A lipoprotein-biomimetic nanoplatform for oncogenic KRAS mutation-specific lysosome-targeting protein degradation

Abstract

Lysosome-targeting protein degradation strategies have evolved to leverage lineage–specific receptors to improve selectivity, yet remain unable to distinguish malignant from normal cells within the same lineage. Since oncogenic mutations are the defining feature of malignant cells, we here develop KRAS-mutated (MT) tumor targeting chimeras (KRASTAC), a lipoprotein-biomimetic nanoplatform that hijacks the enhanced macropinocytosis in KRAS-MT cells for mutation-specific protein degradation. KRASTAC potently degrades diverse proteins including epidermal growth factor receptor and programmed cell death ligand 1 (PD-L1) in KRAS-MT cells at a low antibody concentration of 2.5 nM within 9 h, with minimal impact in KRAS wild-type (WT) cells. Furthermore, KRASTAC exhibits mutation-specific tumor accumulation and protein degradation in vivo. KRASTAC-induced PD-L1 degradation leads to robust antitumor efficacy and reprogramming of the immunosuppressive microenvironment in mice carrying Kras-MT tumors. Our work presents KRASTAC as a potent mutation-specific protein degradation strategy for KRAS-MT cancers and establishes macropinocytosis as a route for lysosome-targeting protein degradation.

Lysosome-targeting protein degradation strategies leverage the lineage–specific receptors to improve selectivity towards cancer cells. Here this group develops a KRAS-mutated tumor targeting chimeras that hijacks the enhanced macropinocytosis in KRAS-MT cells for mutation-specific protein degradation thereby achieving anticancer effects.

Introduction

Targeted protein degradation (TPD) has emerged as a powerful therapeutic strategy that harnesses cellular clearance machinery to eliminate disease-causing proteins¹. Recent advances have expanded these approaches to degrade extracellular and membrane proteins through lysosome-targeting strategies, including lysosome-targeting chimeras (LYTAC)², transferrin receptor-targeting chimeras (TransTAC)³, and cytokine receptor-targeting chimeras (KineTac)⁴. These platforms employ bifunctional molecules that simultaneously engage targeted proteins and lysosome-targeting receptors (LTR), routing them toward lysosomal degradation. Beyond molecular strategies, biocompatible nanoparticles (NPs) represent a promising platform for lysosome-targeting protein degradation, as they can be internalized via multiple endocytic pathways independent of cell-specific receptors and trafficked to lysosomes^5–7. To date, several NP-based TPD platforms, including TPD-NP, MONOTAB, and Nanosphere, have been developed^8–10. These platforms exploit the structural flexibility and ease of surface modification of NPs, enabling versatile degradation of diverse targets.

However, both molecular and NP-based lysosome-targeting degraders often lack cell-type specificity due to either the broad expression of LTR or non-selective cellular uptake of NP, raising concerns about on-target, off-tissue toxicity. To solve this problem, several TPD platforms have been engineered to achieve cellular selectivity. For instance, SelecTAC^B and GalNAc-LYTAC confer cell-type selectivity by engaging the B-cell-specific LTR CD22 and the hepatocyte-enriched asialoglycoprotein receptor, respectively^11,12. Yet these systems remain incapable of distinguishing malignant from normal cells within the same lineage, particularly when they share receptor expression profiles. Moreover, their reliance on cell-type-specific LTRs introduces vulnerabilities, including ligand competition and the risk of resistance through receptor downregulation. Given that most cancers are driven by somatic mutations in key oncogenes and tumor suppressor genes that distinguish them from normal cells^13,14, we reason that a TPD strategy enabling mutation-specific protein degradation could overcome the limitations of current platforms.

KRAS mutations are among the most prevalent oncogenic drivers in lethal tumors yet remain challenging to target because the encoded protein lacks conventional druggable pockets and operates within a complex, adaptive signaling network¹⁵. Although KRAS^G12C inhibitors represent a breakthrough, their clinical utility is limited by the baseline activation and treatment-induced amplification of receptor tyrosine kinases such as epidermal growth factor receptor (EGFR), human epidermal growth factor receptor 2 (HER2)^16–19. In addition, KRAS mutations upregulate the PD-L1 expression and transforming growth factor β/interleukin-6 (TGF-β/IL-6) secretion to create an immunosuppressive tumor microenvironment (TME), which compromises the efficacy of immunotherapy^20,21. Given that these treatment challenges are largely mediated by membrane-associated and extracellular proteins, there is an urgent and unmet need to develop oncogenic KRAS mutation–specific TPD strategies that precisely and efficiently degrade these pathogenic proteins.

