Tumor Necrosis Factor-induced Long Myosin Light Chain Kinase Transcription Is Regulated by Differentiation-dependent Signaling Events

Myosin light chain kinase (MLCK) is expressed as long and short isoforms from unique transcriptional start sites within a single gene. Tumor necrosis factor (TNF) augments intestinal epithelial long MLCK expression, which is critical to cytoskeletal regulation. We found that TNF increases long MLCK mRNA transcription, both in human enterocytes in vitro and murine enterocytes in vivo.5′-RACE identified two novel exons, 1A and 1B, which encode alternative long MLCK transcriptional start sites. Chromatin immunoprecipitation (ChIP) and site-directed mutagenesis identified two essential Sp1 sites upstream of the exon 1A long MLCK transcriptional start site. Analysis of deletion and truncation mutants showed that a 102-bp region including these Sp1 sites was necessary for basal transcription. A promoter construct including 4-kb upstream of exon 1A was responsive to TNF, AP-1, or NFκB, but all except NFκB responses were absent in a shorter 2-kb construct, and all responses were absent in a 1-kb construct. Electrophoretic mobility shift assays, ChIP, and site-directed mutagenesis explained these data by identifying three functional AP-1 sites between 2- and 4-kb upstream of exon 1A and two NFκB sites between 1- and 2-kb upstream of exon 1A. Analysis of differentiating epithelia showed that only well differentiated enterocytes activated the 4-kb long MLCK promoter in response to TNF, and consensus promoter reporters demonstrated that TNF-induced NFκB activation decreased during differentiation while TNF-induced AP-1 activation increased. Thus either AP-1 or NFκB can up-regulate long MLCK transcription, but the mechanisms by which TNF up-regulates intestinal epithelial long MLCK transcription from exon 1A are differentiation-dependent.

Myosin light chain kinases (MLCKs) 3 are essential regulators of actomyosin function (1). Mammalian MLCK is encoded by two separate genes; the first is expressed in skeletal and cardiac muscle whereas the second, smooth muscle MLCK, generates three different proteins, long MLCK, short MLCK, and telokin, as a result of tissue-specific transcription from separate promoters (1)(2)(3)(4)(5)(6). Like all MLCKs, the 130-kDa smooth muscle, or short, MLCK includes a catalytic domain and a separate Ca 2ϩ -calmodulin-dependent regulatory domain.
Long MLCK is a 210-kDa protein originally identified in cultured cells, embryonic tissues, and endothelium (2,4,7,8). It contains the same catalytic and regulatory domains as short MLCK but also includes an N-terminal extension that contains six immunoglobulin-like modules and two tandem actin binding domains (9). Long MLCK must therefore be transcribed from an upstream promoter, although this promoter has not been identified.
Long MLCK is the principal form expressed in intestinal epithelial cells (8) and expression of long MLCK is increased by tumor necrosis factor (TNF) treatment of cultured intestinal epithelial monolayers (10,11). These increases in MLCK protein expression are accompanied by increased myosin light chain phosphorylation which, in turn, leads to disruption of the epithelial barrier (12)(13)(14). This long MLCK-dependent disruption of epithelial barrier function is an important component of intestinal disease (15). For example, long MLCK knock-out mice are protected from TNF-dependent intestinal epithelial MLC phosphorylation, barrier loss, and diarrhea (16). Similar protection is conferred by pharmacological intestinal epithelial MLCK inhibition (16). While the functional significance of increased intestinal epithelial long MLCK expression has not been defined in humans, it is notable that intestinal epithelial cells of patients with inflammatory bowel disease also display increased long MLCK expression (17). This is remarkable, as inflammatory bowel disease patients are known to have epithe-lial barrier defects that can be corrected by anti-TNF therapy (18). Thus, inducible long MLCK expression may be a critical regulator of intestinal epithelial barrier function and disease pathogenesis.
