Loss of TRIM29 Alters Keratin Distribution to Promote Cell Invasion in Squamous Cell Carcinoma

Tripartite motif (TRIM) family proteins have various functions in cellular processes, including intracellular signaling, cell development, apoptosis, protein quality control, and carcinogenesis (1). TRIM family proteins have conserved domains that include a RING, a B box type 1, a B box type 2, and a coiled-coil region. Although most TRIM family proteins have a RING domain, TRIM29 (also known as ataxia-telangiectasia group D complementing protein, or ATDC) lacks it (2). We previously showed that TRIM29 regulates the assembly of DNA repair proteins into damaged chromatin (3). TRIM29 interacts with BRCA1-associated surveillance complex, cohesion, DNA-PKcs, and components of the TIP60 complex, suggesting that TRIM29 functions as a scaffold protein in DNA damage response. Furthermore, a study on TRIM29 knockout mice revealed TRIM29 to be a regulator for the activation of alveolar macrophages, the expression of type I IFNs, and the production of proinflammatory cytokines in the lungs (4). Moreover, TRIM29 has been reported to be overexpressed in several cancers, including lung (5), colorectal (6), and pancreatic cancers (7). A review of the expression of TRIM29 in many cancers revealed an important link between upregulated TRIM29 expression and poor prognosis in patients with malignant neoplasms (8). TRIM29 transgenic mice revealed that TRIM29 upregulates CD44 in pancreatic cancer cells via the activation of β-catenin signaling, leading to the induction of epithelial–mesenchymal transition (EMT) along with the expression of Zeb1 and Snail1 (9). These studies suggest that TRIM29 promotes tumorigenesis and tumor progression in certain cancers. Conversely, in breast cancer, TRIM29 is often silenced because of aberrant gene hypermethylation, which leads to the invasive behavior of breast cancers (10). Furthermore, we previously reported that the TRIM29-positive cells disappear in prostate cancers (11). Thus, the expression levels of TRIM29 in cancers may depend on the cell/tissue types (12). Its role in stratified squamous epithelial cells/tumors has not been elucidated.

Cutaneous squamous cell carcinoma (SCC) is a common cancer in Caucasian populations, accounting for 20% to 30% of skin malignancies (13). The risk of metastasis is low for most patients, not exceeding 5%; however, aggressive SCC is associated with high morbidity and mortality. Although cutaneous SCC can be treated by surgical removal, radiation, or chemotherapy, or by a combination of these therapies, the prognosis of patients with metastatic SCC is poor (14). Even in head and neck lesions, squamous cell carcinoma is the most common histologic type. The risk factors of distant metastasis for head and neck SCC (HN-SCC) are related to age, the site of the primary cancer, local and/or regional extension, and histologic grading (15). Patients with localized HN-SCC are treated with potentially curative therapy using treatment modalities that include surgery, radiotherapy, chemotherapy, and biologic therapy (16). The recurrence rate in early-stage HN-SCC ranges from 10% to 20%, and the recurrence rate in locally advanced HN-SCC exceeds 50%. Patients with metastatic HN-SCC have a poor prognosis with a median overall survival (OS) of less than 1 year (16). Therefore, identifying the molecular mechanisms involved in cutaneous, and head and neck SCC pathogenesis is of vital clinical importance.

In this study, we investigated the role of TRIM29 in the tumorigenesis/progression of SCC. Searches of public databases revealed that TRIM29 is highly expressed in stratified epithelial tissues, including in skin, and head and neck lesions. Immunohistochemically, TRIM29 has lower expression in malignant SCC lesions than in adjacent normal epithelial tissue or benign tumors. DNA methylation analysis revealed the CpG lesion of TRIM29 in primary cutaneous SCC tumors to be aberrantly hypermethylated, whereas that in normal epidermis is not methylated. RNAi-mediated gene knockdown of TRIM29 increases cancer cell migration and invasion; conversely, overexpression of TRIM29 inhibits these. Using nonbiased comprehensive proteomics analysis, keratins and keratin-interacting protein FAM83H have been identified as TRIM29 interactors. Immunohistochemically, TRIM29 colocalizes with keratins in the cytoplasm, and the knockdown of TRIM29 alters the distribution of keratins. Furthermore, in primary tumors, lower levels of TRIM29 expression correlate with altered distribution patterns of keratins. Our findings reveal a critical function of TRIM29 in regulating keratin distribution as well as migration/invasion of SCC.

