BE screen reveals METTL3 S2 dephosphorylation sensitizes gastric cancer cells to oxaliplatin by interfering METTL3-eIF3H interaction

Abstract

The resistance to oxaliplatin (OXA)–based chemotherapies may lead to poor prognosis in patients with gastric cancer (GC). Emerging evidence suggests that resistance is closely associated with phosphorylation modifications. In GC cell line AGS, high-throughput base editor screen identified key phosphorylation sites associated with OXA response. Methyltransferase-like 3 (METTL3) S2 emerged as a notable negative hit. Further investigation revealed that dephosphorylation of METTL3 S2 disrupted the METTL3–eukaryotic translation initiation factor 3 subunit H (eIF3H) interaction, thereby suppressing the translation of oncogenes involved in replication stress responses, including bromine domain protein 4 (BRD4) and serpin family E member 2 (SERPINE2), ultimately enhancing sensitivity to OXA. In addition, clinical investigation showed that METTL3 S2 phosphorylation was highly correlated with the response to GC OXA chemotherapy. In summary, base editor screen provides a versatile approach for exploring the role of phosphorylation sites in cancer chemotherapy. The METTL3-eIF3H interaction may serve as a potential therapeutic target.

Base editor functional screen directly reveals that METTL3 S2 phosphorylation regulates gastric cancer response to oxaliplatin.

INTRODUCTION

As a notable global health issue, gastric cancer (GC) is one of the leading causes of cancer-related mortality (1, 2). Oxaliplatin (OXA) is a platinum-based chemotherapy drug that primarily induces DNA cross-linking and replication stress, ultimately leading to cell death. Clinical practice has confirmed the effectiveness of OXA-based chemotherapy in treating GC patients. However, the development of chemotherapy resistance often results in poor outcomes for GC patients (2, 3). This adverse prognosis is partly driven by protein kinases/phosphatases, which coordinate cell cycle control with DNA damage response (DDR) to enhance DNA repair and enable cells to tolerate replication stress (4, 5). Numerous studies have demonstrated that the inhibition of these pathways can resensitize tumor cells to chemotherapy (6–9). Several inhibitors targeting replication stress response–related kinases such as ataxia telangiectasia and Rad3-related protein kinase (ATR), checkpoint kinase 1 (CHK1), and WEE1 G2 checkpoint kinase (WEE1) inhibitors have shown preliminary potential for combination therapies with DNA-damaging agents (10–13). Current research on OXA chemotherapy resistance has primarily focused on the functions of protein kinases (4). However, the broad regulatory effects and inherent toxicity of kinase inhibitors have substantially limited their clinical application (11, 12, 14). Consequently, our research focuses on identifying more specific phosphorylation sites that are directly linked to chemotherapy resistance.

Protein phosphorylation orchestrates various cellular functions by regulating protein subcellular localization, protein-biomolecular interactions, enzymatic activity, and stability, and increasing evidence underscores its regulatory role in replication stress responses related to signaling pathways (15–18). Phosphorylation is indispensable for exact and precise control of protein activity upon stress-induced stimulation (19, 20). The function of protein phosphorylation is highly complex. Even a single protein can have multiple phosphorylation sites, each contributing to distinct functional roles in regulating various biological processes. Such as p53, an important hub to handle a broad range of cellular stress, have shown that the effect of phosphorylation on p53 (activation or inhibition) is highly variable under different conditions. N-terminal phosphorylation causes p53 stabilization by inhibiting the p53–MDM2 (mouse double minute 2) interaction. In contrast, Ser149 (S149), Thr150 (T150), and T155 activate in unstressed cells and promote p53 degradation (21, 22). Notably, the dysregulation of phosphorylation is frequently associated with therapeutic resistance and tumor progression. Although pioneering studies have demonstrated the importance of phosphorylation in OXA resistance (23–25), the underlying mechanisms are still unclear.

The phosphoproteomics analyses can quantify tens of thousands of phosphorylation sites and track their dynamics upon cell stimulation, drug treatment, or mutational status (26–28). However, current research on phosphorylation sites remains substantially limited, with over 95% of phosphorylation sites lacking identified upstream kinase or specific biological function (19). Given the vast number of phosphorylation sites, a precise, traceable, and high-throughput technique is essential for advancing phosphorylation research. Adenine base editors (ABEs), a tool developed based on CRISPR-Cas9, enable efficient single-base editing without introducing double-strand breaks (DSBs) (29, 30). This tool has been widely used to mutate phosphorylation sites, such as serine, threonine, and tyrosine, to mimic dephosphorylation and elucidate their biological functions. When combined with CRISPR screening technology, we can unbiasedly identify phosphorylation sites related to phenotypes of interest. Recent studies have confirmed the feasibility of base editing for genome-wide screens, proving valuable insights into the distinct roles of individual amino acids in biological processes (31, 32).

In this work, we aim to identify key phosphorylation sites involved in OXA treatment through ABE screen, uncover their underlying mechanisms, and explore potential druggable targets. This research will pave the way to overcome chemotherapy resistance, improve patient outcomes, and guide precision medicine.

RESULTS

Base editor screen of functional phosphorylation sites with OXA treatment

We constructed a genome-wide phosphorylation screening library that included 39,689 guide RNAs (gRNAs) within 7907 biologically functional genes from the UniProt database, with gRNA designs specific to each phosphorylation site (www.uniprot.org/) (table S1). ABE8e-SpRY, a highly efficient ABE with minimal protospacer-adjacent-motif(PAM) constraints (33), was used to construct the stable ABE-expressing GC cell line. The cells were screened under OXA treatment (10 μM) for 14 days, after which surviving cells were collected for genome extraction and identification of enriched gRNAs through targeted deep sequencing (Fig. 1A). We compared pre- and post-OXA treatment by biological duplicates (n = 2). The candidate functional phosphorylation sites were identified and ranked using the model-based analysis of the genome-wide CRISPR-Cas9 knockout (MAGeCK) program (34). Data are shown in table S2.

