Background
Breast cancer represents the most common malignancy in women, with an increased incidence [
1,
2]. In 2020, 2.3 million people were estimated to be diagnosed with breast cancer worldwide, and 685,000 people died of this tumor [
3]. Breast cancer is classified by the gene expression profiles into intrinsic subtypes, such as luminal A, luminal B, HER2 (human epidermal growth factor receptor 2), and basal-like, which possess distinct biological properties, drug responses, and patient outcomes [
1,
4‐
6]. More practically, the prognosis predictions and therapeutic strategies are immunohistochemically determined by the estrogen receptor (ER), progesterone receptor (PgR), and HER2 status, as well as by Ki-67/MIB-1 proliferation index [
7,
8]. Patients positive for either ER, PgR, or HER2 exhibit a relatively favorable prognosis owing to valid medication against these receptors. By contrast, triple-negative breast cancer (TNBC), which accounts for 15–20% of breast cancer, shows limited therapeutic options and poor outcomes [
9,
10].
Cell adhesion molecules, such as E-cadherin, are necessary for multicellular organisms to maintain tissue architecture and homeostasis. They were considered to function as tumor suppressor proteins, but it is an oversimplified principle [
11]. For instance, E-cadherin contributes to collective invasion and/or metastasis in the breast cancer [
12,
13]. In addition to the adhesive activity, cell adhesion proteins exhibit signaling properties that coordinate a wide range of cell behaviors [
14‐
18], theoretically via activation or repression of transcription factors that control the expression of target genes [
19]. However, it remains poorly established how cell adhesion signaling reaches the nucleus and regulates gene expression.
The claudin (CLDN) family is the structural and functional backbone of tight junctions and contains a short cytoplasmic N-terminus, two extracellular loops (EC1 and EC2), and a C-terminal cytoplasmic domain [
20‐
25]. We have recently identified that CLDN6-adhesion signaling regulates the nuclear receptor activity [
20‐
23]. In brief, we uncovered that CLDN6 couples with Src-family kinases (SFKs) in the EC2-dependent and the C-terminal Y196/200-dependent manners. We also showed that the CLDN6/SFK/PI3K/AKT axis targets the AKT phosphorylation sites in the retinoic acid receptor γ (RARγ) and estrogen receptor α (ERα) and stimulates their activities independently of ligands. Importantly, these phosphorylation motifs (RXXS, aa 515 to 518 in human ERα) are conserved in 14 of 48 members of human nuclear receptors, further suggesting the biological relevance of these phosphorylation sites.
CLDNs also possess aberrant expression and/or subcellular localization in a broad range of cancer types [
26‐
28], leading to the promotion or repression of tumor progression, possibly via the dysregulated CLDN signaling [
19]. In breast cancer, the CLDN-low subtype, which is characterized by the low expression of cell–cell adhesion molecules such as CLDN3/4/7, had been considered to show immature cancer properties and poor prognosis [
33‐
35]. However, it has been recently reported that the CLDN-low tumor displays heterogeneous phenotypes but neither an independent intrinsic subtype nor a poor outcome [
33,
34]. On the other hand, there are conflicting reports on the relationship between high CLDN4 expression and patient prognosis in the breast cancer [
35‐
37]. We have recently demonstrated that aberrant CLDN6 signaling accelerates endometrial cancer progression in vitro and in vivo by hijacking the CLDN6–ERα axis [
38,
39]. Among the CLDN family, CLDN4 is evolutionarily close to CLDN6 [
40]. Additionally, Y196 and Y200 in the C-terminal cytoplasmic domain of human/mouse CLDN6 are required to propagate intracellular signals [
18,
39] and are conserved in human/mouse CLND4Y193/197 [
19]. Taken together, we hypothesized that the CLDN4 signaling might regulate breast cancer progression by regulating the nuclear receptor activity in a similar mechanism to CLDN6 in endometrial cancer.
