All mice were maintained in temperature-controlled, air-conditioned, and specific pathogen-free conditions on a 12-h light/dark cycle, and handled in accordance with the Guidelines for Animal Experiments of the Institute of Medical Science (University of Tokyo, Tokyo, Japan). Mice were given access to a standard laboratory diet and water. The animal experiment was approved by the institutional review board of the University of Tokyo (A13-03). p53^−/− mice were provided from the RIKEN BioResource Center (Ibaragi, Japan). Mice were treated with 10 Gy of X-ray irradiation, and tissues were subjected to RNA-seq as described in our previous studies (11,12). Briefly, p53^+/+ and p53^−/− mice were X-ray-irradiated using the MBR-1520R-3 system (Hitachi, Tokyo, Japan), and 24 tissues were collected from each mouse at the 24-h time-point following irradiation (bladder, bone marrow, cerebrum, colon, epididymis, esophagus, eyeball, heart, kidney, liver, lung, muscle, seminal vesicle, small intestine, spleen, stomach, testis, thymus and tongue: male, 6 weeks old, n=3 each; skull, cartilage/knee: male, 1 week old, n=3 each; uterus: female, 10 weeks old, n=3 each; mammary gland and ovary: female, 10 weeks old, n=2 each). Total RNA was recovered from the homogenised tissues using the RNeasy Plus Universal Mini kit or RNeasy Plus Mini kit (Qiagen, Valencia, CA, USA). The samples were selected by the evaluation of the RNA quality and quantity using the Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and Nanodrop (Thermo Fisher Scientific, Waltham, MA, USA). High-quality RNA was subjected to polyA+ selection and chemical fragmentation, and a 100-200-base RNA fraction was used to construct complementary DNA libraries according to Illumina’s protocol. RNA-seq was performed with HiSeq 2500 using a standard paired-end 101-bp protocol. The raw data for the RNA-seq are available on the DNA Data Bank of Japan (DDBJ) database (http://www.ddbj.nig.ac.jp/index-e.html) under the accession no. DRA005768, and bioproject accession no. PRJDB5738. The cDNA microarray of MCF10A and HCT116 cells was performed using methods described in previous studies (12-14). Briefly, the MCF10A and HCT116 p53^+/+ or p53^−/− cells were treated with adriamycin (ADR) for 2 h and cultured with fresh medium until harvest. The concentration of ADR applied was 0.5 and 2 μg/ml for the MCF10A and HCT116 cells, respectively. Total RNA was isolated from the cells using the RNeasy Plus Universal Mini kit (Qiagen) at 0, 12, 24 and 48 h following ADR treatment. Each RNA sample was labeled and hybridised to array slides. Gene expression analysis was performed using a SurePrint G3 Human GE 8×60K microarray (Agilent Technologies) according to the manufacturer’s instructions. The microarray data are available from the NCBI GEO database. The accession no. is GSE98727 for the MCF10A cells and GSE125787 for the HCT116 cells.

In the mouse transcriptome data, genes with a low expression were defined as those with a maximum expression in all groups (W, p53^+/+; WX, irradiated p53^+/+; K, p53^−/−; KX, irradiated p53^−/−) among all 24 tissues inferior to 1 fragment per kilobase million (FPKM), and were excluded from the screening. Subsequently, the gene expression levels in mice in the WX group (n=2 for the mammary gland, ovary and uterus; n=3 for other organs) were compared to the gene expression levels in mice in the W, K and KX groups (same ‘n’ as WX for each group, resulting in a total of n=6 for the mammary gland, ovary and uterus; n=9 for other organs). The expression fold change was calculated with the following formula: Median expression in (WX)/maximum of median expression in (W), (K) or (KX). Genes with a significant expression fold change >3-fold were extracted. One-way ANOVA with a Dunnett’s post hoc test was performed to calculate the P-values between the WX group and each of the other groups (W, K or KX). In the final step of the screening process, genes with a significantly upregulated expression in the WX group in no less than 10 of 24 tissues were selected. In the human cell transcriptome data, the expression of ADR-treated p53^+/+ cells (n=3) was compared to the group gathering p53^−/− cells and untreated p53^+/+ cells (n=5). The expression fold change was calculated by the following formula: Median expression of p53^+/+ cells (ADR 12 h), (ADR 24 h) and (ADR 48 h)/maximum expression in p53^−/− cells (0 h), (ADR 12 h), (ADR 24 h), (ADR 48 h), and p53^+/+ cells (0 h). Genes with an upregulated expression in the ADR-treated p53^+/+ cells were defined by a significant expression fold change >1.5-fold. The two-tailed Student’s t-test was used to calculate the P-values.

