Liver fatty acid-binding protein: A marker for studying cellular differentiation in gut epithelial neoplasms

GASTROENTEROLOGY 1990;99:1727-1735

Liver Fatty Acid-Binding Protein: A Marker for Studying Cellular Differentiation in Gut Epithelial Neonlasms STEVEN L. CARROLL, KEVIN A. ROTH, and JEFFREY I. GORDON From the Departments of Pathology, Medicine, and Biochemistry and Molecular Biophysics, Washington University School of Medicine, St. Louis, Missouri

Human liver fatty acid binding protein is a 127 residue cytoplasmic protein synthesized in liver and in the intestinal epithelium. Previous studies of normal and transgenic mice indicated that the liver fatty acid-binding protein gene is a sensitive marker of enterocytic differentiation. This study shows the use of immunohistochemical methods to examine liver fatty acid-binding protein gene expression in normal human colonic epithelium, colonic villoglandular adenomas, nonmucinous and mutinous adenocarcinomas, and several types of noncolonic epithelial neoplasms. Cells containing liver fatty acid-binding protein were found in normal colonic epithelium, in two thirds of colorectal villoglandular adenomas and nonmucinous adenocarcinomas, and in one third of mutinous adenocarcinomas but not in noncolonic, nonhepatic carcinomas. All liver fatty acid-binding protein-positive colonic adenomas and adenocarcinomas contained patches of immunoreactive cells distributed among histologically identical patches of cells without liver fatty acid-binding protein immunoreactivity. This “mosaicism” was also found in metastases from liver fatty acid-binding proteinpositive colonic adenocarcinomas. Immunostaining of these liver fatty acid-binding protein-positive tissues for carcinoembryonic antigen did not show a mosaic cellular pattern in its expression. These data suggest that within a given neoplasm, differences exist in tbe differentiation programs of monoclonally derived, malignant colonic epithelial cells and that liver fatty acid-binding protein is a useful marker for operationally defining these subpopulations. Liver fatty acid-binding protein is also a potentially useful diagnostic marker for colorectal and hepatic carcinomas.

C

olorectal adenocarcinoma is the second leading cause of cancer death in the United States (1). The multistep hypothesis of cancer evolution suggests that neoplastic transformation results from multiple stepwise genetic lesions (2). Fearon et al. (3) have provided strong evidence that colorectal neoplasms are monoclonally derived.* This conclusion is based on their analysis of cellular patterns of X-chromosome inactivation in tumors isolated from female patients as well as the autosomal restriction fragment length polymorphisms observed in these specimens. An important observation made by this group was that even small adenomas (i.e., t5 mm in diameter] appeared to be monoclonal by these criteria (3). Bos et al. (4) and Forrester et al. (5) both observed a high frequency of K-ras in premalignant villus adenomas and in tumors containing mixtures of adenomas and carcinomas. Their studies provided evidence that mutations in the K-ras proto-oncogene may represent an early step in a multistep evolution to colorectal carcinoma (6-9) and support the conclusion of Fearon et al. (3) that colorectal cancers are monoclonal and arise from preexisting adenomas. Despite these insights, many fundamental questions remain unanswered concerning the pathogenesis of colorectal neoplasms. This reflects, in part, our limited understanding of the factors that regulate cell-specific, *tionoclonality does not necessarily imply that a single colonic epithelial cell has proliferated to form an adenoma. Selective advantage of one cell in a neoplastic population could also produce a monoclonal tumor. Abbreviations used in this paper: CEA, carcinoembryonic antigen; HGH. human growth hormone: L-FABP, liver fatty acidbinding protein. 0 1990 by the American Gastroenterological Association 0019-5095/99/$3.00

1728

CARROLL ET AL.

region-specific, and developmental stage-specific expression of genes in the continuously proliferating gut epithelium. Such information may provide critical clues about why, within the intestine, the colonic epithelium is preferentially susceptible to neoplastic transformation, which cell type undergoes initial neoplastic transformation, and to what extent the normal pathways for commitment and differentiation are recapitulated within these epithelial neoplasms. The adult intestinal epithelium is a complex monolayer that undergoes continuous regeneration of its four principal, terminally differentiated cell types [polarized, absorptive enterocytes, mucin-secreting goblet cells, enteroendocrine cells, and Paneth cells] from multipotent stem cells located in the crypt (lo14). The process of proliferation/differentiation/ translocation/extrusion is very rapid, averaging 3 and 5 days in the mouse and human small intestine, respectively (11). Regional differences in gene expression are established and maintained in this rapidly proliferating epithelium resulting in the formation of gradients of gene expression from the crypts of Lieberktihn to the villi and from duodenum to colon (14). Previous studies have used the liver fatty acid-binding protein (L-FABP) gene as a model to map cis-acting DNA sequencest which regulate cell-specific, region-specific, and developmental stage-specific expression of genes in the gut epithelium (15,16). The L-FABP gene is efficiently expressed in hepatocytes and enterocytes (15,17). Its 15-kilodalton protein product is thought to participate in the uptake of long-chain fatty acids and/or their targeting to sites of metabolic processing (18). Studies of the cellular patterns of L-FABP accumulation in fetal life suggest that it is a very sensitive marker of enterocytic differentiation (19). Analysis of transgenic mouse pedigrees containing different portions of the 5’ nontranscribed region of the L-FABP gene linked to a reporter [the human growth hormone (HGH) gene] showed the following: (a] that several different elements are required to generate an appropriate proximal to distal gradient in gene expression; (b) that distinct mechanisms may be responsible for generating gradients from crypt to villus and from duodenum to colon; and (c) that there appear to be subtle differences in the regulatory environments of some crypt epithelial cell populations that are shown by aberrant patterns of transgene expression involving all cells in a given crypt (15). These differences, which affected up to -1% of colonic crypts in some peditCis-acting DNA sequences regulate the transcriptional activity of the genes to which they are linked. Trans-acting factors (proteins) bind to specific c&acting sequences. Genes contain a variety of c&acting sequences. Production of the appropriate combination of positive and/or negative acting trans-acting factors by a cell will “unlock” (modulate) the transcriptional activity of a given gene.

