adenomas (OR = 1.40, 95% CI = 1.17‐1.68, P < .001). When stratified by genetic polymorphisms of FUT2, LRAU appeared to confer a comparatively greater risk for early‐onset adenomas among participants with rs281377 TT genotype (OR = 1.10, 95% CI = 0.79‐1.52, P = .587, for CC genotype; OR = 1.75, 95% CI = 1.16‐2.64, P = .008, for TT genotype; P interaction = .089). Our study suggested that LRAU during early life is associated with increased risk of early‐onset CRC and adenomas, and the association for adenomas is predominant among individuals with rs281377 TT/CT genotype. Further studies investigating how LRAU contributes together with genetic factors to modify EOCRC risk, particularly concerning the microbiome‐related pathway underlying colorectal carcinogenesis, are warranted.

[Correction added after first online publication on 09 August 2023: There was an error in the order of affiliations and has been corrected in this version.] LRAU during early life and EOCRC risk overall and by polygenic risk score (constructed by 127 CRC-related genetic variants) and Fucosyltransferase 2 (FUT2), a gut microbiota regulatory gene. We also assessed the associations for early-onset colorectal adenomas, as precursor lesion of CRC, to examine the effect of LRAU during early-life and genetic factors on colorectal carcinogenesis. A total of 113 256 participants were included in the analysis, with 165 EOCRC cases and 719 EOCRA cases. LRAU was nominally associated with increased risk of early-onset CRC (OR = 1.48, 95% CI = 1.01-2.17, P = .046) and adenomas (OR = 1.40, 95% CI = 1.17-1.68, P < .001).
When stratified by genetic polymorphisms of FUT2, LRAU appeared to confer a comparatively greater risk for early-onset adenomas among participants with rs281377 TT genotype (OR = 1.10, 95% CI = 0.79-1.52, P = .587, for CC genotype; OR = 1.75, 95% CI = 1.16-2.64, P = .008, for TT genotype; P interaction = .089). Our study suggested that LRAU during early life is associated with increased risk of early-onset CRC and adenomas, and the association for adenomas is predominant among individuals with rs281377 TT/CT genotype. Further studies investigating how LRAU contributes together with genetic factors to modify EOCRC risk, particularly concerning the microbiome-related pathway underlying colorectal carcinogenesis, are warranted.

K E Y W O R D S
antibiotic use, early-life factors, early-onset colorectal cancer, FUT2 gene, polygenic risk

What's new?
Early-onset colorectal cancer is on the rise worldwide, possibly driven by exposure to risk factors in childhood. Long-term or recurrent use of antibiotics, for instance, could cause long-lasting changes in the gut microbiota that lead to cancer susceptibility. Here, the authors investigated the association between antibiotics use and cancer risk, both overall and according to certain genetic factors. They found that long-term or recurrent antibiotic use increased the risk of earlyonset CRC and adenomas, with a stronger effect in people with a particular variant of the gut microbiota regulatory gene, Fucosyltransferase 2 (FUT2).

