COPG1 Is a Selectively Essential Regulator of Cancer Progression and Chemoresistance via Redox Modulation and AKT Signaling

Abstract

The coatomer complex has been implicated in cancer progression; however, a comprehensive pan-cancer analysis is lacking. Therefore, it is essential to identify the critical roles and essentiality of coatomer genes across pan-cancer. We systematically profiled the genetic alterations, expression patterns, prognostic relevance, and functional dependencies of all coatomer subunits across multiple cancers using more than 10,000 tumor samples from The Cancer Genome Atlas, complemented by functional perturbation data from CRISPR (n = 1178) and RNAi (n = 707) screens in DepMap. Functional validation was also performed to identify the essentiality of selectively essential coatomer genes in hepatocellular carcinoma (HCC). Gene amplification, most notably of COPB2, was the most frequent alteration and was associated with poor survival in bladder and esophageal cancers. Mutations in COPA and SEC31A also demonstrated prognostic significance in endometrial carcinoma. Expression analyses revealed broad upregulation of coatomer genes across cancer types, with COPG1 and COPB1 emerging as strong risk-associated genes (HR > 2). Integrative functional dependency analyses identified COPG1 as selectively essential in multiple cancers, and its loss was associated with increased drug sensitivity. Functional validation in hepatocellular carcinoma revealed that COPG1 knockdown impaired malignant phenotypes and reduced tumorigenicity in vivo. Mechanistically, COPG1 depletion induced Golgi disruption and ER stress, increased ROS production, and suppressed PI3K–AKT signaling, thereby sensitizing cells to sorafenib and doxorubicin. Collectively, this pan-cancer analysis reveals the context-dependent roles of coatomer subunits and identifies COPG1 as a novel oncogenic driver and potential therapeutic target in HCC, mediating chemoresistance through redox modulation and PI3K–AKT pathway inhibition.

1. Introduction

The coatomer complex plays a pivotal role in the vesicular transport of proteins and lipids within the endomembrane system [1,2]. It is instrumental in vesicle formation, cargo selection, and the translocation of vesicles across various intracellular compartments, including the endoplasmic reticulum (ER), ER-Golgi intermediate compartment (ERGIC), Golgi apparatus, trans-Golgi network (TGN), early and late endosomes, tubular endosomal network (TEN), endocytic recycling compartment, lysosomes, lysosome-related organelles (LROs), and distinct domains of the plasma membrane. Vesicle formation is initiated by the recruitment of coatomer proteins to the cytosolic side of membranes, a process facilitated by small GTPases such as ARF1 and SAR1. ARF1 primarily regulates COPI-mediated retrograde transport from the Golgi to the ER, whereas SAR1 is essential for COPII-mediated anterograde transport from the ER to the Golgi [3,4,5,6,7]. Upon activation, these GTPases undergo conformational changes that promote membrane association and the subsequent recruitment of coatomer complexes. Coatomer complexes interact with sorting signals located in the cytosolic tails of transmembrane cargo proteins, ensuring their selective incorporation into budding vesicles. Although the coat dissociates following vesicle formation, it continues to influence vesicle targeting and fusion by interacting with specific phosphoinositides, small GTPases, cargo molecules, and other regulatory proteins [8]. Export of oncogenes via the extracellular vesicles promotes cancer progression through activation of multiple oncogenic signaling pathways [9,10].

The origin of the coatomer complex can be traced back to the last eukaryotic common ancestor (LECA) approximately 1 to 1.5 billion years ago [11,12]. Throughout evolution, coatomer complexes have diversified to fulfill specialized roles in intracellular trafficking. The primary classes of coatomer complexes included COPI, COPII, adaptor protein (AP) complexes, and the retromer complex. COPI is primarily involved in retrograde transport from the Golgi apparatus to the ER, whereas COPII mediates anterograde transport of newly synthesized proteins and lipids from the ER to the Golgi complex. AP complexes, including AP-1 through AP-5, facilitate transport between the plasma membrane, endosomes, and the TGN. The retromer complex is responsible for the retrieval of cargo from endosomes to the TGN and plasma membrane. These coatomer complexes are evolutionarily conserved and share structural similarities, reflecting their common ancestry and essential roles in vesicular trafficking [12]. In this study, we focused on COPI and COPII coatomer complexes given their critical functions in maintaining cellular homeostasis and their implications in various diseases, including cancer and neurodegenerative disorders [13,14,15,16,17].

The COPI complex, composed of seven subunits (α-, β-, β′-, γ-, δ-, ε-, and ζ-COP), mediates retrograde transport from the Golgi apparatus back to the ER. This process is initiated by the small GTPase Arf1, which anchors to the Golgi membrane upon GTP binding and recruits the coatomer complex, facilitating vesicle budding. Cargo proteins are incorporated into COPI vesicles through interactions with specific sorting signals, and uncoating occurs after GTP hydrolysis by Arf1, a reaction accelerated by ARF GTPase-activating proteins (ArfGAPs) [8,18]. In contrast, the COPII complex is responsible for anterograde transport of newly synthesized proteins and lipids from the ER to the Golgi complex. COPII assembly begins with the activation of the small GTPase Sar1, which is inserted into the ER membrane upon GTP binding. Sar1-GTP recruits the Sec23/Sec24 complex to form an inner coat layer that selects cargo proteins. Subsequently, the Sec13/Sec31 complex assembles to form an outer coat layer, driving vesicle formation via cage-like polymerization [19,20]. All components of the COPI and COPII subunits are summarized in Table S1.

