Reg4 Interacts with CD44 to Regulate Proliferation and Stemness of Colorectal and Pancreatic Cancer Cells

Gastrointestinal (GI) malignancies account for nearly half of worldwide cancer deaths. Improvements in survival have largely come from screening approaches designed to identify early localized disease, as tumor dissemination predicts very poor survival despite current treatments. Reg4 is a secreted 158 amino acid protein belonging to the C-type lectin superfamily (1). Increases in Reg4 expression is observed in the progression from normal GI tissues to frank malignancy through well-established cancer precursor lesions. Reg4 is expressed in aberrant crypt foci, considered to be the earliest visible preneoplastic colonic lesion (2), adenomatous and sessile serrated polyps (3, 4), and intestinal metaplasia, a recognized precursor of adenocarcinoma of the stomach, esophagus, and gallbladder (5–7). Reg4 is also upregulated in pancreatic cancer precursors. More than 90% of intestinal type intraductal papillary mucinous neoplasms are strongly positive for Reg4 expression (8). Nearly all high-grade pancreatic intraepithelial neoplasia undergoes Reg4 gene amplification during the later stages of precancerous growth and develops into pancreatic cancer (9).

Reg4 expression is also associated with aggressive disease and poor patient outcomes in cancer. In colorectal cancer, higher levels of Reg4 are predictive of drug-resistance (10, 11), shorter survival, and metastatic disease (12). In colorectal cancer tumor xenografts, both Reg4 mAbs and interfering RNAs increased tumor apoptosis and decreased proliferation, resulting in a significantly decreased tumor burden and increased survival (13). In gastric cancer, Reg4 levels are a sensitive disease biomarker (14) and are associated with lymph node spread or distant metastasis and shorter survival time (15). Others reported a correlation of Reg4 expression with invasive depth, intrinsic drug resistance to 5-FU (16) and peritoneal recurrence after curative resection (17). In prostate cancer, serum levels of Reg4 are a useful diagnostic marker (18) and correlate with tumor cell proliferation (8), migration and invasiveness (19), and resistance to gemcitabine (9).

These data suggest that Reg4 may be an important driver of the neoplastic process, rather than merely a passenger event. A barrier to progress in the field has been a lack of scientific knowledge about Reg4 signaling that may underlie tumor growth and responsiveness to treatment. Because Kras mutation is widely reported to be an important regulator of many signaling pathways, we used Kras mutant colorectal cancer (HCT116 and SW480) and pancreatic cancer (PANC-1 and AsPC-1) cell lines and Kras wild-type colorectal cancer (HT29) and pancreatic cancer (BxPC-3) cell lines showing variable expression of Reg4 to analyze results. Here we report that: (i) Reg4 interacts with CD44; a transmembrane protein notable for its expression by populations of colorectal and pancreatic cancer stem cells (CSC); (ii) Reg4 activates γ-secretase–mediated proteolysis of CD44 to release cytoplasmic intracytoplasmic domain (CD44ICD) and its translocation to the nucleus, which has been shown to be a transcriptional activator of D-type cyclins involved in regulation of cancer cell proliferation; (iii) Reg4-mediated increase in nuclear CD44ICD expression facilitates its interaction with cAMP response element (CRE)-binding protein (CREB)/p300 transcriptional complex and increases expression of Krueppel-like factor 4 (Klf4) and SRY-Box Transcription Factor 2 (Sox2) involved in regulating pluripotency of cancer stem cells; and (iv) Reg4 significantly increases colorectal cancer and pancreatic cancer cell proliferation and clonogenic potential in stem cell assays. Thus, this work identifies a previously unknown signaling pathway, active in GI malignancies, which mediates the proliferative and stemness effects of Reg4.

Human colorectal cancer (HCT116, SW480, and HT29) and pancreatic cancer (PANC-1, BxPC-3, and AsPC-1) cell lines were acquired from ATCC. No additional authentication was performed after the cells were purchased. Fresh cell stocks were regularly replenished from the original stocks and were grown in DMEM/RPMI1640 media containing 10% heat inactivated FBS (All from GIBCO by Life Technologies). Growing cells were regularly tested for mycoplasma at every 3 to 4 months using Mycoplasma Detection Kit (InvivoGen). Mycoplasma-free cells used for experiments were from passage number 3 to 30 at approximately 70% to 80% confluency.