| INTRODUCTION
The incidence of colorectal cancer (CRC) in populations older than 50 years is decreasing in many countries (eg, The United States, The United Kingdom), which is largely attributed to the implementation of effective population-based screening for early detection and removal of precursors. 1,2 However, early-onset CRC (EOCRC; diagnosed before 50 years) is increasing worldwide over the past several decades. 3,4 Given the increasing incidence of CRC among younger adults, investigating risk factors for EOCRC is important for enhancing and tailoring primary prevention in this population of special interest.
Accumulating evidence suggests that some potential risk factors for EOCRC may be operative during childhood or adolescence. 5,6 As accessibility to antibiotics increases across low and middle-income countries, antibiotic therapy for common childhood infections or adolescent acne is increasingly widespread. 7,8 However, childhood or adolescence may be more vulnerable periods to the potential effects of overexposure. 5,6 Epidemiology studies have observed significant correlations between long-term use of antibiotic and risk of colorectal neoplasm. [9][10][11] Studies investigating potentially modifiable risk factors for EOCRC have, however, focused on lifestyle factors and exposures occurring during adolescence and/or early adulthood and found associations with several factors such as obesity 12,13 and alcohol consumption. 14 There is thus limited evidence evaluating the effects of LRAU during early life on the risk of early-onset colorectal neoplasm.
Recent genome-wide association studies (GWAS) have expanded the catalog of CRC risk-related single nucleotide polymorphisms (SNPs). 15,16 Polygenic risk score (PRS) combines the effect of these SNPs and provides powerful tools for risk stratification and prediction of CRC risk, especially for EOCRC. 17 It has been suggested that lifestyle factors could interact with PRS to affect EOCRC risk. 18 However, to our knowledge, no previous studies have assessed such potential interactions between early-life exposures and the genetic factors in relation to EOCRC risk.
Early-life events could influence gut microbiota, which have been proposed to play an important role in the initiation and development of CRC. 5,6 LRAU during early life may thus cause long-standing changes in gut microbiota, and irregularities in immunostimulatory bacterial products that can impede normal immune surveillance, increasing CRC risk. 19 The Fucosyltransferase 2 (FUT2) is a gut microbiota regulatory gene and plays an essential role in regulating gut microbiota composition. 20,21 The dysregulation of gut microbiota could be at the crossroads of the above-mentioned risk factors contributing to EOCRC. It could thus be hypothesized that factors that impact gut microbiota content (eg, antibiotic use and genetic polymorphisms of FUT2) might interact together to modify EOCRC risk, but this important question has not been investigated either.
Herein, we used data from the UK Biobank (a) to investigate the associations of LRAU during early life with EOCRC risk overall and according to genetic factors, and (b) to test whether the association between LRAU and risk of EOCRC as well as adenomas differs by the genetic polymorphisms of FUT2.

| Study population
The UK Biobank is a large prospective cohort study with comprehensive health-related phenotypic and genotypic information. It comprises over 500 000 participants, aged 40 to 69 years, recruited from across 22 centers located throughout England, Wales and Scotland between 2006 and 2010. Details of the baseline information, biological samples collection and genome-wide genotyping have been described elsewhere. 22

| Early-life factors and covariates
LRAU during early life was defined as long-term or recurrent use of antibiotics during childhood and/or adolescence. In brief, the participants were asked whether they received long-term or recurrent courses of antibiotics (3+ per year, eg, for acne), and the information was binary (yes/no). The UK Biobank obtained this early-life LRAU data from 174 714 participants, utilizing a digestive health questionnaire disseminated to a subset of the cohort ($335 000 individuals) via email or through the participant website. Of this group, $52% (176 345 participants) completed the questionnaire. In addition to LRAU during early life, we also investigated other early-life factors, encompassing postnatal factors and age at onset of secondary sexual characteristics (more details in Table S1). Information on these early life factors was assessed via the baseline touchscreen questionnaire, verbal interview, or follow-up questionnaire. Information on covariates, including age (continuous in years), sex (men and women), education attainment (college or university degree and above, and high school and below), was assessed in the baseline survey. History of childhood diabetes was ascertained via linkage to the main primary care computer system suppliers in England as well as the hospital inpatient data. Missing data was $1% for education level and 20% for family history of CRC. To avoid a large amount of data loss due to missing information on a particular covariate (eg, family history of CRC), we employed multiple imputation for the missing categorical covariates.