Here, we show that aberrant CLDN4 signaling advances breast cancer metabolism and progression via liver X receptor β (LXRβ), a member of the nuclear receptor family. We also demonstrate that the CLDN4 signaling activates SFK/AKT and targets LXRβS432, resulting in stimulation of the LXRβ activity and malignant behaviors in breast cancer cells. Moreover, we present that the "CLDN4-high/LXRβ-high" group in TNBC cases reveals significantly shorter overall and recurrence-free survival than the "CLDN4-low and/or LXRβ-low" group.
Discussion
In the present study, we demonstrated that CLDN4 accelerates breast cancer progression in vitro and in vivo. This was apparent because KO of the human
CLDN4 gene led to a reduction in cell proliferation, migration, and invasion in two distinct breast cancer cell lines T47D and/or MCF-7. Conversely, these malignant phenotypes were stimulated by the re-expression of CLDN4 in T47D:
CLDN4–/– and MCF-7:
CLDN4–/– cells, as well as by the introduction of the
CLDN4 gene in MDA-MB-231 cells. In addition, using T47D:
CLDN4–/– and MDA-MB-231:
CLDN4 xenografts, it was shown that CLDN4 promotes the tumor growth and cell proliferation of breast cancer cells in vivo. Tumor budding, small clusters of cancer cells, was hindered in T47D:
CLDN4–/– xenografts compared with T47D xenografts, further indicating that CLDN4 functions as a tumor promoter in breast cancer cells. These results are consistent with those of a previous report, which showed that CLDN4 stimulates malignant phenotypes in MCF-7 cells [
47].
We also showed that CLDN4 activates SFK and the downstream AKT in breast cancer cells in the EC2- and Y197-dependent manners, leading to promote their cell proliferation. This conclusion was drawn from the following results: 1) colocalization of CLDN4 and pSFK along cell boundaries was apparently observed in both T47D and MCF-7 cells, whereas it was diminished in T47D:
CLDN4–/– and MCF-7:
CLDN4–/– cells, as well as in C-CPE-treated T47D and MCF-7 cells; 2) the pSFK levels were decreased and increased in T47D:
CLDN4–/– and MDA-MB-231:
CLDN4 cells compared with their parental cells, respectively; 3) the pSFK intensities in T47D:
CLDN4–/–:CLDN4ΔEC2 and T47D:
CLDN4–/–:CLDN4ΔC cells were lower than those in T47D:
CLDN4–/–:WT-CLDN4 cells; 4) the pSFK levels in T47D:
CLDN4–/–:CLDN4Y197A cells were reduced compared to those in T47D:
CLDN4–/–:WT-CLDN4 and T47D:
CLDN4–/–:CLDN4Y193A cells; 5) the CLDN4-triggered cell proliferation was reversed in T47D:
CLDN4–/–:CLDN4ΔEC2, T47D:
CLDN4–/–:CLDN4ΔC, and T47D:
CLDN4–/–:
CLDN4Y197A cells compared with that in T47D:
CLDN4–/–:WT-CLDN4 cells; 6) The CLDN4-provoked cell proliferation was prevented in MDA-MB-231:
CLDN4ΔC cells compared with MDA-MB-231:
CLDN4 cells; 7) the CLDN4-initiated cell proliferation abrogated upon C-CPE, PP2 and AKT inhibitor VIII treatment in T47D cells but not in T47D:
CLDN4–/– cells. We have recently reported that the EC2 domain and Y196/200 of CLDN6 are required to recruit and activate SFKs and to stimulate malignant phenotypes of endometrial cancer cells [
44,
47]. Thus, at least two CLDN subtypes propagate SFKs by similar mechanisms, namely in the EC2- and the C-terminal tyrosine residue-dependent manners. Since CLDN4Y197 and CLDN6Y200 are conserved in human and mouse CLDN1/2/5/9/17/18 [
19], it would be interesting to determine the biological significance of the corresponding tyrosine residues in various types of cancer.