HCT116 (colorectal adenocarcinoma), U2OS (osteosarcoma) and H1299 (lung adenocarcinoma) cells were purchased from the American Type Cell Collection (Manassas, VA, USA) (cat. nos. CCL-247, HTB-96 and CRL-5803, respectively). The HCT116 p53^+/+ and HCT116 p53^−/− cells were gifts from Dr B. Vogelstein (Johns Hopkins University, Baltimore, MD, USA). MCF10A (breast epithelial cells) p53^+/+ and p53^−/− cell lines were purchased from Sigma-Aldrich (St. Louis, MO, USA). HBC4 (breast cancer) cells were a gift from Dr T. Yamori (The Cancer Institute of Japanese Foundation for Cancer Research, Tokyo, Japan). 293T cells (subclone of 293 cells expressing SV40 large T-antigen) were purchased from the RIKEN Cell Bank (Ibaraki, Japan). All cells were cultured in a 37°C incubator with a 5% CO₂ atmosphere. The siRNA oligonucleotides were ordered from Sigma-Aldrich and transfected into the cells using Lipofectamine RNAiMAX (Thermo Fisher Scientific) at the seeding step. The sequences are listed in Table SI. Gene overexpression experiments were performed by transfecting plasmids into cells with FuGENE6 (Promega, Madison, WI, USA). The cells were treated with ADR for 2 h at the 24-h time-point after seeding to induce DNA damage and p53 expression. The U2OS and HBC4 cells were treated with 2 μg/ml of ADR. Total RNA was extracted or whole cell lysates were prepared 48 h following treatment for reverse transcription-quantitative PCR (RT-qPCR) or western blot analysis, respectively. Cell lines stably expressing INKA2 were generated by transfecting the HCT116 cells with pLenti6.3/TO/V5-DEST/INKA2-CDS constructed as previously described (14). Briefly, the full-length INKA2 cDNA fragment was integrated into the destination vector pLenti6.3/TO/V5-DEST using the Gateway LR Clonase II enzyme mix, all provided by Invitrogen/Thermo Fisher Scientific. The plasmid contains the selectable marker G418 and the Blasticidin S-resistance gene. Transfected cells were selected by the addition of Blasticidin and Geneticin^® to the culture medium for 3 weeks.

Following the treatments, total RNA was extracted from the cells using an RNeasy Plus Mini kit (Qiagen) according to the manufacturer’s instructions. SuperScript III reverse transcriptase (Invitrogen/Thermo Fisher Scientific) was used to synthesise cDNA from the collected RNA samples. The resulting cDNA was employed as template for qPCR using SYBR-Green Master Mix and a Light Cycler 480 machine (Roche Applied Science, Basel, Switzerland). The expression level of the housekeeping gene, GAPDH, was used for normalization. The primer sequences are presented in Table SI.

