Regulation of Duodenal Specific Expression of the Human Adenosine Deaminase Gene*

Formation of the mammalian gastrointestinal tract is an ordered process of development and differentiation. Yet, the adult small intestine also retains the plasticity to respond to cues both internal and environmental to modulate intestinal function. The components that regulate this development, differentiation, and modulation at the molecular level are only now being elucidated. We have used the human adenosine deaminase (ADA) gene as a model to identify potential cis-regulatory components involved in these processes within the small intestine. In mammals, high levels of ADA in the small intestine are limited specifically to the differentiated enterocytes within the duodenal region. These studies describe the identification of a region of the human ADA gene, completely distinct from the previously identified T-cell enhancer, which is capable of directing the human intestinal expression pattern in the intestine of transgenic mice. The reporter gene expression pattern observed in these transgenic mice is identical to the endogenous gene along both the cephalocaudal and crypt/villus axis of development. Timing of this transgene activation, however, varies from that of the endogenous mouse gene in that the transgene is activated approximately 2 weeks earlier in development. Even so, this precocious activation is also limited to the epithelium of the developing villi strictly within the duodenal region of the small intestine.

Formation of the mammalian gastrointestinal tract is an ordered process of development and differentiation. Yet, the adult small intestine also retains the plasticity to respond to cues both internal and environmental to modulate intestinal function. The components that regulate this development, differentiation, and modulation at the molecular level are only now being elucidated. We have used the human adenosine deaminase (ADA) gene as a model to identify potential cis-regulatory components involved in these processes within the small intestine. In mammals, high levels of ADA in the small intestine are limited specifically to the differentiated enterocytes within the duodenal region. These studies describe the identification of a region of the human ADA gene, completely distinct from the previously identified T-cell enhancer, which is capable of directing the human intestinal expression pattern in the intestine of transgenic mice. The reporter gene expression pattern observed in these transgenic mice is identical to the endogenous gene along both the cephalocaudal and crypt/villus axis of development. Timing of this transgene activation, however, varies from that of the endogenous mouse gene in that the transgene is activated approximately 2 weeks earlier in development. Even so, this precocious activation is also limited to the epithelium of the developing villi strictly within the duodenal region of the small intestine.
Free nucleotides and nucleosides have been shown to be intimately involved in the normal function and regulation of a wide variety of systems, including but not limited to neurotransmission, vasodilation, platelet aggregation, energy transportation, and the synthesis of nucleic acids. A balance of purine and pyrimidine pools in some cells is critical. Excess amounts of some purine compounds cause cell death in developing thymocytes (1) and in neural cells (2). Insufficient intracellular pools have been shown to be involved in precocious differentiation of cultured intestinal cells (3). In fact, some of the most effective chemotherapy agents are those that are able to interfere with purine metabolism (4). Maintenance of these pools within acceptable limits is accomplished by enzymes of the purine and pyrimidine de novo and salvage pathways. Adenosine deaminase (ADA), 1 a member of the purine salvage pathway, catalyzes the irreversible deamination of adenosine and deoxyadenosine. ADA is expressed in all human tissues, yet levels vary in a specific fashion from the highest levels in thymus (790 nmol/min/mg) and duodenum (570 nmol/min/mg) to much lower levels in tissues such as liver (10 nmol/min/mg; Refs. [5][6][7]. The absence of this enzyme in humans leads to a severe combined immunodeficiency characterized by an absence of both B-and T-cells as well less dramatic changes in other tissues (8). ADA-deficient mice, produced by embryonic stem cell gene-targeting experiments, die at birth with defects of the liver, lungs, and small intestine (9,10). Studies in our laboratory have concentrated on the identification and characterization of cis-regulatory elements controlling the tissue-specific pattern of ADA expression.
Basal promoter activity for the human ADA gene has been mapped in vitro to 81 bp of 5Ј-flanking sequence. This basal promoter has no inherent tissue specificity. The proximal promoter region contains no consensus CCAT or TBP binding sites, but it does possess multiple functional Sp1 binding sites that are necessary for both basal and T-cell enhancer-driven activation of this promoter in vitro (11). Neither this minimal promoter nor an additional 3.7 kb of 5Ј-flanking sequences are capable of consistently activating transcription of a reporter gene in transgenic mice. Even low level activation was absent in a wide variety of tissues examined (7). Consistent, high level transgene expression was observed in the thymus of mice with transgenes containing both the promoter fragment and a Tcell-specific cis-regulatory region from the first intron of the human ADA gene (7,12). This region directs high level, thymus-specific expression of a linked reporter gene in transgenic mice in an insertion site-independent, copy-proportional manner. It is also able to drive variable low level expression in all tissues examined. Characterization of this region revealed an enhancer element of 230 bp capable of elevating reporter gene expression in transient assay systems and containing consensus binding sites for the E box (13), Ap1 (14), c-Myb (15), lymphocyte enhancer-binding factor-1 (LEF-1) (16), and Ets (17) families of transcription factors (12). Similar factor binding sites are present within the recently characterized mouse homolog to this enhancer (18), and transgenic studies of a segment containing this mouse enhancer demonstrate results analogous to those observed with the human enhancer (19). Deletional analysis, footprinting, and site-directed mutagenesis have demonstrated the functional importance of many of these sites within the human enhancer (12,20,21). Flanking the T-cell enhancer are elements termed facilitators implicated in assisting in the formation of a region of stable open chromatin at the enhancer. These facilitators are associated with an locus control region (LCR) effect that results in position-independent expression of the transgenes containing them (12,22,23).
