Homeobox gene expression in the intestinal epithelium of adult mice.

Using a polymerase chain reaction-based strategy, we have detected the expression of nine different homeobox genes in adult mouse intestine. Included among these are the recently described intestine-specific Cdx-1 gene and a new, related gene, Cdx-2. Southern blot experiments show that Cdx-2 is present in a single copy in the mouse genome. Of several adult mouse tissues assayed, intestine was the only one that contained detectable levels of Cdx-2 mRNA. Expression of all nine homeobox genes in different regions of the intestine was quantitated by RNase protection analysis, which revealed a unique expression profile for each gene. These observations suggest that homeobox gene expression may play an important role in cellular differentiation in the adult intestine.

The intestinal epithelium is a tissue endowed with an enormous proliferative and differentiative capacity; in fact, some estimates suggest that over a 70-year lifespan an individual produces some 6-7 tons of intestinal epithelial cells. An additional distinguishing feature of this epithelium, particularly in the colon, is the relatively high frequency of tumor formation (Willett, 1989;Northover, 1989).
The basic structural unit of the intestinal epithelium is the crypt, which in the mouse contains some 300-500 cells. Each crypt can be divided into several zones, namely a stem cell zone, a proliferative zone, and a zone that contains the mature cells. The location of these zones relative to each other differs in the small and large intestine, but in general the immature proliferative cells are found in the lower half of the crypt and the postmitotic differentiated cells in the upper region (Wright and Alison, 1984;Potten and Hendry, 1983). The unitarian hypothesis proposed over 20 years ago (Copson, 1966;Rubin et al., 1968) and largely substantiated by many subsequent experiments (Chang and Leblond, 1971;Cheng and Leblond, 1974;Ponder et al., 1985;Griffiths et al., 1988;Schmidt et al., 1985;Schmidt et al., 1988;Winton et aZ., 1988) suggests that all of the cell types in the adult crypt are derived from nonmigratory self-renewing stem cells. Committed precursors derived from these stem cells migrate and begin to differentiate, progressively adopting morphological and biochemical characteristics of either goblet, columnar, endocrine, Paneth (small intestine), or caveolated (mostly large intestine) cells (Wright and Alison, 1984). In addition to the cellular diversity within individual crypts, there exist clear qualitative and quantitative differences in patterns of gene expression in the same cell types in different regions of the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom correspondence should be addressed.
intestinal tract (Rubin et al., 1989;Sjolund et aL, 1983). How these differentiation programs are regulated is unclear, although recent experiments by Gordon andco-workers (Sweetser et aL, 1988a, 1988b;Hauft et al., 1989;Roth et al., 1990) using transgenic mice are beginning to provide some insights into these complex processes.
Our understanding of how the differentiated state of a particular cell type in vitro and in vivo is established and maintained is still at an early stage. Recent data from many laboratories investigating a range of cell types and organ systems suggest that a class of nuclear proteins (either by themselves or in concert with others) plays a pivotal role in the differentiation process. These proteins clearly fall into different families based on their amino acid sequences (Dresder, 1989;Dressler and Gruss, 1988;Davis et al., 1987;Wright et aL, 1989;Braun et al., 1989) but are thought to act via their ability to bind specifically to DNA sequences and either repress or activate transcription of adjacent genes (Thali et al., 1988;Jaynes and O'Farrell, 1988;Driever and Nusslein-Volhard, 1989;Winslow et al., 1989;Krasnow et al., 1989;Samson et al., 1989;Biggin and Tjian, 1989;Dearlof et al., 1989;Lassar et al., 1989). The most thoroughly characterized of these is the homeobox gene family, initially described in Drosophila, but now shown to be present in many species (Scott et aL, 1989). The importance of homeobox genes in mammalian development can be inferred from detailed mapping of transcript levels in particular tissues during important developmental phases in mice (for review see Holland and Hogan, 1988;Akam, 1989). More direct evidence may be forthcoming as a result of recent studies creating loss of function or gain of function mutant animals (Zimmer and Gruss, 1989;Joyner et aL, 1989;Wolgemuth et aL, 1989;Balling et al., 1989).
We chose to examine whether members of the homeobox gene family are expressed in murine intestinal epithelium and, if so, whether individual family members are found in particular anatomical locations. These experiments should allow us to address the role of these genes in specifying cell types within the crypt, and thereby in a more general sense the polarity of the adult intestinal tract. Using a PCR'-based strategy we have shown that nine different homeobox genes are expressed in the intestine, including one which has not previously been described.
Quantitative analysis of transcript levels for each homeobox gene along the intestine showed clear proximal-to-distal differences; this is consistent with the original hypothesis that products of these genes may play a role in establishing cellular identity in the gut.

