2.1 Overexpression of Usp22 accelerates c-Myc/NRasGV12-induced HCC in mice

A well-established mouse HCC model induced by hydrodynamic co-expression of c-Myc and NRasGV12 oncogenes (c-Myc/NRasGV12) in the liver was used to determine the role of Usp22 in hepatotumorigenesis (Figure 1A). Briefly, mice were hydrodynamically injected with c-Myc, NRasGV12, and Flag-tagged Usp22 plasmids (c-Myc/NrasGV12/Usp22) as well as Sleeping Beauty (SB) transposon. The Usp22 plasmid was replaced with an empty multiple cloning site (MCS) vector in the control group (c-Myc/NRasGV12/MCS). Noticeably, overexpression of Usp22 reduced the survival time of the c-Myc/NRasGV12 mouse model of HCC (Figure 1B). The median survival time was 11 days for c-Myc/NRasGV12/Usp22-injected mice (n = 9) and 17 days for c-Myc/NRasGV12/MCS-injected mice (n = 7) (p = 0.0239). Additionally, liver-specific Usp22 knockout mice were used to explore the effects of Usp22 ablation. As expected, Usp22^fl/fl; Alb-Cre mice (n = 6) survived significantly longer than Usp22^fl/fl mice (n = 5) after c-Myc/NRasGV12 injection (33 days vs. 22.5 days, p = 0.0018) (Figure 1C). The experiment was repeated, and 2 weeks after injection, all mice were euthanized and dissected to observe the livers. The results showed that overexpression of Usp22 accelerated tumorigenesis, whereas Usp22 ablation decelerated it in the mouse HCC model (Figure 1D,E). Hematoxylin–eosin staining revealed that HCC developed successfully in all mice. Immunohistochemistry (IHC) of Flag-tag confirmed ectopic injection of Usp22 in the liver, and Ki-67 demonstrated the cell proliferation status in each group (Figure 1D,E). The weight of the livers, percentage of tumor area, and positive Ki-67 results are shown in Figures 1F and S1. These results further validated that the tumors with Usp22 overexpression were significantly larger and more aggressive than those in the control group, whereas Usp22 ablation diminished this effect.

2.2 USP22 activates the mTORC1 signaling pathway

To explore the downstream pathway, RNA sequencing was performed using human Huh-7 oeNC and Huh-7 oeUSP22 cell lines. Heatmaps and volcano plots of differentially expressed genes in the two cell lines are presented in Figure 2A,B. Gene set enrichment analysis (GSEA) revealed that the overexpression of USP22 increased mammalian target of rapamycin complex 1 (mTORC1) activity (Figure 2C,D). Furthermore, we explored critical phosphorylated proteins in the mTORC1 pathway using four HCC cell lines (Figures 2E,F and S2) and c-Myc/NRasGV12 mouse HCC tissues (Figure 2G,H). The levels of these key phosphorylated proteins in the mTORC1 pathway were increased by USP22 overexpression and decreased upon its knockdown. Therefore, USP22 activates the mTORC1 signaling pathway both in vitro and in vivo.

2.3 USP22 promotes tumorigenic potential and sensitizes HCC toward rapamycin in vitro and in vivo

To investigate USP22 function in HCC cells, we performed gain- and loss-of-function studies. In the Huh-7 and Hep-3B cell lines, knockdown of USP22 slowed cell proliferation, while overexpression accelerated it (Figure 3A). The susceptibility of HCC cells to mTOR inhibitors was also investigated due to the altered mTORC1 pathway. Compared to control transfected cells (shNC or oeNC), treatment of HCC cells with different concentrations of rapamycin slightly affected the proliferation of USP22-silenced (shUSP22) cells, but significantly reduced the proliferation of USP22 overexpression (oeUSP22) cells (Figure 3B). Consistent with the results of the cell counting kit-8 (CCK-8) assay, the cell counting data showed that oeUSP22 cells were more sensitive to rapamycin (Figure 3C). In addition, the colony formation abilities and migratory properties of the two HCC cell lines changed after silencing or overexpression of USP22 (Figure 3D–G), presenting the same trends as the CCK-8 assays.

