Colorectal carcinomas display a high degree of heterogeneity as incidence, malignancy and clinical outcome differ widely in relation to age, gender and tumor location in the colorectum [1,2]. We had raised the question whether variations in disease susceptibility reflect changes in the activities of intrinsic systems that inhibit or promote cancer cell growth. There is suggestive evidence of this with respect to the components of the autocrine/paracrine vitamin D system of the gut, which produces the potent anti-mitogen 1,25-dihydroxyvitamin D₃ [1,25-(OH)2D3], and of inflammation-induced cyclooxygenase-2 (COX-2), which catalyzes the conversion of arachidonic acid to growth stimulatory prostaglandin E₂. Previous studies of the significance of vitamin D insufficiency and chronic inflammation on colorectal cancer development clearly indicated that maintenance of cellular homeostasis in the large intestinal epithelium requires balanced interaction of 1,25-(OH)₂D₃ and prostaglandin cellular signaling networks (e.g., [3]) as illustrated in Figure 1.

The hormonally active vitamin D compound, 1,25-(OH)₂D₃, when bound to the high-affinity vitamin D receptor (VDR), functions as a potent anti-mitogen in many normal and neoplastic cells [6]. Endogenous synthesis of 1,25-(OH)₂D₃ from 25-hydroxyvitamin D (25-(OH)D) in the gut mucosa plays an important role in control of cellular growth and differentiation in the colorectal mucosa [5]. A compromised vitamin D status, characterized by low serum levels of 25-(OH)D, may increase the risk of colorectal cancer due to limited availability of 25-(OH)D as substrate for the colonic CYP27B1-encoded 25-hydroxyvitamin D-1α-hydroxylase. Hence, activity of the enzyme may not be high enough to achieve steady-state concentrations of 1,25-(OH)₂D₃ in the gut mucosa, which are necessary for efficient control of cell growth. In addition, increased activity of the vitamin D catabolic enzyme, the CYP24A1-encoded 25-hydroxyvitamin D-24-hydroxylase, may also curtail anti-proliferative actions of endogenously produced 1,25-(OH)₂D₃ [7]. CYP24A1 has therefore been considered to function as an oncogene for instance in the colon [8].

Anti-inflammatory properties of vitamin D include the ability to down-regulate expression of inflammation-associated PTGS2-encoded cyclooxygenase-2 (COX-2) [4,9]. Vitamin D insufficiency may therefore contribute to the persistence of chronic inflammation in the large intestine, which plays an important role in the development not only of colitis-associated but also of sporadic colorectal cancer [10,11]: Increased COX-2 activity enhances the risk of neoplastic transformation in ulcerative colitis by inducing pro-oncogenic mutations in colonic crypts that may give rise to dysplastic and cancerous lesions [12]. In addition, overexpression of COX-2 accelerates growth of adenomas [13], promotes tumor progression through the adenoma/carcinoma sequence [10,14] and appears to be related to poor survival in colon cancer patients [14]. Both primary and secondary prevention with COX inhibitors lead to decreased incidence of colorectal carcinoma (see e.g., [15]).

In the present study we performed a patient level analysis of CYP27B1, VDR, CYP24A1 and COX-2 mRNA expression in colorectal carcinomas. We were able to demonstrate for the first time by a statistical method also applied in systems biology that variations in the expression of the growth inhibiting CYP27B1/VDR system relative to tumorigenic COX-2 and CYP24A1 could account for the significant age- and gender-related influence on incidence and malignity of cancers at different anatomical locations within the colorectum, which we had reported previously [2].

Results from mRNA expression analysis (Figure 2) show that transcriptional activities of CYP27B1, CYP24A1 and COX-2 are significantly elevated in cancerous lesions compared to mucosal tissue outside of the tumor area (cf. [21,20]). During tumor progression elevated CYP27B1 remains constant, whereas COX-2 peaks in low grade and CYP24A1 in high grade cancers. As 1,25-(OH)₂D₃ induces CYP24A1 in colon carcinoma cells [7], this could explain, at least in part, the parallel rise of the vitamin D catabolic enzyme during tumor progression to low grade malignancy. However, the extraordinarily high level of CYP24A1 activity in highly malignant cancers probably does not result from up-regulation by the 1,25-(OH)₂D₃/VDR system but may be due to overexpression of the gene. Taken together, expression of the two vitamin D hydroxylases and of COX-2 relative to each other depends very much on the degree of tumor cell differentiation.

