We wanted to characterize the expression of putative markers of stem cells in the early phases of colon carcinogenesis. To this aim we studied the expression of LGR-5, MSI-1, DCAMKL-1, CD133 and ALDH1-A1 in normal mucosa, in MDF (microscopic precancerous lesions) and in macroscopic tumours (adenomas) of DMH-induced rats using immunohistochemistry. We also studied the co-localization of these markers with nuclear β-catenin, an additional proposed feature of stemness. To the best of our knowledge, this is the first study in which these markers are evaluated all together along the carcinogenesis process in this relevant model of colon cancer.
LGR-5, a Wnt target gene, was firstly hypothesized as a stem cell marker because of its profile of expression at the base of the crypts, compatible with that of a stem cell . Further investigations with lineage–tracing experiments confirmed it as probable stem cell marker in both the small intestine and colon of mice . In the present study we observed a low expression of LGR-5 in the normal rat mucosa. Only in some crypts we did observe positive cells or a diffuse staining localized mainly in the lower compartment of the crypt, although a small percent of the total labelled cells (13%) was also observed in the upper compartment of the crypt. Although we do not know the origin and nature of these rare positive cells in the luminal compartment, the low expression of LGR-5 in the normal mucosa, together with its prevalent expression at the base of the normal crypts, agrees with previous studies in humans using immunohistochemistry techniques as in the present study [28, 29], and suggest that LGR-5 is a marker of stem cells also in the rat. On the other hand, the seminal papers locating LGR-5 positive cells at the base of the crypt [21, 30], identified those cells by evaluating LGR-5 gene expression (through RNA in situ hybridization) or through the expression of a chimeric protein (LGR-5-EGFP) using an antibody directed against EGFP [21, 30], and thus it is difficult to compare these studies with ours.
It has been hypothesized that stem cells are more abundant in tumours than in normal mucosa [2, 19] and we also observed that there was a clear overexpression of LGR-5 compared to the normal mucosa in MDF and tumours. This result, especially the overexpression in MDF, suggests that LGR-5 expression identifies putative stem cells, which are involved in the process of carcinogenesis from the very beginning, i.e. in precancerous microscopic lesions. We also observed, in agreement with previous reports in humans [21, 28, 31], an overexpression of LGR-5 in macroscopic tumours (i.e. the adenomas), confirming overexpression of the LGR-5 gene that we previously reported in DMH-induced colon tumours .
We also studied the expression of two additional stem cell markers of colonic stem cells, i.e. Musashi-1 (MSI-1) and doublecortin and calcium/calmodulin-dependent protein kinase-like-1 (DCAMKL-1). MSI-1, firstly identified in Drosophila as an RNA binding protein associated with asymmetric divisions in neural progenitor cells, has been proposed by Potten and colleagues as a intestinal stem cell marker . These authors reported that MSI-1 is localized in the stem compartment of the intestinal crypts and is overexpressed in small-intestinal adenomas of the Min mouse, a genetic model of intestinal carcinogenesis . However, colonic staining in both mouse and human samples was weak . Here we show that intense nuclear stained cells were present in the lower third of normal colonic crypts; in contrast, although a diffuse cytoplasmic overexpression was observed in some MDF and tumours, the nuclear staining was not observed and the majority of lesions were negative. Previous studies in normal human colon mucosa showed both cytoplasmic and nuclear staining, but the mechanisms regulating the intracellular localization of this protein are not clear . Altogether, our results do not support MSI-1 as a robust cancer stem cell marker. Accordingly, in our previous transcriptomic analysis of DMH-induced colon tumours, Msi-1 was actually downregulated compared to normal mucosa .
DCAMKL-1, a microtubule-associated kinase expressed in post-mitotic neurons has also been proposed as a stem cell marker . However, the stem cell nature of DCAMKL-1 positive cells has been questioned [33, 34]. Accordingly, DCAMKL-1 has been shown to be a specific marker of tuft, caveolated cells known for many years [33, 35], but poorly characterized until now. Gerbe and colleagues studied DCAMKL-1 positive cells in mouse small intestine demonstrating that they are secretory cells expressing COX-1, COX-2 and β-endorfin . Our results show that DCAMKL-1 positive cells were scattered along the entire length of the crypt in the normal mucosa, with only a slight prevalence in the lower compartment. The percentage of DCAMKL-1 cells was slightly, although not statistically significantly, reduced in MDF, while significantly diminished in macroscopic adenomas. This result is in agreement with Gerbe et al., who report a reduction of DCAMKL-1 cells in human adenocarcinomas, and with our previous results in DMH-induced tumours showing a decreased expression of Dcamkl-1 gene . In contrast, Sureban and colleagues found overexpression in human colon cancers .