Total protein from cells was extracted using Bolt LDS sample loading buffer (Novex by Life Technologies). Total protein from cytoplasmic and nuclear fractions of cells was extracted using NE-PER nuclear and cytoplasmic extraction reagents (Pierce Biotechnology). Western blotting was performed using specific primary antibodies including α-Reg4 and α-CD44ICD (generated at GeneScript), α-CD44, α-Cyclin D1, α-Cyclin D3, α-CREB (phosphorylated and total), and α-Klf4 (all from Cell Signaling Technology), α-p300/CBP (Thermo Fisher Scientific), and α-β-actin (Santa Cruz Biotechnology) and protein expression was analyzed by detecting specific bands using enhanced chemiluminescence system (GE Healthcare). β-Actin, once thought to be an exclusively cytoplasmic protein, is now known to have its presence and important functions within the nucleus, hence served as a loading control for proteins in whole cell, cytoplasmic, and nuclear extracts (20, 21). Intensity of individual protein bands was estimated by densitometry scanning using Kodak Digital Science Image Station 440 (Eastman Kodak Company) and Carestream Molecular Imaging software.

A modified pull-down assay was performed to identify cell surface binding partner of Reg4 protein. Cells were first treated with biotinylated rhR4 (or biotinylated BSA as a control), and then surface binding proteins were cross-linked using a water-soluble amine-to-amine cross-linker BS³ [bis(sulfosuccinimidyl)suberate]. Following cell lysis, a complex of biotinylated rhR4 and its cell surface binding partner was pulled down with NetrAvidin agarose and were digested in separate reactions with Lys-C or trypsin. Peptides were analyzed for differential protein expression by high-resolution tandem mass spectrometry. The interaction of Reg4 and its binding partner was also confirmed by immunoprecipitation using specific antibodies followed by Western blot analysis, and flow cytometry assays using BD Accuri Cflow sampler (BD Biosciences).

The rate of colorectal cancer and pancreatic cancer cell proliferation was determined by using Click-iT EdU flow cytometry assay (Invitrogen/Molecular Probes). EdU (5-Ethynyl-2′-deoxyuridine), a nucleoside analog to thymidine is incorporated into DNA during active DNA synthesis. EdU in newly synthesized DNA was detected using Alexa Fluor 488 dye. The percentage of Alexa Flour-positive cells in S-phase cell population was analyzed by BD Accuri Cflow sampler (BD Biosciences).

Duolink In Situ PLA using reagents and probes from Millipore/Sigma was performed to exhibit protein–protein interaction in colorectal cancer/pancreatic cancer cells. Brief description of assay includes steps for binding of proteins of interest with primary antibodies and their subsequent detection using secondary antibodies. Primary antibodies conjugated with oligonucleotides (PLA probes MINUS and PLUS) bind to their respective protein targets and proximity is determined by ligation and amplification.

The nuclei from human colorectal cancer cells (HCT116 and HT29; both control and Reg4-treated) were isolated using protocol as described previously by Rosner and Hengstschlaeger (22). Briefly, lysed cells in PBS were pelleted and resuspended in cytoplasmic extraction buffer [20 mmol/L Tris-pH 7.6, 50 mmol/L 2-mercaptoethanol, 0.1 mmol/L EDTA, 2 mmol/L MgCl₂, 0.5 mmol/L NaF, and 0.5 mmol/L Na₃VO₄; 1× protease inhibitor mix (PIM), and 1× PMSF were added shortly before use]. Cytoplasmic extract was separated from intact nuclei by low-speed centrifugation. The remaining pellet containing crude nuclei was washed in buffer supplemented with 1% Nonidet P-40 substitute. The nuclear pellet was then resuspended in nuclear extraction buffer (20 mmol/L Tris-pH 7.6, 420 mmol/L NaCl, 1.5 mmol/L MgCl2, 0.2 mmol/L EDTA, 1 mmol/L PMSF) and nuclei were pelleted by low-speed centrifugation. Paraformaldehyde (16%, v/v) was added to isolated nuclei in PBS to obtain a final concentration of 1.5% (v/v) paraformaldehyde and nuclear samples were fixed while incubating them on a roller mixer for 10 min at room temperature under constant, mild agitation. The fixed nuclei were then permeabilized by slowly adding cold (−20°C) 100% methanol to prechilled samples, while gently vortexing to obtain a final concentration of ∼0.5 × 10⁶ nuclei/mL 100% methanol. Following centrifugation, the pelleted nuclei were resuspended in staining buffer. After a short blocking at room temperature, unconjugated primary antibodies specific to proteins of interest were added and incubated for 60 minutes at room temperature. Fluorochrome-conjugated secondary antibody was finally used for flow cytometric detection using BD Accuri C6 flow cytometer. The percentage of Reg4/CD44ICD-positive nuclei was analyzed using flow cytometry.