GASTROENTEROLOGY Vol. 99, No.6

grees, raised the possibility that there may be variations in the levels of positive or negative trans-acting factors in the monoclonal (12) stem cell population and/or that these differences are manifested during cellular commitment/differentiation (15). We have now examined L-FABP expression in human colorectal neoplasms to determine if L-FABP was either a sensitive or specific marker for human colorectal carcinoma, to correlate its expression with the state of differentiation of these tumors, and finally to determine if the cellular patterns of L-FABP accumulation within a given neoplasm could provide clues about the evolution of this type of cancer.

Materials and Methods Source of Neoplasms Formalin-fixed, paraffin-embedded tissues from 54 colorectal adenocarcinomas [representing 53 patients, 48% male, age range 30-88 years (median, 88.4 years)], 10 colorectal adenomas, and 38 cases of noncolonic epithelial neoplasms were selected from the files of the Barnes Hospital (St. Louis, MO) Department of Pathology. Among the colorectal adenocarcinomas, 15 were classified as Duke’s stage A, 19 as stage B, 15 as stage C, and 5 as stage D (distant metastases]. Metastases were examined in the 5 patients from this latter group. Note that the designation of colorectal adenocarcinomas as nonmucinous or mutinous was based on the histological features defined in reference 20. Similarly, classification of adenomas as villoglandular followed well established pathological criteria (20).

lmmunocytochemical

Methods

The immunogold-silver staining technique (21) was used to examine the cellular patterns of L-FABP and carcinoembryonic antigen (CEA] expression/accumulation in these tissues. Carcinoembryonic antigen is a highmolecular weight glycoprotein that is expressed in a variety of neoplasms (22-24). A polyclonal rabbit anti-L-FABP serum was used. It was raised against rat L-FABP expressed in Escherichia coli using a recombinant prokaryotic expression vector (25). The preparation and specificity of this antiserum has been described in other publications (15,18,28). A polyclonal, rabbit anti-CEA serum was obtained from Dako (Santa Barbara, CA, Catalog #A115). Five-micron thick sections were cut and dried at 80°C. Following removal of paraffin and rehydration, these sections were preincubated at room temperature for 20 minutes in “blocking buffer” [phosphate-buffered saline (PBS) containing 5% normal goat serum, 2% bovine serum albumin, and 0.3% Triton X-100]. They were then treated overnight at 4’C with L-FABP antiserum (diluted 1:500 in blocking buffer] or CEA antiserum (1:lOOO dilution]. Following several rinses with PBS, antigen-antibody complexes were visualized by incubating sections with a 1:160 dilution of colloidal gold-conjugated goat anti-rabbit antibody (Janssen Life Science Products, Piscataway, NJ] for 1 hour at room

December 1990

DIFFERENTIATIONIN GUT EPITHELIALNEOPLASMS 1729

temperature followed by silver enhancement (Janssen Life Science Products]. Sections were then lightly counterstained with hematoxylin or H&E.

