Abstract
It is becoming increasingly clear that epigenetic silencing of gene transcription plays a critical role in the regulation of gene expression in many biological processes. Tight regulation of immunomodulatory substances that are important for the initiation of the inflammatory cascade, such as chemoattractive cytokines, is essential to prevent initiation of unrestrained immune activation. Using the Caco-2 intestinal cell line as a model, we reveal two distinctly different mechanisms by which the gene for the neutrophil chemoattractive cytokine IL-8 is silenced. Nuclear run-on studies, as well as stably transfected reporter and marked minigene constructs, demonstrate that cellular differentiation inhibits immune-activated transcription of the IL-8 gene, a mechanism that is dependent on histone deacetylase activity. Unexpectedly, this silencing mechanism does not involve previously described regulatory elements in the IL-8 promoter but rather cis-acting regions located at a distance from the IL-8 gene locus. Genomic elements distant to the immediate IL-8 locus are also required to silence aberrant basal transcriptional activity of the IL-8 promoter in the absence of immune activation. However, in this case, silencing occurs in a histone deacetylase-independent fashion. These findings were confirmed in transgenic mice in which, in the absence of these elements, aberrant IL-8 gene activity was present primarily in the intestinal tract. Epigenetic silencing of cytokine gene transcription through distant genomic elements is an important level of gene regulation that may be relevant to the pathogenesis of immunologic disease states.
Chemoattractive cytokines, also known as chemokines, are small secreted proteins that play critical roles in the regulation of the inflammatory response. Minute quantities of chemokines can elicit potent cellular responses by acting in an autocrine, paracrine, or hormonally derived fashion (1, 2). Therefore, the expression of these chemokines is tightly regulated, primarily at the level of gene transcription. Although much is known about how these genes are transcriptionally activated in response to immune stimulation, very little is known about the mechanisms by which chemokine genes are repressed or silenced. The silencing or repression of chemokine transcription can occur in two scenerios. First, we hypothesize that mechanisms must exist that inhibit transcriptional activation of chemokine genes in the absence of inflammatory stimulation, a state that can be referred to as basal transcriptional activity. Expression of cytokine or chemokine genes in the absence of immune stimulation must remain completely silent to prevent unrestrained immune activation. Indeed, various animal models demonstrate the consequential inflammatory response that results from the aberrant or ectopic expression of these genes (3, 4, 5, 6). In particular, dysregulation of TNF-α gene expression has been recently shown to cause both chronic inflammatory arthritis as well as intestinal inflammation with characteristics similar to patients with Crohn’s disease (7).
Promoter analysis of various cytokine genes using a reporter gene approach has demonstrated significant basal transcriptional activity in the absence of immune stimulation in which expression of the endogenous gene is silent (8, 9). In part, this aberrant promoter activity may be due to constitutive expression of transcription factors shown to activate these promoters (10). For example, constitutive expression of nuclear proteins capable of binding to both the CCAAT/enhancer binding protein and/or the AP-1 element(s) of the IL-8 promoter account for the basal activity of this promoter in colon cancer cell lines. However, we have shown that the POU-homeodomain transcription factor, Oct-1, plays a role in repressing this basal activity by binding independently to an element overlapping that of the CCAAT/enhancer binding protein (10). Recently, a multiprotein complex that binds to the promoter of the monocyte chemoattractant protein-1 (MCP-1)3 gene has been shown to repress platelet-derived growth factor-induced gene transcription in the absence of a heptad sequence located in the 3′ untranslated region (UTR) (11). Finally, AT-rich elements in the 3′ UTR of the TNF-α gene have been shown to attenuate basal expression, presumably by altering mRNA stability (12).
Second, mechanisms exist that inhibit the activation of chemokine gene transcription in response to immune stimulation. For example, the state of cellular differentiation may alter the regulation of immune response genes. The intestinal epithelium is spatially segregated into a proliferating, undifferentiated compartment and a nonproliferating, differentiated compartment, which are both in the small and large intestine (13). Emerging evidence suggests that cellular differentiation of the intestinal epithelium attenuates the activation of the immune response. Colonic inflammation induces the expression of the proinflammatory cytokine IL-lβ and the injury response gene, manganese superoxide dismutase, in rat colonic epithelial cells located only in the undifferentiated crypt compartment (14, 15). Furthermore, epithelial neutrophil-activating protein 78, a neutrophil chemoattractant, has been shown to be expressed at high levels only by colonocytes in the proliferative crypt compartment in patients with inflammatory bowel disease (16, 17). Studies of transgenic chimeric mice, which disrupt the epithelial cell adhesion molecule E-cadherin along the entire crypt villus axis but not in the villus epithelium alone, produced an inflammatory bowel disease resembling Crohn’s disease (18). Finally, it has been shown that IL-1β signal transduction is disrupted in methotrexate-induced differentiation of HT-29 colon cancer cells (19).
