Small proline rich protein 2a in benign and malignant liver disease


  • Yoshiaki Mizuguchi,

    1. Thomas E. Starzl Transplantation Institute, University of Pittsburgh Medical Center, Pittsburgh, PA
    2. Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, PA
    Search for more papers by this author
  • Kumiko Isse,

    1. Thomas E. Starzl Transplantation Institute, University of Pittsburgh Medical Center, Pittsburgh, PA
    2. Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, PA
    Search for more papers by this author
  • Susan Specht,

    1. Thomas E. Starzl Transplantation Institute, University of Pittsburgh Medical Center, Pittsburgh, PA
    2. Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, PA
    Search for more papers by this author
  • John G. Lunz III,

    1. Thomas E. Starzl Transplantation Institute, University of Pittsburgh Medical Center, Pittsburgh, PA
    2. Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, PA
    3. Department of Surgery, Divisions of Transplantation, University of Pittsburgh Medical Center, Pittsburgh, PA
    Search for more papers by this author
  • Natasha Corbitt,

    1. Thomas E. Starzl Transplantation Institute, University of Pittsburgh Medical Center, Pittsburgh, PA
    2. Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, PA
    Search for more papers by this author
  • Toshihiro Takizawa,

    1. Department of Molecular Anatomy and Medicine, Nippon Medical School, Tokyo, Japan
    Search for more papers by this author
  • Anthony J. Demetris

    Corresponding author
    1. Thomas E. Starzl Transplantation Institute, University of Pittsburgh Medical Center, Pittsburgh, PA
    2. Department of Pathology, University of Pittsburgh Medical Center, Pittsburgh, PA
    • Address reprint requests to: A.J. Demetris, M.D., Thomas E. Starzl Transplantation Institute, University of Pittsburgh Medical Center, Pittsburgh, PA 15260. E-mail:

    Search for more papers by this author

  • Potential conflict of interest: Dr. Demetris consults for DCL Medical Lab and AbbVie.

  • This work was supported by the Thomas E. Starzl Professor of Pathology Endowment Fund and the Ministry of Education, Science, Sports, and Culture of Japan and Grants-in-Aid for Scientific Research ((C) #24590559), Research Fellowships for young scientists and the Core Research Project for Private University: matching fund subsidy.


STAT3-driven expression of small proline rich protein 2a (SPRR2a), which acts as an src homology 3 (SH3) domain ligand, induces biliary epithelial cell (BEC) epithelial-mesenchymal transition (EMT), which, in turn, promotes wound healing. SPRR2a also quenches free radicals and protects against oxidative stress and DNA damage in nonneoplastic BEC. Sprr2a-induced EMT also increases local invasiveness of cholangiocarcinomas (CC), but prevents metastases. Understanding SPRR2a regulation of EMT has potential for therapeutic targeting in both benign and malignant liver disease. Molecular mechanisms responsible for SPRR2a-induced EMT were characterized, in vitro, and then evidence for utilization of these pathways was sought in human intrahepatic CC, in vivo, using multiplex labeling and software-assisted morphometric analysis. SPRR2a complexes with ZEB1 and CtBP on the microRNA (miR)-200c/141 promoter resulting in synergic suppression of miR-200c/141 transcription, which is required for maintenance of the BEC epithelial phenotype. SPRR2a induction promotes dephosphorylation and nuclear translocation of the SH3-domain containing protein GRB2 and an SH3-domain ligand in ZEB1 is required for SPRR2a-induced synergic suppression of miR-200c/141. Multiplex protein labeling of CC and morphometric analyses showed: 1) up-regulation of ZEB-1, and 2) down-regulation of CK19 in intrahepatic CC compared to nonneoplastic BEC, consistent with previous CC proteomic studies showing EMT during cholangiocarcinogenesis. Conclusion: SPRR2a modulates ZEB-1 signaling by way of miR-200c/141-associated EMT through SH3-domain networks and contributes to benign and malignant BEC wound-healing responses. (Hepatology 2014;59:1130–1143)


bile duct ligation


biliary epithelial cells


epidermal differentiation complex


epithelial to mesenchymal transition








src homology 3


smad interacting domain


small proline rich protein 2a

Small proline rich protein (SPRR) 2a is one of 14 SPRR family member genes encoded in a 170-kb region of the epidermal differentiation complex (EDC).[1] SPRR2a is coordinately expressed with other EDC genes during epidermal differentiation: these highly conserved proteins are critical for formation and maintenance of the epidermal cornified cell envelope, an effective barrier against the environment.[1, 2] However, noncoordinate/independent SPRR2a up-regulation occurs during a variety of pathophysiological conditions involving nonsquamous epithelia, indicating non-EDC functions.[3, 4] For example, SPRR2a is highly up-regulated during inflammation, stress, and infection involving barrier epithelia protecting the lung, skin, and intestine (reviewed[3-6]). In liver, we first showed SPRR2 importantly contributes to biliary epithelial wound repair and cytoprotection,[3, 7] an observation later confirmed in skin.[5, 6] In fact, in mice genetically predisposed to “fast” wound healing, SPRR2a was the most differentially up-regulated gene compared to “slow” wound healers.[4]

Interestingly, when comparing humans to chimpanzees, a “hot spot” for nonsynonymous substitutions on human chromosome 1 occurs between SPRR2A and SPRR2F,[8] with the EDC being the most rapidly diverging gene cluster.[9] SPRR diversity lies primarily within the regulatory regions,[1] which is thought to enable SPRR2 proteins to precisely modify barriers in response to environmental stressors.[1] The rapid divergence of SPRR loci suggests they are major contributors to natural selection during recent evolution.