Compared with the information regarding regulation of human ADA in thymocytes, much less has been discovered about its regulation in other human tissues. A region in the 5Ј-flank of the mouse ADA gene has been described which is involved in expression in mouse forestomach and fetal placenta (24,25). A placenta-specific regulatory segment was identified recently within that region (26). ADA is also known to be extremely high in the maternal decidua of mice and other mammals (27,28) and in human, mouse, and other mammalian small intestine (5,28,29). Previously, no region of either the mouse or the human ADA genes has been identified which is involved in driving high level expression in either maternal decidua or small intestine. We have now identified and begun characterization of a region of the human ADA gene completely distinct from the T-cell enhancer region which is capable of driving high level, duodenum-specific expression in transgenic mice.

EXPERIMENTAL PROCEDURES
Plasmid Constructions-p5Јacba (12) was derived from pADACAT4.0 (7) by adding an SalI site at the BamHI site. All subsequent clones are manufactured from this plasmid or one of its derivatives. A 2.3-kb SalI-HindIII fragment (GenBank accession number M13792, residues 8327-10584; Ref. 30) of the human ADA gene containing the human T-cell enhancer and facilitators was isolated from pSph2.3-(11). The ends were filled with dNTPs and Klenow, and BamHI linkers were added. The resulting BamHI fragment was subcloned into the BamHI site of p5Јacba to create p5Јacba enh. The 13.0-kb SalI insert from ADA4 (GenBank accession number M13792, residues 26353-36741 ϩ 2.6 kb of additional unsequenced 3Ј-flanking sequences; Ref. 30) was cloned into the SalI site of p5Јacba enh to create p5Јacba enh L4. The 13.0-kb SalI fragment from ADA117 (GenBank accession number M13792, residues 13518 -26509; Ref. 30) was subcloned into the SalI site of p5Јacba or p5Јacba enh to generate p5Јacba L117 and p5Јacba enh L117, respectively. No positive clones could be detected after repeated transformation into rubidium chloride-treated Escherichia coli (DH5␣, Life Technologies, Inc.). Poor transformation efficiency by this method seemed to be directly related to the large size of these plasmids. Ligation products were introduced successfully into E. coli by electroporation (Bio-Rad), and this method was used routinely. A 7,685-bp HindIII fragment from p5Јacba L117 was isolated and ligated to Hin-dIII-cut pUC 18. A 7,685-bp SphI fragment was removed from the resulting plasmid and ligated to SphI-cut p5Јacba to create p5Јacba L117 ⌬1. p5Јacba L117 was digested with ApaLI, filled using T4 polymerase and dNTPs, and subsequently digested with SalI to isolate a 8,765-bp fragment. A 49-bp SalI-SphI fragment from pGEM-5Zf-(Promega) was ligated to SalI-SphI-cut p5Јacba. The resulting plasmid was digested with EcoRV and SalI and ligated to the 8,765-bp fragment described above to create p5Јacba L117⌬2. A 9,612-bp ApaLI fragment from p5Јacba L117⌬1 was filled with T4 and dNTPs, digested with SacI, and a 5,723-bp fragment was isolated and ligated into p5Јacba plasmid that had been BamHI cut, T4 filled, and then SacI cut to generate p5Јacba L117⌬3. All clones were chosen by restriction analysis such that the ADA genomic segments that they contain are in the same relative position and orientation to each other as in the human ADA gene. Plasmid DNAs for transgene isolation were prepared as described previously (12).