We have been studying the regulation of IL-8 gene transcription in the Caco-2 colon cancer cell line. In the preconfluent proliferative state, these cells are relatively undifferentiated. However, several days after confluency, Caco-2 cells spontaneously develop markers of a differentiated phenotype, including the expression of digestive enzymes, certain ion transporters, tight junctions, a well-developed brush border, and a polarized morphology (20). We have shown that expression of multiple proinflammatory cytokines, including IL-8, are inhibited by spontaneous differentiation of the Caco-2 colon cancer cell line by growth to a postconfluent state (21).
We report herein that cellular differentiation of Caco-2 cells inhibits IL-8 gene expression by preventing transcriptional activation in response to immune stimulation. Unexpectedly, stably transfected reporter gene constructs, as well as marked IL-8 minigenes, demonstrate that the cis-acting regulatory elements required to silence IL-8 gene transcription upon Caco-2 cell differentiation are not contained within the immediate IL-8 gene locus. These same constructs also demonstrate significant aberrant basal activity that can be dramatically enhanced by histone acetylation, conditions under which the endogenous IL-8 gene remains completely silent. This observation was confirmed in vivo through the analysis of transgenic mice created with a similar IL-8 minigene construct. These animals demonstrated aberrant basal activity of the transgene predominantly in the intestinal tract. In contrast, Caco-2 cells stably transfected with a 75-kb genomic fragment containing the IL-8 gene demonstrate that this construct fully recapitulates the endogenous pattern of IL-8 gene expression with the absence of basal gene expression and the repression of expression by cellular differentiation. Therefore, epigenetic mechanisms involving cis-acting elements that are distant from the immediate IL-8 locus are required to silence both basal- and immune-stimulated IL-8 gene transcription.
Materials and Methods
Cell culture conditions
Caco-2 cells (obtained from American Type Culture Collection, Manassas, VA) were plated at a density of 4 × 104 cells/cm2 in 10-cm dishes containing DMEM with 10% FBS and penicillin-streptomycin as previously described (22). Preconfluent cells refer to Caco-2 cells on day 5 after plating, whereas cells grown to day 14 after plating are referred to as postconfluent.
RNA isolation, nuclear run-ons, and RT-PCR
Total RNA for Northern blots were isolated by the guanidinium thiocyanate-CsCl gradient method (23). The IL-8 gene transcription rate was determined by nuclear run-on assay using conditions described previously (24). Nuclei were isolated by Nonidet P-40 lysis and Dounce homogenization (25) from pre- and postconfluent Caco-2 cells either in the resting state or after 25 min of stimulation with IL-lβ (5 ng/ml). The DNA plasmids used were the 3′ UTR of the IL-8 cDNA (1118-bp insert amplified by RT-PCR and cloned into pKS− (Stratagene, La Jolla, CA)), a previously described cDNA of the coding region for IL-8 (21) also used for Northern blots, pKS−, pHFBA-l (human β-actin cDNA) (26), and GAPDH (27).
RT-PCR of IL-8 used the primers IL-8 (+103) and IL-8 (exon 4) (21) and the following amplification conditions: 94°C for 1 min, 55°C for 1 min, and 72°C for 1 min for 25 cycles. Using the same conditions, IL-8transcripts were amplified by RT-PCR in transgenic animal tissues using the following primers: 5′-ACTTCCAAGCTGGCCGTGGC-3′ and 5′-CAAAAACTTCTCCACAACCC-3′.
Cell transfections, reporter assays, and EMSA
Luciferase reporter gene constructs containing either 135 or 1469 bp of the IL-8 5′ flank were amplified, cloned into the pGL-2 basic reporter plasmid (Promega, Madison, WI), and sequenced as previously described (21). Stable transfectants of these constructs in Caco-2 cells were developed using a modified calcium phosphate method of transfection along with selection using G418 by cotransfection with pRc/CMV-neo (21). Transient transfections were performed using the same method of transfection with luciferase activity normalized to β-galactosidase activity (10). Conditions for nuclear protein isolation and EMSAs of NF-κB have been described previously (10).
Development of hemagglutinin (HA)-tagged IL-8 minigene constructs
The strategy for development of the marked IL-8 minigene construct, as well as its analysis in Caco-2 cells, is outlined in Fig. 3⇓. A genomic clone for human IL-8 in a P1 plasmid of ∼75 kb in size was obtained by PCR screening of a P1 genomic library (Research Genetics, Huntsville, AL). Using a previously described restriction map of IL-8 (28), the IL-8 gene locus was isolated as two EcoRI fragments cloned into the pKS− (Stratagene) plasmid, which had the SacI site in the polylinker region destroyed by Klenow blunt-end formation and subsequent relegation. Using partial digests, these two EcoRI fragments (3.3 and 1.8 kb in size) were combined to reconstitute the IL-8 gene locus in a single plasmid. This construct contains 1469 bp of the 5′ flank, the entire structural gene, and the entire 3′ UTR extending beyond the polyadenylation signal (28, 29) (Fig. 3⇓A) and was named pE3.3/E1.8. A 33-bp HA tag was then inserted into a SacI site located in exon 3 of the IL-8 gene using previously described methods, creating pE3.3(HA)/E1.8 (30). Orientation of the insert was confirmed by sequencing. A second construct containing 5 kb of the IL-8 5′ flank was developed by ligating a 5.0 PstI fragment extending from exon 1 at the 3′ end to ∼5 kb upstream into a PstI digest of pE3.3/E1.8. Orientation of the PstI fragment in this construct, named pPst5.0(HA)/E1.8, was confirmed by sequencing.
