A Conserved Tissue-Specific Homeodomain-Less Isoform of MEIS1 Is Downregulated in Colorectal Cancer

Colorectal cancer is one of the most common cancers in developed nations and is the result of both environmental and genetic factors. Many of the genetic lesions observed in colorectal cancer alter expression of homeobox genes, which encode homeodomain transcription factors. The MEIS1 homeobox gene is known to be involved in several hematological malignancies and solid tumors and recent evidence suggests that expression of the MEIS1 transcript is altered in colorectal cancer. Despite this potential connection, little is known about the role of the gene in the intestines. We probed murine gastrointestinal tissue samples with an N-terminal Meis1 antibody, revealing expression of two previously described isoforms, as well as two novel Meis1 products. A 32 kD Meis1 product was expressed in the nuclei of non-epithelial cells in the stomach and colon, while a 27 kD product was expressed in the cytoplasm of epithelial cells in the proximal colon. Our data suggest that the 27 kD and 32 kD Meis1 proteins are both forms of the Meis1d protein, a homeodomain-less isoform whose transcript was previously identified in cDNA screens. Both the MEIS1D transcript and protein were expressed in human colon mucosa. Expression of the MEIS1D protein was downregulated in 83% (10/12) of primary colorectal cancer samples compared to matched normal mucosa, indicating that MEIS1D is a biomarker of colorectal tumorigenesis. The decreased expression of MEIS1D in colon tumors also suggests that this conserved homeodomain-less isoform may act as a tumor suppressor in human colorectal cancer.


Introduction
Colorectal cancer accounts for approximately 10% of both cancer diagnoses and mortalities in the United States (www. cancer.org). The mortality rate has decreased in the previous two decades, primarily reflecting higher levels of compliance with recommended colonoscopy screenings as well as the improved efficacy of therapeutic options [1]; however, colorectal cancer still has the second highest mortality incidence in the United States, trailing only lung cancer. Like many types of cancer, colorectal cancer has both environmental and genetic components [2,3,4]. Both sporadic and hereditary forms of the disease are associated with an accumulation of multiple genetic lesions [5,6], which can result in decreased levels of apoptosis [7], loss of cell cycle regulation [8], and constitutive activation of the WNT signaling pathway [9]. As regulators of downstream transcriptional activation, transcription factors are frequently dysregulated in colorectal cancer [10,11].
Homeobox genes encode proteins with DNA-binding domains known as homeodomains. These homeodomain proteins act as transcription factors, binding promoter regions and activating transcription [12]. HOX gene family members are the prototypical homeobox genes. HOX genes are involved in embryonic segmentation and patterning as well as the development of numerous organ systems, including the gastrointestinal tract [12].
Developmental genes are often ectopically expressed during carcinogenesis [13] and several HOX genes have been linked to colorectal cancer. HOXB6, HOXB8, HOXC8, HOXC9, and HOXD13 are overexpressed in both colorectal cancer cell lines and primary colon tumors [14,15,16]. HOXB13, however, is normally expressed in colonic mucosa but is downregulated in tumor tissue [10]. Deregulation of non-HOX homeobox genes has also been observed in colorectal carcinomas. CDX1 and CDX2 are downregulated during colorectal carcinogenesis [17,18], while aberrant PROX1 expression in the colon results in increased dysplasia [19].
Homeodomain transcription factors often bind promoter regions as either homodimeric or heterodimeric complexes [20]. This dimerization provides increased specificity of transcriptional activation [21]. The MEIS1 homeodomain protein is a known binding partner of several other homeodomain proteins, including HOXA7, HOXA9, and PBX1 [22]. MEIS1 plays a role in the normal development of the hematopoietic lineage and is overexpressed in a subset of acute myeloid leukemias [23]. Increased MEIS1 expression has also been observed in neuroblastomas and the expression level of the MEIS1 transcript is a prognostic indicator in breast cancer [24]. Downregulation of total MEIS1 transcript was observed in colorectal adenomas, suggesting a role for MEIS1 in intestinal tumorigenesis [25].
Due to the links between homeobox genes and colorectal cancer, we examined the status of Meis1 in the colon. In this study, we describe two novel Meis1 products expressed in the murine gastrointestinal tract. These two proteins are expressed in different cell types and subcellular compartments. Both proteins are translated from the Meis1d transcript, a homeodomain-less splice variant of Meis1. MEIS1D, the human homolog of Meis1d, was identified in normal human colon tissue at both the mRNA and protein level. Furthermore, we observe the downregulation of MEIS1D expression in human colorectal cancers. These data suggest that MEIS1D is a novel suppressor of colorectal tumorigenesis and a potential therapeutic target for colon cancer.