SPRR2a contributes to biliary wound healing by promoting biliary epithelial cells (BEC) epithelial-mesenchymal transition (EMT), defined as impaired E-cadherin expression, which causes epithelial cells to undergo morphological changes marked by loss of cell polarity and cell-cell contacts and enhances migration during embryogenesis (type 1 EMT), wound healing (type 2 EMT), and carcinogenesis (type 3 EMT) (reviewed[10, 11]). Exuberant biliary wound healing or ductular reactions (type 2 EMT),[3, 7] however, can provoke fibrosis by way of activation of nearby portal myofibroblasts (reviewed[10]) in diseases such as biliary atresia,[12] primary biliary cirrhosis (PBC),[13] and others.[14] Acquisition of BEC mesenchymal characteristics, however, is only transient.[3] In human cholangiocarcinoma cell lines SPRR2a expression enhances local invasiveness by promoting motility, but prevents metastases by blocking re-epithelialization, consistent with enhanced metastatic capabilities of humans cancers lacking SPRR2a expression.[10, 15]

A recent transcriptome analysis of human intrahepatic cholangiocarcinomas suggests that a ZEB1-miR200c feedback loop regulates an EMT signaling axis that might be critical to tumor progression.[19] The transcriptional repressors ZEB1 and ZEB2 regulate EMT by way of microRNA (miR)−200 family members[16-18]: high miR-200 suppresses ZEB1 production by way of binding to its 3′ untranslated messenger RNA (mRNA). Low ZEB1 levels, in turn, increase E-cadherin expression maintaining an epithelial phenotype. Conversely, ZEB1 levels rise during EMT, inhibiting miR-200 expression and directly suppressing E-cadherin.[10, 18]

SPRR2a functions as a ligand for Src homology 3 (SH3) domains,[3] which play important roles in intercellular communication and intracellular signal transduction where protein-protein interactions are mediated by noncatalytic conserved domains.[20] Each SH3 domain is a small, conserved sequence of about 60 amino acids that interacts with proteins containing specific amino acid motifs, the minimal being xPxxP.[20] SH3 domains often act as adapter molecules, recruiting other downstream proteins in signaling cascades. We hypothesized, therefore, that since SPRR2a acts as an SH3 domain ligand and induces BEC EMT it might mediate this effect by way of ZEB1-miR200c cycle and thus provide a point of therapeutic intervention.

Materials and Methods

SPRR2a Stable Transfectants

Maintenance of human intrahepatic cholangiocarcinoma cell lines (SG231, HuCCT-1) and methods to obtain stable transfectants were previously published.[3]


We used the C-terminal His-V5-tagged human SPRR2a expression vector previously described.[3] Human Halo-tagged CtBP and its enhanced green fluorescent protein (EGFP) expression control vectors were purchased from Genecopoeia (Germantown, MD) and the human Halo-tagged ZEB1 and ZEB2 vectors obtained from Promega (Madison, WI). The luc-3UTP-promoter constructs were kindly gifted by Dr. Tong Wu (Dept. Pathology, University of Pittsburgh School of Medicine). The miR-200 family promoter constructs were generated by polymerase chain reaction (PCR) according to previous reports,[17, 21] and subcloned into a pGL3 vector (Promega). The 5′ and 3′ SPRR2a and ZEB1 deletion constructs were generated by annealing synthesized oligonucleotides, whereas the xPxxP domain constructs were generated by PCR and subcloned into the pTracer (SPRR2a) and pCDNA3 (ZEB1) vectors. Mutations in the ZEB1-xPxxP motif were introduced into Halo-tagged ZEB1 construct using the Quikchange Lightning multisite-directed mutagenesis kit (Stratagene, La Jolla, CA). The primers and oligonucleotides used are listed in Supporting Table S1.

Biotinylated Oligonucleotide Precipitation Assays

The probes for DNA pull-down assays are listed in Fig. 5F. The assays were carried out as described[22] 24 hours after cell transfection. Oligo-specific bound proteins were collected and identified by western blotting.

Image Analysis

A tissue micro array (TMA) grid with 18 CC cases and four normal livers was stained using a modified method (Supporting Materials).[23] The TMA slide was imaged with a Mirax MIDI WSI scanner equipped with a Plan-Apochromat 40×/.95N.A. objective lens, AxioCam MRm digital CCD camera (Carl Zeiss, Jena, Germany) and specifically selected excitation/emission Qdot filters (Omega Optical, Brattleboro, VT) as described.[23] Tissue cytometric analysis was done using IAE-NearCYTE ( as described.[23] Briefly, cholangiocytes in regions of interest (ROIs) were segmented by Hoechst signal as objects. Objects were given a unique ID and classified into ZEB1 nucleus +/−, SPRR2 nucleus and cytoplasm +/−, and CK19 cytoplasm +/− with IAE-NearCYTE.