The human cutaneous SCC A431 cells were purchased from the ATCC. The human cutaneous SCC DJM-1 cells were isolated from human skin SCC (17). The human immortalized keratinocyte cell line (HaCaT) was purchased from Cell Lines Service. The SAS human HN-SCC cell line was obtained from the Japanese Collection of Research Bioresources Cell Bank. Cell lines were cultured in DMEM supplemented with 10% FBS. All cells were authenticated by short tandem repeat profiling (Promega, August 2018) and were used within 6 months of continuous passage. All cells were tested and checked for the absence of Mycoplasma (VenorGeM).

Antibodies against TRIM29/ATDC (mouse: A-5, Santa Cruz Biotechnology; rabbit: HPA020053, Sigma), FAM83H (rabbit: HPA024604, Sigma), keratin 5 (rabbit: PRB160P, Covance), keratin 14 (rabbit: PRB155P, Covance), pan-cytokeratin (rabbit: 10550, Progen; mouse: AE1/AE3, Dako; mouse: 34bE12, Abcam), FLAG (mouse, M2, F1804, Sigma), HA (rat, 3F10, Roche), β-actin (mouse, AC-74, Sigma), and horseradish peroxidase (HRP)-conjugated secondary antibodies (GE Healthcare) were purchased from the indicated sources. Predesigned pooled siRNAs directed against human TRIM29 and the negative scramble control were purchased from Dharmacon. 5-Azacytizine (A2033) was purchased from Tokyo Chemical Industry.

For the transient knockdown, cells were transfected with siRNA duplexes by a reverse transfection method using Lipofectoamine RNAiMAX (Life Technologies).

GIPZ-TRIM29 shRNA#1 (TGTGCTCCTGGAACATGCA), shRNA#2 (TGGGTGTCAGGTACATGGA), and nonsilencing scramble control were purchased from Dharmacon. pLenti PGK Blast V5-LUC (w528-1) was a gift from Eric Campeau and Paul Kaufman (Addgene plasmid, catalog no. 19166; ref. 18). Lentiviral supernatants were generated according to an established protocol (19). Cells were selected with 1 μg/mL puromycin (Thermo Fisher Scientific) or 10 μg/mL blasticidin (Wako), and then expanded.

cDNAs encoding full-length TRIM29 and various mutated versions of TRIM29 were inserted into a pQCXI-puro vector. Retroviral supernatants were generated according to an established protocol (20). Briefly, the murine EcoVR was first introduced into cells by using an amphotropic virus to make the human cells susceptible to the subsequent infection by ecotropic viral vectors. Infected cell populations were selected using puromycin (MP Biomedicals) and then expanded.

We isolated total RNA from cultured cells or fresh-frozen sections using the RNeasy Plus Mini Kit (Qiagen). RNA concentrations were measured spectrophotometrically and samples were stored at −80°C until use. We reverse transcribed RNA using SuperScript IV VILO Master Mix (11756050, Thermo Fisher Scientific). Complementary DNA samples were analyzed by the SYBR Green System (Takara). The sequences for primers specific for human TRIM29 and the control housekeeping genes for human GAPDH are as follows:

Human TRIM29: forward: 5′-TTCCAGGAGCACAAGAATCA-3′; reverse: 5′-GCAATGACAGCTCCGTCTC-3′

Human GAPDH: forward: 5′-GAAGGTGAAGGTCGGAGTC-3′; reverse: 5′-ATGGGATTTCCATTGATGAC-3′

All experiments were performed in duplicate and normalized with respect to GAPDH levels.

Cells were plated at a density of 3–5 × 10³ cells per well, and cultured for 24, 48, or 72 hours. MTT assay was performed according to the manufacturer's protocol (CellTiter 96 Non-Radioactive Cell Proliferation Assay, Promega). Also, we assessed cell growth by direct cell counting. A total of 1.0 × 10⁵ cells with shRNA were plated onto 60-mm diameter plates. The cells were counted 1, 3, and 5 days after seeding.

For analysis of cell motility, cells were seeded onto 6-cm diameter plates in DMEM with 10% FBS overnight. The injury line was made with a yellow tip on the confluent cell monolayer. After 4 to 24 hours, the lengths of movement were measured (21). The cell invasion assay was performed using the Corning BioCoat Matrigel invasion chamber with 8.0-μm pore size (Corning 354480). Seventy-two hours after the top well was seeded, the bottom well was fixed and stained with 0.4% crystal violet solution. Invading cells were photographed in four randomly selected fields and counted.

