Self‐Activated Cascade‐Responsive Sorafenib and USP22 shRNA Co‐Delivery System for Synergetic Hepatocellular Carcinoma Therapy

Abstract Resistance to sorafenib severely hinders its effectiveness against hepatocellular carcinoma (HCC). Cancer stemness is closely connected with resistance to sorafenib. Methods for reversing the cancer stemness remains one of the largest concerns in research and the lack of such methods obstructs current HCC therapeutics. Ubiquitin‐specific protease 22 (USP22) is reported to play a pivotal role in HCC stemness and multidrug resistance (MDR). Herein, a galactose‐decorated lipopolyplex (Gal‐SLP) is developed as an HCC‐targeting self‐activated cascade‐responsive nanoplatform to co‐delivery sorafenib and USP22 shRNA (shUSP22) for synergetic HCC therapy. Sorafenib, entrapped in the Gal‐SLPs, induced a reactive oxygen species (ROS) cascade and triggered rapid shUSP22 release. Thus, Gal‐SLPs dramatically suppressed the expression of USP22. The downregulation of USP22 suppresses multidrug resistance‐associated protein 1 (MRP1) to induce intracellular sorafenib accumulation and hampers glycolysis of HCC cells. As a result, Gal‐SLPs efficiently inhibit the viability, proliferation, and colony formation of HCC cells. A sorafenib‐insensitive patient‐derived xenograft (PDX) model is established and adopted to evaluate in vivo antitumor effect of Gal‐SLPs. Gal‐SLPs exhibit potent antitumor efficiency and biosafety. Therefore, Gal‐SLPs are expected to have great potential in the clinical treatment of HCC.


Introduction
Hepatocellular carcinoma (HCC), which has a continuously increasing incidence rate, remains a stumbling obstacle in increasing human life expectancy. [1] Although appreciable progress in the treatment of HCC has occurred over the past few years, current therapeutic effects are still unsatisfactory, and the 5-year survival rate varies from 14% to 18%. [1a,2] Sorafenib, a kind of tyrosine kinase inhibitors, was approved by the FDA in 2007 for clinical applications as the first-line drug against advanced HCC. [3] Sorafenib inhibits cell proliferation and cancer angiogenesis by blocking the RAF/MEK/ERK pathway and tyrosine kinases of VEGFRs/PDGFRs. [4] However, sorafenib often encounters clinical failure owing to drug resistance. Cancer stemness is closely associated with resistance to sorafenib and accounts for the clinical failure of current therapeutics against HCC. The glycolysis features

Construction of Polyplexes and Gal-SLPs
Polyplexes with various N/P ratios were prepared through electrostatic interactions between cationic B-PDEAEA and anionic USP22 shRNA (shUSP22). B-PDEAEA condensed shUSP22 and formed spherical polyplexes with uniform sizes of ≈50 nm and zeta potentials from +20 to +25 mV at N/P ratios ranging from 1a-b After a 24-h incubation with 200 µm H 2 O 2 , the zeta potential of polyplexes rapidly dropped from +24.3 to −4.6 mV and their size increased from 58.0 nm to ≈1 µm, which indicated a quick hydrolysis of the polymer and disintegration of polyplexes upon H 2 O 2 (Figure 1c). Accordingly, the ROS-responsive charge-reversal polyplexes released the DNA after a 1-h incubation with 0.5 mm H 2 O 2 , as detected by gel electrophoresis (Figure 1d). Gene transfection efficiency and cytotoxicity assessments were performed to determine the optimum N/P ratio for further in vitro and in vivo experiments. B-PDEAEA/shUSP22 polyplexes efficiently knocked down USP22 in both Huh-7 and BEL-7402 cells (Figure 1e-f, Figure S2, Supporting Information). In particular, ≈40% of USP22 was downregulated by B-PDEAEA/shUSP22 polyplexes at an N/P ratio of 17 in Huh-7 cells, which was significantly higher than that of conventional commercial vehicles (polyethylenimine [PEI] and Lipofectamin 2000 [Lipo2000]). The efficiency of the polyplexes was further evaluated using an EGFP-shUSP22 plasmid. Likewise, polyplexes at an N/P ratio of 17 transfected optimally as determined by flow cytometry and fluorescence microscopy ( Figures S3 and  S4, Supporting Information). Moreover, polyplexes containing negative control (NC) plasmids showed negligible toxicity to Huh-7 and BEL-7402 cells among a wide range of N/P ratios (Figures S3a and S4a, Supporting Information).
