Dual-site spatiotemporal and simultaneous inhibition on PIN1 via arsenic-retinoic albumin nanoparticles enables synergistic oncotherapy.

Introduction
A central signaling mechanism shared by many heterogeneous oncogenic pathways is proline-directed phosphorylation, which is tightly regulated by a variety of kinases and phosphatases. The functional consequences of these phosphorylation events are further governed by a single, unique enzyme: peptidyl-prolyl cis–trans isomerase NIMA-interacting 1 (PIN1). PIN1 is the only known isomerase that specifically catalyzes the conformational transformation of phosphorylated Serine/Threonine-Proline (pSer/Thr-Pro) motifs, acting as a master molecular switch that profoundly impacts the fate of numerous phosphoproteins (Scheme 1). Functionally, PIN1 amplifies oncogenic signaling by upregulating over 50 oncogenes or proliferation-promoting factors, while simultaneously inhibiting more than 20 tumor suppressors or proliferation-restraining proteins [1]. Elevated PIN1 activity is frequently observed in a broad range of human cancers, where it contributes to tumor progression, maintenance of cancer stem cells, and metastatic dissemination by disrupting the oncogene-tumor suppressor balance.
For decades, the predominant model in anticancer drug development has been the “one target: one drug” approach using small molecules or biologics [2]. However, this paradigm often fails in complex diseases like cancer, where redundant and interconnected signaling pathways can easily bypass single-target inhibition. In clinical settings, such limitations have driven a shift toward multi-targeted therapies that offer improved efficacy and reduced resistance [3]. Given that PIN1 regulates multiple, independent oncogenic cascades, we hypothesized that effective inhibition of PIN1 could block a wide spectrum of downstream tumor-promoting pathways, thereby achieving synergistic suppression of cancer growth and metastasis with minimized toxicity.
We previously identified all-trans retinoic acid (ATRA) as a direct inhibitor of PIN1 through mechanism-based screening [4]. ATRA binds to the WW domain of PIN1, thereby blocking its interaction with pSer/Thr-Pro motifs and disrupting multiple oncogenic signaling cascades. Clinically, however, ATRA has been approved only for the treatment of acute promyelocytic leukemia (APL), a hematologic malignancy, due to its differentiation-inducing activity [5–7]. It has not been effective or approved for use in solid tumors.
Arsenic trioxide (ATO) has also been identified as a PIN1 inhibitor, interacting non-covalently with the catalytic domain of the protein [8]. In combination with ATRA, ATO exhibits complementary inhibitory effects on PIN1 [8]. Despite this mechanistic synergy, the clinical application of ATRA and ATO remains limited to APL. Outside the hematologic setting, particularly in solid tumors, their therapeutic use is hindered by significant pharmacological challenges: ATRA is poorly soluble in water, unstable under light and heat, and rapidly cleared from circulation. ATO suffers from low tumor cell uptake, fast systemic clearance at subtherapeutic doses, and dose-limiting off-target toxicity [9–11]. Moreover, although ATRA and ATO are co-administered for APL treatment, their combination has produced limited efficacy and substantial systemic side effects [9, 10]. Currently, to our knowledge, there have been no reports on the combination of clinical drugs and ATO for the treatment of solid tumors, despite the existence of a very few reports on combinations with ATRA [12]. The combination of multiple drugs for cancer therapy presents significant challenges including unpredictable synergy, complex mechanisms, overlapping toxicities, lack of clear targets, drug delivery hurdles, and intricate trial design, particularly when combining two drugs with opposite solubility properties [13–15].
To date, no formulation has successfully co-delivered ATRA and ATO in a single, integrated platform for therapeutic targeting of solid tumors. Notably, co-encapsulation of ATRA and ATO within albumin-based nanoparticles has not yet been investigated as a strategy for spatiotemporal and simultaneous multi-domain PIN1 inhibition. Such a unified nanocarrier system may overcome the individual pharmacokinetic limitations of each agent [16], enhance tumor-selective delivery, and unlock synergistic, multi-pathway PIN1 inhibition in solid tumors.
To achieve this goal, the present study aimed to develop a novel synergistic nanomedicine by co-conjugating ATRA and ATO onto human serum albumin (HSA), resulting in the dual-site Pin1 inhibition via arsenic-retinoic albumin nanoparticles (ATRA-ATO-NPs). HSA was chosen as the carrier due to its excellent biocompatibility, long circulation half-life (~ 19 days), and unique binding capacity for diverse molecules [17, 18]. Unlike other materials used for the preparation of delivery system (e.g., liposomes, cationic polymers, micelles), HSA exhibits minimal immunogenicity and toxicity. Structurally, HSA contains one free cysteine (Cys34) and several binding pockets that allow efficient drug conjugation [19]. Our prior work demonstrated that aptamer-conjugated HSA nanoparticles significantly extended circulation time in mice by threefold [20], further validating HSA as a multifunctional drug carrier.
Building on this rationale, we successfully synthesized ATRA-ATO-NPs through site-directed conjugation of ATRA and ATO to HSA (Scheme 1). These nanoparticles displayed favorable physicochemical properties, including uniform spherical morphology, controlled release behavior, and effective tumor-targeting via the enhanced permeability and retention (EPR) effect. Functionally, ATRA-ATO-NPs exhibited significantly enhanced synergistic inhibition of hepatocellular carcinoma cell proliferation and metastasis compared to free ATRA, free ATO, or their co-administration. Mechanistically, they more potently suppressed the expression of PIN1 and its downstream effectors—β-catenin, NF-κB, Cyclin D1, CDK2, and C-myc in HuH7 and PLC cells—alongside key metastatic and epithelial–mesenchymal transition (EMT) markers, including N-cadherin, E-cadherin, MMP2, and SNAIL. In vivo murine studies confirmed that ATRA-ATO-NPs significantly suppressed tumor growth and lung metastasis without inducing acute toxicity. Proteomic and bioinformatic analyses revealed that their antitumor activity stems from coordinated inhibition of multiple PIN1-regulated pathways, including oncogenic signaling, metabolism, and cell cycle control. Detailed results of the study are presented below.