Mouse colony
C57BL/6J (B6) mice were obtained from The Jackson Laboratory (Bar Harbor, ME). Mice were maintained in the AALAC-accredited TJU animal facility. This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol, 343C, was approved by the Institutional Animal Care and Use Committee of Thomas Jefferson University, permit number A3085-01.

Retroviral Infection
Hct116 cells were obtained from the American Type Culture Collection (Cat. No. CCL-247), where the identity of the cell line was confirmed by STR analysis. MSCV-Meis1d and MSCV-Neo plasmids were transfected into Phoenix cells using the Profection Calcium Phosphate kit (Promega, Madison, WI). Cells were incubated at 37uC for 24 hours to produce viral media. Viral media was added to Hct116 cells in six well plates and spun at 1,800 rpm for 45 minutes. The cells were then incubated at 32uC for 3 hours, the viral media was removed, and new viral media was added. The cells were spun again at 1,800 rpm for 45 minutes and incubated at 32uC for 3 more hours. Viral media was replaced with DMEM and the cells were moved to 37uC for 48-72 hours to allow translation of the inserted sequence.

Subcellular fractionation
Tissue and cell samples were washed in 16PBS and homogenized in lysis buffer (10 mM Tris-HCl, pH 7.4; 5 mM MgCl 2 ; 1 mM DTT; 1 mM PMSF) with protease inhibitors (Roche Applied Science, Indianapolis, IN). Lysates were then passed through a 25Kg needle and centrifuged at 6006g at 4uC for 10 min. The supernatant was labeled the cytoplasmic fraction. The pellet was resuspended in lysis buffer and labeled the nuclear fraction.

Epithelial cell isolation
Proximal and distal colon samples were cut into 1 mm wide strips. The samples were incubated in 16 PBS with 0.15% DTT for 30 min at room temperature with constant agitation by stir bar to remove mucus. Mucosal strips and stir bar were washed in 16 PBS and then incubated in 16 PBS with 1 mM EDTA, pH 7.2 stirring for 60 min at room temperature to remove epithelial cells. The EDTA solution was collected and centrifuged at 4706g for 5 min. Remaining tissue was incubated in the EDTA solution to remove residual epithelial cells and then homogenized in lysis buffer as described above. The epithelial pellet was resuspended in RPMI 1640 and incubated with collagenase (Sigma-Aldrich, St. Louis, MO) at 50 units/mL for 30 min at 37uC, vortexing gently every five minutes. Samples were centrifuged at 2006g for 5 min. The pellet was resuspended in 16 PBS, centrifuged again, and resuspended in RPMI. The RPMI cell suspension was overlaid on a 50% Percoll/PBS mixture. The Percoll gradient was centrifuged at 4706g for 20 min. Epithelial cells were taken from the top of the gradient, diluted with RPMI, and centrifuged at 8306g for 5 min. The cell pellet was resuspended in RPMI and centrifuged again at 4706g for 5 min. The final cell pellet was homogenized in western blot lysis buffer using a 25Kg needle.

RT-PCR
Tissue samples were homogenized in 1 mL TRI Reagent Solution (Ambion, Inc, Austin, TX) and centrifuged at 12,0006g for 15 min at 4uC. 200 mL chloroform were added. The samples were mixed for 15 sec, incubated at room temperature for 3 min, and centrifuged at 12,0006g for 20 min at 4uC. The top layer was mixed with 500 mL of 2-propanol, incubated at room temperature for 10 min, and centrifuged at 12,0006g for 20 min at 4uC. The RNA pellet was washed with 75% ethanol and centrifuged at 7,5006g for 5 min. The RNA was then air dried and resuspended in DEPC-treated water. SuperScript II Reverse Transcriptase (Invitrogen, Carlsbad, CA) was used to generate cDNA. All Meis1 transcripts were amplified in murine samples, while MEIS1Dspecific primers were used in human samples.