Despite data indicating the potential biological and therapeutic potential of harnessing inducible long MLCK expression, the factors that regulate this process are unknown. Using the model of TNF-induced long MLCK expression in intestinal epithelia, we now show this process to be caused by transcriptional activation. TNF increased long MLCK mRNA transcripts in cultured human intestinal epithelial cells in vitro and in mouse intestinal epithelial cells in vivo. Two novel human long MLCK transcriptional start sites, corresponding to novel exons, were identified and the promoter corresponding to one was cloned. Moreover, the sites necessary for basal and TNFinducible transcription have been defined. Remarkably, the requirements for TNF-inducible long MLCK transcription differ in undifferentiated and well differentiated intestinal epithelial cells, suggesting that long MLCK transcription can be regulated by multiple mechanisms. These data therefore define regulation of long MLCK transcription under basal conditions as well as following cytokine stimulation and suggest that the factors responsible for long MLCK induction are dependent on cellular differentiation state.

EXPERIMENTAL PROCEDURES
Cell Culture and Cytokine Treatment-Caco-2 BBE cells (19) expressing the intestinal Na ϩ -glucose cotransporter SGLT1 (20) were maintained and cultured as monolayers as described previously (8). For studies of differentiation and cytokine responses, monolayers grown on Transwell semipermeable supports (Corning-Costar, Acton, MA) were used 3-14 days after reaching confluence, as indicated. Recombinant human IFN-␥ and TNF (R&D Systems, Minneapolis, MN) were routinely added to the basal chamber without manipulating the apical media, as described previously (10). IFN-␥-priming was performed by incubation for 18 h with 10 ng/ml IFN-␥, and TNF was used at 2.5 ng/ml, both in the basal compartment (10).
Animal Model-7-10-week-old wild-type C57BL/6 female mice were injected intraperitoneally with either 5 g of recombinant murine TNF (PeproTech Inc., Rocky Hill, NJ) in 200 l of phosphate-buffered saline or with phosphate-buffered saline alone, as described previously (16). Animals were sacrificed 4 h after injection and jejunal epithelia isolated as previously described (16). Briefly, a section of intestine was opened lengthwise, and washed at 4°C in Ca 2ϩ -and Mg 2ϩ -free HBSS (CMF-HBSS), transferred to CMF-HBSS containing 10 mM dithiothreitol and 50 nM calyculin A (Calbiochem, San Diego, CA), incubated for 30 min at 4°C, and transferred to a fresh tube containing CMF-HBSS with 1 mM EDTA and 50 nM calyculin A. After incubation at 4°C for 1 h, epithelial cells were dislodged by vigorous shaking, harvested by centrifugation, and placed in TRIzol (Invitrogen, Carlsbad, CA) for mRNA analysis or SDS-PAGE sample buffer for immunoblot. Animal studies were carried out in accordance with National Institutes of Health guidelines under protocols approved by the Institutional Animal Care and Use Committee at The University of Chicago.
5Ј Rapid Amplification of cDNA Ends (5Ј-RACE)-Total RNA was isolated chromatographically from Caco-2 BBE cells (Qiagen, Valencia, CA), and mRNA was purified on poly-T beads (Qiagen). Antisense primer (5Ј-GAATCGTCCCATCT-GTC-3Ј) positioned within exon 4 was used for first-strand cDNA synthesis. Subsequent amplification was performed using the 5Ј-RACE system for rapid amplification of cDNA ends (Invitrogen). In brief, cDNA was tailed with recombinant TdT and linker (dC) oligonucleotide. 5Ј-RACE was performed by incubating with an aliquot of first-strand cDNA, a RACE primer positioned within exon 3 (5Ј-CCGCTGGTGATGGGT-TGC-3Ј), and the abridged anchor primer. PCR conditions were initial denaturation at 94°C for 3 min, 38 cycles of 30 s at 94°C, 1 min at 58°C, and 1 min at 72°C. The final extension step at 72°C was carried out for 10 min. The resulting products were cloned into pCR2.1-TOPO (Invitrogen) and sequenced to determine the transcriptional start site(s).
In Silico Sequence Analysis-Genomic DNA sequences were downloaded from the NCBI data base. In silico transcription factor binding analysis was carried out using the TFSearch program with a threshold of 85. This program searches highly correlated sequence fragments versus the TFMATRIX transcription factor binding site profile data base within the TRANSFAC resources (21).