As a further step aiming at the identification of putative cells with the most stemness features, we carried out co-localization experiments with LGR-5, MSI-1, DCAMKL-1 and nuclear β-catenin, an additional putative marker of stemness, especially in colon cancer [10, 11]. Previous studies on co-localization between LGR-5 and nuclear β-catenin, performed in EGFP-creERT2/Apc flox/flox mice, showed that cells expressing both LGR-5 (chimeric) and intense nuclear β-catenin staining give rise to daughter cells which still accumulate nuclear β-catenin but lose the EGFP-LGR-5 positiveness, suggesting that the simultaneous expression of both markers identify cancer stem cells . A positive correlation between LGR-5 and β-catenin expression (both in cytoplasm and nucleus) has also been reported , but co-localization of both markers in the same histological sections of a tumour has not been performed so far, nor has it been studied in the early phases of carcinogenesis. Here we show that in both MDF and tumours a sub-population of LGR-5 positive cells accounting for a small percentage of the lesion cells (about 1%) exhibits nuclear β-catenin staining, suggesting, on the basis of the previous considerations, that these cells could be neoplastic stem cells. The finding that the percentage of these putative neoplastic stem cells, i.e. of cells expressing the two putative markers, is unchanged between microscopic MDF and larger adenomas suggests that MDF represent the step in which stem cells overpopulation already occurs.
Co-localization between DCAMKL-1 and nuclear β-catenin shows that some cells in the tumours (about 0.4%) co-expressed DCAMKL-1 and nuclear β-catenin; this figure is similar to that observed for the co-expression of LGR-5 and β-catenin (Figure 10), but unfortunately we do not know whether these cells are the same. Therefore, also considering the lower number of DCAMKL-1 positive cells in tumours, a result not in line with the expected increase for a cancer stem marker, we can not draw any conclusions on the significance of these DCAMKL-1-nuclear β-catenin co-expressing cells.
Non-steroidal anti-inflammatory drugs (NSAIDS) are well documented as having chemopreventive activity in colon carcinogenesis [37, 38]. However, besides their ability to block COX, their mechanism(s) of action at a molecular level and the specific population of cells targeted by these drugs have not been completely clarified. Accordingly, recent data show that the protective effect of NSAID may vary depending on genetic and phenotypic characteristics of the tumour . Previous studies have documented a reduction of nuclear β-catenin in tumours from subjects taking daily aspirin or ibuprofen for up to 25 years . Similarly, nuclear β-catenin translocation was reduced or abolished in tumours from DMH-induced rats chronically treated with various NSAIDS . Accordingly, PGE2, a downstream product of COX-2, has been shown to enhance Wnt signaling  suggesting that variation in PGE2 within the stem cell niche may affect cell stemness properties. Recently, it has been shown that short-term pre-operative treatment with the NSAIDS indomethacin or celecoxib, (a selective COX-2 inhibitor), reduces the expression of the stem cell marker CD133 . Moreover, a short treatment with the NSAID sulindac induced apoptosis in LGR-5 expressing cells and reduced nuclear β-catenin in the intestine of Min mice , suggesting that PGE2 modulates the expression of these markers and that it may be possible to down-regulate stemness properties with pharmacological treatments. To verify this hypothesis we treated tumour-bearing rats for two weeks with celecoxib to determine if the treatment was able to reduce the percentage of cells with putative stemness features which we considered those expressing both LGR-5 and nuclear β-catenin. The dose regimen of celecoxib we used has been shown to reduce COX-2 activity and colon carcinogenesis when chronically administered to AOM-induced rats  and is equivalent to the human dose when adjusted for the appropriate scaling factor . In vitro studies showed that celecoxib reduces Wnt-activity in colorectal cancer cells . Our results show that celecoxib causes only a slight reduction in the number of cells expressing nuclear β-catenin, which is not statistically significant; we also observed that it does not reduce cells co-expressing LGR-5 and nuclear β-catenin, indicating that, at least in these experimental conditions, the expression of these putative cancer stem markers is not affected by treatment.