p63-Dependent Dickkopf3 Expression Promotes Esophageal Cancer Cell Proliferation via CKAP4

There are four Dickkopf (DKK) family members in vertebrates, including DKK1, DKK2, DKK3, and DKK4, all of which are secretory proteins and contain two cysteine-rich domains (CRD1 and CRD2; ref. 1). DKK1 was originally identified as a head inducer in Xenopus embryos and the most extensively studied among DKK family proteins. DKK1 antagonizes β-catenin–dependent Wnt signaling by binding to and internalizing low-density lipoprotein receptor–related protein (LRP) 5 or 6, which are Wnt coreceptors (1–4). The Dkk1-null mutation is embryonic lethal in mouse as DKK1 plays important roles in several developmental processes, including fetal anterior–proximal axial patterning and limb formation (5) as well as in postnatal stages, such as bone formation (6). DKK2 and DKK4 also bind to LRP6 via CRD2 and inhibit Wnt signaling, similar to DKK1 (2, 7). Dkk2-null–mutant mice were viable but blind due to a complete transformation of the cornea epithelium into the stratified epithelium (8). No information regarding Dkk4 knockout mice are available at present.

DKK3 exhibits structural divergence from the rest of the DKK family (1). Overall protein sequence homology between DKK1, DKK2, and DKK4 ranges around 50%, but that between DKK3 and other DKKs is less than 40% (9). Two CRDs are separated by a nonconserved linker region that spans 50–55 aa in DKK1, DKK2, and DKK4, but only 12 aa in DKK3. In addition, the soggy domain is found only in DKK3, but not in other DKK proteins. DKK3 neither interacts with LRP6 nor antagonizes Wnt signaling unlike DKK1, DKK2, and DKK4 (1, 2, 10). Therefore, DKK3 is a divergent member of the DKK family and possesses functions independent of Wnt signaling (11). Importantly, no corresponding cell surface receptor for DKK3 has been identified to date.

Dkk3 knockout mice exhibit no obvious phenotype during the developmental stages, but do differ in hematologic and immunologic parameters as well as pulmonary ventilation (12). DKK3/REIC has also been shown to exhibit reduced expression gene in human immortalized cells (13) and its expression is frequently suppressed by promoter hypermethylation in human cancer cells, including the highly aggressive basal breast cancer (14), non–small cell lung cancer (15), hepatocellular carcinoma (16), gastric cancers, and colon cancers (17). DKK3 consistently suppressed cancer cell proliferation when ectopically overexpressed in various cancer cell types (14, 18). In this context, DKK3 seems to function as a tumor-suppressor. In contrast, it has also been reported that DKK3 may have tumor-promoting functions. For instance, DKK3 was overexpressed in esophageal adenocarcinoma and oral squamous cell carcinoma (SCC) tissues, promoting cancer cell proliferation and migration (18, 19). DKK3 also induced stromal proliferation and differentiation in prostate cancer, influencing the angiogenesis program (20). Thus, the role of DKK3 in cancer development remains to be clarified.

Esophageal cancer is the sixth leading cause of cancer-related death worldwide (21, 22). There are two histologic types of esophageal cancer: SCC and adenocarcinoma (23). Esophageal SCC (ESCC) is believed to be affected by environmental factors, including alcohol and tobacco as well as genetic factors, such as a somatic mutation in p53 or overexpression of EGFR (21, 23). Recent genomic analysis has revealed that p63, a p53-related transcriptional factor, is a major oncogenic protein in esophageal cancer; the gene locus is frequently amplified in ESCC and its expression in ESCC is significantly higher than nontumor and adenocarcinoma tissues (24). Because of alternative splicing, there are at least six distinct p63 variants with two different N-termini (TA or ΔN) and three different C-termini (α, β, or γ; ref. 25). ΔNp63 and TAp63 show very different expression patterns, depending on the source of cell lines and tissues (26). ΔNp63α is the main p63 isoform expressed in ESCC and is required for ESCC cell proliferation (27), but the relationship between p63 and DKK3 remains to be clarified.

We recently found that cytoskeleton-associated protein 4 (CKAP4) is a novel DKK1 receptor and that simultaneous expression of DKK1 and CKAP4 is negatively correlated with prognosis in pancreatic, lung, and esophageal cancers (28, 29). Here we show that the DKK3–CKAP4 and the DKK1–CKAP4 signaling axes are activated in distinct populations of ESCC tumors and that ΔNp63α induces the expression of DKK3 in cells with mutations of Kras and p53.