Defects in coatomer proteins are associated with various congenital diseases. Moreover, the altered expression of coatomer proteins has been reported during cancer development and progression. Coatomer protein complex subunit beta 2 (COPB2) is a potential oncogene in multiple cancer types and is often overexpressed in breast, ovarian, and kidney cancers, where it correlates with poor patient prognosis. Its expression levels are associated with tumor progression and chemotherapy sensitivity, making it a candidate for targeted therapies. Depletion of COPB2 has been shown to induce autophagic cell death in cancer cells [16,21,22]. Another coatomer protein, COPZ1, is essential for the survival of certain tumor cells but dispensable for normal cells. This characteristic makes COPZ1 a promising therapeutic target, as its inhibition can selectively eliminate tumor cells without affecting healthy tissues. Studies have shown that silencing COPZ1 leads to cell death in various cancer types, including thyroid and prostate cancers [23,24,25].

To address the lack of a comprehensive understanding of coatomer biology in cancer, the present study systematically evaluated the genetic alterations, expression patterns, prognostic relevance, and functional dependencies of all COPI and COPII subunits across 33 cancer types using The Cancer Genome Atlas (TCGA) and DepMap datasets. This integrated pan-cancer approach aimed to identify functionally essential coatomer components with potential diagnostic or therapeutic significance.

Based on this rationale, we focused on COPG1, a COPI subunit hypothesized to play a key role in cancer progression and therapeutic response, for mechanistic and functional investigation in hepatocellular carcinoma (HCC). Cancer cells are known to adapt cellular mechanisms and remodel them to their advantage, thereby enabling resistance to chemotherapeutic drugs and promoting metastasis [26,27,28,29,30]. The objective of this study was to investigate the role of COPG1 in tumor progression and therapeutic response and to define the biological and clinical impacts of COPI dysregulation in HCC.

2. Results

2.1. Pan-Cancer Genetic Profiling Reveals Frequent Amplifications and Mutations of COPI and COPII Subunits with Prognostic Relevance

To investigate genetic alterations, including in-frame, missense, nonsense, fusion, amplification, and nonstop mutations in COPI and COPII subunits (Supplementary Table S1), we analyzed TCGA data from 33 cancer types using cBioPortal. Amplification was the most frequent alteration across the 20 coatomer genes (Figure 1A). The average alteration frequencies per cancer type were 8.84%, 9.33%, 2.39%, and 1.98% for mutations, amplifications, deep deletions, and multiple alterations, respectively. Notably, cutaneous skin melanoma (SKCM, 42.99%) and uterine corpus endometrial carcinoma (UCEC, 42.34%) showed the highest overall alteration rates, whereas thyroid carcinoma (THCA) exhibited the lowest rate (2.81%; Figure 1A). Cholangiocarcinoma (CHOL) showed the highest frequency of amplification (25.00%), whereas THCA showed the lowest frequency (0.40%; Figure 1A). Among individual genes, COPA displayed the most frequent alterations (5.02%; Figure 1B), primarily due to amplification (3.21%) (Figure 1B). COPB2 also showed notable alteration rates (2.87%), with amplification accounting for 1.67% (Figure 1B). COPA alterations were most frequent in CHOL (16.67%), while COPB2 alterations peaked in lung squamous cell carcinoma (LUSC) (10.06%; Figure 1C). In contrast, SAR1A showed the lowest alteration rate (0.79%; Figure 1B). DNA amplification and copy number gain in COPA and COPB2 were associated with significantly elevated mRNA expression (Figure 1D,E).

Mutation frequency varied among cancer types, with SKCM showing the highest mutation rate (31.90%) and thymoma (THYM) the lowest (0.81%; Figure 1A). SEC31B was the most frequently mutated subunit (1.84%; Figure 1B), whereas COPZ1 was the least frequently mutated (0.26%; Figure 1B). Missense mutations were predominant in both SEC31A (0.47%) and COPA (0.24%; Figure 1B,F).

Survival analysis of these genetic alterations indicated that mutations in COPA and SEC31A were prognostically significant in uterine corpus endometrial carcinoma (UCEC) according to Kaplan–Meier analysis (n = 516, p = 0.048 and p = 0.032, respectively; Figure 1G). In addition, COPB2 amplification showed a significant association with survival in patients with BLCA (n = 407, p = 0.048) and esophageal carcinoma (ESCA, n = 182, p = 0.046; Figure 1G).

2.2. Differential Gene Expression Analysis Between Cancer and Normal Tissues Identifies Cancer-Associated COPI and COPII Subunits

To explore differences in gene expression between tumor and normal tissues, we analyzed RNA-Seq data from TCGA. Vesicle trafficking genes showed broadly elevated expression in tumors (Figure 2A,B). Co-expression analysis revealed frequent co-expression patterns among COPI and COPII subunits, except COPZ2 and SEC31B (Figure 2C). Most of the 32 cancer types with RNA-Seq data from matched normal tissues showed frequent overexpression (Figure 2B,D). Acute myeloid leukemia (LAML) exhibited less frequent overexpression, with 4/20 subunits upregulated (Figure 2B,D). Among the subunits, COPZ2 showed the highest fold change in LAML (fold change = 3.47; Figure 2F). SEC31B was not overexpressed in most cancer types, except THYM and LAML (Figure 2B,E).

