Stem cells, quiescence and rectal carcinoma: an unexplored relationship and potential therapeutic target.

Stem cells are responsible for maintaining differentiated cell numbers during normal physiology and at times of tissue stress. They have the unique capabilities of proliferation, self-renewal, clonogenicity and multi-potentiality. It is a widely held belief that stem-like cells, known as cancer stem cells (CSCs), maintain tumours. The majority of currently identified intestinal stem cell populations appear to be rapidly cycling. However, quiescent stem cell populations have been suggested to exist in both normal intestinal crypts and tumours. Quiescent CSCs may have particular significance in the modern management of colorectal cancer making their identification and characterisation a priority. In this review, we discuss the current evidence surrounding the identification and microenvironmental control of stem cell populations in intestinal crypts and tumours as well as exploring the evidence supporting the existence of a quiescent stem and CSC population in the gut and other tissues.

Conditional expression of mutated K-ras accelerates intestinal tumorigenesis in Msh2-deficient mice.

K-ras mutation occurs in 40-50% of human colorectal adenomas and carcinomas, but its contribution to intestinal tumorigenesis in vivo is unclear. We developed K-ras(V12) transgenic mice that were crossed with Ah-Cre mice to generate K-ras(V12)/Cre mice, which showed beta-naphthoflavone-induction of Cre-mediated LoxP recombination that activated intestinal expression of K-ras(V12) 4A and 4B transcripts and proteins. Only very occasional intestinal adenomas were observed in beta-naphthoflavone-treated K-ras(V12)/Cre mice aged up to 2 years, suggesting that mutated K-ras expression alone does not significantly initiate intestinal tumourigenesis. To investigate the effects of mutated K-ras on DNA mismatch repair (MMR)-deficient intestinal tumour formation, these mice were crossed with Msh2(-/-) mice to generate K-ras(V12)/Cre/Msh2(-/-) offspring. After beta-naphthoflavone treatment, K-ras(V12)/Cre/Msh2(-/-) mice showed reduced average lifespan of 17.3+/-5.0 weeks from 26.9+/-6.8 (control Msh2(-/-) mice) (P<0.01). They demonstrated increased adenomas in the small intestine from 1.41 (Msh2(-/-) controls) to 7.75 per mouse (increased fivefold, P<0.01). In the large intestine, very few adenomas were found in Msh2(-/-) mice (0.13 per mouse) whereas K-ras(V12)/Cre/Msh2(-/-) mice produced 2.70 adenomas per mouse (increased 20-fold, P<0.01). Over 80% adenomas from K-ras(V12)/Cre/Msh2(-/-) mice showed transgene recombination with expression of K-ras(V12) 4A and 4B transcripts and proteins. Sequencing of endogenous murine K-ras showed mutations in two out of 10 tumours examined from Msh2(-/-) mice, but no mutations in 17 tumours from K-ras(V12)/Cre/Msh2(-/-) mice. Expression of K-ras(V12) in tumours caused activation of the mitogen-activated protein kinase and Akt/protein kinase B signaling pathways, demonstrated by phosphorylation of p44MAPK, Akt and GSK3beta, as well as transcriptional upregulation of Pem, Tcl-1 and Trap1a genes (known targets of K-ras(V12) expression in stem cells). Thus, mutated K-ras cooperates synergistically with MMR deficiency to accelerate intestinal tumorigenesis, particularly in the large intestine.

Wnt ligands influence tumour initiation by controlling the number of intestinal stem cells.

Many epithelial stem cell populations follow a pattern of stochastic stem cell divisions called 'neutral drift'. It is hypothesised that neutral competition between stem cells protects against the acquisition of deleterious mutations. Here we use a Porcupine inhibitor to reduce Wnt secretion at a dose where intestinal homoeostasis is maintained despite a reduction of Lgr5+ stem cells. Functionally, there is a marked acceleration in monoclonal conversion, so that crypts become rapidly derived from a single stem cell. Stem cells located further from the base are lost and the pool of competing stem cells is reduced. We tested whether this loss of stem cell competition would modify tumorigenesis. Reduction of Wnt ligand secretion accelerates fixation of Apc-deficient cells within the crypt leading to accelerated tumorigenesis. Therefore, ligand-based Wnt signalling influences the number of stem cells, fixation speed of Apc mutations and the speed and likelihood of adenoma formation.

Mutagenesis of mouse intestine in vivo using the Dlb-1 specific locus test: studies with 1,2-dimethylhydrazine, dimethylnitrosamine, and the dietary mutagen 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline.