Immunohistochemistry
Tissues were fixed in Formalde-Fresh 10% buffered formalin solution (Fisher Scientific) overnight at 4uC and stored in 70% EtOH. Fixed tissues were embedded in paraffin and tissues were cut and mounted on glass slides by the KCC Translational Research Core Facility. Tissues were deparaffinized using citrate buffer, pH 6.0 (Vector Laboratories, Berlingame, CA) and stained with Meis1-N antibody overnight at 4uC. Anti-rabbit secondary antibody (Vector Laboratories, Burlingame, CA) was used at a 1:200 dilution for 1 hour at 37uC followed by ABC Linker Kit (Vector Laboratories, Berlingame, CA) for 30 min at 37uC. Samples were developed using the DAB Peroxidase Substrate Kit (Vector Laboratories, Burlingame, CA). Staining was visualized on a Nikon Eclipse E600 microscope at either 46 or 206 magnification.

Ethics
The patients were consented using a written surgical consent form which states the waste from surgery would be used for research purposes. All of the samples were obtained at Thomas Jefferson University Hospital. The Thomas Jefferson University Institutional Review Board (IRB) exempted this study from review because they deemed that the study did not constitute human subjects research and waived the need for consent due the fact that the samples received were coded to remove identification.

Results
Two novel Meis1 isoforms are present in the murine gastrointestinal tract Although downregulation of MEIS1 in human colorectal cancer has recently been observed, the role of the gene in gastrointestinal (GI) homeostasis and tumorigenesis is poorly understood. To determine the expression of murine Meis1 isoforms in the GI tract, a western blot was performed on C57BL/6J (B6) GI lysates using a polyclonal antibody targeted to the N-terminus of Meis1 (Meis1-N). Products of 43 kD and 51 kD were observed in most samples (Figure 1a). These molecular weights correspond with the known sizes of two previously described isoforms, Meis1a and Meis1b, respectively. Two additional bands were observed in some of the samples: a ,32 kD protein present in the stomach and colon and a ,27 kD protein expressed in the proximal colon and cecum ( Figure 1a). The ,27 kD product is the same as the predicted weight for Meis1d, another Meis1 isoform. The transcript for Meis1d was previously identified during screens of murine cDNA [26,27]. The isoform lacks all of exon 8, resulting in a premature stop codon in exon 9 and a theoretical truncated protein product (Figure 1c, d). The existence of an in vivo protein product, however, has never been confirmed.

Meis1d, a truncated Meis1 isoform, is expressed in the murine cecum and proximal colon
If the 27 kD Meis1 product is Meis1d, then the Meis1d transcript should be detectable in the proximal colon. To determine if the Meis1d transcript is present, Meis1 splice variants were amplified from B6 proximal colon cDNA samples using RT-PCR. Primers were designed to amplify across exon 8, distinguishing Meis1d from the full length isoforms, Meis1a and Meis1b. The PCR reaction amplified products of approximately 758 and 904 bp in length, the predicted size of products from Meis1d and Meis1a/b transcripts (Figure 1b). The size difference between the two products corresponds with the known length of exon 8 (146 bp), suggesting that the smaller product is lacking that exon. Extraction and sequencing of the smaller PCR product confirmed the absence of exon 8 from the PCR product (data not shown), demonstrating the presence of the Meis1d transcript in the murine proximal colon.
Removal of exon 8 from the Meis1d transcript causes a frameshift in exon 9. The result of this frameshift is the production of truncated protein product with five novel amino acids on the Cterminus (Figure 1d). Since these residues are not present on full length Meis1 isoforms, a custom antibody targeting the unique epitope (Meis1d-C) was generated. To determine if the custom antibody recognizes the ,27 kD Meis1 product, colon lysates from B6 mice were probed with the Meis1d-C and Meis1-N antibodies. Proximal colon and distal colon lysates were used as positive and negative controls, respectively. As expected, the Meis1-N antibody picked up the previously identified ,27 kD band in the proximal colon, but not the distal colon (Figure 2a). The same expression pattern was observed in the samples probed with the Meis1d-C antibody, demonstrating that the Meis1d-C antibody detects the ,27 kD Meis1 product and indicating that the product is Meis1d (Figure 2a). To determine the spatial expression pattern of the Meis1d protein, lysates from a panel of B6 organs were probed with the Meis1d-C antibody. Meis1d expression was limited to the cecum and proximal colon (Figure 2b, c).