Two xenograft mouse models were used to examine the role of USP22 in tumor growth in vivo. According to in vitro studies, Huh-7 oeUSP22 cells promoted tumor development, and this effect was inhibited by rapamycin administration in mice (Figure 4A–E). Additionally, patient-derived tumor xenograft (PDX) models with different expression levels of USP22 were used to evaluate the antitumor efficacy of rapamycin (Figure 4F). We found that PDX 25 with high USP22 expression exhibited larger tumor volumes and was more sensitive to rapamycin than PDX 99 with low USP22 expression (Figure 4G–K). These findings suggested that activation of the mTORC1 pathway after USP22 overexpression induces therapeutic susceptibility to rapamycin.

2.4 USP22 regulates FKBP12 to activate mTORC1 through deubiquitylation

Because USP22 was shown to activate mTORC1 activity, we hypothesized that USP22 interacts with certain proteins associated with mTORC1, including regulatory-associated protein of mTOR (Raptor), mTOR-associated protein, LST8 homolog (MLST8), mTOR, and FK506-binding protein 12 (FKBP12). Immunoprecipitation (IP) was performed using an anti-USP22 antibody to detect these proteins in SNU-449 cells. The results showed that Raptor, MLST8, and mTOR did not interact with USP22 (Figure S3). Moreover, it is well-established that FKBP12 inhibits mTORC1 activity by directly binding to rapamycin, and we found that USP22 overexpression induces therapeutic susceptibility to rapamycin. We examined FKBP12 and found that USP22 was an upstream regulator of FKBP12 (Figure 5A). Next, we assessed the effect of FKBP12 on the mTORC1 pathway. In Hep-3B cells, the siRNA-mediated knockdown of FKBP12 downregulated the effector proteins of the mTORC1 pathway (Figure 5B). Intriguingly, the activation of the mTORC1 pathway in oeUSP22 cells was rescued by the additional knockdown of FKBP12 (Figure 5C). These results indicated that the upregulation of USP22 activated the mTROC1 pathway via FKBP12.

Moreover, we identified an interaction between USP22 and FKBP12. Flag-tagged USP22 and Myc-tagged FKBP12 plasmids were used to transfect HEK293T cells. Co-IP was performed in whole cell extracts with an anti-Myc or anti-Flag antibody (Figure 5D). The interaction between endogenous FKBP12 and USP22 in SNU-449 cells is shown in Figure 5E, while the co-localization of FKBP12 and USP22 was observed by confocal microscopy (Figure 5F). USP22 has a zinc finger domain at the N-terminus and a ubiquitin-specific peptidase domain at C19. We constructed USP22 point mutants (Figure 5G), and found that a point mutation in the ubiquitin-specific peptidase domain of USP22 could not upregulate FKBP12 expression (Figure 5H). This finding implies that the ubiquitin-specific peptidase domain is the key catalytic residue of USP22 involved in the regulation of FKBP12. USP22 is a deubiquitinating protease and might regulate FKBP12 function via deubiquitinase activity. Therefore, we explored the effect of USP22 on FKBP12 protein stability. USP22 silencing significantly facilitated FKBP12 protein degradation (Figure 5I), whereas USP22 overexpression delayed FKBP12 protein degradation (Figure 5J), suggesting that USP22 stabilized FKBP12. Besides, the proteasome inhibitor MG132, rather than the autophagy inhibitor 3-methyladenine prevented the FKBP12 protein degradation in shUSP22 cells (Figure 5K), indicating that FKBP12 is degraded via the proteasomal pathway. In addition, knockdown of USP22 enhanced polyubiquitination-induced FKBP12 degradation (Figure 5L), whereas USP22 overexpression reduced polyubiquitination-induced FKBP12 degradation (Figure 5M). We further explored USP22 deubiquitination using a series of ubiquitin mutants (K6, K11, K27, K33, K48, and K63). The results indicated that the K63-linked ubiquitination site alone did not influence FKBP12 polyubiquitination (Figure S4). In summary, USP22 activates mTORC1 by deubiquitinating FKBP12.

2.5 Activated mTORC1 further inhibits the autophagic degradation of USP22

To further determine whether mTORC1 in turn regulates USP22, we inhibited it with rapamycin and found downregulated USP22 with enhanced autophagy activity (Figure 6A). In addition, activated mTORC1 mediated by the knockdown of tuberous sclerosis 1 (TSC1) stabilized USP22 and inhibited autophagy activity, and rapamycin treatment also reduced the upregulation of USP22 (Figure 6A). Importantly, the rapamycin-induced decrease in USP22 expression was restored by blocking autophagy (Figure 6B). This indicates that activated mTORC1 stabilizes USP22 by inhibiting autophagic degradation. In addition, autophagosomes were observed by transmission electron microscopy (Figure 6C), and autophagic flux was detected by fluorescent LC3 in mRFP-GFP-LC3 transfected cells (Figure 6D). These results demonstrate that activated mTORC1 further stabilizes USP22 by inhibiting its autophagic degradation.