Cells were harvested with RIPA buffer (Wako) or IP buffer (1% NP40 buffer, Thermo Fisher Scientific). Cells were left on ice for 20 minutes and then centrifuged at 14,000 × g for 10 minutes. The Bio-Rad Protein Assay Kit (Bio-Rad) was used to determine protein concentrations. Proteins were separated with SDS-PAGE 4%–15% gradient gels (Life Technologies) and transferred onto polyvinylidene difluoride membranes. The membranes were blocked for 1 hour in Tris-buffered saline (TBS) with 5% nonfat dry milk and then incubated overnight at 4°C with primary antibodies. The membranes were rinsed three times in TBS and incubated with secondary HRP-conjugated antibodies for 1 hour at room temperature. An enhanced chemiluminescence (ECL) method (GE Healthcare) was used for antibody detection. To distinguish between soluble and insoluble proteins, cells were washed with PBS and resuspended in 1% NP40 buffer. Cell extracts were then centrifuged at 14,000 × g for 10 minutes, the soluble fraction collected, and the pellets resuspended with 1× Laemmli SDS sample buffer as the insoluble fraction.

For immunoprecipitation (IP), cells were lysed in IP lysis buffer (Thermo Fisher Scientific) containing 20 mmol/L Tris-HCl, pH 7.5, 150 mmol/L NaCl, 0.1 mmol/L EDTA, 1% Nonidet P-40, 5 mmol/L NaF, and an EDTA-free cOmplete protease cocktail tablet. Three milligrams of protein lysate were used for IP by incubation with 2 μg of the antibody for 2 hours at 4°C. A total of 30 μL of Dynabeads Protein G (#10004D, Life Technologies) was added for 1 hour at 4°C, and the IPs were washed four times with lysis buffer. Sample buffer was then added, and the beads were heated for 10 minutes at 70°C. Samples were then analyzed by SDS-PAGE followed by immunoblotting.

Cells were lysed in a solution containing 50 mmol/L Tris-HCl (pH 7.6), 300 mmol/L NaCl, 10% glycerol, 0.2% NP-40, 10 mmol/L iodoacetamide (Sigma-Aldrich), 10 mmol/L n-ethylmaleimide (Sigma-Aldrich), 0.5 mmol/L 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride (AEBSF, Roche), 10 μmol/L MG132 (Merck), and PhosStop phosphatase inhibitors (Sigma-Aldrich). The cell lysates were sonicated and centrifuged at 1.6 × 10⁴ g for 10 minutes at 4°C, and the resulting supernatant was incubated with anti-FLAG M2 agarose (Sigma-Aldrich) for 2 hours at 4°C. The resin was separated by centrifugation, washed five times with ice-cold lysis buffer, and eluted with 250 μg/mL of FLAG peptides (Sigma-Aldrich). The eluate was then treated with 0.1μ of Benzonase Nuclease (Sigma-Aldrich) for 30 minutes at 37°C. After precipitation with trichloroacetic acid, proteins were dissolved in 50 mmol/L ammonium bicarbonate (Wako), reduced in 5 mmol/L dithiothreitol (Thermo Fisher Scientific) for 5 minutes at 95°C, and alkylated with 10 mmol/L iodoacetamide (Thermo Fisher Scientific) for 20 minutes at room temperature. Reduced and alkylated proteins were digested overnight with 5 μg of Trypsin Gold (Promega) at 37°C with rotation. After tryptic digestion, samples were acidified with TFA and desalted by solid-phase extraction using GL-Tip GC and GL-Tip SDB (GL Sciences) before LC-MS analysis.

Desalted tryptic digests were analyzed by nanoflow ultra-HPLC (EASY-nLC 1000; Thermo Fisher Scientific) on-line coupled to the Orbitrap Elite instrument (Thermo Fisher Scientific). The mobile phases were 0.1% formic acid in water (solvent A) and 0.1% formic acid in 100% acetonitrile (solvent B). Peptides were directly loaded onto a C18 Reprosil analytical column (3-μm particle size, 75-μm i.d., and 12-cm length; Nikkyo Technos) and separated using a 150-minute two-step gradient (0%–35% in 130 minutes, 35%–100% in 5 minutes, and 100% in 15 minutes of solvent B) at a constant flow rate of 300 nL per minute. For ionization, 1.6 kV of liquid junction voltage and a capillary temperature of 200°C were used. The Orbitrap Elite instrument was operated in the data-dependent MS/MS mode using Xcalibur software (Thermo Fisher Scientific), with survey scans acquired at a resolution of 120,000 at m/z 400. The 10 most abundant isotope patterns with charges ranging from 2 to 4 were selected from the survey scans with an isolation window of 2.0 m/z and fragmented by collision-induced dissociation with normalized collision energies of 35. The maximum ion injection time for the survey and the MS/MS scans was 60 ms, and the ion target values were set to 1e6 for the survey and MS/MS scans. Ions selected for MS/MS were dynamically excluded for 60 seconds for binding protein identification.