Real Time RT-PCR-Cell extracts in TRIzol were sonicated, and RNA was extracted with chloroform, precipitated with isopropyl alcohol, and resuspended in diethyl pyrocarbonatetreated water. RNA was further purified using a RNeasy mini kit (Qiagen). cDNA was generated from 2 g of RNA using Thermoscript reverse transcriptase (Invitrogen) and random hexamer primers in a 25-l reaction. Long MLCK mRNA expression was determined by SYBR green real-time PCR through 50 cycles using the MyiQ Real-Time PCR Detection System (Bio-Rad). GAPDH was used as a reference. Primers were 5Ј-CAC-CGTCCATGAAAAGAAGAGTAG-3Ј and 5Ј-GAGAGGCC-CTGCAGGAAGATGG-3Ј for detection of human long MLCK, 5Ј-GCGTGATCAGCCTGTTCTTTCTAA-3Ј and 5Ј-GCCCCATCTGCCCTTCTTTGACC-3Ј for detection of murine long MLCK, and 5Ј-CTTCACCACCATGGAGAA-GGC-3Ј and 5Ј-GGCATGGACTGTGGTCATGAG-3Ј for detection of human and murine GAPDH.

SDS-PAGE and Immunoblot
Analysis-Cell lysates were separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes, as described previously (10,16). MLCK was detected using the K36 monoclonal antibody (Sigma). After incubation with peroxidase-conjugated secondary antibodies (Cell Signaling Technology, Danvers, MA), blots were visualized by enhanced chemiluminescence as described previously. Densitometry of immunoblot data were performed using Metamorph 6.2 (Universal Imaging Corp., Downingtown, PA).

TNF Up-regulates Long MLCK Transcription and Protein
Expression in Vitro and in Vivo-TNF treatment of IFN-␥primed Caco-2 intestinal epithelial cell monolayers causes bar-rier dysfunction via increased myosin II regulatory light chain (MLC) phosphorylation (10,12). We and others (10,11) have recently shown that these increases in MLC phosphorylation are accompanied by increased long MLC kinase (MLCK) protein expression. Typically, a 1-2-fold increase in MLCK protein, as detected by SDS-PAGE immunoblot, is induced by TNF ( Fig. 1A; p Ͻ 0.02). Quantitative real time RT-PCR analysis shows that this is accompanied by 3.0 Ϯ 0.4-fold increase in long MLCK transcription ( Fig. 1B; p Ͻ 0.05).
To determine whether these TNF-induced increases in MLCK transcription and expression also occur in vivo, mice were injected intraperitoneally with recombinant murine TNF. Jejunal epithelial cells were isolated 4 h after TNF injection and analyzed by SDS-PAGE immunoblot and quantitative real time RT-PCR. MLCK protein content of jejunal epithelial cells increased 4.2 Ϯ 0.5-fold over this period (Fig. 1C, p Ͻ 0.01) and MLCK mRNA transcript number increased 3.4 Ϯ 0.8-fold over the same period ( Fig. 1D; p Ͻ 0.01). Thus, TNF activates long MLCK transcription both in vivo and in vitro. Given that increased MLCK expression is associated with barrier dysfunction (10) and has been observed in patients with inflammatory bowel disease (17), we sought to identify the regulatory elements that control long MLCK expression.
Identification of Multiple Long MLCK 5Ј-Untranslated Regions and Alternative Transcriptional Start Sites-The data above show that long MLCK expression is transcriptionally upregulated by TNF. We therefore aimed to define the long MLCK promoter. Although the position of the MLCK transcriptional start site has been proposed based on models of gene structure (22), newly identified cDNA sequences (e.g. BC064695) that extend beyond the proposed transcriptional start site have raised uncertainty regarding the accuracy of this identification. We therefore performed 5Ј-RACE on mRNA isolated from Caco-2 cells using a long MLCK 3Ј-PCR primer within the coding region of exon 3. Two reaction products were identified and sequenced (DQ642691 and DQ642692). BLAST searches against the human genome localized both of these sequences to the long MLCK (MYLK) gene on chromosome 3q21. Neither has been previously reported among human cDNA sequences. Therefore, these sequences each represent a new transcriptional start site and corresponding first exons, exon 1A and 1B, for long MLCK ( Fig. 2A). Notably, the translational start site and coding sequences, located within exon 2, were identical for the two transcripts. The portion of exon 3 within each sequence was also identical. Thus exons 1A and 1B represent alternative transcriptional start sites for long MLCK in Caco-2 intestinal epithelial cells.