Fragment Isolation and Transgenic Mouse Preparation-Transgene I has been reported previously as ADACAT4/12 (7) and as 3.7/0-IV (12). Transgene II also has been published previously as 0.3/II-IIIab (12). Both of these transgene fragments were analyzed in C57/C3H hybrid mice. Transgene III is a 17,364-bp BssHII-PvuI fragment from p5Јacba enh L4. Transgene IV is a 20,732-bp NdeI-PvuI fragment from p5Јacba enh L117. Transgene V is an 18,464-bp NdeI-PvuI fragment from p5Јacba L117. Transgene VI is a 14,237-bp NdeI-PvuI fragment from p5Јacba L117⌬1. Transgene VII is a 13,149-bp NdeI-PvuI fragment from p5Јacba L117⌬2. Transgene VIII is an 8,915-bp NdeI-PvuI fragment from p5Јacba L117⌬3. All transgene fragments were isolated from low melting point agarose by digestion with ␤-agarase, phenol extraction, precipitation, and purification over an Elutip-D column (Schleicher & Schuell). Transgenic mice prepared with Transgenes III-VIII were produced from FVB/N mice. Tail DNA samples from F 0 mice bearing Transgenes III or IV were digested with EcoRI, and those containing Transgenes V-VIII were digested with EcoRI and BamHI. DNA was electrophoresed, Southern blotted onto MAGNANT membranes (MSI), and probed with a radiolabeled 1.4-kb EcoRI fragment from pBLCAT6 (31) encompassing most of the CAT coding sequence. Mice that possessed a band of the appropriate size, 1.9 kb for Transgenes III and IV and 1.4 kb for Transgenes V-VIII, were mated with non-transgenic littermates to establish lines from each founder. Offspring were analyzed by both polymerase chain reaction and Southern blot for transmission of the transgene.
Analysis of Transgenic Mice-F 1 transgenic mice were analyzed for CAT and/or ADA activity between 4 and 12 weeks of age except where noted. Small intestine segments were isolated from duodenum (2-cm section adjacent to pyloric sphincter), jejunum (2-cm section from the center of the small intestine), and ileum (2-cm section anterior to the cecum). Other tissues routinely assayed included thymus, liver, tongue, esophagus, colon, and stomach. An extended panel of tissues which was assayed once for each transgene also included quadriceps muscle, lung, heart, ovary/testes, kidney, bone marrow, and brain. For studies examining maternal decidua, transgenic F 1 females were superovulated and mated to non-transgenic males. Superovulated but unmated transgenic females were used as control. The uterus was harvested 9.5 days postcoitus and assayed for ADA and CAT. Protein concentrations, CAT activity, and ADA activity were assessed as described previously (7). Liver DNA was isolated from the same mice used to determine CAT activity and digested with EcoRI (Transgenes III and IV) or EcoRI and BamHI (Transgenes V-VIII). Samples of 20, 10, 5, 3, and 1 g of each digested liver DNA and 1, 2, 3, 5, 10, and 25 copy equivalents of a CAT-containing plasmid were electrophoresed, Southern blotted, and probed with the radiolabeled 1.4-kb EcoRI fragment from pBLCAT6. Blots were scanned and bands quantitated using a PhosphorImager (Molecular Dynamics, Inc.). A copy number standard curve was generated, and the copy number was determined by averaging the estimated copy number from each lane.
Placental Expression Studies-Female transgenic mice containing Transgene III, line 1, or Transgene IV, line 2, were superovulated, mated to non-transgenic males, and sacrificed on day 9.5 postcoitus. Superovulated, unmated transgenic females were used as control. The uterus from a mouse containing Transgene III, line 1, was homogenized in toto and assayed for ADA and CAT and the results compared with the control. Although a 10-fold activation of ADA was observed in pregnant versus non-pregnant uterus, no significant change in CAT reporter gene activity was observed between the two. Low levels of CAT were observed for each, in the range observed for other low level tissues. This low level expression was attributed to the ubiquitous expression generated by the T-cell enhancer/facilitators. Mice from Transgene IV, line 2, were subjected to finer dissection in their analysis. Embryo and yolk sac were dissected from the uterine implantation site. Tissue samples containing the implantation site were dissected away from the nonimplantation uterus. Implantation site uterine extracts contained ADA activity (5,240 nmol/min/mg) that was 65 times higher than that from either non-implantation or non-pregnant uterus. Both implantation site and non-implantation site extracts, however, had similar CAT activities that were very low (0.1-0.2% of that in thymic or duodenal extracts).
In Situ Hybridization-Probes for in situ hybridization were prepared from a pGEM-4Z plasmid containing a 550-bp HindIII-NcoI fragment from pSV0-CAT (32) which encompasses the 5Ј-end of the reporter gene. 35 S-or 33 P-labeled probes were synthesized from linear templates with T7 or SP6 polymerases using an in vitro transcription system (Promega). Duodenal samples from transgenic mice were isolated, rinsed in 1 ϫ phosphate-buffered saline, and fixed in 4% paraformaldehyde in 1 ϫ phosphate-buffered saline at 4°C overnight, followed by overnight incubation at 4°C in 30% sucrose, 1 ϫ phosphate-buffered saline. Tissues were frozen in M1 embedding matrix (Lipshaw) and cut into 10-m sections. Tissue sections were then fixed, acetylated, prehybridized, and hybridized (33, 34) with a solution containing 5 ϫ 10 5 cpm/l 35 S-or 33 P-labeled CAT riboprobe. After overnight hybridization at 48°C, sections were washed at high stringency and treated with RNase A (50 g/l, Worthington Biochemical Corp.) and RNase T1 (50 units/ml, Life Technologies, Inc.) at 37°C for 30 min. Sections were dehydrated and exposed to Kodak NTB-2 emulsion for 1 week at 4°C. Histological staining of sections was performed with hematoxylin and eosin. Sections were observed by light-and dark-field microscopy.
