Background
Colorectal cancer (CRC) remains the third most common cancer worldwide with more than one million new cases reported in 2008 [
1]. The liver is the most common site of CRC metastasis with 50-60% of CRC patients eventually developing liver disease [
2]. Metastasis, in common with growth and invasion of established tumours, is dependent on tumour cells acquiring a migratory and invasive phenotype as part of a highly conserved cellular trans-differentiation programme, the epithelial-mesenchymal transition (EMT) [
3,
4]. Prostaglandins (PG), in particular PGE
2, have previously been implicated in EMT of CRC cells [
5].
Prostaglandins (PG) are fatty acid signaling molecules known to have a range of physiological functions including vascular homeostasis, reproduction and immune regulation [
6]. PGE
2 is the most abundant PG in the human colon and levels of PGE
2 are increased in colorectal neoplasia compared with normal colorectum [
7]. Elevated PGE
2 levels are known to promote colorectal carcinogenesis at various stages, including CRC growth and metastasis [
6]. Recently, PGE
2 has been implicated in promotion of EMT
in vitro[
8]. PG G/H synthase (also known as cyclooxygenase [COX]) controls the rate-limiting step in PGE
2 synthesis, upstream of PGE synthases [
6]. There are two COX isoforms; the constitutive isoform COX-1 and the inducible isoform COX-2, which is up-regulated in CRC and is a putative target for anti-CRC therapy [
9,
10]. Nicotinamide adenine dinucleotide (NAD+)-linked 15-hydroxyprostaglandin dehydrogenase (15-PGDH) controls the rate-limiting step in PGE
2 catabolism by conversion of PGE
2 to 15-keto-PGE
2 coupled to the reduction of NAD+ to NADH [
11]. Initial studies suggested that 15-PGDH expression is reduced in bladder cancer, lung cancer and CRC compared with paired normal tissue and has tumour suppressor properties [
12]. However, subsequent reports have highlighted elevated 15-PGDH expression in breast and ovarian cancer [
12]. Moreover, there are conflicting data on 15-PGDH expression in gastric cancer [
12]. Heterogeneity of 15-PGDH expression in human cancers may reflect tissue-specific differences in regulatory pathways upstream of 15-PGDH [
12], but may also be related to sampling variation secondary to intra-tumoral genetic and micro-environmental influences on 15-PGDH expression [
13,
14].
There has been relatively little investigation of changes in 15-PGDH
activity, as opposed to gene expression, in human cancer. Although NAD+, and its phosphorylated form NADP, act as critical hydride-accepting co-enzymes in cellular metabolism, NAD+ is also a substrate for three classes of NAD+ −consuming enzymes, which are the poly(ADP-ribose) polymerases (PARPs), cADP-ribose synthases and sirtuins, expression and activity of which can be altered in cancer cells [
15,
16]. Therefore, a valid hypothesis is that NAD+ availability is rate-limiting for 15-PGDH activity and PGE
2 catabolism in CRC cells.
Regional hypoxia is common in many cancers including CRC [
17], in which established markers of tumour hypoxia have been linked to worse prognosis [
18]. Central tumour areas are believed to be more hypoxic than peripheral tumour tissue [
19] as demonstrated in CRC liver metastases (CRCLM) by dynamic con-trast-enhanced magnetic resonance imaging [
20] and immunohistochemistry (IHC) for carbonic anhydrase IX [
21]. Hypoxia is associated with increased PGE
2 production and release by several human cell types, including human CRC cells,
in vitro[
22‐
24]. This is believed to occur via induction of the COX-2-PGE synthase axis, with no change in 15-PGDH expression, although 15-PGDH activity and NAD+/NADH levels were not measured in these studies [
23,
25]. Expression of NAD+ −consuming enzymes such as SIRT1 is increased in hypoxic cells [
26] and overall NAD+ levels have been demonstrated to be reduced in ischaemic tissue, as well as a reduction in the NAD+/NADH ratio [
27].
Given the potential micro-environmental influence of hypoxia and co-factor availability on PGE2 metabolism, we tested the hypothesis that there are regional differences in PGE2 levels within human CRCLM, which are related to differential expression and activity of 15-PGDH and COX-2 within tumours. To this end, we collected and analysed human CRCLM tissue from peripheral and central areas of tumours in a systematic, protocol-driven manner, comparing our tissue findings with observations in human CRC cells in vitro, including those from the LIM1863 human CRC cell model of EMT.
Discussion
This is the first study to report regional differences in the levels of PGE
2 and 15-PGDH in human colorectal tumours. This was made possible by employing a strict protocol for rapid and uniform processing of orientated tumour tissue
ex vivo. Herein, we report that PGE
2 levels are higher towards the centre of CRCLM compared with more peripheral cancer tissue. Paradoxically, this was associated with increased levels of 15-PGDH
protein at the centre of CRCLM. However, we demonstrated that the 15-PGDH
activity level in the centre of CRCLM is reduced and is associated with low NAD+/NADH levels.
In vitro studies confirmed that NAD+ availability drives 15-PGDH activity in human CRC cells. We believe that consideration of regional differences in PGE
2 metabolism and micro-environmental influences on PGE
2 metabolism related to enzyme co-factor availability and/or hypoxia is a paradigm shift in the field of eicosanoid cancer research and is consistent with latest understanding of genetic and epigenetic intra-tumoral heterogeneity [
14,
40]. Consideration of intra-tumoral differences in PGE
2 metabolism is essential for development of optimal anti-CRC therapy aimed at the COX-PGE
2-15-PGDH axis.
