Circulating amino acid levels and colorectal cancer risk in the European Prospective Investigation into Cancer and Nutrition and UK Biobank cohorts

Colorectal cancer is the third most common cancer globally, with around 1.9 million cases diagnosed in 2020, and the second most common cause of cancer-related death [1]. There is great potential to reduce this burden since most colorectal cancer cases are sporadic [2] and are associated with modifiable risk factors such as body fatness [3], alcohol intake [4], and diet [5]. Colorectal cancer development is also influenced by metabolic factors [6, 7]. For example, insulin and insulin-like growth factors are thought to play causal roles in colorectal tumorigenesis [8], likely through the promotion of cell proliferation and growth signaling pathways [9]. Broad metabolic dysfunction may lead to perturbed small-molecule metabolism, which in turn elicits bioactivity at the level of tissues and organs.

Amino acids are among the most abundant circulating metabolites and serve as building blocks of proteins, precursors of many signaling molecules, and an important energy source via the citric acid cycle. Certain amino acids may also fuel cancer development [10], and marked changes in blood amino acid concentrations have been extensively observed in colorectal cancer patients [11]. For example, levels of amino acids such as glutamine, citrulline, alanine, and histidine have been inversely associated with advancing disease stage [12, 13], while valine and leucine were among the metabolites that distinguished colorectal cancer cases using a discovery-replication strategy [14]. Similarly, the concentrations of several blood amino acids distinguished early-stage colorectal cancer cases from controls in Japanese patients, most notably aspartic acid [15], as well as ornithine and lysine [16]. Glutamine was a notable discriminant in patients newly diagnosed with colorectal cancer compared to controls in a Chinese hospital-based study [17]. Overall, amino acid levels were generally inversely associated with prevalent colorectal neoplasia, suggesting a depletion of serological concentrations in cases compared to healthy individuals. Amino acid profiling could therefore potentially help identify early-stage disease [18], as well as providing insights into mechanisms of carcinogenesis.

Despite these observations, few prospective studies have been conducted to test the hypothesis that pre-diagnostic amino acid concentrations are associated with colorectal cancer risk. Two such studies of nested case-control design that analyzed pre-diagnostic serum or plasma by untargeted metabolomics found limited dysregulation of lipophilic metabolites only [19, 20], while in a case-control study nested in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort that measured tryptophan and serotonin levels, tryptophan was inversely associated with colon cancer [21]. The aim of the current study was thus to test these associations in a larger and more comprehensive analysis. We first employed the EPIC nested case-control study as a discovery cohort, which measured between 13 and 21 amino acids in fasting plasma or serum in relation to colorectal cancer. In a replication step, we tested those amino acids associated with colorectal cancer risk in EPIC in the UK Biobank cohort, in which 9 overlapping compounds had been measured in over 111,000 participants. Together, the two cohorts allow for the largest and most detailed investigation of circulating amino acids and colorectal cancer risk performed to date.

A median follow-up of 14.4 years was observed for the 654 colorectal cancer cases and 654 controls in EPIC while, during a median follow-up of 10.7 years in the UK Biobank, 1221 incident cases of colorectal cancer occurred among the 111,323 participants with available amino acid measurements. The EPIC and UK Biobank populations were of similar ages at baseline and at colorectal cancer diagnosis although, in EPIC, most participants (77.4%) were from Italian or Spanish centers. Full baseline characteristics are shown in Table 1. Glutamine, alanine, and glycine were at the highest circulating concentrations overall, as quantified in EPIC (Fig. 2). Fasting concentrations of amino acids in cancer-free participants were almost always positively correlated, with the following correlated clusters noted in EPIC: glycine and serine; arginine, methionine, and tryptophan; valine, isoleucine, and leucine; and histidine and phenylalanine (Fig. 3). In the UK Biobank, valine, isoleucine, and leucine concentrations (branched-chain amino acids) were strongly intercorrelated.

In this analysis of pre-diagnostic circulating amino acid levels and colorectal cancer risk, histidine was found to be robustly inversely associated and glutamine borderline inversely associated with colorectal cancer risk via a discovery-replication strategy in two large prospective cohorts. In addition, odds ratios and hazards ratios for these amino acids were attenuated minimally by the exclusion of cases diagnosed within 10 years of follow-up. This study provides strong evidence that lower levels of histidine, and possibly glutamine, are associated with subsequent risk of colorectal cancer, even up to 10 years before a colorectal cancer diagnosis.

Circulating levels of several amino acids have previously been found to be inversely associated with colorectal neoplasia, but in studies of cross-sectional design only. Glutamine, for example, was one of several amino acids found to be lower in colorectal cancer patients compared to healthy controls [28], while histidine was lower among stage IV colorectal cancer cases than stage I cases [12] and even inversely correlated with tumor stage [29]. Untargeted metabolomics studies using discovery and validation cohorts demonstrated leucine and the dipeptide glutamine-leucine to be among those metabolites that distinguished cases from controls [14, 30]. Nevertheless, few studies have analyzed pre-diagnostic samples to investigate whether amino acid dysregulation precedes tumorigenesis. Two other prospective case-control studies on colorectal cancer with some amino acid measurements also found no significant associations [19, 31]. Our study is therefore the first to observe inverse associations of amino acid levels with colorectal cancer risk in a prospective setting and in independent studies. Although the above evidence suggests that tumor energy requirements give rise to the depletion of circulating glutamine and histidine, the levels of these amino acids among colorectal cancer cases in our prospective cohorts were associated with colorectal cancer risk at least 10 years prior to a colorectal cancer diagnosis, suggesting that alterations in the metabolism of these compounds may either reflect etiological pathways associated with the development of disease or metabolic changes linked to early events in colorectal tumorigenesis.