Results and discussion

Synthesis and characterization of ATRA-ATO-NPs
We utilized HSA as an ideal carrier for loading and co-delivering both ATRA and ATO due to its inherent advantages such as fully biocompatibility, safety, and long-circulation [20]. As shown in Scheme 1A, addition of glutathione (GSH) to the HSA solution helped to reduce the disulfide bonds of HSA to the free thiol group. We then added ATO to the reduced HSA solution. The ATO fused into HSA and conjugated with the latter. Next, the ATRA-containing ethanol was injected into the ATO-HSA solution followed by the albumin self-assembly technique under desolvation condition to facilitate the formation of intermolecular disulfide bonds of HSA, resulting in encapsulation of ATRA and ATO into the hydrophobic cavity of HSA as demonstrated by the changes in the spectral scan analysis after each step of biochemical reactions (Fig. 2A). As shown in Fig. 2B, the blank NPs and ATRA-ATO-NPs exhibited good spherical shape, smooth surface and uniform particle size. Dynamic light scattering measurements showed that blank NPs exhibited an average particle size of 123.6 ± 6.39 nm with a narrow polydispersity index (PDI) of 0.175 ± 0.024. In comparison, the particle size of ATRA-ATO-NPs slightly increased to 176 ± 9.26 nm, with a PDI of 0.167 ± 0.013(Fig. 2C and D). These low PDI values indicate good dispersity of the nanoparticles. The particle size of approximately 176 nm for ATRA-ATO-NPs suggests their potential to concentrate in the tumor sites via the enhanced permeability and retention (EPR) effect to target tumor cells [21]. Zeta potential of blank NPs was − 6.57 ± 0.25 mV, and that of ATRA-ATO-NPs decreased to −25.95 ± 1.72 mV, indicating that ATRA-ATO-NPs exhibit superior stability compared to blank nanoparticles (Fig. 2E). To evaluate the drug encapsulation performance, drug loading (DL) and encapsulation efficiency (EE) of ATRA-ATO-NPs were further tested. The DL and EE of ATRA loaded in ATRA-ATO-NPs nanoparticle were 4.1 ± 0.3% and 72 ± 5%, meanwhile the DL and EE of ATO loaded in ATRA-ATO-NPs nanoparticle were 0.3 ± 0.02% and 39 ± 10% (Fig. 2F and G). These results suggested that ATRA-ATO-NPs possessed excellent properties for drug delivery. To confirm the molar ratio of ATRA to ATO in ATRA-ATO-NPs system, we tested the molar content of ATRA and ATO in ATRA-ATO-NPs. As shown in Fig. 2H, the molar ratio of ATRA to ATO encapsulated in ATRA-ATO-NPs was 10.6 ± 0.8, which was close to the expected ratio of 10:1.
The release of ATRA and ATO from ATRA-ATO-NPs were carried out in three different media, namely, phosphate-buffered saline (PBS), Dulbecco’s Modified Eagle Medium (DMEM) and serum-containing DMEM (Fig. 2I and J). ATRA release from ATRA-ATO-NPs in the three media gradually increased with time. However, the release rates of ATRA were significantly different. In PBS, the cumulative release of ATRA was 1.3% in the first day, and finally reached about only 11.4%. In DMEM, the cumulative release of ATRA was 4.8% on day 1 and 11.8% on day 20, whereas in 5% serum-containing DMEM, about 58.5% of ATRA were released on the first day, and 75.4% by the end of day 20. The results suggest that ATRA-ATO-NPs could slowly release ATRA, while serum and release medium itself distinctly affect the release rate of ATRA (Fig. 2I). The differences in the release rate of ATRA in the presence and absence of serum might be caused by the retinol binding proteins in the serum [19]. In contrast, Fig. 2J showed that the cumulative release of ATO was 36.3% at 2 h and 57.6% at 48 h in PBS, meanwhile, the cumulative release of ATO in DMEM was 46.5% at 2 h and 55.8% at 24 h. In 5% serum-containing DMEM, the cumulative release of ATO was 52.5% at 2 h and 63.6% at 48 h. These results indicated that ATO was able to be released from ATRA-ATO-NPs, but the release rate of ATO was significantly faster than that of ATRA, which might be due to the low concentration of ATO in ATRA-ATO-NPs. Overall, ATRA-ATO-NPs could release ATRA and ATO slowly in the tested media.

Cell uptake of ATRA-ATO-NPs
We used fluorescein isothiocyanate isomer I (FITC) loaded HSA nanoparticles (FITC-NPs) to evaluate their uptake in cancer cells. The distribution of FITC-NPs in HuH7 cells at 4, 12, 24 and 48 h were shown in Fig. 2K. At 4 h of incubation, the images showed distinct green fluorescence in the cells, indicating that FITC-NPs had entered into the cytoplasm and nucleus. This fluorescence intensity increased significantly by 12 h, demonstrating that more FITC-NPs entered into the cells over time. However, after 24 h of incubation, only a few fluorescent spots were observed in the cells, indicating that the fluorescent signal began to markedly decline, and the fluorescence in the cells had largely disappeared by 48 h. The above results revealed the time-dependent cellular uptake of the nanoparticles in cancer cells. The results were consistent with our previous findings that nanoparticles can be taken into the cytoplasm via endocytosis [22]. We further evaluated the cellular uptake of FITC-NPs by macrophage RAW 264.7 cells. As shown in Fig. S1, only a very few fluorescent spots were detected at 4 and 12 h post-incubation, and these signals had completely disappeared by 24 and 48 h. The absence of intracellular fluorescence suggested that the nanoparticles were not internalized by the RAW 264.7 cells, thereby indicating negligible phagocytic uptake. Based on the fact that ATRA-ATO-NPs can slowly release ATRA and ATO, and delivered them into the cancer cells, we expected that ATRA-ATO-NPs might synergize the pharmacological capability of ATRA and ATO to restrain cancer cell growth.

ATRA-ATO-NPs enhanced synergistic inhibition of ATRA and ATO on tumor growth in vitro
To further verify whether ATRA-ATO-NPs enhanced cellular uptake of ATRA and ATO, intracellular concentrations of ATRA and ATO were quantified in HuH7 cells treated with ATRA, ATO, and ATRA-ATO-NPs for 4, 24, and 48 h (Fig. 3A and B). The intracellular ATRA concentration in ATRA-treated cells decreased over time, from 4.78 ± 0.33 µg/10⁶ cells at 4 h to 4.2 ± 0.7 µg/10⁶ cells at 24 h, and sharply declined to 1.76 ± 0.19 µg/10⁶ cells by 48 h. Whereas in the ATRA-ATO-NPs-treated group, the level rose from an initial 2.38 ± 0.26 µg/10⁶ cells to 3.52 ± 0.8 µg/10⁶ cells and was maintained at 3.35 ± 0.85 µg/10⁶ cells at 48 h, remaining stable with no significant decrease (Fig. 3A). The intracellular ATO concentration in the ATO group showed a sharp decline over time, from 144.69 ± 4.55 ng/10⁶ cells at 4 h to 71.15 ± 36.67 ng/10⁶ cells at 24 h, and was maintained at 61.28 ± 7.38 ng/10⁶ cells at 48 h. Conversely, in the ATRA-ATO-NPs-treated group, the intracellular ATO content remained at around 80 ng/10⁶ cells within 48 h (Fig. 3B). Collectively, the ATRA-ATO-NPs exhibit a sustained-release profile, in sharp contrast to the rapid clearance of ATRA or ATO.
The inhibitory effects of free ATRA and ATO on cell growth were tested by the cell viability assays after incubation of cells with various doses of ATRA (5, 10, 20, 30, 40 µM) or ATO (0.25, 0.5, 1, 1.5, 2 µM) for 24 h, 48 h and 72 h, respectively. As a single agent, ATRA produced dose-dependent inhibition of cell growth of two HCC cells (HuH7 and PLC), but ATRA did not significantly inhibit growth of the human normal liver cells L-02 (Fig. 3C-E). ATO also caused dose-dependent inhibition of HuH7 and PLC, but did not affect the L-02 cell viability (Fig. 3C, D, and F). To determine whether the combination of ATRA and ATO, at the molar ratio of 10:1 for ATRA to ATO, has a synergistic inhibitory effect on HCC cells, we treated HuH7 and PLC cells with either ATRA (5, 10, 20 µM), ATO (0.5, 1, 2 µM), separately, or their combination under the corresponding concentration for 72 h. To further confirm whether other molar ratios for ATRA to ATO also result in synergism, we treated HuH7 cells with either ATRA (5, 10, 15, 30 µM), ATO (1, 2 µM), separately, or their combination (Fig. 3K). The combination of ATRA and ATO at the molar ratio of 10:1 for ATRA to ATO could significantly enhance the inhibitory effect of these two drugs on HuH7 and PLC cell growth in a dose-dependent manner (Fig. 3C and D). In contrast to the cancer cells, the combination of ATRA and ATO showed no significant inhibitory effect on the growth of normal liver L-02 cells (Fig. 3G). When the molar ratio for ATRA to ATO was 10:1, the combination index ranged from 0.3 to 0.7, indicating a synergistic effect (Fig. 3L and M), which was consistent with our previous study [8]. Oppositely, at other molar ratios for ATRA to ATO such as 5:1 or 15:1, the combination index was above 0.7, indicating the slight synergism, or nearly additive effect. Therefore, the molar ratio of 10:1 for ATRA to ATO was the optimal drug ratio in ATRA-ATO-NPs system. Figure 3N and O showed the cell viability of HuH7 and Hepa1-6 cells treated with ATRA, ATO, ATRA + ATO, ATRA-NPs, ATO-NPs or ATRA-ATO-NPs for 72 h. Compared with the control group, ATRA and ATO only had a slight inhibitory effect on the viability of HuH7 and Hepa1-6 cells. In contrast, the ATRA + ATO and ATRA-ATO-NPs showed significant inhibition on the viability of HuH7 and Hepa1-6 cells, indicating that the combination of ATRA and ATO had a synergistic effect on inhibiting cell growth. Meanwhile, ATRA-NPs and ATO-NPs also showed the significant inhibitory effects on cell viability, which was comparable to the inhibitory effect of ATRA + ATO. More importantly, the inhibitory effect of ATRA-ATO-NPs on cell viability was significantly higher than that of ATRA + ATO, ATRA-NPs, and ATO-NPs. These results confirm that ATRA-ATO-NPs can significantly enhance the synergistic inhibitory effect of ATRA and ATO on cancer cell growth. The blank NPs almost did not cause the inhibition on HuH7, PLC, and L-02 cells even under the high concentration of 0.6 mg mL− 1 (corresponding to the concentration of ATRA-ATO-NPs containing 40 µM ATRA and 4 µM ATO; Fig. 3H-J), suggesting its great biosafety. All the results suggest that ATRA-ATO-NPs could significantly enhance the synergistic inhibition of ATRA and ATO on cancer cell growth, and the nanoparticles themselves had excellent biocompatibility. By the controlled release of ATRA/ATO and internalization of nanoparticles in cancer cells to overcome the defects of ATRA and ATO such as short half-life, poor water solubility and side effects, ATRA-ATO-NPs not only greatly enhance the anti-tumor efficacy of ATRA and ATO, but also are expected to greatly improve the safety of the two drugs in clinical practice.