Our RNA sequencing analysis, using T47D and T47D:
CLDN4–/– cells as well as MCF-7 cells and MCF-7:
CLDN4–/– cells, first suggested a link between CLDN4 and LXRs signalings in breast cancer cells. We also showed that LXRβ but not LXRα protein is expressed in T47D, MCF-7, and MDA-MB-231 cells, as well as in breast cancer analyzed. Therefore, we subsequently generated T47D:
CLDN4–/–:LXRβ–/– (T47D:dKO) and a series of rescue cell lines. Consequently, by comparing phenotypes in T47D:dKO:
CLDN4 cells with those in T47D:dKO cells, we demonstrated that the CLDN4 signaling enhances cell growth, migration, and intracellular cholesterol and triglyceride levels via LXRβ. LXRs play a key role not only in maintaining cholesterol homeostasis and fatty acid metabolism but also in cellular proliferation [
48]. Taken together with the findings showing that LXRα controls the growth of skin and oral squamous cell carcinoma by altering the cholesterol homeostasis [
49,
50], our results suggest that the CLDN4 signaling promotes cell proliferation in breast cancer cells, possibly by LXRβ-mediated control of genes involved in cancer metabolism. More importantly, CLDN4-enhanced cell proliferation, migration, and tumor growth, as well as intracellular levels of cholesterol and triglyceride, were prevented in T47D:dKO:
CLDN4:LXRβS432A cells compared with that in T47D:dKO:
CLDN4:LXRβ cells, indicating that LXRβS432 is responsible for the CLDN4/SFK/AKT-accelerated breast cancer metabolism and progression. Furthermore, AKT and SGK1 formed a complex with LXRβ in 293T cells, reinforcing the conclusion. We previously demonstrated that the CLDN6/SFK/AKT signaling directs S379 and S518 in mouse RARγ and human ERα, respectively [
18,
19,
39]. Hence, these results revealed that the cell-adhesion signaling targets the AKT-consensus phosphorylation sites in at least three nuclear receptors.
Another conclusion of our study is that the CLDN4 signaling LXRβ-dependently and independently regulates a range of gene expressions in breast cancer cells. RT-qPCR analysis, using T47D, T47D:CLDN4–/–, T47D:dKO, T47D:dKO:CLDN4, T47D:dKO:CLDN4:LXRβ, and T47D:dKO:CLDN4:LXRβS432A cells, uncovered that the CLDN4-controlled genes are categorized into at least six groups in terms of distinct requirement of LXRβ and LXRβS432. Among 23 CLDN4-regulating genes whose products are associated with tumor progression in various cancers, eight genes were upregulated via LXRβ, six of which were activated in an LXRβS432-dependent manner. Because both LXRβ and LXRβS432 were essential for the CLDN4-accelerated cell proliferation and tumor growth in breast cancer cells, the LXRβ- and LXRβS432-dependent CLDN4-controlling gene products would be critical to promote breast cancer progression.
Clinicopathologically, we found that the "CLDN4-high/LXRβ-high" and "CLDN4-low and/or LXRβ-low" TNBC subjects possess poor and relatively favorable outcomes, respectively. This is reasonable because a series of our analyses disclosed that the CLDN4 signaling stimulates breast cancer progression through LXRβ. Thus, evaluating expression levels of both CLDN4 (input signal) and LXRβ (output signal) is required to predict distinct prognoses in TNBC cases. Along this line, clinicopathological analysis using expression levels of either CLDN4 [
35‐
37] or LXRβ [
51] does not seem to be enough to predict a prognosis in breast cancer, especially in TNBC. Since there is no valid medication for TNBC, it should also be worth noting that the CLDN4/LXRβ axis may be a promising therapeutic target for TNBC. For instance, an LXR inverse agonist is effective against various types of cancer without obvious side effects [
52]; therefore, it should be determined whether the LXRβ-targeting treatment could be a therapeutic option in the "CLDN4-high/LXRβ-high" TNBC cases.
Methods
Antibodies
The antibodies used in this study are listed in Additional file
1: Table S2.