DNase I Hypersensitivity-Tissues were isolated from four adult transgenic mice, pooled, and nuclei isolated as described previously (7) with the following exceptions. Duodenal mucosa was removed by scraping, and nuclei were isolated from the mucosa only. Polyamine buffer was supplemented with 50 mM N-acetylL-cysteine. PEFABLOC (Boehringer Mannheim) was substituted for phenylmethylsulfonyl fluoride at the same concentration in all solutions. DNase I digestion was carried out on 10 7 nuclei using 2 units of DNase I in a total volume of 0.4 ml. DNA was digested with XbaI and Southern blotted. A region from the human ADA gene (GenBank accession number M13792, residues 23293-23603; Ref. 30) was used as a probe.

RESULTS
No Tissue Specificity Is Observed with 5Ј-Flanking Sequences-A 40-kb locus containing the human ADA gene located at 20q13.11 has been cloned and sequenced previously (30). A schematic is shown in Fig. 1. A number of fragments from this locus have been included previously in constructions with the CAT reporter gene to identify cis-regulatory elements by testing them in either transient transfection assays or transgenic mouse systems (7,11,12,22). A 0.2-kb minimal promoter, defined by transient transfection, is capable of low level reporter gene activation in a number of cell types (7). This low level activation is dependent upon the presence of multiple Sp1 binding sites because no TATA binding site is present within this 200-bp fragment (11). The addition of more of the 5Јflanking region up to 3.7 kb did not improve activation levels above this low level observed in transient transfection assays. Neither of these promoter fragments, 0.2 or 3.7 kb, is able to activate even low level reporter gene expression consistently in transgenic mice in a large panel of tissues assayed (7).
Thymus-specific Expression Depends on a First Intron Enhancer-When a 12.0-kb fragment from the ADA first intron (a in Fig. 1) was included 3Ј of the CAT reporter in transgenes with either the 3.7-kb (Transgene I, Fig. 1; Ref. 7) or the 0.2-kb promoter fragment (12), consistent, high level, CAT activity was observed in thymus. Reporter gene expression for these transgenes is both insertion site-independent and copy num-ber-proportional in this tissue (7,12). Transgenic mice containing fragment a also expressed the transgene at variable low levels in all non-lymphoid tissues assayed, including duodenum (Table I and Ref. 7). A 2.3-kb fragment (b in Fig. 1) is the smallest subregion of fragment a which was able to generate a transgene expression profile like that seen with Transgene I (Transgene II in Table I Transgenes I and II are as described previously (7,12). Transgene III contains a 232-bp ADA promoter, the CAT reporter gene, the 2.3-kb T-cell enhancer/facilitator fragment, b, and a 13-kb ADA genomic fragment, d, which includes exons 3-12 and 3,389 bp of 3Ј-flanking sequences. Transgene IV contains the larger 3.7-kb ADA promoter fragment along with the CAT reporter gene, fragment b, and a 13-kb fragment, c, from the 3Ј-end of the first intron through exon 3 of the human ADA gene. Transgenes IV and V are identical with the exception of fragment b, which is absent from Transgene V. Transgene VI contains fragment e, a 7,685-bp fragment from the 5Ј-end of c. Transgene VII contains a 8,765-bp fragment from the 3Ј-end of c. Fragment g, contained in Transgene VIII, represents the overlap between e and f and contains the multiple duodenal hypersensitive sites described in Fig. 5. Transgenic mice were obtained by pronuclear injection and multiple lines established for each transgene. position-independent expression of the transgene (12,22). High level reporter gene expression, comparable to that observed for the endogenous ADA gene, was notably absent in the small intestine of transgenic mice bearing either Transgene I or Transgene II. High Level Duodenal Expression Is Achieved in Mice Containing Fragment c-To identify region(s) responsible for high level ADA expression in other tissues, the remaining uncharacterized segment of the human ADA gene was analyzed as two overlapping fragments of 13 kb each (c and d in Fig. 1). These fragments were subcloned independently into expression vectors containing the human ADA promoter and the CAT reporter gene. It was not known if non-thymic, cis-regulatory regions of the ADA gene might also require the LCR-like function of the facilitator segments for insertion site-independent expression of a transgene. Because the facilitator regions have not been dissected away from the T-cell enhancer and shown to maintain function, the entire 2.3-kb thymic enhancer/facilitator fragment, b, was included in initial transgenic constructions III and IV. Three independent lines of mice were generated with Transgene III and analyzed for CAT expression. CAT activities from thymic and duodenal extracts are shown in Table I. High level CAT expression was observed only in the thymus. Thymic CAT activities fall within the predicted range for transgenic mice containing the T-cell enhancer (range 16,000 -63,000 pmol/h/100 g/copy; mean ϭ 33,000 Ϯ 13,000 pmol/h/100 g/copy; Ref. 12). Duodenal CAT activity levels are very low (Table I) and similar to the values obtained for all other non-thymic tissues assayed from these mice (data not shown). This pattern of expression is similar to that observed for tissues from mice bearing Transgenes I and II (7,12).