Fig. 1. Base editor (BE) screen of functional phosphorylation sites with OXA treatment.

(A) The workflow of ABE screening under OXA pressure in AGS; n = 2. (B) The volcano plot of the AGS ABE screen. The negative top three hits are indicated by red triangles, whereas the positive top three hits are denoted by blue squares. (C) Mixed-sample competition assay of the negative top three hits with OXA treatment for 14 days; n = 3. (D) PTM-SEA terms for the results of AGS ABE screen. (E) Cell viability of OXA and OXA + adavosertib (Ada) (200 nM) in AGS and HGC27 cells; n = 3. The data are expressed as means ± SD. LFC, log₂ fold change. EGFP, enhanced green fluorescent protein; sgRNA, single-guide RNA.

The top three negative and positive screening hits were presented in Fig. 1B. We designed and transfected corresponding gRNAs into AGS and HGC27 cells, then cocultured the transfected cells with wild-type (WT) cells to assess their responses to OXA. As shown in Fig. 1C and fig. S1A, all these sites consistently exhibited OXA resistance or sensitivity in both AGS and HGC27 cells, aligning with the AGS ABE screening results. The posttranslational modification signature enrichment analysis (PTM-SEA) (35) enriched a large number of proteins related to cancer progression, cell cycle, and DNA damage repair. It also highlighted several signaling pathways closely linked to OXA resistance, such as the WNT pathway (Fig. 1D) (36, 37). Gene ontology (GO) analysis revealed that the targeted proteins enriched from the phosphorylation library were primarily involved in cell cycle, protein phosphorylation, cell division, apoptotic process, and other cell cycle–related biological processes. These proteins are mainly associated with molecular functions such as protein binding, RNA binding, and protein kinase activity (fig. S1B).

Studies have shown that the phosphorylation of WEE1 S53 and cyclin-dependent kinase 1 (CDK1) T14 plays a critical role in the replication stress response (4). The positive third-site hit, WEE1 S53, is phosphorylated by PLK1, which affects its protein stability (38). In addition, the negative third-site hit, CDK1 T14, is phosphorylated by WEE1, leading to the inhibition of CDK1 activity (39). To validate these screening hits, we combined WEE1 inhibitor (adavosertib) with OXA, which demonstrated a significant inhibitory effect on GC cells (Fig. 1E). The exceptional performance of these known sites further reinforces our confidence in the ABE screens and suggests a potential synergistic chemotherapy of OXA and adavosertib.

The top-ranked sites are likely to play a critical role in the response of GC cells to OXA. Extensive studies have shown that methyltransferase-like 3 (METTL3) is closely related to GC progression, and its inhibition has been reported to enhance chemotherapy sensitivity (40–42). While current research primarily focuses on METTL3’s regulation of oncogene expression via N⁶-methyladenosine (m⁶A) modifications, emerging evidences suggest that METTL3 may also promote GC progression and chemoresistance through m⁶A-independent mechanisms (43). This prompted us to investigate the underlying mechanisms of its phosphorylation sites. Hence, we selected METTL3 S2 for further investigation into its potential role in chemotherapy sensitization.

Dephosphorylation of METTL3 S2 enhances sensitivity to OXA in vitro and in vivo

We performed multiple sequence alignment using the ESPript 3.0 online tool (https://espript.ibcp.fr/ESPript/ESPript/) (44) to investigate the conservation of the METTL3 S2 residue. The analysis revealed a high degree of conservation at and around the S2 site of METTL3, suggesting its potential importance in protein function. (fig. S2A). To verify whether phosphorylation at the S2 site of METTL3 plays a crucial role under OXA treatment, we established both-copy monoclonal cell lines in AGS and HGC27 cells, where the serine phosphorylation residue was mutated to alanine (S2A) (fig. S2B). Cell counting kit 8 (CCK8) assays were conducted to assess the effect of METTL3 WT and S2A on cell proliferation under OXA treatment. As shown in Fig. 2A, dephosphorylation of METTL3 S2 significantly increased the sensitivity of GC cells to OXA, and this effect was more pronounced than that observed with other chemotherapy drugs (fig. S2C). Similarly, colony formation (Fig. 2B) and cell apoptosis assays (Fig. 2C) demonstrated that, compared with WT cells, the METTL3 S2A cells exhibited a markedly reduced proliferation rate and an increased apoptosis rate following OXA treatment. We also established monoclonal cell lines with serine mutations to aspartic acid (S2D) and glutamic acid (S2E) (fig. S3A). However, neither the D nor E mutation effectively mimicked the constitutive phosphorylation, and no significant differences in proliferation rates were observed between S2D and S2E mutants and WT cells (fig. S3, B and C).

Fig. 2. Dephosphorylation of METTL3 S2 enhances sensitivity to OXA in vitro and in vivo.

(A) CCK8 assay showing the median inhibitory concentration (IC₅₀) and the cell viability at 10 μM OXA in METTL3 WT/S2A of AGS and HGC27; n = 3. (B) Colony formation assay of METTL3 WT/S2A in the absence or presence of OXA (10 μM); n = 3. (C) Flow cytometry analysis of the apoptosis rates [fluorescein isothiocyanate (FITC)⁺ cells] treated with OXA (50 μM, 48 hours); n = 3. (D) Photograph of harvested tumors for the four groups; n = 6. (E) Growth curves of the tumor for the four groups; n = 6. (F) Tumor weight of the four groups; n = 6. (G) IHC staining showing Ki-67 expression in the xenograft tumor of the four groups; n = 3. The data are expressed as means ± SD.