The cells were plated in 24-well culture plates with glass coverslips on the bottom. Cells were transfected with the siRNAs at the same time as the seeding step. For drug treatments, 2 μg/ml of ADR or 1 μg/ml of doxycycline (Dox) were added at 24 h after plating. The ADR treatment lasted for 2 h and Dox remained in the medium until the cells were harvested. At 48 h after the addition of the drugs, the cells were fixed with 4% paraformaldehyde and permeabilised with 0.2% Triton X-100 in PBS. A 3% BSA solution was applied in the blocking step, and the cells were incubated with antibodies in the same blocking solution. The antibodies were diluted and used for incubation according to each manufacturer’s protocol. Details for the antibodies used in the present study are presented in Table SII. Finally, the samples were mounted in Vectashield Mounting Medium with DAPI (Vector Laboratories, Burlingame, CA, USA) and kept in 4°C until photography. A confocal microscope (Olympus FluoView FV1000) was used to visualise the mounted cover-slips.

ChIP data targeting human and mouse p53 proteins were downloaded from the ReMap 2018 (15) and ChIPBase (16) databases, respectively. Overlapping peaks were merged into a single peak for analysis. The DNA sequence from -10 to +10 kb of the INKA2 gene was submitted to the binding motif scanning tool in the JASPAR database (17) to identify the candidate p53 binding sites. Binding scores were calculated and provided by the database. p53 binding sites reported for human (18) or mouse (19) CDKN1A were subjected to JASPAR binding site scanning analysis to confirm the authenticity of the prediction (Table SIII). Likewise, the p53 binding motif used from JASPAR (MA0106.3) was compared with a refined consensus p53 binding motif of high prediction accuracy (18). Large resemblance was observed between the two models, and both were based on the known p53 binding motif composed of two repeated palindromic motif RRRCWWGYYY (Fig. S1).

Whole-cell lysates were prepared from adherent cells with RIPA buffer (Thermo Fisher Scientific) containing 1 mM PMSF, 0.1 mM DTT and 0.1% Protease Inhibitor Cocktail Set III (Calbiochem, La Jolla, CA, USA). Cell lysates were sonicated 15 times for 30 sec each using a Bioruptor UCD-200 (Cosmo Bio, Tokyo, Japan). The protein concentrations of the resulting lysates were measured using the BCA™ protein assay (Thermo Fisher Scientific). Fifteen micrograms of protein samples were separated using 10 or 12% SDS-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes (Hybond™ ECL™; Amersham, Piscataway, NJ, USA). For the blocking step, we incubated the membranes with 5% non-fat dry milk in 0.05 or 0.1% Tween-20 TBST, as recommended by each antibody’s manufacturer, for 1 h at room temperature. After washing, the proteins were probed with specific primary antibodies overnight at 4°C. On the following day, the membranes were incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies (Santa Cruz Biotechnology, Santa Cruz, CA, USA) for 1 h at room temperature. Details for the antibodies used in this study are provided in Table SII. Proteins were detected by chemiluminescence (ECL, Amersham™, or Immobilon; Millipore, Billerica, MA, USA) with the LAS-4000 mini machine (GE Healthcare, Freiburg, Germany). The bands of the targeted proteins from the obtained images were quantified with ImageJ software (20) and were normalized to the quantified signal intensity of β-actin.

DNA fragments carrying p53-responsive elements, as predicted by the JASPAR database, were amplified by PCR and cloned into the pGL4.24 [luc2P/minP] plasmid (Promega) (primes are presented in Table SI). The known p53 binding site in the promoter region of CDKN1A was used as positive control for human binding site constructs: GAACATGTCCCAACATGTTG (18). For point mutations, the 6th, 13th and 16th nucleotides of the binding site were changed to ‘T’ using site-directed mutagenesis. The H1299 cells were co-transfected with the pGL4.24 control vector or pGL4.24 containing the target p53 binding site, accompanied by the empty vector pcDNA3.1, wild-type p53, or mutant p53 (p53R175H for human p53 and p53R172H for mouse p53) expression vector. All samples were co-transfected with pGL4.74 to normalise the transfection efficiency. Luciferase activities were detected at 48 h following transfection using the Dual-Glo Luciferase Assay System (Promega) according to the manufacturer’s instructions.

