Human liver cancer cell lines (PLC/PRF/5, CSQT-2, HepG2, Huh7, MHCC-97H) along with the normal liver cell line HHL5 were acquired from the Shanghai Cell Bank, Chinese Academy of Sciences. All cell lines were maintained in Dulbecco’s Modified Eagle Medium (DMEM; Basal Media) supplemented with 10% fetal bovine serum (FBS; Gibco) and antibiotic-antimycotic solution (NCM Biotech). Cultures were incubated at 37°C in a humidified atmosphere containing 5% CO₂.

HCC specimens and matched adjacent non-tumor tissues were obtained from patients undergoing resection at Eastern Hepatobiliary Surgery Hospital. The study protocol received ethical approval from the hospital’s Institutional Review Board (protocol code EHBHKY2018-1-001 and approved on the 6 June 2018). Written informed consent was obtained from all participants prior to sample collection. All specimens were anonymized and processed according to standard ethical guidelines.

Cell viability was quantified using the Cell Counting Kit-8 (CCK-8; Absin) following manufacturer specifications. Briefly, cells were seeded in 96-well plates (1–1.5 × 10³ cells/well) and cultured for indicated durations. Following cell adhesion, CCK-8 reagent (10 μL/well) was added and incubated for 1 h at 37°C before measuring absorbance at 450 nm using a Synergy H1 microplate reader (BioTek) to obtain the Day 0 value. Repeat the assay at 24 h, 48 h, 72 h, 96 h, and 120 h to collect data for Day1 through Day 5. Data were normalized to baseline measurements (Day 0) and expressed as fold-change relative to controls.

For colony formation analysis, 1.5 × 10³ cells were plated in 6-well plates and cultured for 7–14 days with medium replacement every 72 h. Colonies were fixed with 4% paraformaldehyde (15 min), stained with 0.1% crystal violet (15 min), and quantified the stained colonies by measuring the clone area using image analysis software. Results represent fold-change compared to control groups.

The lentiviruses of short hairpin RNA (shRNA) targeting PPP1R12B and scramble shRNA (ShCtrl) were obtained (Genechem Co.), and transfected into PLC/PRF/5, CSQT-2 and HHL5 cell lines according to the manufacturer’s instructions. ShRNA sequences were as follows: ShPPP1R12B:5′-CcggAAGGATCTTCTTCTGGAGCAACTCGAGTTGCTCCAGAAGAAGATCCTTTTTTTg-3′; ShCtrl: 5′-CcggTTCTCCGAACGTGTCACGTCTCGAGACGTGACACGTTCGGAGAATTTTTg-3′. Overexpression lentiviral particles of PPP1R12B (NCBI Reference Sequence: NM_002481) and control lentiviral particles were constructed (Genechem Co.) and transfected into Huh7, HepG2 and MHCC-97H cell lines according to the manufacturer’s instructions. For overexpression studies, the full-length PPP1R12B coding sequence was amplified using primers (Sense: 5′-CCAACTTTGTGCCAACCGGTCGCCACCATGGCGGAACTGGAGCACCTAGGAGGG-3′; Antisense: 5′-GTCAATGCCAACTGAGCTTCTTGGACAGTTTGCTGATGAC-3′) and cloned into the GV341 lentiviral vector (AgeI/NheI sites). Since the lentiviral vector carries a puromycin resistance gene, transfected cells were selected with puromycin (2–5 μg/mL, concentration optimized for each cell line) for 3–10 days. Protein lysates of transfected cells were subsequently collected and subjected to Western blot analysis to confirm PPP1R12B overexpression/knockdown efficiency. Puromycin selection pressure was applied intermittently (every other passage) to maintain transduction efficiency (maintenance at half the working concentration).

PAK2 knockdown was achieved using siRNA duplexes (Sense: 5′-CCGGAUCAUACGAAAAUCAATT-3'; Antisense: 5′-UUGAUUUCGUAUGAUCCGGTT-3′) transfected with Lipofectamine 3000 (Invitrogen) according to standard protocols. Control cells received non-targeting siRNA (Sense: 5′-UUCUCCGAACGUGUCACGUTT-3'; Antisense: 5′-ACGUGACACGUUCGGAGAATT-3′).