results that expression of the L-FABP gene in normal human colonic epithelium only occurs after commitment of the descendants of the pluripotent stem cells to differentiate into enterocytes, resulting in a gradient of L-FABP accumulation from colonic crypt to surface epithelial cuff. Sixty-eight percent of the nonmucinous adenocarcinomas examined contained epithelial cells that reacted with our antibody (Table 1). Figure 1C and D demonstrates the cellular patterns of L-FABP accumulation in a moderately differentiated nonmucinous adenocarcinoma which was adjacent to the normal colonic epithelium shown in Figure 1A and B. As was typical for all 28 L-FABP-positive nonmucinous adenocarcinomas studied, the distribution of immunostaining was remarkably focal. Patches of intensely staining cells were seen adjacent to regions of histologically identical, nonimmunoreactive cells. Within a given patch of L-FABP immunoreactive cells, there was typically little variation in staining intensity between cells. This “mosaic” [patchy) expression was not an artifact of histological processing as there was considerably less cellular variation in CEA immunoreactivity in the 93% of L-FABP-positive nonmucinous adenocarcinomas that expressed this marker [compare Figure 1D and E which show the patterns of L-FABP and CEA staining, respectively, in the same tumor]. Cellular CEA levels were also very homogeneous in L-FABP-negative nonmucinous adenocarcinomas (data not shown]. The specificity of the L-FABP immunoreactivity in nonmucinous adenocarcinomas was confirmed by the absence of staining when the primary antiserum was replaced with preimmune serum or when it was preabsorbed with purified rat L-FABP (data not shown). The identity of the antigen as L-FABP was further established by Western-blot analysis of lysates from L-FABP-positive [Figure 2, lane 3) and -negative (Figure 2, lane 5] tumors. Liver fatty acid-binding protein expression was most apparent in well and moderately differentiated nonmucinous colorectal adenocarcinomas. However, it was not limited to these types: epithelial cells in poorly differentiated nonmucinous adenocarcinomas also demonstrated patches of L-FABP-positive cells (Figure 1G). Liver fatty acid-binding protein-positive cells were also detected in five of the thirteen mucinous colorectal adenocarcinomas we studied (Figure 1H and Table 1) Once again a mosaic pattern of immunostaining was consistently observed. Metastases from L-FABP-positive neoplasms also showed this mosaicism [comparable distant metastases from L-FABP-negative nonmucinous adenocarcinomas did not express this intracellular fatty acid-binding protein). The mosaic pattern of L-FABP staining in adenocarcinomas does not appear to be a

Western-Blot

Hybridization

Frozen surgical specimens of normal colon, nonmucinous colorectal adenocarcinomas, and mutinous colorectal adenocarcinomas were thawed and homogenized at 0°C with a motor-driven Teflon probe in lysis buffer (PBS, 1% Triton X-100, 1 mmol/L phenylmethylsulfonylfluoride; 15 pL/mg wet weight of tissue]. Following a lo-minute incubation on ice, cellular debris was removed by centrifugation at ~O,OOO x g for 5 minutes. The concentration of protein in the resulting supernatants was determined using the method of Bradford (27). Samples of lysate proteins (50 pg] were then reduced, denatured, fractionated by electrophoresis through a 15% polyacrylamide gel containing sodium dodecyl sulphate (281,and transferred to nitrocellulose filters by electroblotting (29). Filters were rinsed in a solution of NaCl (0.15 mol/L], Tris (10 mmol/L, pH 7.5), and Tween-20 (0.04%], then dried, and stained with Ponceau S (Sigma Chemical Co., St. Louis, MO) to visualize proteins. The filters were subsequently destained in distilled water and treated for 30 minutes with Blotto (301. After the blocking reaction was completed, filters were incubated with our polyclonal antirat L-FABP serum (diluted 1:lOOOin Blotto). Antigenantibody complexes were visualized using “?-protein A as previously described (31). Results Figure 1A and B shows the cellular pattern of L-FABP accumulation in a region of normal colonic epithelium located adjacent to a nonmucinous adenocarcinoma. Immunoreactive protein was only detected in columnar epithelial cells (enterocytes) located near the crypt orifice and in the surrounding cuff of surface epithelium. Goblet cells located in the upper portion of the crypt were not stained (see arrows in Figure 1B) nor were epithelial cells situated within the deeper regions of the crypt [Figure 1A). The 127 residue long human and rat L-FABPs share 82% sequence identity (32). We established that the cellular pattern of staining shown in Figure 1A and 1B arose from the specific interaction of our rat L-FABP antibody with human L-FABP by performing several control experiments. First, no signal was produced when the primary antibody was replaced with preimmune serum. Second, antiserum preabsorbed with an excess of purified rat L-FABP (26) did not show staining (data not shown). Last, Western-blot hybridization analysis of total cell lysates prepared from normal human colon showed that the antibody reacted with a unique protein [Figure 2, lane 4) that comigrated with mouse L-FABP [lane 6), purified rat L-FABP (lanes 1 and 21, and purified human L-FABP (data not shown]. Thus we could conclude from the immunostaining

1730

CARROLL ET AL.