Development of a HA-tagged P1 IL-8 minigene construct by homologous recombination
To study long-range DNA effects on IL-8 gene expression in vitro, a HA tag was inserted into exon 3 of the IL-8 gene within the context of the P1 clone described above. The method used was a slightly modified version of a previously described method for the modification of bacterial artificial chromosomes by homologous recombination in Escherischia coli (31). This procedure is described schematically in Fig. 6⇓. The PstI site in exon 1 of pE3.3(HA)/E1.8 was converted to a SalI site with the addition of a linker. This plasmid was subsequently digested with SalI to release a fragment of the IL-8 genomic clone containing the HA tag in exon 3. The SalI fragment was cloned into the shuttle vector pSV1.RecA, which contains the RecA gene as well as a temperature-sensitive origin of replication (kindly provided by Dr. N. Heintz, The Rockefeller University, New York, NY). The shuttle vector (tetracycline resistant) was transformed into competent E. coli containing the P1 clone (kanamycin resistant) with the IL-8 gene locus. Cointegrates of these two plasmids were created by growing the bacteria on tetracycline and kanamycin plates at a restrictive temperature of 43°C. Three cointegrates were identified by Southern blot analysis (Fig. 6⇓B). Insertion of the HA tag in exon 3 destroys the SacI restriction site, resulting in a larger 3.3-kb band on Southern blot. Resolution of the cointegrate to complete the homologous combination by removing the pSV1.RecA plasmid was performed by growing the cointegrates on a kanamycin plate containing fusaric acid at 37°C for 3 days. The resolved P1 clones, named P1(IL-8), were then analyzed by two successive Southern blots to verify that homologous recombination had occurred and was complete. At the 5′ end, continuity of the IL-8 locus from the original P1 clone with the IL-8 genomic fragment from the shuttle vector was verified by a Southern blot identical with that described in Fig. 6⇓B. This blot also confirmed presence of the HA tag. To verify continuity of the original P1 IL-8 locus with the HA tagged genomic fragment at the 3′ end, a Southern blot of the resolved P1 clone and the original P1 plasmid was performed after digestion with either BamHI or KpnI. This blot, probed with an EcoRI/KpnI fragment of the IL-8 3′ UTR, showed an identical size band from the two plasmids digested with either restriction enzyme (data not shown).
Stable transfectants of this plasmid in Caco-2 cells were developed with NotI linearized P1 plasmid cotransfected with pRc/CMV-neo and selected by G418 resistance as described above. Southern blots were used to determine transgene copy number in two different clones, as shown in Fig. 7⇓.
Development and analysis of IL-8-transgenic mice
Mice and rats do not have an IL-8 gene. Therefore, transgenic animals can be developed using an IL-8 minigene construct without any modifications in the genomic sequence that could alter patterns of gene expression. The IL-8 minigene was released from pE3.3/E1.8 by digestion with NotI and SalI and purified by gel electrophoresis and Elutip (Schleicher & Schuell, Keene, NH) column chromatography. Following ethanol precipitation, the minigene insert was dissolved in Milli-Q water and delivered to the transgenic core facility at the University of Pennsylvania School of Medicine (Philadelphia, PA). The male pronuclei of zygotes were microinjected with 2000–4000 copies of the recombinant DNA (32). The morulae were transferred to the uteri of pseudo pregnant female mice. Tail DNA isolated from the offspring of these females were analyzed after digestion with EcoRI by Southern blot using an EcoRI fragment of the IL-8 3′ UTR (E1.8) as a probe to identify founder animals and to determine relative copy number. Founder animals were crossed with normal C57BL/6J mice to produce F1 progeny for analysis. The transgenic lines were mated and maintained at the University of Pennsylvania animal facilities.
Transgenic mice were sacrificed under pentobarbital sedation and cervical dislocation. The following tissues were removed: brain, lung, heart, kidney, skeletal muscle, testis, liver, pancreas, spleen, stomach, small intestine (jejunum and ileum), and colon. RNA was isolated from each tissue, and RT-PCR was performed to determine the abundance and distribution of mRNA expression for IL-8. To verify that the IL-8 minigene could be activated appropriately by an inflammatory stimulus, colonic inflammation was induced in both transgenic and normal littermates by the administration of an acetic acid enema (14). A flexible catheter was inserted 2 cm into the colon and followed by the administration of either 0.5 ml of either PBS or 10% acetic acid in PBS. After 6 h, the animals were sacrificed and the distal half of the colon isolated for RNA extraction and RT-PCR for IL-8 as described above.