In Silico Analysis of Putative Long MLCK Promoters-The data above show that long MLCK expression is transcriptionally up-regulated by TNF in both human-derived Caco-2 intestinal epithelial cells and murine intestinal epithelial cells. We took advantage of this to preliminarily characterize sequence upstream of the two 5Ј-untranslated regions we identified for long MLCK expressed in Caco-2 intestinal epithelial cells. Because TNF-induced transcriptional activation is largely mediated by AP-1 and NFB (25), we compared potential binding sites for these transcription factors in sequences upstream of human exons 1A and 1B and the predicted mouse transcriptional start site (Fig. 2B). All three sequences contained numerous AP-1 and NFB sites, but only the sequence upstream of exon 1A contained NFB sites within 2 kb of the transcriptional start site. Given that both NFB-dependent and -independent long MLCK up-regulation has been reported after TNF treatment of Caco-2 intestinal epithelial cells (10,11), we therefore focused on the sequence upstream of exon 1A. We restricted our analysis to 4 kb, because all predicted AP-1 and NFB sites were within that region (Fig. 2B).
Identification of Elements Necessary for Transcription from the Human Long MLCK Promoter-The sequence upstream of exon 1A was cloned into a luciferase reporter vector and included four putative AP-1 sites 1115-, 3433-, 3713-, and 3771-bp upstream of the transcriptional start site and two putative NFB sites 1415-bp and 1584-bp upstream of the transcrip- tional start site (Fig. 3A). To assess the potential requirement for other regulatory sequences in unstimulated long MLCK transcription, a series of truncated promoter constructs was analyzed (Fig. 3B). Transcription from the 4-kb construct, the MLCK(ϩ145) construct with a 145 bp 3Ј extension, and a 2-kb construct were all similar. In contrast, both 500-and 120-bp constructs demonstrated transcriptional activity more than twice as great as the 4-kb construct (Fig. 3B). The 1-kb construct supported transcription at a level intermediate between 2-and 500-bp constructs. These data show that the 120-bp construct contains all of the elements necessary for unstimulated long MLCK transcription and further suggests that the region between 500 bp and 2 kb suppresses transcription.
To further define the sites of this suppressor activity, a series of deletion mutants were analyzed. The suppressor activity was present in a 4-kb promoter with the Ϫ2416to Ϫ1588-bp region deleted, but was partly lost when this deletion was extended to Ϫ863 bp. The activity of a ⌬(Ϫ1715 to Ϫ618)-bp construct was similar to the ⌬(Ϫ2416 to Ϫ863)-bp construct. Together with the intermediate activity of the 1-kb promoter, these data suggest that sites between Ϫ1588 and Ϫ433 bp contribute to the observed suppressor activity.
Further analysis of deletion mutants lacking varying lengths of sequence between Ϫ2416 and Ϫ177 bp demonstrate that these sequences are necessary for unstimulated long MLCK transcription. Thus, together with the 120-bp construct, which extends from Ϫ278 to Ϫ155 bp, these data show that the region from Ϫ278 to Ϫ177 bp relative to the transcriptional start site is essential for long MLCK transcription.
Human Long MLCK Transcription Requires Two Sp1 Sites-To evaluate the role of the Ϫ278 to Ϫ177-bp region, this sequence was evaluated in silico (Fig. 4A). Although no TATA box was identified, two putative Sp1 sites were present 275-and 218-bp upstream of the transcriptional start site. Site-directed mutagenesis of either site-reduced transcription to only 6.0% Ϯ 0.4% and 7.6% Ϯ 0.5% of the wild-type 4-kb construct, a level similar to that of the promoterless pGL3 vector (Fig. 4B). To determine if Sp1 binds to this DNA region in vivo, ChIP analysis was used. Sp1 was immunoprecipitated from Caco-2 cell nuclear extracts and the associated DNA assessed by PCR amplification. The DNA sequence including the two Sp1 sites, Ϫ433 to Ϫ201 bp, was specifically immunoprecipitated with anti-Sp1, but not control, antisera (Fig. 4C). This suggests that these sequences are bona fide Sp1 binding sites. Together with the mutagenesis data, these results show that DNA sequence 5Ј to the long MLCK transcriptional start site binds to Sp1 and that these Sp1 sites are required for long MLCK transcription.