All statistical analyses were performed using SigmaStat software. P < 0.05 was considered statistically significant, and all tests were two-tailed. All interval values are expressed as mean ± SD. Normally distributed variables were compared using the unpaired Student t-test (two groups) and one-way analysis of variance (ANOVA) (three or more groups). The Mann-Whitney Rank Sum Test was used for comparison of the morphometric analysis of analyte expression (e.g., cytokeratin-19, ZEB-1, etc.). The complementary DNA (cDNA) microarray data have been deposited in GEO online (accession number GSE22560). See the Supporting Materials and Methods for additional information.


Expression of SPRR2a and mir-200 Family

We previously showed that forced SPRR2a expression in a cholangiocarcinoma cell line (SG231) down-regulates E-cadherin and up-regulates mesenchymal components, such as vimentin and fibronectin.[3] This suggests that SPRR2a is an upstream effector of EMT likely influencing the miR-200 family. In normal adult human tissues, real-time PCR shows high SPRR2a expression in organs normally containing squamous epithelia (cervix and esophagus), but low or absent expression in mesenchymal tissues such as heart, skeletal muscle, and adipose (Fig. 1A) and normal adult liver. Consistent with its role as a stress-responsive gene we showed that biliary tree injury and repair noncoordinately up-regulate BEC SPRR2a.[3, 7]

Figure 1.

Normal and diseased tissue distribution of SPRR2a and miR-200 family. (A) SPRR2a expression levels in adult human tissues (pools from three people) with real-time PCR. (B,C) Inverse expression pattern of SPRR2a and miR-200c in human BEC grown in complete serum free medium (C-SFM) ± IL-6 using real-time PCR. (B) Comparison of SPRR2a, miR200c, and ZEB1 expression in a bile duct ligation (BDL) injury model. Wild-type mice injected with saline (black; n = 3) were compared to IL-6−/− mice injected with IL-6 (dark gray; n = 3) or saline (light gray; n = 3) starting 6 weeks after BDL and continuing until 12 weeks post-BDL (Baseline = prior to BDL; wild-type n = 5; knockout n = 3). (C) Primary human BEC cultures were treated with (n = 4) or without (n = 4) IL-6 (100 ng/mL) in complete serum free medium (C-SFM) for 24 hours. (D) Mouse BECs were seeded (25, 50, 100, and 200 × 103 cells/well) and miR-141 transcripts evaluated by real-time PCR after 48 hours. (E) Normal and diseased liver samples obtained during transplantation surgeries were evaluated by real-time PCR for gene expression (normal: n = 8; PBC: n = 8; other diseased livers: n = 5; see Supporting Materials for additional information). PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; *P < 0.05; **P < 0.01; ***P < 0.001.

We also showed that BEC maintenance of biliary barrier integrity and wound-healing responses were dependent on (interleukin-6) IL-6/gp130/SPRR2a signaling (reviewed[24]). In a bile duct ligation (BDL) model, IL-6−/− mice did not up-regulate SPRR2a and showed inadequate biliary remodeling/repair following injury, implicating SPRR2a as a downstream effector of IL-6/STAT3 signaling.[7] As a preliminary screening investigation, we exploited the IL-6 induction of SPRR2a by using archived RNA from these mice[7] to study the relationship between SPRR2a and the miR-200 family in BEC.

Confirming our previous observations,[7] IL-6+/+ BECs express low but significantly higher levels of SPRR2a mRNA compared to IL6−/− mice, and this expression increased only in IL-6+/+ mice following BDL. Recombinant IL-6 injections recovered the SPRR2a response to BDL in the IL6−/− mice (Fig. 1B). An inverse relationship is seen between SPRR2a and miRNA-200c expression. IL6−/− mice (low SPRR2a) still have significantly higher expression of miRNA-200c even 6 weeks after BDL. Interestingly, all mice have very low baseline (prior to BDL) levels of ZEB1, which dramatically increased following BDL and/or IL-6 injection. IL-6 stimulation also increased SPRR2a and decreased miR-200 expression in primary cultured human BEC (Fig. 1C). We also confirmed in subconfluent BEC cultures where individual BEC display a “spreading” phenotype similar to cells at the leading edge of a wound that miR-200 family expression is lower (Fig. 1D). This observation agrees with our previous finding that confluent BEC monolayers down-regulate SPRR2a[3] as cell-to-cell junctions form.

Finally, to test whether similar trends occurred in human livers, we confirmed that diseased/injured livers have higher SPRR2a and lower miR-200c expression compared to normal livers, although strong statistical significance was not achieved in all sample categories (Fig. 1E), which is likely attributable to the endstage nature of tissue samples and the relatively low contribution of BEC RNA to total liver RNA. Regardless, these preliminary findings provide strong circumstantial evidence for interactions among SPRR2a, miR-200c, and ZEB1 during cellular remodeling and EMT in BEC, in vivo.