Transgene IV, shown in Fig. 1, contains fragments b and c and was used to generate four independent lines of transgenic mice. These mice exhibited high level CAT activities in both thymus and duodenum (Table I). As with Transgene III, thymic levels fall within the expected range. However, duodenal CAT activities are consistently 100 -1,000-fold higher than duodenal values reported for any previous ADA transgenic construction (Table I and Ref. 7). Ubiquitous low level transgene expression was also observed in the other tissues of these mice. All of the mice from this transgene had CAT activity in the duodenum at least 100 times greater than that found in low level tissues such as liver.
Fragment c Does Not Require the T-cell Enhancer/Facilitator for High Level Duodenal Expression-The ability of fragment c alone to direct high duodenal CAT expression without the T-cell enhancer/facilitator fragment was assessed by creating Transgene V (Fig. 1), which lacks fragment b. Transgenes IV and V are identical with the exception of the thymic enhancer/ facilitator segment. Eleven lines of transgenic mice were generated and analyzed for CAT activity. Duodenal and thymic CAT activities are shown in Table I. All 11 lines of mice express high CAT activity within the duodenum. Although the range of expression levels observed in duodenum is more variable than with Transgene IV, duodenal CAT activities remain very elevated, in the range of 2000 -31,000 pmol/h/100 g/copy, and are independent of the T-cell enhancer/facilitator fragment for this activation. Low level expression is also maintained in all tissues examined, including thymus. This is reminiscent of the T-cell enhancer/LCR yet quite independent of it. Both high level duodenal and ubiquitous expression are associated with the presence of fragment c within the transgene, suggesting the presence of a duodenum-specific enhancer and possibly LCRlike elements within this segment of DNA.
Fragment c Directs the Correct Pattern of Transgene Expression along the Anterior/Posterior Axis-ADA expression is high in human and murine small intestine (28), but within murine small intestine, high level ADA expression is limited to the proximal small bowel (29,35). To delineate better the anterior/ posterior boundaries of this high level expression and to determine if expression of the transgene mimics that of the endogenous gene, we examined both CAT and ADA enzymatic levels in tissue extracts from the same transgenic mouse. A portion of the gastrointestinal tract from the stomach to the jejunum was removed and dissected into 2-cm sections beginning at the pyloric sphincter of the stomach and extending caudally 18 cm, well into the distal portion of the jejunum. Extracts from each were assayed for both ADA and CAT enzymatic activity. Both Transgene IV and Transgene V show strikingly similar enzymatic profiles. Results for Transgene V, line 10, are shown in Fig. 2. ADA and CAT enzymatic activities are low in the stomach. Increases in both are abrupt and quite large as sections progress into the duodenum. Maximal expression of both is seen within the first 2-4 cm of the small intestine, the region commonly demarcated as the murine duodenum. Return to low level gene expression is also quite distinct, 6 -8 cm caudal to the pylorus expression of both is quite low again. Overall, the longitudinal profile of reporter gene expression closely mimics that seen with both human and murine ADA in proximal small intestine. Thus, fragment c contains all of the necessary sequences to regulate transgene expression along the anterior/ posterior axis in a pattern similar to that of the endogenous gene.
Fragment c Directs the Correct Crypt/Villus Patterning-Human and murine ADA expression in small intestine is limited to the duodenal epithelium and of the four cell types within the epithelium to only the enterocyte cell lineage (29,36). ADA mRNA is first detected just after the differentiating enterocytes emerge from the crypt. It increases rapidly and is present at high levels in terminally differentiated enterocytes along the entire length of the villus. In situ studies were performed to localize reporter gene expression within the transgenic duodenum. Sections of transgenic duodena from Transgene IV, line 2, and Transgene V, line 10, were incubated with radiolabeled antisense or sense CAT RNA. The presence of the thymic enhancer in the transgene had no observable effect, as identical results were obtained from both transgenes. The results with Transgene V, line 10, which contains only the newly identified duodenal regulatory region, are shown in Fig. 3. Longitudinal sections through the duodenum were probed with sense ( Fig.  3A) and antisense CAT (Fig. 3, B and C) probes. No signal was seen with the sense probe (Fig. 3A). A strong signal was seen with the antisense CAT probe over all of the duodenal villi within the field. A higher magnification of the villi (ϫ 100, Fig.  3C) reveals that the signal is specific to the epithelial layer and absent from crypts, lamina propria, and underlying intestinal structures, a pattern similar to that seen with endogenous mouse or human ADA in the duodenum. The deposition of signal seen is consistent with expression in the enterocytes which constitute 95% or more of the cellular population of the villous epithelium (37). It is uncertain if the transgene expression is restricted specifically to the enterocyte cell lineage or if it is also expressed in the other differentiated epithelial cell lineages found along the villus.