To assess the impact of METTL3 S2A on the efficacy of OXA treatment in vivo, we performed subcutaneous transplantation of HGC27 METTL3 WT and S2A cells, with or without OXA treatment (fig. S3D). No significant difference in body weight was observed among the four groups (fig. S3E). Compared to the METTL3 WT (OXA) group, the METTL3 S2A (OXA) group exhibited a significant reduction in both tumor volume and weight (Fig. 2, D to F). Immunohistochemical (IHC) staining of tumors revealed a marked decrease in the expression of Ki-67, a marker of proliferation, in the METTL3 S2A (OXA) group compared to the METTL3 WT (OXA) group (Fig. 2G), indicating reduced tumor cell proliferation. Collectively, these findings demonstrate that dephosphorylation of METTL3 S2 significantly enhances sensitivity to OXA in GC, both in vitro and in vivo.

Dephosphorylation of METTL3 S2 enhances OXA-induced DNA damage

OXA exerts its antitumor effects by inducing DNA cross-links, causing DNA damage and repair failure, leading to disrupted cell division and apoptosis. Here, comet assay was performed to assess the extent of DNA damage after OXA treatment in METTL3 WT/S2A group. The results revealed a higher proportion of damaged DNA in the METTL3 S2A (OXA) group compared to the METTL3 WT (OXA) group (Fig. 3A). Simultaneously, we conducted immunofluorescence (IF) staining for γH2AX, a marker of DNA damage. Dephosphorylation of METTL3 S2 led to a significantly increase in γH2AX immunostaining after OXA treatment (Fig. 3, B and C). Western blotting (WB) analysis showed that the total protein levels of METTL3 remained unchanged across all groups. However, after OXA treatment, phosphorylation at the METTL3 S2 site (p-METTL3 S2) was significantly elevated in the WT group, whereas the METTL3 S2A group exhibited no detectable p-METTL3 S2 expression, confirming that the S2A mutation effectively mimicked the dephosphorylation of METTL3 S2. The validation of the primary antibody is shown in fig. S4 (A and B). Consistently, γH2AX levels were significantly higher in the METTL3 S2A (OXA) group, indicating more severe DNA damage (Fig. 3D). These findings collectively indicate that dephosphorylation of METTL3 S2 enhances DNA damage in response to OXA treatment.

Fig. 3. Dephosphorylation of METTL3 S2 enhances OXA-Induced DNA damage.

(A) Representative IF images of comet tails in the absence or presence of OXA (10 μM, 24 hours); n = 3. The extent of DNA damage was quantified by measuring TailDNA%. (B) Representative IF images of γH2AX and nuclear [4′,6-diamidino-2-phenylindole (DAPI)] in METTL3 WT/S2A in the absence or presence of OXA (10 μM, 24 hours); n = 3. (C) Number of γH2AX foci per cell of METTL3 WT/S2A in the absence or presence of OXA (10 μM, 24 hours); n = 3. (D) Representative WB images of METTL3, p-METTL3 S2, and γH2AX of METTL3 WT/S2A in the absence or presence of OXA (10 μM, 24 hours), with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) serving as a loading control; n = 3. The data are expressed as means ± SD.

Phosphorylation of METTL3 S2 is regulated by MEK1 and influences its interaction with eIF3H

Protein phosphorylation is predominantly mediated by kinases. The kinase library (https://kinase-library.phosphosite.org/site) (45, 46) can accurately predict the direct kinases responsible for regulating specific protein substrates. Using this tool, mitogen-activated protein kinase kinase 1 (MEK1) was identified as the most likely upstream kinase responsible for METTL3 S2 phosphorylation (fig. S4C). Molecular docking also predicted a strong binding affinity between METTL3 and MEK1, with a docking score of −641.07 kcal/mol (fig. S4D). Consistently, coimmunoprecipitation (CoIP) experiments revealed the interaction between METTL3 and MEK1 (fig. S4E). Subsequently, we used two MEK1 inhibitors, selumetinib and trametinib, in combination with OXA to assess their synergistic effects on GC. The results of CCK8 assay showed significant inhibition of GC cell growth (fig. S4F). Selumetinib notably reduced p-MEK1 levels, followed by a corresponding decrease in p-METTL3 S2 levels. In contrast, the extracellular signal–regulated kinase inhibitor (ulixertinib) did not induce such changes, confirming that MEK1 can directly phosphorylate the S2 site of METTL3 (fig. S4G). Furthermore, γH2AX levels were significantly elevated in the combination treatment group (selumetinib and OXA) compared to the control (CON) and monotherapy groups, indicating increased DNA damage (Fig. 4A). These findings indicate that inhibiting the upstream kinase MEK1 reduces METTL3 S2 phosphorylation directly, leading to increased DNA damage and sensitivity to OXA.

Fig. 4. Phosphorylation of METTL3 S2 is regulated by MEK1 and influences the interaction with eIF3H.

(A) Representative WB images of METTL3, p-METTL3 S2, MEK1, p-MEK1, and γH2AX in the absence or presence of OXA (10 μM, 24 hours) and selumetinib (100 nM, 24 hours), with GAPDH serving as a loading control; n = 3. (B) CoIP of Myc-METTL3 WT/S2A and Flag-METTL14. Flag-mCherry was included as a negative control; n = 3. (C) The volcano plot shows significant differential proteins in the IP-MS results of AGS METTL3 WT and S2A; n = 3. (D) CoIP of METTL3 and eIF3H in METTL3 WT/S2A cells; n = 3. (E) Reciprocal CoIP of eIF3H and METTL3 in METTL3 WT/S2A cells; n = 3. (F) CoIP of METTL3 and eIF3H in the absence or presence of OXA (10 μM, 24 hours); n = 3. (G) Reciprocal CoIP of eIF3H and METTL3 in the absence or presence of OXA (10 μM, 24 hours); n = 3. The data are expressed as means ± SD. ns, not significant.