MATERIALS AND METHODS
Crypt Isolation and Cell Culture-Crypts were isolated from adult mouse small intestine and colon according to published procedures (Whitehead et al., 1987) with minor modifications. Briefly, tissue was isolated from the animal, cut along its length, and washed several times in phosphate-buffered saline (pH 7.4) to remove luminal contents. The intestinal segments were then incubated at 37 "C for 15 min in 10 volumes of phosphate-buffered saline (pH 7.4) that contained 3 mM EDTA and 0.5 mM dithiothreitol. The buffer was decanted and replaced with fresh solution, and the incubation was continued at 37 "C for a further 45 min. The remainder of the protocol was as described. Isolated crypts were either used immediately or stored frozen at -70 "C.
LIM 1863 cells were maintained as described (Whitehead et al., 1987), and cell pellets were kindly provided by Dr. R. Whitehead (Ludwig Institute for Cancer Research, Melbourne Branch).
RNA Isolation, cDNA Synthesis, and PCR Amplifkution-Total RNA was extracted from tissues or cells by the low temperature guanidine thiocyanate procedure (Han et al., 1987) and in some cases further purified by precipitation with 3 M LiCl. The integrity of all RNA preparations was checked by agarose gel electrophoresis (Finley et al., 1989) Poly(A') mRNA was isolated by chromatography on oligo(dT)-cellulose (Aviv and Leder, 1972). Oligo(dT)-primed cDNA was synthesized from poly(A+) mRNA (5 pg) using a cDNA synthesis kit (Amersham Corp.). The helix 1 oligonucleotide mixture contained 64 different 22-mers with an EcoRI recognition site at one end to facilitate cloning. The helix 3 oligonucleotide mixture contained 48 different 25-mers with a BamHI recognition sequence at one end. Five percent of the resultant cDNA was combined with 300 pmol of each of the synthetic oligonucleotide mixtures; 10 p1 of 10 X amplification buffer (1 X buffer is 20 mM Tris-HC1, pH 9.3, 50 mM KCl, 2 mM MgC12, 1 mM 0-mercaptoethanol); 8 pl of 2.5 mM each dATP, dCTP, dTTP, and dGTP; and 2.5 units of Taq polymerase (Perkin-Elmer Cetus Instruments) in a final volume of 100 pl. The amplification reaction, elongation (68 "C, 2 min), denaturation (95 "C, 1 min), and annealing (37 "C, 1 min) were repeated 30 times.
Each PCR reaction was then phenol/chloroform-extracted and the amplified DNA recovered by ethanol precipitation. This DNA was then digested with restriction endonucleases BarnHI and EcoRI (Promega Biotec, Madison, WI) and size-fractionated on 5% polyacrylamide gels. DNA bands were visualized after staining the gel with ethidium bromide, and the region corresponding to a size of 120 base pairs was excised. DNA was recovered from these gel slices by passive diffusion overnight (Maxam and Gilbert, 1980). The recovered DNA was then ligated into BamHIIEcoRI-digested vector pGEM-3 (Promega Biotec) according to standard methods (Maniatis et al., 1982).
Characterization of Subcloned PCR Products-Miniprep DNA was prepared (Holmes and Quigley, 1981) from overnight cultures of individual colonies and analyzed for insert size by restriction endonuclease digestion (EcoRI plus BamHI) followed by acrylamide gel electrophoresis. Double-stranded DNA from selected clones was then sequenced using the Sequenase DNA sequencing kit (U. S. Biochemical Corp.) according to the manufacturer's instructions.
RNase Protection Assay-Probes were synthesized from pGEM-3 or pGEM-4 vectors linearized with either BarnHI or EcoRI, using T7 or SP6 RNA polymerase (Promega Biotec) and ["PIUTP (3000 Ci/ mmol, BRESA Pty. Ltd.) as described (Melton et al., 1984). Removal of template DNA from probes was achieved either by DNase I digestion (Cdr-1 and Cdr-2) and subsequent chromatography on Sephadex G-50 (Pharmacia LKB Biotechnology Inc.) or by gelpurifying the radiolabeled transcript. The most effective procedure for template removal was provided by the latter technique and was essential for detecting the less abundant mRNAs.
Assays were performed (Thompson and Gillespie, 1987;Firestein et al., 1987) by incubating 25 pg of total RNA in 10 pl of 5 M guanidine thiocyanate, 0.1 M EDTA, pH 7.0, together with 1-2.5 X lo5 cpm of antisense riboprobe at room temperature overnight. Samples were then diluted by the addition of 0.3 ml of buffer (0.3 M NaCl, 1 mM EDTA, 10 mM Tris-HC1, pH 7.5) containing 50 pg/ml RNase A and then incubated at 30 "C for 60 min. Each reaction was then digested with proteinase K (final concentration of 0.2 mg/ml) for 15 min at 37 "C, and RNase-resistant products were isolated after phenol/ chloroform extraction by isopropyl alcohol precipitation. Samples were then electrophoresed together with labeled size markers (Sau3Al-digested pGEM-3Z) on 6% polyacrylamide sequencing gels. Dried gels were exposed to preflashed RX film (Fuji) or XAR-5 film (Kodak) with one or two intensifying screens (Du Pont Lightning Plus). Densitometry of several different exposures was performed using a Chromoscan 3 densitometer (Joyce Loebel).
Southern Blot Analysis-High molecular weight DNA was prepared (Boothby et al., 1981) from the livers of several mice and digested to completion with restriction endonucleases according to the supplier's instructions (Promega Biotec). Digested DNA was electrophoresed on 1% agarose gels and transferred by capillary blotting to Hybond N+ membrane (Amersham Corp.) in 0.4 M NaOH.