Fragment c Activates Reporter Gene Expression Precociously-Extensive characterizations of fetal and newborn expression patterns of intestinal ADA in humans have not been reported. Studies in murine small intestine have shown that ADA increases from low newborn levels to adult levels between 2 and 3 weeks postnatally (29,38). We suspected that this increase in intestinal ADA would be the result of increases in ADA principally in duodenum, and therefore we examined ADA levels in mouse small intestine as a function of development. Samples of duodenum, jejunum, and ileum were harvested from mice as a function of age and assayed for ADA activity. ADA activities in the duodenum, jejunum, and ileum at 1 and 6 weeks are shown in Fig. 4A. A dramatic increase in ADA activity is observed during this time in the duodenum. Increases in ADA activity were also observed in jejunum and ileum, but these activities were significantly lower than those seen in the duodenum. The majority of the activation of ADA in mouse small intestine is apparently accounted for by increases in the duodenum. A finer estimation of the timing of this increase in duodenal ADA was ascertained by sacrificing mice at 2-4-day intervals and examining ADA levels. Fig. 4B shows that the increase in duodenal ADA occurs between 11 and 18 days, thus paralleling the increase in whole small intestine observed previously at 2-3 weeks (38).
To compare transgene expression to endogenous ADA expression in small intestine, samples of duodenum, jejunum, and ileum were harvested at varying intervals from Transgene IV, line 1. In all mice tested, the jejunal and ileal CAT activities were less than 1% of that in the duodenum. Duodenal CAT activities for these mice are shown in Fig. 4B. Unlike endogenous ADA, which exhibits only low level activity at birth, the transgene is expressed in neonates at levels similar to those observed in adult transgenic mice. Examination of embryonic transgene expression within the duodenum revealed a sharp increase in CAT activity between embryonic day 18 (E18) and birth. Thus, transgene activation precedes activation of high level endogenous ADA expression by about 2 weeks. Similar results were observed with Transgene IV, line 2, and Transgene V, line 2 (data not shown). Therefore, the T-cell enhancer/ facilitator segment had no effect in altering this precocious expression in transgenic duodenum.
Because elevated ADA was not detected in these tissues, in situ hybridization was carried out on Transgene IV, line 2, mouse duodenum at 1 week to compare distribution of transgene expression with that of the adult. Signal deposition is distributed in a pattern similar to the adult transgenic duodenum. Transgene expression in these mice is limited to the epithelium found on the duodenal villi or developing villi, which, at this stage of murine intestinal development, can often be found side by side (data not shown).
Specific DNase I-hypersensitive Sites Are Present in Fragment c in Duodenal DNA-Many transcriptionally active enhancers lie in regions of altered chromatin structure characterized by their increased sensitivity to DNase I (39). To identify potential regions of such altered chromatin, DNase I hypersensitivity analysis was executed on a variety of transgenic tissues. Nuclei from tissues containing either Transgene IV or V were isolated and treated with DNase I for increasing time intervals. The resulting DNAs from these tissues were examined for cleavage via an indirect end-labeling procedure (7,40). The entire fragment c and 5Ј-flanking regions were scanned, and a single region of ϳ3 kb containing multiple DNase I-hypersensitive sites was identified in duodenal DNA. A representative Southern blot demonstrating these sites, indicated by arrows, is shown in Fig. 5B. Shown is DNase Itreated, XbaI-digested Transgene IV, line 2, nuclear DNA from liver and duodenum probed with a hypersensitivity probe (striped bar in Fig. 5A). The position of these sites (labeled A-G) within to the human ADA gene is shown in 5A. These sites were either absent or greatly reduced in all non-expressing tissues examined, including a more distal part of the small intestine. Identical sites were mapped for multiple lines of mice containing either Transgene IV and V, which suggests that these hypersensitive sites are not the result of transgene insertion site, but rather are inherent to fragment c.