2.3. Cox Regression Analysis Reveals Prognostic Value of COPI and COPII Subunits

Prognostic analysis based on overall survival (OS) (Figure 3A) and progression-free survival (PFS) (Figure 3B) revealed that elevated expression of coatomer genes was generally associated with poor outcomes. COPI complex subunits showed particularly strong associations with OS (Figure 3C) and PFS (Figure 3D). ARCN1 exhibited the highest oncogenic impact in OS across approximately seven of 33 cancer types, whereas COPA displayed the strongest oncogenic association in PFS across approximately nine of 33 cancer types. Most subunits exhibited unfavorable prognostic patterns, though SAR1B and SEC31B showed favorable patterns more frequently. SEC23B showed favorable prognostic patterns in both OS and PFS (Figure 3C,D), highlighting heterogeneity within the coatomer family.

Across 20 cancer types, adrenocortical carcinoma (ACC) (6/20 genes), kidney chromophobe (KICH) (11/20), low-grade glioma (LGG) (10/20), and uveal melanoma (UVM) (6/20) demonstrated higher frequency of unfavorable OS patterns (Figure 3E), whereas ACC (14/20) and bladder urothelial carcinoma (BLCA) (8/20) exhibited higher frequency of unfavorable PFS patterns (Figure 3F). ACC, HNSC, KICH, LUSC, mesothelioma (MESO), and stomach adenocarcinoma (STAD) showed predominantly unfavorable patterns, while other cancer types showed mixed patterns. KIRC exhibited predominantly favorable associations (Figure 3E,F).

Volcano plot analyses using log₂(HR) and −log₁₀(p-value) identified multiple COPI members—particularly COPA, COPB1, COPG1, and ARCN1—as significantly associated with unfavorable outcomes in both OS (Figure 3G) and PFS (Figure 3H).

2.4. Pan-Cancer Analysis of the Dependency of Individual COPI and COPII Subunits

To examine the perturbation effect (knockout or knockdown) of each subunit on each cancer type, we analyzed the Chronos scores, which represent the effects of CRISPR-mediated knockouts, and the DEMETER2 scores, which represent the effects of RNAi-mediated knockdowns, for each subunit in the DepMap portal. Chronos scores were available for 48 subunits across 54 cancer types (Table S5), and DEMETER2 scores were available for 47 subunits across 46 cancer types. Subunits or cancer types lacking relevant data were excluded from analysis.

In the Chronos score analysis, eight subunits showed significant average dependency across cancer types (Chronos score < −1), while 10 subunits did not show significant dependency in any cancer type (Figure 4A). Depending on the percentage of dependency across cell lines, we classified subunits into a common (significant dependency in ≥90% of 1178 cell lines) or selective (significant dependency in <90% of cell lines) gene. Although many subunits have been classified as common essential genes (COPA, COPB1, COPB2, SEC13, ARCN1, COPZ1), two subunits (COPG1 and COPE) demonstrated strong selective lethality (Figure 4A, Table S6). Among these, COPA exhibited the strongest average dependency in ACC (Chronos score = −2.675).

In the DEMETER2 score analysis, nine subunits showed significant average dependency across cancer types (DEMETER2 score < −1; Figure 4B), whereas 10 others showed no significant dependency in any cancer type (Figure 4B). Only ARCN1 was categorized as a common essential gene, showing significant dependency in ≥90% of the 707 analyzed cell lines (Table S7). Eight subunits (COPA, COPB1, COPB2, COPG1, COPZ1, SEC13, COPE, SEC24D) showed strong selectivity. Among these, COPA displayed the strongest average dependency in UVM (DEMETER2 score = −1.861). Notably, two subunits (COPG1 and COPE) showed strong selective dependency in both Chronos and DEMETER2 analyses.

To assess the sensitivity of each cancer type to perturbation of COPI and COPII subunits, we counted the number of subunits with dependency scores < −1 per cancer type. The CRISPR score datasets revealed that most cancer types harbored seven to eight dependent coatomer subunits, whereas the RNAi score datasets identified three to six dependent subunits per cancer type (Figure 4C).

Next, we assessed the frequency of each subunit that exhibited dependency across cancer types. Based on the Chronos scores, eight subunits showed dependency in more than 30 cancer types, whereas 12 subunits did not show dependency in any type (Figure 4D). Based on the DEMETER2 scores, five subunits were dependent in more than 30 cancer types, whereas 11 subunits did not show dependency in any type (Figure 4D). Both the CRISPR and RNAi datasets showed that COPA, COPB1, COPB2, COPG1, ARCN1, COPE, COPZ1, and SEC13 were consistently expressed in a broad spectrum of cancers (Figure 4D).

These results demonstrated that several COPI and COPII subunits are indispensable for cancer cell survival, with stronger dependency patterns observed in the CRISPR datasets than in the RNAi-based screens. Furthermore, bar-plot analyses were generated to quantify the number of coatomer genes with dependency scores < −1 across cancer types (Figure 4C). When comparing the number of cancer types in which each coatomer gene exhibited strong dependency (score < −1), genes such as COPA, COPB1, COPB2, COPG1, ARCN1, COPE, COPZ1, and SEC13 were found to be dependent in more than 50 cancer types, although the magnitude of dependency varied between the Chronos and DEMETER2 datasets (Figure 4D).