Proteome Discoverer software (version 1.4; Thermo Fisher Scientific) was used to generate peak lists. The MS/MS spectra were searched against the UniProt Knowledgebase (version 2015_09) using the SequestHT search engine. The precursor and fragment mass tolerances were set to 10 ppm and 0.6 Da, respectively. Methionine oxidation, protein amino-terminal acetylation, Asn/Gln deamidation, Ser/Thr/Tyr phosphorylation, and diglycine modification of Lys side chains were set as variable modifications, and Cys carbamidemethyl modification was set as a static modification for database searches. Peptide identification was filtered at 1% FDR. To identify binding proteins, the results of samples (cells expressing FLAG-TRIM29 or not) were assembled into one multiconsensus report using Proteome Discoverer software. Cumulative protein scores were compared on the basis of the protein sequence coverages and the total numbers of identified sequences.

Immunofluorescence staining was performed as described previously (19). Cells were seeded on 8-well chamber culture slides (Lab-Tek II, Thermo Fisher Scientific). After appropriate treatment or transfection, the cells were fixed with 4% paraformaldehyde (Wako) for 20 minutes at room temperature and permeabilized with 0.5% Triton-X100 in PBS buffer for 15 minutes. The cells were incubated overnight at 4°C with primary antibodies diluted in PBS with 3% BSA. After three 10-minute washes in PBS, the cells were incubated for 60 minutes at room temperature with fluorochrome-conjugated secondary antibodies. Culture slides were washed three times for 10 minutes in PBS. For imaging, cells were stored in Vectashield hard set mounting medium with DAPI (H-1500, Vector Laboratories). Cell imaging was accomplished with a BZ-9000 microscope (Keyence) or a FV-1000 confocal microscope (Olympus).

All animal experiments were approved by the Institutional Animal Care and Use Committee of Hokkaido University (Hokkaido, Japan; 17–0100). In in vivo imaging experiments, luciferase-tagged A431 cells (A431-LUC) with or without shRNA targeting TRIM29 cells (5 × 10⁵ cells, GFP-tagged) were injected into the tail vein of athymic (nu/nu) mice, and metastatic incidence was determined using an IVIS imaging system over time. On the final day (day 23), we assessed the number of metastases in all extracted lungs by GFP channel using fluorescence microscopy (BZ-9000, Keyence). Metastatic colonies in the lungs were detected by GFP expression in fresh tissues. Similar experiments were performed using GFP-tagged DJM1 cells containing shRNA (control or targeting-TRIM29-#2).

Live animal images were captured using the IVIS system, Xenogen (Caliper Life Sciences). d-Luciferin potassium salt (VivoGlo Luciferin, Promega) was injected intraperitoneally (300 mg/kg) and animals were anesthetized in an oxygen-rich induction chamber with 2% isoflurane (Abbie). Images were collected 10 minutes after d-luciferin injection using the IVIS System (exposure time: 2 minutes). Photons were quantified using Living Image Software (Xenogen).

Dewaxed tissue sections (4.0–5.0 μm) were immunostained as reported previously using primary antibodies (22). The application of the primary antibody was followed by incubation with goat anti-mouse or rabbit polymer-based EnVision-HRP-enzyme conjugate (Dako, Japan). DAB chromogen was applied to yield a brown color. For the evaluation of IHC staining, the total score (value from 0 to 6) was calculated by measuring the staining intensity (negative = 0, weak = 1, moderate = 2, strong = 3) plus the proportion of immunopositive tumor cells (0% = 0, 1%–25% = 1, 26%–50% = 2, >50% = 3).

