The tissue-dependent keratin 19 gene transcription is regulated by GKLF/KLF4 and Sp1.

Keratins play critical roles in cellular differentiation and cytoskeletal organization. Keratin 19 (K19) is unique because it has been implicated as a marker of stem cells in some tissues, such as the hair follicle in the skin. It is also associated with malignant transformation in esophageal and pancreatic cancers. Here, we show that the K19 promoter is active in a subset of gastrointestinal cancer cells derived from esophageal and pancreas but inactive in other contexts. This activity was mapped to a short region containing an overlapping binding site for gut-enriched Krüppel-like factor (GKLF/KLF4) and Sp1. GKLF has a higher binding affinity and is the predominant binding factor in cells with low Sp-1 protein levels. Pancreatic acinar cells normally do not express K19, but overexpression of GKLF and Sp1 in these cells leads to aberrant expression, similar to what is observed in pancreatic cancer. These results demonstrate the functional interaction of ubiquitous and tissue-restricted transcription factors in determining tissue- and neoplasm-specific patterns of gene expression.


Introduction
Cytokeratins belong to the family of intermediate filaments (IFs) and are associated with cellular differentiation and cytoskeleton organization. Keratin proteins can be further subdivided in two different types, the acidic type I (K9-20) and basic type II (K1-8) keratins (1)(2)(3). There is at least one pair of acidic and basic keratins expressed as heterodimers in any given epithelial tissue throughout development and differentiation. For example, in stratified epithelia, basal cells are characterized by the expression of K5 with K14/K15. Upon differentiation, stratified keratinizing epithelia express K1 and K10, whereas non-keratinizing stratified epithelia (e.g. the human esophagus) express K4 and K13. Simple epithelia express K8 and K18 (4). Regulation of keratin expression in the differentiating stratified squamous epithelium is orchestrated by a complex interplay of both ubiquitous and tissue-specific transcription factors (5,6).
Analysis of the molecular mechanisms underlying provides insights into the normal program of differentiation as well dedifferentiation during malignant transformation.
Moreover, possible progenitor stem cells can be potentially identified or defined by studying keratin gene expression (7) as exemplified by the hair follicle.
The most striking exception to the keratin-pair rule is the smallest known acidic keratin, namely keratin 19 (K19). K19 has no known basic type II keratin partner, although its expression is often found in cells which express K8 (2,8). K19 has the highly conserved alpha-helical central domain which is essential for filament formation.
However, this keratin lacks the C-terminal nonhelical "tail domain" present in other acidic keratins. K19 is fully competent in filament formation with a basic keratin partner in vitro, although these complexes are unstable (9). Indeed, in some tissues, K19 can substitute for K18, the usual partner of K8 (10). In K18 knockout mice, K19 compensates for the loss of K18 expression (11). Stasiak et al. (1989) proposed that K19 might act as "neutral" keratin in terms of differentiation. By dimerization with any early-synthesized basic type II keratin and with resulting delayed expression of the usual type I keratin partner, a cell could be held in a "flexible state of differentiation" (12).
The human K19 gene was sequenced by Bader et al. (1986) (13) and Stasiak and Lane (1987) (14) and the gene localizes to chromosome 17q21-q22 (15,16). The murine K19 gene shows high homology to its human counterpart (17) and the gene is located in the acidic keratin cluster on mouse chromosome 11 (17,18). Precursor cells in different tissues display high K19 levels and upon differentiation, K19 expression is epithelial cell-specific. For example, in the developing pancreas, duct-like precursor cells harbor high K19 expression. After these precursor cells transdifferentiate into endo-and exocrine compartments, K19 expression resides predominantly in ductal epithelial cells, whereas islet cells show markedly decreased K19 expression and acinar cells are K19 negative (19). Likewise, the fetal liver shows high K19 expression. Upon differentiation, hepatocytes lose expression of K19, whereas its expression is retained in bile ducts (20). Sites of K19 expression in the human adult include the esophagus, stomach, pancreas, small intestine and colon. In the human skin, K19 expression is limited to a small portion of cells in the outer root sheath of the hair follicle, but no expression is evident in the interfollicular suprabasal epidermis (12). These K19 positive cells in the hair follicle demonstrate features of slow-cycling cells and may represent stem cells (21).
Many premalignant and malignant tissues display K19 expression, such as dysplasia and carcinoma of squamous epithelia, adenocarcinoma of the lung, breast, pancreas, stomach and colon (22).
Little is known about the regulation of K19 expression. We hypothesized that the transcriptional regulation of K19 is important in understanding differentiation programs in different cell types, namely squamous epithelial and pancreatic ductal cells, and that differences may arise from the manner in which K19 is regulated by different transcriptional factors. In this study we describe the tissue specific activity and regulation The PCR products for each mutant in this first amplification step were gel purified (Qiagen) and combined in a subsequent "fusion" PCR reaction. The overlapping ends of these fragments anneal and serve as template and as primer for the second amplification reaction. Using the outer K19-1970 5' primer and the K19+46 3' primer, the different mutated full-length K19 promoter fragments were generated. PCR reaction components and amplification conditions for first and second steps were the same as by guest on July 23, 2018 http://www.jbc.org/ Downloaded from described above. The mutated full-length K19 promoter fragments were then subcloned into the pGL3 basic vector as outlined above. All constructs were subjected to confirmatory sequencing.

