Background
Colorectal cancer (CRC), the second and third most commonly diagnosed cancer in women and men, respectively, presents a variable geographical incidence that has been attributed to differences in diet and environmental exposure [
1]. These variations occur in conjunction with a genetically determined background of susceptibility. The incidence of CRC is substantially higher in men and women greater than 50 years of age, although the basis for this difference is unknown. Sex steroids have been considered a contributing factor because high parity, early age at first pregnancy, oral contraceptive use and oestrogen replacement therapy are associated with decreased risk of CRC [
2‐
5]. In addition, younger women (<45 years of age) with CRC have better overall survival than men of the same age [
6].
Familial adenomatous polyposis syndrome (FAP), a precancerous colorectal condition, is an inherited disease caused by a germline mutation in the adenomatous polyposis coli (
APC) gene [
7]. In FAP, many neoplastic lesions of different stages are typically found within the same individual, thus providing a better understanding of the adenoma-carcinoma progression. The fact that these polyps appear in the second decade of life (i.e., during puberty) and show increased size and number during adolescence (i.e., at the peak of production) also suggests that sex steroids may act as a cofactor in the development of colorectal polyps and in carcinogenesis linked to FAP [
8].
Evidence from several studies suggests that among the sex hormones, oestrogen may play a role in colorectal cancer, particularly through oestrogen receptor-β (ERβ) [
9‐
13]. However, since the discovery of this receptor in 1996 [
14], research efforts have been focused on defining its biological function, which remains poorly understood. Several alternative splicing isoforms of the oestrogen receptor beta gene (
ESR2) occur (transcript variants a, b, d, f, g, h, i, k, l; ERβ isoforms 1–5), and these isoforms have been detected in both normal and malignant cells [
15‐
18]. The expression of
ESR2 splicing variants in colon cancer cells was first reported in 2001 [
19]. To our knowledge, this is the first study in which the expression levels of
ESR2 variants in FAP, adenomatous polyps and sporadic colorectal carcinomas have been evaluated and compared with normal mucosa, tumour stage and prognosis data. The goal of this study was to provide a better understanding of oestrogens and their potential effects on sporadic and hereditary colorectal tumours.
Discussion
The effects of oestrogens on tissue are mediated by members of the nuclear oestrogen receptor subfamilies ERα and ERβ. In normal colorectal tissues, ERβ is predominant and appears to play an important role in the biological mechanisms of action of sex steroids [
25,
26]; its expression decreases at higher Dukes’ stages, which display little or no ERα expression [
9‐
11,
19]. Despite being important in the development of breast cancer, ER
α is found in low levels in normal colorectal tissue [
26]. Using different strategies, several studies reported no differences between the expression levels of ERα comparing colorectal tumours and normal mucosa [
12,
13,
27‐
29]. Therefore, the ERβ variant isoforms may be the key to understand the ERβ signalling in different tissues [
17].
The expression of splicing variants in colon cancer cells was first reported by Campbell-Thomson et al. in 2001 [
19]. In normal human colon tissue,
ERβ1, ERβ2 and
ERβ5 are expressed, whereas
ERβ3 and
ERβ4 have been found only in testicles. In 2005, Wong et al. reported the expression of
ERβ isoforms in colorectal carcinomas [
30]. In that study, the expression levels of
ERβ1 and
ERβ2 were completely lost in 22% and 49% of primary colorectal carcinomas, respectively. In contrast, a truncated isoform referred to as
ERβ5 was found in all colorectal carcinomas and in tumour cell lines. Despite the limited number of cases, the authors suggested a prognostic role of
ERβ1 and
β2 isoforms.
A few alternative
ERβ1 isoforms (
ERβ2/ERβcx, ERβ3, ERβ4 and
ERβ5) that result from alternative splicing of the last receptor-encoding exon have been reported. These proteins have a truncated domain or are otherwise altered in the C-terminal region [
31]. The C-terminal region of the ER contains a ligand-binding domain that is involved in transcription activation, receptor dimerization, nuclear translocation and interaction of the receptor with transcription co-regulators [
32]. Therefore, modifications in this region can affect the activity and biological function of the ER. Leung et al. (2006) reported that ERβ2–4-5 isoforms are deficient in intrinsic ligand-dependent transactivation activity, making ERβ1 (wild type) the only functional receptor isoform [
17]. However, ERβ2–4-5 isoforms form heterodimers with ERβ1 and may increase ERβ1-induced transcription activity at physiological oestrogen concentrations.