To meet the requirement for intravenous administration, the positive charge of B-PDEAEA/shUSP22 polyplexes should be shielded with polyanion or a lipid layer. Inspired by our previous report, a fusogenic lipid, consisting of 1,2dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE), cholesteryl hemisuccinate (CHEMS) and distearoyl phosphoethanolaminepolyethylene glycol (DSPE-PEG)-galactose was used to enable the formation of lipopolyplex with stealth properties, HCCtargeting ability and membrane fusion feature. [9a,12] After decoration with a sorafenib-loaded lipid layer, the zeta potential of Gal-SLPs decreased to -5.6 ± 0.8 mV and the size increased to 95.6±5.2 nm (Figure 1g-h). Transmission electron microscopy (TEM) images of polyplexes and Gal-SLPs (Figure 1b,g) also verified the successful decoration of the lipid layer. The drug loading content (DLC, %) and drug loading efficiency (DLE, %) of sorafenib in Gal-SLPs were 3.6% and 74.5%, respectively. Further, sorafenib was released faster in the acidic environment than in the neutral environment, which indicated more favorable enhanced intratumoral sorafenib release (Figure 1i). Scheme 1. Schematic illustration of a self-activated cascade-responsive co-delivery system (Gal-SLP) for synergetic cancer therapy. i) B-PDEAEA condenses USP22 shRNA (shUSP22) plasmids to form polyplexes. Polyplexes are further coated with a galactose-decorated sorafenib-loaded lipid layer to afford Gal-SLPs. After intravenous injection, Gal-SLPs accumulate in the tumor region, where the Gal ligands bind to the asialoglycoprotein receptors (ASGPRs) overexpressed on the cell membrane. The lipid layer fuses with the cell membrane and polyplexes are ejected into the cytosol. ii) Sorafenib is quickly released from Gal-SLPs and induces elevated intracellular reactive oxygen species (ROS), which triggers dissociation of polyplexes and release of shUSP22. iii) The released shUSP22 enters the nucleus for transcription to specifically degrade USP22 mRNA and further suppress USP22 expression. iv) The downregulation of USP22 not only suppresses multidrug resistance-associated protein 1 (MRP1), enhances the intracellular accumulation of sorafenib, and finally inhibits cell proliferation and cancer angiogenesis, but also inhibits glycolysis in hepatocellular carcinoma (HCC) cells, enhancing sorafenib chemosensitivity and impairing cell metabolism.
sorafenib and shUSP22 gene therapy. Obviously, Gal-SLPs exhibited much higher cytotoxicity against HCC cells (Figure 2a,c). The half-maximal inhibitory concentrations (IC 50 values) of Gal-SLPs against Huh-7 cells were 2.8 µм (in terms of sorafenib dose) and 0.28 µg well −1 (in terms of shUSP22 dose), compared with 6.1 µм for sorafenib and 0.91 µg well −1 for galactose-decorated liposomes without sorafenib (Gal-LPs), a 2.2fold decrease and a 3.3-fold decrease, respectively. The same phenomenon was observed in BEL-7402 cells. Based on the the cell viability profiles in Figure 2a,c, combination index (CI 50 ) of sorafenib and shUSP22 gene therapy was calculated via Equation (1) according to median-effect analysis. [13] CI 50 of sorafenib and shUSP22 therapy against Huh-7 and BEL-7402 cells were 0.759 and 0.713, respectively, which indicated a synergetic effect of sorafenib and shUSP22 combinational therapy (Gal-SLP) ( Table 1). The cell death-inducing capability of Gal-SLPs was further studied by a propidium iodide (PI) staining assay (Figure 2b,d and Figure S5, Supporting Information). The percentage of dead cells upon Gal-SLP treatment in Huh-7 cells was 16.3 ± 0.25%, significantly higher than that of sorafenib (11.5±0.67%) and Gal-LPs (9.5±0.82%) along, which was more obvious in BEL-7402 cells. Gal-SLPs led to 13.7 ± 2.0% BEL-7402 cells death. In contrast, PI-positive cell proportions in the sorafenib and Gal-LP groups were 7.6 ± 1.3% and 5.7 ± 2.1%, respectively.
The ability of Gal-SLPs to inhibit cell proliferation and colony formation was evaluated in HCC cells via 5-ethynyl-2'deoxyuridine (EdU) incorporation and colony formation assays. EdU, a type of thymidine nucleoside analog, can be incorporated into DNA and label cells undergoing DNA replication. All the treatments inhibited cell proliferation (Figure 2e-f). It should be noted that Gal-SLPs displayed an obviously better proliferation inhibition ability in contrast to the single treatment of sorafenib or Gal-LPs. The clonogenic assay revealed that Gal-SLPs induced a less colony formation and showed a long-term synergistic antitumor effect (Figure 2g-i). All these data demonstrated that Gal-SLPs were more efficient at decreasing the viability, proliferation and colony formation of HCC cells than sorafenib or Gal-LPs alone.