Cell lines and cell culture
Breast cancer cell lines T47D (HTB-133) and MDA-MB-231 (HTB-26TM) were purchased from the American Type Culture Collection (ATCC). MCF-7 (RCB1904) and SKBR-3 (RCB2132) were obtained from RIKEN Bioresource Center. These cell lines were maintained in Dulbecco's Modified Eagle Medium (DMEM) with 10% fetal bovine serum (FBS; Sigma-Aldrich) and 1% penicillin–streptomycin-amphotericin B suspension (161-23181, FUJIFILM). For assays, the cells were grown in a phenol red-free medium with charcoal-treated FBS to exclude fat-soluble ligands. For the preparation of charcoal-treated FBS, 500 ml of FBS was treated with 0.5 g of charcoal dextran-coated (Sigma) overnight at 4ºC, followed by filtration using 0.22 µm cellulose acetate filter membranes. The cells were treated for 24 h with 1 µg/ml of C-CPE, 10 µM of PP2 (529573, Sigma-Aldrich), 0.1 µM of AKT inhibitor VIII (CS-0001, Funakoshi), 1 mg/ml of Cholesterol-Water Soluble (C4951-30MG; Sigma-Aldrich), and 1–25 µM of T0901317 (71810, Cayman Chemical) 24 h after plating. C-CPE production and purification were performed as described previously by using
E.coli BL21 and the expression vector pET16b coding C-CPE194-319 [
42].
Genome editing
We used the CRISPR technique to establish the
CLDN4 and
LXRβ (NR1H2) knockout cell lines. Annealed oligos, including targets described in Additional file
1: Fig. S2, were cloned into the
Esp3I site of lentiCRISPR v2 plasmid (#52961 Addgene). Although lentiCRISPR v2 was originally designed to be packaged into lentivirus, the plasmids were directly and transiently transfected into the parental cells by Lipofectamine 3000 (15292465, Thermo Fisher Scientific) in the present study. Twelve h after transfection, the cells were exposed to 10 µg/ml of puromycin for 24 h, followed by limiting dilution and genotyping by genomic PCR. Knockout of
CLDN4 and
NR1H2 genes was verified by DNA sequencing after TA-cloning of genomic PCR products.
Expression vectors, transfection, and establishment of stable cell lines
The protein-coding regions of human CLDN4 or NR1H2 were cloned into the NotI/ BamHI site of the CSII-EF-MCS-IRES2-Venus (RDB04384, RIKEN) plasmid. Expression vectors of mutant genes, including CLDN4ΔEC2, CLDN4ΔC, CLDN4Y193A, CLDN4Y197A, and LXRβS432A, were established using a standard PCR-based site-directed mutagenesis protocol. The overexpression or rescued cell lines were established by lentiviral transfection. First, lentiviral vectors were generated by transfecting 1.0 × 107 cells of 293 T with 10 µg of the CSII plasmids containing the target genes, 5 µg of packaging plasmids psPAX2 (#12260, Addgene), and pCMV-VSV-G (#8454, Addgene) using Polyethylenimine Max (PEI Max; 24765-1, Cosmo Bio). Culture media containing recombinant lentiviruses were collected 72 h after transfection and directly added to the cell culture medium of T47D:CLDN4–/–, T47D:CLDN4–/–:LXRβ–/– (T47D:dKO), MCF-7:CLDN4–/–, and MDA-MB-231 cells. After more than 7 days and three times passages, the cells were used for further analysis. T47D:dKO:CLDN4, T47D:dKO:CLDN4:LXRβ, T47D:dKO:CLDN4:LXRβS432A cells were single-cell cloned by limiting dilution.
Cell proliferation, migration, invasion, and apoptosis assays
Cell proliferation index was evaluated by incorporation of bromodeoxyuridine (5-Bromo-2-DeoxyUridine, BrdU; 19–960, Sigma-Aldrich). 24–48 h after passage, cells were exposed to BrdU for 30 or 60 min. The specimens were fixed with 4% paraformaldehyde and 0.1% Triton-X, followed by immunostaining with the anti-BrdU antibody and its standard protocol.
Total viable cell counts were quantified by CellTiter 96 AQueous One Solution Cell Proliferation Assay (MTS) Kit (G3582, Promega). One thousand cells were seeded on 96-well plates, and the reagent was added to each well after 48 h, followed by measurement of absorbance at 490 nm.
To evaluate cell migration, wound areas were generated by scratching with disposable 200 µl pipette tips 48 h after passage. Photographs of the wound areas were taken at the same locations after scratching by the indicated intervals, using a phase-contrast microscope. Wound healing was calculated as the percentage of the remaining cell-free area compared with the initial wound area using ImageJ software (Wayne Rasband National Institutes of Health).