Results
Immune-activated transcription of the IL-8 gene is silenced upon Caco-2 cell differentiation by a histone deacetylase-dependent mechanism
Stimulation of the colon cancer cell line, Caco-2, with IL-1β induces high level expression of IL-8 mRNA (33). In contrast, isolated colonic epithelial cells, like Caco-2 cells, do not respond to stimulation with TNF-α, perhaps due to the absence of TNF-αR expression (33, 34, 35, 36). We have previously shown that spontaneous differentiation of Caco-2 cells during growth to a postconfluent state inhibits both the mRNA and protein expression for IL-8 (21). Nuclear run-ons show that immune stimulation of preconfluent Caco-2 cells strongly activates transcription of the IL-8 gene, a response that is inhibited during the process of cellular differentiation in postconfluent Caco-2 cells (Fig. 1⇓A). Many mechanisms that result in the repression of gene transcription involve covalent modification of nucleosomes through the removal of acetyl groups from histone proteins through histone deacetylases (37, 38). Inhibition of histone deacetylase activity by either butyrate or trichostatin A has been shown to induce the activation of genes repressed by histone deacetylases (39). Therefore, the activation of IL-8 mRNA expression in postconfluent Caco-2 cells with increasing concentrations of butyrate is consistent with a histone deacetylase-dependent mechanism of silencing (Fig. 1⇓B).
Transcriptional activation of the IL-8 gene is inhibited in postconfluent Caco-2 cells by a histone deacetylase-dependent mechanism. A, Nuclear run-on experiments performed using nuclei isolation resting and IL-1β stimulated pre- and postconfluent Caco-2 cells. Arrows point to two different DNA targets specific for IL-8. B, Northern blot for IL-8 using RNA isolated from pre- and postconfluent Caco-2 cells stimulated with IL-1β. Postconfluent cells were treated with the indicated concentrations of sodium butyrate for 48 h before immune stimulation.
Neither the IL-8 promoter nor the immediate IL-8 gene locus contains elements required to silence IL-8 gene transcription upon Caco-2 cell differentiation
All previous described cis-acting regulatory elements of the IL-8 gene have been located within the first 135 bp upstream of the transcriptional initiation site (10, 40). Therefore, to determine whether cellular differentiation of Caco-2 cells alters the transcriptional activation of this promoter, clonal populations of Caco-2 cells stably transfected with luciferase reporter genes regulated by the IL-8 5′ flank were studied. We have previously shown that the stimulation of these clones with IL-1β induces high-level luciferase activity (21). Although IL-8 mRNA expression is repressed when these clones are grown to a postconfluent state, induction of luciferase activity is actually greater in postconfluent cells (Fig. 2⇓A). This clearly demonstrates that the elements in the IL-8 gene required to silence expression in postconfluent Caco-2 cells are not located within the immediate 5′-flanking region of the gene. In fact, the observed increase in promoter activation in postconfluent cells is consistent with the increase in NF-κB binding activity observed by EMSA using nuclear extracts isolated from postconfluent cells (Fig. 2⇓B).
Luciferase reporter gene analysis of the IL-8 promoter shows greater activation in postconfluent Caco-2 cells. A, IL-1β induced promoter activation of luciferase reporter genes containing either 135 or 1469 bp of the IL-8 5′ flank in permanently transfected Caco-2 cells in either the pre- or postconfluent state. Activation of endogenous IL-8 mRNA expression in these same cells is shown below. B, EMSA of NF-κB in pre- and postconfluent Caco-2 cells stimulated with IL-1β. NS, nonspecific.
To determine whether the cis-acting regions of the IL-8 gene necessary to inhibit IL-8 gene transcription were located anywhere within the immediate IL-8 gene locus, an IL-8 minigene marked with a 33-bp HA tag in exon 3 was stably transfected into Caco-2 cells. This strategy, shown in Fig. 3⇓A, allows the direct comparison of the IL-8 transgene to endogenous IL-8 gene expression in the same cell line. This approach has been useful in identifying cis-acting elements important for the regulation of another chemokine gene (41). Analysis by RT-PCR was performed on six independent clones containing either 1469 bp (E3.3(HA)/E1.8) or ∼5 kb (Pst5.0(HA)/E1.8) of the 5′ flank, the entire structural gene, as well as the entire 3′ UTR including the endogenous polyadenylation signal. The results showed that IL-8 mRNA expression of the marked minigene was either equal to or greater in postconfluent compared with preconfluent cells (Fig. 3⇓B). Although different from the endogenous IL-8 expression pattern, it is consistent with the pattern observed in the luciferase reporter studies (Fig. 2⇑A), demonstrating that even the entire immediate IL-8 gene locus lacks elements required to silence IL-8 expression upon Caco-2 cell differentiation.
Development and analysis of IL-8-marked minigene constructs in pre- and postconfluent Caco-2 cells. A, Two IL-8 minigene constructs were marked in exon 3 with a HA tag resulting in a 33-bp increase in RT-PCR amplicon size compared with that amplified from endogenous IL-8 mRNA expression. These two minigenes, named E3.3(HA)/E1.8 and Pst5.0(HA)/E1.8, contain 1469 bp and ∼5 kb of the IL-8 5′ flank, respectively. B, RT-PCR analysis of IL-1β stimulated IL-8 gene expression in two independent clones of Caco-2 cells, permanently transfected with marked IL-8 minigene constructs, grown to either the pre- or postconfluent state.