Evaluation of Intracellular ROS and Gene Trafficking
The ROS-responsive charge-reversal gene delivery system was utilized for shUSP22 delivery. The concentration of ROS in tumor cells could reach 0.5 nmol 10 4 cells −1 h −1 , which is higher than that in normal cells and avoids the therapeutic agents from off-target effects. [10c] However, intratumoral ROS are not high enough for the rapid and effective release of therapeutic agents and become an obstacle for the further development of ROSresponsive delivery systems. [14] Herein, we chose sorafenib, the first-line treatment for unresectable HCC, as an ROS inducer. Sorafenib induces ROS production and ROS-related cell death more selectively in HCC cells than normal cells both in vitro and in vivo, as indicated by previous studies. [11,15] The function of sorafenib as an inducer of ROS production was studied by flow cytometry and visualized by confocal laser scanning microscopy (CLSM). Dihydroethidium(DHE), which can be oxidized to fluorescent ethidium, was used as an ROS probe. Gal-SLPs rapidly triggered ROS production after a 6-h treatment, comparable to The data are presented as the mean ± SD (n = 3). d) Gel retardation assay of polyplexes at different N/P ratios after an 1-h incubation at 37°C with or without H 2 O 2 . e,f) Representative Western blot analysis (e) of USP22 downregulation in Huh-7 cells treated with polyplexes at different N/P ratios. Band intensities were semiquantified using IMAGE J software and normalized with -actin. Means are presented under the USP22 band (e) and data are presented as the mean ± SD (n = 3) (f). Polyethylenimine (PEI) and Lipofectamin 2000 (Lipo2000) were used as positive controls. Statistical differences and p value between two groups (versus N/P = 17) were shown as *p < 0.05, **p < 0.01, ***p < 0.001, and # p > 0.05. g) The size distribution measured by DLS and the morphology observed by TEM of Gal-SLP (0.2 µmol lipid/µg DNA; N/P = 17). The scale bar is 200 nm. h) The zeta potential of the B-PDEAEA/shUSP22 polyplex (N/P = 17) and Gal-SLP. The data are presented as the mean ± SD (n = 3). i) In vitro sorafenib release from Gal-SLPs in PBS containing 1% Tween 80 at pH 7.4 or 5.0. The data are presented as the mean ± SD (n = 3).
the results of free sorafenib (Figure 3a,b). Moreover, strong red fluorescence was detected in the confocal images of Huh-7 cells treated with sorafenib and Gal-SLPs (Figure 3c), which suggested a high intracellular ROS concentration. In contrast, Huh-7 cells in the control group had very weak fluorescence. These findings demonstrated that sorafenib released from Gal-SLPs in tumor cells can induce a sharp elevation in intracellular ROS, which is expected to oxidize B-PDEAEA and thus trigger a fast shUSP22 release for transcription.
We subsequently examined the USP22 downregulation efficiency of Gal-SLPs using a Western blot assay. USP22 expression in Huh-7 cells treated with Gal-SLPs was much lower than that in Gal-LP-treated cells, and USP22 was almost knocked out in Gal-SLP-treated BEL-7402 cells (Figure 3d), indicating that more shUSP22 was released from Gal-SLPs as a result of sorafenibinduced ROS elevation and was transcribed effectively to suppress USP22 expression. The intracellular dissociation and gene trafficking of Gal-SLPs in Huh-7 cells was investigated to reveal the mechanisms underlying the efficient USP22 downregulation, using doubly labeled lipopolyplexes consisting of Cy5labeled DNA ( Cy5 DNA, red) and FITC-labeled B-PDEAEA ( FITC B-PDEAEA, green) (Figure 3e,f). After a 2-h incubation with Gal-SLPs, some red dots were found in the cytosol, even in the nuclei, indicating fast DNA release from Gal-SLPs and entry into nuclei for transcription, although more red and green fluorescence still overlapped (yellow dots). And with incubation   expanding to 4 and 8 h, more red fluorescence was separated from the green fluorescence, and great red fluorescence was observed in the nuclear regions ( Figure 3e). However, even after 8 h incubation with Gal-LPs, few separated red dots were found in the nuclei, indicating slow polyplex dissociation and DNA release (Figure 3f). To further validate the role of sorafenib-or Gal-SLPinduced ROS in the quick release of shUSP22, N-acetylcysteine (NAC) was utlized as a typical ROS scavenger. As shown in Figure  S6a,b, Supporting Information, NAC efficiently lowered the concentration of Gal-SLP-induced ROS. More importantly, results of Western blot assay showed that ROS ablation by NAC inhibited the knockdown efficiency of USP22 by Gal-SLP ( Figure S6c, Supporting Information), which further supported that Gal-SLPinduced ROS participated in the quick release of shUSP22 for transcription. These results well proved that the charge-reversal B-PDEAEA/DNA polyplexes could dissociate quickly upon oxidation by the sorafenib-induced elevation in ROS levels, and release DNA for quick nuclear entry and subsequent efficient transcription.