For cutaneous SCC, Bowen's disease, verruca vulgaris, seborrheic keratosis, keratoacanthoma (KA), KA-like SCC, and HN-SCC analyses, specimens were procured retrospectively under an Institutional Review Board–approved protocol at the Hokkaido University Hospital (Hokkaido, Japan), which deemed this retrospective analysis appropriate for a waiver of informed consent (016–0435, 017–0263). These studies were conducted in accordance with the Declaration of Helsinki. Thirty-six patients with SCC, 10 patients with Bowen's disease, 5 patients with verruca vulgaris, 5 patients with seborrheic keratosis, 4 patients with KA, 6 patients with KA-like SCC, and 35 patients with HN-SCC were enrolled. Normal epidermis was harvested from the patients with SCC (N = 29). The patients’ attributes are shown in Supplementary Table S1.

To identify the methylation status of CpG islands in the TRIM29 promoter, published methods were used (10). Briefly, a total of 0.5 μg of genomic DNA extracted from each frozen tissue sample using the QIAmp DNA Mini Kit (Qiagen) was treated with the EZ DNA Methylation-Gold Kit (Zymo Research) for bisulfate reaction. The promoter region of the TRIM29 genes was amplified by PCR using the following gene-specific primers.

Left primer: 5′-TTAGGTGGGGTTTGAGATGTAGT-3′

Right primer: 5′-CCAACTAAAAACTACCAAAAAACCA-3′

PCR products were cloned into the pCR4-TOPOVector (Life Technologies). To define the methylation status of the TRIM29 promoter region, six SCC samples and two normal epidermal tissues were sequenced with the 3130xl Genetic Analyzer (Applied Biosystems).

Tissue-specific gene expression data were downloaded from the Human Protein Atlas (https://www.proteinatlas.org/; ref. 23).

Statistical analysis was performed using the Excel add-in software Statcel (OMS Ltd.). Means and SD were calculated statistically from three determinations. The data are expressed as mean ± SD. We used t tests (Student or Welch t) to assess the statistical significance of differences between various samples. Kaplan–Meier survival curves were calculated for the two groups (high TRIM29 or low TRIM29), and the log-rank test was used to compare OS. Fisher exact test was used in the analysis of contingency tables. P < 0.05 was considered significant.

To clarify the kinds of tissues in which TRIM29 is highly expressed, we accessed the Human Protein Atlas public database (https://www.proteinatlas.org/). Investigation using public databases revealed that TRIM29 is highly expressed in stratified epithelial cells, including skin, esophagus, and tonsil (Supplementary Fig. S1). On the basis of the previous reports on the roles of TRIM29 in certain cancers, we assessed the levels of TRIM29 expression in primary cutaneous SCC tumor specimens by IHC. Expression levels of TRIM29 were high in normal epidermis and were markedly lower in cutaneous SCC (Fig. 1A). Similar results were observed in Bowen's disease (Supplementary Fig. S2A). We further examined the expression of TRIM29 in patients with metastatic cutaneous SCC. The expression levels of TRIM29 were lower in the samples of lymph node metastasis than in those of primary tumors (Fig. 1B). Moreover, the expression levels of TRIM29 in benign tumors (seborrheic keratosis, verruca vulgaris, and keratoacanthoma) were high (Supplementary Fig. S2B and S2C; Supplementary Fig. S3). Also, we assessed the levels of TRIM29 in HN-SCC (primary tongue cancer, N = 35). TRIM29 expression levels in HN-SCC tumors were lower than in adjacent epithelium (Fig. 1C–G). In the cases with lower expression of TRIM29 in primary tumors (N = 23), the expression level of TRIM29 in invasive tumors was even lower than that in microinvasive lesions (Fig. 1E, F, and H). The prognostic value of TRIM29 was assessed in the patients with stage III/IV tongue tumors. Low TRIM29 protein levels significantly correlated with lower OS (Fig. 1I).

Next, we assessed the mRNA expression of TRIM29 in cutaneous SCC samples. mRNA expression of TRIM29 in six fresh-frozen cutaneous SCC tumors was lower than in normal epidermis (Supplementary Fig. S4A). Furthermore, we analyzed epigenetic silencing of TRIM29 using frozen cutaneous SCC samples in bisulfite sequence testing, which revealed that aberrant DNA methylation in the CpG island of TRIM29 was observed in all of the cutaneous SCC samples, whereas very few methylations were detected in normal epidermis (Supplementary Fig. S4B–S4D). Also, we assessed the DNA methylation of cultured cells. DNA methylation in the CpG lesion of TRIM29 was observed in A431 SCC cells, whereas none was detected in normal keratinocytes or in HaCaT cells (Supplementary Fig. S5A). We treated A431 cells with the DNA-demethylating drug 5-azacytidine for 10 days and harvested RNA and genomic DNA. qRT-PCR analysis revealed a 5- to 7-fold rise in TRIM29 transcript (Supplementary Fig. S5B). Consistent with gene reexpression, 10 days of 5-azacytidine produced a decrease in TRIM29 methylation of A431 cells (Supplementary Fig. S5C).