Madin-Darby canine kidney (MDCK) cells were provided by Dr. S. Tsukita (Osaka University, Osaka, Japan). T24 bladder cancer cells and U-251 MG were purchased from the National Institutes of Biomedical Innovation, Health, and Nutrition (Osaka, Japan). TE-1, TE-4, TE-5, TE-6, TE-8, TE-9, TE-10, TE-11, and TE-14 ESCC cells were obtained from the Riken Bioresource Center Cell Bank in November 2008 (TE-5 and TE-11), May 2009 (TE-6, TE-9, and TE-14), or January 2015 (TE-1, TE-4, TE-8, and TE-10). KYSE-410 and KYSE960 cells were obtained from the Japanese Cancer Research Resources Bank in May 2015. SW480 and DLD-1 colorectal cancer cells and TMK-1, KKLS, MKN1, and MKN45 gastric cancer cells were provided by Dr. W. Yasui (Hiroshima University, Hiroshima, Japan) in September 1997 (SW480 and DLD-1), September 2006 (TMK-1), August 2006 (KKLS), February 2006 (MKN1), or September 2005 (MKN45). HCT116 and Caco-2 colorectal cancer cells were provided by Dr. T. Kobayashi (Hiroshima University, Hiroshima, Japan) in November 2003 and were purchased from RIKEN Bioresource Center Cell Bank in April 2013, respectively. AGS gastric cancer cells were provided by Dr. M. Hatakeyama (Tokyo University, Tokyo, Japan) in April 2014. HepG2, HLE, and HLF hepatic cancer cells were purchased from the ATCC in July 2017 (HepG2) and the Japanese Collection of Research Bioresources in June 2015 (HLE and HLF), respectively. A549 and Calu-6 lung adenocarcinoma cells were provided by Dr. Y. Shintani (Osaka University, Suita, Japan) in January 2014 and Shionogi Pharmaceutical Research, respectively.

S2-CP8, SUIT-2, and PANC-1 pancreatic cancer cells were purchased from the Cell Resource Center for Biomedical Research, Institute of Development, Aging, and Cancer of Tohoku University in April 2014 (S2-CP8 and SUIT-2) and the Riken Bioresource Center Cell Bank in October 2014 (PANC-1), respectively. Lenti-X 293T (X293T) cells were purchased from Takara Bio Inc. in October 2011. TE-5, TE-6, TE-9, TE-11, and TE-14 cells were authenticated in February 2017 using short tandem repeat analysis. Mycoplasma testing was not conducted.

Additional methods and antibodies, siRNAs, and shRNA used in this study are described in Supplementary Methods and Supplementary Table S1.

Polarized secretion assays of DKK molecules were performed as described previously (30). MDCK cells expressing DKK-FLAG (2 × 10⁵ cells) were seeded on Transwell polycarbonate filters (Corning Costar). DKKs secreted into the culture medium from the apical and basolateral chambers were collected and probed with anti-FLAG antibody. For DKK4, the same conditioned medium was precipitated with anti-FLAG antibody and protein G beads. After washing three times with NP40 buffer (20 mmol/L Tris-HCl pH 8.0, 10% glycerol, 137 mmol/L NaCl, and 1% NP40), the precipitates were probed with anti-FLAG antibody.

Proliferation assays were performed as described previously (28, 31) with modification. MDCK cells (2 × 10⁵ cells) were grown and polarized on Transwell polycarbonate filters (Corning Costar) in growth medium and subsequently starved by changing to DMEM containing 0.1% BSA. Then, DKK-FLAG conditioned medium was added to MDCK cells apically or basolaterally.

Single guide RNA oligo against canine CKAP4 (GTCGGGCGGCGCGGATGAC) was inserted into pX330 (Addgene plasmid #42230) that expresses hCas9. The cloned plasmids were introduced into MDCK cells using Viafect reagent (Promega) with blasticidin-resistant gene expressing plasmids. After transfection, MDCK cells were replated at low density in growth medium containing blasticidin (10 μg/mL); cells were allowed to grow until single-cell colonies became visible. Then, single colonies were picked mechanically, amplified, and analyzed.