MRP1 Expression and Intracellular Sorafenib Accumulation
The expression of USP22 dramatically affects the expression of MRP1, which mediates MDR in numerous carcinomas by pumping anticancer agents out of cells. [16] Downregulation of USP22 could inhibit the AKT/GSK-3 pathway and further suppress the expression of MRP1 (Figure 4a  differences in sorafenib accumulation among cells treated with sorafenib, Gal-SLP-NC and Gal-SLP-shUSP22. However, with the treatment extending to 24 and 48 h, intracellular sorafenib accumulation in the Gal-SLP-shUSP22 treatment group was ≈1× higher than that in the sorafenib and Gal-SLP-NC groups. These results together provided strong evidence that Gal-SLPs could induced an enhanced intracellular sorafenib accumulation through downregulating the expression of MRP1, thus hampering the efflux of sorafenib.

Glycolysis in Huh-7 Cells with Different Treatments
We established USP22-knockdown (SH) and USP22overexpressing (OE) HCC cells via lentiviral infection to explore the role of USP22 in sorafenib sensitivity. The results of Western blot assays verified the successful generation of Huh-7 and BEL-7402 cells with different expression levels of USP22 ( Figure S8a,b, Supporting Information). The expression of USP22 was associated with the sensitivity to sorafenib ( Figure  S8c-e, Supporting Information). Compared with USP22-NC SH and USP22-NC OE HCC (Huh-7 and BEL-7402) cells, USP22-SH HCC cells exhibited more sensitive to sorafenib, while USP22-OE cells were more resistant to sorafenib. Meanwhile, glycolysis stress assay revealed that overexpression of USP22 increased the extracellular acidification rate (ECAR) of Huh-7 cells (Figure 5a), whereas downregulation of USP22 decreased the ECAR (Figure 5c,d). Thus, we speculated that the downregulation of USP22 by Gal-SLPs could block the glycolysis and further suppress stemness features in HCC cells. Accordingly, a glycolysis stress assay was performed in Huh-7 cells with different treatments. Sorafenib greatly stimulated glycolysis of Huh-7 cells, as evidenced by the greatly enhanced ECAR (Figure 5e,f) and lactate production (Figure 5g), which was one of the main reasons for the acquired drug resistance in clinical. [17] Meanwhile, downregulation of USP22 by Gal-LPs could reduce the ECAR and Gal-SLPs offset the increase of ECAR triggered by sorafenib. Furthermore, a cytotoxicity assessment of sorafenib, Gal-LPs, Gal-SLPs, 2-Deoxy-d-glucose (2-DG) and a combination of 2-DG and sorafenib was executed (Figure 5h). 2-DG, a typical glycolysis inhibitor, sensitized HCC cells to sorafenib. It is worth noting that Gal-SLPs inhibited more cell proliferation than sorafenib and Gal-LPs alone. The inhibitory effect on glycolysis and the synergistic cell-killing activity strongly indicated that Gal-SLPs have an enhanced antitumor effect in a glycolysis-inhibited manner.
to further enhance tumor accumulation. [19] The in vivo realtime distribution of lipopolyplexes was tracked on HCC PDXbearing BALB/c mice after a single intravenous injection of DiR-loaded SLP (SLP/DiR) or Gal-SLP (Gal-SLP/DiR). As exhibited in Figure 6a, a strong fluorescence derived from DiR was observed throughout the whole body and gradually accumulated in the tumor regions in both the SLP/DiR and Gal-SLP/DiR groups. Notably, the fluorescence signals of SLP/DiR and Gal-SLP/DiR decayed slowly, which indicated that lipopolyplexes have a long in vivo circulation period. Compared with SLP/DiR, obviously enhanced fluorescence signal was detected in the tumor region in the Gal-SLP/DiR group 9 h post-injection and the high contrast between the tumor area and surrounding tissues sustained throughout the following experimental period. At 72 h postinjection, tumors and organs were excised from the mice and ex vivo imaging was performed (Figure 6b). The tumor accumulation of DiR in the Gal-SLP/DiR group was clearly higher than that in the SLP/DiR group, demonstrating an effective tumortargeting capability of galactose-functionalized Gal-SLP.
