Graphical abstract graphic file with name fx1.jpg [41]Open in a new tab Highlights * • Developed sorafenib-resistant HCC PDX models preserving primary tumor features * • Phosphoproteomics found active cell cycle pathway in sorafenib-resistant HCC PDX * • KSEA analysis revealed active CDK1 and PRKDC in sorafenib-resistant HCC PDX * • Sorafenib+kinase inhibitor combo showed synergistic anti-resistance effects __________________________________________________________________ Integrative aspects of cell biology; Cancer; Proteomics Introduction Hepatocellular carcinoma (HCC), the most common type of liver cancer, is projected to reach 1 million new cases by 2025.[42]^1 Despite the increasing utilization of surgical and locoregional therapies globally estimates indicate that approximately 50%–60% of HCC patients ultimately require systemic treatments.[43]^2 Targeted therapies have created an encouraging trend in the management of advanced HCC.[44]^3 Sorafenib, a molecular targeted therapeutic agent, was the first drug approved for unresectable HCC by the US Food and Drug Administration (FDA). However, its clinical efficacy was unsatisfactory, with a short improvement in survival time (about 3–6 months), and acquired drug resistance typically develops within six months of treatment initiation.[45]^4^,[46]^5^,[47]^6 Understanding the mechanisms of drug resistance is crucial for developing effective strategies to improve therapeutic outcomes in HCC. Ongoing trials are investigating combination therapies, including immune checkpoint inhibitors,[48]^5 tyrosine kinase inhibitors,[49]^7 and even combinations of two immunotherapy regimens.[50]^2 Phosphorylation, an essential post-translational modification found in at least 75% of the proteome,[51]^8 is regulated by protein kinases. This modification impacts various cellular processes, including cell cycle progression and signaling.[52]^9^,[53]^10^,[54]^11^,[55]^12 Dysregulation of signaling networks often leads to diseases, rendering protein kinases prime drug targets.[56]^13 There are over 73 small molecule inhibitors approved by the FDA, and 175 compounds are undergoing clinical trials.[57]^14 These targeted inhibitors typically reduce the activity of known oncogenic signaling pathways, but comprehending their therapeutic mechanisms extends beyond genetic and protein levels.[58]^14 The advancement of liquid chromatography coupled with tandem mass spectrometry (LC-MS/MS) has facilitated the identification and quantification of sample-specific proteins in a high-throughput manner. Recently, quantitative proteomic and phosphoproteomic analyses have helped identify intrinsic cancer signaling networks and their activated kinases in a variety of cancers, such as HCC,[59]^15 glioblastoma,[60]^16 T cell acute lymphoblastic leukemia,[61]^17 and urothelial carcinoma.[62]^18 To date, only one study has utilized tumor tissues from sorafenib-treated HCC patients.[63]^19 This study employed proteomics and phosphoproteomics analyses on biopsied tissues from HCC patients before and during sorafenib treatment, revealing a wide range of signaling pathways involved in tumor progression and drug resistance. However, the limited availability of non-renewable biopsy material poses a challenge to obtaining sustainable large-scale proteomic data. Given the scarcity of clinically resistant patient samples, patient-derived xenograft (PDX) models have played an increasingly pivotal role in clinical research.[64]^20 In the PDX model, the human stroma is replaced by host murine components, but human tumor proteins can be predominantly distinguished from murine stromal proteins through the identification of species-specific peptide sequences.[65]^21 These models retain the key characteristics of the primary tumor, including histology, genome, transcriptome, proteome, and phosphoproteome,[66]^20^,[67]^22 making them a reliable option for preclinical studies. Moreover, phosphoproteomics analyses of PDX models have revealed potential tumor suppressors, mechanisms of drug resistance, novel targets, and successful multidrug combination regimens in various diseases.[68]^16^,[69]^21^,[70]^23 Our previous study has successfully established the sorafenib-resistant PDX models.