MDCK cells (1 × 10⁵ cells) were seeded into 3.5-cm dishes and cultured in DMEM supplemented with 1% FBS. TE-11 and KYSE960 cells (1 × 10⁵ cells) were seeded in a 3.5-cm dish and cultured in growth medium. Cell numbers were counted on the indicated days.

Three-dimensional cell proliferation assays were performed as described previously (28, 31) with modification. Forty microliters of Matrigel (BD Biosciences) were solidified on a round coverslip by incubating for 30 minutes at 37°C. MDCK cells (5 × 10⁴ cells) suspended in DMEM containing 2% Matrigel and 10% FBS were seeded on the solidified Matrigel and cultured for the indicated days.

Detection of cell surface CKAP4 was performed as described previously (28). MDCK cells treated with DKK-FLAG, HCT116, TE-11, and KYSE960 cells were incubated with 0.5 mg/mL sulfo-NHS-LC-biotin (Pierce Biotechnology) for 30 minutes at 4°C. Biotinylated cells were lysed in 500 μL of NP40 buffer with protease inhibitors. Lysates containing biotinylated proteins were precipitated using NeutrAvidin agarose beads (Pierce Biotechnology). Then, beads were washed twice with NP40 buffer and once with 10 mmol/L Tris-HCl pH 7.5; the bound complexes were probed with anti-CKAP4 antibody.

Patients with ESCC (n = 72) who underwent surgery at Osaka University Hospital (Osaka, Japan) from January 2006 to May 2012 were examined in this study. All experimental protocols were approved by the Ethical Review Board of the Graduate School of Medicine, Osaka University (Osaka, Japan; no. 13455) under Declaration of Helsinki, and were performed in accordance with the Committee guidelines and regulations. The written informed consent was obtained from all patients. Among them, 48 patients overlapped with patients used in our previous study (29), and were reexamined in this study.

All tissue sections (5 μm thickness) were stained using a DakoRealTMEnVision Detection System (Dako) in accordance with the manufacturer's recommendations. After deparaffinization and heat-induced antigen retrieval, endogenous peroxidase activity was quenched with peroxidase-blocking solution (Dako) and the sections were blocked with 5% goat serum in buffer (5% BSA, 0.5% Triton X-100 in PBS) for 16 hours at 4°C. Then, the blocked specimens were stained with anti-DKK1 (1:100), anti-CKAP4 (1:100), or anti-DKK3 (1:50) antibody for 16 hours at 4°C. After washing, the specimens were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies for 1 hour and subsequently detected with diaminobenzidine (DAB; Dako). All tissue sections were also counterstained with 0.1% (w/v) hematoxylin. For DKK1 and CKAP4, tumors in which the positively stained area covered >5% were classified as positive cases. Because DKK3 was expressed specifically in the marginal regions of tumor nests near stroma, we regarded tumor areas in which marginal regions were positively stained (> 5%) as DKK3-positive tumor nests. Four researchers including a pathologist evaluated the staining.

The xenograft tumor formation assay was performed as described previously (28, 31) with modification. KYSE960 cells (1 × 10⁷ cells) were transplanted using dorsal subcutaneous transplantation of 6-week-old male BALB/cAnNCrj-nu immunodeficient mice (Charles River Laboratory Japan Inc.). Control cells (KYSE960 stably expressing control shRNA) were transplanted on the left dorsal region while the other cells (KYSE960 stably expressing DKK3 shRNA and CKAP4 shRNA) were on transplanted on the right dorsal region. The mice were sacrificed at 40 days after transplantation and mice with a control tumor size of 100 mm³ or less at 18 days after transplantation were excluded from the study.

To examine antitumor effects of the anti-CKAP4 antibody, mice were divided into anti-CKAP4 antibody or control IgG administration groups when the tumor size reached 100 mm³. Two hundred and fifty μg antibody was intraperitoneally administrated twice a week. The mice were sacrificed at 22 days after antibody injection. Tumor volumes were calculated using the following formula: (major axis) × (minor axis) × (minor axis) × 0.5.

X293T cells were transfected with pGL4.27 (Promega)-containing wild-type (WT) or mutated (Mut) genomic regions of −8466 to −8118 from transcription start site (TSS) of DKK3 and the indicated constructs. The cells were lysed at 24 hours posttransfection, and the luciferase activity was measured with PicaGene reagent (Toyo Ink) as described previously (32). β-Galactosidase activities were measured to standardize the transfection efficiency.