| Genetic susceptibility
To expand the catalog of CRC risk loci, Huyghe et al 15 and Law et al 16 have conducted latest GWAS meta-analyses respectively based on 125 478 and 106 006 individuals of European ancestry and identified 127 CRC risk-related SNPs (P < 5 Â 10 À8 , Table S2). All risk variants were free of linkage disequilibrium (LD, r 2 > .001) and were available in the UK Biobank. We constructed PRS based on the 127 SNPs by summing the product of the number of CRC risk alleles (0, 1 and 2) and corresponding effect size on CRC genetic liability for each risk variant. The CRC PRS was dichotomized as low (≤ the median PRS of non-CRC cases) and high (above the median PRS of the non-CRC cases) to assess the genetic risk for early-onset CRC and its precursory lesion of adenomas. 17 Three SNPs, including rs35866622, rs601338 and rs281377, which are different from the above 127 CRC risk variants, have been shown to be genetic polymorphisms of FUT2 gene that regulates gut microbiome. 27 We obtained the genotyping data of these SNPs in the UK Biobank to test whether the association between LRAU and the risk of EOCRC and its precursory lesion of adenomas is modified by the genetic polymorphisms of FUT2 gene.

| Statistical analysis
The distribution of baseline characteristics of the study population according to colorectal neoplasm status was evaluated in descriptive analyses and differences were tested for statistical significance using the Pearson χ 2 test for categorical variables (eg, sex) and the t-test for continuous variables (eg, age at recruitment). We used the logistic regression to estimate adjusted odds ratios (ORs) and 95% confidence intervals (95% CIs) for the association of LRAU and other early life factors (Tables S3 and S4) with EOCRC risk. Two levels of statistical adjustments were applied: model 1 was adjusted for age at recruitment, sex, PRS and the first five principal components, and model 2 was additionally adjusted for education level, family history of CRC and childhood diabetes. We also investigated whether and to what extent the association between LRAU and EOCRC risk differs by PRS levels (low/high), sex (male/female), family history of CRC (yes/no) and tumor subsites (proximal colon, distal colon, rectum) (Table S5). Here, we used a likelihood ratio test to calculate P values for interactions between early-life factors and the genetic factors.
Moreover, the joint association of LRAU and PRS levels with EOCRC risk was evaluated (Table S6). Furthermore, we conducted gene-antibiotic interaction analysis to test whether the association between LRAU and EOCRC risk varies by the expression of FUT2 gene.
For the genetic polymorphisms of FUT2 that were marginally associated with EOCRC risk (ie, rs281377), subgroup analyses were additionally performed based on the rs281377 genotype to assess potential effect modification of the association between LRAU and EOCRC risk by this factor. Also, we estimated the cumulative incidence of EOCRC according to LRAU status by the genetic polymorphisms of FUT2 gene. Additionally, we used the above analysis to further explore the relationships among LRAU, genetic factors and risk of early-onset CRC precursors.
We conducted sensitivity analysis to test the robustness of the associations by (a) using multivariable Cox proportional hazards regression to validate the associations regarding LRAU (detail in Supplementary Methods in Data S1, Tables S7 and S8); (b) using 55 years as the age cutoff to define EOCRC (Tables S9 and S10). All analyses were performed in the R software (version 4.1.1), and all statistical tests were two-sided. Bonferroni correction was applied to account for the multiple testing, where P-value <.005 was regarded as statistically significant and P-value <.05 was regarded as nominally significant.  Table 1. Overall, early-onset CRC and adenomas cases were more likely to be men and have a family history of CRC and a history of LRAU.   (Tables S3 and S4).
For the precursor lesion of CRC, LRAU was significantly associated with increased risk of EOCRA (OR = 1.40, 95% CI = 1.17-1.68, P < .001 for incident and prevalent cases combined, Table 2), regardless of genetic liability (OR = 1.41, 95% CI = 1.05-1.90, P = .021 for low and OR = 1.39, 95% CI = 1.11-1.74, P = .005 for high PRS group, Table 3). Again, the association between LRAU and EOCRA risk was not statistically significant for incident cases only (OR = 1.15, 95% CI = 0.84-1.57, P = .402). Sensitivity analysis using multivariable Cox proportional hazards regression showed very similar associations between LRAU and risk of EOCRC and EOCRA overall and by genetic risk (Tables S7 and S8). In sensitivity analysis where the initial follow-up age of EOCRC was set to 19 years in the LRAU-related analysis, similar results as those reported for the main analysis were observed (Supplementary Methods in Data S1, data not shown).