[71]^22 Here, we further utilized a multiplex iTRAQ MS-based proteomics approach to comprehensively quantify the proteome and phosphoproteome of four paired sorafenib-resistant and sorafenib-sensitive PDX models. KEGG enrichment analysis and kinase substrate enrichment analysis (KSEA) identified CDK1 and PRKDC kinases in the cell cycle pathway as potential targets for overcoming sorafenib resistance in HCC. We then evaluated the therapeutic efficacy of combining sorafenib with Ro-3306 (a CDK1 inhibitor) or KU-57788 (a PRKDC inhibitor) against sorafenib-resistant HCC using both cell lines and PDX models. Both combination therapies were found to be synergistic in overcoming drug resistance. This study provides critical insights into the mechanisms of tumor progression and resistance to cancer therapy, emphasizing the valuable role of PDX models in identifying and validating potential biomarkers and signaling pathways for enhancing cancer treatment. Results Workflow of quantitative proteomics and phosphoproteomics analysis The scarcity of tumor specimens from sorafenib-resistant HCC patients poses a significant challenge in studying drug resistance mechanisms. To comprehensively understand the development of acquired sorafenib resistance in patients with advanced HCC, we conducted quantitative proteomic and phosphoproteomic analyses of four pairs of sorafenib-sensitive/resistant PDX tumors. The schematic workflow is presented in [72]Figure 1. Tumor tissue samples were subjected to protein extraction using 8M urea buffer, followed by protein reduction, alkylation, and tandem Lys-C/trypsin digestion to generate peptides. The resulting peptides were labeled with 8-plex iTRAQ reagents and fractionated using basic pH reversed-phase liquid chromatography with a non-continuous pooling strategy.[73]^24 A total of 96 fractions was collected and then combined into 24 fractions, which consisted of early, intermediate, and late eluting fractions.[74]^24^,[75]^25 Of these 24 fractions, 6% were used for global proteomics, whereas the remaining 94% were combined into 13 fractions that were enriched for phosphopeptides using immobilized metal ion affinity chromatography. The obtained peptides were subsequently analyzed using LC-MS/MS with a timsTOF Pro instrument. Three technical replicates were performed for each fraction to ensure analytical accuracy and reliability. Figure 1. [76]Figure 1 [77]Open in a new tab Experimental setup for quantitative proteomics and phosphoproteomics (A) Flowchart of establishing sorafenib-resistant HCC patient-derived xenograft models obtained from consecutive sorafenib treatment. (B) The schematic workflow of proteome and phosphoproteome analysis in four pairwise PDX mouse tumor tissues using multiplex iTRAQ. Sor, sorafenib; Ctrl, phosphate-buffered saline. See also [78]Table S1. Comprehensive proteomic and phosphoproteomic analysis reveals extensive protein and phosphosite diversity in HCC PDX models The comprehensive iTRAQ-based global proteomic and phosphoproteomic analysis identified 17,096 proteins and 67,586 peptides, including 50,593 phosphopeptides. Of these, 9,366 proteins (with an average of 8,200 per sample) and 33,621 phosphopeptides matched the human proteome ([79]Figures 2A, 2B, and [80]S1A). Of 33,621 human phosphopeptides, 26,214 were unique, with 88.1% having a single phosphosite, 9.5% having two, 2.1% having three, and 0.24% having more than three phosphosites ([81]Figure 2B and 2C). These unique phosphopeptides were derived from 4,934 phosphoproteins and contained 20,127 phosphosites (with an average of 14,363 per sample) that had high confidence (Ascore >20) ([82]Figures 2B and [83]S1B). The phosphosites consisted of 16,904 serine residues, 3,089 threonine residues, and 134 tyrosine residues ([84]Figure 2D). The abundance distribution of proteins and phosphosites was consistent across all samples as indicated by the boxplots ([85]Figures S1C and S1D). Figure 2. [86]Figure 2 [87]Open in a new tab Data statistics (A) Proteomic dataset was filtered and the numbers are displayed. (B) The phosphoproteomic dataset was filtered at different levels, and the numbers are displayed. (C) Distribution of phosphosites observed per phosphopeptide in this experiment. (D) The proportion of the phosphorylated serine, threonine, and tyrosine on the identified phosphopepetides. See also [88]Figures S1 and [89]S2. This study used HeLa cell lysates to evaluate the performance of the LC-MS instrument. Spearman’s correlation coefficients ranged from 0.75 to 0.88 for the HeLa cell lysates, with