An in vivo pharmacokinetic study was conducted to investigate drug retention in the blood circulation. Blood was collected from ICR mice after a single intravenous injection of sorafenib, SLP and Gal-SLP at a sorafenib dose of 5 mg kg −1 and sorafenib content in the blood was quantified by HPLC analysis. The concentration-time profile and pharmacokinetic parameters of sorafenib are presented in Figure 6c and Table 2. Compared with the rapid clearance of sorafenib, both SLPs and Gal-SLPs exhibited substantially extended circulation in vivo. The area under the curve (AUC 0-t ) values of SLP and Gal-SLP were 71.4 ± 14.6 and 58.7 ± 8.98 µg × h mL −1 , respectively, while the AUC 0-t value of free sorafenib was relatively low (16.3 ± 3.48 µg × h mL −1 ). These data supported that lipopolyplexes helped prolong the blood circulation of sorafenib, which was in accordance with the in vitro sorafenib release kinetics and biodistribution study.

In Vivo Antitumor Effects on Sorafenib-Insensitive HCC Patient-Derived Xenografts
The patient-derived xenografts (PDX) applied in this study was originated from a patient, who didn't have any chemotherapeutic or locoregional treatments before the surgery. The patient was treated with transarterial chemoembolization (TACE) thrice, radiofrequency ablation (RFA) once and long-term oral administration of sorafenib immediately after HCC recurrence and reduced the sorafenib dosage to half due to the severe handfoot skin reaction. The time to progression (TTP) of this patient was 5.5 months, less than the median TTP (6.3 months for the combination of TACE and sorafenib) reported previously, which indicated that the patient responded poorly to the combinational therapy. [20] Finally, the patient changed treatment plans due to rapid tumor progression. The illustration of establishing sorafenib-insensitive HCC PDX models and tumor progression of Patient-25 was shown in Figure 7a,b.
We explored whether our strategy to co-delivery sorafenib and shUSP22 via Gal-SLP was promising for enhancing curative effect on the sorafenib-insensitive HCC PDX model. Saline, sorafenib, Gal-LP and Gal-SLP were intravenously injected every other day for 6 times at a sorafenib dose of 5 mg kg −1 and shUSP22 dose of 0.5 mg kg −1 . Compared with saline, all sorafenib, Gal-LP and Gal-SLP led to significant tumor growth inhibition, although a single treatment of sorafenib or Gal-LP showed very limited anticancer efficacy with a fast tumor rebound during treatment-free period (Figure 7c,d, Figure S9, Supporting Information). However, Gal-SLPs exhibited much better antitumor efficiency, strongly suppressing the tumor growth during the whole experimental period, resulting in a tumor-growth inhibition rate (TIR %) of 82.7 ± 4.9% in terms of tumor weight at the end of experiment, significantly higher than that of sorafenib (53.4 ± 15.5%) and Gal-LPs (55.0 ± 7.8%) (Figure 7d,e, Figure S10, Supporting Information). Histological analysis via hematoxylin and eosin (H&E) staining of the tumors tissue sections showed numerous apoptotic cells with extensive vacuolization, severe nucleus shrinkage and decreased cellularity in the Gal-SLP-treated group, compared with the tightly-packed tumor cells in the saline group and much reduced apoptotic characteristics in sorafenib and Gal-LP groups, supporting the strong tumor inhibition activity of Gal-SLP (Figure 7g). The results of immunohistochemistry (IHC) staining revealed that Gal-SLPs efficiently suppressed the expression of Ki67, a marker for tumor proliferation, further confirming the excellent antitumor capability of Gal-SLP. Besides, Gal-SLPs downregulated in vivo USP22 expression to a much greater extent than Gal-LPs (Figure 7g), which was well correlated with the in vitro Western blot analysis, convincingly proving the self-activated cascade-responsive process and synergy of sorafenib and shUSP22.
For biosafety evaluation, organs were excised from nude mice on day 30 and analyzed by H&E staining. There was no notable damage in any of the treatment groups ( Figure S11, Supporting Information). The administration of sorafenib, Gal-LPs or Gal-SLPs did not influence the weights of mice (Figure 7f). In vivo safety evaluation, performed in the healthy ICR mice, revealed no renal or liver damage ( Figure S12, Supporting Information). These data supported the excellent safety profile of Gal-SLPs.