The full-length INKA2 cDNA with stop codon was cloned into the PCAGGS vector containing an HA-tag (pCAGGSnHC) (primers are presented in Table SI). Cells were seeded (densities of 10,000 cells/well and 5,000 cells/well for HBC4 and U2OS cells, respectively) and grown in 6-well culture plates. On the following day, the PCAGGS/INKA2 or the PCAGGSnHc mock plasmid was transfected into the cells. The U2OS and HBC4 cells were cultured in the presence of 1.0 or 0.4 mg/ml Geneticin^® (Wako Pure Chemical Industries, Ltd., Osaka, Japan), respectively, for 1 to 2 weeks. Colonies were washed with PBS, fixed with methanol at room temperature for 2 h, and were counted with ImageJ software (20) following overnight staining at room temperature with crystal violet (Sigma-Aldrich). The two-tailed Student’s t-test was used to calculate the P-values for the differences between the mock and INKA2-expressing replicates (n=3).

The full-length encoding INKA2 and PAK4 cDNAs were cloned into PCAGGS vectors containing a multiple cloning site and HA-tag (pCAGGSnHC) or FLAG-tag (pCAGGSn3FC) (primers are presented in Table SI). Plasmids were transfected into 293T cells to induce the overexpression of the target genes. Whole-cell lysates were collected at 48 h following transfection with a cell lysis buffer consisting of 20 mM Tris (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 1 mM PMSF and 0.1% Protease Inhibitor Cocktail Set III (Calbiochem). Samples were sonicated 3 times for 5 sec each using a Bioruptor UCD-200 (Cosmo Bio). A total of 45 μl of the supernatant in each sample was aliquoted as the input sample. Sonicated cell lysates were precleared by incubation with Protein G-Sepharose 4B (Invitrogen/Thermo Fisher Scientific) and normal mouse IgG (1 μg/sample, sc-2027; Santa Cruz Biotechnology) for 1 h at 4°C. Precleared cell lysates were immunoprecipitated with mouse monoclonal anti-HA antibody-conjugated agarose beads or mouse monoclonal anti-FLAG antibody-conjugated agarose beads (25 μg/sample, cat. nos. A2095 and A2220; Sigma-Aldrich). The obtained samples were washed 4 times with 1 ml of ice-cold cell lysis buffer (without PMSF and Protease Inhibitor) and then boiled for 2 min following the addition of Laemmli Sample Buffer (Bio-Rad Laboratories, Hercules, CA, USA). For deletions, the RNRQPLVLGDN (150-160) residues of the INKA2 iBox domain and the catalytic domain (326-591) of PAK4 were deleted by site-directed mutagenesis.

The pCAGGS mock plasmid or pCAGGS/INKA2-HA plasmid were transfected into the 293T cells on the 2nd day after plating in a 10-cm dish. Whole-cell lysates were collected and immunoprecipitated with an antibody against HA-tag (Table SII). The resulting lysates were subjected to mass spectrometry analysis as previously described (21). Briefly, the samples were reduced in 10 mM tris(2-carboxyethyl)phosphine with 50 mM ammonium bicarbonate (both from Sigma-Aldrich) for 30 min at 37°C and alkylated in 50 mM iodoacetamide (Sigma-Aldrich) with 50 mM ammonium bicarbonate for 45 min in the dark at 25°C. Trypsin/Lys-C (Promega) solution was added followed by incubation at 37°C for 12 h. The resulting peptides were extracted from the gel fragments and analysed with an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific) combined with the UltiMate 3000 RSLC nano-flow HPLC system (Thermo Fisher Scientific) in HCD MS/MS mode. The MS/MS spectra were searched against the Homo sapiens protein sequence database in SwissProt using the Mascot or Sequest search engine in the Proteome Discoverer 2.2 software (Thermo Fisher Scientific), in which peptide identification filters were set at ‘false discovery rate <1%’ and ‘Mascot expectation value <0.05 or Sequest XCorr >2.0’.