Notably, KRAS-MT tumors exhibit a metabolic dependency on macropinocytosis and subsequent lysosomal degradation to fuel their growth in nutrient-poor environments^22–24. We reasoned that this distinct metabolic profile between KRAS-MT and -WT cells could be leveraged to achieve mutation-specific lysosome trafficking and subsequent protein degradation. Supporting this concept, we and others have shown that lipoprotein-biomimetic NPs formed by the adsorption of apolipoprotein E (APOE) onto lipid nanoparticles (LNP) are selectively internalized via macropinocytosis and trafficked to lysosomes in KRAS-MT cells^25–27. These findings position APOE-based lipoprotein-biomimetic NPs as a promising vehicle for oncogenic KRAS mutation-specific lysosome-targeting protein degradation.

Here, we present KRASTAC, a TPD nanoplatform that enables oncogenic KRAS mutation-specific protein degradation by leveraging tumor-intrinsic endocytic vulnerabilities. KRASTAC demonstrates superior degradation efficiency against multiple therapeutic targets, including EGFR and PD-L1, at substantially lower concentrations than existing platforms. Notably, it achieves potent and mutation-specific tumor accumulation and degradation in vivo, with durable anti-tumor response observed in syngeneic models through PD-L1 degradation and reversal of immunosuppressive microenvironment. The modular assembly via click chemistry enables flexible and versatile tailoring of KRASTAC for diverse target proteins. Mechanistically, we reveal that the mutation-specific TPD of KRASTAC is driven by macropinocytosis and critically depends on the upregulated expression of LDLR and SLC35B2 in KRAS-MT tumors. Collectively, this work presents KRASTAC as an efficient, versatile, and mutation-specific TPD strategy with broad therapeutic potential for KRAS-MT cancers and establishes macropinocytosis as a route for lysosome-targeting protein degradation.

Results

Design of the KRASTAC platform

To overcome the lack of selectivity in current lysosome-targeting degradation platforms, we engineered KRASTAC to achieve oncogenic KRAS mutation-specific protein degradation by exploiting oncogene-driven macropinocytosis (Fig. 1). In this design, targeting antibodies are conjugated to the surface of lipoprotein-biomimetic nanoparticles, which are formed by the adsorption of apolipoprotein E (APOE) onto LNP. The antibodies serve to specifically capture membrane proteins of interest, while the APOE-LNP drive robust cellular uptake and lysosomal trafficking via macropinocytosis. Because upregulated macropinocytosis is a distinct metabolic dependency of KRAS-MT tumors, KRASTAC selectively routes the engaged target proteins for rapid lysosomal degradation specifically in KRAS-MT cells. In contrast, this process remains highly inefficient in KRAS-WT cells, thereby safely sparing healthy tissues and minimizing off-target toxicity.

Fig. 1. Schematic illustration of the oncogenic KRAS mutation-specific lysosome-targeting protein degradation platform.

KRASTAC consists of lipoprotein-mimetic NPs functionalized with targeting antibodies. In this design, the lipoprotein-mimetic NPs promote cellular uptake and lysosomal trafficking, and the antibody captures the protein of interest. In KRAS-MT tumor cells, enhanced macropinocytosis activity facilitates efficient internalization of the KRASTAC–target protein complex, leading to robust lysosomal delivery and degradation. In contrast, KRAS-WT cells exhibit low macropinocytosis, resulting in minimal uptake, negligible lysosomal localization, and degradation, thereby conferring mutation-specific protein degradation. Created in BioRender. Xixi, H. (2026) https://BioRender.com/haw4ol2.