The Duodenal Enhancer Lies within Fragment g, a 3.4-Kb Region Containing the Hypersensitive Sites-In an effort to confirm that enhancer function was associated with these hypersensitive sites Transgenes VI, VII, and VIII were created. These transgenes contain fragments e, f, and g (Fig. 1) which overlap each other such that all possess a common 3.4-kb region containing all of the duodenal DNase I-hypersensitive sites. Several lines of mice were established containing each transgene. All three transgenes maintain duodenum-specific reporter gene activity (Table I) similar to that observed for Transgene V, which contains the entire c fragment. In every line of mice from Transgenes VI-VIII, CAT expression is at least 100-fold higher in duodenum than in any other tissue, including jejunum, ileum, and colon. In most lines, duodenal CAT activity is greater than 1,000-fold higher than the next highest CAT-expressing tissue. These results map the duodenal regulatory region to the 3.4-kb segment encompassed by fragment g, which contains all of the identified hypersensitive sites A-G (see Fig. 5A).
As observed for Transgene V, there is significant variation in normalized CAT activity between lines. These line-to-line vari- ations are likely related to transgene insertion site. Three of the lines (Transgene VII, line 1, and Transgene VIII, lines 6 and 7) actually have normalized duodenal CAT activities that are lower than any of the lines from Transgene V. Because all three of these lines have a relatively high copy number, this lower activity may be at least partially the effect of squelching due to high copy number. This has been a common observation in transgenic mouse studies, as has variation in expression related to insertion site. The actual normalized CAT value observed for any one line is likely the result of the combined effects of both insertion site and copy number. However, regardless of these effects, all mouse lines from Transgenes VI-VIII do express high levels of CAT activity specifically in the duodenum (6,000 -200,000 total units).

DISCUSSION
This present study, along with a number of previous studies, indicates that the ubiquitous expression of ADA is highly regulated and that this regulation occurs via regulatory modules. In humans, we have shown that distinct intragenic regulatory modules govern expression in thymus and duodenum. In mouse, a segment in the distal promoter controls expression in fetal placenta and forestomach (24). Regulatory regions that control expression in these two tissues are likely distinct because a discrete segment specifically regulating placental expression was identified recently (26). It thus appears that multiple distinct regulatory modules are responsible for activation of ADA expression in at least those tissues that express high levels. The one-to-one correspondence of modules in the human and mouse ADA genes is at present unclear. The functional T-cell enhancer/LCR module is the only module that has been identified in both human (7,12) and mouse (18,19). Although there are many similarities in mouse and human patterns of ADA expression, there are also some significant differences. Tongue and esophagus express very high levels of ADA in adult mouse but not in humans (7). Sequences activating this expression in mouse have not been identified.
Studies employing transgenic mice containing various sections of the human adenosine deaminase gene have been very successful in identifying regions important for the high level, tissue-specific expression of ADA. A portion of the first intron has been shown previously to direct transgene expression that recapitulates the specific expression pattern observed in human thymus. However, this segment is not sufficient to activate transcription in transgenic mice in other tissues known to have high levels of ADA, especially in the small intestine. In this study we have identified a segment of the human ADA gene which is able to direct appropriate intestinal expression of a reporter gene in transgenic mice.
The putative duodenal enhancer bears some similarity to the human ADA T-cell enhancer, which by analogy suggests a possible LCR-like property of this enhancer. Both are capable of consistent tissue-specific activation, both drive ubiquitous low level expression, and both require additional sequences for full activation. Removal of the T-cell enhancer/facilitator from the transgene appears to result in somewhat lower reporter gene activity in general and greater variability in copy-normalized expression in the duodenum. This suggests that the duodenum-specific enhancer element identified may require sequences within the enhancer/facilitator region or as yet unidentified sequences for which the T-cell enhancer/facilitator fragment can compensate, to generate true copy-proportional expression in transgenic mice. Each of these enhancer segments is also able to generate low level transgene expression in all tissues assayed. The mechanism by which this generalized tissue expression occurs at low levels is not clear. A qualitative change in the promoter region might occur when either is present which allows for the ubiquitous expression observed. Additional study of these two tissue-specific enhancer regions may provide insight into this question.