The 4-kb MLCK Promoter Is Responsive to TNF-Although the data above suggest that only the first 278-bp upstream of the transcriptional start site are required for long MLCK transcription, the preliminary in silico analyses suggested that relevant regulatory sites may be present as much as 4-kb upstream (Fig. 3A). These include NFB and AP-1 sites that might be activated by TNF. To determine the functional domains necessary for TNF-dependent MLCK transcriptional up-regulation, we first asked whether the 4-kb putative MLCK promoter was sufficient to confer TNF responsiveness. TNF alone induced a modest 0.6 Ϯ 0.05-fold increase in transcription (Fig. 5). This is consistent with our previous demonstration that TNF alone caused only small statistically insignificant increases in MLCK protein expression (10). IFN-␥ alone also failed to activate transcription from the 4-kb long MLCK pro-   Fig. 2A). B, luciferase expression from a series of 4-kb long MLCK constructs shows that the region from Ϫ278 to Ϫ177 bp relative to the transcriptional start site contains the minimal elements necessary for transcriptional activity. Data shown are mean Ϯ S.E. of triplicate samples, normalized to pGL3 (dashed line). The activity of the 4-kb promoter is indicated for reference (solid line). Data are representative of three independent experiments. moter (Fig. 5). However, when monolayers were treated with IFN-␥ and TNF, transcription increased by 1.8 Ϯ 0.2-fold (Fig.  5). The data indicate that the 4-kb MLCK promoter is responsive to cytokine treatment in a manner that parallels endogenous MLCK expression. This suggests that the 4-kb sequence cloned contains the functional elements necessary for regulated MLCK expression.
The 4-kb MLCK Promoter Is Activated by AP-1 and NFB-Major transcription factor signaling pathways used by TNF to induce gene expression include AP-1 and NFB (25). To determine if either of these transcription factors is able to activate transcription from the 4-kb MLCK promoter, cells were cotransfected with the promoter construct and vectors that increase AP-1-or NFB-dependent transcription. AP-1dependent transcription was activated using a constitutively active JNKK2-JNK1 fusion construct (23). This increased luciferase expression from an AP-1 consensus promoter by 2.6 Ϯ 0.3-fold (Fig. 6A, p Ͻ 0.01). The JNKK2-JNK1 fusion construct has no significant activating effect on expression from a pro-moterless luciferase construct or from a luciferase reporter vector with a consensus NFB promoter (Fig. 6A). When MLCK promoters were analyzed, the JNKK2-JNK1 fusion construct did not augment expression from either a 1-or a 2-kb MLCK promoter (Fig. 6B), but did increase expression from the 4-kb MLCK promoter reporter by 1.5 Ϯ 0.1-fold ( Fig. 6B; p Ͻ 0.01). Thus, the 4-kb MLCK promoter is responsive to AP-1 and critical regulatory sequences are present between 2and 4-kb upstream of the transcriptional start site.
NFB-dependent MLCK transcription was analyzed similarly using an NFB p50 -65 fusion protein construct (24). Transfection of Caco-2 cells with this p50 -65 construct increased luciferase expression from an NFB consensus promoter by 2.5 Ϯ 0.2-fold (Fig. 7A, p Ͻ 0.01). The p50 -65 fusion construct had no effect on expression from a promoterless luciferase construct, but markedly suppressed expression from a luciferase reporter vector with a consensus AP-1 promoter (Fig. 7A, p Ͻ 0.01), consistent with the emerging model of NFB-AP-1 suppressive cross-talk (26). When MLCK promoters were analyzed, the p50 -65 fusion construct did not augment expression from a 1-kb MLCK promoter, but did increase transcription from the 2-and 4-kb MLCK promoters by 4.1 Ϯ 0.1-fold and 2.0 Ϯ 0.2-fold, respectively ( Fig. 7B; p Ͻ 0.01). Thus, the 4-kb MLCK promoter is responsive to NFB. Moreover, these data indicate that the critical regulatory sequences necessary for the NFB response are present between 1-and 2-kb upstream of the transcriptional start site.