Following 24-h serum starvation for cell cycle synchronization, the cells were re-fed with complete medium and incubated for an additional 24–36 h (HHL5 and PLC/PRF/5 cells for 24 h, and MHCC-97H cells for 36 h) prior to harvesting. Synchronized cells were fixed in 75% ethanol overnight, treated with RNase, and stained with propidium iodide using the Cell Cycle Assay-PI/RNase Staining Kit (Dojindo). DNA content analysis was performed on a BD LSRFortessa flow cytometer, with data processed using FlowJo software (v10.6.2).

Nuclear and cytoplasmic extracts were prepared using the Nuclear/Cytoplasmic Extraction Reagent Kit (Beyotime) supplemented with protease inhibitors. Protein localization was confirmed by immunoblotting using compartment-specific markers (Lamin B1 (12987-1-AP, Proteintech) for nuclear; GAPDH (AC033, ABclonal) for cytoplasmic fractions).

Total protein lysates were prepared using RIPA buffer (Beyotime) containing protease and phosphatase inhibitors. Protein concentration was determined by BCA assay (Thermo Fisher). Samples were separated by 10% SDS-PAGE, transferred to nitrocellulose membranes, and probed with primary antibodies (4°C, overnight) followed by appropriate secondary antibodies. The primary antibodies used in Western blotting included PPP1R12B antibody (13366-1-AP, Proteintech), PAK1 Polyclonal antibody (21401-1-AP), gamma-PAK2 antibody (sc-373740, Santa Cruz), Cyclin D1 antibody (2978, CST), beta-catenin antibody (37447, CST), Phospho-beta-catenin (Ser675) antibody (4176, CST), beta-actin antibody (AC004, ABclonal). The secondary antibodies used were IRDye 800CW Goat anti-Mouse IgG (H + L) and IRDye 800CW Goat anti-Rabbit IgG (H + L) from LI-COR. Detection was performed using an Odyssey Sa Imaging System (LI-COR).

Cells were collected and lysed in IP buffer with protease inhibitors (Beyotime). Subsequently, Protein A/G Magnetic Beads (MCE) was added to total protein, incubated at 4°C for 1 h with mild rotation. After centrifugation at 12,000 rpm for 15 min, the harvested supernatant was incubated with 2 μg anti-FLAG (66008-4-IP, Proteintech) and 2 μg anti-IgG (sc-2025, Santa Cruz), or 2 μg anti-PAK2 antibody (19979-1-AP, Proteintech) and 2 μg anti-IgG (12–370, Sigma-Aldrich) at 4°C overnight with gentle rotation. A magnetic rack was used to recover the precipitated protein complex. Immune complexes were captured with magnetic beads, washed extensively, and analyzed by Western blotting. 10% of total protein was applied as input control.

Formalin-fixed, paraffin-embedded sections (3–5 μm) underwent antigen retrieval before incubation with PPP1R12B antibody (13366-1-AP, Proteintech) (4°C, overnight). Detection employed HRP-conjugated secondaries (Supervision) with DAB chromogen (DAKO). Slides were counterstained with hematoxylin, scanned (Leica Aperio AT2), and analyzed using Aperio ImageScope (v12.4.6) and ImageJ. Staining intensity (0: negative, 1: weak, 2: moderate, 3: strong) and positive cells distribution (0%–100%) were combined to generate H-scores (range 0–300) (Liu W. et al., 2015).

Patients were stratified by PPP1R12B expression (low/medium/high) based on IHC scores. Combined with overall survival information of HCC patients, Kaplan-Meier survival curves were generated using GraphPad Prism 10, with statistical significance assessed by log-rank test.