Esophageal organoid culture was performed as described previously (33) with modification. The esophagus mucosa from wild-type or KP (KrasG12D^LSL from National Cancer Institute–Frederick, Bethesda, MD and p53^lox/lox from The Jackson Laboratory) adult mouse was physically separated and minced into small pieces. Cells were dissociated in DMEM/F12 medium containing 0.25% Trypsin-EDTA and 100 U DNase I (Sigma, D4527) for 30 minutes at 37°C with agitation. After passing the cells through a 40-μm sterile filter (BD Biosciences, 352340), 5,000 esophageal epithelial cells were suspended in ice-cold 20 μL Matrigel solution (BD Biosciences). Matrigel-containing esophageal epithelial cells were placed as a droplet in a 24-well tissue culture plate and solidified by incubating for 20 minutes at 37°C. Complete growth medium was prepared at the time of use and consisted of 1:1 mixture of DMEM and F12, 1 × N-2 (Life Technologies, 17502048), 1 × B-27 Supplements (Life Technologies, 17504044), 1 × Glutamax (Life Technologies), 10 mmol/L HEPES/NaOH (pH 7.4), 50 ng/mL EGF (R&D Systems, 236-EG), 100 ng/mL Noggin (R&D Systems, 1967-NG), 100 ng/mL R-Spondin 1 (R&D Systems, 7150-RS), 100 ng/mL FGF10 (PeproTech, 100–26), and 10 μmol/L Y27632 (WAKO, 253-00513).

Experiments were performed repeatedly at least three times and results are presented as the mean ± SD. A paired or unpaired Student t test with a P < 0.05 and Mann–Whitney U test with P < 0.001 was used to determine statistical significance.

The protocol for utilization of human specimens was approved by the ethical review board of the Graduate School of Medicine, Osaka University (Osaka, Japan; no. 13455). All protocols used for animal experiments in this study were approved by the Animal Research Committee of Osaka University (Osaka, Japan; no. 21-048-1).

To examine whether DKK family proteins other than DKK1 act on CKAP4, FLAG-tagged DKK family proteins were stably expressed in MDCK cells (Fig. 1A). When MDCK cells were cultured on a membrane filter to allow them to be polarized two dimensionally, DKK2-FLAG, DKK3-FLAG, and DKK4-FLAG were primarily secreted apically as efficiently as DKK1 (Fig. 1A; ref. 28). CKAP4 was expressed on the apical cell surface of MDCK cells (28). The addition of DKK proteins to the apical side (Ap), but not the basolateral side (Bl), promoted cell proliferation as assessed using Ki-67 positivity (Fig. 1B). Like DKK1, DKK2, DKK3, and DKK4 also bound to CKAP4 (Fig. 1C) and induced the internalization of CKAP4 from the cell surface membrane (Fig. 1D). DKK(s)-induced cellular proliferation was inhibited in CKAP4 knockout (KO) MDCK cells, which were generated by using CRISPR/Cas9-based genome editing (Fig. 1E and F). In three-dimensional culture conditions using Matrigel, MDCK cells formed cysts (34). The size of MDCK cysts increased and Ki-67–positive cells were simultaneously increased by the expression of DKKs (Fig. 1G). CRD1 was required for the binding of DKK1 to CKAP4 (28). CRD1 deletion mutants (ΔCRD1) of DKK2, DKK3, and DKK4 did not increase the size of the cyst or increase cellular proliferation as assessed using Ki-67 (Fig. 1G). These results suggest that all DKK family proteins promote MDCK cell proliferation through CKAP4, suggesting that CKAP4 is a common receptor of DKK family proteins.