The most pronounced finding of the current study was a robust inverse association between circulating histidine and colorectal cancer risk in both EPIC and UK Biobank. Histidine, an essential amino acid derived from the diet, is converted by histidine decarboxylase to the biogenic amine histamine [32], a signaling molecule that mediates an acute inflammatory response by binding to specific receptors [33]. Histidine decarboxylase activity may be upregulated in tumor cells and is thought to accelerate cell proliferation and angiogenesis [34]. For instance, the enzyme was found to be more active in colon cancer cells, particularly metastatic tumor cells, than normal colonic cells [34]. Given that an inverse association between histidine and colorectal cancer risk was apparent as long as 10 years before diagnosis, perturbations to specific etiologic pathways may be hypothesized. Prior evidence suggests that higher histidine concentrations mitigate metabolic dysregulation; for example, dietary supplementation with histidine was found to improve insulin sensitivity, possibly via the suppression of pro-inflammatory cytokine expression, in women with metabolic syndrome [35]. These findings suggested that histidine may even hold potential as a therapeutic agent against metabolic disease. However, levels of histidine have also been found to be positively associated with breast cancer [36]. Therefore, additional laboratory studies are needed to elucidate the potential role of this amino acid in carcinogenesis.

Glutamine was borderline inversely associated with colorectal cancer risk in both cohorts. Glutamine is among the most abundant small-molecule metabolites in circulation and plays a central role in amino acid metabolism. It is used by proliferating cancer cells as an energy source [37] and is likely an important substrate throughout colorectal tumorigenesis [38]. The reasons for lower circulating glutamine in individuals who went on to develop colorectal cancer compared to those who remained cancer-free are uncertain. Firstly, given the slow development of colorectal cancer, lowered glutamine may reflect the undetected presence of polyps or early cancerous lesions in cases at baseline [28]. Secondly, regardless of the presence of such lesions and even controlling for major risk factors, abnormally low glutamine levels may reflect cancer-promoting metabolism. For instance, as well as being directly related to the tumor stage, glutamine levels have been inversely associated with serum C-reactive protein and inflammatory cytokines [13, 39]. Also, lowered glutamine and the glutamine-glutamate ratio was reported to be associated with incident type II diabetes [40], an established risk factor for colorectal cancer [41]. Lowered glutamine may represent dysregulation of the glutamine-glutamate axis. Although our study measured glutamate in a subset of EPIC cases and controls only, some evidence for a positive association of glutamate with colorectal cancer risk was observed in the categorical analysis. Glutamine concentrations may influence multiple mechanisms related to cancer development which deserve further investigation in experimental models.

The inverse associations observed between amino acid concentrations and colorectal cancer risk may also reflect cancer-promoting dysbiosis of the gut microbiota. Similar to the protection against colorectal cancer afforded by short-chain fatty acids produced from dietary fiber by microbiota via the mitigation of an inflammatory microenvironment [42], certain components of the gut microbiota may act upon amino acids in the lumen to influence inflammation and tumorigenesis [43]^,[44]. For example, the production of histidine decarboxylase by gut microbes has been suggested to decrease intestinal inflammation via the binding of histamine to the receptor HR2 in the gut lumen [45]. Also, specific components of the microbiota are also believed to mediate the relationship between branched-chain amino acids and insulin resistance [46]. It is therefore plausible that the associations of histidine and glutamine with colorectal cancer risk in the current study may reflect variability in gut microbial activity and its interaction with host metabolism. Further mechanistic research is needed to investigate links between histidine and glutamine metabolism, the gut microbiota and colorectal tumorigenesis.

The main strengths of this study include the prospective design, the use of large-scale cohorts with extensive participant data, and robust amino acid measurements in participants of well-characterized fasting status. We excluded non-fasted participants from the outset in the EPIC discovery cohort to minimize the effects of recent dietary intake upon amino acid levels which could have complicated interpretation of the results. As an additional safeguard against this bias, we performed sensitivity analyses for fasting status in EPIC and for major dietary sources of amino acids in both cohorts, which did not appreciably attenuate risk estimates. This is consistent with a recent study in EPIC that found weak or no correlations between amino acid intake and their blood concentrations [47].

In terms of limitations, only 9 of the 21 amino acids were measured in all EPIC and UK Biobank samples, while only 13 were measured all EPIC samples, with limited statistical power for the remainder. It is plausible that levels of amino acids other than glutamine and histidine are associated with colorectal cancer and we note that HR or OR point estimates were lower than 1 for most compounds in both cohorts. With greater statistical power, particularly in UK Biobank, other amino acids would likely have been found inversely associated with colorectal cancer risk via the discovery-replication strategy. Also, measurements of amino acids were taken at the study baseline only, and the technical and biological reproducibility of measurement was therefore not accounted for. However, studies calculating intra-class correlation in blood samples suggest that polar metabolites such as amino acids are measured reproducibly, particularly in fasted participants [48]. Statistical power was limited for individual colorectal cancer subsites, particularly rectal cancer. Also, we were not able to consider amino acids in tissue samples, which may better represent the tumor microenvironment and provider deeper insight into the biological implications of our findings.

Circulating histidine levels were robustly inversely associated with colorectal cancer risk in two independent prospective cohorts with similar, albeit slightly weaker, evidence for glutamine. This knowledge should contribute to a better understanding of the underpinnings of colorectal cancer and metabolism and could potentially support new prevention or early detection strategies. Further research using experimental models to assess potential causality of the identified associations is now needed.