Cells fixed with 4% paraformaldehyde were permeabilized (0.1% Triton X-100), blocked with 2% BSA, and incubated with primary antibodies (4°C, overnight) followed by fluorophore-conjugated secondaries. The primary antibodies used in Immunofluorescence were same as in Western blotting, plus PAK2 antibody (19979-1-AP, Proteintech). The fluorophore-conjugated secondary antibodies were bought from Invitrogen (A-11008, A-11001, A-21428, and A-21422). Nuclei were counterstained with DAPI (Beyotime) before imaging on a Leica TCS SP8 confocal microscope.

TOPFlash/FOPFlash reporter assays were conducted by co-transfecting 250 ng of TOPFlash or FOPFlash reporter plasmid with 20 ng pRL-TK (Promega) using Lipofectamine 3000 (Invitrogen). Luciferase activity was measured 48 h post-transfection using the Dual-Luciferase Assay System (Promega). Data were normalized to Renilla luciferase activity, with the TOPFlash/FOPFlash ratio representing relative luciferase activity.

All animal procedures were approved by the Institutional Animal Care Committee at Eastern Hepatobiliary Surgery Hospital. Male BALB/c nude mice (4–5 weeks; GemPharmatech) were housed under specific pathogen-free conditions. Huh7 overexpression and control cells (5 × 10⁶) suspended in Matrigel/medium (3:2) were injected subcutaneously. Tumor dimensions were measured every 3 days, with volume calculated as 0.5 × length × width². Mice were euthanized when tumors reached 1,500 mm³.

Phosphoproteomic sequencing analysis was performed by Jingjie Biotechnology Co., Ltd. At first, protein samples lysed in urea buffer (8M urea, 1% protease/phosphatase inhibitors) via ultrasonication and precipitated with acetone, digested with trypsin (1:50 w/w), reduced (DTT), alkylated (IAA), and desalted (Strata X SPE). Next, phosphopeptides were enriched using IMAC microspheres with wash (50% ACN/0.5% acetic acid → 30% ACN/0.1% TFA) and elution (10% NH₄OH) before LC-MS/MS analysis on an Orbitrap Exploris 480 mass spectrometer coupled to an EASY-nLC 1200 UPLC system. And then, data were processed using Spectronaut (v18) against the UniProt human database (Homo_sapiens_9606_SP_20231220).

Data analysis employed GraphPad Prism 10 and SPSS 22.0. Continuous variables are presented as mean ± SEM and compared using Student’s t-test. Categorical variables were analyzed by χ ² test. Survival data were evaluated by Kaplan-Meier method with log-rank test. Correlation analyses used Spearman’s rank test. All tests were two-tailed with α = 0.05.

The Cancer Genome Atlas (TCGA) datasets were used to assess PPP1R12B expression patterns. The results revealed significant downregulation of PPP1R12B mRNA in HCC tissues relative to adjacent non-tumor controls (Figure 1A). Western blot and immunohistochemistry (IHC) experiments of 29 paired samples revealed significantly lower PPP1R12B protein expression in HCC tumor tissues compared to matched adjacent non-tumor tissues (Figures 1B–D). Moreover, we subsequently performed survival analysis using HCC tissue microarray (TMA). All 228 tissue spots were individually scored by IHC and categorized into high (n = 76), medium (n = 76), and low (n = 76) PPP1R12B expression groups based on IHC scores (Figure 1E), followed by Kaplan-Meier survival curve analysis with clinical follow-up data. The survival analysis demonstrated that high PPP1R12B expression correlated with significantly better survival outcomes (Figure 1F). Hence the low and medium expression groups showed nearly identical survival patterns, we combined these two groups and then analyzed the differences in clinicopathological characteristics between the high versus low/medium expression groups. HCC patients with high PPP1R12B expression were apt to have higher age, recurrence-free survival time, overall survival time, and better final clinical outcome as shown in Table 1.