EMT in SPRR2a-Expressing Cells

Most aggressive cancers, including the cholangiocarcinoma lines SG231 and HuCCT-1, idiopathically down-regulate SPRR2a expression.[3] These cells lines, therefore, provide a suitable vehicle to study downstream pathways of SPRR2a by way of plasmid transfection. To identify genes commonly involved in SPRR2a-induced EMT, cDNA microarray analysis was done using SG231 and HuCCT-1 SPRR2a stable transfectants. EMT marker genes, such as desmoplakin (DSP), E-cadherin (CDH1), claudin-7 (CLDN7), laminin subunit beta-3 (LAMB3), and mucin-1 (MUC1) were down-regulated by SPRR2a induction, whereas ZEB1 was up-regulated when compared to the vector control cells (Fig. 2A). Specific RNA and protein expression levels of E-cadherin and ZEB1 were then verified (Fig. 2B,C). Although the cDNA microarray did not detect significant changes in ZEB2 expression, real-time PCR analysis showed ZEB2 significantly increased in both SPRR2a-expressing cell lines. In contrast, vimentin, a mesenchymal marker, is significantly up-regulated in HuCCT-1 but less so in SG231. SNAIL1, another important regulator of EMT,[25] did not appear to contribute to SPRR2a-induced EMT (Fig. 2C).

Figure 2.

Confirmation of EMT by SPRR2a induction. (A) cDNA microarray followed by hierarchical clustering was done on HuCCT-1 and SG231 stable transfectants. The complete dataset is available (GEO online: accession number GSE22560). EMT-associated genes whose expression changed in the same direction are tabulated beside the heat map using an average ratio. (B) Immunoblot showing loss of E-cadherin during SPRR2a expression—a hallmark of EMT. HuCCT-1, but not SG231, also shows changes in vimentin. (C) Real-time PCR of EMT-associated genes shows SPRR induced changes in E-cadherin, ZEB1, and ZEB2 for both cell lines. HuCCT-1 also shows significant changes in vimentin. (D,E) Immunofluorescence staining of control and SPRR2a-expressing cells for E-cadherin, vimentin, and DAPI (D) and for ZEB1 and DAPI (E) (scale bars = 25 μm). (F) Immunofluorescence shows that BEC nuclear ZEB1 staining (red) increases during liver inflammation. Human liver tissue with a ductular reaction shows BEC, marked by CK-19 staining (green), have more ZEB1 than BEC from normal uninflammed liver tissue. Nuclei stained with DAPI. Yellow arrows: high expression in BEC; White arrows: high expression in mesenchymal cells. *P < 0.05; **P < 0.01; ***P < 0.001.

Immunostaining HuCCT-1 for E-cadherin confirms the microarray/PCR data showing a loss of membranous E-cadherin (Fig. 2D), increased nuclear ZEB1 staining (Fig. 2E), and stronger vimentin staining of the cytoplasmic bundles in SPRR2a-expressing cells (Fig. 2D). In addition, cholangiolar BECs involved in ductular reactions in human livers showed higher levels of nuclear ZEB1 expression than normal noninflamed BEC lining the bile ducts (Fig. 2F). The intensity of ZEB-1 expression in cholangiolar BEC participating in ductular reactions, however, was noticeably lower than in nearby myofibroblasts (Fig. 2F), consistent with a change in epithelial morphology, but not a complete conversion to mesenchymal cells. These findings suggest a contributory role for ZEB1 in biliary remodeling, ductular reactions, and liver fibrogenesis.

Since ZEB1 expression is completely silenced in HuCCT-1 vector control cells and induced only upon SPRR2a transfection (Fig. 2C,E), we selected HuCCT-1 to study the involvement of miR-200 family with SPRR2a-induced EMT in BEC.

miR-200 Family in SPRR2a-Induced EMT

The miR-200 family is the result of two coding clusters: miR-200b/a/429 on chromosome 1 and miR-200c/141 on chromosome 12 (Fig. 3A). As recently reported,[16, 17] the promoter region for both coding clusters contains E-box and Z-box elements with conserved sequences for ZEB1 binding upstream of the transcription starting site. Real-time PCR showed that SPRR2a induction suppressed expression of primary transcripts for the miR-200 family, particularly miR-200c and miR-141 on chromosome 12 in both HuCCT-1 and SG231 (Fig. 3B; Supporting Fig. S1A-B). Expression of other miRNAs, such as miR-22, was not affected by SPRR2a induction (Fig. S1C).

Figure 3.

Involvement of miR-200 family in SPRR2a-induced EMT. (A) Primary microRNA structures of miR-200c/141 (left) and miR-200b/200a/429 (right) with mature miRNAs (gray), Z-box, E-box, transcription starting sites (TSS), and polyadenylation (Poly A) signals. The primer pairs that were used in (B); Fig. S1A,S1D are also indicated. (B) HuCCT-1 SPRR2a stable transfectants have reduced miR200 family expression. Real-time PCR was done to evaluate primary miRNA (pri-miR) transcripts for miR-200c/141 and miR-200b/200c/429, as well as precursor (pre-miR) and mature microRNA 200c and 200b. The primer pairs for pri-miR-200c/141 and pri-miR-200b/200c/429 were 4 and C (Fig. 3A), respectively. Complete results can be found in Fig. S1A. (C) Silencing of ZEB1 and ZEB2 in HuCCT-1 SPRR2a stable transfectants recovered miR200c expression, but only ZEB1 siRNA resulted in clear EMT changes: i.e., decreasing E-cadherin and increasing vimentin. (D) Microscopic observation of HuCCT-1 SPRR2a cells after ZEB1 or ZEB2 silencing verifies that ZEB1 silencing allows for recovery of an epithelial morphology (scale bars = 100 μm). *P < 0.05; **P < 0.01; ***P < 0.001.