Basal transcriptional activity of the IL-8 gene is silenced by a mechanism independent of histone deacetylation that requires elements distant from the immediate gene locus
Additional analysis of the Caco-2 clones permanently transfected with the marked IL-8 minigenes showed that they all demonstrated expression of transgene activity in the absence of immune activation with IL-1β (Figs. 3⇑B and 4D). This aberrant basal activity is present in the absence of any endogenous gene expression. We have previously shown that stimulation of Caco-2 cells with IL-1β induces activation of the IL-8 promoter, as measured using a luciferase reporter assay, by 50- to 100-fold (10). However, even in the absence of immune stimulation, the IL-8 promoter shows substantial basal activity that is >100-fold greater than the background activity observed with the promoterless vector (Fig. 4⇓A). Indeed, the magnitude of this basal IL-8 promoter activity approaches that observed with the potent sucrase-isomaltase intestinal promoter characterized specifically in the Caco-2 cell line (22). This aberrant basal activity, as observed with either luciferase reporter constructs or IL-8-marked minigenes, can be dramatically enhanced by inhibiting histone deacetylase activity using either butyrate or trichostatin A (Fig. 4⇓, B and D). In contrast, endogenous IL-8 gene expression remains completely absent under the same conditions (Fig. 4⇓, C and D), demonstrating that the silencing of basal activity is independent of histone deacetylase activity.
Aberrant basal activity of the IL-8 promoter, observed in either luciferase reporter constructs or a marked IL-8 minigene constructs, is prevented in the endogenous IL-8 gene by a histone deacetylase-independent mechanism. A, Luciferase reporter studies comparing basal activity of the IL-8 promoter to that of the promoter for an intestinal gene, sucrase-isomaltase (SI). B, Comparison of IL-1β-stimulated and butyrate-induced basal IL-8 promoter activity in Caco-2 cells permanently transfected with the −135(IL-8) luciferase (Luc) reporter gene construct. C, Northern blot of endogenous IL-8 gene expression in Caco-2 cells, with or without IL-1β stimulation, treated with various concentrations of sodium butyrate for 48 h. D, Regulation of endogenous IL-8 vs marked IL-8 minigene (E3.3(HA)/E1.8, clone 2) expression, analyzed by RT-PCR, in the presence of either sodium butyrate or trichostatin A.
Butyrate had a different effect on immune-activated IL-8 gene transcription. Upon activation of gene transcription by IL-1β-mediated immune stimulation, expression of either the IL-8 reporter construct or the endogenous IL-8 gene was enhanced by sodium butyrate (Fig. 4⇑, B and C). This response has also been observed for the macrophage-inflammatory protein (MIP)-2 gene in a different intestinal cell line (42).
The immediate IL-8 gene locus shows aberrant transcriptional activity in transgenic mice
To confirm that the IL-8 transgene would also demonstrate aberrant basal activity in vivo, as well as to determine whether such an activity occurs in a tissue-restricted fashion, transgenic mice were developed using the IL-8 minigene construct pE3.3/E1.8. Because mice do not have an endogenous IL-8 gene, the IL-8 transgene used for the development of these animals did not require insertion of the HA tag used in the cell culture experiments. This provided a unique opportunity to exclude any artifactual results of this tag on patterns of IL-8 gene expression.
Two founders were identified by Southern blot, F#8 had a 6-fold higher copy number than F#3 (Fig. 5⇓A). Transcripts for IL-8 were determined by RT-PCR using total RNA isolated from various tissues of animals derived from these two founder lines (Fig. 5⇓B). Basal activity of the IL-8 trangene was detected in several tissues from both the high and low copy number transgenic mice. Interestingly, expression was not detected in all tissues but was rather restricted to a subset of tissues. Although there was some variability in the pattern of IL-8 expression between the two founder lines, detectable expression in the small intestine (jejunum and ileum) and colon was consistently present in both transgenic lines. It is also interesting that no expression was detected in the stomachs of either transgenic line. Finally, in Fig. 5⇓C, we show that the IL-8 minigene can be activated by acetic acid-induced colonic inflammation in both the low and high copy number transgenic animals. These results also show that the relative levels of basal to activated IL-8 minigene expression are similar to those observed in the Caco-2 cell culture model system (Figs. 3⇑B and 4D). In total, the results in transgenic mice validate and extend the findings in cell culture with respect to aberrant basal activity.
IL-8 minigene construct (E3.3/1.8) exhibits aberrant basal gene expression in transgenic animals. A, Southern blot showing relative transgene copy number in two founder lines. B, Expression of IL-8 transgene expression, assessed by RT-PCR, in various tissues isolated from two different founder lines (Sp, spleen; K, kidney, T, testis; F, fat; M, muscle; Br, brain; Sk, skin; St, stomach; J, jejunum; I, ileum; C, colon; Lu, lung; H, heart; Li, liver; and P, pancreas). C, IL-8 minigene expression in transgenic animals in the presence of colonic inflammation induced by acetic acid.