We further explored potential targets influenced by METTL3 S2 phosphorylation. Given the critical role of METTL3 in m⁶A modification (40, 47), we first examined the interaction between METTL3 and METTL14. Unexpectedly, the S2A mutation did not affect the interaction between METTL3 and METTL14, nor did it alter the level of m⁶A methylation modification (Fig. 4B and fig. S5A). Subsequently, we performed immunoprecipitation mass spectrometry (IP-MS) on METTL3 WT and S2A AGS cells. As shown in Fig. 4C, METTL3 S2A mutant exhibited a significant reduction in binding to eukaryotic translation initiation factor 3 (eIF3) subunit proteins. The IP-MS data are provided in table S3. GO analysis also enriched in pathways related to the formation of cytoplasmic translation initiation complex and the eIF3 complex, primarily associated with translation initiation factor activity and RNA binding (fig. S5B). In addition, as with the previous description (48) and recent study (49, 50), translation/ribosomal pathways may be involved.

The direct interaction between METTL3 and eukaryotic translation initiation factor 3 subunit H (eIF3H) has been proven in a previous study (51). Therefore, we hypothesized that METTL3 S2 phosphorylation may influence its binding to eIF3H. To validate this, we performed CoIP and found that the binding of METTL3 S2A to eIF3H was notably weaker than that of METTL3 WT, indicating that dephosphorylation of METTL3 S2 weakens its interaction with eIF3H (Fig. 4, D and E, and fig. S5C). OXA treatment enhanced the binding of METTL3 to eIF3H (Fig. 4, F and G), further supporting the key role of METTL3 S2 phosphorylation in this interaction. Moreover, based on the mutational profile from the library ABE screening, we constructed METTL3 S2P mutants too and observed similar effects to the S2A (fig. S6).

Dephosphorylation of METTL3 S2 suppresses tumor growth through inhibiting replication stress response

The METTL3-eIF3H interaction is essential for enhancing the translation of specific oncogenes, forming densely packed polyribosomes, and facilitating oncogenic transformation (51). After confirming the regulatory role of METTL3 S2 phosphorylation in its interaction with eIF3H, we detected the expression levels of these related oncogenes, and there was significant down-regulation of their protein expression levels in the METTL3 S2A mutant, but no change in mRNA was observed by quantitative polymerase chain reaction (qPCR; fig. S7). Furthermore, OXA treatment significantly increased the protein levels of these oncogenes in the METTL3 WT without affecting their mRNA levels. These findings suggested that METTL3 S2 phosphorylation enhances the translation of specific oncogenes by increasing its binding with eIF3H.

Previous studies have demonstrated that bromine domain protein 4 (BRD4)–ATR–CHK1 plays an indispensable role in replication stress-induced checkpoint signaling. Inhibition of key DNA replication stress response genes ATR and CHK1 significantly enhances OXA’s cytotoxicity against GC cells (fig. S8A). Inhibition of BRD4 leads to a rapid, time-dependent reduction in ATR-CHK1 phosphorylation and aberrant DNA replication re-initiation, and sensitizes cancer cells to various replication stress-inducing agents (52–54). Our results showed that BRD4 expression was reduced in METTL3 S2A, accompanied by a corresponding decline in p-ATR and p-CHK1 levels, suggesting that S2 dephosphorylation of METTL3 inhibits the BRD4-ATR-CHK1 axis (Fig. 5A and fig. S8B). Cell cycle analysis revealed a significantly lower proportion of S-phase cells in METTL3 S2A cells compared to WT cells following OXA treatment (Fig. 5B), suggesting that the METTL3 S2A inhibits the BRD4-ATR-CHK1 axis, leading to death in S phase and slippage of damaged cells into mitosis, then mitotic catastrophe. IHC further confirmed reduced BRD4 expression in METTL3 S2A (Fig. 5C).

Fig. 5. Dephosphorylation of METTL3 S2 inhibits replication stress responses.

(A) Representative WB images of BRD4, ATR, p-ATR, CHK1, and p-CHK1 in the absence or presence of OXA (10 μM, 24 hours), with GAPDH serving as a loading control; n = 3. (B) Representative images and quantification of flow cytometry cell cycle in the absence or presence of OXA (10 μM, 12 hours); n = 3. (C) IHC staining of BRD4 expression in the xenograft tumor of the four groups; n = 3. (D) Cell viability of OXA and OXA + JQ1 (100 nM) in AGS and HGC27 cells; n = 3. (E) Representative WB images of γH2AX in the absence or presence of OXA (10 μM, 24 hours) and JQ1 (100 nM, 24 hours), with GAPDH serving as a loading control; n = 3. (F) Representative IF images of γH2AX and nuclear (DAPI) in the absence or presence of OXA (10 μM, 24 hours) and JQ1 (100 nM, 24 hours); n = 3. The extent of DNA damage was quantified by measuring γH2AX foci per cell. (G) Photograph of harvested tumors for the four groups; n = 6. (H) Growth curves of the tumor for the four groups; n = 6. (I) Tumor weight of the four groups; n = 6. The data are expressed as means ± SD.

These findings prompted us to explore the antitumor potential of combining OXA with JQ1, a BET bromodomain inhibitor targeting BRD4. CCK8 assays showed that the combination of OXA and JQ1 significantly inhibits GC cell growth (Fig. 5D and fig. S8C). As shown in Fig. 5 (E and F), γH2AX levels were significantly increased in the combination group, compared to the CON and monotherapy groups, indicating a substantial increase in DNA damage with the combined treatment. Furthermore, JQ1 inhibits the ATR-CHK1 signaling pathway activated by OXA treatment (fig. S8D). To verify its efficiency in vivo, we established HGC27 WT xenograft models and randomly divided them into four groups: CON, JQ1, OXA, and JQ1 plus OXA (fig. S9A). The results demonstrated that the combination group significantly inhibited tumor growth compared to the CON or monotherapy groups (Fig. 5, G to I), accompanied by decreased Ki-67 expression in xenograft GC tissues (fig. S9B). There were no significant differences in body weight across the groups (fig. S9C). Overall, the combination of JQ1 and OXA showed enhanced antitumor efficacy.