Together, these findings highlight that multiple subunits of the COPI complex and SEC13 function as core essential genes across multiple cancer types, and that their essentiality is more robustly detected by CRISPR-based perturbation than by RNAi-based approaches.

Because both COPG1 and COPE subunits showed strong selective dependency in both Chronos and DEMETER2 analyses and showed consistent dependency patterns across a broad spectrum of cancers, we focused on both subunits in the remainder of this study. To validate both subunits as therapeutic targets, we compared their expression statuses, prognostic values, and dependency scores. Both subunits were significantly expressed in all cancer types that showed significant dependency on the CRISPR score (CRISPR < −1; Figure S2A,B). However, COPG1 showed prognostic significance in only two cancer types (LGG and LIHC; Figure S2C), and COPE in only three cancer types (ACC, CESC, and LIHC), showing significant dependency on the CRISPR score (CRISPR < −1; Figure S2C,D).

2.5. Pan-Cancer Analysis Shows Significant Association Between COP Subunit’s Dependency and Molecular Features of Cancer-Related Genes

To identify the molecular features associated with the dependency of individual COPI and COPII subunits, we examined the correlations between Chronos or DEMETER2 scores and molecular features, including gene expression, copy number variation, and mutation profiles (Figure S3). Associations between the COPI and COPII subunits and their molecular features were excluded from the analysis.

In the Chronos dataset, the strongest negative correlation was between COPG1 dependency and S100A11 expression (R = −0.350, p = 0.000), and the strongest positive correlation was between COPG1 dependency and GPC2 expression (R = 0.349, p = 0.000; Figure S3C). Notably, many cancer-related genes, including RRAS, HAUS1, RBMS, and ZKSCAN2, were significantly associated with COPG1 dependency (Figure S3C).

In the DEMETER2 analysis, the strongest negative correlation was between COPG2 dependency and USP24 damaging mutation (R = −0.360, p = 0.000; Figure S3D), and the strongest positive correlation was between COPB1 dependency and CALCA copy number variation (R = 0.557, p = 0.000; Figure S3D). Furthermore, higher S100A11 expression correlated negatively with stronger dependency on COPG1 (R = −0.297, p = 0.000; Figure S3D), which is consistent with the Chronos dataset. ARCN1 dependency was significantly associated with cancer-related gene alterations, including NDST1 expression and copy number alterations in TREH and PHLDB1 (Figure S3D). COPB2 dependency also showed a significant association with copy number alteration in SLC25A36 (Figure S3D).

2.6. Pan-Cancer Analysis Reveals a Significant Association Between COP Subunit Dependency and Drug Resistance

To investigate the relationship between subunit dependency and cellular signaling activity, we analyzed the correlations between Chronos or DEMETER2 scores and ssGSEA enrichment scores across cancer cell lines (Figure S4).

In the Chronos dataset, we identified numerous significant associations between drug resistance (Trabectedin, Dasatinib, Cisplatin, Mitoxantrone, and Fluorouracil) and dependence on specific subunits (SEC13, SEC24D, and COPG1; Figure S4). Similarly, the RNAi dataset revealed significant associations between drug resistance (Doxorubicin, Trabectedin, Imatinib, and Dasatinib) and subunit dependence (SEC24D, COPG1, SEC23A, and COPG2). Notably, COPG1 dependency was negatively associated with drug resistance, whereas the other subunits exhibited positive associations (Figure S4C,D).

2.7. Role of COPG1 in Malignant Phenotypes of the HCC

According to our comprehensive pan-cancer analysis, most COP subunits were significantly overexpressed in cancers and showed prognostic associations with multiple cancer types. Integrating loss-of-function datasets indicated that while the majority of COP genes behaved as common essentials (essential in >90% of cell lines), COPG1 and COPE exhibited selective essentiality. Because our study focused on hepatocellular carcinoma (LIHC) and COPG1 remains comparatively under-characterized among the COPI components, we prioritized COPG1 for functional and mechanistic validation in liver cancer models. For validation, we conducted a loss-of-function study using siRNA, which showed a knockdown efficiency > 70% across mRNA and protein expression levels (Figure S1A,B).

To evaluate the differential expression status of COPG1, databases like TCGA (HCC tumor samples) and GTEX (normal samples) were utilized and their mRNA expression status were evaluated using boxplot (Figure S5A). Furthermore, the protein expression levels of COPG1 were also evaluated by referring to websites like Proteomics Pathway Commons, which collects datasets from Clinical Proteomic Tumor Analysis Consortium (CPTAC), International Cancer Proteogenomic Consortium (ICPC) and Applied Proteogenomic Organizational Learning and Outcomes (APOLLO) (Figure S5B).

In multiple HCC cell lines, knockdown of COPG1 resulted in a significant reduction in proliferation rates (Figure 5A), consistent with the dependency data in the DepMap dataset. Moreover, COPG1 knockdown significantly impaired the colony formation ability of HCC cell lines (Figure 5B–D). Furthermore, in vivo studies showed that the tumor-forming capacity of the HCC cell lines to survive and proliferate in nude mice was deeply hampered. As a result, no growth of tumors was observed in the knockdown groups after two weeks of injection, whereas the control groups developed tumors nearly reaching 1 cm³ (Figure 5E,F). The knockdown efficiency of the injected cells has been examined, which demonstrated a significant reduction in the COPG1 expression level in the injected cell line (Figure S1B).