The protein expression levels of DKK2, DKK3, and DKK4 in cancer cells was examined. DKK3 was highly expressed in KYSE960 and TE-11 esophageal, T24 bladder, U-251 glioblastoma, PANC-1 pancreatic, and Caco-2 colon cancer cells; moderately detected KKLS gastric, HLE liver, S2-CP8 and SUIT-2 pancreatic, SW480 colon, and Calu-6 and A549 lung cancer cells (Fig. 2A); and hardly detected endogenously in other cell lines, including TMK1, AGS, MKN1, and MKN45 gastric, HLF and HepG2 liver, and DLD-1 and HCT116 colon cancer cells in addition to X293T and MDCK nontumor cells (Fig. 2A). Transcriptome data from matched tumor and nontumor tissues (12 paired cases) in The Cancer Genome Atlas (TCGA) showed that DKK3 is expressed in ESCC tumor lesions more significantly than in nontumor tissue (Fig. 2B). The TCGA dataset also revealed that DKK3 is expressed highly in ESCC (86 cases) rather than esophageal adenocarcinoma (85 cases) and nontumor esophageal tissues (12 cases; Fig. 2B).

Because the antibodies for DKK2 and DKK4 were not available for Western blotting, a public database analysis was performed to test their expression levels. When compared with DKK1 and DKK3, the mRNA levels of DKK2 and DKK4 were quite low in cultured cell lines from various tissues (Supplementary Fig. S1; http://medical-genome.kribb.re.kr/GENT/). In addition, DKK2 and DKK4 were barely detectable in various tissues and cancer cell lines (RefExA, http://157.82.78.238/refexa/main_search.jsp). Therefore, the role of DKK3 in ESCC was further investigated.

Among ESCC cell lines, DKK3 was highly expressed in TE-6, TE-11, and KYSE960 cells; moderately expressed in TE-8, TE-9, and TE-10 cells; and little expressed in TE-1, TE-4, TE-5, TE-14, and KYSE410 cells (Fig. 2C). In ESCC cases, DKK3 and CKAP4 were detected in 37 of 72 (51.4%) and 47 of 72 (65.2%) cases, respectively, whereas positive expression was minimally detected in nontumor regions under our staining conditions (Fig. 2D).

Clinicopathologic examination revealed that the protein expression levels of DKK3 or CKAP4 are not associated with various clinicopathologic parameters including tumor invasion, lymph node metastasis, or venous invasion (Supplementary Table S2). Overall survival was decreased in patients positive for DKK3 compared with patients that were negative for DKK3 (P = 0.0402; Fig. 2E). In addition, ESCC cases positive for both DKK3 and CKAP4 significantly shortened the duration of overall survival and relapse-free survival as compared with ESCC cases positive for either DKK3 or CKAP4 (P = 0.0139 and 0.0333, respectively; Fig. 2E and F). Univariate analysis demonstrated that patients with pN1-3, lymphatic vessel invasion, or DKK3 and/or CKAP4 expression were associated with inferior overall survival (Supplementary Table S3). Multivariate analysis identified that, along with lymph node metastasis (pN1-3), being both DKK3 and CKAP4 positive was indicative of poor prognosis (Supplementary Table S4). Taken together, these results suggest that the expression of DKK3 and CKAP4 negatively correlates with prognosis in ESCC.

Because DKK1 was also shown to be expressed in ESCC (29, 35), the expression levels of DKK1 and DKK3 were compared in ESCC. The results revealed that the expression of DKK1 and DKK3 is mutually exclusive in ESCC cell lines that we examined (Fig. 3A). Both DKK1 and DKK3 were expressed in 27% of the ESCC cases used in Fig. 2 (Fig. 3B); furthermore, they were present in different tumor lesions in the same patient (Fig. 3C). Whereas DKK1 was expressed throughout the tumor lesion, DKK3 showed a tendency to be localized to peripheral regions of the tumors (Fig. 3C). ESCC cases were classified according to the expression of ligands (DKK1 and/or DKK3) and a receptor (CKAP4; Fig. 3D). Interestingly, patients expressing both CKAP4 and ligands (either DKK1 and/or DKK3) showed poorer overall survival and relapse-free survival compared with other patients (Fig. 3E). Thus, DKK1 and DKK3 can be expressed in ESCC independently, while the simultaneous expression of both ligands and CKAP4 is associated with increased aggressiveness of ESCC.

