During colon carcinogenesis, histological modifications associated with genetic abnormalities are accumulated in a stepwise manner.1 This model postulates an adenoma-carcinoma sequence with a progressive evolution of “early adenomas” toward “intermediate adenomas” and “late adenomas.” While genetic modifications (mutations and deletions) during carcinogenesis have been well documented, little is known about their direct consequences on cell differentiation. However, alterations of mucin expression have long been associated with early colon carcinogenesis in the human as well as in the rat. Thus, the expression of sialomucin as well as of the gastric M1/MUC5AC mucin was observed in adenomas, hyperplastic polyps,2, 3 in the transitional mucosa4, 5, 6 and more recently in aberrant crypt foci (ACF).7, 8 Such markers of colon carcinogenesis may be useful for an early cancer detection.
Using another approach, Bird9 has described in 1987 for the first time a marker called “ACF” appearing early during colon carcinogenesis. ACF have been postulated to precede the development of adenomas, the precursor of colorectal cancer,10 and are now believed to represent the earliest step in colorectal carcinogenesis.11 This ACF/adenoma transition is now well supported by the presence of mutations shared between ACF and adenoma.12 Moreover, dysplasia, a marker indicating an increasing risk for cancer progression, has been observed in ACF induced with DMH or AOM,13, 14 and more recently in rats treated with 2-amino-1-methyl1-6phenylimidazo[4,5-b]pyridine (PhIP).15 Direct evidence of the tumors arising from large ACF has been observed in the AOM induced rat colon carcinogenesis showing a large aberrant crypt focus with a microscopic invasive carcinoma.16 This observation was confirmed later.17 Recently, we have demonstrated that human ACF expressed the MUC5AC mucin.8 We have adopted the rat colon carcinogenesis protocol using intrarectal instillations of Methyl-N′-Nitro-N-Nitroso-Guanidine (MNNG)18 to study the early lesions that ultimately give rise to colon tumors. Intrarectal instillation of MNNG, a directly-acting carcinogen is a reliable method for inducing tumors selectively in the area exposed to the carcinogen. MNNG does not require metabolic activation or alteration and acts on the colon mucosa.19 Consequently, in the area exposed to MNNG, it was possible to observe the initial premalignant lesions which lead to the formation of tumors. In rats treated intrarectally with MNNG there were no additional tumors in other organs, in contrast with rats injected with DMH and AOM. Moreover, MNNG represents a standard alkylating agent that preferentially methylates the O6 position of deoxyguanosine residues in DNA and there is much epidemiological evidence indicating the relevance of this compound to human cancer.20
Using a polyclonal antibody against human gastric M1 mucin which cross-reacted with rat gastric mucin, we demonstrated that the M1 mucin was an early marker of the DMH induced rat carcinogenesis.21 Indeed, M1-positive goblet cells, absent from the normal colon, were observed in the colon 2 weeks after the carcinogen treatment and their number increased during carcinogenesis until the appearance of tumors. For a better characterization of the M1 mucin, we have produced 12 monoclonal antibodies (Mabs) against human gastric mucin by an original immunohistological screening method using their specific reactivity on adenoma and their nonreactivity on the normal colon.22, 23, 24 Subsequently, we demonstrated that the M1 immunoreactivity was encoded by the MUC5AC gene25 and 8 of the 12 M1 epitopes were mapped on the MUC5AC apomucin.26 Unfortunately, most of these Mabs immunoreacted with precancerous rat colon mucosae only after deglycosylation of mucin using periodate.27 For this reason, using immunohistology we selected the Mab 660, which immunoreacted with the native gastric mucin expressed in the rat colon during rat carcinogensis as the polyclonal anti-M1 mucin.28
In our study, we show that the 660 epitope is encoded by the rMuc5AC gene and we describe in detail the sequential development large ACF towards adenomas/adenocarcinoma with the abnormal expression of MUC5AC mucin.
Material and methods
The animals were housed according to the European Union Regulations on the Care and Use of Laboratory Animals. Carcinogenesis was induced in 7–8-weeks-old male Wistar rats. Rats (n = 86) were given 4 successive intrarectal deposits of 0.5 ml solution of MNNG (5 mg/ml) (Sigma-Aldrich, Steinheim, Germany) twice a week for 2 weeks, at 7 cm of the anal margin, using a bulb-tip needle. Groups of rats were sacrificed by CO2 inhalation at week 10 (n = 20), 14 (n = 20), 22 (n = 20), 30 (n = 10) and 66 (n = 16) after the beginning of treatment. In addition, groups of three rats were treated in the same manner with 0.5 ml phosphate buffered saline (PBS) or MNNG, and another group of three rats was treated with 0.5 ml of Pristane (2,6,10,14-tetramethylpentadecan) (Sigma-Aldrich, Steinheim, Germany), as an additional control for inflammatory reaction. These groups were sacrificed at Week 4.