Discussion and Conclusion
Sorafenib is the first FDA-approved MKI for unresectable HCC and remains the first-line therapy. However, only ≈40% of HCC patients can benefit from sorafenib owing to cancer stemness and other reasons. [1a,2b,3] Methods for reversing cancer stemness and sensitizing HCC patients to sorafenib remain some of the largest concerns for improving the prognosis of HCC patients. USP22, as a cancer stem cell (CSC) marker, plays a critical role in cancer stemness and has drawn more and more attention in the recent researches. [5a,6,7b,21] USP22-targeted gene therapy seems to be a promising way out of the current dilemma in HCC therapy. To this end, we developed a self-activated cascade-responsive sorafenib and shUSP22 co-delivery system (Gal-SLP) for synergetic HCC therapeutics.
Characterizations of the sorafenib/shUSP22 co-delivery nanoplatform showed that Gal-SLPs, with an average diameter of ≈100 nm, exhibited excellent ROS responsiveness and high gene transfection efficiency. Sorafenib, entrapped in the Gal-SLPs, displayed sustained and acid-favored release kinetics. Once Gal-SLPs entered HCC cells, sorafenib was released immediately and worked as an ROS inducer to trigger the rise of intracellular ROS to further induce a rapid release of shUSP22 from polyplexes. Only a few Huh-7 and BEL-7402 cells treated with Gal-SLPs expressed USP22. Accordingly, a series of in vitro cytotoxicity experiments demonstrated that Gal-SLPs were more efficient than sorafenib and Gal-LPs suppressing the growth, proliferation and colony formation of HCC cells. Additionally, cloaking the surface with a galactose-decorated lipid layer not only extended the blood circulation of Gal-SLPs, but also enabled the nanoplatform to specifically co-deliver sorafenib and shUSP22 to the tumor region.
In our study, USP22 shRNA gene therapy was utilized to reverse HCC stemness and sensitize HCC cells to sorafenib. On the one hand, our previous work has indicated that USP22 is closely associated with HCC stemness through the modulation of glycolysis. [5a] In this study, we observed that the downregulation of USP22 by lentiviral infection or shUSP22 gene therapy (Gal-LP) was able to suppress glycolysis of Huh-7 cells. However, glycolysis was activated by sorafenib administration and this phenomenon was believed to connect tightly with cancer stemness and to further mediate resistance to sorafenib. [22] Gal-SLPs inhibited the potential activation of glycolysis by sorafenib and sensitized HCC cells to sorafenib, which was familiar with the sorafenib sensitabling effect of 2-DG. On the other hand, USP22 was proven to participate in MDR by modulating MRP1 through ii) Postoperative changes in the liver. iii) One month after surgery and tumor recurrence. iv) Half a year after surgery and tumor progression. The white arrow indicates the tumor region, the black arrow indicates the titanium clip after the surgical resection, and the red arrows indicate the tumor recurrence and progression. c) The tumor growth curves after intravenous injection of saline, sorafenib (formulated in Cremophor EL/ethanol, v/v = 1:1), Gal-LP and Gal-SLP (n = 6, sorafenib dose, 5 mg kg −1 ; shUSP22 dose, 0.5 mg kg −1 ) according to a q2 × 6 regimen. d) Images of the whole-body of the mice at the experimental endpoint. e) The average tumor weight of the excised tumor in each group at the experimental endpoint (n = 6). f) Body weight variation in the tumor-bearing mice during the experimental period (n = 6). g) Representative images of IHC staining (USP22 and Ki67) and H&E staining of tumors on day 30. *p < 0.05, **p < 0.01, ***p < 0.001, and # p > 0.05. the AKT/GSK-3 pathway. [8b] Our study provided strong evidence that the downregulation of USP22 by Gal-SLPs suppressed the expression of MRP1 and caused high intracellular sorafenib accumulation. Sorafenib accumulation not only suppressed the cell proliferation and cancer angiogenesis, but also generated an ROS-responsive positive feedback loop to trigger more shUSP22 release and sorafenib accumulation. Our study showed that Gal-SLP-shUSP22 induced higher intracellular sorafenib accumulation in Huh-7 cells after 24 and 48 h incubation. Furthermore, recent studies indicated USP22 played an important role in modulating of PD-L1 expression in cancer cells and maintaining T reg cell suppression function, [7a,21,23] which suggested that USP22targeted gene therapy was not only essential to sensitive cancer cells to sorafenib, but also meaningful for modulating the cancer immune microenvironment and enhancing the efficiency of anticancer immunotherapy. This is a focus area of our future work.