ATRA-ATO-NPs enhanced synergistic inhibition of ATRA and ATO on tumor metastasis in vitro
To determine the anti-metastasis ability of ATRA-ATO-NPs, we selected high Pin1 expression and widely used HCC cells HuH7 as a model to investigate the inhibition of nanoparticles on tumor cell motility using a wound healing assay after 72 h treatment. As shown in Fig. 4A and C, HuH7 cells treated with ATRA-ATO-NPs had the lowest wound healing rate among all treatment groups, indicating a significantly stronger capability to inhibit cell migration compared to ATRA, ATO, and their combination. We further used a transwell assay to measure the inhibition of ATRA-ATO-NPs on HuH7 cell migration and invasion. Figure 4B showed that ATRA-ATO-NPs significantly inhibited cell migration and invasion more effectively than the combination of ATRA and ATO. Quantitative analysis revealed that the number of migrated HuH7 cells in the control, ATRA, ATO, ATRA + ATO, and ATRA-ATO-NPs groups were 670 ± 76, 468 ± 83, 506 ± 44, 357 ± 28, and 202 ± 35, respectively (Fig. 4D). The corresponding number of invaded cells were 624 ± 66, 503 ± 64, 320 ± 134, 220 ± 13, and 56 ± 16, respectively (Fig. 4E). Overall, these results confirmed that ATRTA-ATO-NPs could significantly enhance the synergistic inhibition of ATRA and ATO on cancer cell migration and invasion. Once again, our results proved that the controlled release of ATRA and ATO via ATRA-ATO-NPs potentiates their anti-metastasis efficacies, thereby decreasing drug doses to reduce their toxicity. Therefore, ATRA-ATO-NPs offers a promising non-toxic and clinical usable formulation to fight solid tumor growth and metastasis.

ATRA-ATO-NPs enhanced synergistic effect of ATRA and ATO on PIN1 and its signaling pathway proteins and metastasis related proteins in tumor cells
To reveal the mechanism of ATRA-ATO-NPs against tumor growth and metastasis, we investigated the effects of nanoparticles on the expression of Pin1 and its signaling pathway proteins and metastasis related proteins in tumor cells. HuH7 and PLC cells were incubated with the blank NPs for 72 h at concentrations of 0.1, 0.2, 0.4, and 0.6 mg mL− 1. Figures 5A-C showed that the blank NPs did not affect the expression of PIN1 at the treated concentrations even as high as 0.6 mg mL− 1. We used the Western Blot assay to measure the expression of PIN1 and its signaling pathway proteins including β-catenin, NF-κB, Cyclin D1, CDK2, and C-myc in HuH7 and PLC cells treated with ATRA, ATO, ATRA + ATO or ATRA-ATO-NPs for 72 h (Fig. 5D). Quantitative analysis data was shown in Fig. 5E. The results showed that either ATRA or ATO did not significantly inhibit the expression of PIN1, β-Catenin, NF-κB, Cyclin D1, CDK2, and C-myc in HuH7 and PLC cells at the concentration of 10 µM (ATRA) or 1 µM (ATO). The combination of ATRA and ATO increased the inhibitory effects on the expression of PIN1 and signaling pathway proteins. Importantly, ATRA-ATO-NPs significantly down-regulated the expression of PIN1 and signaling pathway proteins compared to ATRA, ATO, and even the combination of ATRA and ATO. Overall, these results confirmed that ATRA-ATO-NPs could significantly enhance the synergistic inhibitory effect of ATRA and ATO on the expression of PIN1 and its signaling pathway proteins in cancer cells, while blank nanoparticles had no inhibitory effect on Pin1 expression.
Figures 5F and G showed the expression levels of PIN1 and metastasis marker proteins including N-cadherin, E-cadherin, MMP2, SNAIL in HuH7 cells after different treatments. We found that when HuH7 cells were treated with 10 µM ATRA or 1 µM ATO, there was no significant downregulation of PIN1, N-cadherin, MMP2, and SNAIL, nor significant upregulation of E-cadherin. The ATRA + ATO increased the down-regulation of PIN1, N-cadherin, MMP2 and SNAIL and the up-regulation of E-cadherin. In comparison, ATRA-ATO-NPs much more significantly increased the down-regulation of PIN1, N-cadherin, MMP2 and SNAIL by ATRA and ATO and the up-regulation of E-cadherin. These results confirmed that ATRA-ATO-NPs could significantly enhance the synergistic regulatory effect of ATRA and ATO on the expression of cancer metastasis marker proteins.
Our previous studies have demonstrated that Pin1 plays a key role in the growth and metastasis of liver cancer cells, and is the main target of Pin1 inhibitor ATRA in inhibiting the growth and metastasis of liver cancer [22, 23]. In this study, we also measured the effect of ATRA-ATO-NPs on PIN1 expression in Pin1-knockdown (shPin1) PLC cells (Fig. S2). In PLC cells expressing empty vector (PLC-shV), where ATRA or ATO treatment resulted in little downregulation of PIN1, and the combination of ATRA and ATO resulted in a slight downregulation of PIN1, whereas ATRA-ATO-NPs greatly downregulated PIN1. Therefore, ATRA-ATO-NPs could greatly enhance the synergistic inhibition of ATRA and ATO on PIN1 expression in cancer cells, consistent with the above experimental results. In contrast, ATRA, ATO, ATRA + ATO and ATRA-ATO-NPs treatments did not cause significant changes in PIN1 expression in PLC-shPin1 cells. These results are consistent with our previous studies [22, 23].
Given Pin1 activates more than 50 oncogenes [1, 24], inhibition of Pin1 may lead to the inhibition of its multiple oncogenes. Indeed, in our results, Pin1-activated oncogenes such as β-Catenin, NF-κB, Cyclin D1, and C-myc were significantly downregulated in ATRA-ATO-NPs treated cancer cells. β-catenin is a principal component in the WNT pathway [25], and C-myc and Cyclin D1 are the WNT pathway-related proteins [26]. Overactivation of the WNT/β-catenin signaling pathway is involved in the promotion of cancer growth and dissemination [27]. Constitutive activation of the NF-κB has been shown to induce tumor initiation, progression and distant metastasis [28]. Cyclin D1 and CDK2 play an important role in cell progression [29], and overexpression of Cyclin D1 is contributed to cancer development [30]. Downregulation of β-Catenin, NF-κB, Cyclin D1, CDK2, and C-myc may result in the inhibition of above cancer signaling pathways, which in turn inhibits tumor cell growth. As expected, ATRA-ATO-NPs significantly downregulated the expression of β-Catenin, NF-κB, Cyclin D1, CDK2, and C-myc, resulting in the strong inhibition of cancer cell growth. Our results once again proved our previous findings that inhibiting Pin1 expression can inhibit the multiple cancer pathways [23].
In cancer, epithelial–mesenchymal transition (EMT) is associated with the metastasis progress of solid tumors, such as tumor initiation, tumor invasion, tumor cell migration, intravasation to the blood, and forming distant metastatic tumor [31]. Owing to its vital role in metastasis, EMT has become a promising target for cancer metastatic therapy. The downregulation of E-cadherin and upregulation of N-cadherin indicates cancer EMT, meanwhile, the increased expressions of MMP2 and transcription factor SNAIL imply the initiation and progression of EMT [32]. Oppositely, the decreased expression of N-cadherin, MMP2 and SNAIL, and increased expression of E-cadherin suggest the inhibition on EMT. These EMT markers such as E-cadherin, N-cadherin, MMP2 and SNAIL, are also regulated by Pin1 [1, 22, 33]. As expected, ATRA-ATO-NPs greatly decreased the expression of N-cadherin, MMP2 and SNAIL, and increased E-cadherin expression. Moreover, ATRA-ATO-NPs significantly enhanced the synergistic regulation of EMT marker protein expression by ATRA and ATO, suggesting that ATRA-ATO-NPs could greatly inhibit the progression of cancer EMT. This may be the reason why ATRA-ATO-NPs significantly enhance the synergistic inhibitory effect of ATRA and ATO on cancer metastasis. In short, above results consistently demonstrated that ATRA-ATO-NPs simultaneously block multiple signaling pathways and cancer EMT progression to inhibit cancer metastasis via downregulating PIN1 expression.