The ability of three model carcinogens, 1,2-dimethylhydrazine, dimethylnitrosamine, and 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline, to induce mutation in a novel in vivo assay in mouse intestine has been examined. The assay is based on mutations at the Dlb-1 locus which determines the tissue specific pattern of expressio of the binding site for the lectin Dolichos biflorus agglutinin. In C57BL/6J x SWR F1 mice Dlb-1 mutants are recognized as clones of epithelial cells not staining with a peroxidase conjugate of D. biflorus agglutinin. Chronic administration of 1,2-dimethylhydrazine (20 mg/kg/week s.c. for 10 weeks) induced Dlb-1 mutants, whereas administration of a single dose did not. Similarly, chronic dimethylnitrosamine treatment p.o. (0.001% in drinking water for 8 weeks) induced Dlb-1 mutants, but acute administration did not. In contrast, neither chronic nor acute treatment of the mice with 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline induced Dlb-1 mutations. The activities of 1,2-dimethylhydrazine, dimethylnitrosamine, and 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline in the Dlb-1 assay more accurately reflect their carcinogenic potential than do many in vitro bioassays.

2-Amino-1-methyl-6-phenylimidazo[4,5-b]pyridine is a potent mutagen in the mouse small intestine.

Mutations in long lived stem cells are critical events in carcinogenesis. The Dlb-1 assay detects intestinal stem cell mutation at the Dlb-1 locus in Dlb-1a/b heterozygous mice by visualizing mutated clones of epithelial cells in situ which do not bind the lectin Dolichos biflorus agglutinin. We have used this assay to show that the food-derived heterocyclic amine 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) is a potent intestinal mutagen when administered either i.p. or p.o. This contrasts with the inactivity of the structurally related mutagen 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline in the assay which we have described previously. Immunocytochemical localization of the P-450 enzyme CYP1A2, which is responsible for the primary activation of these mutagens, shows that in untreated mice it is present in liver hepatocytes and in occasional villus epithelial cells but is absent from the target intestinal stem cell population. In addition, liver microsomes, unlike intestinal microsomes, were able to convert PhIP to the proximate mutagen N-hydroxy-PhIP. CYP1A2 immunoreactivity in beta-napthoflavone-induced animals was elevated in liver hepatocytes and increased to a lesser extent in duodenal villus epithelial cells. Treatment with beta-napthoflavone produced an unexpected 46% decrease in the number of Dlb-1 mutations in response to PhIP. Following treatment with PhIP, there was no difference in the number of Dlb-1 locus mutations between the proximal and distal ends of the small intestine in uninduced animals, indicating that the bile duct is unlikely to be responsible for transport of mutation inducing metabolites of PhIP to the small intestine. Our results demonstrate that metabolic activation of an indirect acting genotoxic agent can occur at a site other than the target tissue, and absence of the enzymes required for activation of a mutagen does not necessarily protect that tissue from its genotoxic effects.

p53, mutation frequency and apoptosis in the murine small intestine.

Normal function of the p53 gene is integral to the cellular response to genotoxic stress. One prediction arising from this is that p53 deficiency results in an increased mutation frequency. However, limited evidence has been produced in support of this idea. In order to further investigate the in vivo role of p53 in surveillance against mutation, and particularly to address the significance of p53-dependent apoptosis, we scored mutation frequency at the Dlb-1 locus within cells of the intestinal epithelium of animals which were wild type, heterozygous or null for p53 and heterozygous (a/b) at the Dlb-1 locus. Using this assay we have shown that loss of a p53-dependent apoptotic pathway is associated with the detectable acquisition of mutations, but only at high levels of DNA damage. These results question the significance of the immediate 'wave' of p53-dependent apoptosis seen in this tissue, particularly as there was a delayed p53-independent apoptotic pathway. We conclude that loss of p53 function only becomes relevant to the in vivo acquisition of mutations and thus tumorigenesis in certain circumstances.

Msh-2 suppresses in vivo mutation in a gene dose and lesion dependent manner.