We also pay attention to serpin family E member 2 (SERPINE2), an oncogene known to influence the accumulation of phosphorylated ataxia-telangiectasia mutated (p-ATM) and downstream DNA repair protein RAD51 homolog 1 (RAD51) during replication stress processes (55). In our work, SERPINE2 decreased significantly after METTL3 S2 dephosphorylation, while markedly increased following OXA treatment. Similarly, the protein levels of p-ATM, phosphorylated checkpoint kinase 2 (p-CHK2), and RAD51 also showed a notable decline (fig. S10, A and B). These findings suggest that dephosphorylation of METTL3 S2 suppresses SERPINE2-ATM-RAD51 axis in GC cells, which is associated with OXA-induced replication stress response.

The phosphorylation levels of METTL3 S2 were significantly elevated in nonresponsive GC samples following OXA chemotherapy

To further evaluate the clinical significance of METTL3 S2 phosphorylation, we analyzed GC samples from patients treated with the XELOX regimen, which includes OXA. METTL3 S2 phosphorylation levels were significantly higher in nonresponsive patients compared to those who responded to XELOX. Moreover, nonresponsive patient samples exhibited more cytoplasmic accumulation of METTL3 (Fig. 6A). The previous study has shown that the cytoplasmic/nuclear expression ratio of METTL3 in clinical samples is directly correlated with the translational efficiency of oncogenic mRNAs (43). These results further support the finding that METTL3 enhances the translation of oncogenes through its interaction with eIF3H in the cytoplasm, independent of m⁶A modification. In summary, OXA-induced p-MEK1 enhances METTL3 S2 phosphorylation to promote oncogenic translation and replication stress response, reducing OXA sensitivity, whereas METTL3 S2 dephosphorylation reverses these effects, increasing OXA sensitivity (Fig. 6B).

Fig. 6. METTL3 S2 phosphorylation was significantly elevated in nonresponsive GC samples after OXA chemotherapy.

(A) Representative images of H&E and IHC staining showing p-METTL3 S2 and METTL3 expression in clinical GC samples. (B) Mechanism of METTL3 S2 dephosphorylation enhancing sensitivity to OXA. The data are expressed as means ± SD. CAP, cap-binding protein.

DISCUSSION

Functional screening technology based on CRISPR-Cas9 provides a powerful platform for gene function research and drug target discovery (56). Traditional CRISPR knockout (KO) screening methods rely on Cas9-mediated DSBs, which can easily induce cell death and affect the accuracy of gene function screens (57, 58). In contrast, base editor (BE) screen does not produce DSBs, thereby reducing gene-independent cytotoxicity induced by editing tools and offering substantial advantages in screening precision and accuracy (59). Given that GC chemotherapeutic drugs typically exert their effects by inducing DNA damage (60), BE screens, which do not rely on DSBs, are particularly well suited for chemotherapy-related research. In addition, BE screen can effectively edit a single base and accurately simulate point mutation (61). This approach facilitates the transition from protein-level alterations to genetic modifications at the nucleic acid level (62, 63). In recent years, several remarkable studies have demonstrated the versatility of BE screens (31, 32, 62–65). In our work, we used ABE8e-SpRY, a near-PAMless ABE variant (33), to overcome PAM restrictions and maximize genome-wide phosphorylation site coverage, ultimately achieving stable and reliable results.

Dysregulation of phosphorylation is directly related to cancer development and chemotherapy resistance. A rising number of studies have demonstrated that phosphorylation sites play diverse and critical roles in OXA chemotherapy, highlighting the necessity of exploring chemotherapy resistance/sensitivity from the perspective of phosphorylation (23, 24, 66). Phosphorylation modifications primarily occur on three amino acids: serine, threonine, and tyrosine. ABE can mimic the dephosphorylation of these residues through precise single-base editing, making them ideal tools for studying the functions of phosphorylation sites. Unlike traditional CRISPR KO screens, which eliminate target protein expression and result in the complete loss of its functions, ABE screens preserve protein expression, allowing for greater precision in identifying the roles of PTMs at specific sites (62, 63, 67, 68). This targeted approach provides a more refined tool for investigating phosphorylation-dependent signaling and condition-specific protein activities, overcoming the limitations of KO screens that often obscure nuanced functional insights. In this study, we used an ABE screening approach to directly identify protein phosphorylation sites critical for the OXA response, shifting the focus from overall protein function to specific regulatory modifications. Previous studies have demonstrated the feasibility of ABE screens in investigating functional phosphorylation sites, confirming their utility in uncovering phosphorylation-dependent regulatory pathways (31, 32). By directly targeting specific phosphorylation sites, this approach bypasses the need to explore from proteins to sites, offering a more efficient and precise method for identifying functional regulatory mechanisms.

Existing BE screens of protein phosphorylation sites primarily focus on mapping the comprehensive landscape. We aimed to investigate the underlying mechanisms by which the identified phosphorylation sites contribute to OXA resistance and identify promising therapeutic targets to improve clinical outcomes for GC patients. In our work, CDK1 T14 and WEE1 S53 were proven to influence replication stress response and contribute to OXA resistance (38, 39), highlighting the feasibility of further exploring the mechanisms underlying single phosphorylation site. Based on the screening and validation results, we focused on the S2 site of the METTL3, a prominent research hotspot, for further investigation.