Stable integration of the IL-8 gene within the context of a 75-kb genomic fragment faithfully recapitulates both differentiation-dependent and basal silencing the IL-8 gene
The studies above show that two independent mechanisms of IL-8 gene silencing cannot be recapitulated even when a marked IL-8 minigene is stably integrated into the chromatin of Caco-2 cells. Therefore, genomic elements that reside outside the immediate IL-8 gene locus must play a critical role in silencing this gene. Indeed, experiments from both Drosophila and mouse transgenics have shown that distant elements such as enhancers, locus control regions, or insulators are sometimes required to fully recapitulate patterns of endogenous gene expression (43, 44, 45, 46). These elements may reside 50 kb or more away from the immediate gene locus.
Therefore, we inserted a HA tag into exon 3 of the IL-8 gene within the context of a 75-kb P1 human genomic clone to determine whether it was possible to develop a model system in which a transgene could recapitulate the silencing patterns of the endogenous IL-8 gene. The strategy to create this extended minigene, named P1(IL-8) and described in Fig. 6⇓A, used homologous recombination in a fashion similar to that described previously (31). Southern blots were used to identify cointegrates (Fig. 6⇓B) as well as to confirm appropriate resolution of the homologous recombination resulting in the production of an IL-8 transgene tagged in exon 3 (data not shown). Clonal populations of Caco-2 cells stably transfected with this construct were subsequently produced. Southern blot analysis was used to determine copy number insertion of all the IL-8 minigene constructs permanently transfected into Caco-2 cells (Fig. 7⇓). To verify that the patterns of IL-8 minigene expression were copy number- and insertion site-independent, multiple Caco-2 clones were analyzed with varying minigene copy numbers, and several minigene pools were also analyzed.
Insertion of a HA tag into exon 3 of the IL-8 gene locus in a P1 clone by homologous recombination. A1, Homologous recombination of an IL-8 genomic fragment containing the HA tag in the shuttle vector pSV1. RecA with the IL-8 locus in the P1 clone results in the cointegrate shown in A2. A3, Resolution of the cointegrate by growth of bacteria in fusaric acid removes the shuttle vector and completes the homologous recombination. B, Strategy for identifying cointegrate clones by Southern blot. Three independent cointegrates were identified in lanes 1, 2, and 4.
Determination of minigene copy number by Southern blot. Strategy to distinguish the IL-8 minigene from the endogenous IL-8 gene through the use of three restriction enzymes, XbaI, SacI, and EcoRI, and a Southern probe that hybridizes to the 5′ end of these fragments. Copy number was determined in Caco-2 cell clones containing three different IL-8 minigene constructs: E3.3(HA)/E1.8, Pst5.0(HA)/E1.8, and P1(IL-8).
Analysis of two different clones of Caco-2 cells containing different copy numbers of the P1(IL-8) minigene show that the transgene pattern expression now recapitulates that of the endogenous gene (Fig. 8⇓A). There is a significant decrease in expression of both endogenous and minigene mRNAs in postconfluent cells. In postconfluent cells, the induction of DRA, a ion transport gene expressed specifically in the differentiated surface epithelium of the colon (27), verified the differentiated phenotype of clone 2. Kinetic studies have shown that the inhibition of IL-8 gene expression corresponds temporally with the induction of DRA mRNA expression (data not shown). Expression of the minigene was also dramatically reduced in pooled Caco-2 transfectants, verifying that silencing of IL-8 minigene expression in postconfluent cells occurs in a copy number- and insertion site-independent fashion (Figs. 7⇑ and 8⇓A). Also apparent in individual or pooled colonies of Caco-2 cells transfected with the P1 minigene for IL-8 was the complete absence of minigene expression in the unstimulated (basal) state (Fig. 7⇑, A and B). Furthermore, unlike the shorter IL-8 minigene constructs (Fig. 4⇑D) or the luciferase reporter constructs (Fig. 4⇑D), aberrant basal expression of the P1 minigene could no longer be significantly enhanced by butyrate (Fig. 8⇓B) or trichostatin A (data not shown). This was true for both individual and pooled colonies of the P1 minigene transfectants (Fig. 8⇓B). Consistent with a histone deacetylase-independent mechanism of silencing, inhibition of DNA methylation using 5-aza-2′deoxycytidine did not induce basal expression of either endogenous IL-8 gene or a P1 minigene (Fig. 8⇓C). Therefore, in summary, the P1(IL-8) minigene recapitulated endogenous IL-8 gene silencing upon Caco-2 cell differentiation and in the basal unstimulated state.
Within the context of the P1 clone, IL-8 minigene expression parallels that of the endogenous IL-8 gene in postconfluent Caco-cells and upon treatment with sodium butyrate. A, Comparison of P1(IL-8) minigene to endogenous IL-8 gene expression in both individual clones and pooled transfectants of Caco-2 cells in the pre- and postconfluent state. B, Comparison of sodium butyrate induced basal expression of the P1(IL-8) transgene to that of the endogenous gene in two stably transfected clones and pooled colonies of Caco-2 cells. C, Effect of 5-aza-2′deoxycytidine (48 h) on the basal expression of endogenous IL-8 and a P1(IL-8) minigene in Caco-2 cells.