Mice deficient for the mismatch repair (MMR) gene Msh2 show accelerated tumourigenesis and a reduced apoptotic response to DNA damage of methylation type. Here we examine the effect of mutation for Msh2 on in vivo mutation frequencies in the intestine as determined by loss of function at the Dolichos biflorus (Dlb-1) locus. Spontaneous mutation frequencies were scored in cohorts of ageing mice either wild type or mutant for Msh2. In mice less than 1 year old, mutation frequencies were only elevated in Msh2 null mice. However, beyond this age heterozygous Msh2 mice showed significantly higher mutation frequencies than controls. These findings implicate a gene dose dependent requirement for Msh2 in mutation suppression and prompted an analysis of young Msh2 mutants following exposure to DNA damage. Following exposure to N-methyl-N'-nitro-N-nitrosoguanidine (MNNG), Msh2 deficient mice show a reduced apoptotic response and an increase in mutation frequency. Heterozygotes did not differ from controls. Following exposure to cisplatin, no significant elevation was seen in mutation frequencies, even within homozygotes. This is particularly surprising given the association between cisplatin resistance and MMR deficiency. These findings therefore demonstrate a complex reliance upon functional Msh2 in mutation surveillance. We have identified three separate scenarios. First, where retention of both Msh2 alleles over an extended period of time appears critical to the suppression of spontaneous mutation; second, 3 weeks following exposure to MNNG, where only complete loss of Msh2 results in elevated mutation; and finally following cisplatin exposure, where induced levels of mutation are independent of Msh2 status.

C-cell and thyroid epithelial tumours and altered follicular development in transgenic mice expressing the long isoform of MEN 2A RET.

Gain-of-function mutations in the gene encoding the receptor tyrosine kinase RET have been identified as the aetiological factor for multiple endocrine neoplasia type 2A (MEN2A). MEN2A is a dominantly-inherited cancer predisposition syndrome characterized by medullary thyroid carcinoma, a tumour of the calcitonin-producing thyroid C-cells. There are three isoforms of RET: RET9, RET43 and RET51, and although in vitro evidence suggests they vary in cellular transformation activities, little is known about their function in tumorigenesis in vivo. To address this, we used RET51 cDNA to construct mice in which the most frequent MEN2A mutation, Cys-634-Arg, was expressed under the control of the human calcitonin promoter (CT-2A mice). These mice developed C-cell tumours resembling human MTC and follicular tumours resembling human papillary thyroid carcinoma (PTC) depending on the founder line examined. One founder line developed compound MTC/PTC at low frequency (8%) and pancreatic cystadenocarcinoma. CT-2A mice also displayed a developmental defect in thyroid follicular structure, in which much of the thyroid was occupied by large irregular cystic follicles thought to be derived from the ultimobranchial body, a developmental precursor of the thyroid gland. The CT-2A mice will provide a suitable model to further study the effects of the MEN 2A RET mutation in vivo.

Stem-cell organization in mouse small intestine.

We have investigated stem-cell organization in mouse small intestine (SI) by using a cellular marker induced by somatic mutation. In small intestinal whole mounts from heterozygous Dlb-1b/Dlb-1a mice stained with a peroxidase conjugate of Dolichos biflorus agglutinin (DBA-Px), mutations of Dlb-1b in stem cells result in loss of DBA-Px binding and so are recognizable as wholly or partly unstained crypts. The frequency of these clonal patterns can be measured during the accumulation of spontaneous mutations in untreated mice, or after treatment with ethylnitrosourea (ENU). The results show that there is a single infrequently dividing stem cell that maintains the epithelium of each crypt through a population of transit stem cells. The entire crypt epithelium is renewed approximately every 12 weeks.

Gene-mutation assays in lambda lacZ transgenic mice: comparison of lacZ with endogenous genes in splenocytes and small intestinal epithelium.

Comparison of results derived from transgenic animal gene-mutation assays with those from mutation analyses in endogenous genes is an important step in the validation of the former. We have used lambda lacZ transgenic mice to study alkylation-induced mutagenesis in vivo in (a) lacZ and hprt in spleen cells, and (b) lacZ and Dlb-I in small intestine from lambda lacZ+/0/Dlb-Ia/b mice. Induction of mutations by ethyl- and methylnitrosourea (ENU, MNU) and ethyl methanesulphonate (EMS) was investigated at 7 weeks after a single i.p. dose of each of these chemicals. In the small intestine, treatment with various dosages of ENU (10-150 mg/kg) resulted in a linear dose-response in both lacZ and Dlb-I. MNU (30 mg/kg) was also mutagenic in lacZ and Dlb-I, while EMS (250 mg/kg) did not significantly induce mutations in either gene. In spleen, ENU gave a linear dose-related response in both lacZ and hprt, MNU induced mutation sin both lacZ and hprt, and EMS was only positive for lacZ. No differences in response were observed between single and split-dose treatment with ENU (1 x 50 or 5 x 10 mg/kg with a 1- or 7-day interval), both in spleen and small intestine, except for lacZ in small intestine, where the single high dose gave a significantly higher induction than the split dose with the 7-day interval. The overall results suggest that mutagenic effects of fractionated doses are generally additive. In most cases, the induction factor (ratio treated over controls) for mutations in lacZ was lower than that for hprt and Dlb-I, presumably due to a higher background in lacZ and/or a lower mutability of lacZ. The general concordance between the data for lacZ and the endogenous genes indicates that lambda lacZ transgenic mice are a suitable model to study induction of gene mutations in vivo.