Consistent with the ssGSEA indications for adhesion/ECM pathways, COPG1 knockdown reduced the wound-healing ability of the cells compared to control cell lines (Figure 5G,H). In addition, Boyden chamber (Figure 5I) and transwell invasion assays (Figure 5I) showed that COPG1 knockdown impaired migration and invasion abilities of HCC cell lines. Collectively, these results demonstrated that COPG1 knockdown profoundly impairs proliferative, tumorigenic, migratory, and invasive capabilities of HCC cells.

2.8. COPG1 Knockdown Induces the Structural Change in Golgi and Golgi Stress

Based on our comprehensive analysis identifying COPG1 as a prognostically significant selective essential coatomer gene, and a functional validation study showing that loss of COPG1 significantly impaired malignant phenotypes of HCC cells both in vitro and in vivo, we sought to determine the underlying mechanism behind COPG1’s role in HCC progression.

RNA sequencing in PLCPRF5 cell revealed that pathways such as the unfolded protein response and regulation of Golgi organization were highly enriched and associated with high COPG1 expression (Figure 6A). Because COPI is involved in retrograde vesicle transport from the Golgi to the endoplasmic reticulum, anterograde transport within intra-Golgi compartments, and maintenance of Golgi structural integrity, we hypothesized that loss of COPG1, a crucial COPI subunit, could disrupt Golgi intra-structure and trigger stress-related pathways. A heat map of stress-regulated target genes showed that all stress-associated genes were highly enriched in COPG1 knockdown cells (Figure 6B).

Immunocytochemical analysis using the cis-Golgi marker GM130 and the trans-Golgi marker TGN46 demonstrated morphological changes upon siCOPG1 knockdown. COPG1 depletion caused marked Golgi structural alterations and increased expression of Golgi stress-related genes, including LAMP2, SEC24D, GABARAPL1, IRE1, and ATF6β (Figure 6C,D).

As Golgi stress may trigger ER stress, immunoblotting was performed to examine ER stress markers. COPG1 knockdown increased phosphorylated eIF2α and ATF4 levels in both HLF and PLCPRF5 cells (Figure 6E,F). Consistent with reports that ER stress can inhibit PI3K-AKT signaling [16], COPG1 depletion reduced phosphorylated AKT levels (Figure 6G,H) and decreased AKT luciferase activity in SNU886 and SNU761 cells (Figure 6I,J).

Collectively, these results indicate that COPG1 maintains Golgi architecture, thereby preventing Golgi and ER stress and sustaining AKT signaling, underscoring its role in cellular homeostasis.

2.9. COPG1 Knockdown Leads to ROS Accumulation and Hypoxia-Related Pathway Activation

Experimental validation demonstrated that COPG1 plays an important role in maintaining Golgi-ER homeostasis, possibly by preserving Golgi structural integrity and facilitating retrograde transport of essential ER-resident proteins. Notably, previous studies have shown that the ER and mitochondria form contact sites called mitochondria-associated membranes (MAMs), whereby stress in one organelle can affect the other [31,32]. Gene-set enrichment analysis of RNA-seq identified the hypoxia pathway as highly enriched (Figure 7A), and expression changes in hypoxia-related markers between siSCR and siCOPG1 groups were visualized in a heatmap (Figure 7B).

To test whether ER stress propagates to mitochondria and increases mitochondrial reactive oxygen species (ROS), we performed immunocytochemistry using DCFDA, 2′,7′-dichlorofluorescin diacetate (ROS indicator) and MitoTracker (mitochondria marker). Hydrogen peroxide (H₂O₂), a known ROS-inducer, served as a positive control. ROS were significantly elevated in siCOPG1-transfected cells after 48 h (Figure 7C), with green ROS signal overlapping red mitochondrial signals, indicating mitochondrial ROS accumulation.

To assess the impact of ROS on cell viability, cells were treated with the ROS scavenger NaSH·xH₂O (Sodium hydrosulfide hydrate, 50 µM). siSCR-transfected cells treated with the scavenger showed similar viability to naive cells, whereas siCOPG1-transfected cells treated with the scavenger exhibited restored proliferation, rescuing the reduction caused by COPG1 knockdown (Figure 7D).

2.10. A Novel Role of COPG1 in the Drug Sensitivity

Genome-wide CRISPR and RNAi-based ssGSEA pathway analyses (Figure 7A,B) revealed that COPG1 expression negatively correlated with multiple chemotherapeutic resistance signatures, including cisplatin, imatinib, dasatinib, and fluorouracil resistance pathways. The strong negative enrichment scores suggest that loss of COPG1 may enhance sensitivity to chemotherapeutic agents.

Given that reduced phospho-AKT and elevated ROS levels are associated with increased chemosensitivity, functional assays were performed to validate these observations. COPG1-knockdown and COPG1-overexpressing HCC cell lines were treated with increasing concentrations of sorafenib and doxorubicin (Figure 8). Cells with COPG1 knockdown exhibited increased drug sensitivity, reflected by a leftward shift in the IC₅₀ curve relative to siSCR controls. Conversely, COPG1-overexpressing cells displayed increased resistance, indicated by a rightward IC₅₀ shift.