GASTROENTEROLOGY Vol. 99, No. 6

December

DIFFERENTIATION

1990

consequence of malignant transformation. Figure 1F shows the heterogenous pattern of cellular L-FABP accumulation in a colorectal villoglandular adenoma. As with the adenocarcinomas, this degree of mosaicism was not observed when adjacent sections of the same neoplasm were stained for CEA (data not shown]. Our survey of nonmucinous adenocarcinomas indicated that 32% did not contain L-FABP immunoreactive cells. By contrast, 93% of the 41 tumors sampled were CEA positive. This latter value is comparable to previous surveys of CEA expression in nonmucinous colonic adenocarcinomas (33). No morphological differences were noted between L-FABP-positive and L-FABP-negative nonmucinous (or mutinous) colorectal adenocarcinomas. The absence of L-FABP immunoreactivity in nonmucinous or mutinous adenocarcinomas was not merely an artifact of fixation because, in all cases, adjacent normal colonic epithelium demonstrated a normal pattern of cellular L-FABP expression. Similar percentages of nonmucinous adenocarcinomas (68%) and colorectal adenomas (60%) were L-FABP-positive [see Table 1). This result suggests that the pattern of L-FABP expression in adenocarcinomas is established before malignant transformation. The mosaic pattern of L-FABP expression does not appear to represent a nonspecific consequence of malignant transformation: we failed to observe L-FABP immunoreactivity in 38 noncolonic epithelial neoplasms. This latter group included neoplasms from eight other tissues (breast, kidney, stomach, pituitary, pancreas, endometrium, prostate, and lung], 39% of which contained CEA (Table 1). As expected, L-FABP immunoreactivity was found in hepatocytes and hepa-

IN GUT EPITHELIAL NEOPLASMS

1731

tocellular carcinomas (unpublished observations). These latter observations raise the possibility that L-FABP may represent a useful diagnostic marker for colorectal adenocarcinomas and hepatocellular carcinomas. Discussion Our studies indicate that L-FABP is expressed in approximately two thirds of colorectal villoglandular adenomas and nonmucinous adenocarcinomas. in approximately one third of colorectal mutinous adenocarcinomas, and rarely, if ever, in the several types of noncolonic, nonhepatic neoplasms we surveyed. Within a given colonic adenoma or adenocarcinoma, a remarkably mosaic pattern of L-FABP expression was noted among morphologically similar cells. In contrast, the cellular patterns of CEA immunoreactivity were more monotonous. Our findings suggest several conclusions. First, L-FABP immunoreactivity shows a remarkable diversity in the degree of phenotypic differentiation of cells that are represented within a given colonic tumor. Therefore, the L-FABP gene is a useful reporter that can be exploited to operationally define different subpopulations of cells within a monoclonally-derived colonic neoplasm. At the present time, L-FABP appears to be a unique marker for these subpopulations. An earlier set of immunohistochemical studies of blood group antigens and several, poorly described carcinoma-associated antigens showed heterogeneity in their cellular staining patterns in both rectal adenocarcinomas (35) and adenomas (36). These antigens were detected in a few scattered positive cells, in

Figure 1. Cellular patterns of GFABP expression in normal colonic epithelium and in colorectal adenocarcinomas. The immunogold silver-staining technique (21) was used to visualize GFABP immunoreactivity as a dark black precipitate. Nuclei were counterstained with hematoxylin. A and B. Expression of L-FABP in sections of normal human proximal colon. Immunoreactive protein is confined to enterocytes located in the upper portions of the crypt and a cuff of surface epithelium surrounding the crypt orifice. Liver fatty acid-binding protein is not detectable in the goblet cell indicated by the arrow in B [original magnification xl60 and x400, respectively]. Note that comparable cellular patterns of L-FABP staining also occur in the distal colon (17). C. Heterogenous tion x 250).

patterns of L-FABP accumulation

in a moderately well-differentiated,

nonmucinous

adenocarcinoma

[original magnifica-

D. Higher power (x400) view of lesion shown in C emphasizing the intense staining present in clusters (patches] of epithelial cells. E. Same nonmucinous adenocarcinoma that showed a patchy pattern of cellular L-FABP accumulation in D, now stained with antibodies directed against CEA [original magnification x400). Note the similar levels of immunoreactive protein between cells. Within cells, CEA shows a polarized distribution. This has been noted previously in normal colonic epithelial cells (23). F. Heterogenous

L-FABP staining in a villoglandular

adenoma [original magnification

x 400).

G. A less-well differentiated nonmucinous adenocarcinoma still demonstrates patchy patterns of cellular L-FABP accumulation [original magnification x400). The adjacent normal epithelium [not seen in this field] showed a pattern of L-FABP expression identical to that illustrated in A and B. H. Mosaic L-FABP staining in a mutinous colorectal adenocarcinoma (original magnification x250). The closed arrow shows a cluster of L-FABP-positive cells while the open arrows show a patch of morphologically similar cells without detectable levels of L-FABP.

1732 CARROLL ET AL.

I

2

GASTROENTEROLOGY Vol. 99, No. 6

3

4

5

6

Figure 2. Western-blot analysis of intestinal extracts using an anMsera raised against purified E. co&derived rat LFABP. Proteins contained in tissue lysates prepared from normal colonic mucosa, an LFABP-positive nonmucinous colorectal adenocarcinoma, and an LFABP-negaMve mutinous colorectal adenocarcinoma were denatured, reduced, and fractionated by SDSpolyacrylamide gel electrophoresis before transfer to nitrocellulose filters. Blots were incubated with a rabbit anti-rat LFABP serum and antigen-antibody complexes visualkd by labeling with lzsI protein A as described in Methods. An autoradiograph of the blot is shown after a 2day exposure at -70%. Lanes 1 and 2 show 500 and 100 ng of rat LFABP purified from E. coli containing a recombinant expression vector derived from pPI245 (25) in lane 3, 50 wg of protein were contained in a tissue extract of a human nonmudnous colorectal adenocarcinoma that showed a mosaic cellular pattern of LFABP immunoreactivity; lane 4 shows 50 r(8 of total protein contained in a lysate of normal human colonic mucosa; lane 5 shows 50 fig of protein present in an extract of a human mutinous colorectal adenocarcinoma that contained no LFABP-positive cells; and in lane 6, extract prepared from normal mouse liver (10 pg of protein) is shown. Note that the entire length of the gel is represented in the autoradiograph.