ß-Catenin activation synergizes with PTEN loss to cause bladder cancer formation.

Although deregulation of the Wnt signalling pathway has been implicated in urothelial cell carcinoma (UCC), the functional significance is unknown. To test its importance, we have targeted expression of an activated form of ß-catenin to the urothelium of transgenic mice using Cre-Lox technology (UroIICRE(+) ß-catenin(exon3/+)). Expression of this activated form of ß-catenin led to the formation of localized hyperproliferative lesions by 3 months, which did not progress to malignancy. These lesions were characterized by a marked increase of the phosphatase and tensin homologue (PTEN) tumour suppressor protein. This appears to be a direct consequence of activating Wnt signalling in the bladder as conditional deletion of the adenomatous polyposis coli (Apc) gene within the adult bladder led rapidly to coincident ß-catenin and PTEN expression. This PTEN expression blocked proliferation. Next, we combined PTEN deficiency with ß-catenin activation and found that this caused papillary UCC. These tumours had increased pAKT signalling and were dependent on mammalian target of rapamycin (mTOR). Importantly, in human UCC, there was a significant correlation between high levels of ß-catenin and pAKT (and low levels of PTEN). Taken together these data show that deregulated Wnt signalling has a critical role in promoting UCC, and suggests that human UCC that have high levels of Wnt and PI3 kinase signalling may be responsive to mTOR inhibition.

Fine structural alterations induced in the rat hepatocyte by adrenalectomy: influence of deoxycorticosterone acetate or fasting.

Electron microscopy was used to investigate the changes induced by bilateral adrenalectomy (ADX) in rat hepatocytes. Animals were killed at 1 week post-ADX, either with or without a prior 24 h fast. Further studies were carried out at 4 to 5 weeks after ADX to investigate the effect of administering deoxycorticosterone acetate (DOCA) for 4 weeks on the ADX-induced alterations. At 1 week, fasting had a profound effect on the response of hepatocytes to ADX in that either fasting or ADX alone caused proliferation of smooth endoplasmic reticulum (ER) and reduced glycogen but did not alter the rough ER, while ADX together with fasting resulted in the total depletion of glycogen, the virtual absence of smooth ER and the disaggregation of the rough ER into single cisternae. In contrast to the changes seen at 1 week after ADX in unfasted animals, at 4 weeks the hepatocytes contained increased glycogen in large, dense areas and had a scanty smooth ER, suggesting that some endogenous source of steroid hormones active in carbohydrate metabolism becomes available in male ADX rats by this time. This interpretation was consistent with the finding that DOCA-treatment of ADX rats prevented the glycogen accumulation seen at 4 weeks and also induced some smooth ER proliferation.

Determination of spatial patterns of DNA damage and repair in intestinal crypts by multi-cell gel electrophoresis.

We have developed a method to quantitate DNA strand breaks as a measure of DNA damage and repair in intact, isolated intestinal crypts. The assay is a modified form of the single-cell gel electrophoresis or 'comet' assay. By maintaining the spatial relationship between the cells we were able to characterise the repair response and the susceptibility to DNA damage of cells as a function of their position in the crypt. All cells were equally repair competent over the first 30 minutes of the repair of UV-C and gamma-radiation induced lesions. DNA damage was equally distributed following gamma-radiation but following incubation with the topoisomerase II inhibitor etoposide, damage was greater in the lower crypt with an unusual component to the comet tall which was tapered, implying an incremental change in susceptibility by cell position. This tapered component of the comet tail resolved rapidly after removal of etoposide. The pattern of damage produced by hydrogen peroxide was dose dependent with lower doses producing more strand breaks in the base of the crypt-an effect lost at higher doses. The assay has the ability to detect differences between cells in their susceptibility to DNA damage and their subsequent repair response which may vary with their proliferative or differentiative status.