Altogether, these bioinformatics and experimental results identify COPG1 as a key determinant of chemoresistance. Its depletion sensitizes HCC cells to anticancer drugs, whereas overexpression confers resistance, potentially via mechanisms involving ROS production and AKT pathway.

3. Discussion

Coatomer complex proteins, particularly those constituting the COPI and COPII vesicular trafficking systems, play essential roles in maintaining cellular homeostasis. Increasing evidence indicates that these components contribute substantially to the development and progression of cancer. Disruption of the COPI complex has been linked to Golgi fragmentation, the accumulation of immature autophagosomes, defective autophagy, and subsequent cell death [4,33,34]. Additionally, COPI subunits have been implicated in major oncogenic signaling pathways, including PI3K/Akt and Wnt, underscoring their potential role in tumorigenesis [15]. In our comprehensive analysis of COPI and COPII subunit expression across 33 cancer types using TCGA, we identified distinct expression patterns that were strongly associated with patient outcomes. These results highlight the potential utility of individual coatomer subunits as diagnostic biomarkers and therapeutic targets. Furthermore, functional perturbation data from DepMap reinforced their cancer type-specific essentiality, supporting their relevance as therapeutic vulnerabilities. Collectively, these results emphasize the critical involvement of vesicular trafficking machinery in cancer biology and underscore the need for further investigations that integrate protein interaction networks and clinical datasets to better define their roles in tumor progression.

Pan-cancer genetic profiling revealed frequent amplifications and mutations in COPI and COPII subunits with prognostic relevance. CNAs in COPA and COPB2 were associated with elevated mRNA expression (Figure 1D,E). Notably, COPB2 amplification demonstrated prognostic significance in BLCA and ESCA (Figure 1G). Overexpression of COPB2 has been linked to enhanced tumor cell proliferation, survival, invasion, and metastasis across multiple cancer types, including BLCA and ESCA [33,35]. Recent studies have identified COPB2 as a modulator of the Hippo signaling pathway, influencing the nuclear translocation of Yes-associated protein 1 (YAP1), a key regulator of cell growth and survival. In HCC, COPB2 affects cancer cell sensitivity to cisplatin by regulating YAP1 localization, suggesting a potential mechanism of chemoresistance [22,23,33,35]. Conversely, mutations in COPA have been associated with improved survival in patients with UCEC (Figure 1G). These mutations may lead to aberrant protein trafficking, which, in turn, affects cancer cell survival and proliferation [36]. Additionally, COPA mutations have been linked to increased expression of pro-inflammatory cytokines, such as interleukin-1β (IL-1β) and interleukin-6 (IL-6), which could contribute to immune cell activation and influence tumor progression [15,36,37]. COPA is a promising prognostic biomarker in cervical cancer and can enhance erdafitinib sensitivity through p16 and p21 [38,39]. Sec31A often forms oncogenic gene fusions, referred to as circSec31A, which are linked to aggressive tumor metastasis and act as potential diagnostic biomarker candidates [19,20]. These findings underscore the complex roles of COPA and COPB2 in cancer biology and highlight their potential as biomarkers and therapeutic targets. Further research is warranted to elucidate the mechanisms that contribute to tumor development and progression.

Global transcriptomic analysis revealed the widespread upregulation of COPI and COPII coatomer subunits across diverse cancer types, suggesting an elevated demand for vesicular trafficking, secretion, and membrane recycling to sustain the metabolic and proliferative needs of malignant cells (Figure 2A,B). A coordinated overexpression pattern was evident in almost all cancer types. However, LAML showed a relatively low frequency of overexpression. Correlation analysis further highlighted the highly co-regulated nature of most subunits, whereas SEC31B did not show overexpression across cancer types and was weakly or negatively correlated with other subunits (Figure 2C,E), possibly reflecting its tissue-restricted roles, suggesting that its loss may be tolerated in cancer cells. Notably, the correlation analysis revealed that the paralogous pairs COPG1–COPG2 and COPZ1–COPZ2 showed weak or negative co-expression patterns across cancers, suggesting that these paralogs may have diverged to perform lineage-specific non-redundant functions. In contrast, COPE, despite being frequently upregulated across cancer types, exhibited poor correlation with other COPI subunits, implying that it may also perform functions beyond its canonical role in the COPI complex. Collectively, these observations underscore that while most coatomer subunits are co-upregulated to support oncogenic demands, certain members, such as COPG2, COPZ2, COPE, and SEC31B, may follow divergent regulatory trajectories that could contribute to cancer-type-specific adaptation.

The present study identifies a novel mechanistic link between COPG1 and chemoresistance in hepatocellular carcinoma (HCC). COPG1 (Coatomer Protein Complex Subunit Gamma 1) encodes the γ1 subunit of the coatomer protein complex I (COPI), which plays a critical role in retrograde transport from the Golgi to the endoplasmic reticulum (ER), as well as in intra-Golgi trafficking [34,40]. COPI-coated vesicles are essential for preserving Golgi structure and for recycling ER-resident proteins that have mislocalized to the Golgi. Consistent with this function, COPG1 knockdown in our study resulted in disrupted Golgi morphology and activation of Golgi stress responses (Figure 6C,D).