The spheroid forming culture is the most widely used in vitro functional assay for assessing growth of cancer stem cells. Colorectal cancer and pancreatic cancer cells were seeded in media containing EGF (1 ng/mL; R&D Systems), insulin (0.4 μg/mL; Life Technologies), and Rock Inhibitor (10 μmol/L; ATCC) into the wells of low adhesion culture plate. Cells were grown 14 days in above-mentioned culture media and the number and size of growing spheroids were quantified microscopically. Growing spheroids were microscopically photographed in a fixed focal area to represent visible changes in the size of spheroids, whereas total number of growing spheroids were microscopically counted in each well and then normalized to the number of spheroids/1,000 seeded cells to represent graphical changes in the respective experimental groups.

Total RNA isolated from colorectal cancer (HT29 and SW480) and pancreatic cancer (PANC-1 and AsPC-1) cells using TRizol was reverse transcribed using random hexanucleotide primers and SuperScript IV Reverse Transcriptase (Invitrogen). cDNAs were then mixed with SsoFast Evagreen Supermix to perform real-time RT-PCR analysis using CFX96 Real-Time System (Bio-Rad). Crossing threshold values for individual genes were normalized to β-actin. Changes in mRNA expression were expressed as fold change relative to control. Primers used in this study include: β-actin: 5′-ATCATTGCTCCTGAGCG-3′ and 5′-GCTGATCCACATCTGGAA-3′, Reg4: 5′-TGAGCTGCCTGGCCAAA-3′ and 5′-AAGTAACCATAGCAATTGGACTTGTG-3′, Klf4: 5′-AGAGACCGAGGAGTTCAACG-3′ and 5′-CGGATCGGATAGGTGAAGC-3′, and Sox-2: 5′-CAGCGCATGGACAGTTACG-3′ and 5′-AGCCGTTCATGTAGGTCTGC-3′.

All values were expressed as the mean ± SEM. Data with unequal variances were analyzed using a two-tailed t test with Satterthwaite method. A “P” value of less than 0.05 was considered to indicate statistical significance.

The data that support the findings of this study are available from the corresponding author upon reasonable request. Supplementary data for this article are available at Molecular Cancer Research Online (http://mcr.aacrjournals.org/).

Although we previously reported a Reg4-mediated induction of EGFR activity in colorectal cancer cells (23), a direct interaction between Reg4 and EGFR was not observed. We accordingly extended our studies to identify cell surface proteins interacting with Reg4 and determined their possibilities to serve as a receptor for mediating Reg4 signaling in cancer cells. Biotinylated rhR4 and biotinylated BSA (as a control) were added to HCT116 cell suspensions. A pull-down assay using Lys-C or trypsin digestion followed by a high-resolution tandem mass spectrometry was performed to analyze differential expression of peptides interacting with Reg4. A list of proteins pooled down with biotinylated Reg4 and identified by mass spectrophotometric analysis is shown in Supplementary Table S1. CD44R1 and CD44 molecule (Indian blood group) were identified as high scoring proteins that bound to biotinylated rhR4 but not to control biotinylated BSA. Considering CD44, a cell surface protein, highly expressed in many cancers and known to be a receptor for hyaluronan and several other ligands including, osteopontin (OPN), collagens, and matrix metalloproteinases, is identified as a possible receptor for Reg4 protein. To both confirm the interaction between Reg4 and CD44, and examine a broader group of GI cancer cells, we performed coimmunoprecipitation of total protein extract from multiple colorectal cancer (HCT116, HT29, and SW480) and pancreatic cancer (PANC-1 and AsPC-1) cells using antibodies specific to CD44 and Reg4 proteins. As assessed by Western blot analysis, addition of Reg4-specific antibody to cancer cell extracts increased coimmunoprecipitation of CD44, and addition of CD44-specific antibody increased coimmunoprecipitation of Reg4 (Fig. 1A). These results suggest a direct interaction of Reg4 with CD44 receptor complex in several colorectal cancer and pancreatic cancer cells.