2.6 Clinical significance of USP22 in patients with HCC after LT

To explore the clinical significance of USP22, we detected the protein expression of USP22 and FKBP12 in 205 patients with HCC after LT using IHC. USP22 and FKBP12 were found to have a strong positive correlation (Figure 7A,B), further demonstrating their positive relationship.

Sirolimus is a common immunosuppressant administered after LT. Given that LT recipients fulfilling the Milan Criteria already had a satisfactory prognosis, we explored the role of USP22 in recipients beyond the Milan Criteria and especially investigated the effect of sirolimus on prognosis depending on USP22 expression. Table S1 presents the clinical features of the 115 LT recipients who did not meet the Milan Criteria, there were no notable variations in demographic factors. In LT recipients diagnosed with HCC beyond the Milan Criteria, those with high USP22 expression presented a lower 3-year overall survival (OS) rate than those with low USP22 expression (74.8% vs. 50.6%, p = 0.027) (Figure 7C). Univariate and multivariate Cox regression analyses indicated that the use of sirolimus was an independent protective factor for OS (Hazard Ratio (HR) = 0.465, p = 0.035), whereas vascular invasion (HR = 2.513, p = 0.015) and Child–Pugh score (high) (HR = 1.327, p = 0.023) were independent risk factors (Table S2).

Further subgroup analysis was performed based on USP22 expression. The clinical features of these recipients are shown in Table S3. In recipients with high USP22 expression, sirolimus improved the 3-year OS (62.4% vs. 41.3%, p = 0.012) (Figure 7D) and was an independent protective factor for OS (HR = 0.268, p = 0.005) (Table S4). However, sirolimus did not improve OS in recipients with low USP22 expression (Figure 7E). Therefore, high USP22 expression in tumor tissues could predict the benefit of sirolimus in patients with HCC after LT, which supports the significant role of USP22 in activating mTORC1 in HCC.

In summary, in HCC, USP22 overexpression increased mTORC1 activity and sensitized HCC toward rapamycin. Mechanistically, USP22 interacted with FKBP12 to activate the mTORC1 pathway through deubiquitylation. Increased mTORC1 activity further stabilized USP22 by inhibiting autophagic degradation (Figure 7F).

5.1 Generation of mice and genotyping

Usp22 conditional knockout mice were purchased from Cyagen Biosciences. To produce targeted conditional knockout offspring, the gRNA for Usp22 gene in mice (Table S5), along with the donor vector containing loxP sites, and Cas9 mRNA, were simultaneously injected into fertilized mouse eggs. F0 founder animals were identified using polymerase chain reaction followed by sequence analysis, and then bred with wild-type mice to assess germline transmission for F1 animal generation. To generate F2, an F1 targeted mouse was crossed with a tissue-specific Alb-Cre deletion mouse. Heterozygous Cre⁺ mice were then bred with homozygous mice. Genotyping was performed using the Mouse Tail Quick Genotyping Kit (Beyotime) with appropriate primers (Table S6) and some results are shown in Figure S5.

5.2 Plasmid and lentiviral construction

The plasmids (RiboBio) used in animal and cell experiments are listed in Table S7. All plasmids were purified using an Endo-Free Plasmid Max Kit (Omega Bio-Tek). Lentivirus overexpressing USP22 and shRNAs against USP22 were provided by Shandong Vigene. Four separate siRNAs targeting USP22 were created, validated, and effectively packaged in a four in one shUSP22 lentivirus to prevent off-target effects (Figure S6). All siRNA sequences are listed in Table S8.

5.3 Hydrodynamic tail vein injection models

C57BL/6J mice (6−8 weeks, male) were obtained from the Shanghai Experimental Animal Centre, Chinese Academy of Science, and Usp22^fl/fl; Alb-Cre mice were purchased from Cyagen Biosciences. Briefly, mice were administered a hydrodynamic tail vein injection of plasmid mixture in 2 mL saline within 5–9 s as previously described.⁴⁰ Each plasmid mixture contained the following: c-Myc 10 μg, NRasGV12 10 μg, SB100 0.8 μg, MCS 20 μg, and Flag-USP22 20 μg. Body weight, abdominal palpation, and physiological parameters of the mice were recorded daily. Mice were euthanized when a high burden of HCC manifested, for example, an abdominal mass.