groups of cells, in large foci and/or in combinations of the above. Unlike L-FABP, a similar heterogeneous pattern of immunoreactivity was also observed in normal (rectal) mucosa. Antigen expression was neither cell type- nor region-specific, occurring at any point from the base of the crypt to the surface epithelium (36). These antigens are expressed in a variety of neoplasms (37,35). Thus, the heterogeneous cellular patterns of blood group- and carcinoma associatedantigen expression do not appear to be related to enterocytic differentiation, and may not be comparable to the mosaic pattern of L-FABP immunoreactivity we observed in colorectal adenomas and adenocarcinomas. We speculate that the (regulatory) mechanisms that produce blood group- and carcinoma associated-

antigen heterogeneity are distinct from those that underlie L-FABP mosaicism. Liver fatty acid-binding protein expression also appears to differ from that of other oncofetal markers such as crypt cell antigens (38). These antigens are not found in normal colonic epithelial cells but are patchily expressed in colonic adenocarcinomas (301. A mucin-type glycoprotein is produced in goblet cells in normal colon and focally in colonic adenocarcinomas (39). Unlike L-FABP, expression of this antigen is lost in poorly differentiated adenocarcinomas. Moreover, focal staining does not occur in adenomas. The second conclusion from our findings is that the diversity in cellular L-FABP expression is manifest at one of the earliest stages in the multistep evolution to colorectal cancer, i.e., in adenomas. The mosaic pattern of L-FABP expression in adenomas implies that during expansion of these monoclonal tumors, foci of cells either retain or lose their ability to support synthesis of this cytoplasmic protein. There are a number of potential mechanisms that could produce this mosaicism. Because the mosaicism in L-FABP-positive colonic adenocarcinomas was also present in their distant metastases, it appears that the factor(s) that are responsible for this phenomenon are intrinsic to the neoplasm and not dependent upon direct interactions with luminal (or antiluminal) derived substances that may affect either the normal intestinal epithelium or the primary tumors. The presence or absence of L-FABP could arise from mutations in the L-FABP structural gene or in genes that encode positive or negative trans-acting factors that influence L-FABP expression. If such an event(s) occurs early in the evolution of these lesions, an adenoma containing Table 1. Expression of Liver Fatty Acid-Binding Protein and Carcinoembryonic Antigen in Colorectal Adenocarcinomas, Villoglandular Adenomas, and Noncolonic,

Nonhepatic

Type Colorectal adenocarcinomas Nonmucinous Mutinous Colorectal villoglandular adenomas Noncolonic neoplasmsb

Carcinomas

No. of neoplasms examined

L-FABP

CEA”

41 13

28(68%) 5(38%)

38(93%) 10(77%)

10 38

6 (60%] 0

8(80%) 15 (39%)

“No. of neoplasms containing immunoreactive cells (percent positive neoplasms). these include formalin-fixed, paraffin-embeddedtissues from four cases of renal cell carcinoma, five cases of invasive ductal breast carcinoma, one gastric adenocarcinoma, one case of endometrial adenocarcinoma, one case of gastric carcinoid, three pancreatic neuroendocrine neoplasms, eight pituitary adenomas. one case of prostatic adenocarcinoma, and fourteen cases of lung carcinoma (one undifferentiated, three epidermoid. six neuroendocrine, and four adenocarcinomas].