To investigate the effect of TRIM29 knockdown on cutaneous SCC cells, we performed gene knockdown experiments using two shRNAs. qRT-PCR and immunoblotting confirmed the knockdown of mRNA and protein levels by both of the TRIM29-targeting shRNAs in A431 SCC cells (Fig. 2A). Through direct cell-counting and MTT proliferation assays, we determined that the growth rates of the TRIM29 knockdown cells did not differ significantly from those of the control cells (Supplementary Figs. S6A–S6C and S7A and S7B). In the cell migration assay, we observed greater cell migration into the wound in the TRIM29 knockdown A431 cell lines than in the control lines (Fig. 2B). Conversely, the cell migration of A431 lines overexpressing TRIM29 was lower than that of mock A431 lines (Fig. 2C and D). Similar results were obtained for the SCC DJM-1 and immortalized HaCaT keratinocytes (Supplementary Figs. S8A–S8D and S9A–S9D). Invasion assays using A431 cells with or without A431 knockdown were performed using a Matrigel chamber assay, which showed that TRIM29 knockdown enhanced the invasion activity (Fig. 3A). The cell invasion of A431 cells overexpressing TRIM29 was lower than those of mock cell lines (Fig. 3B). Similar results were obtained for DJM1, HaCaT cells, and oral SCC SAS cells (Supplementary Figs. S10A–S10D and S11A and S11B). To extend these studies to an in vivo context, we used luciferase-tagged A431 SCC cells (A431-LUC) containing control shRNA or TRIM29-shRNA (#2) in a xenograft model. Immunocompromised (nu/nu) mice were injected intravenously with A431-LUC cells, and lung metastases were observed over time. A live-animal imaging system (IVIS) revealed that TRIM29 knockdown enhanced the cancer cell metastasis to the lungs (Fig. 3C and D). Extracted lungs also showed that the number of tumor metastases was greater for the TRIM29 knockdown A431 cells than for the control cells (Fig. 3E and F). Similar results were observed in the in vivo metastasis model using GFP-tagged DJM1 cells (Supplementary Fig. S12A and S12B).

To clarify the molecular function of TRIM29, we first assessed the subcellular localization of TRIM29. Immunofluorescence staining of TRIM29 revealed that TRIM29 was mainly expressed in the cytosol, where the expression patterns were filamentous or perinuclear (Fig. 4A–C). Also, immunoblotting using whole-cell lysates (1× Laemmli buffer) and 1%NP40 lysate showed that TRIM29 mainly presents in a detergent-insoluble fraction (Fig. 4D). Next, we established HaCaT and A431 cells that stably expressed FLAG-tagged TRIM29. To detect the TRIM29-interacting protein, we immunopurified FLAG-TRIM29 and TRIM29-associated proteins from cell lysates (Fig. 4E) and performed mass spectrometry analysis. From screening by mass spectrometry analysis, 67 proteins were identified as interacting specifically with FLAG-TRIM29 (Supplementary Fig. S13A and S13B). IP analysis using HaCaT cell lysates revealed that endogenous TRIM29 interacts with FAM83H and keratins (Fig. 4F–H). Furthermore, immunofluorescence analysis using confocal microscopy showed that TRIM29 colocalizes with keratin 5, 14, and FAM83H in the cytosol (Fig. 4I–K; Supplementary Figs. S14A–S14E and S15). These results indicate that the TRIM29 forms a complex with keratins and keratin-interacting proteins. To identify the domains of TRIM29, which are necessary for interaction with keratins/FAM83H, we generated HaCaT cells that stably express the series of truncated mutants of FLAG-TRIM29 (Supplementary Fig. S16A). IP/SDS-PAGE analysis revealed that the zinc finger, B-box, coiled-coil, and C-terminal domains of TRIM29 were necessary for the formation of the TRIM29–keratin–FAM83H complex (Supplementary Fig. S16B–S16D), which was consistent with the findings of immunofluorescence staining (Supplementary Fig. S16E).