Cells stably expressing mock vector or INKA2 were cultured in medium with 1 μg/ml of Dox for at least 48 h before seeding. Cells were seeded on 24-well plates and were cultured for 4 days. Cell viability was evaluated by Cell Titer-Glo Luminescent Cell Viability assay (Promega) at the 24-, 48-, 72- and 96-h time-points according to the manufacturer’s instructions. Briefly, tge culture medium was removed and cells were incubated with a mix of 100 μl of Cell Titer-Glo reagent and 100 μl of culture medium for 10 min. The luminescence of the cell lysates was measured using the ARVO X3 plate reader (Perkin-Elmer, Waltham, MA, USA).

Whole-transcriptome sequencing data, transcriptome profiling microarray data, and p53 mutation status were obtained from TCGA database using cBioPortal (22,23). We downloaded and analysed datasets for 20 cancer cohorts: Bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), cervical and endocervical cancers (CESC), cholangiocarcinoma (CHOL), colorectal adenocarcinoma (COAD/READ), glioma (GBM/LGG), head and neck squamous cell carcinoma (HNSC), pan-kidney cohort (KIPAN), acute myeloid leukaemia (LAML), liver hepatocellular carcinoma (LIHC), lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), ovarian serous cystadenocarcinoma (OV), pancreatic adenocarcinoma (PAAD), pheochromocytoma and paraganglioma (PCPG), prostate adenocarcinoma (PRAD), skin cutaneous melanoma (SKCM), stomach adenocarcinoma (STAD), thyroid carcinoma (THCA) and uterine corpus endometrial carcinoma (UCEC). The two-tailed Student’s t-test was used to calculate the P-values for differences between the compared groups. The number of samples is presented in Table SIV. We excluded the cancer type when the number of samples was not relevant. Data regarding the DNA methylation status of the INKA2 promoter region were downloaded from the MethHC database (24).

In this study, we applied known DNA double-strand break induction, such as X-ray irradiations and ADR, to activate p53 expression and the transcription of its downstream targets. We first screened the RNA sequencing (RNA-seq) data obtained from 4 groups of mice, p53 wild-type mice (W), p53 knockout mice (K), treated with (WX and KX) or without 10 Gy of whole-body X-ray irradiation. For each mouse, the total RNA extracted from 24 tissues was subjected to RNA-seq analysis (11,12). Among 23,813 genes, we first excluded transcripts with an extremely low expression in all organs (FPKM <1). Subsequently, we compared the median expression of the WX group with the maximum median expression among the rest (K, KX or W). Transcripts displaying a significantly upregulated expression in the WX group by at least 3-fold in 10 or more organs were selected (Table SV). In total, 22 genes fulfilled our criteria, including 13 genes that were previously reported as p53 downstream targets: Mgmt (9), LIF, interleukin 6 family cytokine (Lif) (25), sestrin 2 (Sesn2) (26), D630023F18Rik (27), BCL2 binding component 3 (Bbc3) (28), cyclin G1 (Ccng1) (29), zinc finger protein 365 (Zfp365) (30), 1700007K13Rik (31), pleckstrin homology like domain family A member 3 (Phlda3) (32), proline and serine rich coiled-coil 1 (Psrc1) (33), cyclin dependent kinase inhibitor 1A (Cdkn1a) (18), growth differentiation factor 15 (Gdf15) (34) and ectodysplasin A2 receptor (Eda2r) (35). Notably, well-known p53 downstream targets, such as Cdkn1a and Eda2r, exhibited the highest number of organs satisfying the screening criteria. The rest of the resulting genes consisted of 2 non-coding RNAs, one gene without any corresponding human homolog, and 6 genes remaining as candidates: Fam212b (Inka2), 4632434I11Rik (Ddias), Celf5, 9030617O03Rik, Cd80 and Plcd4 (Table I).