In our cohort we observed a distinct site-specificity, since there is more COX-2 expression in proximal tumors compared with distal and rectal carcinomas. Sub-group analysis according to gender revealed a significantly higher expression of CYP27B1, VDR and COX-2 in the tumors located in the distal colon of male patients compared with that of female patients. In the rectum, only CYP27B1 showed gender-dependent differences, such that expression was higher in women. COX-2 showed an even stronger gender-dependency in low-grade tumors with significantly higher expression levels in men compared with women (Figure 3 and Figure 4).

By factor analysis, the eight mutually correlated variables (see Figure 5) could be assigned to four latent, independent variables, so-called factors. These serve as markers for higher categories that cannot be measured directly but represent core features of colorectal cancer pathogenesis. For example, high loadings of CYP27B1 and VDR as well as of COX-2 on Factor 1_TU (Table 1) indicate that (i) 25-hydroxyvitamin D-1α-hydroxylase-mediated synthesis of 1,25-(OH)₂D₃ and its genomic actions via the nuclear VDR are apparently not disrupted in colorectal tissue irrespective of subsite, age, gender, and tumor grading, and (ii) that expression of the 1,25-(OH)₂D₃/VDR system and of the inflammation marker COX-2 are strongly correlated. This can be interpreted as activation of the 1,25-(OH)₂D₃/VDR complex in response to inflammatory stimuli in colorectal cancer tissue. Gustafsson Asting et al. [22] showed that high COX-2 expression in colorectal cancer is associated with expression of a host of inflammatory cytokines and mitogenic signaling factors. Particularly the latter have been implied to mediate the up-regulation of the CYP27B1/VDR system in early malignancy [5,21]. Taken together, Factor 1_TU stands for joint activity of the CYP27B1/VDR system and COX-2 in an inflammatory cancerous environment.

Factor 2_TU captures high COX-2 expression in proximal colon cancers (Figure 2), which is more pronounced in men than in women. In the analysis, “subsite” was defined as ordinal variable with increasing indices assigned to proximal colon, distal colon, and the rectum, respectively. Likewise, as for the dichotomous variable “gender”, female sex was assigned a higher index than male sex, explaining the negative sign of COX-2 expression loadings (Table 1). Altogether, this indicates an increase in inflammatory activity in cancerous lesions in the colorectum in the distal to proximal direction, particularly in men. By the same token, the interpretation of an inflammatory component decreasing towards the rectum, particularly in female patients, is equally valid. Factor 2_TU therefore reflects a subsite- and gender-related inflammatory component in colorectal cancer development. The proximal colon is certainly the prime site of tumorigenic actions of bile acids. This probably causes a higher incidence of cancers at this site compared to more distal locations in the colorectum. Bile acids such as deoxycholic and lithocholic acid are able to induce pre-neoplastic lesions, so-called aberrant crypt foci, in the colonic epithelium. These are “hot spots” of hyperproliferating pluripotent, undifferentiated cells, which are likely to develop into cancerous lesions (for review, [23]). There is also evidence that bile acids specifically induce COX-2 expression and, at the same time, suppress expression of 15-hydroxyprostaglandin dehydrogenase, the key enzyme responsible for catabolism of prostaglandins [24]. Notably, some bile acids also interact with the vitamin D system: For example, by binding to the VDR, lithocholic acid stimulates expression of CYP24A1 [25] and might thus reduce the availability of locally synthesized anti-mitogenic 1,25-(OH)₂D₃.

The lower incidence of cancers in the distal parts of the colorectum in women had been linked to direct and indirect anti-proliferative effects of estrogens, which are known to modulate COX-2 and CYP27B1/VDR expression: Thus, 17β-estradiol inhibits prostaglandin E₂-induced COX-2 expression [26] and increases VDR and CYP27B1 expression and activity in cultured human colonocytes [27,28,29] as well as in human rectal epithelium in vivo [30].