Posttranslational modification of Meis1d correlates with subcellular localization
Homeodomain-less isoforms of other homeodomain proteins have been shown to function through two mechanisms. The homeodomain-less proteins can sequester full length isoforms in the cytoplasm through dominant negative interactions. Homeodomain-less isoforms can also act as nuclear binding partners for other transcription factors, altering the activation of downstream target genes. Therefore, the subcellular localization of Meis1d may indicate the functional role of the isoform in the colon. To determine the subcellular localization of Meis1d, B6 proximal colon lysates were separated into nuclear and cytoplasmic fractions and probed with Meis1-N antibody. As expected, full length Meis1 was present in the nuclear fraction (Figure 3a). The 32 kD Meis1 product was also present in the nucleus. The 27 kD Meis1d protein, however, was only observed in the cytoplasmic fraction ( Figure 3a). These localization data indicate that Meis1d is not acting as a transcription factor or sequestering full length Meis1 outside the nucleus, suggesting that the isoform has novel cytoplasmic functions.
To study the effects of Meis1d expression in vitro, the open reading frame (ORF) of Meis1d was retrovirally infected into the HCT116 colon cancer cell line. Western blot analysis of cell lysates revealed the presence of a 32 kD protein that was detected by the Meis1-N antibody, but not the Meis1d-C antibody (Figure 3b). This in vitro Meis1d product is ,5 kD larger than the expected 27 kD protein predicted by the Meis1d transcript and observed in vivo (Figure 3b). However, the molecular weight matches the novel 32 kD Meis1 isoform previously observed in B6 stomach and colon lysates (Figure 1a). The molecular weight of in vitro Meis1d was confirmed in both murine 3T3 and simian Cos-1 cells, indicating that the increased size is not cell line specific (data not shown). Posttranslational modification of the in vitro Meis1d product may be the cause of the increased molecular weight, as well as the inability of the Meis1d-C antibody to detect the protein.
Unlike the 27 kD Meis1d product, the 32 kD Meis1 isoform is localized to the nucleus in the proximal colon. To determine if in vitro Meis1d is also localized to the nucleus, HCT116 cells expressing Meis1d were fractionated. The 32 kD in vitro Meis1d product was present in the nuclear lysates, recapitulating the in vivo subcellular localization of the novel 32 kD isoform (Figure 3c).   Nuclear and cytoplasmic Meis1d are mutually exclusive in the colon The colon consists of a single layer of epithelial cells covering the lamina propria and several muscle layers. Colon cancer is derived from stem cells located in the epithelial layer [28]. To determine if either Meis1d 27 or Meis1d 32 are expressed in the colonic epithelium, epithelial cells were isolated from B6 proximal and distal colon samples. Lysates from the isolated epithelial cells and lamina propria/muscle layers were probed with Meis1-N antibody. The Meis1d 27 protein was expressed exclusively in the epithelial cells of the proximal colon (Figure 4a). Meis1d 32 , however, was expressed in the lamina propria/muscle layers from both the proximal and distal colon (Figure 4a). The full length Meis1 isoforms, Meis1a and Meis1b, were also present in only the lamina propria/muscle samples (Figure 4a). These data indicate that the nuclear and cytoplasmic forms of Meis1d are not present in the same cell populations in the murine colon. Furthermore, Meis1d 27 is expressed in colonic epithelial cells in the absence of full length Meis1. To further confirm the epithelial cell expression of Meis1d 27 , B6 colon samples were stained with Meis1-N antibody. Strong cytoplasmic staining was observed in the epithelium of the proximal colon, but not the distal colon ( Figure 4b). This expression pattern recapitulates the expression of Meis1d 27 , again suggesting that Meis1d 27 is present in the epithelial cells of the proximal colon.