Identification of ACF
At sacrifice, the colons were opened longitudinally. The colon mucosae were flattened between filter papers, fixed in 95% ethanol and kept at 4°C in 95% ethanol for several days.8
ACF were detected in the rat colons and characterized according to Bird.9 The mucosae were stained by 0.05% methylene blue in distilled water for 1 min. The mucosal side was then observed at 40× magnification. ACF were distinguished from the surrounding normal crypts (NC) by their slit-like opening, increased size, blue staining and pericryptal zone. ACF were classified into 2 groups according to the number of ACs: small (<10 Acs) and large (≥10 Acs). Small colon trapezoidal ACF-containing fragments of ∼1 cm2 were dissected from the entire colon mucosa for immunohistological examination. The trapezoid sample shape allowed to orient the ACF during paraffin embedding (Fig. 1a), to obtain longitudinal sections of colon crypts. The remaining mucosae were coiled into “Swiss rolls” (Fig. 1b) around the tumor (when present), fixed and embedded separately in paraffin.
Monoclonal antibody (660 Mab) raised against rat gastric mucin,28, 29 which reacts with an oncofetal mucin marker of rat colon carcinogenesis, has been previously used to study rat ACF.8 To verify that the 660 Mab recognized the mucin encoded by the rMuc5AC gene, we used the 463M Mab, an anti-rat gastric mucin that cross-reacts with human gastric mucin epitope located in the D4 domain of the C-terminal region of the human MUC5AC gene.26 By immunohistology, in the normal gastrointestinal tract, the 660 Mab stained the surface gastric epithelium exclusively and did not stain the gastric and Brünner's glands, neither the intestinal epithelium where the MUC6, MUC2 and MUC5B genes are expressed.28 Mab 660 did not stain other rat tissues (lung, ovary, liver, spleen, kidney).28 Mabs 660 and 463M were used as purified immunoglobulins from ascites fluids after affinity isolation on protein A Sepharose™ (Amersham Pharmacia Biotech AB, Uppsala, Sweden) according to the supplier's instructions. Purified 660-IgG was biotinylated using Pierce EZ-link NHS-LC-biotin according to the supplier's instructions (Pierce, Rockford, IL). Purified IgG were used for the ELISA, exclusively.
Enzyme-linked solid-phase immunoassay
Microtiter plates (Nunc, Wiesbaden, Germany) were coated overnight at 4°C with 100 μl/well of 2 μg/ml IgG1 Mab 463M in 0.1 M sodium bicarbonate buffer pH 9.5. The plates were aspirated, washed 3 times with PBS/0.5% Tween-20 (Sigma-Aldrich, Steinheim, Germany), and incubated with 3% BSA (Sigma-Aldrich, Steinheim, Germany) in PBS for 1 hr at 37° to block nonspecific adsorption sites. After washing 3 times with PBS/Tween-20, the plates were incubated with increasing quantities of rat gastric mucin extract (scrapings of rat gastric mucosa) for 2 hr at 37°C. After 3 washes, the wells were incubated with 100 μl of biotinylated IgG2A of Mab 660 (1 μg/ml), washed 3 times with PBS/Tween-20, and incubated with avidin-linked alkaline phosphatase (Sigma-Aldrich, Steinheim, Germany) for 30 min at 37°C. Phosphatase activity was measured with p-nitrophenylphosphate (Sigma-Aldrich, Steinheim, Germany). The optical density was measured at 450 nm.