Homeobox Gene Expression in Adult Mouse
Intestinal Epithelium-Sequence analysis of homeobox genes, particularly in Drosophila, has revealed that certain regions within the homeodomain are more highly conserved than others. This suggested to us that, by using appropriate synthetic oligonucleotide primers in a polymerase chain reaction with cDNA made from epithelial crypt mRNA, we would have a sensitive and rapid way of ascertaining which members of this family are expressed in these cells.
We chose to use degenerate oligonucleotide primers based on two regions of five amino acids ( Fig. 1) found in helix 1 and helix 3, respectively, which are highly conserved among different classes of Drosophila homeobox genes. As mentioned previously, all of the dividing and differentiating epithelial cells are found in crypts. To enrich for PCR products derived from these cells we extracted RNA from purified crypts rather than whole tissue fragments or mucosal scrapings, where the crypt epithelial cells represent a much smaller percentage of the total cell population. Whole crypts can be isolated by incubating the intestine in an isotonic buffer containing dithiothreitol and EDTA (Whitehead et al., 1987). The resulting preparation does contain single cells but consists almost entirely of individual, intact crypts (Fig. 2). As a positive control for cell type (epithelial) specificity we decided to include in our experiments mRNA from a cell line (LIM 1863) that was isolated from a human colonic tumor (Whitehead et al., 1987). These cells grow in culture as organoids made up of several moderately well differentiated cell types, which express markers characteristic of normal intestinal epithelium.
Gel electrophoresis of products from all PCR reactions showed the presence of a band with a mobility that corresponded to the predicted size of approximately 120 base pairs. This DNA was recovered from preparative acrylamide gels and subcloned in preparation for sequence analysis. Such an analysis of randomly selected clones revealed that a variety AbaE:

Q T L E L E K E F L F N -Y -----R -E -A R -L -L -E R Q K I W F Q R R M K -K K -lab: Q L T E L E K E F H F N -Y L T R A R R -E I A --L Q L N E T Q K I W F Q R R M K Q K K R -
Antp:

Q T L E L E K E F H -N -Y -T R R R -----------E R Q l K l W F Q R R M K -K K --
Dld:

Q -L E L E K E F H -N R Y L T R R R R -E I A H -L -L -E R Q -K I W F Q
r n RRMKWKKD- FIG. 1. Sites used for preparation of PCR primers. Sequence alignment of the four major families of Drosophila homeobox-containing genes is shown. Degenerate oligonucleotide probes were synthesized based on the amino acid sequences enclosed in bores. References are as follows: AbdB (Regulski et al., 1985), lab (Hoey et al., 1986), Antp (Scott and Weiner, 1984), Dfd (Regulski et al., 1987).