GKLF and Sp1 expression vectors
Construction of the expression vectors containing human GKLF and Sp1 cDNAs was described previously. Briefly, the cDNA of human GKLF was subcloned into pcDNA Horseradish peroxidase activity was detected with a chemiluminescence system (ECL system, Amersham Pharmacia Biotech).

Electrophoretic Mobility Shift Assays (EMSAs)
Nuclear extracts from different cell lines were prepared as described by Schreiber loaded on a 6% polyacrylamide, 0.25x Tris borate gel and electrophoresed at 10 V/cm for 2 h. The gels were dried and exposed to x-ray film (Kodak) at 80 °C for 1 to 12 hrs.
For competition experiments, the nuclear extract was co-incubated with 100-fold excess of unlabeled double-stranded oligonucleotides. Sequences of wild-type and mutated oligonucleotides are summarized in Table 2. Immune supershift assays were performed using the Sp1 and GKLF antibodies. The antibody was preincubated with the nuclear extract at room temperature for 15 min. prior to the addition of the [α-32 P]-labeled oligonucleotide DNA probe. Generation and purification of recombinant GKLF protein was described previously (24). Recombinant Sp1 protein was obtained from Promega.
Other conditions for EMSAs are described above.

Binding of Sp1 and GKLF to the -67 to -56 element
To characterize DNA-protein interactions between transcriptional factors and cisregulatory elements that reside within the K19 promoter sequence at -62 bp, we performed electromobility shift assays (EMSAs) using nuclear extracts from TE-12 and Panc-1 cells. A radioactive labeled double-stranded oligonucleotide probe spanning from -81 to -48 bp was incubated with 5 µg of nuclear extracts (Figures 4 A and B).
Sequences of wild-type and mutated double stranded oligonucleotides are indicated in Table 2. Computer-based analysis revealed that this region contains putative binding sites not only for Sp1, but also for GKLF/KLF4 which is important in epithelial cell differentiation and also GATA of which certain isoforms are expressed in the gastrointestinal tract.
Using TE-12 nuclear extracts, we detected two specific bands (designated as I, II) binding to the wild-type oligonucleotide probe (Figure 4  However, none of the bands observed were super-shifted by a Sp3 antibody (data not shown). Of note, there was a nonspecific band that migrated between bands I and II in both TE-12 and Panc-1.
To further elucidate which nucleotides in the -81 to -48 bp element bind Sp1 and GKLF, we performed EMSAs using different unlabeled mutated oligonucleotides as competitors ( Figure 4B)

Recombinant protein assays show predominant binding of GKLF
To independently confirm the results from the EMSAs with nuclear extracts, we performed EMSAs with purified recombinant proteins for GKLF (24) and Sp1 (Promega).
Purified human GKLF protein was obtained from cells transfected with a HIS-tagged GKLF expression vector and purified over a Nickel column as previously described (24).
The purified protein was separated on a SDS-PAGE and probed with the GKLF antiserum. A band of the expected size was detected and the protein was not degraded.
(data not shown).  suggesting that GKLF likely has higher binding activity to this sequence than Sp1. In addition, we performed EMSAs using nuclear extracts from TE-12 cells ( Figure 5C).
After addition of 150 ng purified recombinant GKLF protein into the assay, we observed complete loss of Sp1 binding (band I; lane 2). Titration of recombinant Sp1 protein (300 and 600 ng) to the nuclear extracts did not abolish GKLF binding (band II; lane 3 and 4).
The same titration experiments were performed using nuclear extracts from Panc-1 cells with GKLF as the predominant DNA binding protein to the wild-type sequence (data not shown).

Disruption of the Sp1 and GKLF binding site leads to loss of K19 promoter activity
The EMSAs indicate that four guanosines within the GGGCGGGGAAGT element of the K19 promoter are especially important for binding of Sp1 and GKLF. We next elucidated the functional consequences on K19 promoter activity with the introduction of mutations in these nucleotides. In addition to the Sp1-62 mutant already mentioned, we generated two GKLF-58 mutants (MT1 and MT2) by a PCR-based site-directed mutagenesis approach in the context of the full length (-1970+46) K19 promoter.
Promoter activity was determined after transient transfection of TE-12 ( Figure 6) and Panc-1 cells and compared to the wild-type full-length K19 promoter . resulted in a significant loss of K19 promoter activity in TE-12 ( Figure 6). Mutation of the adjacent two guanosines (GGGCGGGGAAGT to GGGCGGTTAAGT, designated as GKLF MT1), which resulted in disruption of both Sp1 and GKLF binding, lead to dramatic loss of promoter activity in TE-12 ( Figure 6). Of note, nearly identical results were observed in Panc-1 cells (data not shown). GKLF MT2 (GGGCGGGGAAGT to GGGCGGGGGGGT), which preserves the DNA binding of Sp1 and GKLF, accordingly did not alter activity of the full-length K19 promoter ( Figure 6).