The expression of some miRNAs is controlled posttranscriptionally through RNA editing or modulation of enzymes involved in miRNA maturation.[26] To determine whether SPRR2a-induced suppression of miRNA was the result of posttranscriptional modification, we compared the levels of primary miR-200 family gene transcripts (pri-200b/a/429 and pri-200c/141), to the processed miR-200 forms (precursor or mature) by real-time PCR (Figs. 3B; S1A). SPRR2a-induced suppression was equivalent in primary, precursor, and mature transcripts. Furthermore, SPRR2a-expressing cells strongly suppressed all phases of miR-200c and 141, whereas miR-200b, 200a, and 429 were only moderately suppressed (Figs. 3B; S1A). Using primer pairs for pri-miRNAs (Fig. S1D), we discovered: 1) transcription in SPRR2a-expressing cells is strongly suppressed compared to vector control cells, and 2) these primary transcripts are the same as previously reported[16] except for the far 3′ end of pri-miR-200c/141. This is probably due to a difference in cell types (biliary epithelium versus mammary gland).

Next we determined whether ZEB1 is a downstream component of SPRR2a and controls miR-200 family expression. Successful silencing of ZEB1 was accomplished by transfecting SPRR2a-expressing HuCCT-1 cells with target specific siRNA. Two weeks after ZEB1 siRNA transfection, SPRR2a-expressing cells recovered the expression of pre- and mature-miR-200c (Figs. 3C; S1E) as well as the other miR-200 family members (Fig. S1F). Concomitant with recovery of miR-200c expression, ZEB1 silencing in SPRR2a cells increased E-cadherin, decreased vimentin (Fig. 3C), and caused reversion to an epithelial morphology (Fig. 3D). All of these observations suggest that ZEB1 is a downstream effector of SPRR2a during EMT. Transfection of siZEB2 into SPRR2a-expressing cells was less effective at reversing SPRR2a-induced EMT (Fig. 3C,D).

SPRR2a Is a Functional Partner of ZEB1

The above findings suggested that miR-200 family genes might be transcriptionally regulated by SPRR2a through suppression in the miR-200 promoter region. DNA analysis showed HuCCT-1 does not contain any promoter region mutations of either miR-200c/141 or miR-200b/a/429, and have intact Z-box and E-box elements (data not shown). Also, luciferase reporters showed functional suppression of promoter Z-box/E-box regions in SPRR2a-expressing cells between nucleotides −126 to −17 on miR-200c/141 and −321 to −64 on miR-200b/a/429 (Fig. 4A). These effects were not observed with a control vector that lacks the Z-box/E-box sequence (data not shown). Although SPRR2a alone suppresses the miR-200c/141 promoter, there is augmentation of suppression when SPRR2a is combined with ZEB1 (Fig. 4B). Moreover, the synergic effect of SPRR2a on ZEB1 inhibition of promoter activity can be attenuated by siZEB1 or siZEB2 molecules (Fig. 4C), indicating that promoter suppression by SPRR2a is ZEB1-dependent. Therefore, ZEB1 is essential for suppression of the miR-200 promoter and the same region is likely targeted by SPRR2a.

Figure 4.

SPRR2a is a functional partner of ZEB1 and CtBP that suppress miR-200 family transcription. (A) Structure of luciferase reporter construct for 5′ deletions of miR-200c/141 (Top) and miR-200b/200a/429 (Bottom) promoters with E-box, Z-box, and TSS. Luciferase activities and the ratio (Vector control/SPRR2a) are tabulated. (B) Cells were transiently transfected with both the miR-200c/141 reporter (−126/152-luc) and increasing amounts of SPRR2a plasmid (HuCCT-1: 50, 100, 200 ng/well; SG231: 50, 400 ng/well) followed by transfection of ZEB1 plasmid (black bar) or the control plasmid (open bar). Luciferase activity was measured 24 hours later. (C) Cells were transiently transfected with ZEB1 or ZEB2 siRNA at the time of cell seeding. Forty-eight hours later, the cells were cotransfected with the miR-200c/141 reporter (−126/152-luc) and increasing amounts of SPRR2a plasmid (described in (B)). Luciferase activity was measured 24 hours later. (D) HuCCT-1 vector or SPRR2a cells were transfected with ZEB1 (0.1 μg/well), SPRR2a (0.2 μg/well), and CtBP (0.2 μg/well) expression vectors, alone and in combination. Open bar: untreated cells. *P < 0.05; **P < 0.01; ***P < 0.001.

Figure 5.