DKK3 and CKAP4 formed an endogenous complex in TE-11 cells (Fig. 4A). When compared with HCT116 colon cancer cells, CKAP4 was significantly localized to the cell surface membrane of TE-11 and KYSE960 cells (Fig. 4B). Knockdown of DKK3 or CKAP4 using siRNA in TE-11 and KYSE960 cells resulted in the inhibition of cell proliferation under two-dimensional culture conditions (Fig. 4C–F). Impaired cell proliferation caused by DKK3 or CKAP4 knockdown was rescued by the expression of DKK3-FLAG or CKAP4-HA, respectively (Fig. 4G–J). A DKK3 mutant lacking the CRD1 domain (DKK3-ΔCRD1) did not rescue the phenotype (Fig. 4K and L). Depletion of DKK3 or CKAP4 in TE-11 and KYSE960 cells inhibited AKT activity, but not Src and ERK activities (Fig. 4G–I; Supplementary Fig. S2A and S2B). AKT activity was also rescued by the expression of DKK3 and CKAP4 in siRNA-transfected cells (Fig. 4G–I). These results suggest that the binding of DKK3 to CKAP4 activates AKT and is required for DKK3-mediated ESCC cell proliferation.

KYSE960 cells exhibited xenograft tumor formation when they were inoculated into immunodeficient mice; KYSE960 cells stably expressing shRNAs for DKK3 or CKAP4 showed decreased cell proliferation ability in vitro and tumor-forming ability in vivo (Fig. 5A and B; Supplementary Fig. S3A and S3B). Rabbit anti-CKAP4 polyclonal antibody (anti-CKAP4 pAb) inhibited the binding of DKK3 and CKAP4 (Fig. 5C; Supplementary Fig. S3C). In DKK3-depleted KYSE960 cells, anti-CKAP4 pAb suppressed DKK3-dependent internalization of cell surface CKAP4 (Fig. 5D) and activation of AKT (Fig. 5E). After xenograft tumors were allowed to grow to a size of approximately 100 mm³, anti-CKAP4 pAb or control IgG was administrated intraperitoneally and tumor development was observed for an additional 22 days. The mice receiving anti-CKAP4 pAb reduced tumor volumes (Fig. 5F). These results suggest that the DKK3-CKAP4 axis is required for tumor growth in vivo and that anti-CKAP4 antibody is effective as an anticancer agent against ESCCs with DKK3 and CKAP4 expression.

The underlying mechanism that induces high expression of DKK3 in ESCC is not known. p63, a p53-related transcription factor, is frequently amplified in ESCC and is required for cell proliferation (24, 27). p63 was expressed in the basal layer of esophageal epithelium in nontumor regions as well as throughout entire tumor lesions (Fig. 6A). DKK3-positive areas seemed to partially overlap with p63-positive areas in the peripheral legions of tumors (Fig. 6A). ΔNp63α was primarily expressed in TE-11 and KYSE960 cells (Fig. 6B; Supplementary Fig. S4). ΔNp63β and ΔNp63γ were also weakly expressed in TE-11 cells. Knockdown of p63, using a common siRNA for ΔNp63 isoforms, in TE-11 and KYSE960 cells reduced the expression of DKK3; DKK3 expression was rescued by exogenously expressed ΔNp63α in KYSE960 cells (Fig. 6B). In addition, although overexpression of DKK3 did not increase proliferation, knockdown of p63 inhibited the proliferative ability of TE-11 and KYSE960 cells; the inhibition was partially rescued by ectopic expression of DKK3 (Fig. 6C).

A consensus binding motif for p63 was found in the approximately 8-kb upstream promoter region of the DKK3 transcription starting site (Fig. 6D). A chromatin immunoprecipitation (ChIP) assay revealed that p63 binds to the predicted binding region in KYSE960 cells (Fig. 6E). Loss-of-function mutations in p53 have been frequently found in ESCC and ESCC cell lines as well (24, 36, 37). It was examined whether the p63-binding genomic region that we identified is involved in the transcription of DKK3 using the reporter gene assay. ΔNp63α expression stimulated luciferase activity under the control of the genomic region including −8466 to −8118 of DKK3 in X293T cells depleting p53 (Fig. 6F). When the possible p63-binding sites were mutated, ΔNp63α-dependent luciferase activity was diminished (Fig. 6D–F). Similarly, although the expression of ΔNp63α or depletion of p53 alone did not induce the endogenous DKK3 expression, their combination stimulated the DKK3 expression (Fig. 6F). These results suggest that p63 could function as a transcription factor to stimulate DKK3 expression in ESCC, likely in the setting of a p53 mutational background; thus, one reason why DKK3 is overexpressed in ESCC may be due to the expression of ΔNp63α.