In vivo anti-tumor effects of ATRA-ATO-NPs
To confirm the anti-tumor growth efficacy and mechanism of ATRA-ATO-NPs, we further investigated the inhibitions of ATRA-ATO-NPs on xenograft tumors in mice (Fig. 6A). Figure 6B-D showed the tumor size, volume and weight of mice after treated with saline, blank NPs, ATRA, ATO, ATRA + ATO, ATRA-NPs, ATO-NPs or ATRA-ATO-NPs for 3 weeks. Compared with the saline group, the blank NPs did not show inhibitory effect on tumor growth, indicating that the blank NPs themselves had no anti-tumor effect. ATRA and ATO only had a slight inhibitory effect on tumor growth. In contrast, the ATRA + ATO and ATRA-ATO-NPs showed significant inhibitory effects on tumor growth, indicating that the combination of ATRA and ATO had a synergistic anti-tumor effect. Meanwhile, ATRA-NPs and ATO-NPs also showed the significant inhibitory effects on tumor growth, which was comparable to the inhibitory effect of ATRA + ATO. More importantly, the inhibitory effect of ATRA-ATO-NPs on tumor growth was significantly higher than that of ATRA + ATO, ATRA-NPs, and ATO-NPs. These results confirm that ATRA-ATO-NPs can significantly enhance the synergistic inhibitory effect of ATRA and ATO on tumor growth, which is consistent with the in vitro anti-tumor growth results obtained above. Compared with the saline group, blank NPs, ATRA, ATO, ATRA-NPs, ATO-NPs, and ATRA-ATO-NPs did not cause a significant reduction in the body weight of mice (Fig. 6E), indicating good biocompatibility of these tested biomaterials.
To confirm the mechanism of ATRA-ATO-NPs against tumor growth, the expression of PIN1 and its substrate oncogenes in xenograft tumors was further analyzed by Western blot. As shown in Fig. 6F and G, treatment with blank NPs, ATRA, or ATO did not result in significant downregulation of PIN1 and its substrate oncogenes including β-catenin, NF-κB and Cyclin D1 in tumors compared to the saline control. In contrast, the ATRA + ATO and ATRA-ATO-NPs significantly down-regulated PIN1, β-catenin, NF-κB and Cyclin D1, indicating that the combination of ATRA and ATO had a synergistic inhibitory effect on the expression of PIN1 and its substrate oncogenes. Meanwhile, ATRA-NPs could significantly down-regulate the expression of PIN1, NF-κB and Cyclin D1, which was comparable to the inhibitory effect of ATRA + ATO on the expression of these proteins (Fig. S3A, B, D and E). ATO-NPs also significantly down-regulated the expression of PIN1 and NF-κB, which was comparable to the inhibition effect of ATRA-NPs and ATRA + ATO on the expression of these proteins (Fig. S3A-E). More importantly, the inhibitory effect of ATRA-ATO-NPs on PIN1, β-catenin, NF-κB and Cyclin D1 was significantly higher than that of ATRA + ATO, ATRA-NPs and ATO-NPs, indicating that ATRA-ATO-NPs could significantly enhance the synergistic inhibitory effect of ATRA and ATO on the expression of PIN1 and its substrate oncogene in tumors.
We further analyzed the expression of PIN1 and the metastasis related EMT marker proteins including N-cadherin and MMP2 in tumors after different treatments by Western blot. As shown in Fig. S4, treatment with free ATRA or ATO alone induced only a mild or moderate inhibitory effect on the expression of PIN1 and N-cadherin in tumors compared to the saline group. In contrast, the ATRA + ATO, ATRA-NPs, ATO-NPs and ATRA-ATO-NPs treatments significantly down-regulated PIN1 and N-cadherin in tumors (Fig. S4B and C). The inhibitory effects of ATRA-NPs and ATO-NPs on the expression of PIN1 and N-cadherin were comparable to that of the ATRA + ATO. ATRA-ATO-NPs treatment also significantly down-regulated the expression of MMP2 in tumors (Fig. S4D). These results demonstrated that the combination of ATRA and ATO had a synergistic inhibitory effect on PIN1 expression and cancer EMT progression in tumors. Notably, ATRA-ATO-NPs further enhanced this synergy, leading to the most potent suppression of PIN1, N-cadherin, and MMP2 in tumors.
Overall, ATRA-ATO-NPs can significantly enhance the synergistic inhibitory effects of ATRA and ATO, not only on PIN1 expression and cancer signaling pathways but also on cancer EMT progression in tumors. These in vivo findings align with the anti-tumor growth mechanism previously observed in vitro.