Nuclear SPRR2a acts as a component of the transcriptional suppressing complex for miR-200c/141. (A) Immunoblot analysis of cytoplasmic and nuclear fractions from cells transiently cotransfected with SPRR2a and ZEB1 vectors show ZEB1 is nuclear and that SPRR2a is present in both the cytoplasm and nucleus. (B,C) HuCCT-1 (B) or SG231 (C) cells were transiently transfected with Halo-tagged ZEB1 with or without cotransfection of V5-tagged SPRR2a. SPRR2a/ZEB1 interactions were verified through immunoprecipitation using a magnetic bead for the designated tag. (D) Immunoprecipitation analysis of SPRR2a and endogenous ZEB1 in SPRR2a HuCCT-1 stable transfectant. NC; nonimmune mouse IgG control. (E) Probes for the DNA pull-down assay with E-box and Z-box elements in the promoter of miR-200c/141. The mutations are indicated with colored and underlined letters. (F) DNA-pull-down assay followed by ZEB1 immunoblotting with HuCCT1-SPRR2a or control cells. (G) HuCCT-1 cells were transfected with indicated constructs. After 24 hours, cell lysates were subjected to DNA-pull-down assay.

ZEB1 transcriptional suppression is mediated through the molecule's repressor domain that contains several binding motifs. Among these are zinc finger domains that bind to DNA E-box elements and a binding site for the corepressor C-terminal binding protein (CtBP).[27] Stably transfected HuCCT-1 cell lines (SPRR2a and vector control) were transiently transfected with ZEB1, SPRR2a, or CtBP plasmids to determine the individual dose that would induce an appropriate suppression of the miR-200c/141 reporter. Using these dosages, cotransfection with various plasmid combinations was conducted and the miR-200c/141 reporter response measured (Fig. 4D). SPRR2a and/or CtBP plasmids significantly potentiated the ZEB1-dependent suppression of the miR-200c/141 promoter, and maximum effects were conferred with all three vectors (Fig. 4D). CtBP/SPRR2a cotransfection did not suppress the miR-200c/141 promoter as much as ZEB1/SPRR2a or ZEB1/CtBP, again illustrating that enhanced promoter suppression for SPRR2a and CtBP is ZEB1-dependent.

SPRR2a Is a Cosuppressor of ZEB1 for miR-200c/141

The above results indicate that SPRR2a is a functional partner of ZEB1 and CtBP for miR-200/141 promoter suppression. In agreement with previous reports,[28, 29] SPRR2a protein is distributed in both the cytoplasm (perinuclear) and nucleus (Fig. 5A), and therefore, is in position to interact with ZEB1, in vivo. ZEB1 also coimmunoprecipitated with endogenous and exogenously expressed SPRR2a (Fig. 5B-D).

We next constructed biotinylated double-stranded oligonucleotide probes, which mimic wild-type or mutational sequences in the E-box and Z-box binding sites for ZEB1, to determine whether SPRR2a acted directly to suppress the miR-200/141 promoter or as part of a suppressing complex (Fig. 5F). Cell lysates were incubated with the biotinylated probes and resulting protein-DNA complexes precipitated with streptavidin beads that were then visualized by western blotting. Lysates from HuCCT1-SPRR2a cells show ZEB1 binding to the wild-type promoter with intact E-box/Z-box sequences, while vector control cells show no ZEB1 binding, indicating the importance of ZEB1 binding during SPRR2a-induced EMT (Fig. 5G). Moreover, robust binding of SPRR2a, ZEB1, and CtBP to the miR141/200c-promoter was observed in extracts from cells overexpressing all three proteins (Fig. 5H). Oligonucleotides of miR-141/200c with Z-box and E-box mutations did not precipitate this complex. These results suggest that SPRR2a complexes with ZEB1 and CtBP to suppress the miR-200c/141 promoter in HuCCT-1 cells.

Mapping of the Interacting Regions

Using Flag epitope-tagged human ZEB1 constructs (Fig. 6A) in transfected HuCCT-1-SPRR2a cells, we determined that native ZEB1 and its N-terminal to Homeodomain (5′T-HD) construct contained the SPRR2a-interacting region (Fig. 6B). Mapping shows the interacting region resides between the Smad-interacting domain (SID) and HD on ZEB1 (Fig. 6A). Only the 5′T-HD construct retained the ability to enhance suppression of the miR-200c/141 response concurrent with SPRR2a (Fig. 6C). Furthermore, deletion constructs of SPRR2a (Fig. 6D) demonstrated that the xPxxP-containing region binds ZEB1 (Fig. 6E) and is required to suppress miR-200c/141 promoter activity (Fig. 6F). Interestingly, the xPxxP construct is more suppressive than the full-length SPRR2a in this assay. The ZEB1 binding region may be occluded in the basal state, but require an activation signal to become amenable to the full-length SPRR2a.

Figure 6.

Mapping of the interacting regions. (A) ZEB1 and its deletion mutants with a putative SH3 domain ligand (xPxxP) motif are shown schematically. Constructs that elicit synergic suppression with SPRR2a on the miR-200c/141 promoter and immunoprecipitate with SPRR2a are summarized. (B) FLAG-ZEB1 constructs were transiently transfected into HuCCT-1-SPRR2a cells. At 24 hours, cells were subjected to immunoprecipitation analysis. (C) HuCCT-1 cells were transfected with the miR-200c/141 reporter together with the deletion mutants alone (open bar) or deletion mutants and SPRR2a (black bar), and then luciferase assays were performed. (D) Protein sequence of SPRR2a and the deletion mutant constructs are shown schematically. The SH3 domain ligand regions (xPxxP) are indicated. (E) HuCCT-1 was transfected with the full-length or xPxxP domain of SPRR2a, and was subjected to immunoprecipitation analysis after 24 hours. NC, untransfected control. (F) HuCCT-1 cells were transfected with the miR-200c/141 reporter (−126/152-luc) together with the deletion mutants shown in (D). Luciferase activity was measured 24 hours later (N; untransfected control).