Salient positive loadings of gender and age on Factor 3_TU can be interpreted to account for gender-specific influence of age on colorectal cancer pathogenesis. In our cohort, female patients, particularly those with high grade colorectal cancer, were older than their male counterparts (for details, [2]). It has also been reported that younger women with metastatic colorectal cancer survive longer than younger men [31]. The question why women are protected particularly from more aggressive cancer in early life cannot be easily answered. It has been argued that this may be due to long-time exposure to sex hormones before menopause, or during hormone replacement therapy thereafter. It is well established that women on hormone replacement therapy have an approximately 40% lower risk of colorectal cancer, particularly when estrogen therapy is complemented with progestin [32,33]. Because sporadic colon cancers progress stepwise from adenoma to carcinoma, with a latency period that may last decades [34] and with highest incidence during advancing age, it is conceivable that tumors start developing slowly before menopause, but rapidly progress with cessation of ovarian estrogen production.

Factor 4_TU reflects a strong positive correlation between CYP24A1 and histopathological tumor grading. Factor 4_TU therefore accounts for the relevance of vitamin D degradation for progression to a higher degree of malignancy.

The plausibility of the presented factor analytical model was confirmed by cluster analysis of all variables. Cluster analysis aims at sorting different objects into groups in a way that the degree of association between two objects is maximal if they belong to the same group and minimal otherwise. The results can be presented in form of a “hierarchical tree” as for example in Figure 6 (see Legend for details): The cluster that can be formed from CYP27B1/VDR and COX-2 at a short linkage distance, indicating a high degree of association between the variables, corresponds to Factor 1_TU of the factor analytical model. Likewise, the clusters built at a somewhat higher distance from age and gender or from CYP24A1 and tumor grading reflect Factors 3_TU and 4_TU, respectively. A combination of the single branch “subsite” with a cluster containing the variables with salient loadings on two factors, i.e., COX-2 and gender, corresponds to Factor 2_TU (Figure 6).

We want to emphasize that Factors 1_TU–4_TU are specific features of malignancy as none of them could be extracted from variables with respect to adjacent mucosal tissue. In fact, Factors 1_AM–3_AM seem to reflect a basically normal situation as they are apparently not influenced by the grade of malignancy of the adjacent tumor: Factor 1_AM indicates activity of the vitamin D endocrine system in normal mucosa. Factor 2_AM denotes a positive association between female gender and CYP24A1-encoded 25-(OH)D-24-hydroxylase. Factor 3_AM shows that age is an independent variable in the histologically normal mucosa. Only Factor 4_AM can be seen as indication that progression to higher grades of malignancy makes its imprint on tissue out of the tumor area, as COX-2 and VDR in conjunction with tumor grading became salient (Table 1).

Our results have some important implications for prevention of colorectal cancer. Since inhibition of COX-2-mediated prostaglandin synthesis protects from colorectal cancer only to a certain degree, modulation of colonic vitamin D hydroxylase activities by nutrient factors, as proposed by Cross et al. [5,35], should be considered as an additional strategy for prevention. Enhancement of 25-(OH)D-1α-OHase activity will add to availability of 1,25-(OH)₂D₃ in the gut epithelium and consequently lessen the tumorigenic potential of COX-2, while simultaneous down-regulation of CYP24A1 could reduce the risk of further tumor progression.

Inhibition of COX-2 cannot be suggested for colorectal cancer prevention in general, because the use of selective as well as non-selective COX-2 inhibitors bears a strong risk of thrombotic cardiovascular events in individuals with a high risk for cardiovascular disease (for review, see [32]). In this case, cancer prevention with vitamin D could turn out as a viable alternative to COX-2 inhibition by non-steroidal anti-inflammatory drugs, because 1,25-(OH)₂D₃ has the potential to inhibit COX-2 [33] and to protect from adverse cardiovascular events [36].

Combined results from factor and cluster analysis suggest that (i) the antagonism between COX-2 and the anti-proliferative activity of the vitamin D system could be an important determinant of cancer cell growth in the colorectal epithelium; (ii) differences in COX-2 expression could be, at least in part, responsible for variations of cancer incidence at different anatomical subsites; (iii) gender-specific differences in cancer incidence correlate mainly with advancing age; (iv) progression from well differentiated to undifferentiated cancers correlates with the activity of the vitamin D catabolic enzyme, the CYP24A1-encoded 25-hydroxyvitamin D-24-hydroxylase.