Gene expression profile and genomic alterations in colonic tumours induced by 1,2-dimethylhydrazine (DMH) in rats

Colorectal cancer is one of the most common neoplastic diseases in the Western World [1], developing through a multistage process which involves the accumulation of genetic and epigenetic alterations. Experimental models mimicking this disease in rodents, such as 1,2-dimethylhydrazine (DMH) or azoxymethane (AOM)-induced carcinogenesis, provide a tool for the understanding of the molecular alterations arising in human colon cancer. Inbred strains such as F344 rats, which are frequently used in these experiments, are relatively uniform and the tumors developing after induction show phenotypic and genotypic features similar to those observed in human sporadic colon cancers, notably activation of Wnt signaling pathway and mutations in Kras and Apc genes [2]. Importantly, AOM/DMH-induced tumours develop almost exclusively in the colon, at variance with other experimental models in which carcinogenesis develops mainly in the small intestine, a site rarely involved in human cancer. Since AOM/DMH tumours are affected by the same dietary variations known to affect human colon carcinogenesis, this model is among the most used and reliable for identification of potentially chemopreventive agents [3].

The advent of genome-wide technologies, such as gene expression microarray and array-based comparative genomic hybridization (a-CGH), has provided the possibility to achieve a comprehensive view of the alterations involved in cancer. Although several papers have been published on human colon tumours [4] and some also in AOM-treated mice [5, 6], no studies applying genome-wide technology to DMH-induced cancers in rats have been reported. Given these considerations, we thought it of interest to study the gene expression profile of DMH-induced colon tumours in rats with the 44K Agilent rat arrays, representing the whole rat genome. Moreover, we were also interested in studying genomic alterations in this experimental model. In fact, although genomic instability in AOM-induced colon cancer in rats was previously reported [7], the technique in that study (Random Amplified Polymorphic DNA (RAPD analysis)) did not give information on the affected loci. On the other hand, using low resolution comparative genomic hybridization (CGH) [8], it has also been reported that AOM-induced tumours in mice have a low level of genomic alterations, making it difficult to draw any conclusions. Given these considerations and the possibility of achieving a comprehensive view of genomic alterations involved in tumors with high-resolution genome-wide technologies, we analyzed DMH-tumours with a-CGH using high resolution arrays (2 × 105K Agilent rat genome array).

The analyses described in this paper were carried out in colon cancers harvested from 14 rats induced with DMH as described in the Methods. The mean number of total tumours (both adenomas and cancers)/rat was 2.5 ± 0.34 (SD); adenomas/rat were 1.14 ± 1.09 while cancers/rat were 1.36 ± 0.49; all rats developed tumours. All the cancers analyzed (n = 19) were adenocarcinomas (Fig. 1D), except for one that was a mucinous adenocarcinoma (sample #14). Two samples were graded as moderately differentiated (samples #2 and #14), while the remaining 17 samples were well differentiated cancers.

Bild vergrößern

This is the first report describing variations in gene expression and genomic alterations in DMH-induced rat colon adenocarcinomas using microarray technologies. We found that many transcripts were differentially expressed in colon tumors, 7400 of them being differentially expressed in tumours with a FC ≥ 2 [3724 (13.6%) and 3750 (13.7%) up-regulated or down-regulated in tumours].

The most up-regulated gene was Defcr4, also known as enteric defensin α5, belonging to the family of the α-defensin antimicrobial peptides secreted by Paneth cells [17]. Other markers of Paneth cells such as defensin alpha 6 (Defa6; FC = 18.7), matrix metallopeptidase 7 (Mmp7), lysozyme (Lyz; FC = 13.1), secretory phospholipaseA2 (Pla2 g2a; FC = 20), were also up-regulated in our samples, suggesting that these cells are a cellular component of colon tumours, in agreement with previous reports [17‐19]. Although it has been suggested that the presence of Paneth cells in colon cancer is a fortuitous consequence of the Wnt pathway activation [17, 20], it is interesting to note that Mmp7 up-regulation, as observed in our samples, has been correlated with metastasis and apoptosis resistance in cancer [21]. Besides Mmp7, other metalloproteinases (MMPs) involved in the degradation of extracellular matrix in the late phases of carcinogenesis were up-regulated in our samples (Mmp3, 9, 10, 12 and 14; FC = 4.3, 18.1, 13.1, 65.1, 3.2, respectively) (see also Additional file 2) [18], a result reinforced by the significant up-regulation of the MMP pathway pointed out by GenMapp analysis. Linked to MMPs in the process of invasion, we also found up-regulation of genes involved in the urokinase-type plasminogen activator system such as urokinase-plasminogen activator u-PA (Plau in rat; FC = 9.6), u-PA receptor (Plaur; FC = 2) and plasminogen (Plg; FC = 13.6) (see Additional file 2) [22]. Interestingly, it has also been suggested that u-PA might catalyze the activation of the mesenchymal-derived cytokine hepatocyte growth factor/Scatter Factor (HGF/SF), thus facilitating invasiveness [22, 23], an observation in line with the activation of HGF signaling in our tumours, as pointed out by the GenMapp analysis.