To determine whether PPP1R12B represents an independent prognostic factor in HCC patients, we then performed multivariate COX regression analysis based on the HCC TMA results and clinic characteristics data of HCC patients (Supplementary Table S1). Initially, univariate Cox analysis was used to identify insignificant prognostic factors which were excluded from further modeling (p ≥ 0.05). As results, the following factors were included in further multivariate analysis: PPP1R12B expression, tumor size, lymph node metastasis, distant metastasis, TNM stage, bile duct tumor thrombus, and vascular invasion. Given that TNM stage inherently incorporates data on lymph node metastasis and distant metastasis, we conducted separate statistical analyses for two scenarios: with and without TNM stage inclusion. Using forward stepwise regression to construct optimal models, we found that: When TNM stage was included as a covariate, tumor size and vascular invasion emerged as risk factors while PPP1R12B served as a protective factor (Figure 2A); and when TNM stage was excluded, PPP1R12B remained protective, while tumor size, lymph node metastasis and distant metastasis became significant risk factors (Figure 2B). These results collectively indicate that PPP1R12B represents an important protective factor in HCC patients.

Based on the baseline expression levels of PPP1R12B across hepatocellular carcinoma cell lines (Supplementary Figure S1A), we constructed stable PPP1R12B-overexpressing cell lines in Huh7, HepG2, and MHCC-97H (with relatively low natural expression), and stable knockdown cell lines in PLC/PRF/5 and CSQT-2 (with relatively high endogenous expression). Besides, human normal liver cell lines HHL5 was also constructed as a PPP1R12B knockdown cell lines. The CCK-8 assays and plate colony formation experiment demonstrated that PPP1R12B knockdown significantly enhanced proliferative capacity, whereas overexpression of PPP1R12B suppressed cell proliferation (Figures 3A–D). Using PPP1R12B-overexpressing Huh7 cells and controls cells, we performed xenograft tumor experiments in nude mice, periodically measuring tumor volume to compare growth rates between groups. Xenograft experiments and quantitative analysis revealed a significant reduction in mean tumor weight in PPP1R12B-overexpressing mice relative to control groups (Figure 3E). Meanwhile, each inoculation including Huh7 cells with PPP1R12B-overexpression developed into a slowly growing and small tumor than controls (Figure 3F). These overexpression and knockdown results suggested PPP1R12B might function as tumor suppressor in HCC cell proliferation.

Given the well-established relationship between proliferation and cell cycle regulation, we hypothesized that PPP1R12B might influence HCC cell proliferation by modulating the cell cycle. To further investigate this, PPP1R12B overexpressing MHCC-97H cells and downregulated PLC/PRF/5 and HHL5 cells was chosen for cell cycle analysis by flow cytometry. After serum starvation for cell cycle synchronization and an additional 24–36 h incubation, flow cytometry analysis revealed that altered PPP1R12B expression indeed affected HCC cell cycle distribution, primarily impacting the G0/G1 and S phases. More specifically, PPP1R12B knockdown accelerated progression from G0/G1 to S phase (Figures 4A,B), whereas overexpression caused cell cycle arrest at the G0/G1-S transition boundary (Figure 4C).

Phosphoproteomic sequencing analysis using PPP1R12B-overexpressing Huh7 cells and controls was performed to identify the critical molecules involved in PPP1R12B-mediated HCC proliferation suppression (Figure 5A). Combined with protein interaction network analysis, the potential interaction between PPP1R12B and PAK1/PAK2 were found (Figure 5B). Then we examined the protein expression levels of both PAK1 and PAK2 in PPP1R12B-overexpressing Huh7 and MHCC-97H cell lines. Western blot analysis revealed that PAK1 expression remained relatively unchanged, whereas PAK2 expression showed a significant decrease (Fig. S1B). Consequently, we selected PAK2 as the focus for subsequent investigations.

We first confirmed this interaction through co-immunoprecipitation (Co-IP) using Flag antibodies in Flag-tagged Huh7 and HepG2 overexpression cell lines (Figure 5C). Subsequent reciprocal Co-IP with PAK2 antibodies in Huh7 overexpression cells further verified the PPP1R12B-PAK2 interaction (Figure 5C). Moreover, confocal microscopy clearly demonstrated co-localization of PAK2 and PPP1R12B in Huh7 and HepG2 cells, providing additional evidence for their interaction (Figure 5D).