METTL3 is an RNA methyltransferase primarily responsible for m⁶A methylation modifications. Abnormal up-regulation of METTL3 has been observed in various cancers, and METTL3 deficiency sensitizes cancers to chemotherapy (40, 69). Recent studies have indicated that METTL3 regulates gene expression by catalyzing m⁶A modifications on mRNA (including mRNA involved in DDR), thereby contributing to tumor resistance (70, 71). For example, METTL3 stabilizes poly(adenosine 5′-diphosphate–ribose) polymerase 1 mRNA through m⁶A modification, increasing the activity of the base excision repair pathway and promoting resistance to OXA (72). A series of small molecule inhibitors targeting the METTL3-METTL14 binding capability have been developed to investigate the role of METTL3-regulated methylation modifications in chemotherapy, showing promising results in the treatment for leukemia and other cancers (73). Recently, it has been reported that METTL3 phosphorylation directly participates in the DDR, with ATM-mediated phosphorylation of METTL3 at S43 site being specifically activated by DSBs and recruited to DSBs to promote DDR (74). Additionally, the cytoplasmic METTL3 has been found to enhance the translation of epigenetic mRNAs in an m⁶A-independent manner, identifying it as an oncogenic driver in cancer progression (43). These findings highlight the diverse functions of METTL3 and its potential roles in chemotherapy resistance, although the precise and comprehensive mechanisms remain unclear. Our study reveals that phosphorylation of METTL3 S2 plays a critical role in chemotherapy-induced replication stress response, primarily through its interaction with eIF3H. These findings provide insights into the pivotal role of METTL3 in chemotherapy resistance and cancer progression, paving the way for a deeper understanding of its therapeutic potential.

In the AGS ABE screen, METTL3 S2 dephosphorylation sensitized GC cells to OXA. Subsequent results indicated that METTL3 S2 dephosphorylation down-regulated the protein expression of oncogenes associated with replication stress responses, such as BRD4 and SERPINE2, by disrupting the METTL3-eIF3H protein interaction, ultimately resulting in enhanced sensitivity to OXA. The phosphorylation of METTL3 S2 shows potential as a prognostic biomarker in clinical practice. Previous studies have demonstrated the tumorigenic role of the METTL3-eIF3H interaction in lung and colorectal cancers (51, 75). Our findings further reveal its critical role in GC and chemotherapy resistance, suggesting that this interaction may broadly promote tumorigenicity and affect multiple stages of cancer progression. Thus, the METTL3-eIF3H interaction may serve as a potential therapeutic target.

Several limitations and challenges are also acknowledged in our study: (i) Our work primarily focuses on individual phosphorylation sites, serving as an important starting point for understanding the broader, coordinated, and systemic roles of phosphorylation in cancer chemotherapy. Future efforts will prioritize constructing a functional landscape to achieve a more comprehensive understanding. (ii) Our research is currently based on OXA and GC, which lays the foundation for future investigations into the role of the METTL3-eIF3H interaction across various cancer types and DNA-damaging chemotherapeutic agents.

In conclusion, genome-wide phosphorylation screens by base editing provide a powerful platform to eliminate confounding factors, enabling the direct and precise identification of key phosphorylation sites associated with chemotherapy resistance/sensitivity. This strategy not only advances our understanding of the molecular determinants of cancer but also facilitates the discovery of potential therapeutic targets, accelerating the path to precision medicine.

MATERIALS AND METHODS

Cell culture

Human GC cell lines HGC27 (TCHu 22) and AGS (TCHu232) were purchased from the Culture Collection of the Type Culture Collection of the Chinese Academy of Sciences and validated by mycoplasma testing and short tandem repeat profiling analyses. The cells were cultured using RPMI 1640 (11875093, Gibco) medium with 10% fetal bovine serum (10099141C, Gibco) and penicillin-streptomycin (15140122, Gibco) in 5% CO₂ at 37°C. PCR was used to test the cell lines for mycoplasma contamination.

Reagents and plasmids

Reagents: OXA (S1224, Sellek), selumetinib (HY-50706, MCE), trametinib (HY-10999, MCE), adavosertib (HY-10993, MCE), 5-fluorouracil (HY-90006, MCE), cisplatin (HY-17394, MCE), irinotecan (HY-16562, MCE), ulixertinib (HY-15816, MCE), VE-821(HY-14731, MCE), AZD-7762 (HY-10992, MCE), and JQ1 (T2110, TargetMol).

Plasmid: Lenti-single-guide RNA (sgRNA)-mCherry-puro and Lenti-EF1a-8e-SpRY were used for adenine base editing. pMD2.G, psPAX2, and Lenti-EF1a-8e-SpRY were used for lentiviral packaging. CMV-Csy4-PE and prime editing guide RNA (pegRNA)-Csy4site-sgRNA-GFP were used to mutate METTL3 S2A. Myc-METTL3 WT/S2A, Flag-mCherry, Flag-MEK1, Flag-eIF3H, and Flag-METTL14 were used to overexpress tagged proteins.

ABE screen under OXA pressure

AGS cell line stably expressing ABEs was established and maintained with Blasticidin S (10 μg/ml; A1113902, Thermo Fisher Scientific). These cells (5 × 10⁷) were then infected with lentiviral gRNA library at a low multiplicity of infection (MOI: 0.3) to achieve >300× coverage, followed by selection with puromycin (2 μg/ml; A1113802, Thermo Fisher Scientific) after 2 days. After 7 days of infection, a portion of the cells was collected as the CON group, while the experimental group was treated with 10 μM OXA to allow gRNA enrichment or depletion. After 14 days of OXA treatment, surviving cells were collected as the experimental group. Two biological replicates were collected for each experimental condition.

Genomic DNA isolation and sequencing

Cells were lysed overnight at 55°C with lysis buffer containing proteinase K (100 μg/ml; 10412ES, Yeasen). After digestion, DNA extraction buffer was added to the mixture, which was then vortexed thoroughly and centrifuged. The upper phase was transferred, mixed with isopropanol, and centrifuged again. The pellet was washed with 75% ethanol, air-dried, and dissolved in prewarmed ddH₂O (75°C). DNA concentration was then measured.

Genomic DNA was used as a template for gRNA amplification via two rounds of PCR using Q5 High-Fidelity DNA Polymerase (M0491L, NEB). In the first round, multiple 100-μl reactions with 10 μg of genomic DNA were performed under the following conditions: 95°C for 3 min, then 24 cycles of 98°C for 10 s, 64°C for 20 s, and 72°C for 30 s. PCR products were pooled, and 10 μl was used as the template for a second 100-μl reaction with 17 cycles under the same conditions. Purified products were sent to Novogene for analysis using NovaSeq X Plus.