Slc30a2, the second most up-regulated gene in our set of colon adenocarcinomas, codes for the protein Znt2 which has been reported to promote zinc efflux into intracellular vesicles when Zn concentration rises in the cell [24]. Accordingly, a previous study showed that Zn concentration was higher in rat colon tumours than in normal mucosa [25].

The Lum gene, coding for the protein Lumican, is the third most expressed gene in our set of tumours (Table 1). Lumican plays an important role in collagen fibrillogenesis and its over-expression has been reported in human colorectal cancer [26]. Another dramatically up-regulated gene in our samples is Igfbp5 (Insulin-like growth factor-binding protein 5) (Table 1), a component of the IGF axis [27], implicated in colon carcinogenesis with contradictory results [28, 29]. Indeed, while Igfbp5 is up regulated in adenomas from Apc^Min/+ mice, it moves in the opposite direction in human colorectal cancers and cancer cell lines [28]. It has also been shown that IGFBP5 can be cleaved by MMP7 forming IGF-II which in turn can act as a growth factor for colonic myofibroblasts [29], suggesting an important role for IGFBP5 and MMP-7 in the tumoural epithelial-mesenchymal transition. Recent evidence also indicates that Igfbp5 may act as a tumour suppressor by inhibiting angiogenesis [30]. These results suggest the need to further investigate the role of Igfbp5 in colon cancers.

Among the most up-regulated genes we also found Nos2 (Table 1) coding for the pro-inflammatory enzyme inducible nitric oxide synthase (i-NOS). Many reports document high i-NOS activity in colorectal cancers [31, 32]; recently, we also reported that DMH-induced colorectal tumours have high i-NOS expression [33]. Interestingly, S100A8 and S100A9, whose products form the heterocomplex calprotectin, are strongly up-regulated in our study (Table 1), in agreement with reports on inflammation-associated cancer and in human colorectal carcinomas [34, 35]. Accordingly, we also observed an up-regulation of genes involved in prostaglandin synthesis, such as cyclooxygenase 1 and 2 (Ptgs1 and Ptgs2; FC = 5.4 for both genes, see Additional file 2) and prostaglandin E synthase (Ptges; FC = 2.2) associated with strong down-regulation of the gene for the enzyme degrading prostaglandins (15-hydroxyprostaglandin dehydrogenase, Hpgd; FC = -29), as demonstrated in both the array and RT-PCR experiments and in agreement with previous reports showing a down-regulation of this gene in colon carcinogenesis [36]. Moreover, GenMapp analysis pointed out that the TNFα-NF-kB pathway was up-regulated in our samples, as also demonstrated by the over-expression of p65 (RelA) observed with both immunohistochemistry (Fig.1, panels A-C) and array technique (FC of RelA: 2.2, see additional file 2). All these data confirm that DMH-induced tumours show activation of inflammatory process pathways, as documented in rats and human colon cancer [32, 33, 37].

Aberrant activation of the Wnt signaling has been demonstrated in both experimental and human carcinogenesis [38]. Accordingly, many Wnt-target genes were up-regulated in our samples (Ccnd2 (FC = 5.9), Lef1 (FC = 21.5), Mmp7, Axin2 (FC = 3.9), CD44 (FC = 3.4), Bmp4 (FC = 5.1), Dkk3 (FC = 14.3), Sox9 (FC = 2.9), Fn1 (FC = 31.4), Mmp9 (FC = 18.1), Stra6 (FC = 21.3), Ptgs2 (FC = 5.4), Postn (FC = 11.8), Sfrp2 (FC = 6.3)) [39] (see also Additional file 2). Notably, Lgr5, a marker of stem cells in colon epithelium [18] was also up-regulated in the arrays (FC = 10) as well as in RT-PCR experiments, demonstrating that cells presenting this stem marker are over-represented not only in Min mice intestinal tumours and human colon cancers [40, 41], but also in DMH-induced colonic tumors.

As expected, we also found an up-regulation of pathways related to nucleotide metabolism, DNA replication, cytoplasmic ribosomal protein, translation factors, proteasome degradation, mRNA processing and cell cycle (Table 2).