In vivo anti-tumor metastasis, pharmacokinetics and biodistribution of ATRA-ATO-NPs
To confirm the anti-metastatic efficacy of ATRA-ATO-NPs, the inhibitory effect of ATRA-ATO-NPs on lung metastasis of cancer cells was further investigated in mice (Fig. 7A). Because the lung metastasis ability of mouse hepatocellular carcinoma cells Hepa1-6 is stronger than that of human hepatocellular carcinoma cells HuH7, Hepa1-6 was selected as a model cell to construct lung metastasis in mice. Figure 7B-D showed photos of the lungs, hematoxylin and eosin (H&E) staining of lung tissue, and the number of metastatic nodules in the lungs of the mice after different treatments. Compared with healthy lungs, many metastatic nodules were observed in the lungs of the saline group, indicating that a mouse model with cancer lung metastases was successfully constructed. Compared with the saline group, ATRA and ATO had little inhibitory effect on lung metastases. In contrast, the ATRA + ATO and ATRA-ATO-NPs had obvious inhibitory effects on the number and size of metastatic nodules in the lungs, indicating that the combination of ATRA and ATO had a synergistic inhibitory effect on tumor metastasis. ATRA-NPs and ATO-NPs also showed significant inhibitory effects on the number and size of metastatic nodules in the lungs, which was comparable to the inhibitory effect of ATRA + ATO. More importantly, ATRA-ATO-NPs can significantly enhance the inhibitory effect of ATRA and ATO on lung metastasis. These results confirm that ATRA-ATO-NPs can significantly enhance the synergistic inhibition of ATRA and ATO on cancer metastasis, which is consistent with the in vitro anti-tumor metastasis results obtained above.
To further understand why ATRA-ATO-NPs are able to enhance the synergistic anti-tumor efficacy of ATRA and ATO, we determined the pharmacokinetics of ATRA-ATO-NPs and compared them to the pharmacokinetics of ATRA and ATO. Compared with the ATRA group, the ATRA-ATO-NPs group showed a significant increase in plasma ATRA concentration (Fig. 7E). Compared with the ATO group, the plasma concentration of ATO in the ATRA-ATO-NPs group could be maintained at a stable high level within 4 h (Fig. S5). Pharmacokinetic assays of ATRA-NPs and ATO-NPs showed that they were able to significantly increase the concentrations of ATRA and ATO in plasma (Fig. 7F and G). Furthermore, pharmacokinetic parameter analysis showed that compared with the ATRA or ATO groups, the ATRA-NPs, ATO-NPs, and ATRA-ATO-NPs groups significantly increased the maximal plasma concentration (Cmax), the area under concentration-time curve (AUC), and the half life time (t1/2), and significantly reduced the clearance rate (CL) (Tables S1 and S2). These results demonstrated that the pharmacokinetic performances of ATRA-NPs, ATO-NPs, and ATRA-ATO-NPs were significantly superior to that of ATRA and ATO. Moreover, compared to the commercial ATRA slow-releasing pellet in our previous study [23], ATRA-NPs and ATRA-ATO-NPs exhibited superior pharmacokinetic properties. Therefore, these results confirm that ATRA-NPs, ATO-NPs and ATRA-ATO-NPs can sustainably release ATRA and ATO in vivo. In addition, we examined the biodistribution of ATRA-ATO-NPs in tumor-bearing mice. After 4 h and 24 h of ATRA-ATO-NPs administration, the content of ATRA in the main organs and tumor tissues of mice except for the liver remained stable, which further confirmed that ATRA-ATO-NPs were able to release the drug slowly in vivo (Fig. 7H). However, compared with 4 h of administration, the content of ATO in the main organs and tumor tissues of mice was significantly reduced after 24 h of ATRA-ATO-NPs administration (Fig. 7I). The biodistribution of ATRA-NPs and ATO-NPs in tumor-bearing mice were also examined. Similarly, after 4 h and 24 h of ATRA-NPs administration, the content of ATRA in the main organs and tumor tissues of mice remained stable, confirming that ATRA-NPs were able to release the drug slowly in vivo (Fig. 7J). However, compared with 4 h of administration, the content of ATO in the main organs and tumor tissues of mice was obviously reduced after 24 h of ATO-NPs administration except for the spleen (Fig. S6). This might be due to the low ATO content in ATRA-ATO-NPs (the ATO content is equivalent to only 6.6% of the ATRA content), which makes ATO prone to loss. Furthermore, we calculated the relative ATRA and ATO distribution in different tissues after 4 h and 24 h of ATRA-ATO-NPs administration (Fig. S7). After 4 h and 24 h of ATRA-NPs, ATO-NPs and ATRA-ATO-NPs administration, approximately 15% of ATRA and ATO accumulated in tumor tissues, and their levels remained stable. These results confirmed that the nanoparticles can successfully co-deliver ATRA and ATO to the tumor tissue and maintain stable drug levels. Overall, ATRA-NPs, ATO-NPs and ATRA-ATO-NPs can slowly release ATRA and ATO, and can be internalized by cancer cells, which explains the excellent anti-tumor efficacy of ATRA-ATO-NPs.
To verify the mechanism of ATRA-ATO-NPs against tumor metastasis, the expressions of PIN1 and EMT marker proteins including E-cadherin, N-cadherin, and MMP2 in lung tissues of mice after different treatments were analyzed by Western blot. As shown in Fig. 7K and L, PIN1, N-cadherin, and MMP2 were significantly up-regulated in the lungs of the saline group compared to healthy lungs, while E-cadherin was significantly down-regulated. This indicated that the cancer EMT program in the lungs of the saline group was activated. ATRA or ATO treatment had only a mild or moderate effect on the expression of PIN1, E-cadherin, N-cadherin, and MMP2 in lung tissue compared to the saline group. In contrast, the ATRA + ATO and ATRA-ATO-NPs treatments significantly down-regulated PIN1, N-cadherin and MMP2 in lung tissue, while E-cadherin was significantly up-regulated. Additionally, ATRA-NPs and ATO-NPs significantly down-regulate PIN1, N-cadherin, and MMP2, while E-cadherin was significantly up-regulated (Fig. S8). The effects of ATRA-NPs and ATO-NPs on the expression of these proteins were comparable to those of ATRA + ATO. These results demonstrated that the combination of ATRA and ATO had a synergistic inhibitory effect on PIN1 expression and cancer EMT progression in lung metastases. More excitingly, ATRA-ATO-NPs could further enhance the synergistic inhibitory effect of ATRA and ATO on PIN1, N-cadherin and MMP2. Overall, ATRA-ATO-NPs were able to significantly enhance the inhibitory effects of ATRA and ATO on Pin1 expression and cancer EMT progression in lung metastasis, consistent with the in vitro anti-tumor metastasis mechanism obtained above.

In vivo biosafety evaluation of ATRA-ATO-NPs
To evaluate in vivo biosafety of ATRA-ATO-NPs, we tested the biochemistry indexes in blood sample collected from mice after 48 h treatment of ATRA-ATO-NPs (150 and 600 mg kg− 1). Fig. S9 showed the blood biochemical indexes measured in blood samples of mice. We found the levels of total protein (TP), serum albumin (ALB), hemoglobin concentration (HGB), alkaline phosphatase (ALP), alanine aminotransferase (ALT), aspartate aminotransferase (AST), lactate dehydrogenase (LDH), total bilirubin (TBIL), creatinine (CREA), and blood urea nitrogen (BUN) in treatment groups were not significantly different from those of the control group. The counts of red blood cell (RBC), platelet (PLT), white blood cell (WBC) and the percentage of neutrophil (NEUT%), lymphocyte (LYMPH%), monocyte (MONO%), eosinophilic granulocyte (EO%) and basophilic granulocyte (BASO%) in the white blood cells were shown in Fig. S9B. The RBC counts, PLT counts, WBC counts, NEUT%, LYMPH%, MONO%, EO% and BASO% did not show significance between groups of the control and the ATRA-ATO-NPs treatment. These results confirmed that ATRA-ATO-NPs had no significant effect on the blood biochemistry indexes in mice. Given that TP, ALB, HGB, ALP, ALT, AST, LDH, TBIL, CREA, and BUN were functional indexes of the heart, livers or kidneys, the above results proved that ATRA-ATO-NPs did not affect the functionalities of mouse hearts, livers and kidneys, indicating ATRA-ATO-NPs’ in vivo biosafety. The H&E staining images of heart, liver, spleen, lung, kidney and stomach tissues of mice gave additional evidences to demonstrate the excellent in vivo biosafety of ATRA-ATO-NPs: no obvious histological abnormalities or damages were detected in the heart, liver, spleen, lung, kidney and stomach of mice after ATRA-ATO-NPs treatment even at high concentrations of 150 and 600 mg kg− 1 (Fig. S9C). Overall, ATRA-ATO-NPs exhibited the excellent biosafety in vivo.
Once ATRA-ATO-NPs is used as the clinical drug, a vital question will be the toxicity to human organism [34]. Considering that the cytotoxicity of ATRA-ATO-NPs depends on the nanoparticle type, properties, size, surface and concentration [20], we assessed the toxicity of ATRA-ATO-NPs (Fig. S9). HSA, a protein from plasma, served as a nano carrier for ATRA and ATO due to its non-toxic, non-immunogenic, and long-circulating properties [35]. As respected, hemocompatibility evaluation showed that ATRA-ATO-NPs had no significant side effects on the blood biochemistry of tested mice and exhibited good hemocompatibility. Moreover, H&E staining images of mouse organs confirmed the excellent in vivo biosafety of ATRA-ATO-NPs.