Characterization of AP-1 and NFB Binding Sites within the 4-kb MLCK Promoter-The data above show that both AP-1 and NFB are able to increase transcription from the 4-kb MLCK promoter. Together with the in silico analyses showing that the 4-kb sequence contains four potential AP-1 sites and two potential NFB sites (Fig. 3A), these data suggest that one or more of these sites may be involved in MLCK transcriptional regulation. To assess the functionality of these potential transcription factor binding sites, we first used electrophoretic mobility shift assay (EMSA) to determine if the sites were able to bind nuclear protein. Nuclear extracts caused a probe containing the AP-1 (1) site at Ϫ1115 to shift (Fig. 8A). This shift was specifically and efficiently competed by unlabeled AP-1 (1) oligonucleotide or an AP-1 consensus oligonucleotide. Simi- A, sequence analysis of the 321-bp upstream of the translational start site identifies two putative Sp1 sites but no TATA box. The indicated intronic sequence includes exon 1B, which is not used in this transcript. B, although the 4-kb long MLCK promoter was sufficient to drive expression of luciferase from a reporter construct, site-directed mutagenesis of either Sp1 site reduced transcription to levels similar to a promoterless control vector (pGL3). Data shown are mean Ϯ S.E. of quadruplicate samples, normalized to the 4-kb long MLCK construct, and are representative of two similar experiments. C, chromatin from Caco-2 monolayers was immunoprecipitated using antibodies against Sp1 protein. PCR analysis using primers surrounding the Sp1 sites shows that this DNA sequence (Ϫ433 to Ϫ201 bp) is specifically immunoprecipitated, indicating that Sp1 binds to sites within this sequence.
larly, probes including the AP-1 (2), AP-1 (3), and AP-1 (4) sites, at Ϫ3433, Ϫ3713, and Ϫ3771 bp, respectively, were shifted in EMSA analysis and these shifts could be competed with both site-specific and AP-1 consensus oligonucleotides. These data suggest each of these putative AP-1 binding sites binds to AP-1 in vitro.
EMSA analysis of the two putative NFB sites, Ϫ1415 and Ϫ1584 bp, identified within the MLCK promoter also demonstrated shifts when co-incubated with nuclear extracts (Fig.  8C). These bands could be specifically and efficiently competed using either site-specific cold oligonucleotides or NFB consensus oligonucleotides, suggesting that NFB binds to each of these sites in vitro. To assess the ability of NFB to interact with sites in Caco-2 cells, ChIP analysis was performed using NFB immunoprecipitates, but NFB(1/2) DNA was not recovered (Fig. 8D). This suggests that TNF does not induce NFB binding to these sites within the MLCK promoter in Caco-2 cells.

Site-directed Mutagenesis of Specific Transcription Factor Binding Sites Ablates Transcriptional Activation-Together
with the responsiveness of the 4-kb promoter to AP-1 and NFB activation by JNKK2-JNK1 and p50 -65 fusion proteins, respectively, the EMSA and ChIP data suggest that the transcription factor binding sites identified may be involved in regulation of MLCK transcription. To determine the specific roles of these sites in MLCK transcriptional regulation each site was disrupted by site-directed mutagenesis. Disruption of the AP-1 (1) site resulted in only mild reduction of luciferase expression (Fig. 9A). In contrast, mutagenesis of AP-1 (2), AP-1 (3), or AP-1 (4) sites markedly reduced the ability of JNKK2-JNK1 fusion protein to increase transcription (Fig. 9A). These data indicate that AP-1 sites 2, 3, and 4 are each required for transcriptional regulation of MLCK by AP-1. In contrast, the AP-1 (1) site appears to be at least partially dispensable.  Similar site-directed mutagenesis of the NFB (1) and NFB (2) sites showed that mutation of either site completely blocked transcription by the p50 -65 fusion protein (Fig. 9B). This suggests that both NFB binding sites identified are required for MLCK transcriptional activation by NFB.

TNF-induced MLCK Promoter Activation Only Occurs in Well Differentiated Intestinal Epithelial Monolayers-
The transcription factors involved in TNF-induced MLCK up-regulation have been a subject of controversy, with conflicting results in separate studies using different Caco-2 cell clones grown under different conditions (10,11). In comparing these studies one obvious difference is the differentiation state of the Caco-2 monolayers used. We used a well differentiated Caco-2 BBE subclone (19) grown for at least 2 weeks after confluence to allow for complete epithelial differentiation (10,(27)(28)(29). In contrast, others used the less well differentiated Caco-2 parental line (11). We therefore considered the possibility that differences in results obtained could be caused by differences in epithelial differentiation state.