Effect of fasting on aggregation of hepatocyte rough endoplasmic reticulum in adrenalectomized and 3'MeDAB-treated rats: quantitative electron microscope study.

A method for the quantitative analysis of the degree to which hepatocyte rough endoplasmic reticulum (ER) is arranged into parallel arrays was used to study the effect of fasting on rough ER aggregation in rat liver cells following either bilateral adrenalectomy or administration of the carcinogenic azo dye 3'-methyl-4-dimethylaminoazobenzene (3'MeDAB). One group of male inbred Leeds strain rats was subjected to bilateral adrenalectomy (ADX): after 1 week half of the animals were fasted for 24 h whereupon the whole group was killed. A second group of rats, fed for 4 weeks on a diet containing 0.06% of the carcinogenic azo dye 3'MeDAB, was similarly divided into two groups that were killed either with or without a prior 24-h fast. Untreated control groups of rats, both fasted and unfasted, were also killed. A quantitative electron microscope study was carried out to investigate the effect of each treatment on the degree to which the hepatocyte rough ER was aggregated into parallel arrays. ADX alone was without effect but caused a dramatic fall in rough ER aggregation when combined with fasting. At least as great an effect was induced by 3'MeDAB, with or without fasting, while fasting alone had a significant but much more modest effect than either ADX or the carcinogen. Thus, two disparate treatments induced morphologically identical responses in hepatocyte rough ER. The implications of this are discussed in terms of known interrelations between glucocorticoids and chemical carcinogenesis in the rat liver.

Quantitative electron microscopy of carcinogen-induced alterations in hepatocyte rough endoplasmic reticulum. II. Modulation of the effects of 3'MeDAB by adrenalectomy and adrenal corticosteroids.

A quantitative electron microscope study was made of the effects of bilateral adrenalectomy (ADX) and deoxycorticosterone acetate (DOCA) on the state of aggregation of hepatocyte rough ER cisternae in both untreated and 3'MeDAB-fed Leeds strain rats. Disaggregation of hepatocellular rough ER appears to be a common response of the rat liver to hepatocarcinogenic insult, while ADX and DOCA treatment are known to inhibit the chemical induction of neoplasia in this species. In untreated animals both ADX and DOCA significantly increased the mean number of cisternae per array or stack, while an even more pronounced effect was obtained from a combination of the two. The carcinogen 3'MeDAB induced a highly significant loss of aggregation, which was prevented by the combination of ADX and DOCA. In a separate experiment, a single dose of cortisone acetate was also found to partially reverse 3'MeDAB-induced rough ER disaggregation. In the rat hepatocyte, rough ER stacking or aggregation appears to be at least partially under hormonal control.

Analysis of DNA damage and repair accompanying differentiation in the intestinal crypt.

The ability to process damaged DNA may vary between cells depending on their differentiated status. However, there is little in vivo data available and it is not intuitively obvious how the activity of specific repair pathways may vary between different subpopulations (e.g. stem cells and proliferative, committed and differentiated cells) of a particular tissue. To obtain such information for the intestinal epithelium, we have developed an assay that detects differences in the way different regions of the crypt (stem, proliferative and maturation zones) respond to DNA damage. The assay is a variant of the 'comet' assay, which detects DNA strand breaks by measuring the proportion of DNA migrating from individual cells, or in this case intact isolated crypts, in an electrophoretic field. The method is quantitative, with the amount of migrating DNA being proportional to the number of strand breaks. Isolated crypts are repair competent and spatial differences are apparent with some agents. The assay has the potential to characterize the repair properties of cells at different stages of differentiation within the crypt, determine the characteristics that might predispose them to damage and may help in understanding the route of stem cell mutation.

Polyclonal origin of mouse skin papillomas.

We show that, from the earliest morphologically recognisable stages of development, mouse skin papillomas induced by a chemical initiation-promotion regime are polyclonal. We have demonstrated polyclonality directly by immunohistochemical staining of the mosaic cell populations in embryo aggregation chimaeras, a method which removes some of the uncertainties of previous conclusions based on analysis using electrophoretic polymorphisms. The findings may imply that initiation within a single cell and promotion of its clonal descendents is not a sufficient explanation for the origin of these tumours and that interaction between cells of more than one clone is involved.

Development of the pattern of cell renewal in the crypt-villus unit of chimaeric mouse small intestine.