The signaling paradigm, known as regulated intramembrane proteolysis (RIP), has recently been recognized as a unique signaling mechanism used by a number of Type 1 transmembrane proteins, including Notch, Delta, and CD44. CD44 is proteolytically cleaved at the ectodomain through membrane-associated metalloproteases in various cancer cell lines to produce a membrane-bound cleavage product (24). The remaining membrane bound CD44 protein is further cleaved by γ-secretase to release the CD44ICD fraction into the cytoplasm and then transported to the nucleus. To determine a role of Reg4 in RIP signaling, we first generated data showing that Reg4 induces γ-secretase–mediated proteolytic cleavage of CD44 leading to an increased expression of CD44ICD in the nuclear fraction of the human colorectal cancer cells. In an initial set of experiments, colorectal cancer cells (HCT116 and SW480) were transfected with DDK-tagged CD44v3 DNA (a common variant of CD44 present in many GI cancer cells) and a time-dependent effect of Reg4 (200 nmol/L) treatment on proteolytic cleavage of CD44 was determined in nuclear fraction of cells using α-DDK antibody following Western blot analysis. A phorbol ester 12-0-tetradecanoylphorbol-13-acetate (TPA) was used as a positive control to show an induced proteolytic cleavage of CD44 generating CD44ICD fraction of ∼16 kDa size. HCT116 and SW480 cells transfected with DDK-tagged CD44v3 DNA were treated with Reg4 (200 nmol/L) for different time periods (5, 15, 30, and 60 minutes) and the expression of CD44ICD in nuclear extract of cells was analyzed by Western blot. Similar to TPA, Reg4 treatments potently induced proteolytic cleavage of CD44 to generate 16 KDa fraction of CD44 representing the intracytoplasmic domain observed in the nuclear extract of cells (Fig. 1B: left and middle). In a separate set of experiments, SW480 cells were pretreated with γ-secretase specific inhibitor (2 μmol/L; γ-secretase inhibitor IX:DAPT, N-[N-(3,5-difluorophenacetyl-L-alanyl)]-S-phenylglycine t-butyl ester—Calbiochem, Millipore Sigma) for 1 hour before adding Reg4 (200 nmol/L) for 60 minutes. The protein band of ∼16 kDa exhibited an inhibitor-specific decrease, hence confirmed a pathway inducing γ-secretase activity leading to generation of nuclear CD44ICD (Fig. 1B, right). On the basis of the results of Fig. 1B showing Reg4-mediated time-dependent increase in CD44ICD expression, we further extended our experiment by selecting a time point of 60 minutes to reconfirm Reg4-mediated increases in CD44ICD expression in colorectal cancer (HCT116 and SW480) and pancreatic cancer (PANC-1) cells using α-DDK antibody following Western blot analysis (Fig. 1C). For next set of experiments, we generated affinity purified goat pAb against CD44ICD peptide (sequence: AVEDRKPSGLNGEAC; GenScript), and used to directly detect Reg4-mediated increase in endogenous CD44ICD expression in CRC (HCT116, SW480, and HT29) and PC (PANC-1 and BxPC-3) cells (Fig. 1D). To confirm the involvement of γ-secretase-mediated RIP of CD44 generating CD44ICD and its association with Reg4-mediated signaling, we pretreated the CRC (HCT116, SW480, and HT29) and PC (PANC-1 and BxPC-3) cells with γ-secretase inhibitor (GSI; 2 μmol/L) for 1 hour before adding Reg4 for 60 minutes, and the expression of endogenous CD44ICD was detected by Western blot analysis using goat α-CD44ICD pAb. These results again showed that Reg4 treatment results in increased nuclear CD44ICD expression, whereas pretreatment with GSI blocked this effect in colorectal cancer and pancreatic cancer cells (Fig. 1D).

We next generated a unique protocol of nuclei isolation and a sensitive measurement of nuclear CD44ICD in small subpopulation of cells using flow cytometry techniques. To confirm Reg4-mediated increase in nuclear CD44ICD, we used control and Reg4-treated (200 nmol/L; 1 hour) HCT116 and HT29 cells to isolate intact nuclei followed by the detection of CD44ICD. Detection was performed using goat α-CD44ICD pAb or isotype control and FITC conjugated secondary antibody. Top panels of Fig. 2A and B show gating of the isolated unstained nuclei by SSC/FSC. Gates for staining were set so that unstained nuclei and isotype control showed <0.3% positive staining in FL1. As shown in Fig. 2A, exogeneous Reg4 treatment in HCT116 cells resulted in 2.6% CD44ICD-positive nuclei in comparison with 0.8% in endogenous Reg4 control (untreated). Similarly, exogenous Reg4 treatment in HT29 cells resulted in 3.6% CD44ICD-positive nuclei in comparison to 1.4% in endogenous Reg4 control (untreated). These results further demonstrate that Reg4-mediated pathway for increasing nuclear CD44ICD expression is active in a subpopulation of cancer cells.