As shown in some murine and human tissues, including gastrointestinal organs, liver, and pancreas (38), organoid cultures were generated from adult mouse esophagus (33). The organoids consisted of an outer layer of basal cells positive for cytokeratin (CK) 14 and inner differentiated cells expressing CK13 (Fig. 7A). Growth factors, including EGF, noggin, R-spondin, and FGF10, were required for esophageal organoid proliferation (Fig. 7B).

It has been shown that the activation of RAS/MAPK pathway was found in human esophageal cancer in addition to genomic mutations in p53 locus (24). Furthermore, expression of Kras^G12D in the foregut is able to generate esophageal cancer in mice treated with genotoxic agents (39). Therefore, it was examined to generate tumor-like organoids by introducing Kras^G12D and depleting p53. Esophageal organoids were generated from KrasG12D^LSL/p53^lox/lox (KP) mice–derived esophageal epithelial cells and Cre recombinase was transduced into KP organoids to generate another type of organoid (KPC). KP organoids were morphologically and histologically similar to wild-type esophageal organoids and required complete growth media, including EGF and FGF10 (Fig. 7C). In contrast, KPC organoids grew rapidly and exhibited increased organoid size in the absence of EGF and FGF10 (Fig. 7C).

CKAP4 was confirmed to be expressed on the cell surface membrane of the organoid cells when single cells were prepared from the organoids (Fig. 7D). Additional expression of ΔNp63α, DKK3, or DKK1 into KPC organoids further increased the size of organoids, and expression of ΔNp63α increased levels of DKK3 mRNA but not that of DKK1 mRNA (Fig. 7E). These organoids were able to differentiate as they expressed CK13, but CK13 expression was distributed irregularly in places other than the inner region (Fig. 7F), suggesting that these organoids may lose the epithelial polarity and exhibit tumor-like characteristics. Notably, anti-CKAP4 pAb inhibited the growth of the KPC organoids expressing ΔNp63α (Fig. 7G). These results suggest that DKK3 promotes organoid growth via CKAP4 under conditions in which driver genes, such as Kras and p53, are genetically manipulated.

In our preceding study, we found that CKAP4 is a DKK1 receptor (28, 29, 40); our current study suggests that CKAP4 also functions as a receptor for other DKK family proteins including DKK2, DKK3, and DKK4. This finding is especially important for DKK3, because its receptor has been unknown to date. We showed that DKK3 binds to CKAP4, activates AKT, and induces the internalization of CKAP4. Furthermore, we also demonstrated that DKK3-ΔCRD1 does not rescue the phenotypes induced by DKK3 knockdown. These results strongly support the conclusion that CKAP4 on the cell surface membrane is a receptor for DKK3.

Possible functions of DKK3 in other pathologic conditions have been reported. For instance, DKK3 regulated MMP2/9 levels to control normal acinar morphogenesis in prostate epithelial cells and inhibited TGFβ-induced prostate cancer cell migration and invasion (41). DKK3 was also expressed in macrophages found in atherosclerotic plaques, while Dkk3 and ApoE simultaneous knockout mice exhibited decreased atherosclerotic pathogenesis (42). Therefore, it is intriguing to speculate that DKK3 has pleiotropic functions and that CKAP4 mediates several pathologic functions of DKK3.

Despite recent advances in endoscopic detection of early esophageal tumor lesions and in therapeutic advances such as endoscopic resection, surgery, radiation, and chemotherapy, the survival of patients with ESCC is still poor. Thus, a novel molecularly targeted therapy is required. Taken together with our previous report that the DKK1-CKAP4 signaling is a therapeutic target for ESCC (29), our current study demonstrates that DKK3 activates AKT via CKAP4 like DKK1 and that the DKK3-CKAP4 and DKK1-CKAP4 axes are activated in distinct populations of ESCC cells, because DKK1 and DKK3 are separately expressed in the different cultured ESCC cells and ESCC cases and in the different tumor lesions of the same cases.