We have previously shown that the epithelium of each adult intestinal crypt in chimaeric mice is derived from a single progenitor cell. Whether the crypts are monoclonal from the outset-that is, are formed by the proliferation of a single cell-or whether their formation is initiated by several cells was not known. Here we report that many crypts contain cells of both chimaeric genotypes in the neonatal period indicating a polyclonal origin at this stage of morphogenesis. The cellular organization of the early neonatal crypt is therefore different from that of the adult crypt, which includes a zone of 'anchored' stem cells above the crypt base. Within 2 weeks, however, the crypt progenitor cell and its descendants displace all other cells from the crypt and the crypt attains monoclonality. The distribution of enterocytes on chimaeric villi in the neonate shows a mottled pattern of mosaicism which is progressively replaced by coherent sheets of cells from the crypts, and within two weeks the orderly adult clonal pattern is established.

Effect of gamma radiation at high- and low-dose rate on a novel in vivo mutation assay in mouse intestine.

We have previously proposed that an in vivo mutagenicity assay could be based on the detection of mutations affecting the Dlb-1 locus in the small intestine of C57BL/6J x SWR F1 mice. F1 mice are heterozygous Dlb-1b/Dlb-1a and have only a single allele (Dlb-1b) which specifies expression of the binding site for the lectin Dolichos biflorus agglutinin (DBA) in intestinal epithelia. In whole-mount preparations of small intestine stained with a DBA--peroxidase conjugate, mutated stem cells and their progeny can be recognized as ribbons of unstained cells against a stained background. These DBA-negative ribbons can be quantified. The present investigation characterizes the responses of F1 mice to high- and low-dose rate gamma radiation (1.8 and 0.01 Gy/min respectively) and shows that the induction of ribbons is dose dependent in both cases. Dose sparing is evident at the lower dose rate: 20 ribbons/10(4) villi can be induced by 4 Gy at high-dose rate and by 7.1 Gy at low-dose rate, a dose sparing of 1.76. Age-matched untreated mice had a background level of 3.7 ribbons/10(4) villi. As expected, no ribbons were observed in homozygous (Dlb-1b/Dlb-1b) C67BL/6J mice, in which each stem cell has two alleles specifying lectin binding. The results further validate this system as a sensitive in vivo mutagenicity assay and suggests that the target cells are capable of repairing radiation damage.

Quantitative electron microscopy of carcinogen-induced alterations in hepatocyte rough endoplasmic reticulum. I. Chronic effect of 3'MeDAB and short-term effects of azo dyes of different carcinogenic potentials.

A new method has been developed for the quantitative analysis of an ultrastructural change in hepatocyte rough endoplasmic reticulum (ER) that is induced by liver carcinogens. Hitherto only subjective observations of this alteration had been made. Male inbred Leeds rats were fed a diet containing 0.06% of the carcinogenic azo dye 3'-methyl-4-dimethyl-aminoazobenzene (3'MeDAB). Groups of treated rats, together with untreated controls, were sacrificed after 10 days, 4 weeks, 10 weeks and 17 weeks. Samples of liver tissue from each animal and from 10 3'MeDAB-induced hepatocellular carcinomas (HCC), were examined by electron microscopy. A quantitative study was carried out to investigate the effect of chronic exposure to 3'MeDAB upon the state of aggregation of the hepatocyte rough ER into parallel arrays. As early as 10 days after the start of treatment, the rough ER showed a highly significant degree of disaggregation: the mean number of ER cisternae per array fell from the control value of 6.20 to 3.73. This change was sustained throughout the experiment. At 17 weeks, comparison of the mean array size in the HCC cells with that in the surrounding hepatocytes revealed a further significant decline, from 3.46 to 2.12. Further groups of rats were fed other azo dyes for 4 weeks and subjected to the same assay. The carcinogens 4'-methyl-4-dimethylaminoazobenzene (4'MeDAB) and N,N-dimethyl-4-amino-N-acetyl-N-monomethyl-4-aminoazobenzene (DAAMAB) resembled 3'MeDAB with respect to the degree of rough ER disaggregation they induced. In contrast, the non-carcinogen 3'-trifluoromethyl-4-dimethylaminoazobenzene (3'TFMeDAB) had no such effect, while the weak initiator 2-methyl-4-dimethylaminoazobenzene (2MeDAB) induced disaggregation to a lesser degree than the strong carcinogens. At least with the azo dyes used in this study, the extent of rough ER disaggregation appears to be related to hepatocarcinogenesis.