BioCoat Matrigel Invasion Chamber (#354480, Corning) was used for assessing cell invasion by following the provider's protocol. Briefly, 2.0 × 104 cells were transferred to each well. Twenty-four h after passage, the samples were fixed with 100% methanol and stained with crystal violet. The invasion index was calculated by dividing by cell numbers in negative control membranes, which consist of empty mesh but do not contain matrigel.in situ Cell Death Detection Kit (11684795910, Sigma-Aldrich) was used for the evaluation of cell apoptosis.
Measurement of cholesterol and triglyceride content
Cholesterol and triglyceride in cell lysates were measured by Cholesterol/Cholesterol Ester-Glo Assay (J3190, Promega) and Triglyceride-Glo Assay (J3160, Promega), respectively. One thousand cells were plated in 96-well plates, and the cholesterol and triglyceride contents were quantified by the kits following the manufacturer’s protocols after 24 h.
Xenograft model
Xenograft studies were performed in 8-week-old CB17/IcrJcl-Prkdcscid female mice (CLEA Japan). 5.0 × 106 cells were subcutaneously injected into the back of anesthetized mice. Twenty-eight days after injection, the mice were ethically sacrificed, and tumor tissues were collected. The samples were immediately fixed with 10% neutral buffered formalin solution for 24–36 h. Hematoxylin–eosin staining and Ki-67 staining were performed following standard protocols optimized for human tissues. Ki-67 index was assessed in the hot spot of each tumor. Tumour budding was analyzed by counting tumor clusters consisting of two to six cells at the five high-power fields of invasion fronts of each specimen.
Immunoprecipitation and immunoblot
Total cell extracts were collected by using CellLytic MT Cell Lysis Reagent (C3228, Sigma-Aldrich) and were subsequently sonicated with three or four bursts of 5–10 s. Immunoprecipitation was performed using Immunoprecipitation Kit Protein G (11719386001, Sigma-Aldrich), following the manufacturer's protocol. 1 µg of ChromPure Rat IgG (012-000-003, Jackson Immunoresearch Laboratories) was used as a negative control. Whole-cell lysates or immunoprecipitated samples were mixed with sample loading buffer containing 2-mercaptoethanol and incubated for 10 min at 95 °C. They were resolved by one-dimensional SDS-PAGE and electrophoretically transferred onto a polyvinylidene difluoride membrane. The membranes were saturated with PBS containing 4% skimmed milk or PVDF Blocking Reagent for Can Get Signal (NYPBR01, TOYOBO) for 30 min. After rinsing in TBS containing 0.1% Tween 20, the membranes were incubated with a primary antibody solution diluted in PBS or Can Get Signal Solution 1 (NKB-101, TOYOBO) for 1 h at room temperature or overnight at 4 °C, followed by 1-h incubation with horseradish peroxidase (HRP)-conjugated secondary antibodies diluted in PBS or Can Get Signal Solution 2 (NKB-101, TOYOBO). An anti-GFP antibody was used for detecting Venus, of which amino-acid alignment completely matches the antigen region of GFP. They were rinsed again and exposed to EzWestLumi One (ATTO). After rinsing with 10% H
2O
2 to inactivate HRP, each membrane was hybridized with HRP-conjugated anti-beta actin antibody as loading controls. Total cell extract of F9 cells [
53] was used as the positive control shown in Additional file
1: Fig. S1A. Each signal was quantified by ImageJ software (Wayne Rasband National Institutes of Health) and divided by the corresponding actin levels.
Immunofluorescence and imaging
Cells were grown on coverslips coated with Cellmatrix Type I-A (Nitta gelatin). The samples were fixed in 4% paraformaldehyde and 0.2% Triton-X for ten min at room temperature. After washing with PBS, they were preincubated in PBS containing 5% skimmed milk. They were subsequently incubated overnight at 4 °C with primary antibodies diluted in Signal Booster Immunostain F (BCL-ISF, Beacle), then rinsed again with PBS, followed by a reaction for 1 h at room temperature with appropriate secondary antibodies. All samples were examined using a laser-scanning confocal microscope (FV1000, Olympus). Photographs were processed with Photoshop CC (Adobe) and ImageJ software (Wayne Rasband National Institutes of Health).