Mixed-sample competition assay and deep sequencing

The gRNAs targeting the top three hits from the pooled screen were constructed individually, with the spacer of gRNAs listed in table S4. These gRNAs were cotransfected with an ABE to introduce mutations at the corresponding phosphorylation sites. Mutant and WT cells were cocultured under 10 μM OXA pressure for 14 days to mimic screening conditions. Cells were collected on days 0, 7, and 14, and genomic DNA was extracted using QuickExtract DNA Extraction Solution (QE09050, Lucigen). PCR amplification of ~200–base pair (bp) fragments containing the target sites was performed using Phanta Max Super-Fidelity DNA Polymerase (P505, Vazyme). The primers used are listed in table S5. These PCR products were sent to Annoroad Gene Technology for deep sequencing using Illumina NovaSeq (PE150). Sequencing reads were demultiplexed using AdapterRemoval (v2.2.2), and the pair-end reads with 11 bp or more alignments were combined into a single consensus read. All processed reads were mapped to the target sequences using the BWA-MEM algorithm (BWA v0.7.16). Editing efficiency was calculated as the percentage of (number of reads with the desired edit)/(total mapped reads).

Establishment of mutant cell lines

The pegRNAs were designed using the PrimeDesign online tool (https://github.com/pinellolab/PrimeDesign) to introduce a serine-to-alanine mutation at the METTL3 S2 site. The designed pegRNA and prime editor plasmids were cotransfected into AGS and HGC27 cell lines. Three days after transfection, single-cell clones were sorted by fluorescence-activated cell sorting (FACS; BD FACSAria III), including a gate to exclude doublets. The sorted clones were expanded for 14 days, followed by PCR amplification and Sanger sequencing. Mutant clones with the desired S2A/D/E mutation were expanded for subsequent experiments: pegRNA spacer oligo top, ggtgtcagggctgggagact; pegRNA spacer oligo bottom, agtctcccagccctgacacc; pegRNA extension (S2A) oligo top, acgtgtccgCcatcctagtctcccagc; pegRNA extension (S2A) oligo bottom, gctgggagactaggatgGcggacacgt; nick sgRNA spacer oligo top, gagagtccagctgcttcttg; nick sgRNA spacer oligo bottom, caagaagcagctggactctc.

CCK8 assay

Cells were counted and seeded into 96-well plates. After 24 hours, these cells were treated as indicated for 48 hours. Then, 10 μl of CCK8 (C0038, Beyotime) was added to each well, and the plates were incubated for 2 to 4 hours. A microplate reader (SpectraMax M2, Molecular Devices) was used to read the absorbance at 450 nm.

Colony formation assay

GC cells were counted and seeded into six-well plates with 500 cells per well. After treatment with 10 μM OXA for 48 hours, the drugs were removed and replaced with fresh medium for 14 days. Then, the cells were fixed and stained, followed by colony counting.

Comet assay

To evaluate DNA damage, we performed an alkaline comet assay using a comet assay kit (C2041, Beyotime) according to the manufacturer’s instructions. Ultimately, the slides were stained with propidium iodide (PI), and pictures were captured with a fluorescence microscope (Olympus). The DNA damage was calculated by measuring the tail DNA% (percent DNA in the tail).

IF staining

To evaluate DNA damage, we used a DNA Damage Assay Kit by γH2AX Immunofluorescence (C2035S, Beyotime) according to the manufacturer’s instructions. Pictures were captured with laser confocal microscopy (Leica).

Cell apoptosis and cell cycle

Samples were harvested, washed, and stained with annexin V/fluorescein isothiocyanate (FITC) and PI dye (AP101C, MultiSciences) for 10 min. The cell cycle was determined by staining with DNA staining and permeabilization solution (CCS012, MultiSciences). The final apoptotic rate and cell cycle distribution were detected using flow cytometry (CytoFLEX, Beckman Coulter). During flow detection, the acquisition voltage, rate, and gating settings were consistent for each group of cells.

WB analysis

Samples were lysed with RIPA (BL504A, Biosharp) containing phenylmethylsulfonyl fluoride (BL1426A, Biosharp) and phosphorylase inhibitor (G2007, ServiceBio) at a ratio of 100:1:1, and protein concentrations were analyzed using a BCA Protein Assay Kit (P0012, Beyotime). Then, the proteins were boiled, separated, transferred to polyvinylidene difluoride membranes (Merck Millipore), and blocked with 5% fat-free milk (GC310001, ServiceBio) for 60 min. After that, primary antibodies were incubated at 4°C overnight. The antibodies used are listed in table S6. The next day, the membranes were washed and incubated with secondary antibodies (Proteintech) for 1 hour at room temperature. An enhanced chemiluminescence (ECL) detection kit (1705060, Bio-Rad Laboratories) was used to visualize the protein bands.

CoIP assay

For endogenous CoIP, anti-METTL3/eIF3H antibody (1:50) and magnetic protein A/G beads (B23201, Selleck) were incubated at 4°C for 6 hours with gentle rotation. Normal immunoglobulin G (IgG) was used as a negative IP control. Cells were collected, lysed using IP lysis buffer (SL1030, Coolaber) supplemented with protease inhibitor cocktail (P8430, Sigma-Aldrich) for 30 min on ice, and centrifuged at 12,000g for 15 min. After protein quantification using a BCA Protein Assay Kit (P0012, Beyotime), lysates were incubated with the beads that bind to the antibody overnight at 4°C. The beads were washed at least five times with wash buffer, and proteins were eluted by boiling in 1× loading buffer at 100°C for 10 min and then used for WB.

For exogenous CoIP, the corresponding plasmids with Myc- and Flag-tags were overexpressed in AGS cells. Cell lysates were immunoprecipitated with an anti-Myc antibody, followed by WB analysis using an anti-Flag antibody. Flag-mCherry was included as a negative control.