Deparaffinised sections were incubated 2 hr with citrate buffered solution (pH 6) including a pretreatment for 15 min at 120°C in the “Retriever 2100™” device (Pickcell, Amsterdam, The Netherlands). After a 3 min preincubation in PBS containing 0.1% tween-20, sections were incubated for 30 min with the undiluted Mab 660 hybridoma supernatants. After 3 rinses in PBS-tween-20, the sections were incubated for 30 min with peroxidase-labelled anti-mouse IgG (H + L) (Santa Cruz, CA) diluted 1/200 in PBS-tween-20. After 3 rinses in PBS-tween-20, peroxidase activity was revealed with amino-ethylcarbazol (Sigma-Aldrich, Steinheim, Germany). Cell nuclei were stained during 5 min with 1% hematoxylin (Sigma-Aldrich, Steinheim, Germany). The specificity of the immunoreactivity was controlled by inhibition of staining after absorption of 1 ml of the hybridoma supernatant with 10 mg (dry weight) of lyophilized rat gastric mucus.
Data bases were established using Excel software for each ACF or NC including the number of the rat, the time after the beginning of MNNG treatment and for ACF the quantity of aberrant crypts (ACs). The statistics used included the Student's t-test. The increase in the percentage of large ACF or the number of MUC5AC positive NCs was measured by calculating the Pearson correlation coefficient. A p value ≤ 0.05 was considered to be statistically significant.
Rat MUC5AC immunoreactivity of the Mab 660
When the anti-rat MUC5AC apomucin antibody (463M Mab) was adsorbed to the solid phase and incubated with the rat gastric mucin extract, the 660 epitope was immunofixed as revealed by the biotinylated MAb 660 (Fig. 2). This immunofixation increased with the amount of mucin extract. This result demonstrates that the 660 epitope is associated with the MUC5AC rat apomucin.
Macroscopic observation of the colon mucosa during carcinogenesis
Until Week 22 after treatment with MNNG, the rat colonic mucosae appeared macroscopically normal, and only the lymphoid structures normally present throughout the normal left rat colon were seen. This observation was confirmed by histological examination. At the Week 30, 8 out of the 10 rats examined have developed tumors (12 tumors in total) (mean number of tumors/rat, 1.2 ± 1.03). Ten tumors were very small (<0.2 cm diameter). Among the 16 rats sacrificed at the Week 66, a total of 12 tumors in 9 rats were noted (mean number of tumors /rat, 0.75 ± 0.86). All the tumors were located in the distal part of the colon.
Evolution of the ACF during carcinogenesis
The mean number of ACs per focus before 22 weeks was smaller (5.54) than after 22 weeks (9.26) (p < 0.0001). According to our ACF classification and taking into account the number of ACs in each focus, the number of small and large ACF per rat increased until Week 22 and 30, respectively (Fig. 3). At Week 66, the number of ACF per rat declined and was approximately the same for each size group. The percentage of large ACF per rat increased with the time of carcinogenesis (Fig. 4) (r = 0.839; p < 0.0001). The colon of the control rats (intrarectal deposits of PBS or pristane) did not show ACF. In contrast, the 3 MNNG treated rats displayed small ACF already at 15 days after instillation.
Evolution of the MUC5AC-positive NC
The distal part of control rat colon (intrarectal deposits of PBS or pristane) did not react with the anti-MUC5AC Mab (Mab 660). MUC5AC-positive NCs were observed exclusively in the distal colon in the areas where the colon mucosa has been in contact with the carcinogen MNNG. The proximal colon was not stained with the Mab 660. Most of the positive glands were isolated, scattered throughout the distal colon. Their number decreased during carcinogenesis between Weeks 10–66. (r = 0.394; p = 0.0007) (Fig. 5).
Thirty weeks after MNNG treatment, 8 out of the 10 rats have developed a total of twelve small tumors, detected by methylene blue staining of the entire colon. Eight of these tumors had large ACF (more 20 crypts in each) in their adjacent areas. The tumors showed severe dysplastic glands and looked like in situ carcinomas and no parietal invasion was observed. Four tumors were located in the middle of ACF (Fig. 6a), while the 4 others were at the border of ACF (Fig. 6b). This pattern suggested that these tumors arose from large ACF. The 4 tumors, which did not show evidence of adjacent ACF were the largest (>0.2 cm). Serial sections of the entire 10 small tumors were obtained and stained for histological and immunohistochemical examination (Fig. 6c). In all cases, cell proliferation showing severe dysplasia was observed in deep glands as well as in the surface region covering the associated remainders of ACF, which were positive for MUC5AC. The NC at a distance from the tumor areas were not stained with the anti-M1/MUC5AC Mab. Moreover, in 3 ACF, MUC5AC positive bifurcating glands were observed in the deep glands (Fig. 6d). Such glands were not observed in the normal colon mucosae coiled into Swiss rolls of rats treated with MNNG. In the 12 tumors studied, both the adjacent ACF and tumors were strongly stained with the anti-MUC5AC Mab. In the tumors, most of the goblet cells (75–95%) strongly expressed the MUC5AC mucin. Among the tumors observed, 2 (Weeks 30 and 66) showed parietal invasion and were classified as adenocarcinomas.