an average correlation coefficient of 0.82 ([90]Figure S2A), indicating good stability of the proteomics platform. The correlation coefficients for proteomic samples ranged from 0.77 to 0.96, with average values of 0.94 for technical replicates and 0.92 for paired samples ([91]Figure S2B). And the correlation coefficients for phosphoproteomic samples ranged from 0.52 to 0.94, with average values of 0.72 for technical replicates and 0.77 for paired samples ([92]Figure S2C). These findings suggested that the phosphoproteome exhibit greater variation than the proteome. Furthermore, paired samples exhibited higher correlation coefficients than unpaired samples, indicating molecular heterogeneity in PDX tumor tissues. Supervised partial least squares discriminant analysis highlighted that proteomic and phosphoproteomic levels could differentiate between paired drug-resistant and control tumor samples ([93]Figure S3), providing a basis for further investigation into the molecular mechanisms of drug resistance in tumors. Pathway enrichment analysis reveals dysregulated cell cycle pathway in sorafenib-resistant tumor This study identified 2,120 phosphorylation sites that were significantly altered, with 944 upregulated and 1,176 downregulated in resistant tissues (Wilcoxon signed-rank test, p < 0.05, fold change 1.25, [94]Figure 3A). To gain deeper biological insights, we utilized Metascape[95]^26 to identify KEGG pathways associated with the differentially expressed phosphosites. The resulting top 20 pathways were visually represented in the heatmap ([96]Figure 3B). Furthermore, utilizing Metascape enrichment network visualization, we illustrated intra- and inter-cluster similarities of enriched terms and predicted interactions among biological pathways ([97]Figure S4A). The results indicated that the drug-resistant group exhibited activation of pathways associated with cell proliferation, including spliceosome and the cell cycle. In contrast, sorafenib-sensitive tumors exhibited enrichment of pathways involved in cell communication, endocrine system function, and energy metabolism, which are essential for normal liver function. These pathways encompassed tight junction, adherens junction, insulin signaling pathway, thyroid hormone signaling pathway, vascular smooth muscle contraction, endocytosis, and glucagon signaling pathway. These results suggest a higher degree of tumor malignancy in the drug-resistant group. Similarly, the proteome exhibited comparable results to the phosphoproteome (Wilcoxon signed-rank test, p < 0.05, fold change 1.25, [98]Figures S5A and S5B). The notable enrichment of cell cycle signaling pathways in both the proteomic and phosphoproteomic data of drug-resistant tumor tissues prompted us to map differentially expressed phosphoproteins to the KEGG cell cycle pathway using Pathview[99]^27 ([100]Figure 3C). The network analysis highlighted several upregulated key node proteins in drug-resistant tumors. This observation indicates that the cell cycle pathway is highly active in drug-resistant tumors, suggesting a strong tumor proliferative capacity. Figure 3. [101]Figure 3 [102]Open in a new tab Bioinformatic analysis of differential expressed phosphoproteins (A) Volcano plot of the −log[10](p value) against the log fold-change (log[2]FC) of the differentially regulated phosphosites for sorafenib-resistant group versus sensitive group. Vertical dashed line indicates |log[2]FC| = 0.322; horizontal dashed line indicates p = 0.05. (B) Heatmap of the top 20 significantly enriched KEGG pathways in the resistance and matched control groups using Metascape. The cells are colored based on their p values, with blank cells indicating non-enrichment in the corresponding gene list. (C) Activation of the cell cycle signaling pathway (KEGG hsa04110) in the sorafenib resistance group, visualized using the Pathview Web tool. The relative expression of each phosphoprotein in the resistance group versus the control is shown by the color scale of red (high) to green (low), whereas white represents no significant change or not detected. See also [103]Figures S3–S5. Identification of dysregulated kinases associated with cell cycle signaling pathway in sorafenib-resistant HCC PDX models Protein kinases have emerged as important drug targets over the past two decades, primarily due to their frequent dysregulation in various diseases.