Mammalian ADA has a well defined profile of expression along each axis of intestinal development (for a review of intestinal development, see Ref. 41). An anterior/posterior (A/P) gradient of expression is observed where expression in the duodenum is significantly higher than in the remaining regions of the small intestine. Along the crypt/villus (C/V) axis of the duodenum, ADA is expressed in a lineage-specific fashion. It is not found in the dividing stem cell compartment nor in the mature Paneth cell population, both of which are located in the crypt. It is expressed at high levels only in the differentiated enterocytes along the villus but not the enteroendocrine or goblet cells also found there. The identified intragenic regulatory fragment from the human ADA gene is capable of directing a transgene expression pattern that recapitulates most of the intestine-specific ADA expression profile. High reporter gene activity was limited to the duodenum along the A/P axis. The C/V expression pattern was also maintained. Villous expression was limited to the epithelium and excluded from the crypt, lamina propria, and underlying structures of the intestine. This pattern is consistent with the location of the differentiated enterocytes in which ADA has been reported to be highest. In situ studies suggest that this segment of the human ADA gene may be capable of restricting expression to the enterocyte cell lineage, but we have not eliminated the possibility that it also allows expression in the other differentiated epithelial cell types on the villus. Development and differentiation along both the A/P and C/V axes occur in the mammalian small intestine independently. There has been some evidence that distinct mechanisms may control each of these processes (42)(43)(44). Differentiation and transition from the dividing stem cell compartment of the small intestine crypt to apical extrusion from the villus occur in 3-5 days (45). Because of the unidirectional and processional nature of enterocyte differentiation and migration along the villus, a cross section through the C/V axis represents a continuum of this process. In situ experiments showed villi of average size and cellularity which expressed the transgene along their entire length. CAT mRNA first appears, just as the ADA mRNA does, as cells emerge from the crypt onto the villus. These results suggest that the C/V pattern of temporal expression is maintained. Development and differentiation along the cephalocaudal axis of the small intestine occur in a wave beginning at the duodenum and continuing caudally. In humans this process begins at about 8 weeks of gestation when nascent villi are first observed in the duodenum and is completed at roughly 12 weeks (46). In mice this process begins shortly before birth and continues postnatally until about the time of weaning (47). Timing of the increase in murine ADA seems to coincide with this differentiation process and may be linked to the final maturation of the epithelium. The ability of the transgene to be activated to full expression before the endogenous gene presents the intriguing possibility of a difference in temporal regulation between mouse and human homologs. Human ADA gene expression in fetal duodenum has not been studied, but the appearance of transgene expression in the nascent villi of very young mice suggests that human ADA may be regulated by factors involved in villus development and might be expressed quite early in human gestation when villi are first observed in the duodenum. There is some evidence to support the idea that homologous genes can be expressed at relatively different stages of intestinal development in rodents versus humans (48) and may, therefore, rely on different activational cues for the timing of these events in gestation. Interpretation of the temporal expression pattern of the transgene in mice is limited at this point without additional information regarding activation of the ADA gene in human duodenum. We have shown that although expression of the human ADA-CAT transgene does not follow the same timing as the endogenous murine ADA gene for activation along the A/P or C/V axis, the eventual adult expression pattern achieved is the same. At no time was high level precocious transgene activation observed in any section of the mouse small intestine besides duodenum, the region that would eventually support high levels of endogenous ADA. Inherent differences in temporal regulation are not the only possible explanation for the observed activation. The inability of mouse factors to recognize human sequences or the absence from the transgenes of specific temporal regulatory element(s), such as a repressor, might also explain these results. All of these possibilities are quite interesting and could provide a unique mechanism for examining the temporal events of intestinal gene regulation.
Maternal decidua has also been reported to have high levels of ADA in mice and other mammals (27,28), although a similar elevation of decidual ADA has not been described in humans. Levels of ADA in mouse decidua at the implantation site have been reported to be elevated above those observed in nonpregnant uterus and non-implantation pregnant uterus at about 9.5 days postcoitus (24,49). Treatment of pregnant mice with deoxycoformycin, an adenosine analog and potent ADA inhibitor, just before the time of this ADA elevation results in complete embryo resorption (50) and suggests a role for ADA in the maintenance of pregnancy in mice and possibly other mammals. Recent findings with ADA knockout mice showing partial rescue by expressing ADA at the fetal interface of the placenta would support the idea that in mice ADA is necessary for embryo survival (9,10,25,26). Our transgenic studies presented here indicate that neither the human promoter nor intragenic fragments b, c, or d contain sequences that activate high levels of expression specific to the maternal decidua (see placental studies under "Experimental Procedures").
This intestinal regulatory segment from the ADA gene joins a growing list of intestine-specific elements from other genes (51)(52)(53)(54)(55)(56). Many of these genes have expression patterns that differ along the A/P and/or the C/V axis from each other. Further characterization and comparison of the regulatory regions of these genes and the duodenal regulatory element of the ADA gene may give us insight into how positional identity is specified within the small intestine. These studies will also elucidate the role that all of these genes and the factors that regulate their expression play in normal gastrointestinal tract development and function. It is possible that the various hypersensitive sites identified within the human ADA duodenal regulatory region may be associated with regulatory segments with distinct functions. As mentioned earlier, studies with other genes expressed in the intestine have suggested that different elements may govern positional expression along the C/V axis than those involved in regulation along A/P axis. Our studies hint that temporal regulation may be distinct and separate as well. Future studies on the ADA duodenal regulatory region will allow us to determine if such separations of function exist and what factors and elements are associated with particular regulatory functions. A study of the gastrointestinal cis-regulatory elements of the ADA gene may also provide information about the regulation of the other purine metabolic genes that seem to be activated in a coordinate fashion within the intestine (36). It may also aid in the delineation of regulation of other duodenum-specific genes.