First description of phorbol ester-induced proteolytic release of CD44ICD was reported in glioblastoma cells (25). CD44ICD activated the 12-otetradecanoylphorbol 13-acetate (TPA)-responsive element (TRE) in glioma cells and involved the transcriptional co-activators CBP (CREB binding protein) and p300. Initially the mechanism was unclear because they were not able to show a direct interaction between CBP/p300 and CD44ICD (25). Reg4-mediated increase in CREB phosphorylation was initially reported in PC cell proliferation and invasion through EGFR/Akt/CREB pathway (26), a signaling pathway similar to our finding showing Reg IV activates the EGFR/Akt/AP-1 signaling pathway in colon adenocarcinomas (23). Recently, CD44ICD was shown to transactivate CRE (cAMP-responsive element) reporters. CREB is activated by phosphorylation on serine 133 and then binds p300 to become an active transcription factor. CD44ICD binds to CREB in the complex and blocks dephosphorylation, thereby maintaining a transcriptionally active complex (27). To establish the involvement of Reg4-mediated increase in CD44ICD and its effect on the CBP/p300 transcription factor complex, we examined cultures of CRC (HT29 and SW480) and PC (PANC-1 and BxPC-3) cells. Western blot analysis showed a Reg4-mediated rapid induction of CREB phosphorylation at Ser133 (Fig. 3A). In addition, we observed similar increases in p300 expression in response to rhR4 treatment. To establish a possible link between Reg4-mediated increase in CD44ICD expression and CREB/p300 activities, PLAs were performed using pAbs specific to CD44ICD, CREB, and p300 in HT29 cells treated with rhR4 for 1 hour. By PLA analyses, we observed an increased colocalization/binding of CD44ICD with CREB and p300 following Reg4 treatment (Fig. 3B; bottom panels shown by the presence of red fluorescent color) in a very small population of cells. These results show a role of Reg4-mediated CD44ICD signaling for activation of CREB/CBP/p300 transcriptional complex.

Cancer cells share the ability to proliferate beyond the constraints limiting growth in normal tissue. Therefore, we next determined the role of Reg4–CD44 interaction and subsequent release of CD44ICD in regulation of colorectal cancer and pancreatic cancer cell proliferation. Colorectal cancer (HT29) and pancreatic cancer (PANC-1 and BxPC-3) cells were treated with rhR4 (200 nmol/L) for 24 hours and expression of previously reported Reg4-regulated proteins involved in cancer cell proliferation (Cyclin D3 and D1) was estimated by western blot analysis. As reported previously, treatment of colorectal cancer cells with Reg4-induced expression of Cyclin D3 and/or D1, we next examined if this induction was blocked by γ-secretase inhibition. Pretreatments of GSIs (2 μmol/L) 1 hour prior to rhR4 addition blocked Reg4-mediated cyclin induction (Fig. 4A). We next examined the effects of Reg4 and GSI on colorectal cancer and pancreatic cancer cell proliferation using EdU incorporation determined by flow cytometry. Reg4 treatment (200 nmol/L: 24 hours) significantly increased colorectal cancer (HT29 and HCT116) and pancreatic cancer (PANC-1 and BxPC-3) cell proliferation, whereas pretreatment of GSI (2 μmol/L: 1 hour prior to rhR4 addition) blocked Reg4-induced effects (Fig. 4B).

Stem cells are capable of unlimited self-renewal ability, and unique subsets of cancer cells that acquire stem cell properties have the ability to form de novo tumors when grown under low-attachment conditions with minimal growth factor supplementation. Tumorigenic efficiency of CSCs can be determined on the basis of number of spheroids that emerge from single cells. Therefore, spheroid forming assay is one of the most widely used in vitro technique for assessing clonogenic growth potential of CSCs (28). To determine the role of Reg4 and its signaling pathway through CD44 interaction, we performed spheroid forming assays using colorectal cancer (HT29 and SW80) and pancreatic cancer (PANC-1 and AsPC-1) cells. After plating the cells into the wells of low-adhesion culture plates containing minimal essential media supplemented with EGF (1 ng/mL), insulin (0.4 μg/mL), and rock inhibitor (10 μmol/L), cells were allowed to grow and form spheroids for 14 days. Before seeding the cells for spheroid formation, cancer cells were treated with or without rhR4 (200 nmol/L; 24 hours) and with or without pretreatments of the GSI (2 μmol/L; 1 hour). The number and size of growing spheroids were noted microscopically, and representative photographs and graphical presentation of numbers are shown in Fig. 5A and B. Because growing spheroids were microscopically photographed in a fixed focal area, number of larger sized spheroids are visibly lesser in number, whereas smaller sized spheroids are visibly greater in number in comparison with their respective controls (top panels of Fig. 5A and B). To take care of these visible discrepancies, total number of growing spheroids were microscopically counted in each well and values normalized to the number of spheroids/1,000 seeded cells are used in representative graphs to show changes in the number of spheroids in respective experimental groups (bottom panels of Fig. 5A and B). Results demonstrated Reg4-mediated increase in size and number of spheroids grown from colorectal cancer (Fig. 5A) and pancreatic cancer (Fig. 5B) cells, whereas pretreatment of GSI blocked the Reg4-mediated induction of spheroid growth.