Proteomic and bioinformatic analysis of the synergistic anti-tumor effect of two Pin1 inhibitors
Although both ATRA and ATO are inhibitors of Pin1, the molecular mechanism of their synergistic anti-tumor effect is still unclear. In order to systematically understand the mechanism by which these two Pin1 inhibitors synergistically inhibit tumor growth and metastasis, we performed proteomic and bioinformatic analyses on ATRA and ATO-treated hepatocellular carcinoma cells (Fig. 8A). Due to the shortcomings of free ATRA and ATO and weak anti-tumor efficacy, we chose ATRA and ATO-loaded nanoparticles for experiments. Meanwhile, cancer cells treated with blank nanoparticles were used as the control group. Figure 8B showed that ATRA-NPs, ATO-NPs, and ATRA-ATO-NPs treatments significantly down-regulated Pin1 expression in cancer cells. In comparison, ATRA-ATO-NPs had the strongest inhibitory effect on Pin1. Proteomic analysis showed that 8441, 8298 and 7254 proteins were identified in the ATRA-NPs, ATO-NPs and ATRA-ATO-NPs groups compared with the control group (Fig. 8C). Among them, the differentially expressed proteins (abundance change > 1.5-fold and p value < 0.05) in the ATRA-NPs, ATO-NPs and ATRA-ATO-NPs groups were 98, 110 and 2239, respectively. The number and expression patterns of differentially expressed proteins caused by ATRA-NPs and ATO-NPs treatments were similar (Figs. 8C-E). In comparison, the number of differentially expressed proteins caused by ATRA-ATO-NPs was much higher than that of ATRA-NPs and ATO-NPs (Fig. 8D). Specifically, ATRA-ATO-NPs treatment resulted in 1431 down-regulated proteins and 808 up-regulated proteins. Subcellular localization analysis showed that the differentially expressed proteins in the ATRA-NPs, ATO-NPs, and ATRA-ATO-NPs groups were mainly located in the nucleus and cytoplasm (Figs. 8F-H). In comparison, the number of differentially expressed proteins in the nucleus and cytoplasm of the ATRA-ATO-NPs group was much higher than that of the ATRA-NPs and ATO-NPs groups. KEGG pathway analysis showed that the nanodrug treatment, especially ATRA-ATO-NPs, regulated a large number of biological pathways. Figure 8I showed the regulation of 24 biological pathways related to cancer, cell signaling, cellular energy metabolism, cell growth and metastasis by nanodrug treatment. It was evident that the number of biological pathways regulated by ATRA-ATO-NPs treatment, as well as the number of differentially expressed proteins in these pathways, were greater than those of ATRA-NPs and ATO-NPs treatments. In order to understand whether these biological pathways were activated or inhibited, we further performed GSEA analysis on the KEGG pathway in which differentially expressed proteins participated. Table 1 compared the biological pathways obtained by GSEA KEGG analysis in each group. It was found that ATRA-NPs, ATO-NPs, and ATRA-ATO-NPs all exerted inhibitory effects on cancer pathways, but their effects on cell signaling, cellular energy metabolism, cell growth and metastasis-related pathways were different. ATRA-NPs activated calcium signaling, phosphatidylinositol signaling, mTOR signaling and cell cycle, and inhibited mitogen-activated protein kinase (MAPK), p53, Wnt, TGF-beta, JAK-STAT, and focal adhesion signaling. ATO-NPs inhibited calcium signaling, phosphatidylinositol signaling, Notch signaling, and cell cycle. In contrast, ATRA-ATO-NPs inhibited all of these pathways related to cell signaling, cellular energy metabolism, cell growth and metastasis. It was known that Pin1 could promote cancer cell invasion and metastasis by activating p53, STAT3, NOTCH and WNT/β-catenin, promote cancer cell proliferation and resist cell death by activating p53 and WNT/β-catenin, and promote cellular energy metabolism by inhibiting HIF-1α and Myc [24]. The MAPK signaling pathway involved in various cellular functions, including cell proliferation, survival and migration [36, 37]. Focal adhesion plays essential roles in important biological processes including cell motility, cell proliferation, and cell survival [38, 39]. Overall, these results demonstrated that the two inhibitors of Pin1, ATRA and ATO, could synergistically inhibit tumor growth and metastasis by complementary inhibition of downstream signaling pathways of Pin1, including cancer, cell signaling, cellular energy metabolism, and cell growth and metastasis (Fig. 9).

Conclusion
The key finding of this study is that a “multi-target approach” to designing treatments that act, spatiotemporally and simultaneously, on multiple intracellular components and signaling pathways of tumor can improve the potency and safety of oncotherapy. We tested this hypothesis by conjugating ATRA and ATO to uniform HSA nanoparticles, ATRA-ATO-NPs, to simultaneously inhibit Pin1 from multiple angles, aiming for high synergistic inhibition of Pin1 and its downstream pathway biomarkers at low drug doses. As expected, the controlled release of ATRA and ATO via ATRA-ATO-NPs significantly enhanced their anti-tumor growth and metastasis efficacy in vitro and in vivo, allowing for reduced drug doses and decreased toxicity. ATRA-ATO-NPs deliver synergistic effects from both ATRA and ATO by interacting with multiple targets of Pin1 and related biomarkers (β-catenin, NF-κB, Cyclin D1, CDK2, and C-myc) as well as metastatic markers (N-cadherin, E-cadherin, MMP2, SNAIL), outperforming individual or combined ATRA and ATO treatments. Proteomic and bioinformatic analyses further systematically revealed that ATRA-ATO-NPs synergistically inhibited tumor growth and metastasis by complementary inhibition of the Pin1 downstream signaling pathways. Blood chemistry tests and tissue staining demonstrated the excellent safety profile of ATRA-ATO-NPs, supporting their potential for future clinical trials. Thus, our approach offers a promising, non-toxic, and clinically viable formulation to combat solid tumor growth and metastasis.

Materials and methods

Materials
HSA was purchased from Solarbio Biotech Co., Ltd (Beijing, China). ATRA, ATO, sodium deoxycholate, GSH and FITC were purchased from Sigma-Aldrich (Missouri, USA). Fetal bovine serum (FBS) was purchased from PAN-Biotech (Aidenbach, Germany). DMEM was obtained from ThermoFisher. 1% penicillin–streptomycin solution was obtained from HyClone. Primary antibodies, including PIN1, NF-κB, Cyclin D1, MMP2, E-cadherin were purchased from ImmunoWay Biotechnology Company (Texas, USA). Other primary antibodies, including β-catenin, CDK2, C-myc, N-cadherin, SNAIL, and β-actin were purchased from Cell Signaling Technology (Massachusetts, USA). 3-(4,5-dimethylthiazol-2-yl)−2,5-diphenyltetra-zolium bromide (MTT), 4, 6-diamidino-2-phenylindole (DAPI) kit, cell lysis buffer for IP, protease inhibitor cocktail, H&E kit, and BCA protein assay kit were purchased from Beyotime Biotechnology (Shanghai, China). Hepatocellular carcinoma cells and macrophage were obtained from the Cell Bank (Shanghai, China). Ultra-purified water was provided using a Milli-Q Synthesis System (Millipore, Merck KGaA, Darmstadt, Germany). Other reagents with analytical or chemical purity standards were provided by commercial suppliers.