The anti-DKK1 antibody DKN-01, a humanized therapeutic mAb against DKK1, underwent phase I evaluation for advanced esophageal cancer in a combination with paclitaxel (43). Given that DKK1- or DKK3-expressing tumors compose a different subset of cancer cells even in the same patient with ESCC, anti-CKAP4 antibody effectively suppresses both types of DKK signaling in ESCC. Therefore, therapeutic agents directed against CKAP4 could be more useful than antibodies against each DKK family member. The AKT activity and proliferation of TE-14 cells, where CKAP4 is expressed on the cell surface membrane but DKK1 and DKK3 are hardly expressed, were not decreased by the anti-CKAP4 antibody (29). These results support the idea that anti-CKAP4 antibody specifically affects cancer cells that express both DKKs and CKAP4.

The underlying mechanism by which DKK3 expression is increased in cancer tissue has not yet been elucidated. We found that p63 knockdown reduces the expression of DKK3 in ESCC cells and the reduced expression of DKK3 was rescued by the reexpression of ΔNp63α. Furthermore, the predicted enhancer region was associated with p63, suggesting that p63 is a possible transcription factor that promotes DKK3 expression. On the other hand, ΔNp63 is physiologically required for epidermal development and homeostasis (44) and ΔNp63, but not DKK3, is expressed in normal basal layer tissue, suggesting that ΔNp63-dependent DKK3 expression occurs in a tumor-dependent manner. It is possible that in the environment of a tumor, there may be unknown coactivators of ΔNp63 that can promote DKK3 expression, because overexpression of ΔNp63α alone did not affect DKK3 expression in X293T cells but increased DKK3 expression in p53-depleted cells. In addition, the expression of DKK3 partially rescued the inhibition of cell proliferation in p63-depleted cells, suggesting that not only DKK signaling but other pathways are activated in ESCC expressing ΔNp63.

Recent progress in organoid technology has allowed the establishment of epithelial organoid culture from many murine and human tissues including gastrointestinal organs, liver, and pancreas (33, 38). Esophageal organoids retain a similar cell-surface phenotype as compared with primary tissues, including a nonquiescent stem cell population residing in the basal epithelium. By expressing Kras^G12D and depleting p53, the organoids were able to grow in the absence of EGF and FGF10 in vitro. Organoid growth was promoted by the expression of either DKK3, DKK1, or ΔNp63α, while treatment with anti-CKAP4 pAb suppressed organoid growth. Thus, these tumor-like organoids could be useful for the evaluation of the effects of new anticancer drugs, such as anti-CKAP4 antibody. It is also possible to establish organoids from surgically resected human intestinal tissues and endoscopic biopsies of patients with cancer (45). It would be interesting to evaluate the efficacy of anti-CKAP4 antibody in esophageal cancer organoids from patients that are positive for both DKK3 and CKAP4.

No potential conflicts of interest were disclosed.

Conception and design: C. Kajiwara, K. Fumoto, M. Yamasaki, A. Kikuchi

Development of methodology: C. Kajiwara, K. Fumoto, K. Asano, A. Kikuchi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Kajiwara, K. Fumoto, K. Asano, K. Odagiri, M. Yamasaki, E. Morii

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Kajiwara, K. Fumoto, H. Kimura, M. Yamasaki, E. Morii

Writing, review, and/or revision of the manuscript: K. Fumoto, M. Yamasaki, E. Morii, A. Kikuchi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Fumoto, S. Nojima, M. Yamasaki, H. Hikita, T. Takehara, Y. Doki

Study supervision: K. Fumoto, A. Kikuchi

Other (analyzed the expression of DKK1, DKK3, and CKAP4 in cancer specimens): H. Kimura

We would like to thank Drs. S. Tsukita, M. Hatakeyama, Y. Matsuura, K. Matsumoto, A. Shintani, Y. Watanabe, and K. Shinjo for donating cells and helping data analysis. K. Asano is supported by the Osaka University Medical Doctor Scientist Training Program. This work was supported by Grants-in-Aid for Scientific Research (2016-2021; grant no. 16H06374 to A. Kikuchi), and Grants-in-Aid for Young Scientists (Start-up; 2016-2017; grant no. 16H06944 to H. Kimura) from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and also supported by the Project Promoting Support for Drug Discovery (2016-2018; grant no. DNW-16002 to K. Fumoto) and the Project for Cancer Research and Therapeutic Evolution (P-CREATE; 2016–2017; grant no. 16cm0106119h0001 and 2018–2019; grant no. 18cm0106132h0001; to A. Kikuchi) from the Japan Agency for Medical Research and Development, AMED, and by grants to A. Kikuchi from the Yasuda Memorial Foundation.

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.