To clarify the physiologic role of TRIM29 in stratified epithelial cells, we performed siRNA transfection experiments. HaCaT cells were reverse-transfected with pooled siRNA targeting TRIM29 or a negative control. Immunoblotting revealed that TRIM29 knockdown did not alter the expression levels of keratin 5 (Fig. 5A). Immunofluorescence staining showed that TRIM29 knockdown led to a perinuclear distribution of keratins, whereas control cells showed cytosolic diffuse keratin expression (Fig. 5B). We also established HaCaT keratinocytes that stably contained HA-tagged TRIM29. The lysates from HaCaT cells containing HA-tagged TRIM29 did not show altered expression levels of keratin 5, either (Fig. 5C). Confocal immunofluorescence images showed that forced-expressed TRIM29 led to the diffuse or unbiased distributions of keratins, which colocalized with TRIM29 (Fig. 5D). These data suggest that TRIM29 regulates the cellular distribution of keratins.

We further examined the cutaneous SCC samples on the basis of the keratin staining patterns (Supplementary Fig. S17). In adjacent normal epidermis, keratin staining patterns were diffuse. Also, SCC samples with high TRIM29 expression showed relatively diffuse keratin patterns. In turn, cutaneous SCC with low TRIM29 expression showed perinuclear or random “nondiffuse” keratin patterns (Fig. 6A). “Nondiffuse” distribution patterns were significantly more common in low-TRIM29 tumor tissue than in high-TRIM29 tumors (Fig. 6B). Similar results were observed in HN-SCC samples (Fig. 6C and D). Moreover, we analyzed the expression of keratins in an in vivo lung metastasis model. TRIM29 knockdown altered the keratin expression of lung metastatic tumors, whereas control xenograft tumors showed relatively diffuse expression of keratins (Supplementary Fig. S18). These findings indicate that the loss of TRIM29 correlates with the altered distribution of keratins.

We herein reported that TRIM29 plays a crucial role in cutaneous and HN-SCC cell migration and invasion by regulating keratin distribution (Fig. 7). Clinically, the loss of TRIM29 was observed in only malignant tissues, and not in normal epithelium or benign tumors. The expression levels of TRIM29 in invasive or advanced lesions showed greater decreases than in primary tumors. These results strongly suggest that TRIM29 is a novel diagnostic and prognostic biomarker for stratified epithelial tumors.

In this study, IHC and immunofluorescence analyses revealed that TRIM29 localizes mainly in the cytoplasm. Previous studies suggested that cytosolic TRIM29 has oncosuppressive effects in breast and prostate cancers (10–12), whereas nuclear TRIM29 has prooncogenic effects in pancreatic, lung, bladder, and gastric cancers (9, 24–27). Thus, the localization (cytosolic or nuclear)-related physiologic function of TRIM29 was determined by the kinds of cells/tissues. Our results indicate that TRIM29 localizes mainly in the cytoplasm in stratified epithelial cells, which is consistent with TRIM29 having an oncosuppresive role in such cells.

Keratins are the intermediate filament-forming proteins of epithelial cells. Keratin filaments play a role in the formation of an insoluble structural framework within the cytoplasm, which plays a critical role in cell protection. Today, keratins have been recognized as regulators of other cellular functions, including polarization, motility, and signaling (28). Furthermore, keratins are involved in cancer cell invasion and metastasis, as well as in treatment responsiveness. It has been reported that keratin distribution regulates cell migration in cancer cells as well as in normal cells. In injury experiments using human volunteers, injury to human epidermis prompted the keratin filaments to reorganize from a diffuse cytoplasmic distribution to a perinuclear distribution, which lead to keratinocyte migration to the wound (29). Furthermore, human epithelial tumor cells including pancreatic cancers treated with sphingosylphosphorylcholine (SPC) show the reorganization of keratins to perinuclear areas and enhanced cancer cell migration (30, 31). A study by Seltmann and colleagues using keratin-knockout mice showed that keratin-knockout keratinocytes migrate faster than wild-type keratinocytes (32). These results suggest that the absence of expression (or the juxtanuclear localization) of keratins could facilitate cell migration. In the SPC model, the phosphorylation of keratins by the MEK–ERK pathway is thought to be a molecular mechanism in keratin reorganization. In our study, we analyzed the phosphorylation of keratins (K5 and K14) using lysates of HaCaT with or without TRIM29 knockdown; however, phos-tag SDS-PAGE (Wako) found no obvious differences. Also, we analyzed the dimerization of keratins (K5, K14) through native page analysis (33), which showed no obvious differences between TRIM29 knockdown and control HaCaT cells. Thus, the detailed mechanism of TRIM29-mediated keratin reorganization remains unknown.