Oncogenic KRAS mutation-specific lysosome targeting by KRASTAC

To evaluate the lysosomal delivery capability of the KRASTAC platform, we required a traceable model cargo. For this purpose, we synthesized biotin-modified LNPs (LNP-biotin) and incubated them with APOE to generate KRASTAC-biotin (Fig. 2a). Streptavidin (SA) was then employed as a model cargo, which binds to the biotin on the NP surface and allowed us to directly monitor the delivery of SA to lysosomes (Fig. 2b). Flow cytometry analysis showed that KRAS-MT HCT116 cells incubated with KRASTAC-biotin showed significantly higher percentage of AF647⁺ DiI⁺ cells and AF647 fluorescence intensity than those treated with LNP-biotin (Fig. 2c–e). Consistent with these observations, confocal microscopy of KRAS-MT HCT116 cells expressing LAMP1-zsGreen revealed pronounced punctate co-localization of DiI-labeled KRAS-TAC-biotin, AF647-SA, and lysosomal markers LAMP1 following co-incubation, demonstrating efficient cellular uptake and lysosomal localization. In contrast, cells treated with LNP-biotin exhibited negligible SA trafficking into lysosomes (Fig. 2f).

Fig. 2. Validation of oncogenic KRAS mutation-specific lysosome targeting of KRASTAC.

a Schematic illustration of the synthesis of KRASTAC-biotin. Created in BioRender. Xixi, H. (2026) https://BioRender.com/a0sgndu. b Assay of the KRASTAC-biotin-induced internalization and lysosomal trafficking of AF647-SA. Cells were incubated with 30 μg/mL AF647-SA and 10 μg/mL KRASTAC-biotin or LNP-biotin for 2 h and subjected to flow cytometry and confocal microscopy. Created in BioRender. Xixi, H. (2026) https://BioRender.com/emkrrvh. c–e Flow cytometry evaluating the uptake of SA in HCT116 cells treated with KRASTAC-biotin or LNP-biotin. n = 3 independent experiments. f Confocal microscopy images of HCT116 cells expressing LAMP1-ZsGreen treated with AF647-SA and DiI-labeled KRASTAC-biotin or LNP-biotin for 1 h. Scale bar, 5 μm. n = 3 independent experiments. g,h Confocal microscopy images and relative AF647 fluorescent intensity of multiple human cancer cells treated with AF647-SA and KRASTAC-biotin for 1 h. n = 5 independent experiments. Scale bar, 10 μm. i,j Confocal microscopy images and relative AF647 fluorescent intensity of Kras^G12D and parental MC38 cancer cells treated with AF647-SA and KRASTAC-biotin for 1 h. Scale bar, 5 μm. n = 5 independent experiments. Data are presented as mean ± s.d. p values were determined by (d,e,j) unpaired two-tailed Student’s ttest or (g) one-way analysis of variance (ANOVA) with Dunnett’s multiple comparisons test. Source data are provided as a Source Data file.

The oncogenic KRAS mutation-specific lysosome targeting specificity of KRASTAC was further validated across multiple KRAS-MT cell lines derived from pancreatic, colorectal, and lung cancers, compared to their WT counterparts. Quantitative analysis demonstrated significantly higher SA fluorescence intensity in KRAS-MT cells compared to WT cells (Fig. 2g, h). Furthermore, using CRISPR/Cas9-engineered Kras^G12D MC38 cancer cells (Supplementary Fig. 1a) and their isogenic WT controls, we found that Kras^G12D MC38 cells treated with KRASTAC-biotin exhibited significantly higher SA fluorescence intensity, consistent with observations in above human cancer cells (Fig. 2i, j). Together, these data demonstrate that KRASTAC achieves lysosomal delivery specifically in KRAS-MT cells.