In tissues such as prostate, breast and endometrium, a relationship between oestrogen signalling and the initiation of carcinogenesis, cancer progression and therapeutic response have been suggested [
16,
33‐
38]. Oestrogenic action and the quantitative variation in the proportion of ERβ in different tissues indicate that the ERβ subtypes have different functions [
35,
39]. Although the specific roles of ERβ subtypes in colorectal cancer are still uncertain, it is plausible that the varying co-expression of different isoform groups in normal mucosa, polyps and tumours of the colon, as shown in this study, can alter carcinogenesis. These findings should be taken into account in the design of future studies and screening protocols. In addition, the frequency of screening examinations should be re-evaluated in subjects who are undergoing hormone replacement or hormone deprivation therapy, treatments that are commonly used in breast cancer. These patients have a higher risk of developing colorectal cancer, and the hormonal exposure they receive may be involved in this increased risk. The effect of the use of hormone replacement therapy in patients with CRC or FAP syndrome must therefore be carefully evaluated.
The analysis to evaluate
ERβ isoforms levels was performed in fresh frozen tissues (sporadic tumours) and in FFPE samples (sporadic and FAP polyps). One limitation of our study is the use of FFPE samples, which presents RNA with lower quality compared to fresh frozen tissue samples. However, the comparison of the expression levels of different
ERβ isoforms using normal mucosa (fresh frozen tissue) and normal FAP mucosa (FFPE) revealed decreased
ERβ expression in tumour samples and polyps. This result was confirmed grouping all samples independently of the tissue type (fresh tissue or FFPE). To validate the data, RNA-Seq data from 285 colon adenocarcinomas and 41 non-tumour colon tissue samples extracted from the TCGA dataset were used. In agreement with our data, this analysis revealed that mRNA ERβ expression levels were significantly lower in tumour tissue than in non-tumorous tissue samples (
p < 0.001; Mean Diff 0.984 ± 0.146; 95% CI 0.697 to 1.271). Nguyen-Vu et al. (2016) using RNA-seq data of 233 colon adenocarcinomas and 21 non-tumor colon tissues from The Cancer Genome Atlas (TCGA) dataset. The ERβ expression was decreased in the cancerous state compared to non-cancerous tissues [
40].
In our study,
ERβ1, ERβ2 and
ERβ4 variants (primer sets 1 and 3) showed downexpression in ST compared to NS. Similarly,
ERβ1, ERβ2, ERβ4 and
ERβ5 variants (primer sets 1, 2 and 3) presented downexpression in FAP polyps compared with NFAP. These findings were confirmed with TCGA dataset results in
ERβ2 and
ERβ4 performed in sporadic tumours. However, these transcripts demonstrated very low expression by RNA sequencing and some variants was not even detected by TCGA (i.e.:
ERβ1 and
ERβ5). Although RNA sequencing is a suitable method to identify and discriminate mRNA variants, the sensitivity of RT-qPCR is superior, which is considered a gold standard methodology to evaluate transcripts [
41‐
43].
Primer pair 2, which amplifies the transcript variants a and g (encoding ERβ1 and ERβ5) showed the highest expression levels of ESR2 in normal FAP mucosa (NFAP). FAP polyps emerge during puberty and increase in number during adolescence, which is the period of peak oestrogen production. Based on these findings, it appears that the transcript variants have the potential to act as cofactors in the development of polyps and FAP-related colorectal carcinogenesis. Interesting, comparing sporadic tumours (ST) with sporadic normal mucosa (NS), the ERβ1 and ERβ5 levels were not significant.
A primary chemoprevention trial [
44] consisting of a double-blinded, randomised four-year study using sulindac included subjects with
APC mutations as well as phenotypically unaffected subjects. The authors reported total eradication of the polyps in the placebo group, which included one patient who received an occasional administration of oral contraceptives and who had developed polyps [
44]. The patient follow-up comprised flexible rectosigmoidoscopy every four months for 48 months, and an increased prevalence and recurrence of polyps was observed after the discontinuation of oral contraceptives.