We then examined the expression of the remaining 6 candidate genes in human cells using cDNA microarray performed in two different cell lines (12,13). The induction of gene expression was induced by ADR instead of irradiation for experimental convenience. The p53 wild-type (p53^+/+) or knockout (p53^−/−) HCT116 colorectal carcinoma and MCF10A breast epithelial cells were treated with ADR for 2 h. Total RNA was extracted at 4 time-points following treatment, namely at 0, 12, 24 and 48 h, and subjected to a cDNA microarray analysis. We considered the p53^−/− group and untreated samples (0 h) as a cluster. The maximum expression in the cluster was compared to the median expression of the p53^+/+ ADR-treated group (p53^+/+ ADR 12, 24 and 48 h). We selected genes displaying a significant upregulation in expression of at least 1.5-fold in the p53^+/+ ADR induction group for further investigation (Fig. 1A and B). Notably, INKA2 was the only gene which exhibited an induction in expression by ADR treatment in the p53^+/+ cells. The expression of this gene was significantly increased by 1.61- and 2.19-fold in the HCT116 and MCF10A cells, respectively (Fig. S2A).

We obtained INKA2, also known as FAM212B, as the only gene satisfying all criteria in the screen. In the mouse transcriptome data, Inka2 exhibited an increase expression in the WX group that was significantly induced >3-fold in 12 of 24 organs (Fig. 1C). Its expression was upregulated over 5-fold in the colon, liver, mammary gland and calvaria (skull), over 10-fold in the bone marrow, ovary and spleen, and over 20-fold in the kidney (Fig. S2B). The induction was insignificant or significant, but <3-fold in the other mouse tissues (Fig. S2B and C).

We further investigated the association between INKA2 and p53 expression using human cancer cell lines. Although INKA2 has 2 mRNA isoforms, the second isoform exhibited insufficient expression levels for analysis (Fig. S3). Therefore, we focused our study on isoform 1 of INKA2, the major transcript. Its expression was found with the highest induction in HBC4 human breast cancer cells and U2OS osteosarcoma cells; thus, we conducted further experiments using these two cell lines. The mRNA levels of INKA2 in both cell lines were evaluated by RT-qPCR. A significant increase in the mRNA levels of the INKA2 was observed in the cells treated with ADR. However, the expression was suppressed when p53 expression was silenced using siRNA (Fig. 2A). A siRNA against enhanced green fluorescent protein (EGFP) was used as the negative control. We then we performed a western blot analysis of U2OS cell lysates cultured under the same experimental conditions (Fig. 2B). The amount of p53 protein increased in the cells treated with ADR and decreased in the cells transfected with p53 siRNA. We used CDKN1A as the positive control. INKA2 and CDKN1A exhibited similar expression patterns. Thus, these results suggest that the mRNA and protein expression of INKA2 is regulated via a p53-dependent mechanism.

We designed 2 siRNAs targeting INKA2 to evaluate the quality of the antibody against INKA2. The results of RT-qPCR confirmed the silencing of the mRNAs by the siRNAs (Fig. 2C). The results of western blot analysis also revealed the loss of the band corresponding to INKA2 in the cells transfected with the siRNAs, indicating the specificity of the antibody (Fig. 2C). Finally, we performed immunocytochemistry to validate the induction of INKA2 expression by p53 and the repression of INKA2 by the siRNAs (Fig. 2D).

In addition to the intracellular expression, we analysed the INKA2 mRNA expression levels in tumour tissues from all cancer types of the TCGA database presenting adequate data and sample size. We observed significantly a higher INKA2 expression in p53 wild-type samples compared with p53 mutant samples in 4 cohorts: BLCA, COAD/READ, KIPAN and PRAD (Fig. 2E). Consequently, INKA2 showed a p53-dependent expression pattern at both the cellular and tissue levels, indicating that the expression of this gene is regulated by p53.