5.4 Immunohistochemical analysis

Tumors from an orthotopic HCC mouse model were isolated, treated with formalin, and embedded in paraffin. A total of 205 HCC samples were obtained to establish a microarray. The manufacturer's instructions for the Immunohistochemical Assay Kit (Abcam) were followed during immunohistochemical staining. The antibodies used are listed in Table S9. An Olympus VS200 was used to scan the microarray slides and they were assessed using the immunoreactive score (IRS), which was determined by multiplying the staining intensity grades and the fraction of positive cells. The gradations were ranked as shown in Table S10. IRS (0−7) and IRS (8−12) were analyzed for low and high expression levels, respectively.

5.5 Cell culture, transfection, and lentivirus infection

Human HCC cell lines (Huh-7, Hep-G2, and Hep-3B) and HEK293T cells were acquired from the Shanghai Institute of Cell Biology, Chinese Academy of Sciences, and SNU-449 cells were purchased from Guangzhou Cellbook Biotech. The authenticity of all cells was verified using short tandem repeat profiling. Huh-7 and HEK293T cells were grown in Dulbecco's modified Eagle's medium with high glucose (BI), Hep-G2 and Hep-3B in minimum essential medium, and SNU-449 in RPMI-1640. Fetal bovine serum (FBS, 10%; Gibco), 100 U/mL penicillin, 100 μg/mL streptomycin, and 50 μg/mL gentamicin (Solarbio) were added to the media. All cells were cultured in a 37°C, 5% CO₂ incubator. The manufacturer's protocol (Polyplus) was followed for transient siRNA and DNA transfection using jetPRIME. Lentivirus infection was performed as previously reported.⁴¹

5.6 RNA sequencing and GSEA

The poly (A) selection procedure was used to select mRNA from the HCC cell line Huh-7 after extracting total RNA. RiboBio prepared and sequenced the RNA sequencing library.⁴² DESeq2 was used for differential expression analysis, employing a significance threshold of p < 0.05 and a minimum fold change of 1.5. Cluster analysis was conducted to determine the expression patterns of various genes under diverse experimental conditions.

5.7 Cell viability assay

Cell viability was determined using a CCK-8 assay. Briefly, 2000−5000 HCC cells were seeded in 96-well plates, cultured for 6, 24, 48, and 72 h with or without rapamycin, and treated with CCK-8 solution. After incubation at 37°C with 5% CO₂ for 1–2 h, the absorbance was measured at 450 nm with an EnSpire 2300 Multilabel Reader (PerkinElmer). For cell counting, 2−5 × 10⁴ cells were seeded into 12-well plates. Following a 48-h treatment period, with or without rapamycin, the cells were digested, resuspended, and subsequently stained with trypan blue. The cells were counted using a cell-counting plate under a microscope.

5.8 Colony formation assay

Two thousand HCC cells were seeded in 24-well plates and treated with various concentrations of rapamycin for 10−14 days. The cell colonies were fixed with methanol and visualized using crystal violet staining.

5.9 Cell migration assay

A total of 50,000 HCC cells were diluted in 200 μL of media without serum and containing 0.1% bovine serum albumin (BSA). These cells, with or without rapamycin, were introduced into the upper chamber of a transwell plate (Corning). The lower compartment received 700 μL of media containing 10% FBS, either with or without rapamycin. After 24–48 h, the polycarbonate membranes were washed, fixed, and stained with 0.1% crystal violet. The cells were then examined under a microscope. The experiments were independently repeated three times.

5.10 Western blotting

Total protein was extracted from the cells or tissues using RIPA lysis buffer (FUDE) supplemented with protease and phosphatase inhibitor cocktails (Roche). Electrophoresis was performed on 4%−20% sodium dodecyl–sulfate polyacrylamide gel electrophoresis gels (GenScript). The proteins were then transferred to polyvinylidene difluoride membranes (Millipore). Following gentle shaking in blocking buffer (Tris-buffered saline with 5% nonfat milk or BSA) for 1 h, the blots were incubated with primary antibody at 4°C overnight and subsequently with secondary antibody at 37°C for 1 h. The antibodies used are listed in Table S9. The bands were visualized using ECL chemiluminescence (FUDE) and a Fluor Chem M imaging system (ProteinSimple).