Selective oestrogen receptor modulators (SORMs or SERMs) such as raloxifene and tamoxifen have both oestrogenic and anti-oestrogenic actions depending on the tissue type [
44‐
46]. For example, raloxifene is not only effective in preventing osteoporosis [
47] but has also been shown to be as effective as the SORM archetype tamoxifen in preventing breast cancer [
48]; however, it increases the risk of uterine adenocarcinoma [
49]. Despite the widespread clinical application of these modulators, very little is known about how they may affect colon cancer. It is important to identify oestrogen receptor variant subtypes and to determine their relative expression levels at different stages of tumorigenesis, especially in different tissues, because specific oestrogen receptor subtypes can be involved in the specific agonistic or antagonistic actions of oestrogens. The next step is therefore to determine the correlation between the predominance of specific isoforms and their interactions with various forms of oestrogen (isoflavones, SORM and hormone replacement therapy). The incidence of colorectal cancer is much lower in Asian countries than in the western world, a fact that may be explained by the high intake of phyto-oestrogens, such as soy, in Asian cultures [
50]. This intake offers a protective effect that may be associated with the mediation of ER binding to genistein, which is the main phyto-oestrogen in soy and is thought to be a tyrosine kinase inhibitor, a characteristic that may explain some of its action on the colon [
51,
52].
When invasive tumours were analysed individually, no differences in expression associated with the degree of tumour differentiation were found. This result likely occurred because the majority (87.8%) of the tumours studied herein displayed a moderate degree of differentiation. Immunohistochemistry analysis using an ERβ monoclonal antibody revealed decreased expression in higher grade or more dedifferentiated tumours [
11,
13].
Although the results of a comparison of individual T stages (T1 vs T2 vs T3 vs T4) were not significant, most likely due to the limited number of cases at stages T1 and T4, the grouped analysis revealed that larger (T3/T4) tumours presented decreased ERβ expression levels compared to T1/T2 tumours. These alterations were more evident and significant for the primer pair 1, which amplified the largest number of transcription variants (a, b, d, f, k and l), representing ERβ1, ERβ2 and ERβ4 isoforms. The amplification with the primers 2 and 3 resulted in a significant reduction of the expression levels in normal mucosa to very low values during the colorectal carcinogenesis, which may explain the absence of differences between T1/T2 and T3/T4 tumours. One plausible explanation is that the primer 1 is more sensitive for the detection of these differences, because contains a group with more homologous variants. The loss of ERβ expression was significantly different in adenomatous polyps with low-grade dysplasia, illustrating that differences in the expression of these receptors occur at early stages of carcinogenesis.
Using an expression level threshold of −4.5, we found that low expression levels of the ERβ1 and ERβ5 isoforms (transcript variants a and g: primer pair 2) were independent factors associated with sporadic colorectal cancer recurrence (ERβ > −4.5 presented a 4.08-fold higher risk of relapse (CI 1.69–9.84); p = 0.002). Paradoxically, this sample set of cases displayed no significant difference in loss of normal mucosa ERβ expression. With respect to polyps and smaller and larger sporadic tumours, this subgroup was different in tissues from patients with FAP syndrome; therefore, it is plausible that this isoform group may be correlated with the APC suppressor gene, which is mutated in FAP syndrome. In contrast, sporadic cases are often associated with other variables associated with somatic carcinogenesis.
Cases of FAP cancer were not included in this study, which was designed to evaluate the prognosis of FAP tumours, because FAP is a very rare syndrome and therefore involves a heterogeneous group of patients; as a result, cases of FAP have been historically treated in several ways, which may lead to uncertainties in the prognostic results. Unfortunately, the gold standard for preventive treatment of colorectal cancer in FAP individuals is still total prophylactic colectomy at a young age, preferably prior to the malignant transformation of polyps [
7]. Thus, we investigated mRNA expression of ERβ isoforms in normal mucosa and polyps. Significant loss was demonstrated in all subgroups of ERβ isoforms in adenomatous polyps of PAF compared to normal mucosa (
p < 0.001).
Recently, an alternative mechanism for oestrogenic action was described where the G protein-coupled oestrogen receptor (GPER) mediates the rapid non-genomic effects of oestrogen, phyto-oestrogen and xeno-oestrogen [
53]. GPER activation may inhibit the growth of CRC cells, both in vitro and in vivo, through multiple intracellular signalling pathways. Liu et al. (2017) showed that GPER expression in tumour tissue was markedly lower than in corresponding adjacent normal mucosal tissue. Tumours expressing lower levels of GPER exhibited a significantly lower survival rate than those with higher GPER expression levels [
54]. The physiological significance of these rapid effects and their integration with nuclear responses to oestrogen are important issues that still need to be further investigated.