Synthesis and characterizations of nanoparticles
ATRA-ATO-NPs were synthesized by albumin self-assembly. Scheme 1A illustrates the synthesis procedure for ATRA-ATO-NPs. Briefly, first, HSA was dissolved in water and then reduced with GSH to open the disulfide bonds within the albumin molecule and expose the free thiol groups. Secondly, ATO was dissolved with sodium hydroxide (NaOH) and then added to the reduced HSA solution. Thirdly, ATRA was dissolved in ethanol and then added to the mixture of ATO and HSA by a microsyringe pump (Pump 11 Elite, Harvard Apparatus, Massachusetts, USA). ATRA-ATO-NPs were generated by albumin self-assembly under desolvation conditions. Finally, dried ATRA-ATO-NPs were obtained by washing, centrifugation, and freeze-drying. According to the results of our pilot experiments (data not shown), the HSA concentration, the ratio of GSH to HSA, the temperature of the reduction reaction, the ratio of ATRA to HSA, the ratio of ATO and ATRA, the injection speed of ATRA solution, and the volume ratio of ethanol and water were set to 4%, 38.5%, 40 ℃, 3%, 12.7%, 1.5 mL min− 1, and 5:1, respectively. The absorption values of particles and ATRA in the wavelength ranging from 300 nm to 600 nm were determined by a microplate reader. The surface morphology of particles was observed under a scanning electron microscope (SEM) (Nova Nano SEM 230, FEI CZECH REPUBLIC S.R.O., CZE). The particle size and zeta potential of particles were determined by using Zeta potential and laser particle size analyzer (NanoBrook 90Plus Zeta, Brookhaven Instruments Co., Ltd., USA). Synthesis and characterizations of ATRA-NPs and ATO-NPs were similar to the above procedure of ATRA-ATO-NPs, and physicochemistry and biological characterizations of these nanoparticles followed our previous reports [40].

DL, EE, and in vitro release profiles of nanoparticles
For DL, EE, the molar content of ATRA and ATO, and release of ATRA, the procedures were similar to what we reported [22, 23]. Briefly, we prepared a standard calibration curve of ATRA solution, which was used to determine the amount of ATRA in ATRA-ATO-NPs nanoparticles by absorbance recorded using a microplate reader at 360 nm, and then utilized our previous equations to calculate DL, EE, and the molar ratio of ATRA to ATO. The release of ATRA from ATRA-ATO-NPs was analyzed in three different buffers including PBS, DMEM, and DMEM supplemented 5% serum. To determine DL, EE and release of ATO, we modified the reported procedure to quantify the concentration of ATO loaded in ATRA-ATO-NPs by measurement of arsenic using Inductively Coupled Plasma Mass Spectrometry (ICP-MS) (NexION 350X, PerkinElmer, USA) [11]. The standard calibration curve of arsenic solution was prepared by dilution of 1 µg mL− 1 arsenic mother liquor using water. 5 mg ATRA-ATO-NPs was dissolved in 1 mL water followed by digestion with 2 mL 1 M NaOH until powder dissolved, and fixed the final volume to 5 mL. The digested solution was adjusted pH to 6.8 and filtered with 0.22 μm aqueous filtration membrane before ICP-MS analysis. We also measured the release curve of ATO from ATRA-ATO-NPs in PBS, DMEM, and DMEM supplemented 5% serum.

In vivo pharmacokinetic profiles and biodistribution of nanoparticles
In vivo pharmacokinetic profiles of ATRA-NPs, ATO-NPs and ATRA-ATO-NPs were conducted as we reported previously [22, 23]. Mice were randomly divided into multiple groups and intraperitoneally injected with ATRA, ATO, ATRA-NPs, ATO-NPs and or ATRA-ATO-NPs, where the dose of ATRA was 15 mg kg− 1 and the dose of ATO was approximately 1 mg kg− 1. At each experimental time point, mouse blood was collected and plasma was obtained by centrifugation. The concentration of ATRA in plasma was quantified by liquid chromatography–tandem mass spectrometry [22, 23]. The determination of ATO in plasma was briefly as follows: 100 µL of plasma was mixed with 125 µL of nitric acid, 125 µL of hydrogen peroxide and 100 µL of deionized water under vortex conditions, and then digested in a water bath at 65 °C and 95 °C for 1 h, respectively. The digestion product was diluted with deionized water and filtered with a 0.22 μm filter membrane, and then the arsenic content of the sample was determined by ICP-MS. Then, the pharmacokinetic parameters of each group were analyzed using the pharmacokinetic software DAS 3.2.3. Biodistribution assays of ATRA-NPs, ATO-NPs and ATRA-ATO-NPs were performed according to our previously reported procedures with minor modifications [22]. 4 × 106 HuH7 cells in 200 µL of DMEM were subcutaneously injected into the hind limb of BALB/c nu/nu mice. When tumors grew to about 50 mm3, mice were used for biodistribution assays. After intraperitoneal injection of ATRA-NPs, ATO-NPs and ATRA-ATO-NPs for 4 h and 24 h, mice were sacrificed and organs were collected. The amount of ATRA in the tissues was determined as previously reported [22]. The ATO content in the tissues was determined according to the method above. Briefly, tissues were digested with nitric acid and hydrogen peroxide, and then the arsenic content in the tissues was determined by ICP-MS. We further calculated the relative ATRA or ATO distribution in different tissues by the following equation:
Relative ATRA or ATO distribution (%) = (ATRA or ATO contentsingle tissue/Total ATRA or ATO contentall tested tissues) ×100.
In the above formula, the ATRA or ATO content refers to the amount of ATRA or ATO per 100 mg of tissue.

Cell culture and establishment of Pin1 knockdown cells
Human hepatocellular carcinoma cells HuH7 and PLC, normal human liver cells L-02, human embryonic kidney cells 293 T, mouse hepatocellular carcinoma cells Hepa1-6, and mouse macrophage RAW264.7 cells were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). The culture method of these cells was similar to what we described before [22, 23]. These cells were cultured with DMEM at 37 °C in a humidified atmosphere containing 5% CO2. Culture mediums contained 10% FBS and 1% penicillin-streptomycin. Pin1 knockdown cells were obtained by following our previous method [22, 23]. Briefly, HuH7 and PLC cells were transfected with lentiviruses carrying PIN1 shRNA (shPin1) or scrambled shRNA (shV) that were packaged in 293 T cells, and then selected with 1 µg mL− 1 puromycin and validated by Western blot.

Cellular uptake and intracellular ATRA and ATO content
The method of cell uptake assay was similar to our previous report [22, 23]. In brief, HuH7 cells or RAW264.7 cells were incubated with FITC-NPs (0.13 mg mL− 1) for 4, 12, 24 and 48 h, then the cells were fixed with paraformaldehyde, stained with DAPI, and finally observed under a laser scanning microscope (Zeiss, Germany). The determination of intracellular ATRA and ATO content was as follows. HuH7 cells were treated with ATRA (10 µM), ATO (1 µM), or ATRA-ATO-NPs (the concentration of ATRA-ATO-NPs corresponds to containing 10 µM ATRA and 1 µM ATO) for 4, 24, and 48 h. At each time point, the cells were washed three times with PBS, then collected and counted. For ATRA measurement, 700 µL of ultra-purified water was added to the cell pellet, followed by sonication. Then, four volumes of methanol were added to the cell lysates, vortexed for 5 min, and centrifuged at 10,000 rpm for 10 min at 4 ℃. The ATRA content in methanol was measured using the aforementioned method. The intracellular ATO content was measured as previously described. Briefly, the cell pellets were digested with nitric acid and hydrogen peroxide, and then the ATO content was determined by ICP-MS.