[104]^28 Sorafenib, a potent multikinase inhibitor drug, is closely associated with phosphorylation.[105]^19 To infer changes in kinase activity in drug-resistant tumors, we subjected the differentially expressed phosphosites as substrates to KSEA.[106]^29 The KSEA results indicated that several kinases, including cell cycle-dependent kinases 1/2/4 (CDK1/2/4), ATM, PRKDC, AURKA/B, ATR, NEK2, HIPK2 MAK, PLK1/4, and TTK, showed higher activity in drug-resistant tumors, whereas the kinase activity of PRKG1 and SGK1 was increased in the sorafenib-sensitive tumor group ([107]Figures 4A and 4B). By mapping these kinases to the human kinome tree, we discovered significant dysregulation in the CMGC, AGC, and atypical PIKK protein kinase families ([108]Figure S4B). KEGG enrichment analysis indicated that enriched kinases were mainly associated with the cell cycle signaling pathway ([109]Figure S4C), implying that the onset of sorafenib resistance in HCC could be due to the activation of the cell cycle pathway. These kinases, including ATM, ATR, and PRKDC, play key roles in DNA repair. Cells have evolved an intricate DNA repair mechanism to limit DNA damage progression by triggering cell cycle checkpoints and repairing DNA damage before it interferes with the replication process. In some tumor cells, enhanced DNA repair capacity after replication and genotoxic stress has been observed, leading to accumulated genomic damage and ultimately cell death.[110]^30^,[111]^31 Figure 4. [112]Figure 4 [113]Open in a new tab Kinase substrate enrichment analysis (KSEA) identifies activated kinases and their substrates in sorafenib-resistant HCC (A) The bar plot displays the enrichment Z score of the kinases with significantly up- or downregulated kinase activity in a KSEA comparing resistance with the control group (p < 0.05, number of substrates ≥5). (B) The bubble plot shows representative phosphorylation site substrates of the activated kinases. The substrates are shown in rows, where red and green dots represent up- and downregulation of phosphorylation sites, respectively. The size of the dots is proportional to the |log[2]fold-change| of the phosphorylation site. The transparency of the dots shows an inverse correlation with the p values of the phosphorylation sites. Phosphorylation sites with at least a 2-fold difference between resistance and control tissues are shown. For kinases with <3 substrates with at least a 2-fold difference, the top three substrates with the highest |log[2]FC| are displayed. (C) Venn diagrams showing the overlap between the significantly upregulated phosphoproteins in the cell cycle pathway (cyan) and activated kinases (orange). (D) Representative western blot analysis of p-PRKDC (S2056), total PRKDC, p-CDK1 (T161) and total CDK1 in sorafenib-sensitive tumor (Ctrl) and matched resistant tumor (Res). Densitometric analysis of bands was performed using ImageJ software, with the control group normalized to 1. The numerical values below each lane represent the ratio of the target proteins to the β-actin loading control. See also [114]Figures S4–S6. The activation of signaling pathways is frequently driven by the differential expression of phosphoproteins within those pathways. In the present study, by considering the intersection of highly expressed phosphoproteins and kinases with increased activity in the phosphoproteomic data, we identified two kinases, CDK1 and PRKDC, as potential targets for overcoming the observed resistance phenotype ([115]Figure 4C). Protein immunoblotting and parallel reaction monitoring analysis further confirmed that p-PRKDC and p-CDK1 were highly expressed in drug-resistant tumors ([116]Figures 4D, [117]S6A, and S6B). Notably, even after adjusting for phosphosites based on protein abundance, the enhanced kinase activity of CDK1 and PRKDC could still be inferred ([118]Figures S5C and S5D). Synergistic effect of CDK1 or PRKDC inhibitor in combination with sorafenib overcomes drug resistance in vitro Subsequently, we established a sorafenib-resistant cell line to investigate drug resistance mechanisms and develop effective treatment strategies. Gradual increases in sorafenib concentrations were applied to the culture medium of Huh7 parental cells (Huh7^S) until tolerance reached 10 μM (Huh7^R).[119]^32 Huh7^R cells exhibited a spindle-like morphology, with a loss of cell-to-cell contact ([120]Figure 5A), suggestive of cells undergoing epithelial-mesenchymal transition.