RNA extraction, RT-PCR, and RNA sequencing
Total RNA was isolated from cells using TRIzol RNA Isolation Reagents (15596018, Thermo Fisher Scientific). For RT-qPCR, reverse transcription was performed using SensiFAST cDNA Synthesis Kit (BIO-65054, meridian BIOSCIENCE), and target genes were quantified by THUNDERBIRD SYBR qPCR Mix (QPS-201, TOYOBO) and Step One Real-Time PCR System (Applied Biosystems) using the primers listed in Additional file
1: Table S3. The expression levels of the target genes were normalized to the corresponding
GAPDH expression.
RNA sequencing and mapping were performed by BINDS, a platform project for supporting drug discovery and life science research in Japan. To generate mapped bam files, the index-trimmed single-end 100 bp reads were aligned to the human reference genome (GRCh38 v90). The mapped bam files were imported to SeqMonk software (Babraham Bioinformatics) as single-ended RNA-Seq data. Then they were quantitated by using a standard RNA-Seq quantitation pipeline consisting of TopHat2, CuffLinks2, and CummeRbund. Raw data was uploaded to Gene Expression Omnibus (
https://www.ncbi.nlm.nih.gov/geo/) as GSE207704.
TCGA expression analysis
Gene expression and clinical data of 1100 breast cancer cases in TCGA cohorts were downloaded from cBioPortal (
www.cbioportal.org/). mRNA expression levels of
LXRα (NR1H3) and
LXRβ (NR1H2) were imported as RSEM values [
53] and visualized as a scatter plot by Prism 9 (GraphPad).
Tissue collection and immunohistochemistry
Formalin-fixed paraffin-embedded (FFPE) tissue sections were obtained from 187 patients with breast cancer (age, 27–85 years; average ± SD = 55.7 ± 11.7) who underwent a total or partial mastectomy and sentinel lymph node biopsy or axillary lymph node dissection between 2008 and 2013 at Fukushima Medical University Hospital (Additional file
1: Table S1). The subjects were limited to patients who were confirmed to have at least 5-year outcomes. Detailed information, including postoperative pathology diagnosis reports, age, stage (The UICC TNM classification), histological type, ER, PgR, HER2, recurrence status, recurrence-free survival, and overall survival, was obtained. Patients' backgrounds were anonymized. CLDN4 and LXRβ expression were independently and blindly evaluated by two pathologists and one breast specialist. The signal intensity of CLDN4 was semi-quantified using the Immunoreactive Score (IRS; Additional file
1: Table S4) [
45], and the lowest scores were adopted. Briefly, staining intensity (SI) was classified into four levels (0, negative; 1, weak; 2, moderate; 3, strong) and staining range (percentage of positive cells; PP) into five levels (0, < 1%; 1, 1–10%; 2, 11–30%; 3, 31–50%; 4, > 50%). IRS was calculated by multiplying SI and PP. Scores 0–4 were defined as CLDN4-low, and scores 6–12 as CLDN4-high. On the other hand, LXRβ expression was evaluated by the Allred Score, which is used for assessing the ER and PgR in the breast cancer [
46], and the highest scores were adopted. Scores 0–6 were defined as LXRβ-low and scores 7–8 as LXRβ-high.
Statistical analysis
Statistical significance for cell proliferation was analyzed by the Mann–Whitney test, while those for cell migration, invasion, and growth of xenograft were analyzed by Welch's t-test. GSEA was performed using GSEA v4.2.3 software and hallmark gene sets which are publically available from the Broad Institute [
54]. Kaplan–Meier method was used for survival analyses, and differences between groups were analyzed using the log-rank test. Two-tailed
p values < 0.05 were considered to indicate a statistically significant result when comparing two groups. Benjamini and Hochberg's correction method was used to counteract the multiple comparisons problem when comparing more than three groups. All statistical analyses were performed using GraphPad Prism v9.4.0 software.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.