Preparation of antibody-conjugated KRASTAC platform

Building upon the proof-of-concept findings, we designed a synthetic route for preparing antibody-conjugated KRASTAC via click chemistry (Fig. 3a and Supplementary Fig. 2a). Briefly, azide-modified LNPs produced by microfluidics were reacted with dibenzocyclooctyne (DBCO)-modified antibody followed by incubation with APOE to yield the final construct. To characterize the antibody conjugation, denaturing protein electrophoresis with Coomassie blue staining was performed for cetuximab (Ctx)-conjugated LNP (LNP-Ctx) (Supplementary Fig. 2b) and two bands near 50 kDa and 25 kDa were observed, corresponding to the heavy and light chains of the antibody, respectively. Slight smearing of these bands was noted, which could be attributed to antibody-lipid conjugation. These results confirmed successful antibody conjugation to LNPs. Morphological characterization by transmission electron microscopy (TEM) confirmed the formation of spherical LNP-Ctx (Supplementary Fig. 2c). To characterize and optimize APOE density, we first evaluated APOE adsorption by incubating LNP-Ctx with gradient amounts of APOE. To separate the LNP-Ctx from unbound APOE in the mixture, the LNP-Ctx were pulled down using protein A magnetic beads targeting the surface-conjugated antibody. Subsequent ELISA analysis demonstrated that APOE adsorption is highly controllable by modulating the input ratio (Supplementary Fig. 2d, e). Functional validation via western blot revealed that target EGFR degradation efficiency peaked at an APOE-to-lipid weight ratio of 0.2 (Supplementary Fig. 2f), establishing this as the optimal formulation for all subsequent studies. Finally, dynamic light scattering (DLS) analysis indicated that the resulting KRASTAC particles exhibited a uniform size distribution, with a Z-average diameter of 126.8 ± 5.3 nm and a polydispersity index (PDI) of 0.225 ± 0.026 (Supplementary Fig. 2g).

Fig. 3. KRASTAC facilitates rapid and oncogenic KRAS mutation-specific degradation of PD-L1.

a Preparation of antibody-conjugated KRASTAC. Created in BioRender. Xixi, H. (2026) https://BioRender.com/a0sgndu. b Schematic of KRASTAC-mediated PD-L1 lysosomal degradation. Created in BioRender. Xixi, H. (2026) https://BioRender.com/emkrrvh. c Confocal microscopy images of HCT116 cells expressing LAMP1-ZsGreen treated with DiI-labeled KRASTAC-Env for 1 h. Scale bar, 5 μm. n = 3 independent experiments. d–f Dose escalation experiment showing PD-L1 degradation in (d,e) HCT116 cells, n = 3 independent experiments, data are presented as mean ± s.d. and (f) NCI-H460 cells after 12 h of treatment with KRASTAC-Env. g Time-course experiments showing PD-L1 degradation in HCT116 cells treated with KRASTAC-Env. h Confocal microscopy images of HCT116 cells expressing PD-L1-EGFP after treatment with a 10 nM equivalent antibody dose of Env antibody, LNP-Env and KRASTAC for 24 h. Scale bar, 5 μm. i–m PD-L1 levels in human KRAS-MT (n = 3 independent cell-lines) and -WT cancer cells (n = 3 independent cell-lines) from (d,f,i–l) treated with KRASTAC-Env for 12 h. n PD-L1 levels in NCI-H226 KRAS^G12D OE and vector cells treated with KRASTAC-Env for 12 h. Data are representative of three independent experiments (g–l). Source data are provided as a Source Data file.

Potent and oncogenic KRAS mutation-specific degradation of PD-L1 in vitro

Having established the general preparation method, we next synthesized KRASTAC conjugated with the anti-PD-L1 antibody envafolimab (KRASTAC-Env). DiI-labeled KRASTAC-Env was added to HCT116 cells expressing the LAMP1-zsGreen. Confocal microscopy revealed that KRASTAC-Env exhibits an identical, highly efficient lysosomal colocalization profile to that of unmodified KRASTAC (Fig. 3b, c and Supplementary Fig. 3), indicating lysosomal targeting remained unaffected by antibody coupling.

To test the degradation efficacy of KRASTAC-Env, KRAS-MT colorectal cancer cell line HCT116 were incubated with KRASTAC-Env or control LNP-Env at varying concentrations. Western blot analysis demonstrated that KRASTAC-Env induced significant concentration-dependent PD-L1 degradation, with nearly 70% target protein reduction at 2.5 nM antibody concentration (Fig. 3d, e and Supplementary Fig. 4a, b). In contrast, LNP-Env induced significantly weaker PD-L1 degradation than KRASTAC-Env. In contrast, neither envafolimab antibody alone nor non-targeting antibody-conjugated KRASTAC-IgG induced detectable PD-L1 degradation (Supplementary Fig. 4c, d). Similar concentration-dependent PD-L1 degradation was observed in KRAS-MT NCI-H460 lung cancer cells (Fig. 3f and Supplementary Fig. 4e–g). Time-course experiments further revealed that near 70% PD-L1 degradation occurred as early as 8 h post-treatment (Fig. 3g). Consistent with these findings, confocal imaging of HCT116 cells stably expressing PD-L1-EGFP demonstrated that KRASTAC effectively triggered the internalization of PD-L1 into intracellular vesicles, leading to a profound reduction in membrane fluorescence. In contrast, treatment with the Env antibody or LNP-Env failed to induce significant PD-L1 degradation (Fig. 3h). Furthermore, to demonstrate the real-time degradation kinetics, we established an HCT116 cell model stably co-expressing PD-L1-EGFP and LAMP1-mCherry. Time-course confocal imaging demonstrated that KRASTAC-Env rapidly mediated PD-L1 into lysosomes within 30 min. This was followed by a substantial reduction in the membrane PD-L1 signal at 6 h, leading to significant clearance of membrane fluorescence by 12 h. In contrast, the LNP-Env control exhibited delayed internalization, appearing only around 6 h, and only induced minimal PD-L1 clearance even at 12 h (Supplementary Fig. 5).