Expression of the MEIS1D protein is downregulated in human colorectal cancers
Meis1 is highly conserved in eukaryotes and removal of exon 8 from the human MEIS1 mRNA would encode a predicted protein product with 99% homology to murine Meis1d. Based on the evolutionary conservation of Meis1, we attempted to characterize the expression of MEIS1D in the human colon. A western blot was performed on human colon lysates using the Meis1-N antibody. A band the predicted size of MEIS1D 27 was observed in human colon samples (Figure 5b). To confirm the expression of the MEIS1D transcript, RT-PCR was performed on human colon cDNA (Figure 5a). A 758 bp band was amplified, indicating that MEIS1D mRNA is transcribed in the human colon.
Expression of full length MEIS1 is known to be dysregulated in several solid tumors [29,30]. To determine the status of MEIS1 in colorectal cancer, western blots were performed on matched normal colon and primary colorectal tumor lysates using the Meis1-N antibody. MEIS1A and MEIS1B were expressed in both tumor and normal samples, while MEIS1D 32 was not present in either set of samples (data not shown). However, 83% (10/12) of normal colon mucosa samples expressed detectable levels of MEIS1D 27 protein (Figure 5b). Expression was reduced or completely lost in all matched tumor samples from these patients (Figure 5b). The remaining two  patients expressed undetectable levels of MEIS1D 27 in both normal mucosa and tumor tissue (Figure 5b). These in vivo data indicate that MEIS1D 27 expression is downregulated during intestinal tumorigenesis.