[121]^33 The sorafenib resistance of Huh7^S and Huh7^R cells was assessed by subjecting the cells to different drug concentrations simultaneously ([122]Figures 5B and 5C). The results showed that Huh7^R cells maintained relatively unchanged viability at a 5 μM concentration, whereas Huh7^S cell viability decreased with increasing sorafenib concentration. Further analysis showed that the half-maximal inhibitory concentration (IC[50]) of sorafenib for Huh7^S and Huh7^R cells was 5 μM and 12.3 μM, respectively ([123]Figure 5D). The successful development of a sorafenib-resistant cell model enabled us to evaluate the potential of CDK1 inhibitor (Ro-3306) and PRKDC inhibitor (KU-57788) in overcoming sorafenib resistance. Figure 5. [124]Figure 5 [125]Open in a new tab Characterization of Huh7 liver cancer cell line and its relative growth under different small molecule inhibitors (A) Cell morphology of Huh7^S and Huh7^R cells is shown. Scale bars, 200 μm. (B and C) Relative viability of Huh7^S/Huh7^R cell lines treated with varying concentrations of sorafenib. (D) Dose-response curves of Huh7^S/Huh7^R cell lines generated by treating the cells with varying concentrations of sorafenib for 72 h. The corresponding IC50 values represent the half-maximal inhibitory concentration. (E and F) Relative growth of Huh7^S/Huh7^R cell lines after treatment with sorafenib, Ro-3306, KU-57788, or their combination at indicated concentrations over the entire experiment. Relative growth was determined daily and normalized to day 0. Data are presented as mean ± SEM, n = 3. p values were determined by two-way ANOVA with Tukey’s multiple comparisons. ∗∗∗∗p < 0.0001, ∗p < 0.05. See also [126]Figures S7 and [127]S8. The western blot analysis presented in [128]Figure S7 demonstrates that under sorafenib treatment, the phosphorylation of CDK1 at T161 and PRKDC at S2056 was significantly higher in the Huh7^R cells compared with the Huh7^S cells. Additionally, the CDK1 substrate p-Rb (T373) and the PRKDC substrate p-Vimentin (S56) were elevated in the Huh7^R cells. These findings suggest that the kinase activities of CDK1 and PRKDC were enhanced in the Huh7^R cells. The results of cell proliferation experiments showed that Ro-3306 and KU-57788 could independently inhibit the growth of Huh7 cells ([129]Figures 5E and 5F). Remarkably, when sorafenib was combined with KU-57788, it markedly diminished the viability of Huh7^R cell line, displaying greater efficacy than when either drug was used alone ([130]Figure 5F). To further verify the efficacy of the selected kinase inhibitors, we conducted long-term colony formation assays.[131]^34 Treatment with sorafenib alone led to a decrease in the colony-forming ability of both Huh7^S and Huh7^R cell lines, accompanied by a corresponding reduction in clonal area as the drug concentration increased ([132]Figures 6A–6C). However, at a concentration of 5 μM, Huh7^R cells exhibited a stronger cellular colony-forming ability than Huh7^S. Notably, the combination of sorafenib with either Ro-3306 or KU-57788 resulted in a dose-dependent decrease in colony formation for both cell lines, surpassing the efficacy of sorafenib treatment alone ([133]Figures 6D–6K). Additionally, the combination of Ro-3306/KU-57788 and sorafenib exhibited moderate synergism in the Huh7^R cell line, as indicated by the combination index values from the CCK-8 assay data ([134]Figure S8). These findings suggest that the combination of sorafenib and KU-57788 exerts a synergistic effect in the Huh7^R cell line and is capable of overcoming sorafenib resistance at the cellular level. Figure 6. [135]Figure 6 [136]Open in a new tab Synergistic effect of sorafenib, Ro-3306, and KU-57788 on long-term colony formation of Huh7^S/Huh7^R cell lines Long-term colony-formation assay of Huh7^S/Huh7^R cell lines treated with sorafenib alone (A–C), sorafenib + Ro-3306 (D, E, H, I), or sorafenib + Ro-3306 (F, G, J, K) at the indicated concentrations. The cells were fixed and stained after 10–12 days. Representative data from three independent experiments and bar graphs are the quantification of colony area using ImageJ software. Scale bar, 1 cm. Data are mean ± SEM, n = 3 independent experiments. p values were determined by two-sided unpaired Student’s t test. ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Therapeutic targeting of PRKDC delays sorafenib resistance in vivo The efficacy of combination therapy with sorafenib and KU-57788 was evaluated in sorafenib-resistant HCC PDX models. The validation experiment comprised four groups, each consisting of six tumor-bearing nude mice randomly assigned to the following treatments: control (Ctrl), sorafenib, KU-57788, and sorafenib + KU-57788. The nude mice exhibited uniform weight after drug treatment, with no observed weight loss ([137]Figure 7A), indicating good tolerability. Compared with the control and sorafenib monotherapy groups, the cohort receiving the combined sorafenib and KU-57788 treatment demonstrated significant tumor regression ([138]Figures 7B and 7C). Notably, the KU-57788 monotherapy exhibited comparable efficacy to the combination treatment, suggesting a lack of synergistic activity between KU-57788 and sorafenib in vivo. This finding implies the need to stratify patients who might respond better to the combination therapy or optimize the drug administration regimen. Nonetheless, these results indicate the potential of KU-57788 to delay sorafenib resistance in HCC patients. In summary, this investigation represents a significant step toward developing effective therapeutic strategies for managing sorafenib-resistant HCC, while also highlighting the need for continued research in this critical area. Figure 7. [139]Figure 7 [140]Open in a new tab Therapeutic targeting of PRKDC delays sorafenib resistance in vivo (A) Body weights of mice in the PDX models with the aforementioned treatments were assessed. (B) Representative tumor images of each group of xenografts at the end of treatment. (C) Tumor growth curves were obtained for one HCC PDX model in mice treated with vehicle, sorafenib (30 mg/kg), Ro-3306 (4 mg/kg), KU-57,788 (4 mg/kg), or the specified combination. n = 6 mice per group. Data are mean ± SEM. p values were determined by two-way ANOVA with Tukey’s multiple comparisons. ∗p < 0.05, ns, not significant. Discussion HCC is an aggressive and malignant form of adenocarcinoma.[141]^35 However, there is a limited availability of effective therapies for addressing sorafenib resistance in advanced HCC. Our research specifically addressed the crucial issue of discovering potential therapeutic strategies for overcoming acquired resistance to sorafenib in HCC through comprehensive profiling of proteomics and phosphoproteomics of sorafenib-resistant HCC PDX mouse models. To our knowledge, this study represented the first quantitative phosphoproteomic analysis of tumors from sorafenib-resistant PDX models, providing valuable insights into the underlying mechanisms of acquired sorafenib resistance in patients with advanced HCC. The analyses revealed detailed changes in the proteome and phosphoproteome between sorafenib-resistant and sorafenib-sensitive PDX tumors, leading to the identification of a significant number of altered proteins and phosphosites. Further exploration of the enriched pathways uncovered significant activation of the cell cycle signaling pathway in sorafenib-resistant tumors, suggesting its potential role as a compensatory pathway driving tumor progression. In contrast, sorafenib-sensitive tumors exhibited enriched pathways associated with cell communication, endocrine system function, and energy metabolism. These differences may indicate distinct underlying mechanisms between sorafenib-resistant and sorafenib-sensitive tumors and could potentially account for the variation in clinical responses observed in HCC patients. Nowadays, six systemic therapies have gained FDA approval for HCC based on phase III clinical trials, including atezolizumab plus bevacizumab,[142]^36 sorafenib,[143]^6 lenvatinib,[144]^37 regorafenib,[145]^38 cabozantinib,[146]^39 and ramucirumab.[147]^40 Notably, five of these therapies target protein kinases, highlighting the importance of protein kinase-dependent signaling networks in the progression of HCC. Transitioning our focus onto the role of kinases, this research has identified the dysregulation of several key players, including CDK1/2/4, ATM, PRKDC, and AURKA/B, among others. The significant modifications we have observed in phosphorylation sites, along with the variety of kinases identified through KSEA, illustrate the multidimensional nature of sorafenib resistance. The elevated activity of these kinases suggests their potential roles in driving the resistance process. Notably, CDK1 and PRKDC were identified as potential kinases in this study, suggesting that inhibitors targeting these kinases could represent novel therapeutic options for drug-resistant HCC. CDK1 is recognized as an oncogenic protein due to its regulation of the G2/M cell cycle checkpoint, which allows for cell progression through mitosis. CDK1 is commonly overexpressed or exhibits enhanced kinase activity in various cancers, including lung cancer,[148]^41 HCC,[149]^15^,[150]^42 and urothelial carcinoma of the bladder.