We inspected data obtained from chromatin immunoprecipitation coupled with deep sequencing (ChIP-seq) targeting the p53 protein to examine whether p53 directly regulates INKA2 expression. In human samples, the ChIP-seq results from the ReMap 2018 database (15) revealed p53 accumulation at 3 regions of the INKA2 gene: 15 and 7.5 kb upstream of the first exon of isoform 1 and throughout the first exon of isoform 1 (Fig. S4A). Further investigations using the JASPAR database (17) identified a putative p53 binding site located approximately 600 bases downstream of the first exon of isoform 1 (Fig. 3A and B). The binding site is adjacent to the INKA2 promoter (Fig. S4A) and overlaps with the third ChIP-seq peak region (Table SVI). Moreover, the ENCODE histone marks data (36) integrated in the UCSC Genome Browser (37) revealed an enrichment of H3K4me1 and H3K4me3 histone methylation and H3K27Ac histone acetylation marks in the same region, implicating chromatin remodelling and transcriptional regulation at the putative p53-responsive site (Fig. S4A).

Likewise, we identified a p53-responsive element in the mouse Inka2 sequence using ChIP-seq data of mouse embryonic fibroblasts (MEFs) treated with 0.2 μg/ml of ADR for 12 h, obtained from the ChIPBase database (16) (Fig. S4B). We then employed the JASPAR database to predict p53-responsive elements in the ChIP-seq results. Similar to the human one, the putative p53 binding site was located approximately 600 bases downstream of the first exon of isoform 1 (Fig. 3A and B). The sequence is adjacent to the Inka2 promoter (Fig. S4B) and overlaps with the ChIP-seq peak (Table SVII). The putative p53 binding sites predicted by the JASPAR database comprised 2 repeated palindromic nine-base half-sites. A similarity of 94.4% was observed between the sequences of the 2 species, only differing from the 9th nucleotide where a ‘T’ in the human sequence was replaced with a ‘C’ in the mouse sequence (Fig. 3B). In addition to the JASPAR p53 motif, the putative binding site sequence largely resembles to a high prediction accuracy consensus p53 binding motif (18) (Fig. S1). The absolute scores calculated from the position weight matrix of the JASPAR database were 9.21 and 10.85 for human and mouse, respectively (Fig. 3B). Known p53 downstream targets have been reported to disfavour spacer between the 2 half-sites of the binding site (18). The putative binding sites were consecutive half-sites and this confirmed a prediction specificity. Further inspection of the binding site sequences indicated a high level of conservation across species (Fig. S5), revealing its importance.

We then performed luciferase reporter assays in the H1299 cells expressing wild-type human p53 or the p53R175H mutant, or wild-type mouse p53 or the p53R172H mutant (Fig. 3C). The reported p53 binding site in CDKN1A was used as a positive control (18). The p53 protein interacted with the predicted binding site when both p53 and the binding site sequences were wild-type. Luciferase activity decreased when point mutations, as shown in Fig. 3B, were introduced into the responsive element sequences. Similarly, luciferase activity decreased in the presence of mutant p53. Thus, INKA2 is directly regulated by p53.

As regards the association between INKA2 and tumours, we investigated its role in cancer cell growth by performing colony formation assays using the HBC4 and U2OS cells (Fig. 4A). A total of 20 to 40% fewer colonies were formed by cells overexpressing INKA2 compared with the mock vector. Subsequently, it was concluded that INKA2 reduces cancer cell proliferation to a certain extent.

We further analysed INKA2 expression data from the TCGA database. Among all available cancer datasets, INKA2 was expressed at lower levels in the tumour samples than in the control samples in multiple cancer types: BRCA, COAD, KIRC + KIRP, LUAD, LUSC, PRAD and THCA (Fig. 4B). Additionally, the promoters of the human and mouse INKA2 genes contained CpG islands (Fig. S4). A higher DNA methylation was observed in the human INKA2 promoter region in the tumour samples than in the control samples (Fig. 4C), explaining the down-regulation of INKA2 expression in tumour samples. These results indicate that DNA methylation and p53 mutations synergistically suppresses INKA2 expression in cancer tissues. Furthermore, the p53 mutation status is associated with INKA2 expression, but not with DNA methylation. Taken together with the reduction of cancer cell growth, it is thus suggested that INKA2 may function as a cancer cell growth inhibitor.