Unexpectedly, the region of ZEB1 required for the interaction (between SID-HD) lacks recognizable motifs except for a putative SH3 domain ligand motif PTQPPPLP (555-562) (Figs. 6A; S2A). Similar motifs are also found in mouse and rat sequences (Fig. S2B and Supporting Discussion). Introduction of a mutation (xPxxP mut) in the motif disrupted the interaction between SPRR2a and ZEB1 (Fig. 7A), indicating that PTQPPPLP is required. SPRR2a does not have an SH3 domain to bind PTQPPPLP, and both ZEB1 and SPRR2a contain putative SH3 domain ligands. We hypothesized, therefore, that the ZEB1/SPRR2a interaction involved some other SH3 domain containing protein (-s) that would translocate to the nucleus as a complex upon SPRR2a induction.

Figure 7.

BEC uses SH3 domain networks in the SPRR2a/GRB2/ZEB1 pathway. (A) SPRR2a cells were transfected with native or an xPxxP mutated Halo-ZEB1 and immunoprecipitation analyses were performed. Immunoblots verify the presence of a SPRR2a/GRB2/ZEB1 complex that requires the PTQPPPLP ZEB1 motif; a SPRR2a/GRB2 interaction; and a ZEB1/GRB2 interaction. (B) HuCCT-1 cells were transfected with SPRR2a and the nuclear and cytoplasmic proteins extracted separately, as verified by SAM68. Cell lysates were directly immunoblotted for the SH3-domain proteins GRB2 and c-Yes1. For pTyr GRB2, cell lysates were immunoprecipitated with GRB2 antibody followed by immunoblottting with anti-pTyr antibody. SPRR2a causes dephopshorylation and nuclear relocalization of cytoplasmic GRB2. (C) SPRR2a plasmids were transiently transfected into HuCCT-1 cells, and cells subjected to immunoprecipitation analysis. SPRR2a promotes the interaction of ZEB1/GRB2.

Testing the cellular distribution of proteins containing one or more SH3 domains in the presence and absence of SPRR2a showed that GRB2, an adaptor protein involved in RTK signal transduction,[30, 31] translocated from the cytoplasm to the nucleus following SPRR2a induction. SPRR2a expression also resulted in dephosphorylation of cytoplasmic GRB2 (Fig. 7B).

Further experimentation revealed that SPRR2a promoted the interaction between ZEB1 and GRB2 (Fig. 7C) and the xPxxP mutation in ZEB1 disrupted the ZEB1/GRB2 interaction, while the SPRR2a/GRB2 association remained unaffected (Fig. 7A), suggesting that SPRR2a complexes with GRB2, promoting nuclear translocation by blocking tyrosine phosphorylation. GRB2/ZEB1 binding then occurs through the PTQPPPLP motif. Collectively, SPRR2a partners with ZEB1 through SH3 domain binding, resulting in the modulation of miR-200c/141 transcription and EMT in the cholangiocarcinoma.

CK19 and ZEB1 in Intrahepatic Cholangiocarcinoma Cells Versus Nonneoplastic BEC

Previous reports show that ZEB1 correlates with the differentiation status of intrahepatic CC, reduced CK19 expression,[32] and cells undergoing EMT in hepatocellular cancers,[10] gallbladder cancers,[33] and a cholangiocarcinoma cell line.[32] We previously showed that forced SPRR2a expression decreased CK19 expression[3] and confirmed that BEC SPRR2a up-regulation is associated with EMT during various chronic liver diseases[3] (Fig. 1E) and with nuclear ZEB-1 expression in cholangioles (Fig. 2F).

Using a tissue array consisting of tissue plugs from human intrahepatic cholangiocarcinomas (n = 18: well differentiated = 7, moderate = 5, and poor = 6) and four normal liver controls, we compared CK19 and nuclear ZEB-1 protein expression in malignant CC cells versus nonneoplastic BEC. One ROI from a representative staining area for each plug was subjected to automated image analysis. In agreement with previous laser capture proteomic studies,[34] CK19 protein expression was significantly lower in CC cells (total 2,998 cells, 28.5 ± 31.9, range 0-239.8) than in normal intrahepatic cholangiocytes (Fig. 8; total 51 cells, 49.3 ± 22.8, range 8.4-111.8; P ≤ 0.001). Conversely, nuclear CC ZEB1 intensity (30.1 ± 26.5; range 0-225.7) was significantly higher than seen in normal nonneoplastic cholangiocytes (18.8 ± 14.6; range 0-72.9; P < 0.001).

Figure 8.