Discussion
Alternative splicing is known to produce two Meis1 isoforms, Meis1a and Meis1b, in both mice and humans. Both the Meis1a and Meis1b proteins are full length, containing the two Meinox domains involved in protein-protein interactions and the DNAbinding homeodomain (Figure 2d). The protein products of these splice variants differ in the C-terminus, due to the splicing out of exon 12 from the Meis1b coding sequencing (Figure 2c, d). Evidence suggests that the two similar isoforms may activate transcription of different subsets of genes [31]. Screens of cDNA libraries indicate that a large number of additional alternatively spliced transcripts are produced from the Meis1 locus [26,27,32]. Unlike Meis1a and Meis1b, however, proteins for other Meis1 transcripts have not been observed in vivo. Therefore, it remains unclear whether these transcripts are functionally relevant or simply artifacts generated by incorrect alternative splicing.
In this paper, we have described two protein products of the Meis1d splice variant, an isoform previously identified in cDNA screens. The Meis1d transcript lacks exon 8, resulting in a predicted 27 kD protein lacking the homeodomain (Figure 2c, d). This 27 kD product was observed in the cytoplasm of proximal colon epithelial cells (Figures 3a, 4a). A second, 32 kD Meis1d product occurs in the nuclei of non-epithelial cells in the stomach and colon (Figures 3a, 4a). The cytoplasmic and nuclear forms of Meis1d have been named Meis1d 27 and Meis1d 32 , respectively. The larger molecular weight and nuclear localization of the Meis1d 32 protein are recapitulated in cells transfected with the Meis1d ORF (Figure 3b). Expression of MEIS1D 27 in human colon samples was also observed, demonstrating evolutionary conservation of this novel isoform ( Figure 5). Meis1d is only the third Meis1 isoform for which a translated protein product has been confirmed and is also the first homeodomain-less Meis1 product observed in either mice or human tissues.
Homeodomain-less isoforms have been described for many homeobox genes, including several members of the Hox family. Only a few of these isoforms, however, have been identified in TALE superfamily genes, a group including Meis1. The loss of the homeodomain prevents direct binding to DNA target sequences and normal transcriptional activation. Homeodomain-less isoforms of TALE transcription factors, however, have been previously shown to still regulate transcription through two distinct mechanisms. The truncated isoforms can have dominant negative effects, sequestering full length proteins in the cytoplasm and preventing transcriptional activation of downstream targets [33,34]. The homeodomain-less proteins can also interact indirectly with DNA by binding other transcription factors to form heterodimers. The presence of the truncated proteins alters the DNA binding site of the protein complex, activating a distinct subset of downstream targets [35].
Since homeodomain-less isoforms of related genes have distinct mechanisms in the cytoplasm and nucleus, the subcellular localizations of Meis1d and any potential binding partners may suggest a cellular function. In the present study, Meis1d was observed in either the nucleus or cytoplasm depending on cell type. Meis1d 27 is expressed in the cytoplasm of proximal colon epithelial cells (Figures 3, 4). Full length isoforms are present in the nuclear fraction of proximal colon lysates, indicating that Meis1d 27 does not sequester other Meis1 isoforms in the cytoplasm (Figures 3, 4). Further evidence indicates that Meis1d 27 is not expressed in the same cells as Meis1a or Meis1b, preventing any kind of dominant negative interaction (Figure 4). The cytoplasmic form of Meis1d may still bind other unknown transcription factors and prevent nuclear localization. The isoform may also have a novel homeodomain-independent function unrelated to transcriptional activation.
The potential function of Meis1d 27 is unclear because the data does not fit either known mechanism of homeodomain-less TALE proteins. The protein can not be acting as a dominant negative, since it is not expressed in the same cells as other Meis1 isoforms. Meis1d 27 also can not be activating downstream transcription inside the nucleus, because the protein is localized to the cytoplasm. The nuclear form of Meis1d, however, may still fit either of these mechanisms in the lamina propria or muscle surrounding the colon. The nuclear localization of Meis1d 32 means the protein could be acting as a binding partner for other transcription factors. Meis1d 32 still maintains the motifs required for Pbx interaction and homodimerization [36]. These data suggest that the truncated isoform could still function as a co-factor with these other homeodomain proteins. The Meis1d 32 protein may also be acting as dominant negative isoform inside the nucleus, preventing association between full length Meis1 proteins and promoter regions in the DNA. Further dissection of the lamina propria and the underlying muscle layer is necessary to determine if Meis1d 32 and any potential binding partners are expressed in the same population of cells. Identification of functional roles for both Meis1d 32 and Meis1d 27 will help explain the relevance of Meis1 splicing in intestinal function and homeostasis.
Despite the lack of a defined function, we have provided evidence that MEIS1D 27 is downregulated in primary colorectal cancers compared to normal mucosa (Figure 5b). These data may explain the downregulation of total MEIS1 transcript previously observed in early colon adenomas [25]. Preliminary data also suggests that MEIS1D 27 is not expressed in the Hct116, LoVo, or SW480 cell lines, which are derived from human colon cancers (data not shown). Tumor suppressor genes are functionally inactivated during tumor initiation and progression through somatic mutations or loss of expression. These data suggest that MEIS1D 27 may act as a tumor suppressor in the colon. The loss of MEIS1D 27 during intestinal tumorigenesis, however, may also be indicative of more global changes in the transcriptome of the tumor cell. Alternative splicing is an essential mechanism for generating the protein diversity necessary for cellular function. Splicing provides both direct and indirect regulation of multiple cancer-related pathways, including apoptosis, signal transduction, and differentiation [37,38,39]. As with other regulatory mechanisms, alterations in alternative splicing have been observed in primary human cancers [40,41]. Transcriptome-wide changes are often caused by aberrant expression of splicing factors, resulting in simultaneous downregulation of tumor suppressive isoforms and upregulation of oncogenic isoforms. Newer microarrays can map the transcriptome with exon-level resolution, allowing analysis of these changes in primary tumors. Despite these technological advancements, there are still difficulties identifying causative links between production of specific isoforms and tumorigenesis. Due to the global changes in alternative splicing observed in cancer, MEIS1D may be one of many splice variants whose expression levels change during intestinal tumorigenesis. Future work will need to determine if there is a causative role for MEIS1D 27 in colorectal cancer or if the loss of the cytoplasmic isoform is simply a biomarker of colorectal cancer.