Identification of the most effective, least toxic, and costeffective treatment plan for cancer patients is still the main problem faced by clinicians. PDX models have been widely adopted as a powerful tool in the evaluation of drug efficiency and offer assistance in making individual precision medical plans because PDX models can retain the unique features (such as gene patterns, histological properties and responses to drug treatment) of tumors in human patients. [24] In this study, a PDX model was established as an avatar of a sorafenib-insensitive HCC patient for in vivo antitumor evaluation. Once injected into mice, more Gal-SLPs accumulated in the tumor site owing to the EPR effect and active targeting. And contributed by self-activated cascade-responsive design, shUSP22 encapsulated in Gal-SLPs was apt to release more rapidly than that in Gal-LP. As a result, Gal-SLP dramatically suppressed the USP22 expression in the tumor and lead a great proportion of apoptotic cells. Compared with sorafenib or Gal-LP single treatment, Gal-SLP was more potent in this HCC PDX model to inhibit the tumor growth. The excellent in vitro and in vivo antitumor efficiency of Gal-SLPs suggested that the patient from whom, the PDX originated, might benefit from this sorafenib and shUSP22 combination therapy.
In summary, we developed a self-activated cascade-responsive sorafenib and shUSP22 co-delivery system (Gal-SLP) for synergetic HCC therapeutics. Gal-SLPs exhibited potent antitumor efficiency via a trio synergetic effect and are expected to be promising for HCC therapy.

Experimental Section
Preparation and Characterization of Polyplexes with Different N/P Ratios: HEPES buffer solution (10 mm, pH = 7.4) was used to dilute plasmid DNA to a concentration of 40 µg mL −1 . According to the preset N/P ratios (N/P = 5,9,13,17,21,25), B-PDEAEA was dissolved in HEPES buffer solution (10 mm, pH = 7.4) at various concentrations. And then, an equal volume of DNA solution was added to the B-PDEAEA solution. The mixture was immediately vortexed for 10 s, followed by incubating statically for 30 min. Sizes and zeta potentials of polyplexes were measured by dynamic light scattering (DLS) (Malvern, UK).
USP22 Knockdown Efficiency and Cytotoxicity Assessments of Polyplexes: For USP22 knockdown efficiency assessment, 150 000 HCC cells in 6well plates were treated with polyplexes containing shUSP22 plasmids at a shUSP22 concentration of 8 µg well −1 for 48 h and then harvested. Lipo2000 and branched polyethylenimine (PEI, 25 kDa) were used as positive controls. Western blot assay was performed as described previously. [5a,8b] Bands were incubated with primary antibody dilution overnight, followed by incubation with secondary antibody dilution for 2 h. Finally, bands were visualized with ECL detection reagents (Fude Biological Tech.). All antibodies were diluted according to manufacturers' recommendations and supplied in Table S1, Supporting Information.
Cytotoxicity of polyplexes was determined by CCK-8 assay (Dojindo). 3000 HCC cells were seeded into 96-well plates and incubated overnight. Polyplex contained negative control (NC) plasmids with different N/P ratios was added into medium at a DNA concentration of 0.5 µg well −1 and incubation was continued for an additional 48 h. When the incubation was finished, the medium was replaced with 110 µL fresh medium containing 10 µL CCK-8 solution. After a further incubation for 2 h, the absorbance of 450 nm of each well was measured via microplate reader (ELx800, Bio Tek).
DNA Encapsulation Efficiency and ROS-Responsiveness Assessment of Polyplexes: DNA encapsulation efficiency of polyplexes with various N/P ratios was evaluated by gel retardation assay. The polyplexes were loaded into a 0.8% agarose gel and electrophoresed in TBE buffer at 120 V for 45 min using GT Mini-Gel Casting System (BIO-RAD). There existed Gel Red (Biotium) in the agarose gel for DNA detection. For ROSresponsiveness assessment, polyplexes were treated with 0.5 mm H 2 O 2 at 37°C for 1 h, followed by electrophoresis.
The changes of sizes and zeta potentials of polyplexes were measured using DLS. Polyplexes (N/P = 17) were incubated with 200 µм H 2 O 2 at 37°C. At predetermined timepoints, 200 µL of sample was collected for size and zeta potential measurements.