December

1990

DIFFERENTIATION

IN GUT EPITHELIAL NEOPLASMS

1733

A third explanation for the observed L-FABP mosaa cellular population without any detectable L-FABP icism is that it represents a recapitulation of the could result (33% of adenomas surveyed fall into this normal, early developmental pattern of cytodifferenticlass). If this event occurred later in the evolution of ation of the intestinal epithelium. Examination of the adenomas/adenocarcinomas, a mosaic population of has late gestational fetal rat and mouse gut (16,191 L-FABP-positive and L-FABP-negative cells could be shown that a complex cellular pattern of L-FABP gene generated, In this case, L-FABP would represent a activation is initiated coincident with the first appearmarker for mutations induced in subclones of the ance of a well-organized epithelial monolayer overlymonoclonally-derived lesions (40). The human L-FABP gene maps to the 12p-llq region of chromosome 2 (41) ing nascent villi (this occurs between fetal days 16 and 18 of the 21-2%day gestation). Some columnar epitheand therefore does not fall within regions of frequent lial cells express high levels of L-FABP whereas allelic loss that have been previously identified in morphologically similar cells, occupying similar posisporadic colorectal adenomas and adenocarcinomas tions along the intervillus space-to-villus tip axis, have (i.e., in chromosomes 5, 17p, and 181 (6-9,42). We low to absent levels of the protein. This can be cannot exclude the possibility that the structural gene is located within a region of previously unrecognized recapitulated in fetal transgenic mice when nucleallelic loss or that the known deletions associated with otides -4000 to + 21 of the rat L-FABP gene are linked colorectal adenomas/adenocarcinomas perturb an alThe heterogeneous cellular pattern of to HGH (16). lele operating in trans to regulate L-FABP expression. L-FABP expression resolves by the first postnatal day. As noted in the Introduction, studies with transgenic Both the proximal and distal portions of the gastrointesmice containing domains from the 5’ nontranscribed tinal tract have this phenomenon but its onset and region of the rat L-FABP gene linked to a reporter resolution in the distal half is delayed by l-2 days HGH showed that animals from multiple pedigrees, relative to the proximal bowel. Not all genes that are known to be heterozygotes for the transgenes, exhibit activated at this time in fetal life and expressed in aberrant patterns of HGH expression affecting all the differentiated enterocytes will demonstrate initial hetmonoclonally-derived cells in a given colonic crypt erogeneous patterns of expression (19). Thus, L-FABP (15). Like the mosaicism seen in colorectal adenomas may be a sensitive marker of this phenomena. Unforand nonmucinous adenocarcinomas, this may reflect a tunately, comparative developmental studies of the stochastic phenomenon-one to which the L-FABP cellular patterns of L-FABP and CEA activation in the gene is particularly sensitive. human (fetal] intestinal epithelium have not been A second explanation for the patchy pattern of reported. L-FABP accumulation is that it reflects a cell-cycle The mosaic pattern of cellular differentiation noted effect. Increases in L-FABP levels have been previduring development and in colorectal adenomas and ously noted to occur in normal mouse and rat hepatoadenocarcinomas has also been observed in a cultured cytes undergoing mitosis (15,43). Studies with transhuman cell line derived from a colonic adenocarcigenie mice indicate that the signals that produce this noma. HT-29 cells can,be induced to differentiate into effect are mediated by cis-acting elements located enterocytelike cells by removing glucose as a carbon between nucleotides -596 and +21 of the gene (15). source from medium (45). Even though monoclonally High levels of L-FABP occur throughout interphase derived HT-29 cells all express a colon-specific antiduring proliferation of hyperplastic and malignant rat gen (517) on their apical surfaces when they differenhepatocytes induced by treatment with certain carcintiate (46), only a fraction (20%) express sucraseogens (N-fluorenyacetamide and 3’-methyl-4-dimethisomaltase (47). Thus, this cell line represents one ylaminoazobenzene) (43). These compounds are also model that may help identify factors responsible for bound by L-FABP. Based on these findings, Bassuk et producing the mosaic patterns of intestinal differential. (43) proposed that L-FABP may interact with a class ation noted in colorectal neoplasms. Transgenic mouse of ligands that promote cell division.$ Studies by models of colorectal carcinoma using the L-FABP others (17) have shown that normal colonic enteropromoter to regulate expression of oncogenes in the cytes finally begin to express L-FABP coincident with gut epithelium may allow us to recapitulate this heterloss of their ability to replicate (defined by their ogeneity in vivo and to identify cis-acting elements inability to incorporate tritiated thymidine). Comparathat contribute to this phenomenon (14). ble analyses are needed in L-FABP-positive colonic adenomas and adenocarcinomasto determine if there is a correlation between protein accumulation and the References cell cycle. 1. Cancer

/A homologous protein known as mammary-derived growth inhibitor also affects proliferation of cultured mammary carcinoma cells (441.

facts and figures-1988. New York: American Cancer Society, 1988. 2. Land H, Parada LF, Weinberg RA. Cellular oncogenes and multistep carcinogenesis. Science 1983;222:771-778.

1734

CARROLL ET AL.