This impaired trafficking induced ER stress, likely due to the accumulation of misfolded proteins in the ER lumen, as indicated by elevated ER stress markers (Figure 6A,B,E,F). These findings are consistent with previous studies [41,42] demonstrating that disruption of COPI function leads to ER stress. Moreover, other previous studies have reported that COPI dysfunction disrupts mitochondria-associated ER membranes (MERCs), leading to impaired mitochondrial calcium uptake, defective oxidative phosphorylation, and excess production of reactive oxygen species (ROS) [32,43]. In our study, COPG1 knockdown led to a significant increase in intracellular ROS levels, and the use of a ROS scavenger reversed the growth-inhibitory effects of COPG1 depletion (Figure 7C,D), suggesting that ROS accumulation plays a functional role in mediating the observed phenotype. Although our data do not directly demonstrate that COPG1 regulates ROS transport across the mitochondrial membrane, we propose that COPG1 influences mitochondrial redox balance indirectly, through its role in maintaining organelle communication and intracellular vesicle trafficking. Future studies using compartment-specific ROS indicators (e.g., MitoROX) and high-resolution imaging of ER–mitochondria contact sites will be necessary to further elucidate this mechanism.

Additionally, COPG1 knockdown reduced AKT phosphorylation and transcriptional activity (Figure 6G–J), in line with previous findings that ER stress can suppress PI3K–AKT signaling via CHOP induction and PTEN activation [4,16,44,45]. Given that PI3K–AKT signaling promotes chemoresistance in multiple cancer types [46,47], its attenuation may contribute to the increased drug sensitivity observed in COPG1-deficient cells (Figure 8). Furthermore, elevated mitochondrial ROS has been reported to potentiate the cytotoxicity of chemotherapeutic agents by inducing oxidative damage and triggering apoptosis [17,29,30,48,49,50]. Collectively, these findings suggest that enhanced chemosensitivity in COPG1-knockdown cells results from a combination of suppressed AKT signaling and elevated ROS levels, both of which contribute to reduced survival under chemotherapeutic stress [50,51,52].

In conclusion, our study identifies coatomer complex subunits as important contributors to cancer progression, highlighting diverse prognostic and context-dependent roles across tumor types. While COPB2 and COPA illustrate the tissue-specific nature of coatomer biology and SEC31A demonstrates fusion-associated oncogenic potential, COPG1 emerged as a key candidate associated with hepatocellular carcinoma (HCC) progression. Our data indicate that COPG1 expression is associated with sustained PI3K–AKT signaling and cellular stress adaptation, and that its depletion disrupts organelle homeostasis, leading to ER and mitochondrial stress and suppression of tumor-associated phenotypes (Figure 7). Beyond tumor-intrinsic effects, COPG1 perturbation was associated with altered chemotherapeutic responses, with knockdown sensitizing cells to, and overexpression conferring resistance against, sorafenib and doxorubicin. Together, these findings suggest that COPG1 and related coatomer proteins represent promising but as-yet unvalidated therapeutic vulnerabilities, particularly in tumors that exploit vesicle trafficking pathways for survival and drug resistance.

While the current study integrates transcriptomic and proteomic analyses with in vitro and in vivo functional validation, we acknowledge that it does not fully establish a causal relationship between COPG1 dysregulation and hepatocellular carcinoma (HCC) progression. Our knockdown experiments demonstrate that COPG1 depletion impairs tumorigenic properties such as proliferation, migration, invasion, and in vivo tumor growth, suggesting a functionally relevant role for COPG1 in HCC biology. However, these findings, while compelling, primarily indicate association rather than causality at the systems or clinical level. To establish definitive causality, further studies incorporating causal inference frameworks—such as Mendelian randomization, Bayesian network analysis, or integrative multi-omics modeling—are warranted. For example, recent work by Zhang et al. [53] demonstrates the utility of such approaches in identifying functionally and clinically relevant cancer drivers. Applying similar strategies in future research could help elucidate the upstream regulatory architecture of COPG1 and its direct causal role in HCC pathogenesis and therapeutic resistance.

Although our study highlights the functional significance of COPG1 in HCC tumorigenicity and drug resistance, it does not comprehensively address the potential immunological implications of COPG1 expression. Specifically, we did not perform an in-depth analysis of the tumor immune microenvironment (TIME), including immune infiltration patterns, immune subtypes, or immunotherapy-related markers. This represents an important methodological limitation of the current work. To address this gap, future studies should adopt a more integrative framework to explore the relationship between COPG1 and the immune landscape of HCC. For instance, the approach proposed by Lai et al. [54] provides a systematic methodology to characterize immune subtypes and predict immunotherapeutic responsiveness. Applying this framework, future research should (1) integrate bulk and single-cell transcriptomic datasets to investigate associations between COPG1 expression and immune cell infiltration; (2) evaluate COPG1 expression across well-defined immune subtypes and immunophenotypes; and (3) assess correlations between COPG1 and key immune regulatory factors, including checkpoint molecules and cytokine signaling pathways. Such analyses may offer new insights into the immunomodulatory role of COPG1 and reveal its potential as a biomarker or therapeutic target in the context of immune-based treatments for HCC.