In this study, keratin-interacting protein FAM83H was detected as a novel interactor with TRIM29. Recently, the Kuga group discovered that FAM83H is involved in certain cancer cell migrations via the regulation of keratins. High expression of FAM83H in colorectal cancer cells causes casein kinase 1 alpha to be recruited to keratins and disrupts the keratin cytoskeleton, which leads to cell migration/invasion (34, 35). On the basis of these reports, we investigated the FAM83H expression in primary cutaneous SCCs samples and found no marked differences in FAM83H protein expression. Also, in cell cultures, TRIM29 knockdown or forced expression did not change the expression levels of FAM83H (Fig. 5A and C). At this point, it remains unclear whether FAM83H is directly associated with the interactions between TRIM29 and keratins in stratified epithelial cells.

To date, TRIM29 has been reported to be positively or negatively involved in epithelial–mesenchymal transition (EMT) in cancers (10, 36). RNAi-mediated gene knockdown of TRIM29 in breast cancer cell lines increases the expression of mesenchymal markers (N-cadherin and vimentin), decreases the expression of epithelial markers (E-cadherin), and increases the expression of the oncogenic transcriptional factor TWIST1, which is a key driver of EMT (10). Another study used epigenetic change analysis to identify TRIM29 as a driver candidate during EMT (36). Furthermore, experimental model mice have revealed that TRIM29 facilitates EMT in pancreatic cancer cells (9). On the basis of these studies, we also assessed the protein expression of EMT-related molecules including SNAIL2, TWIST1/2, FAK, p-FAK, and p63 using the lysate of TRIM29 knockdown cells, in which obvious alterations of protein expression were not observed. As such, TRIM29 knockdown–mediated migration/invasion in stratified epithelial cells may be associated with keratin distribution/reorganization, rather than with the EMT process.

In conclusion, we revealed an unrecognized role of TRIM29 in the cell migration/invasion of cutaneous and HN-SCC tumors. These results suggest that TRIM29 could be a novel diagnostic/prognostic marker in stratified epithelial tumors. Also, the TRIM29–keratin axis may be a novel therapeutic target in intractable/metastatic SCCs.

No potential conflicts of interest were disclosed.

Conception and design: T. Yanagi, S. Hatakeyama

Development of methodology: T. Yanagi, M. Watanabe

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Yanagi, M. Watanabe, H. Hata, S. Kitamura, K. Imafuku, H. Yanagi, A. Homma, L. Wang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Yanagi, M. Watanabe, A. Homma, S. Hatakeyama

Writing, review, and/or revision of the manuscript: T. Yanagi, M. Watanabe, A. Homma, S. Hatakeyama

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Homma, H. Takahashi, S. Hatakeyama

Study supervision: A. Homma, H. Shimizu, S. Hatakeyama

We thank Ms. Yuko Tateda and Dr. Kanae Takahashi for their technical assistance. This work was supported in part by KAKENHI (#15H05998, #16K1970106, #18K08259 to T. Yanagi; #17K19506, #18H02607 to S. Hatakeyama) from the Ministry of Education, Culture, Sports, Science and Technology in Japan, and by the Ichiro Kanehara Foundation for the Promotion of Medical Science (to T. Yanagi), the Cosmetology Research Foundation (to T. Yanagi), the Ono Cancer Research Foundation (to T. Yanagi), the Takeda Science Foundation (to T. Yanagi and S. Hatakeyama), the Suhara Memorial Foundation (to T. Yanagi), the Geriatric Dermatology Foundation (to T. Yanagi), the Uehara Memorial Foundation (to T. Yanagi), the Pias Skin Research Foundation (to T. Yanagi), the Life Science Foundation of Japan (to S. Hatakeyama), the Princess Takamatsu Cancer Research Fund (to S. Hatakeyama), the Japan Foundation for Applied Enzymology (to S. Hatakeyama), the Tokyo Biochemical Research Foundation (to S. Hatakeyama), and the Project Mirai Research Grants (to S. Hatakeyama).

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.