Our data highlight significant differences between findings in human cancer tissue
ex vivo and experimental observations using CRC cells
in vitro. Although we propose that differences in 15-PGDH
activity in cancer tissue compared with cultured CRC cells may account for the contrasting relationship between 15-PGDH expression and PGE
2 levels in CRCLM tissue versus cell-conditioned medium, we cannot completely rule out that inadvertent stimulation of PGE
2 synthesis
ex vivo occurred. Avoidance of possible artefactual changes in tissue eicosanoid levels
ex vivo will only be possible with other techniques such as MALDI-MS for measurement of PG distribution in frozen tissue sections [
41].
The tissue microarray comparison of regional differences in 15-PGDH immunoreactivity between CRCLM and the paired primary CRC suggests that 15-PGDH expression, and hence PGE
2 metabolism, in CRCLM differs from that in the primary CRC, from which the CRCLM were derived. This finding is consistent with recent data describing significant genetic differences between primary CRC and synchronous liver metastasis [
40]. Local factors specific to CRCLM may, at least partly, explain regional 15-PGDH expression in CRCLM and the contrast with observations from previous studies of 15-PGDH expression in primary CRCs [
12].
NAD+ and NADH levels were both significantly lower in central rather than peripheral CRCLM tissue, compatible with depletion of the cellular NAD(H) pool. The NAD+/NADH ratios that we observed in human CRCLM tissue are similar to previous studies that have measured tissue NAD(H) levels by the same cycling assay [
42]. However, absolute levels of NAD+ and NADH were low compared with other tissues [
42]. One testable hypothesis is that the NAD(H) pool is depleted because of increased NAD-consuming enzyme activity in CRC cells. Consistent with this notion, sirtuins such as SIRT1 and poly-(ADP ribose) polymerase expression and activity are increased in cancer tissue [
43]. In particular, SIRT1 expression and activity are increased in human hepatoma and fibrosarcoma cells
in vitro[
26].
One weakness of our study is that we do not have direct evidence that the central area of CRCLMs that we studied were hypoxic. However, there is substantial indirect evidence that regional hypoxia exists in tumours including CRCLMs [
19‐
21]. Importantly, the regional difference in functional 15-PGDH protein levels in CRCLMs was not mirrored in primary CRC. Central tumour necrosis is more common in CRCLMs than primary CRC tumours and implies greater degrees of hypoxia in the central regions of CRCLMs, which could account for differential 15-PGDH expression in metastatic tumours. This observation, and the fact that elevated 15-PGDH in CRC cells in the centre of CRCLMs is likely inactive secondary to NAD+ deficiency, help to reconcile our data with the existing literature, which, in general (but not exclusively), implies that 15-PGDH has tumour suppressor activity [
12].
Roberts et al. have reported that acute hypoxia (16 hours) did not alter 15-PGDH protein expression in HT-29 human CRC cells, despite an increase in PGE
2 levels believed to be secondary to COX-2 induction [
25]. It is possible that CRC cell line-specific differences in hypoxia-induced gene expression and NAD+ availability explain the experimental variability in
in vitro models. Nevertheless, our data highlight that it is crucial to confirm the relevance of
in vitro observations in tissue expression studies, which take into account potential micro-environmental influences.
TGFβ-induced attachment and spreading of LIM1863 human CRC cell colonies allowed us to develop a novel semi-quantitative measure of EMT based on an established model [
38]. Using this assay, we have provided support for previous observations that PGE
2 drives EMT of CRC and other human cancer cells
in vitro, which were based on down-regulation of E-cadherin expression, light-microscopic phenotype changes in adherent cells and cell motility assays [
5,
36,
44,
45].
We have contributed to emerging evidence that hypoxia drives EMT [
46]. Interestingly, we observed that 15-PGDH expression was maintained in hypoxic TGFβ-induced LIM1863 human CRC cell colonies
in vitro and CRC cells in the centre of CRCLMs that had an ‘EMT (E-cadherin-low) phenotype’. This is consistent with our observations that hypoxia induces 15-PGDH in other CRC cell lines
in vitro and that 15-PGDH levels are higher in the centre rather than the periphery of CRCLMs. One testable hypothesis is that hypoxia inhibits β-catenin-related signaling [
47], which could lead to de-repression of 15-PGDH [
48]. Further studies are required to understand the rather counter-intuitive finding that the main rate-limiting catabolic enzyme for PGE
2 inactivation is elevated in a tumour microenvironment, in which cell survival would be potentiated by PGE
2. These studies should always take into account NAD+ co-factor availability and measure levels of other lipid mediators, which have anti-proliferative activity, that are also potential substrates for 15-PGDH such as lipoxins [
11].
Previous
in vitro studies have demonstrated that Snail, one of the key transcription factors in EMT, represses 15-PGDH expression in CRC cells via direct binding to conserved E-box elements in the 15-PGDH promoter region [
5]. However, to our knowledge, the effect of hypoxia on human 15-PGDH gene expression has not been formally assessed. The human 15-PGDH gene promoter contains multiple ETS, AP-1 and CREB binding sites [
49], although no hypoxia-responsive elements for direct hypoxia-inducible factor binding are evident. Therefore, a valid, testable hypothesis is that 15-PGDH is a hypoxia-inducible gene in CRC via ETS-dependent transcriptional up-regulation, which is recognised for several hypoxia-inducible genes [
50].
Competing interest
The authors declare that they have no competing interests.
Authors’ contributions
ALY collected all tissue samples and performed in vitro studies; CRC and GH performed CRC cell experiments; SLP performed immunohistochemistry; DT supervised the immunohistochemical scoring; GJT, PJ and MAH conceived of the study, and participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.