Stem cells including CSCs demonstrate increased expression of pluripotency genes. Since, we observed Reg4-mediated increase in CSCs growth, we next performed experiments to determine underlying molecular mechanism mediating Reg4 effects in CSCs. Using real time RT-PCR analysis, we first determined a parallelism between the expression of Reg4 and a panel of pluripotency-associated genes including Oct3, Oct4, Nanog, Lgr5, c-Myc, Wif1, Bmi1, Id1, Sox2 and Klf4 in CRC/PC cells grown as monolayer, and their corresponding CSCs grown as spheroids. We observed significant upregulation and a parallelism between Reg4 and Klf4 in CSCs (Fig. 6A) that bears a potential CREB/p300-binding site in their promoter sequence. In addition, treatments of Reg4 to CRC (HT29 & SW480) and PC (PANC-1 & AsPC-1) cells for different time periods (1, 2, 4, 8, 12 & 24 h) led to rapid (1 & 2 h) induction in Klf4 expression (Fig. 6B). To determine the involvement of CD44ICD in Reg4-mediated induction of Klf4 expression, we inhibited γ-secretase activity by adding specific GSI (2 μmol/L) 1 hour prior to Reg4 treatment and observed a significant blocking of Reg4-mediated increase in Klf4 expression (Fig. 6C). Taken together, these results identified Klf4 as a connecting link mediating Reg4 signaling through γ-secretase–mediated release of CD44ICD and its subsequent interaction with CREB/p300 transcriptional complex to regulate growth and survival of CSCs.

Yamanaka-factors including Oct4, Klf4, Sox2, and c-Myc are identified as crucial pluripotency-associated genes in cancer stem cells involved in a complex regulatory network of signaling molecules, kinases, and miRNAs influencing cell (de-) differentiation on a transcriptional, posttranscriptional, and translational level (29, 30). Reprogramming of any somatic cell type can be achieved by initiating several synergistic processes of these genes. In the process of reprogramming, induced pluripotency elicits several transcriptional waves driven by c-Myc/Klf4 and Oct4/Sox2/Klf4. To determine this complex network of signaling, our next set of experiments identified Sox2 in addition to Klf4 as a crucial factor of pluripotency network as we observed significant upregulation and a parallelism between Reg4 and Sox2 in CSCs (Figs. 6A, left and 7A). In addition, we observed a similar Reg4-mediated increase in Sox2 expression and its inhibition by using GSI (Fig. 7B and C). These results suggested a possible involvement of Klf4/Sox2 network in association of Reg4-mediated induction of CD44ICD-CREB/p300 signaling in CSCs. Therefore, results of this experiment provide possible explanation for blocking this pathway for enhanced therapeutic effect for treatment of colorectal and pancreatic cancers.

Increased expression of Reg4 has been observed in the progression from normal gastrointestinal tissues to frank malignancy through well-established cancer precursor lesions. Reg4 expression is increased in colonic aberrant crypt foci (31), adenomatous (3), and sessile serrated polyps (4). Reg4 also is upregulated in gastric intestinal metaplasia and adenomas (5), intestinal metaplasia of the Barrett esophagus (32), intestinal metaplasia (7) of the gallbladder, and pancreatic cancer precursors: intraductal papillary mucinous neoplasms (IPMN) and high-grade pancreatic intraepithelial neoplasia (PanIN; ref. 9). A role for Reg4 as a driver of the malignant phenotype has also been advanced by a considerable literature showing the association between Reg4 expression and aggressive disease and poor cancer outcomes. We have shown Reg4 to be a potent activator of the EGF receptor/Akt/AP-1 signaling pathway in colorectal cancer cell lines, comparable with EGF, leading to increased expression of Bcl-2, Bcl-XL, Survivin, and Matrilysin, genes associated with a poor prognosis in advanced colorectal cancer (23). Reg4 antagonists also reduced tumor growth and increased apoptosis in human tumor xenografts (13). However, Reg4 does not bind directly to the EGFR and progress toward a mechanistic understanding has been slow. In this study, we show that Reg4 interacts with CD44, induces its proteolytic cleavage, and releases disordered CD44ICD in the cytoplasm which is then translocated to the nucleus.