3. Fearon ER, Hamilton SR, Vogelstein B. Clonal analysis of human colorectal tumors. Science 1987;238:193-197. 4. Bos JL, Fearon ER, Hamilton SR, Verlaan-de Vries M, van Boom JH, van der Eb AJ, Vogelstein B. Prevalence of ras gene mutations in human colorectal cancers. Nature (Land) 1987327: 293-297. 5. Forrester K, Almoguera C, Han K, Grizzle WE, Perucho M. Detection of high incidence of K-ros oncogenes during human colon tmnorigenesis. Nature (Lond) 1987;327:298-303. 6. Vogelstein B, Fearon ER, Hamilton SR, Kern SE, Preisinger AC, Leppert M, Nakamura K. White R, Smits AMM. Bos JL. Genetic alterations during colorectal-tumor development. N Engl J Med 1988;319:525-532. 7. Vogelstein B, Fearon ER. Kern SE, Hamilton SR, Preisinger AC, Nakamura Y, White R. Allelotype of colorectal carcinomas. Science 1989;244:207-211. 8. Kern SE, Fearon ER, Tersmette KWF, Enterline JP. Leppert M, Nakamura Y, White R, Vogelstein B, Hamilton SR. Allelic loss in colorectal carcinoma. JAMA 1989;261:3099-3103. 9. Baker SJ. Fearon ER, Nigro JM, Hamilton SR, Prelsinger AC, Jessup JM, van Tuinen P, Ledbetter DH, Barker DF, Nakamura Y, White R, Vogelstein B. Chromosome 17 deletions and p53 gene mutations in colorectal carcinomas. Science 1989;244:217221. 10. Cheng H. Leblond CP. Origin, differentiation, and renewal of the four main epithelial cell types in the mouse small intestine V. Unitarian theory of the origin of the four epithelial cell types. Am J Anat 1974;141:537-562. 11. Potten CS, Hendry JH. Stem cells in murine small intestine. In: CS Potten, ed. Stem cells. Edinburgh: Churchill Livingstone 1983;155-199. 12.Ponder BAJ, Schmidt GH, Wilkinson MM, Wood MJ, Monk M, Reid A. Derivation of mouse intestinal crypts from single progenitor cells. Nature (Lond). 1985;313:689-691. 13. Schmidt GH. Wilkinson MM, Ponder BAJ. Cell migration pathway in the intestinal epithelium: an in situ marker system using mouse aggregation chimeras. Cell 1985;40:425-429. 14. Gordon JI. Intestinal epithelial differentiation: new insights from chimeric and transgenic mice. J Cell Biol 1989;108:11871194. 15.Sweetser DA, Birkenmeier EH, Hoppe PC, McKee1 DW, Gordon JI. Mechanisms underlying generation of gradients in gene expression within the intestine: an analysis using transgenic mice containing fatty acid binding protein/human growth hormone fusion genes. Genes Dev 1988;2:1318-1332. 16.Hauft SM. Sweetser DA, Rotwein PS. Lajara R, Hoppe PC, Birkenmeier EH, Gordon JI. A transgenic mouse model that is useful for analyzing cellular and geographic differentiation of the intestine during development. J Biol Chem 1989;264:84198429. 17. Suzuki T, Hitomi M. Ono T. Immunohistochemical distribution of hepatic fatty acid binding protein in rat and human alimentary tract. J Histochem Cytochem 19884349-357. 18.Sweetser DA, Heuckeroth R, Gordon JI. The metabolic significance of mammalian fatty acid binding proteins: abundant proteins in search of a function. Annu Rev Nutr 1987;7:337-359. 19. Rubin DC, Ong DE, Gordon JI. Cellular differentiation in the emerging fetal rat small intestinal epithelium: mosaic patterns of gene expression. Proc Nat1 Acad Sci USA 1989;86:1278-1282. 20. Rosai J. Ackerman’s Surgical Pathology. St. Louis: Mosby, 1989;668-611. 21. Holgate CS, Jackson P, Cowen PN, Bird CC. Immunogold-silver staining: new method of immunostaining with enhanced sensitivity. J Histochem Cytochem 1983;31:938-944. 22. Isaacson P, LeVann HP. The demonstration of carcinomembryonic antigen in colorectal carcinoma and colonic polyps using an immunoperoxidase technique. Cancer 1976381348-1356.