FIG. 2. Isolated crypt preparation from adult mouse colon.
Crypt epithelium was prepared from mouse colon by the EDTA/ dithiothreitol method. Note the presence of intact crypts and a low background of single cells. Bar is 1 pm. of homeobox genes as expressed in the different tissues (Table  I). It appears from this table that there is tissue-specific expression of some homeobox genes; however, results of studies detailed elsewhere in this report suggest that the frequency with which a particular sequence is recovered is not necessarily proportional to its abundance in the original RNA preparation. This has been observed by others (He et al., 1989) employing an identical strategy for examining members of the POU family of transcription factors.
One result of these initial experiments was the isolation of a clone encoding a new member of the homeobox family, which we have named Cdx-2, because it is most homologous to the recently described homeobox gene Cdx-1 (Duprey et al., 1988). Alignment of the two sequences (Fig. 3) shows that the predicted amino acid sequences are 88% identical, whereas the nucleic acid sequences are less so (73%).

Cdn-2 Is a Single Copy Gene Expressed Only in Adult
Intestine-The amount of nonhomology between Cdx-1 and Cdx-2 nucleic acid sequences should allow us to discriminate between the two at the level of blot hybridization. To confirm this we performed Southern analysis on mouse DNA. Using a Cdx-1 PCR subclone to generate probe we detected single predominant bands in digests with three different restriction enzymes (Fig. 4A ), which confirm and extend previous results

digested with either EcoRI ( E ) , Hind111 (H), or PstI (P) and probed with Cdx-1 (panel A ) or Cdx-2 (panel B ) . Filters were exposed to Fuji
RX film with two intensifying screens for 15 h. K6, kilobases. (Duprey et al., 1988). The same digests probed with Cdx-2 revealed a different pattern of bands, showing that it was present in the mouse genome in a single copy and that under the conditions of the experiment it did not cross-hybridize with Cdx-1 (Fig. 4 B ) .
Most of the homeobox genes listed in Table I have been described by others to be expressed in a range of tissues at various times during development (see references cited in Table I). The interesting finding which distinguishes Cdx-1 from the others is that its expression, at least at the level of Northern blot and in situ hybridization, has been reported to be confined to the intestinal epithelium (Duprey et al., 1988). We decided to re-examine expression of the Cdx-1 gene because of our particular interest in intestinal epithelial cells and to document Cdx-2 gene expression using a sensitive RNase protection assay.
The result of these experiments (Fig. 5 ) was that protected Cdx-1 and Cdx-2 products of the appropriate size were present only in the intestine, and although the autoradiographs as shown were overexposed, there was no evidence for the presence of either transcript in any of the other tissues sampled. All bands other than those indicated by the asterisk are present in controls lacking RNA and arise because the probes

1
were not gel-purified before use.

Quantitation of Homeobox Gene Expression in Different Regions of the Intestine-The intestine can structurally and . '
functionally be divided into several different regions, namely 40 duodenum, jejunum, ileum, proximal colon, and distal colon. Studies by Gordon and colleagues (Gordon, 1989, and references therein) have shown substantial differences in the expression of several genes along the intestine reflecting this functional diversity. To establish whether the genes listed in Table I are expressed in a region-specific manner, we performed RNase protection experiments on RNA isolated from whole segments of intestine representative of the areas listed above. By using unfractionated tissue any selective loss of cell types that might occur, for example during crypt isolation, was avoided. Tissue from several animals was pooled and extracted for these experiments.
All experiments were performed in duplicate or triplicate using the same tissue RNA samples, and the results varied between analyses by 10% or less. Autoradiographs of representative experiments are shown in Fig. 6 together with the corresponding densitometric data. All of the genes assayed, except for Hox 1.6, gave observable signals, which in most cases (except for Hox 3.2) permitted accurate quantitation, that is gave discrete bands. Not unexpectedly there was a considerable range in the relative mRNA abundance from one homeobox gene to another. Although no effort was made to accurately quantitate this parameter, all probes were of a similar specific activity, and so band intensities and exposure times give some indication as to the range of transcript levels detected. Within the intestine there were clear differences in mRNA levels for each gene in the different regions. Transcripts from the intestine-specific homeobox genes Cdx-1 and I, ileum; CP, proximal colon; CD, distal colon), was hybridized to 1-2.5 X lo5 cpm of antisense riboprobe. The protected RNAs were electrophoresed on polyacrylamide gels, and regions of the autoradiographs corresponding to those protected RNAs are shown together with the corresponding densitometric data. The exposure times for autoradiography are as follows: Cdx-1 and Cdx-2 (18 h, one intensifying screen), Hox 1.3, Hox 3.1, Hox 3.2 (2 weeks, two intensifying screens), Hox 2.5 (1 week, two intensifying screens), Hox 1.7, Hox 2.3 (60 h, two intensifying screens).