Cell viability assay and evaluation of synergistic effects of ATRA and ATO
According to our reported procedure [22, 23], MTT was used to examine the viability of HuH7, PLC, L-02, and Hepa1-6 cells in 96-well plates after treatment with blank NPs, ATRA, ATO, ATRA + ATO, ATRA-NPs, ATO-NPs or ATRA-ATO-NPs for 24, 48–72 h. The method for evaluating synergistic effects of ATRA and ATO was described as before [8]. Briefly, data were analyzed by using the CalcuSyn software (Biosoft, Cambridge, UK), and the combination index (CI) is calculated. When the CI value is less than 0.1, it represents very strong synergism. When the CI value is in the range of 0.1–0.3, 0.3–0.7, 0.7–0.85, 0.85–0.9, or 0.9–1.1, it represents strong synergism, synergism, moderate synergism, slight synergism, and nearly additive effect, respectively.

Wound healing
The procedure was similar to what we reported [22]. 5 × 105 HuH7 cells were cultured in six-well plates for 24 h. After that, cells were incubated with serum-free medium for 12 h. Next, we used a sterile pipette tip to scratch a linear wound through cell layer, and the cells were continuously incubated with medium containing ATRA (10 µM), ATO (1 µM), ATRA + ATO (10 + 1 µM) or ATRA-ATO-NPs (the concentration of ATRA-ATO-NPs corresponds to containing 10 µM ATRA and 1 µM ATO) for 72 h. Finally, wound healing images were acquired and analyzed.

Transwell migration and invasion assay
Assays of transwell migration and invasion were carried out according to our previous report [22]. In brief, the upper chamber was added with 5 × 104 HuH7 cells in 200 µL serum-free medium containing ATRA (10 µM), ATO (1 µM), ATRA + ATO (10 + 1 µM) or ATRA-ATO-NPs (the concentration of ATRA-ATO-NPs corresponds to containing 10 µM ATRA and 1 µM ATO), whereas the lower chamber was added with 500 µL medium containing 10% FBS. After 72 h of incubation, the cells migrating to the lower chamber were fixed, stained, photographed, and counted. The procedure of invasion assay was same to that of migration beside that number of cells in the upper chamber was 3 × 105, and the invaded cells were measured.

Western blot
Western blot was performed as we reported [22, 23]. Briefly, the total proteins from cells, xenograft tumors and tissues were extracted by RIPA buffer and quantified by BCA protein assay kit. The total proteins were separated using polyacrylamide gel, and then transferred onto polyvinylidene difluoride membranes and blocked with 5% nonfat milk. Next, the membranes were incubated with the primary antibody overnight, and then incubated with the secondary antibody for 1 ~ 2 h. Finally, the targeted protein bands were visualized using the chemiluminescence system. β-Actin was the control protein. Note: in in vitro assays, the concentrations of ATRA, ATO, and ATRA + ATO were 10, 1 and 10 + 1 µM, respectively, meanwhile, the concentration of ATRA-NPs, ATO-NPs and ATRA-ATO-NPs corresponds to containing 10 µM ATRA and 1 µM ATO.

In vivo tumor growth and metastasis studies
The protocols of animal experiments were approved by the Experimental Animal Ethics Committee of Fujian Medical University (Issue No. IACUC FJMU 2023 − 0211). Male BALB/c nu/nu mice were purchased from Shanghai Laboratory Animal Center (Shanghai, China). In vivo tumor growth and metastasis studies were conducted based on our previous reports with minor modifications [22, 23]. For anti-tumor growth experiments, 3 × 106 HuH7 cells in 200 µL of DMEM were subcutaneously injected into the hind limb of BALB/c nu/nu mice. When tumors grew to about 50 mm3, mice were randomly divided into six groups, and then intraperitoneally injected with saline, blank NPs, ATRA, ATO, ATRA + ATO, ATRA-NPs, ATO-NPs or ATRA-ATO-NPs twice a week for three weeks, with a dose of 15 mg kg− 1 for ATRA and a dose of about 1 mg kg− 1 for ATO. Tumor volumes and mouse body weights were measured weekly. At the end of drug treatment, mice were sacrificed, and tumors were then collected, photographed and weighed. In anti-metastasis study, normal mice without cell injection were chose as the control, meanwhile, experimental mice were intravenously injected by 200 µL saline containing 2.5 × 106 Hepa1-6 cells, and then randomly separated into five groups of the saline, ATRA, ATO, ATRA + ATO, ATRA-NPs, ATO-NPs or ATRA-ATO-NPs. The seven groups were then intraperitoneally injected with saline, ATRA, ATO, ATRA + ATO, ATRA-NPs, ATO-NPs or ATRA-ATO-NPs twice a week for 3 weeks, with the ATRA dose of 15 mg kg− 1 and the ATO dose of about 1 mg kg− 1. One week after the end of drug administration, mice were sacrificed, lungs were collected and the number of metastatic nodules on the lung surface was counted.

Histology
The experiments were performed as we reported [22, 23]. Briefly, the harvested tissues embedded in paraffin were sliced, fixed, washed and stained with H&E and photographed by a microscope (Zeiss, Germany) as we described previously [41].

Biosafety evaluation
Mice were randomized into three groups (n = 6 per group) and treated with the PBS control, ATRA-ATO-NPs (150 mg kg− 1), and ATRA-ATO-NPs (600 mg kg− 1) by tail vein for 48 h. Then, the mice were sacrificed, and the major organs were collected and analyzed by H&E staining. The plasmas were separated from mice blood samples by centrifugation. The level of blood biochemical indices including TP, ALB, HGB, ALP, ALT, AST, LDH, TBIL, CREA, and BUN, RBC, PLT, WBC and the NEUT%, LYMPH%, MONO%, EO% and BASO% in the white blood cells were measured using Roche Cobas 8000 modular analyzer series.

Proteomic and bioinformatic analyses
Pharmacoproteomic effects of various NPs on cell lines were analyzed as we described previously [42]. HuH7 cells were seeded into cell culture dishes at a density of 1.25 × 104 cells/cm2. After overnight culture, the HuH7 cells were treated with ATRA-NPs, ATO-NPs, or ATRA-ATO-NPs for 72 h. At the same time, HuH7 cells treated with blank NPs were used as the control group. The concentrations of ATRA and ATO in ATRA-NPs, ATO-NPs and ATRA-ATO-NPs were 20 µM and 2 µM, respectively. For each group, the biological experiments were carried out three times. The total protein in HuH7 cells was extracted using lysis buffer (5% sodium deoxycholate, 100 mM Tris-HCl, pH 8.5) and digested with trypsin. The digested peptides from each sample were analyzed by OrbitrapTM AstralTM mass spectrometer (Thermo Scientific) connected to an Vanquish Neo system liquid chromatography (Thermo Scientific) in the data-independent acquisition (DIA) mode. DIA data was analyzed with software DIA-NN 1.8.1. All reported data were based on 99% confidence for protein identification as determined by false discovery rate (FDR) ≤ 1%. Compared to the control group, the change in protein abundance in each group was greater than 1.5-fold and the statistical difference p value was less than 0.05 was defined as significantly differentially expressed proteins. Hierarchical clustering analysis of proteins were performed by Cluster 3.0 (http://bonsai.hgc.jp/mdehoon/software/cluster/software.htm) and Java Treeview software (http://jtreeview.sourceforge.net). CELLO (http://cello.life.nctu.edu.tw/) was used to predict protein subcellular localization. The online Kyoto Encyclopedia of Genes and Genomes (KEGG) database (https://www.genome.jp/kegg/) was used to analyze the biological pathways regulated by differentially expressed proteins. The Gene Set Enrichment Analysis (GSEA) of KEGG pathways regulated by differentially expressed proteins were performed by cluster profile package in R package (version 4.4.4) and p value cutoff = 0.9. The above-mentioned cellular protein extraction and identification were performed at Shanghai Applied Protein Technology Co. Ltd. (Shanghai, China).

Statistical analysis
The results were presented as means ± standard deviation. Student’s t-test or two-way analysis of variance (ANOVA) was conducted for statistical analysis. The mean differences were considered significant for *P < 0.05, and highly significant for **P < 0.01, ***P < 0.001, or ****P < 0.0001.

Supplementary Information