Preparation and Characterization of Galactose-Decorated Sorafenib-Loaded Lipopolyplexes (Gal-SLPs, N/P = 17): Galactose-decorated sorafenib-loaded lipopolyplex was prepared by the thin-film hydration method. [9] The molar ratio of sorafenib/DOPE/CHEMS/ DSPE-PEG2000galactose was fixed at 1.0:6.9:1.8:1.2. Sorafenib (0.23 mg), DOPE (2.53 mg), CHEMS (0.43 mg), and DSPE-PEG2000-galactose (1.84 mg) were dissolved in 2 mL chloroform and the organic solvent was removed by rotary evaporation to afford a thin lipid film. The film was hydrated overnight in 1 mL HEPES (10 mm, pH 7.4) at room temperature with stirring, followed by ultrasonication for 10 min in ice bath. The lipopolyplex solution was then stored at 4°C for 2 h and filtered through a 0.22 µm nylon filter to remove non-encapsulated drug aggregates. Gal-SLPs was prepared as previously published. [9a,25] The previous lipopolyplex solution (500 µL) was added to the polyplexes solution (N/P = 17, 500 µL, 10 µg DNA) and incubated overnight at room temperature. The size and zeta potential of Gal-SLP were assessed using DLS and the morphology of Gal-SLP was observed and imaged by TEM (TECNAL 10, Philips) after negative staining. For the determination of drug loading content (DLC) and drug loading efficiency (DLE), Gal-SLP was lyophilized and redissolved in acetonitrile for HPLC analysis. Release process of sorafenib from Gal-SLP was explored via the dialysis method. Briefly, Gal-SLP solution (2 mL, containing 0.5 mg sorafenib) was sealed in a dialysis bag (MWCO = 3500 Da) and dialyzed against 60 mL of PBS containing 1% tween 80 at different pH (pH = 5.0 or 7.4). At predetermined time intervals, 100 µL of the dialysate was withdrawn for HPLC analysis.
In Vitro Cytotoxicity Assay: In vitro cytotoxicity assay was determined by CCK-8 assay and PI staining assay. HCC cells were seeded into 96-well plates at a density of 3000 cell/well and cultured overnight before treatment. Then, sorafenib, Gal-LP and Gal-SLP were added into medium at virous sorafenib and shUSP22 concentrations. After incubation for 48 h, cell viability was assessed by CCK-8 assay as mentioned above. For PI staining, 100 000 HCC cells (Huh-7 and BEL-7402) were plated into 6-well plates. Sorafenib, Gal-LP and Gal-SLP were added at a sorafenib concentration of 5 µм and shUSP22 concentration of 5 µg well −1 . After treatment for 36 h, cells were harvested and stained with 5 µL PI solution (1 µg mL −1 , KeyGen) for 15 min. PI-positive cells proportion was quantitatively investigated by flow cytometer (BD FACSCanto II).
Synergy of Sorafenib and shUSP22 Gene Therapy: Based on the cell viability profiles in Figure 2a,c, combination index (CI 50 ) was calculated via Equation (1) according to median-effect analysis. [13] CI 50 = IC sorafenib in Gal−SLPs IC single sorafenib + IC shUSP22 in Gal−SLPs IC shUSP22 in Gal−LPs (1) where CI 50 represents the combination index of sorafenib and shUSP22 gene therapy in the combination (Gal-SLP) that inhibits a half of cells; IC sorafenib in Gal-SLPs and IC shUSP22 in Gal-SLPs are the concentration of sorafenib and shUSP22 in Gal-SLPs that inhibits 50% of cells; IC single sorafenib and IC shUSP22 in Gal-LPs are the concentration of sorafenib and shUSP22 (Gal-LPs), respectively, that inhibits 50% of cells individually. The classifications of synergy are additive (CI 50 = 1), synergistic (CI 50 <1), or antagonistic (CI 50 >1). IC 50 value was calculated using GraphPad Prism software. Western Blot Analysis: To detect the USP22 and its related protein (MRP1, p-GSK-3 , p-AKT) expression levels, 200 000 HCC cells were plated into Petri dishes. After treatment with sorafenib, Gal-LPs and Gal-SLPs for 48 h at a sorafenib concentration of 5 µм and shUSP22 concentration of 8 µg dish −1 respectively, cells were harvested, followed by Western blot assay as mentioned above. All antibodies were diluted according to manufacturers' recommendations and supplied in Table S1, Supporting Information.
Cell Proliferation Assay: Cell proliferation was investigated by staining cells with Click-iT EdU Imaging Kit (Invitrogen). EdU staining was carried out according to manufacturer's protocols. In a nutshell, 20 000 HCC cells were seeded in glass bottom petri dishes. After the same treatment as mentioned in the Western blot analysis for 24 h, an equal volume of 20 µм EdU solution was added into medium. Cells were incubated for further 30 min, followed by cell fixation with 4% paraformaldehyde (PFA) fixation solution and permeabilization with 0.5% Triton X-100 solution. Alexa Fluor 488 azides were added and incubated for additional 30 min at room temperature and cell nuclei were stained with Hoechst 33 342 (Thermo) for 30 min. The variations of fluorescence were detected using a fluorescence microscopy.