Cdx-2.
were detectable in all regions assayed but were most abundant in the distal and proximal colon, respectively. This is in stark contrast to Hox 1.7, whose expression was restricted to proximal and distal colon. Even very long exposures of the corresponding gel shown in Fig. 6 failed to reveal any signal above background in other regions of the intestine. The signal obtained for Hox 3.2 transcript was very weak, preventing accurate quantitation, but it is clear that proximal colon had the highest level, followed by ileum and then distal colon.

DISCUSSION
The complex process of cellular differentiation, which begins with a multipotential stem cell and ends with mature Homeobox Gene Expression in the Intestinal Epithelium cells, has been studied at several different levels. Two types of modulators (that for purposes of this discussion are called extrinsic and intrinsic) have been broadly defined, which can influence developmental outcomes within a particular lineage. Extrinsic regulators include growth and differentiation factors, extracellular matrix components, and other cells, whereas intrinsic regulators are the levels or activities of various cellular molecules. Clearly these two categories are not mutually exclusive, as often members of the former can influence the latter (Ruiz i Altaba and . In the intestinal epithelium the molecules, either extrinsic or intrinsic, that regulate commitment and progression down a particular lineage pathway are as yet undefined. The lack of progress in defining factors important for maintaining these cells in vitro currently represents the largest impediment to additional study of the extrinsic factors influencing these cells. In order to define intrinsic factors that may play a role in establishing or maintaining cellular identity in the adult gut, we have chosen to study the homeobox family of transcriptional regulatory proteins. Studies to date have largely focused on the means by which expression of individual members of this gene family influences developmental decisions in the context of the embryo. We have extended these studies to the adult, where tissue identity is clearly established but where important cell lineage decisions are still being made in several tissues. Similar studies are underway in the hemopoietic system (Kongsuwan et al., 1988;Shen et al., 1989).
Using a PCR-based strategy we have detected expression of nine different homeobox genes in the intestinal epithelium of adult mice. This approach represents an efficient way of analyzing expression of individual members of large multigene families with the potential of discovering new genes as demonstrated by our isolation of Cdx-2. However, the technique, as currently employed, is not quantitative, in that we found no correlation between the frequency with which a particular clone was isolated from a population of PCR products and the relative abundance of the corresponding mRNA in tissue as determined by RNase protection analysis. This suggests that we have not sampled the entire repertoire of homeobox gene expression in the tissues examined.
The new homeobox gene family member we have isolated, Cdx-2, is most closely related to a gene, Cdx-1, which was originally cloned from a mouse embryo cDNA library based on its homology to the Drosophila gene caudal (Duprey et al., 1988). The small amount of sequence we have derived from our PCR clones suggests that Cdx-1 and Cdx-2 represent a small subgroup within the homeobox family similar to those of En-1 and En-2 (Joyner and Martin, 1987) and the recently isolated Eux-1 and Eux-2 (Bastian and Gruss, 1990). One distinguishing feature of these subgroups is the high degree of sequence homology, which extends throughout the protein rather than being confined to the homeodomain. Comparison of the predicted amino acid sequences of Cdx-1 and Cdx-2 proteins from recently isolated cDNA clones has revealed extended regions of homology.* Of particular relevance to our interest in intestinal epithelial differentiation was the finding that of several adult mouse tissues surveyed, intestine was the only one which contained detectable Cdx-1 and Cdx-2 transcripts. Our recent in situ hybridization experiments have established that, within the intestine, expression of both genes is confined to the crypt epithelium. This very restricted expression pattern is unusual for homeobox genes and further indicates an important role for the protein products of these two genes in establishing cellular identity in this tissue. ' R. James and J. Kazenwadel, unpublished data.