4. Materials and Methods

4.1. Patient Cohorts and Data Analysis

We systematically analyzed genetic alterations, including mutations and copy number alterations (CNAs), as well as the expression profiles and prognostic significance of all subunits of the COPI and COPII coatomer complexes, using TCGA pan-cancer datasets. These datasets included approximately 11,000 primary and/or metastatic tumor samples from 33 cancer types and 11 pan-organ systems as previously described [22,24]. Genomic and transcriptomic profiling data were obtained from TCGA datasets accessed via cBioPortal to systematically assess COPI and COPII subunit alterations across multiple cancer types. Detailed methods for the patient cohorts and data analysis are described in the Supplementary Methods.

4.2. Mitochondrial Stress and ROS Detection

Cells were seeded onto 96-well plates. Twenty-four h later, the seeded cells were transfected with siRNA at approximately 40–50% confluency. After 48 h, transfected cells were used to detect ROS levels. Hydrogen peroxide (H₂O₂), a known inducer of ROS, was included as a positive control. Non-transfected cells were treated with positive controls for 2 h. Following the treatment period, MitoTracker and DCFDA solutions were co-treated with siRNA-transfected, H₂O₂-treated, and naïve cells in a 1:1 ratio dissolved in serum-free medium and incubated for 30 min. After the incubation, the cells were carefully washed with 1X PBS once and replaced with fresh medium to reduce background signal. Finally, ROS levels were observed using confocal microscopy.

4.3. In Vivo Xenograft Modeling

BALB/c nude mice (Orient, Sungnam, Republic of Korea) were used to establish subcutaneous xenograft models. In total, 4 × 10⁶ cells (PLCPRF5 cells) were prepared for each experimental group, including control (scrambled siRNA-transfected) and COPG1 knockdown (COPG1 siRNA-transfected) groups. Cells were transiently transfected in vitro, and after 48 h, they were harvested and resuspended in a 1:1 mixture of serum-free medium and Matrigel (Corning, Corning, NY, USA) to a final volume of 120 µL. Prior to injection, mice were anesthetized using an appropriate dose determined based on individual body weight following the institutional anesthesia dosing chart. A volume of 100 µL of the cell–Matrigel suspension was injected subcutaneously into the flanks of the mice. Tumor growth was monitored for two consecutive weeks. Once tumors reached approximately 1 cm in diameter, the mice were euthanized and tumors were excised for further analysis. Tumor volumes were measured for two weeks and calculated using the following formula: tumor volume (mm³) = (a × b²)/2, where a = length (mm) and b = width (mm). After seven weeks, the mice were sacrificed, and tumor volume and weight were measured. This study was performed in strict accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals from the National Institutes of Health. All experimental procedures were approved by the Pusan National University Institutional Animal Care and Use Committee approved the experimental procedures (approval number PNU-2025-0545). Tumor growth was allowed based on the ethical committee guidance criteria, and excised tumor size did not exceed 1 cm according to ethics guidelines.

4.4. RNA-Seq and Data Processing

Total RNA was extracted using the RNeasy Mini Kit (Qiagen, Hilden, Germany). RNA quality was assessed by running 1 μL of RNA on a Bioanalyzer system (Agilent, Santa Clara, CA, USA) to ensure the RIN and rRNA ratio. We used 100 ng of total RNA to prepare sequencing libraries using the MGIEasy RNA Directional Library Prep Kit (MGI). Libraries were quantified using an Agilent 2100 BioAnalyzer (Agilent) according to the manufacturer’s library quantification protocol. Following cluster amplification of the denatured templates, paired-end sequencing (2 × 150 bp) was performed using the MGI MGISEQ-T7 platform (MGI, San Jose, CA, USA). The data were processed as previously reported [55]. Briefly, raw sequencing reads were assessed using FastQC (v0.12.1) (RRID:SCR_014583) and aligned with STAR (v2.7.11b) (RRID:SCR_004463) to the human reference genome hg38 with Ensembl annotation. Transcript abundance was quantified using RSEM5 in TPM units. Differential gene expression analysis was performed using DEseq2 (v1.44.0) (RRID:SCR_015687). The FASTQ and RNA-Seq sequences reported in this article were deposited in the NCBI Gene Expression Omnibus database (accession number GSE311074).

4.5. Statistical Analysis

Statistical analyses were performed using R software (v4.5.2). For paired comparisons between cancer and matched normal samples, the Wilcoxon signed-rank test was applied (using the coin R package) (v4.5.2). For comparisons between two unpaired groups, either the Mann–Whitney U test or Student’s t-test was used, depending on data distribution and variance homogeneity. When comparing more than two groups, one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons test was used for parametric data, while the Kruskal–Wallis test was applied for non-parametric data. Survival differences were evaluated using Kaplan–Meier survival curves and compared by the log-rank test. Univariate survival analyses were conducted using the Cox proportional hazards regression model. Values < 0.05 were considered statistically significant. Statistical significance was indicated using the following convention: p < 0.05 (*), p < 0.01 (**), and p < 0.005 (***).

4.6. Supplementary Methods

Additional experimental procedures, including antibodies, reagents, cell culture conditions, siRNA transfection, proliferation assays, chemosensitivity assays, migration and invasion assays, real-time PCR, luciferase assays, Western blotting, and immunocytochemistry, are provided in the Supplementary Information (Tables S2–S4, Supplementary Method).