To model the process of intestinal epithelial differentiation, we studied Caco-2 BBE monolayers 3 days after reaching confluence, as well as at early, 7 days, and late, 14 days, in the differentiation process. This is a well established model of intestinal epithelial differentiation that is thought to recapitulate the process that occurs in vivo as cells migrate from crypt to villus (8, 28 -33). The data show that TNF did not significantly upregulate transcription from the full-length 4-kb MLCK promoter in undifferentiated monolayers studied after 3 or 7 days (Fig. 10A). In contrast, at 14 days, when monolayers were well differentiated, the 4-kb MLCK promoter was fully responsive to TNF after priming with IFN-␥ (Fig. 10A). This suggests that the signaling pathways activated by TNF must differ between undifferentiated and well differentiated intestinal epithelial cells.
To determine if activation of NFBor AP-1-mediated transcription by TNF was modulated during intestinal epithelial differentiation, monolayers transfected with the consensus NFB or AP-1 reporter constructs were studied 3 days, 7 days, and 14 days after confluence. Remarkably, TNF-induced NFB  activation diminished during the process of epithelial differentiation (Fig. 10B). Conversely, TNF-induced AP-1 activation increased during epithelial differentiation (Fig. 10B). These data suggest that signaling pathways activated by TNF are modulated by epithelial differentiation. Moreover, the data may explain differences in the reported roles of specific transcription factor pathways in MLCK up-regulation reported by various laboratories (10,11).

DISCUSSION
MLCK-dependent actomyosin regulation is central to control of many essential cellular processes, including contraction, migration, and epithelial barrier function (27, 34 -39). In many nonmuscle cells, long and short MLCK are co-expressed from two different promoters within a single gene (2,3). When both MLCK proteins are expressed in a single cell, short MLCK tends to associate with myosin IIA, but not F-actin, while long MLCK associates with dense F-actin bundles, but not myosin IIA (4). This suggests that, even when expressed in a single cell, long and short MLCK proteins serve unique functions.
Some cell types primarily express only long or short MLCK. For example, we have recently shown that long MLCK is the principal MLCK expressed in intestinal epithelial cells (8). Moreover, we have shown that long MLCK is associated with the perijunctional ring of actomyosin and that the activity of long MLCK can regulate tight junction permeability, i.e. epithelial barrier function (8,27). Similar roles for MLCK have been reported in other epithelial tissues and vascular endothelia (40 -42). We have recently shown that MLCK activation is necessary for the loss of barrier function induced by TNF, both in vitro and in vivo (12,16). This increased MLCK activity is associated with increased MLCK expression (10,11), and similar increases in MLCK expression and activity are present in intestinal epithelia of inflammatory bowel disease patients (17). Thus, it is likely that inducible long MLCK expression plays an important role in disease pathogenesis. It is, however, impossible to assess the potential therapeutic efficacy of pharmacological agents that modify gene transcription without detailed  understanding of the mechanisms that regulate long MLCK expression.
Our data represent the first comprehensive analysis of the elements necessary for human long MLCK transcription. Thus, while we used cultured human intestinal epithelial cells as our model system, the data are likely to have significant implications in many other cell types. Additionally, our data provide new insights into the structure of the long MLCK, or MYLK, gene. We performed the first 5Ј-RACE analysis to identify the 5Ј-untranslated region of long MLCK. In doing so we discovered two previously unrecognized exons that are alternatively used as the transcriptional start site. This observation has important implications, as 31 kb of intronic sequence upstream of exon 1B may represent a unique promoter for that transcriptional start site.
Because of the location of potential binding sites for AP-1 and NFB, we focused our attention on transcription from exon 1A. We therefore cloned the 4-kb region and used a standard reporter assay to identify the minimal sequence necessary for long MLCK transcription. The data show that this information is contained within the first 278 bp 5Ј to the transcriptional start site and includes 2 Sp1 sites but no TATA box. These analyses also suggest that transcriptional repressors are located within the region from Ϫ1588 to Ϫ618 relative to the long MLCK transcriptional start site. We then determined that the fulllength 4-kb sequence was able to respond to TNF in a manner similar to the endogenous promoter.