Experiments in several different organisms over recent years have shown that during embryogenesis homeobox genes are generally expressed in a graded fashion within a body segment or particular tissue. In a majority of cases the amount of homeobox protein present is proportional to the abundance of the corresponding mRNA. A mechanism by which transcription of other genes is regulated by a given amount of homeobox protein is beginning to emerge (Driever et al., 1989). If homeobox proteins are playing a role in differentiation, we might expect to see different levels of expression in anatomically distinct regions of the gut and perhaps gradients within the crypt. Quantitative mapping throughout the gut of the transcript levels of all genes isolated in the PCR experiments revealed several interesting findings. First, every gene except Hox 1.6 gene gave detectable signals in the RNase protection assay, suggesting that the PCR approach has revealed a spectrum of gene expression, the level of which is likely to be biologically meaningful. The inability to detect Hox 1.6 transcripts may be due either to the extreme sensitivity of the PCR technique (Sarkar and Sommer, 1989) or to expression in a cell type that represents a minor population within the intestine. Second, most genes are expressed at their highest levels in the colon, the exception to this being Hox 3.1. The most extreme example of this is Hox 1.7, whose transcripts are relatively abundant in both proximal and distal colon but undetectable in any regions of the small intestine. The next step in this analysis will be to perform in situ hybridization experiments using transcript-specific probes in order to identify which cell types within the gut express these genes. The intestine-specific genes Cdx-1 and Cdx-2 show graded expression patterns, with transcripts in the colon being 5-10-fold more abundant than in the duodenum. Interestingly, there was little or no change in their abundance between proximal and distal regions of the small intestine. One of the questions that these results raise is whether a gradient of this magnitude is biologically significant. In one of the few cases where homeobox gene mRNA levels have been accurately quantitated, Ruiz i Altaba and Melton (1989aMelton ( , 1989b examined Xhox 3 transcripts in Xenopus embryos. They showed that posterior axial mesoderm has five times more of this particular mRNA than anterior axial mesoderm and that if the gradient is destroyed by artificially overexpressing the gene, development of the anterior portion of the embryo is disrupted. Based on these data it would be of interest to create lines of transgenic mice expressing elevated Cdx-1 or Cdx-2 in the small intestine to assess the importance of the normal in uiuo gradient. The molecular mechanisms responsible for generating these expression gradients are at present unknown, although it appears that there are two classes of homeobox gene. One class is clustered together on a particular chromosome and comprises the majority of gene family members characterized to date, whereas the other class (exemplified by En-1 and En-2 and possibly including Eux-1 and Eux-2 and Cdx-1 and Cdx-2) expresses each member of the pair on different chromosomes. Data from several laboratories have shown that in mice, like Drosophila, those genes which are found in a group are expressed in the animal (anterior-to-posterior) in an order determined by their position within the group. Part of the regulation of these clustered genes is thought to involve selective changes in chromatin structure (Peifer et al., 1987;Gaunt and Singh, 1990). Our data are consistent with this model, in that the anterior boundary of Hox 2.3 expression is the jejunum, whereas Hox 2.5, which is located at the end of the Hox 2 locus, is first detectable in the ileum. Likewise transcripts from the Hox 1.3 gene, which is in the middle of the Hox 1 locus, are found in the duodenum to the colon, whereas Hox 1.7, which is at the end of this locus, is found only in proximal and distal colon. What is novel about these data is that intestine does not arise from a segmented structure during embryogenesis like the nervous system or kidney (Hogan et al., 1985;Lewis, 1989). Both Cdx-1 and Cdx-2 mRNAs are detectable throughout the intestine and, overall, show fairly similar anterior-to-posterior expression profiles. The role of chromosomal location in regulating transcription of these genes remains to be determined.
In conclusion, the data reported in this paper represent the beginning of studies aimed at understanding how one class of molecules, the homeobox family of transcriptional regulatory proteins, may affect cell lineage pathways in the intestine, particularly the epithelial compartment. Additional study of Cdx-i and Cdx-2 genes and their protein products is clearly warranted because of their restricted tissue distribution and graded expression pattern. Several of the other homeobox genes described in this report are of interest. Although they are found in other tissues, they exhibit region-specific expression within the intestine. Having identified these genes, we are in the process of directly testing their biological activity in transfected intestinal epithelial cell lines and transgenic animals.