Mass spectrometry

For IP-MS analysis, the IP samples were separated by SDS–polyacrylamide gel electrophoresis and stained with Coomassie Blue, after which the gel bands were excised into small pieces. Alternately, these pieces were treated with acetonitrile or ammonium bicarbonate to remove the background stain. Subsequently, the proteins were reduced using dithiothreitol (DTT) and alkylated with iodoacetamide (IAM) before being digested overnight with trypsin at 37°C. The following day, the supernatants were collected by centrifugation, desalted, and subjected to liquid chromatography–tandem mass spectrometry (LC-MS/MS) analysis using the Orbitrap Fusion (Thermo Fisher Scientific). Raw data were processed using MaxQuant software, and protein identification was performed against the UniProt database.

For proteomic profiling analysis, the cell samples were added to lysis buffer and disrupted by sonication. Subsequently, the proteins were reduced using DTT and alkylated with IAM before being digested overnight with trypsin at 37°C. The following day, the supernatants were collected by centrifugation, desalted, and subjected to LC-MS/MS analysis using the Orbitrap Astral (Thermo Fisher Scientific). Raw data were processed using MaxQuant software, and normalized using the total protein and the consistently expressed housekeeping protein glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as described in previous studies (76, 77).

Quantitative real-time PCR

Total RNA isolation and cDNA synthesis were conducted according to the instructions (CWBIO). The ΔCt method was applied to calculate the relative expression level, with GAPDH serving as an internal reference. The primers used in this study were as shown in the table S7.

Dot blot assays

The extracted RNA or polypeptides were dotted onto the nitrocellulose membrane (HATF00010, Merck Millipore). The membranes were then blocked and incubated with primary antibody overnight at 4°C. The next day, the membranes were washed and incubated with secondary antibodies for 1 hour at room temperature. An ECL detection kit (1705060, Bio-Rad Laboratories) was used to visualize the membranes.

Enzyme-linked immunosorbent assay

The antigen was coated onto the microplate, followed by washing and blocking. The primary antibody was incubated at 37°C for 1.5 hours. After washing, the secondary antibody was incubated at room temperature for 1 hour. Trimethylboron chromogenic solution was added for 5 to 10 min, then the reaction was terminated, and the data are read with a microplate reader (SpectraMax M2, Molecular Devices).

Molecular docking

HDOCK was used for docking interactions. The structure of MEK1 was predicted via AlphaFold3 (https://alphafoldserver.com/), and METTL3’s three-dimensional structure was obtained from Protein Data Bank (www.rcsb.org/) (ID:6Y4G), with missing residues completed using SWISS-MODEL. Phosphorylation was introduced through site-directed mutagenesis in PyMOL 2.5.7, followed by optimization and adjustment of the protein structure. Molecular docking and conformational scoring were conducted using ITScorePP. A negative score indicates molecular binding, with larger absolute values representing stronger binding affinity.

Animal models

The five-week-old female BALB/c nude mice were purchased from GemPharmatech Co., Ltd. All procedures followed institutional and national guidelines for laboratory animal care and use. To establish the HGC27 tumor model, 2 × 10⁶ HGC27 cells were implanted subcutaneously into the right axilla of mice (this day was marked as day 0). Tumor-bearing mice were randomly divided into four groups (n = 6) at day 8 and received different treatments as indicated at day 9. Mice bearing tumors were administrated every 3 days through intraperitoneal injection. The treatment duration lasts for 18 days. The mice weights and tumor sizes were also measured every 3 days. The tumor volume calculation formula was length × width² × 0.5. At day 28, all mice were sacrificed, and samples were harvested and processed for the subsequent assays. All experiments were permitted by the Laboratory Animal Center of Zhongnan Hospital of Wuhan University (no. ZN2024073).

Hematoxylin and eosin and IHC staining

For hematoxylin and eosin (H&E), tissues were fixed with 4% formalin and dehydrated. After deparaffinization and rehydration, the sections were stained with hematoxylin solution (51275, Sigma-Aldrich) and eosin solution (318906, Sigma-Aldrich).

For IHC staining, the sections were respectively incubated with anti-Ki-67 antibody (1:1000), anti-METTL3 antibody (1:1000), anti-p-METTL3 S2 antibody (1:200), anti-BRD4 antibody (1:100), and anti-SERPINE2 (1:100) antibody for 12 hours at 4°C and then incubated with horseradish peroxidase–conjugated goat secondary antibody at room temperature, followed by 3,3′-diaminobenzidine (P0203, Beyotime) reaction. The results were photographed with light microscope (Olympus). The relative cytoplasmic-to-nuclear expression ratio of METTL3 = the positive area ratio of METTL3 in cytoplasm/the positive area ratio of METTL3 in nucleus.

Human tissue samples

Surgically resected specimens from 12 GC patients who received adjuvant chemotherapy were obtained at Zhongnan Hospital (Wuhan, Hubei, China) from September 2024 to March 2025. These patients underwent three cycles of chemotherapy before surgery. The response rate was assessed according to the RECIST (Response Evaluation Criteria in Solid Tumors) 1.1 guidelines. The tumor response included complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD). CR and PR were considered response to chemotherapy, whereas SD and PD were regarded as nonresponse. All samples were clinically and pathologically diagnosed. The clinical information of all patients is listed in table S8. This study was approved by Zhongnan Hospital’s Protection of Human Subjects Committee (no. 2025034K). Written informed consent was obtained from each patient.

Statistical analysis

Statistical analyses were performed on GraphPad 8.0 software. Student’s t test was used for the analysis of two groups. A one-way analysis of variance (ANOVA) was used to compare the means of more than two groups. Experiments were repeated as biological replicates. ABE screens were analyzed as described in the relevant sections above. The data are expressed as means ± SD. Statistical significance was defined as P < 0.05.

Supplementary Materials

The PDF file includes:

Supplementary Text

Figs. S1 to S10

Tables S1, S3 to S8

Legend for table S2

Other Supplementary Material for this manuscript includes the following:

Table S2