Increasing ZEB1 correlates with loss of BEC differentiation marker in cholangiocarcinoma. (A) Hematoxylin and eosin (H&E) stain of CC cells. (B) cytokeratin-19 (CK19; red) expression in the same CC area as shown in (A). (C) CK19 expression intensity was measured by IAE-NearCyte tissue cytometry after nuclear segmentation. The threshold for CK19-positive staining is verified by a mask overlay and then set to designate the positive staining areas (upper panel). Then the CK19 expression intensity is determined in a predetermined area around each nucleus, measured, and recorded by the software for each individual CC or BEC. (D) Results showed that invading CC cells showed lower CK19 expression than normal-appearing nonneoplastic BEC, as expected, confirming previous proteomic studies using a more time consuming laser microdissection.[34]


We show that SPRR2a, a STAT3-dependent stress-responsive gene involved in epithelial barrier maintenance and repair, is also a novel transcriptional cosuppressor that complexes with ZEB1 and GRB2 by way of SH3 domains to enhance silencing of miR-200c/141 gene expression thereby promoting EMT (Fig. S3). We focused on miR-200c/141 as a ZEB1 target because miR-200c/141 plays a pivotal role in maintaining an epithelial phenotype and suppressing EMT.[10, 18, 35] Specifically: 1) SPRR2a enhances suppression of miR-200c/141 expression; 2) ZEB1, SPRR2a, CtBP, and GRB2 form a complex on the promoter region of miR-200c/141; 3) xPxxP SH3 ligands in both ZEB1 and SPRR2a are critical for ZEB1/GRB2 and SPRR2/GRB2 binding; 4) SPRR2a facilitates nuclear translocation of GRB2 after binding; and 5) these signaling pathways participate in types 2 and 3 EMT responses in human livers, and probably other barrier epithelia.

GRB2, SPRR2a, and ZEB1 proteins undergo nuclear translocation to suppress miRNA transcription, but the molecular mechanism of nuclear translocation needs to be determined (Fig. S3). SPRR2a-induced nuclear translocation was also detected with another SH3 domain-expressing BEC protein, c-Yes1 (Fig. 7B) in HuCCT-1 cells. GRB2, usually a cytoplasmic protein, has one SH2 domain and two SH3 domains enabling it to complex with other proteins.[30] However, similar to our observations after SPRR2a induction, others showed that GRB2 undergoes nuclear translocation when complexed with hnRNP.[36] SPRR2a binds to a GRB2 SH3 domain causing GRB2 activation by way of dephosphorylation and subsequent nuclear translocation (Fig. 7B). Dephosphorylation as the activated state of GRB2 is supported by the observation that phosphorylated GRB2 down-regulates tyrosine kinase signaling[30, 31] (Fig. S3).

SPRRs are among the most highly up-regulated genes during wound healing and inflammation-mediated remodeling of mucosal epithelia in a broad range of tissues, including BEC.[3] Mice genetically predisposed to “fast” wound healing show SPRR2a to be the most differentially up-regulated gene compared to “slow” wound healers.[4] Based on SPRR2a's ability to bind SH3 domains and change the phosphorylation status of a variety of proteins[3] (Fig. 7B), we speculate that SPRR2a affects signaling pathways by acting as a “general activator” of its SH3 domain-containing protein binding partners. The SH3 ligand-dependent interaction between SPRR2a/GRB2 and ZEB1/GRB2 results in the formation of a suppressive complex on the promoter of miR-200c/141 substantiates this contention.

Aberrantly expressed miRNAs and type 3 EMT in cholangiocarcinoma contribute to proliferation, invasiveness, and chemoresistance[10, 18, 35] and the miR-200 family are up-regulated or down-regulated in various human cancers.[35] Interestingly, in the liver,[37] miR-200 is down-regulated in comparison to normal liver tissue with benign lesions, but up-regulated in advanced hepatocellular carcinoma, indicating stage-specific regulation of miR-200. Early, when cancer cells acquire invasive behavior (EMT), miR-200 might be down-regulated, only to be up-regulated again during the re-epithelialization of distal metastases when cells undergo MET,[35] consistent with recent results in our experimental animal model.[15]

SH3 domains are required in many key signaling pathways.[20] For scientific consistency, we addressed SPRR2a only in biliary tract cells, in vivo and in vitro, but the phenomena seen here likely occurs in the other cell types. Since ZEB1 suppression can target many genes, SPRR2a could aid transcriptional regulation of other mRNAs and miRNAs. More important, SH3 domain interactions provide a novel molecular mechanism by which SPRR2a could differentially regulate and “finely tune” responses to stress, as speculated nearly a decade ago.[1] Understanding how SPRR2a is involved in these pathways might yield potential therapeutic targets. For example, STAT3-dependent BEC SPRR2a expression might alter biliary barrier function or augment BEC wound healing in liver allografts or prevent CC metastasis,[15] whereas blocking SPRR2a expression might limit local CC invasiveness.

Author Contributions

S.S. generated stable cell lines and performed cDNA microarray; J.L. III performed and collected samples from mouse experiments in Fig. 1; K.I. and N.C. performed immunofluorescent staining in Fig. 2; Y.M. performed all experiments and analysis; A.D. and T.T. supervised the project; the article was written by Y.M and edited by A.D.