Discussion
In contrast to prokaryotes, the transcriptional activation of genes in eukaryotes can be described as restrictive due to the fact that eukaryotic DNA is packaged into chromatin templates (47). That is to say, basal promoters are virtually inactive at a ground state, and essentially all eukaryotic genes require activators to initiate transcription. However, although chromatin structure per se is responsible for maintaining a restrictive ground state, it is not sufficient to completely block normal or cryptic activators. Indeed, in many biological situations, it is essential to silence completely the expression of particular genes. As a consequence, eukaryotes have evolved specialized mechanisms to repress gene transcription. At the core of most silencing mechanisms is the alteration of chromatin structure with covalent modification of histones through the recruitment of histone deacetylases (38, 48). Such mechanisms include modification of DNA by methylation (49, 50), heterochromatin silencing (varigation effects in Drosophila as well as centromeric and telomeric silencing effects in Drosophila and yeast, respectively) (43, 51), the related polycomb-dependent silencing observed in Drosophila development (52, 53), and the association of histone deacetylases with specific corepressors such as N-Cor, SMRT, Mad, and YY1 (48). More recently, histone-independent silencing mechanisms have also been described (54, 55).
In this report, we characterize a model system that reveals two distinctly different mechanisms by which the IL-8 gene is transcriptionally silenced, both of which require the presence of cis-acting regions at a distance from the immediate IL-8 gene locus. Differentiation of Caco-2 cells inhibits IL-8 gene transcription in response to immune stimulation with IL-1β. Treatment of postconfluent Caco-2 cells with either sodium butyrate or trichostatin A shows that histone deacetylation is required for this repressive effect. Interestingly, none of the previously described cis-acting elements of the IL-8 promoter play a role in this effect (40). Differentiation of Caco-2 cells also represses the expression of other immune response genes such as MCP-1. Although a mechanism has been described that is responsible for the transcriptional repression of this gene (11), it is unlikely to be responsible for the effect observed in Caco-2 cells. In contrast to our cell culture model, repression of MCP-1 transcription as described by Sridhar et al. does not inhibit activation of the endogenous MCP-1 gene in response to platelet-derived growth factor due to a TTTTGTA heptad repeat.
Although further studies will be required to define the precise mechanism by which IL-8 gene transcription is silenced upon Caco-2 cell differentiation, several characteristics are similar to those observed in heterochromatic silencing. First, the silencing effect in postconfluent Caco-2 cells is dependent upon histone deacetylase activity. Second, the cis-acting regions required for silencing of the IL-8 gene are located some distance from the TATA box, >6.5 kb upstream and/or 3.7 kb downstream. This, of course, does not eliminate the possibility that a histone deacetylase activity associated with more classical transcriptional corepressors may be involved (48). However, long-range transcriptional silencing effects spanning many kilobases are certainly more consistent with heterochromatin-mediated effects (43). Finally, heterochromatic effects important for polycomb group-mediated silencing are critical for cellular differentiation during Drosophila development (52), and they are also perhaps important for cellular differentiation in postconfluent Caco-2 cells. The physiologic significance of proinflammatory cytokine gene silencing by chromatin-mediated effects remains to be determined. However, due to the central role that chromatin plays in all biological processes involving DNA, it has been suggested that alteration in chromatin structure may influence many fundamental biological processes, likely resulting in the development of specific diseases (56). In this regard, it is interesting that proteins involved in nucleosome remodeling and heterochromatin formation have been shown to be either mutated or the target of autoantibodies in specific malignant and autoimmune disease processes (57, 58).
The mechanism by which basal activity of IL-8 gene transcription is silenced is distinctly different from the silencing observed in postconfluent Caco-2 cells. First, basal silencing of the IL-8 gene can be completely reversed by immune stimulation with IL-1β. Second, we show that this basal transcriptional silencing is independent of histone deacetylase activity. Previous studies have shown that NF-κB mediates activation of transcription by reconfiguring chromatin structure by nucleosomal alterations (59). Therefore, in the absence of NF-κB activation by immune stimulation with IL-1β, the IL-8 promoter remains in a basally repressed state and cannot be activated by histone deacetylase inhibitors. Significant basal transcriptional activity of cytokine promoter regions has been observed with several promoters analyzed using reporter genes. For example, high-level basal expression has been demonstrated for the NF-κB-dependent gene MCP-1/JE in Rat-1 cells (9). Interestingly, similar to the IL-8 promoter, binding of nuclear proteins to an AP-1 element is essential for this basal expression. High-level basal expression has also been described for the MIP-1β promoter in a macrophage cell line (8). These results demonstrate that aberrant high-level basal expression of reporter gene expression is a more generalized property that is not specific to either the IL-8 promoter or Caco-2 cells.
In the absence of endogenous gene expression, our results suggest that this aberrant transcriptional activity is an artifact of the reporter systems used. This may be due to the absence of AT-rich regions in the 3′ UTR required to destabilize mRNA, as has been demonstrated for the TNF-α gene (12). In contrast, we demonstrate that the significant basal activity observed with our IL-8 transgene construct, which contains these AT-rich regions in the 3′ UTR as well as the polyadenylation signal, is due to aberrant transcriptional activation. It is interesting that this aberrant basal activity can be augmented dramatically by the use of histone deacetylase inhibitors while the endogenous gene remains silent. However, upon immune activation, expression of the endogenous gene is augmented in parallel with the increased expression observed in the luciferase reporter or the minigene constructs. This pattern of cytokine gene expression has also been observed with the MIP-2 gene in a different cell line, IEC 6, suggesting that silencing of basal transcriptional activity may be a general mechanism that may regulate many immune response genes (42).