On the contrary, the up-regulation of the TGF-β pathway in our samples was somewhat unexpected since previous findings in colon cancers showed inactivation of TGF-β signaling [42‐44]. In our samples the genes coding for the ligand Tgfb1 (FC = 5.7), the receptors Tgfrb1 (FC = 2.6) and Tgfrb2 (FC = 5.5) as well as Smad2 (FC = 2.9) and Smad4 (FC = 2.4) were up-regulated, although not Smad3 (FC = -2.3) which was down-regulated. It is interesting to note that although TGF-β is a potent inhibitor of normal colonic epithelial cells, it can also promote the survival, invasion and metastasis of colorectal cancer cells, thereby acting as an oncogene, a paradoxical role which could explain our results [45, 46].

Slc26a3, coding for a Cl(-)/HCO(3)(-) exchanger, is the most down-regulated gene in this report. Although genetic studies linked mutations in the homologous human gene (SLC26A3) to congenital chloride-losing diarrhea, it is interesting to note that Slc26a3 is also known with the alias "down-regulated in adenoma" (Dra), since it was first identified as a gene strongly down-regulated in colon adenomas and adenocarcinomas [47, 48]. Its deficiency has been associated with an increase in cellular proliferation [48], a feature that could partly contribute to tumour development also in our samples. Mptx, the gene coding for mucosal rat pentraxin, is also strongly down-regulated in the present study. Mptx was first described as the most down-regulated gene in the colonic mucosa of rats fed dietary heme and, since it has been suggested to play a role in the clearance of dying cells, its down-regulation could lead to a reduced apoptosis ability, another characteristic of tumour cells [49]. Retlna (resistin-like molecule α), which is also down-regulated in our samples (Table 1 and Fig. 2), belongs to a family of resistin-like molecules implicated in metabolism, energy balance and intestinal inflammation [50].

We also observed a strong down-regulation of the Muc2 gene, coding for Muc2, the most abundant mucin secreted by normal colonocytes, in agreement with experimental and clinical reports documenting defective mucin production in colorectal cancer [51, 52].

Significantly down-regulated pathways were processes such as Krebs-TCA-cycle, electron transport chain, and fatty acid beta-oxidation (Table 2), in agreement with many previous observations showing a metabolic reprogramming of the cancer cell favoring glycolysis even in the presence of oxygen [53].

It has also to be noted that our rats were fed a high-fat diet containing 23% corn oil as a source of fat, a diet currently used in our laboratory since it closely mimics the high-fat diet typical of western countries with populations having a high incidence of colon cancer. Thus one could wonder whether the same results would be observed also in animals fed a low-fat diet. However, since in our experiments we compared tumours with their paired normal mucosa, it is not possible to infer the influence of diet on gene expression profile in tumours. In fact, to carefully address this issue, one would have to perform an additional experiment comparing the tumour profile from low-fat diet rats with that from high-fat diet rats.

In the present study we also evaluated genomic alterations with a-CGH analysis in a different set of tumours harvested from the same rats. This is, to our knowledge, the first attempt to highlight genomic aberrations in DMH-induced colonic tumours in the rat by means of high resolution a-CGH.

Our results show that genomic aberrations are relatively infrequent, with 40% of tumours showing only one or two aberrations (i.e. amplification or deletion), while the remaining 60% show a stable genome. The detected alterations spanned from 66 to 2135 kb and in no case did we observe gains or losses of whole chromosomes. The relatively low level of genomic alterations in this study is in agreement with previous reports on tumours from Min mice and AOM-treated A/J mice [8, 54]. In a previous study from our laboratory, we reported frequent genomic alterations in AOM-induced colon tumours using RAPD analysis [7]; this discrepancy might be due to the fact that high-resolution a-CGH, at variance with RAPD, does not detect small alterations (such as single base mutations), which could be frequent in this model.

Notably, one of the aberrations found in the present study was a deletion on chromosome 18, spanning a 1362 kb region containing the Apc gene, a key gene in colon cancer [38]. Previously, we demonstrated that DMH-induced colon cancers, beside mutations in Ctnnb1 (coding for β-catenin) and Kras, carry single-nucleotide mutations in Apc with a frequency of about 30% [2]. Therefore, the occurrence of an interstitial deletion encompassing this gene could be a further mechanism of Apc inactivation in colon carcinogenesis. Accordingly, allelic loss of one APC gene copy without detectable APC mutation has been reported in human colorectal tumours [55]. It is also interesting to note that among the genes mapping in the region deleted in this sample, we found Catna1 (coding for β-catenin). Unlike β-catenin, the contribution of α-catenin to colorectal carcinogenesis is controversial [56]. In fact some reports suggest that α-catenin is essential for tumour formation [56], while others suggest that α-catenin plays a tumour-suppressive role in the late stages of colorectal carcinogenesis [57, 58]. Notably all the genes deleted on chromosome 18 (including Apc and Catna1), are synthenic with their human homologues genes (i.e. they are also close to each other in the human genome, all of them mapping on the short arm of chromosome 5).