Using an antibody against HA-tag, we performed immunoprecipitation using whole-cell lysates obtained from 293T cells overexpressing HA-tagged INKA2 to determine its function in human cells. The samples were then subjected to mass spectrometric analysis (Table II). The results revealed a high binding capacity of INKA2 to PAK4, a serine/threonine-protein kinase. In fact, INKA1, a member of the same INKA protein family, has been reported as an inhibitor of PAK4 kinase (38). Both INKA1 and INKA2 contain an iBox domain of 38 amino acids (Fig. 5A). On the other hand, PAK4 contains an auto-inhibitory domain (AID) that interacts with the substrate-docking site situated in the catalytic domain via the RPK peptide: R⁴⁹PKPLVDP (39,40) (Fig. 5A and B). The iBox domain also includes a tripeptide PLV that is shared by the RPK sequence in PAK4-AID (Fig. 5A). Thus, the iBox peptide is likely able to bind to the substrate-docking site of PAK4 in the same manner as PAK4-AID. Compared with INKA1, INKA2 maintains the iBox domain (137-174) with a high degree of sequence identity, where the essential PAK4-inhibitory motif is NRQPLVLG instead of SRQPLVLG (Fig. 5A). Moreover, the iBox domain is highly conserved across species (Fig. 5C), indicating the importance of the sequence in the cell.

We immunoprecipitated INKA2 proteins ectopically expressed in 293T cells with an anti-HA antibody to confirm the interaction between INKA2 and PAK4 (Fig. 6A). Subsequent western blot analysis of PAK4 confirmed the binding of INKA2 to PAK4. The same and reciprocal results were obtained from 293T cells overexpressing both HA-tagged INKA2 and FLAG-tagged PAK4 proteins (Fig. 6B). We then constructed 2 plasmids overexpressing partially deleted INKA2 and PAK4 proteins. The PAK4-inhibitory motif was deleted in INKA2 Δ150-160, and part of the C-terminal catalytic domain (326-591) was deleted in PAK4 1-325 (Fig. 5B). These domains have been reported to interact with each other and a deletion will interrupt the bound (38). Following immunoprecipitation, both deletions disrupted the interaction between INKA2 and PAK4, suggesting that the targeted binding motif of iBox was located in the kinase domain of PAK4 (Fig. 6C). Consequently, we hypothesized that INKA2 inhibits the kinase activity of PAK4 in the same manner as INKA1.

Among a large number of substrates, PAK4 interacts with β-catenin and phosphorylates at Ser675 to regulate its nuclear and cytoplasmic transportation (41). Ser675 phosphorylation protects β-catenin from the destruction complex and thus stabilises the protein in the cytoplasm (41). Thus, we hypothesised that INKA2 binds to the substrate-docking site within the catalytic domain of PAK4 and inhibits the phosphorylation of β-catenin Ser675. Subsequently, the protein becomes unstable and hence degraded. We generated a HCT116 cell line with a tetracycline-regulated lentiviral vector system stably expressing INKA2 in the presence of Dox to confirm our hypothesis. The HCT116 cells were selected for the low baseline expression of endogenous INKA2. Western blot analyses indicated decreased intracellular β-catenin levels in cells overexpressing INKA2 (Fig. 6D). Immunocytochemistry produced similar results (Fig. 6E), suggesting that INKA2 potentially inhibits the kinase function of PAK4 in the same manner as INKA1.

We also performed ATP assay using the INKA2-expressing HCT116 cells to evaluate the suppression of cell growth in four days (Fig. 6F). The results revealed an insignificant difference between INKA2-overexpressing cells and other cells. This indicates that INKA2 inhibits cancer cell growth at a longer time-point of approximately 2 weeks, as shown by the colony formation assays.