GASTROENTEROLOGY

Vol. 99, No. 6

23. Ahnen DJ, Nakane PK. Brown WR. Ultrastructural localization of carcinoembryonic antigen in normal intestine and colon cancer: abnormal distribution of CEA on the surfaces of colon cancer cells. Cancer 1982;49:2077-2090. 24. Zamcheck N. Kupchik HZ. Summary of clinical use and limitations of the carcinoembryonic antigen assay and some methodologic considerations. In: N.R. Rose, H. Friedman, eds. Manual of clinical immunology. 2nd ed. Washington, DC: American Society of Microbiology, 1980;919-935. 25. Lowe JB, Strauss AW, Gordon JI. Expression of a mammalian fatty acid binding protein in Escherichia coli. J Biol Chem 1984;259:12696-12704. 26. Lowe JB, Sacchettini JC, Laposata M, McQuillan JJ. Gordon JI. Expression of rat intestinal fatty acid binding protein in Escherichia coli: purification and comparison of ligand binding characteristics with that of Escherichia coli-derived rat liver fatty acid binding protein. J Biol Chem 1987;262:5931-5937. 27. Bradford MM. A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 1976;72:248-254. 28. Laemmli UK. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature (Land) 1970;227:680685. 29. Towbin H, Staehelin T, Gordon J. Electrophoretic transfer of proteins from polyarylamide gels to nitrocellulose sheets: procedure and some applications. Proc Nat1 Acad Sci USA 197976: 4350-4354. 30. Johnson DA, Gautsch JW, Sportsman JR, Elder JH. Improved technique utilizing nonfat dry milk for analysis of proteins and nucleic acids transferred to nitrocellular. Gene Anal Tech 1984;1:3-8. 31. Burnette WN. “Western blotting”: electrophoretic transfer of proteins from sodium dodecyl sulfate-polyacrylamide gels to unmodified nitrocellulose and radiographic detection with antibody and radioiodinated protein A. Anal Biochem 1981;112:195203. 32. Lowe JB. Boguski MS. Sweetser DA, Elshourbagy NA, Taylor JM, Gordon JI. Human liver fatty acid binding protein: isolation of a full length cDNA and comparative sequence analyses of orthologous and paralogous proteins. J Biol Chem 1985;260:34133417. 33. Zamchek N, Moore TL, Dhar P, Kupchik H. Immunologic diagnosis and prognosis of human digestive tract cancer. Carcinoembryonic antigens. N Engl J Med 1972;286:83-86. 34. Enblad P, Glimelius B, Busch C, Pontett J, Pahhnan L. Antigenic heterogeneity and individuality in adenocarcinomas of the rectum and their secondaries. Br J Cancer 1987;55:503-508. 35. Cooper HS, Haesler WE. Blood group substances as tumor antigens in the distal colon. Am J Clin Path 1978:69:594-598. 36. Enblad P, Busch C, Carlsson U, Erelund G. Glimelius B, Lindstr6m C, Ponteti J, Pahlman L. The adenoma-carcinoma sequence in rectal adenomas: support by the expression of blood group substances and carcinoma antigens. Am J Clin Path 1988;90:121-130. 37. Laurence DJR, Neville AM. Fetal antigens and their role in the diagnosis and clinical management of human neoplasms: A review. Br J Cancer 1972;26:335-355. 38. Quaroni A, Weiser MM, Herrera L, Fay D. Crypt cell antigens (CCA]: new carbohydrate markers for human colon cancer cells. Immunol Invest 1989;18:391-404. 39. Gangopadhyay A, Barlow JJ, Petrelli NJ, Tsukada Y, Bhattacharya-Chatterjee M. Expression of a high molecular weight mucin-type glycoprotein in human colon cancer as defined by monoclonal antibody lD,. Hybridoma 1988;7:141-154. 40. Solomon E, Voss R, Hall V, Bodmer WF, Jass JR, Jeffreys AJ, Lucibello FC, Pate1 I, Rider SH. Chromosome 5 allele loss in human colorectal carcinomas. Nature (Lond) 1987;328:616-619.

December 1990

41. Sweetser DA, Birkenmeier EH, Klisak IJ, Zollman S, Sparkes RS, Mohandas T, Lusis AJ, Gordon JI. The human and rodent intestinal fatty acid binding protein genes: a comparative analysis of their structure, expression and linkage relationships. J Biol Chem 1967:262:16060-16071. 42.Winton DJ, Blount MA, Ponder BAJ. A clonal marker induced by mutation in mouse intestinal epithelium. Nature 1988;333:463466. 43. Bassuk JA, Tsichlts PN, Sorof S. Liver fatty acid binding protein is the mitosis-associated polypeptide target of a carcinogen in rat hepatocytes. Proc Nat1 Acad Sci USA 1967;84:7547-7551. 44. Boehmer FD. Kraft R, Otto A, Wernsted C, Hellman U. Kurtz A. Mttller T, Rohde K, Etzold G, Lehmann W, Langen P, Heldin C-H, Grosse R. Identification of a polypeptide growth inhibitor from bovine mammary gland: sequence homology to fatty acid and retinoid binding proteins. J Biol Chem 1987;262:1513715143. 45. Pinto M, Appay MD, Simon-Assman P, Chevalier G, Dracopuli N, Fogh J, Zweibaum A. Enterocytic differentiation of cultured human colon cancer cells by replacement of glucose by galactose in the medium. Biol Chem 1982:44:193-196.

DIFFERENTIATION

IN GUT EPITHELIAL NEOPLASMS

1735

46. LeBivic A, Hirn M, Reggio H. HT-29 cells are an in vitro model for the generation of cell polarity in epithelia during embryonic differentiation. Proc Nat1 Acad Sci USA 1988;85:136-140. 47. Huet C, Sahuquillo-Merino C, Coudrier E, Louvard D. Absorptive and mucus-secreting subclones isolated from a multipotent cell line (HT-29) provide new models for cell polarity and terminal differentiation. J Cell Biol1987;105:345-357.

Received December 14,1989. Accepted June 27.1990. Address requests for reprints to: Jeffrey I. Gordon, Department of Biochemistry and Molecular Biophysics, Washington University School of Medicine, 660 South Euclid Avenue, Box 8231, St. Louis, Missouri 63110. Supported by a grant from the National Institutes of Health (DK30292, J.I.G.) and an award from the Markey Foundation (K.A.R.). The authors thank Carol Langner for her assistance with the Western-blot analysis.