To identify the mechanism by which the 4-kb sequence activated transcription in response to TNF we first asked whether the long MLCK promoter was responsive to NFB or AP-1 activation. These transcription factor signaling pathways are both activated by TNF and, in the case of TNFdependent intestinal epithelial long MLCK protein up-regulation, the role of NFB has been the subject of controversy (10,11). Our previous work has shown that TNF-dependent intestinal epithelial long MLCK protein up-regulation does not require NFB activation (10). This is consistent with the absence of in vivo NFB binding to MLCK promoter sequences, as determined by ChIP analyses. Although Ma et al. (11) also reported that TNF could induce increases in Caco-2 MLCK expression, they found that this up-regulation was blocked by NFB inhibition. Ma et al. recently reported that the a single NFB binding site mediates a 50% increase in transcription in response to TNF (43). We have located this site ϳ150bp upstream of exon 1A, downstream of the 3Ј Sp1 site. While this NFB site was scored as a low probability NFB binding site by the algorithm we used for our initial in silico analysis, we have noted that the region including this site does augment transcription induced by p50 -65 by 0.2 Ϯ 0.05-fold relative to a deletion construct lacking this region ( p ϭ 0.01). 4 Thus, whereas our data are consistent with a minor role for this site, it is unlikely to be the major upstream element critically involved in long MLCK transcriptional regulation from exon 1A.
We used a combination of EMSA, ChIP, and site-directed mutagenesis approaches to identify 2 NFB and 3 AP-1 sites that were required for NFB-and AP-1-mediated long MLCK transcriptional activation, respectively. We found that transcription from the long MLCK promoter could be activated by NFB but was also activated by AP-1. However, TNF was only able to activate transcription from the long MLCK promoter in well differentiated intestinal epithelial monolayers. We also found that an NFB consensus promoter reporter was better activated by TNF in undifferentiated monolayers, while an AP-1 consensus promoter reporter was better activated by TNF in well differentiated monolayers. This result provides a potential explanation for divergent results, as it may be that NFB mediates long MLCK transcriptional activation in undifferentiated cells while AP-1 is more important in differentiated cells.
That is, the differentiation state of the intestinal epithelial monolayers studied may modulate the response to TNF in terms of the transcription factor pathways activated. These data prompt one to ask why the intestinal epithelium would use two different transcription factor pathways to increase long MLCK expression. What is so different about the undifferentiated and well differentiated Caco-2 monolayers? To understand this, it is important to recognize that undifferentiated monolayers are most like intestinal crypt epithelium while well differentiated monolayers model villus epithelium (28 -33). Undifferentiated crypt epithelium is specialized for proliferation and ion secretion. Thus, one might hypothesize that NFB activation in response to TNF would be important, as NFB can activate anti-apoptotic cytoprotective events, in part through down-regulating JNK activation (26), and crypt cell survival is critical to intestinal homeostasis. In contrast, TNF-induced JNK activation can promote apoptosis (26). Such apoptosis may in fact be a physiologically appropriate response in the face of intestinal injury. For example, in mice with an intestinal epithelial cell-specific knock-out leading to an inability to activate NFB, intestinal ischemia-reperfusion injury results in increased villus cell loss compared with wild-type mice (44). Paradoxically, this is beneficial, as the knock-out animals with increased enterocyte damage are protected from the multi-organ failure that follows intestinal ischemia-reperfusion injury (44). Alternatively, the observations that the intestinespecific Cdx-2 homeobox gene is necessary for villus cell differentiation (45,46) and that TNF down-regulates Cdx-2 via an NFB-dependent pathway (47) might explain the reduced TNFdependent NFB activation in well differentiated villus-like cell. Whatever the explanation, the data do show that intestinal epithelial transcription factor responses to TNF are modulated by cell differentiation state.
In summary, our data identify two unique 5Ј-untranslated regions corresponding to two novel alternatively utilized exons and transcriptional start sites for human long MLCK. We are also the first to comprehensively characterize the mammalian promoter for long MLCK transcribed from exon 1A. This is of biological and, potentially, therapeutic importance, given the critical and diverse roles of long MLCK in disease. Moreover, the data suggest that the differentiation state of intestinal epithelial, and perhaps other, cells can profoundly influence cytokine-induced transcriptional activation.