RIP-based signaling has recently become recognized as a signaling mechanism used by a number of type 1 transmembrane proteins including Notch, Delta, and CD44. In a common intramembrane proteolytic event, CD44 undergoes a metalloproteinase-mediated cleavage that results in shedding of the extracellular ectodomain (ectoCD44; ref. 13). Cleavage of the ectoCD44 is particularly prominent in malignancies, where high serum levels correlate with poor clinical outcomes (33). In patients with gastric and colon cancer, ectoCD44 levels in blood correlate with tumor burden and decrease immediately after resection or treatment (34). CD44 species containing variant exons appear to represent a preferred substrate for tumor proteolysis (35). Several signaling pathways including PKC, influx of extracellular Ca2+, Rho family of GTPases and Ras regulate ectodomain cleavage (24, 36). After cleavage of the ectodomain, the remaining membrane bound C-terminal fragment (CD44-CTF) is also a target of intramembrane proteolysis mediated by γ-secretase (25, 27).

Phorbol esters previously have been shown to proteolytically cleave and release CD44ICD in glioblastoma cells, and to transactivate CRE (cAMP-responsive element) reporters. CREB is activated by phosphorylation on serine 133 and then binds p300 to become an active transcription factor. CD44ICD binds to CREB in the complex and blocks dephosphorylation, thereby maintaining a transcriptionally active complex (27). Interestingly, p300 and CREBBP constitute a unique family of lysine acetyltransferases (KAT) that have recently been shown to direct a broad acetylome, acetylating thousands of sites, including signature histone H2B, signaling effectors, and transcriptional regulators. Indeed, it is our hypothesis that CD44ICD acts as a direct regulator of this signaling program. Catalytic inhibitors of p300 and CREBBP acetyltransferases reduce acetylation of key signaling effectors involved in Notch, Hedgehog, Wnt, and Hippo/YAP, among others (37). CD44ICD also increases ATF-1 phosphorylation, suggesting that CD44ICD may also regulate other transcription factors. Cyclin D1 is activated by CREB-p300 and drives proliferation in thyroid carcinoma cells. We recently showed that Reg4 treatment leads to similar increases in Cyclin D1 and/or D3 in colon cancer cells (38). Furthermore, we observed Reg4-mediated increase in Klf4 gene bearing CREB-p300 binding site to be involved in regulating pluripotency of CSCs. In addition, a study using breast cancer models showed that CD44ICD induces MMP-9 transcription by interaction with Runt-related transcription factor 2 (Runx2) and binds to a novel CD44ICD response element (CIRE; ref. 39). In prostate cancer cells, endogenous Reg4 controls invasiveness by upregulating MMP-7 and MMP-9 (24). CIRE is present in promoters for hypoxia-induced transcription factor (Hif1α) and three genes (ALDOC, PDK1, and PFKFB4); it was speculated that this may contribute to high rates of glycolysis and cytosolic lactic acid fermentation that occur in tumors (the “Warburg effect”), that is the basis for PET scans (40). In breast cancer, CD44ICD was reported to interact with CREB and directly activate PFKFB4 transcription, facilitating glycolysis, and stemness (41). We recently published a study documenting the relationship between Reg4, CD44, and nuclear CD44ICD expression in patients with stage II/III colorectal cancer (42). In this study, we used our goat pAb against CD44ICD for IHC on tumor samples from a large, well characterized colorectal cancer patient cohort, and validated the clinical importance of Reg4–CD44–CD44ICD signaling in tumor stage and patient survival. These data suggests that Reg4–CD44ICD signaling may regulate key genes important in tumor proliferation, metastasis, and glycolytic metabolism and adds support to the idea that CD44ICD signaling may represent a common pathway, downstream of multiple oncogenic pathways in cancer.