| Gene-antibiotic interactions for EOCRC risk
In analysis of the three genetic polymorphisms of FUT2 gene (ie, rs35866622, rs601338 and rs281337) in their associations with EOCRC risk, we observed no associations of rs35866622 and rs601338 with EOCRC risk, but a marginal association for rs281377 (OR = 1.23, 95% CI = 0.99-1.54, P = .068) (Table S11). Risk estimates appeared much greater for LRAU with EOCRC risk in participants with rs281377 TT genotype (OR = 2.73, 95% CI = 0.83-8.95, P = .098) than those with CT and TT genotype (Table 4) Table 4 shows the adjusted ORs for the association between LRAU and risk of EOCRC as well as EOCRA, respectively, stratified by the rs281377 genotypes of FUT2.

| DISCUSSION
Given that CRC takes several decades to develop, investigating earlylife factors as potential risk factors for EOCRC is important. We investigated the association between LRAU during early life as well as its interactions and joint associations with genetic risk factors. Results suggested that LRAU was associated with increased risk of both EOCRC and EOCRA. In participants with high PRS and family history of CRC, LRAU was associated with 72% and 134% higher risk of EOCRC, respectively. In addition, when stratified by genetic polymorphisms of FUT2, LRAU appeared to confer a comparatively greater risk for EOCRA among participants with rs281377 TT/CT genotype.
The observed higher EOCRC risk among those with a history of LRAU during childhood lends credence to the findings of a case-control study that showed that antibiotic exposure was strongly associated with EOCRC risk, whereas a modest association was observed with LOCRC risk. 11 However, the authors did not assess the adenomas are asymptomatic and only detected by screening, the associations between exposures and adenomas can be confounded by colonoscopy. 28 However, antibiotics use appeared to be independent of colonoscopy or related symptoms. Notably, LRAU appeared to increase risk of CRC precursors regardless of genetic liability, whereas the increased EOCRC risk was restricted to high PRS group. The latter probably suggested that inadequate power due to a few cases might lead to difference in LRAU with EOCRC risk by PRS levels.
Genetic factors play an important role in risk of early-onset colorectal neoplasm, and it has been shown that the association between genetic factor (eg, PRS) and CRC risk is particularly strong for EOCRC. 17 However, genetic factors are not modifiable and there is limited evidence supporting differential CRC screening among younger adults with high genetic risk such as those with a family history of CRC. 29 Hence, identification of modifiable factors that independently interact with genetic factors to affect EOCRC risk is highly rele- Fusobacterium load, cancer cell proliferation and overall tumor growth of mouse with xenograft. 32 Therefore, influence of antibiotics on colorectal carcinogenesis is probably bidirectional and multiple potential mechanisms remain to be explored. Our findings, however, indicated that novel strategies are still expected to substitute or complement antibiotic therapies, so that pathogens can be targeted selectively without perturbing the microbiota and the beneficial effects they confer to the host. 33 Notably, pathogens necessitating antibiotics use may induce inflammation, a well-established risk factor for CRC, and may mediate the observed association between antibiotics use and EOCRC risk. 34 Besides, a meta-analysis of epidemiological studies involving participants of all ages, for whom the majority are usually older than 60 years, have demonstrated that the association between LRAU and CRC risk is restricted to broad-spectrum antibiotics. 35 20 The FUT2 gene encodes the galactoside 2-α-lfucosyltransferase 2 enzyme and thus influences several types of hostmicrobial interactions. 21 In-vivo studies have also shown that gut microbiota in FUT2-deficient mice is altered structurally and functionally (eg, increased proportional representation of several genera of Bacteroidetes and decreased alpha diversity). 21,43,44 This increases the generation of lysophosphatidylcholine, which in turn promotes inflammation and epithelial barrier damage. 45 Our study has several strengths.