5.11 Animal experiments

For the cell-derived xenograft model, 5 × 10⁶ Huh-7 oeNC and Huh-7 oeUSP22 cells were subcutaneously implanted into 5-week-old male BALB/c nude mice. Details of the observations, rapamycin solubilization, and data analyses were performed as previously described.³⁸ Rapamycin was administered daily via intraperitoneal injection at a dose of 1 mg/kg. For the PDX model, fresh HCC tissues were acquired from patients, diced into 1-mm³ pieces and implanted into the flanks of 5-week-old male NOD-SCID mice purchased from the Shanghai Experimental Animal Centre, Chinese Academy of Science. Mice that successfully developed subcutaneous tumors from the tumors of the original patients were defined as founder 0 (F0). The second generation of PDX model was defined as F1, and so forth. PDX from patient 25 (PDX 25) and patient 99 (PDX 99) had different expression levels of USP22 and were chosen for the assessment of the antitumor effect of rapamycin.

5.12 Co-IP assay

In brief, cell lysates were precleaned using IgG and Protein A+G agarose (Beyotime) for 30 min and then immunoprecipitated with proper antibodies against USP22, FBKP12, Flag, or Myc (Table S9) by incubation with agarose at 4°C for 2 h. Next, the beads were washed three times with the lysis buffer. Precipitated proteins (10 μg) were prepared for western blot analysis, and the procedure was as described above.

5.13 Immunofluorescence

Glass coverslips were attached to six-well plates and the cells were placed on them. Fluorochrome-conjugated secondary antibodies were used to identify target proteins. Nuclear localization was achieved using DAPI (300 nM) labeling. Fluorescence distribution was observed using a laser scanning confocal microscope (Olympus).

5.14 Ubiquitination assay

To explore ubiquitination, the protein synthesis inhibitor cycloheximide and the proteasome inhibitor MG132 (MCE) were used. Cells were disrupted in a solution containing 50 mM Tris (pH 7.4), 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, and 0.1% sodium dodecyl sulfate along with cocktails. The samples were heated at 95°C for 5 min to cause denaturation. IP and western blotting were performed as previously described. The antibodies used are shown in Table S9.

5.15 Patients and data sources

In total, 205 adult patients with HCC who received deceased donor LT at the First Affiliated Hospital of Zhejiang University between January 1, 2015, and December 30, 2022, were enrolled in this part of the study. Data from the China Liver Transplant Registry database (http://www.cltr.org) were collected for further analyses. IHC and IRS, as described above, were used to analyze the expression of USP22 and FKBP12. A total of 115 recipients beyond the Milan Criteria were further divided into high/low USP22 expression and sirolimus (SRL)/non-SRL groups. Patients with pathologically diagnosed HCC and complete follow-up records were included in this study. Exclusion criteria included the use of SRL for less than 30 days or after HCC recurrence, death within 90 days, and multiple-organ transplantation.

5.16 Immunosuppression

The immunosuppressants were the same as detailed previously.³⁸ After LT, tacrolimus/cyclosporine A (CNI) was administered to the non-SRL group, and dose adjustments were made based on blood concentration and liver function. Following surgery, the target level of tacrolimus in the peripheral blood was 4–8 ng/mL. Additionally, clinicians lowered the tacrolimus dose as much as feasible, as long as liver function was normal. In the SRL group, CNI was administered initially, then 30−45 days after LT, SRL was administered. When SRL was introduced, the dose of CNI was halved and stopped when SRL reached the desired blood concentration (4–10 ng/mL). Both groups continued to receive mycophenolate.

5.17 Statistical analysis

IBM SPSS Statistics 23.0 and GraphPad Prism 7.00 were used for the statistical analysis. Data are displayed as mean ± standard deviation. Student's t-test and the Mann–Whitney U-test were used for comparisons between two groups, and Pearson's chi-square test was applied for categorical variables. The Kaplan–Meier method was used to compare OS, while differences between groups were evaluated using a log-rank test. Cox univariate analysis included all factors with p-values less than 0.05, which were then incorporated into multivariate analysis using the Cox proportional hazards model. Significance was defined as p < 0.05 for all two-tailed p-values.