Papillary thyroid carcinomas harboring activated RET/PTC, RAS, or BRAF use CD44ICD to sustain cell proliferation via CREB-dependent transcriptional activation of Cyclin D1 (27). Knockdown of CD44 by RNA interference or GSIs (GSI) blocked proliferation, which was restored by CD44ICD overexpression. CD44ICD is involved in the receptor tyrosine kinase Ret-induced transformation of rat fibroblast cells and transfection of CD44ICD was sufficient to transform fibroblasts and allow anchorage-independent growth in soft agar (43). In pancreatic cancer cells CD44-positive cells are responsible for gemcitabine resistance (44). These data support our finding that the binding of Reg4 to CD44 receptor and a subsequent release of CD44ICD for nuclear translocation are central in the proliferation and apoptosis resistance of GI cancers. Although, pretreatment of GSI blocked Reg4-mediated increase in cell proliferation, treatment of GSI alone was quite inconsistent for inhibiting proliferation of colorectal cancer and pancreatic cancer cells, hence raised an issue while using GSI as a single agent to control colorectal cancer/pancreatic cancer growth and cancer stem cell susceptibility to chemo-/radiotherapies. In addition, GSIs have been hindered in clinical trials by serious grade 3/4 gastrointestinal toxicity; albeit this is not surprising considering multiple type 1 transmembrane proteins have recently been identified as γ-secretase substrates (45). Moreover, our finding that Reg4 treatment, in a γ-secretase dependent fashion, increases CSCs clonogenic potential in sphere forming and soft agar assays show that this pathway is active in the CSCs and suggest CD44ICD as a promising target to antagonize Reg4–CD44 signaling for a better patient outcomes of GI malignancies.

Clonogenic potential of Reg4 in colorectal cancer and prostate cancer cells, and its inhibition by using GSIs and Reg4 antibodies support a role for Reg4 in the CSCs as assayed by an in vitro marker of stemness. This fits well with other published data showing Reg4 silencing resulted in loss of stem-like properties including inhibition of spheroid growth and increased sensitivity to chemoradiation-induced cell death in poorly differentiated gastric cancer cells (46). Reg4 was also required for the tumorigenic activity of ALDH1⁺ enriched CSCs in diffuse-type GC and Reg4 knockdown dramatically reduced soft agar colony growth (47). As we have shown that Reg4 binds CD44 and directs the release of the CD44ICD, this provides a likely mechanistic explanation for a publication several years ago showing CD44 was involved in colorectal cancer stemness, defined by in vitro spheroid growth, in a process dependent on the carboxy terminal end of CD44 (48). Recently another CD44 ligand, osteopontin (OPN), was shown to trigger CD44ICD release and mediate tumor stemness, growth, and radiation resistance in malignant gliomas. CD44ICD transfection alone mimicked the effect of osteopontin, promoting stemness as assessed by (i) exclusion of Hoechst dye in the side population assay; (ii) expression of stem cell markers (nanog, Sox2, Oct4, and Id1); and (iii) the colony-forming ability of glioma cells following irradiation (49). An interaction between osteopontin and CD44 was found to mediate aggressive tumor growth and enhanced stemness of glioblastoma multiforme cells (49). These findings suggest that Reg4 may be one of multiple CD44 ligands capable of supporting RIP of CD44. However, it remains to be tested if osteopontin and other CD44 agonists, like Reg4, also lead to transactivation of the EGF receptor. As our data show, Reg4-mediated transactivation of the EGF receptor is not blocked by current therapeutic anti-EGFR monoclonal antibodies used in cancer therapy, these findings may also have significant therapeutic implications and support the possible utility of simultaneous blockage of CD44 agonists.

M.A. Ciorba reports grants from VA and NIH/NIDDK during the conduct of the study. C.W. Houchen is a cofounder of COARE Holdings Inc. No disclosures were reported by the other authors.

K.S. Bishnupuri: Conceptualization, resources, data curation, software, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. S.K. Sainathan: Data curation. M.A. Ciorba: Funding acquisition, writing–review and editing. C.W. Houchen: Formal analysis, writing–review and editing. B.K. Dieckgraefe: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We thank Dr. Reid Townsend, Professor of Medicine and Director of Mass Spectrometry Resource, Washington University School of Medicine, St Louis, for his help in performing high-resolution tandem mass spectrometry, and Dr. Ling Chen, Assistant Professor of Biostatistics, Washington University School of Medicine, St Louis, for her valuable suggestions in statistical analysis of data. Supported by VA Merit Grant I01 BX003072, NIH grant R01DK060106, Siteman Investment Program grant Pre-RO1 #5383 to B.K. Dieckgraefe; and NIH grant 1R01DK109384 to M.A. Ciorba.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.