Having established this interaction, we investigated whether PAK2 involves PPP1R12B-mediated HCC proliferation suppression. Using siRNA to knockdown PAK2 in PPP1R12B-overexpressing (Huh7, HepG2, MHCC-97H) and -knockdown (HHL5, CSQT-2) cell lines, we observed that PAK2 knockdown attenuated the proliferative changes induced by PPP1R12B modulation (Figure 5E). Control experiments showed that PAK2 positively regulated HCC cell proliferation, as its knockdown reduced proliferative capacity of HCC cells. These findings suggested that PPP1R12B likely exerts its effects on proliferation through PAK2.

To explore PPP1R12B and PAK2 how to regulate HCC proliferation, we re-analysed the phosphoproteomic sequencing data focusing on PAK2. The results revealed its potential association with CTNNB1 (β-catenin) and SRC. Furthermore, the phosphoproteomic sequencing data indicated phosphorylation at β-catenin Ser675, with decreased levels, and phosphorylation at SRC Ser17, with increased levels (Supplementary Figure S1C). However, Western blot analysis in PPP1R12B-overexpressing Huh7 cells revealed no significant alterations in either total SRC protein levels or its phosphorylation status (Supplementary Figure S1D). Consequently, we prioritized β-catenin for subsequent mechanistic investigations. Reduced β-catenin expression following PAK2 knockdown was confirmed by confocal microscopy (Figure 6A). Since Ser675 phosphorylation was reported to enhance β-catenin stability and transcriptional activity (Zhu et al., 2012), we examined β-catenin activity after PAK2 knockdown. Indeed, β-catenin activity was significantly lower in PAK2-knockdown cells compared to controls (Figure 6D). We further demonstrated that PAK2 knockdown reduced total β-catenin, p-β-catenin (Ser675), and the expression of CCND1 (a canonical β-catenin downstream gene) in CSQT-2 and HHL5 cells (Figure 6B). These findings suggested that PAK2 promotes β-catenin expression, Ser675 phosphorylation, and nuclear translocation.

Next, we investigated whether PPP1R12B similarly affects β-catenin localization. In CSQT-2 knockdown cells and HepG2 overexpression cells, PPP1R12B overexpression decreased while knockdown increased the expression of PAK2, β-catenin, p-β-catenin (Ser675) and CCND1 (Figure 6C). β-catenin activity assays showed corresponding changes, with increased activity in knockdown cells and decreased activity in overexpression cells (Figure 6E). Nuclear-cytoplasmic fractionation experiment revealed that nuclear β-catenin levels were elevated in knockdown cells but reduced in overexpression cells compared to controls cells (Figure 6F). These results indicate that PPP1R12B inhibits β-catenin expression and Ser675 phosphorylation, thereby suppressing nuclear translocation and downstream CCND1 expression.

Cyclin D1 (encoded by CCND1) forms complexes with cyclin-dependent kinases 4/6 (CDK4/6) to critically regulate cell cycle progression in G0/G1 phase (Bertoli et al., 2013). Palbociclib, an oral, reversible, selective CDK4/6 inhibitor, was used to test whether CDK4/6-Cyclin D complexes mediate PPP1R12B-mediated HCC proliferation suppression. We found that palbociclib inhibited HCC cell proliferation and counteracted the proliferative changes induced by PPP1R12B modulation (Figure 6G). This indirectly supports that PPP1R12B regulates HCC cell proliferation by affecting Cyclin D1 expression and subsequent cell cycle control.

TCGA HCC cohort data was analyzed to further explore PAK2 and β-catenin expression patterns and prognostic significance. Across four HCC datasets (GSE22058, GSE36376, GSE14520, OEP000321), both PAK2 and β-catenin showed significantly higher expression in tumor tissues versus adjacent non-tumor tissues (Figures 7A,B). Correlation analysis of TCGA-HCC data demonstrated inverse relationships between PPP1R12B and both PAK2/β-catenin expression in HCC tissues (Figure 7C). Combined with our own findings on PPP1R12B expression pattern and prognostic significance, these results showed that PAK2 and β-catenin exhibit opposite expression patterns and prognostic associations compared to PPP1R12B, further supporting our proposed hypothesis.