Endogenous Sp1 and GKLF level correlate with K19 promoter activity
Based upon the results in the EMSAs and protein competition experiments, we next tested the hypothesis that endogenous Sp1 and GKLF protein levels influence K19 promoter activity. While Sp1 was variably expressed in the cell lines, its level is reduced in Panc-1 cells (Figure 7). This is supportive of the notion that we did not observe Sp1 binding in EMSAs with Panc-1 nuclear extracts. GKLF expression also varied but is found to be at high levels in both esophageal squamous and pancreatic ductal cancer cells ( Figure 7). However, GKLF expression was not detected in pancreatic acinar cancer cells (AR42J, ARIP). Thus, it appears that cells with low expression levels of both Sp1 and GKLF have little, if any, K19 promoter activity and K19 protein expression as shown in figure 1.

Overexpression of Sp1 and GKLF restores K19 promoter activity in cells lacking K19 promoter activity
If endogenous expression levels of Sp1 and GKLF contribute differentially to K19 promoter activity and expression of its gene product in different cell types, it is reasonable to postulate that overexpression of these transcription factors would lead to increase of K19 promoter activity. We therefore cotransfected different cell lines with the promoter activity. This supports the notion that Sp1 and GKLF both bind and transactivate the K19 promoter through an additive mechanism.

Analysis of the transcriptional control of keratin genes provides important insights
into cellular differentiation and malignant transformation. K19 is expressed during early embryonic development (33,34) and has a restricted epithelial pattern in the adult (22).
K19 expression has been implicated as a marker for stem cells in the skin (21)  Our biochemical and functional studies suggest that GKLF contributes to specific activation of the K19 promoter in pancreatic ductal epithelial cells, but not pancreatic acinar cells. This may in part be attributable to differences in GKLF protein levels with abundant expression in pancreatic ductal epithelial cells but virtually none in acinar cells, but perhaps even more so, promotes the notion that K19 indeed might be a marker for pancreatic ductal cells which to date has escaped such a specific marker. Furthermore, GKLF expression correlates with K19 promoter activity, but appears to be only minimally dependent upon Sp1 in these cell types, whereas both are required in esophageal squamous epithelial cells. Accordingly, when Sp1 and GKLF are overexpressed in cell lines with low or absent K19 promoter activity, promoter activation can be restored.
The Sp1/KLF family comprises more than 16 different mammalian members of zinc finger transcription factors that bind GC/GT rich DNA elements (37,38). These proposed by several studies (48)(49)(50)(51). GKLF is found in epithelial cells of the gastrointestinal tract in the intestine and esophagus (50), pancreas (Brembeck and Rustgi, unpublished observations), lung and skin (51). Induction of GKLF mRNA levels is observed in the mouse embryo at day 15 (51) and is important for intestinal development (52). Additionally, GKLF is down regulated in late stages of tumorigenesis of the min mouse model (52). Expression of GKLF is high in growth-arrested fibroblasts and almost absent in exponentially growing (50). In addition, GKLF expression is found in multilayered epithelia, where there switch from proliferation to differentiation (50,51).
GKLF binds DNA through a core DNA motif of 5'-(G/A)(G/A)GG(C/T)G(C/T)-3' (53). Only few gene targets for GKLF transactivation have been previously characterized. Our own work identified the esophageal squamous epithelium specific K4 and Epstein-Barr virus ED-L2 promoters as targets for GKLF (24). Both promoters harbor a CACCC-like motif, which has been described as the core motif for EKLF (48).
In the K4 and ED-L2 promoter, this motif is recognized by GKLF resulting in activation of transcription (24). Other studies described the GKLF core motif in the basic transcription element of the cytochrome P-450IA1 promoter CYP1A1. Both Sp1 and GKLF bind and recognize this element. However, GKLF inhibits Sp1-mediated activation of the CYP1A1 promoter by competing for this binding site and by protein-protein interactions with Sp1  Total cell lysates were separated by SDS page and protein expression determined using a specific K19 antibody.        Tables   Table 1: Sequences of K19 promoter specific primers. Nucleotides in bold represent the mutated nucleotides (MT) compared to the wild-type sequence (WT).
Note that oligonucleotides for the K19-Sp1-62 MT, K19-GKLF-58 MT1 and MT2 have been used for EMSAs and for PCR-generation of the respective luciferase reporter gene constructs (see Methods).

Sequencing primers
K19 INT1   Sequences of wild-type (WT) and mutated (MT) oligonucleotides corresponding to position -81 to -48 in the K19 promoter. The potential binding sites for GATA, Sp1 and GKLF are indicated.