The large ACF were particularly carefully studied because of the above observation suggesting that tumors arose from these large ACF. At Week 30, 10 very large ACF containing more than 40 ACs were removed and examined for dysplasia using serial sections. All large ACF showed foci of dysplastic glands: few goblet cells, slight nuclear polystratification and numerous mitoses (Fig. 6e). These 10 ACF showed a strong expression of MUC5AC.
Residual ACF in the mucosa adjacent to large tumor (Week 66)
At Week 66, the mucosa adjacent to the tumors, called transitional mucosa, was characterized by a thicker epithelium because of the elongation of crypts that were sometimes branched. Several glands showed a serrated epithelium, as observed in ACF. In 12 of the 16 tumors studied, the histology of the transitional mucosa was similar to that of ACF, and was also strongly stained by the anti-MUC5AC Mab.
It is generally accepted that the early phase of colon carcinogenesis is characterized by the sequence ACF–adenoma. This sequence has been suggested by a close relationship between the induction of ACF and the carcinogenic effect of chemicals.30 Moreover, the progression of ACF (increase in their size and number) has been extensively used as a biomarker to evaluate the effects of natural products or drugs that enhance or inhibit colon carcinogenesis.30 Based on features shared by ACF and adenomas, 3 types of arguments have been proposed to support the concept of the ACF–adenoma sequence of rat colon carcinogenesis: (i) biochemical and immunohistochemical alterations, such as sialomucin,7 or M1/MUC5AC mucin expression;8 (ii) mutations of genes: K-ras,31Ctnnb115 and Apc,15 suggesting that genetic alterations play a key role in the progression of some ACF toward adenomas; (iii) histopathologic features: the dose-dependent induction of ACF by carcinogens preceeding the growth of adenomas, and the presence of dysplasia in ACF.13 This last argument is not the least and a direct histological proof has been furnished by Pretlow showing the presence of a small carcinoma arising from a large ACF.16 This observation was confirmed later, by a surgical approach of Shpitz showing that some ACF are precursors of adenomas and cancers.17 Biochemical, immunohistochemical, genetic and histopatologic alterations were also observed in human ACF.32
Based on microscopic observation of intact tissue and immunohistological studies, we present evidence that adenomas/adenocarcinomas are growing from large ACF within 30 weeks after the initiation of MNNG treatment. Indeed, microscopic examination using methylene blue staining clearly demonstrated large ACF adjacent to the tumor in eight out of 12 small tumors analyzed. Histological studies confirmed this observation since in all the tumors (12/12), the adjacent area displayed histological features typical of ACF, and was positive for MUC5AC. Also, dysplasia was observed in all the very large ACF (>40 ACs) found at Week 30. This observation supports the conclusion that tumors arise from these dysplastic glands. In fact, these large ACF could be regarded as microadenomas as already suggested,33 and the tumors as genuine in situ carcinomas. Using another carcinogen (MNNG) and another route of carcinogenesis (intrarectal instilllations), more relevant for the human,20 our report is in agreement with the previous findings of Pretlow et al.16 and Shpitz et al.17 who observed the emergence of tumors from ACF using AOM injected by systemic route.
At Week 66, the mucosa adjacent to the tumors, termed “transitional mucosa,”4 displayed epithelium secreting the MUC5AC mucin, suggesting that these glands may be remnants of ACF. Indeed, ACF were the main source of the MUC5AC throughout the colon. The isolated MUC5AC-positive NCs were few, and their number did not increase during carcinogenesis. At this time (Week 66), such MUC5AC-positive transitional mucosae have still been observed in association with 12 out of 16 tumors. This transitional mucosa shares with the ACF also other features such as expression of sialomucin,2, 7KRAS mutations34 and bifurcating glands.2 All tumors expressed MUC5AC.