The silencing of basal activity may be particularly relevant in the large intestine, in which concentrations of butyrate normally range from 20 to 40 mM due to the fermentation of undigested carbohydrates by the colonic bacterial flora (60). In this regard, it is interesting that basal activation of the IL-8 locus is present primarily in the intestinal tract of transgenic animals (Fig. 5⇑, B and C). Additional investigation will be necessary to determine the specific cell type responsible for this aberrant activity. In turn, such studies may provide new insights into tissue-restricted mechanisms by which chromatin structure may be modified. Although basal activity of the IL-8 gene in transgenic animals is easily detected by RT-PCR, the level of activity is relatively low, similar to that observed in cell culture (Figs. 3⇑B and 4D). Nevertheless, the results in transgenic animals provide two important insights. First, the results are a proof of principle that the aberrant activity of the IL-8 promoter observed in cell culture has relevance in vivo. This observation is supported by the presence of aberrant IL-8 mRNA expression in two independent transgenic founder lines with different insertion sites and copy numbers. Second, because the IL-8 transgene does not contain any bacterial DNA, artifactual alterations in gene expression caused by extraneous plasmid DNA are excluded. Indeed, previous studies have shown that prokaryotic vector DNA can alter transgene expression in vivo (61). Finally, because the IL-8 locus used in the development of the transgenic animals did not contain the HA tag used in the cell culture experiments, aberrant activity due to modification of the IL-8 transgene can be excluded.
Although it is possible that recently described histone deacetylase-independent mechanisms of transcriptional silencing may be involved in repressing basal activity of the IL-8 gene (54, 55), the complete reversal of this silencing by immune stimulation with IL-1β makes these mechanisms less plausible. It is more likely that there are flanking boundary elements, such as insulators, which silence basal IL-8 gene transcription induced by interactions with euchromatic chromatin associated with heterologous enhancing elements (46, 62). Therefore, insulators prevent enhancers from promiscuously activating a neighboring gene locus. In this model, either the IL-8 minigene constructs (E3.3(HA)/E1.8 or Pst5.0(HA)/E1.8) or the luciferase reporter genes, which lack insulating elements, would exhibit enhanced basal transcriptional activity induced by heterologous enhancing elements in proximity to the transgene insertion site. It would be predicted that the magnitude of basal activity observed with either chromatin-integrated construct should be insertion site-dependent. Indeed, although we observed basal activity in all of the individual clones of Caco-2 cells permanently transfected with either luciferase reporter genes or the IL-8 minigenes (E3.3(HA)/E1.8 or Pst5.0(HA)/E1.8), there was a significant variation in the magnitude of this basal activity (data not shown). In contrast, the P1(IL-8) minigene, which presumably contains insulating elements flanking the IL-8 gene locus, would not be aberrantly activated. Insulators have been described for several genes to be associated with gene clusters, such as globin genes, in which they may help to prevent aberrant activation by positive regulatory elements associated with adjacent genes (62). Recently, Bell et al. identified a zinc finger DNA-binding protein, CTCF, that exhibits enhancer-blocking activity when bound to vertebrate insulator elements (62). The same insulating function may be necessary for chemokine genes that are also located within gene clusters on chromosomes 4 and 17 (63).
There has been much recent interest in mechanisms involving the silencing of gene transcription. Previous studies have carefully characterized elements within the IL-8 promoter that are critical for the regulation of IL-8 gene transcription (10, 40). Herein, we provide evidence that additional mechanisms, located at a distance from the immediate IL-8 locus, are critical for two distinctly different types of transcriptional silencing. It is likely that these mechanisms are not unique for IL-8 gene regulation but instead are important for the regulation of multiple chemokine genes. Further characterization of the mechanisms responsible for epigenetic silencing of chemokine gene transcription may provide important new insights into the pathogenesis of inflammatory diseases.
Acknowledgments
We thank Dr. N. Huang for her technical expertise in this project, Dr. N. Heintz for the protocol and plasmids used in the development of the plasmid P1(IL-8), and Dr. S. Liebhaber for critical review of this manuscript.
Footnotes
-
↵1 This work was supported by Public Health Service Grants AI39368 and DK54893 (to G.D.W.) and Center Grant DK50306.
-
↵2 Address correspondence and reprint requests to Dr. Gary D. Wu, Division of Gastroenterology, Pennsylvania School of Medicine, 600 Clinical Research Building, 415 Curie Boulevard, Philadelphia, PA 19104. E-mail address: gdwu{at}mail.med.upenn.edu
-
↵3 Abbreviations used in this paper: MCP-1, monocyte chemoattractant protein-1; UTR, untranslated region; HA, hemagglutinin; MIP, macrophage-inflammatory protein.
- Received January 11, 2001.
- Accepted April 3, 2001.
- Copyright © 2001 by The American Association of Immunologists
References
In this issue