[151]^18 Targeting CDK1 has shown promise in improving cancer treatment outcomes. For example, studies have demonstrated that inhibition of CDK1 using Ro-3306 increases the efficacy of sorafenib treatment for HCC.[152]^42 The inhibition of CDK1 is under investigation as a potential therapeutic approach for several other cancers, such as pancreatic ductal adenocarcinoma and intrahepatic cholangiocarcinoma.[153]^43^,[154]^44 PRKDC, also known as DNA-dependent protein kinase, is a core kinase that repairs non-homologous end joining DNA damage. Its high expression is often associated with a poor prognosis in various cancers, including HCC[155]^45^,[156]^46 and breast cancer.[157]^47 In many cancers, the upregulation of DNA repair activity leads to drug resistance, suggesting that PRKDC could be a potential target for overcoming resistance.[158]^48^,[159]^49 The discrepancy observed between the in vitro and in vivo findings can be attributed to several key factors. First, the notable differences between the in vitro and in vivo environments, such as the complex tumor microenvironment and systemic factors present in the PDX models, may lead to divergent drug responses compared with the simplified in vitro conditions.[160]^20 Second, the inherent tumor heterogeneity within the PDX models represents a significant challenge.[161]^20 The presence of diverse subpopulations of cancer cells, each potentially relying on different signaling pathways for survival, could hinder the ability of the combination therapy to elicit a synergistic effect. Furthermore, the study may have been limited in its capacity to fully optimize the drug administration regimen to overcome sorafenib resistance.[162]^23 The complex interplay between the tumor, its microenvironment, and the pharmacokinetics/pharmacodynamics of the drugs requires a more in-depth investigation to unravel the specific mechanisms underlying the observed discrepancy. In summary, the results of KEGG enrichment analysis and KSEA revealed the activation of the cell cycle pathway in sorafenib-resistant tumors, along with enhanced activity of the associated kinases CDK1 and PRKDC. Based on these findings, the two kinase inhibitors were administered with sorafenib in sorafenib-resistant cell line and mice, respectively. The drug combinations exhibited synergistic inhibition of proliferation at the cellular level. These findings have the potential to be translated into clinical applications, enabling precision therapy and more effective treatment for patients. Furthermore, this approach can be extended to investigate resistance patterns in other cancer types and their response to different drugs. By doing so, we can advance our knowledge of carcinogenesis and cancer drug resistance on a broader scale, contributing to the development of more effective treatment strategies. Limitations of the study The sample size in this study was limited. To validate the efficacy of the combination therapies, it is crucial to include a larger cohort of sorafenib-resistant HCC PDX models, considering the heterogeneity of patients. Additionally, we point out here that the detailed mechanism of how kinases function in sorafenib-resistant HCC remains to be explored in future studies. Resource availability Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Haojie Lu (luhaojie@fudan.edu.cn). Materials availability This study did not generate new unique reagents. Data and code availability * • The mass spectrometry proteomics data have been deposited at ProteomeXchange Consortium and are publicly available as of the date of publication.[163]^50 Accession numbers are listed in the [164]key resources table. * • This paper does not report original code. * • Any additional information required to reanalyze the data reported in this paper is available from the [165]lead contact upon request. Acknowledgments