Recently, other putative premalignant lesions have been described in colon carcinogenesis, such as β-catenin-accumulated crypts35, 36 and mucus depleted foci37 However, these lesions were induced using DMH or AOM by systemic route. In our carcinogenesis model, using another carcinogen (MNNG) and another way of administration (intrarectal instillation) we have observed the emergence of tumors mainly from ACF. The apparent discrepancies may be due to the different pathways of carcinogenesis implicating different premalignant lesions in the various experimental conditions used (mode of administration and type of carcinogen). Colon carcinogenesis induced by oral administration of PhIP, acting in part by luminal exposure of the colon mucosa, similar to the MNNG instillation, also induced ACF with dysplasia.15 Both MNNG and PhIP are considered to be relevant for human colon carcinogenesis.20
In agreement with previous publications,32 in our model of MNNG carcinogenesis we observed a decrease in the number of small ACF at a longer time after treatment (after Week 22), suggesting that some ACF were transient in nature. Moreover, the number of “large” ACF (generally defined as 4 or more ACs per focus),14 or crypt multiplicity (the average number of ACs per focus)38 demonstrated a better correlation with tumor development than did the total number of ACF. Our results suggest that tumors grew from large ACF and that the number of these large ACF progressed during carcinogenesis, in contrast with the number of small ACF which declined after 22 weeks. At Week 66, the transitional mucosa adjacent to the tumors4 displayed serrated epithelium secreting MUC5AC mucin as observed in ACF, suggesting that these glands may be remnants of ACF.
In the past, we described the association of the M1/MUC5AC mucin with human adenomas,3 premalignant lesions of the colon. In the present study, we show that ACF are the main lesions, which strongly express MUC5AC in macroscopically normal mucosa, and that this expression increases during carcinogenesis in parallel with the increasing number of large ACF. In contrast, the MUC5AC-positive NCs are transiently induced and then decrease with time after exposure to the carcinogen. Moreover, our study reveals the association of the M1/MUC5AC expression with the ACF–adenoma/adenocarcinoma sequence. Indeed, the expression of the gastric MUC5AC mucin is a phenotypic feature common to ACF and adenoma, most likely as a consequence of the genetic alterations induced by the mutagenic effect of MNNG, and supports the notion that the tumors were actually growing from ACF. Consequently, the emergence of tumors (in situ carcinomas) from strongly MUC5AC-positive large ACF (microadenomas) indicates that this mucin is an excellent biomarker of early colon carcinogenesis.
In the normal rat colon, the expression of the MUC5AC mucin ceases 1 day after birth.21 Its re-expression in ACF suggests a fetal character of this lesion. Such a fetal-type differentiation has been seen in ACF by Scanning electron microscopy, which revealed a “fission mechanism” for the formation of glands located at the base of the crypts, similar to that occurring during fetal development (postnatal period).39 The bottom of the crypts is the area where the functional stem cells are found and it is probable that each fission event begins at the base of crypt in both postnatal glands and ACF. Moreover, in our study we observed such MUC5AC-positive bifurcating glands at the base of the crypts in 3 different ACF. Such glands were not observed in any of the histologically normal colon mucosa at 30 weeks after the beginning of MNNG treatment. Crypt fission, observed in sporadic human colorectal adenomas and hyperplastic polyps, is described as a process implicated in the growth of these polyps.40, 41 The biological significance of ACF could be the regeneration of glands in the colon mucosa in response to an injury elicited by the carcinogen, including, re-expression of proteins characteristic of fetal phenotype. This process could transiently reactivate epithelial redifferentiation of fetal type. We suggest that a dysfunction of this process may produce large ACF (microadenomas).
We suggest a possible regulation of rMuc5AC gene expression by early genetic changes in colon carcinogenesis, including Wnt pathway: APC/β-catenin/TCF, Cdx1, Cdx2 and other genes acting at the bottom of the colonic crypts and affecting cell proliferation and differentiation and stem cell niche.42 Such a regulation needs to be investigated.
In conclusion, using a model of locally applied chemical carcinogen we directly show the growth of tumors from large ACF associated with the expression of MUC5AC. This is in agreement with the earlier reported correlation between the size of ACF and the risk of colorectal cancer.43, 44 In the human, 70% of M1/MUC5AC positive transitional mucosa display ACF-like morphology.6 It is likely that many human adenocarcinomas may in fact originate from ACF. Our results suggest that large ACF need to be checked and removed during endoscopic examination along with adenomas, to reduce the risk of colon cancer.
The authors thank Dr. Jan Mester for his assistance in editing this article for style and usage of English. The technical work of Mrs. Lydie Germain was greatly appreciated.