Having validated the rapid and potent protein degradation capacity of KRASTAC-Env, we next assessed its KRAS mutation-specificity of protein degradation. Beyond HCT116 and NCI-H460 cells, PD-L1 levels were evaluated in KRAS-MT MIA PaCa-2 pancreatic cancer cells and KRAS-WT COLO320 colorectal cancer cells, as well as NCI-H661 and NCI-H226 lung cancer cells, across the same concentration range. KRASTAC-Env consistently induced robust PD-L1 degradation in KRAS-MT cancer cells but showed little effect in KRAS-WT cells (Fig. 3i–m). To further validate mutation-specific degradation, we engineered the human KRAS-WT cell line NCI-H226 to stably express either an empty vector or the KRAS^G12D mutation. We found that KRASTAC-Env induced substantial PD-L1 degradation in KRAS^G12D NCI-H226 cells but had minimal impact in vector cells (Fig. 3n). In addition, we compared the KRASTAC-mediated PD-L1 degradation in Kras^G12D MC38 cells and parental WT cells. Since envafolimab targets only human PD-L1, we substituted it with avelumab (targeting both human and murine PD-L1) to generate KRASTAC-Ave^28,29. Similarly, KRASTAC-Ave induced substantial PD-L1 degradation in Kras^G12D MC38 cells, with minimal degradation observed in wild-type cells (Supplementary Fig. 4h).

These results demonstrate that the KRASTAC platform achieves rapid and efficient PD-L1 degradation while maintaining strict specificity for KRAS-MT cells.

Expansion of KRASTAC-addressable targets

Following validation of PD-L1 degradation by KRASTAC-Env, we further applied KRASTAC-Ctx to degrade EGFR in KRAS-MT tumor cells. Upon adding DiI-labeled KRASTAC-Ctx to HCT116 cells expressing the LAMP1-zsGreen fusion protein, confocal microscopy confirmed punctate co-localization between KRASTAC and lysosomes (Fig. 4a, b and Supplementary Fig. 3), confirming the lysosomal targeting capability of KRASTAC-Ctx.

Fig. 4. KRASTAC mediated rapid and oncogenic KRAS mutation-specific degradation of EGFR.

a Schematic of KRASTAC-mediated EGFR lysosomal degradation. Created in BioRender. Xixi, H. (2026) https://BioRender.com/emkrrvh. b Confocal microscopy images of HCT116 cells expressing LAMP1-ZsGreen treated with DiI-labeled KRASTAC-Ctx for 1 h. Scale bar, 5 μm. c,d Dose escalation experiment showing EGFR degradation in HCT116 cells after 12 h of treatment with KRASTAC-Ctx. n = 3 independent experiments. e,f Confocal imaging and quantification of EGFR degradation in HCT116 cells after treatment with a 10 nM equivalent antibody dose in KRASTAC for 24 h. n = 10 independent experiments. Scale bar, 5 μm. g,h Dose escalation experiment showing EGFR degradation in NCI-H460 cells after 12 h of treatment with KRASTAC-Ctx. n = 3 independent experiments. i–k EGFR levels in multiple human KRAS-MT and -WT cancer or normal cells treated with KRASTAC-Ctx for 12 h. Data are presented as mean ± s.d., p values were determined by one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test. Data are representative of three independent experiments (i–k). Source data are provided as a Source Data file.

KRAS-MT HCT116 colorectal cancer cells were incubated with KRASTAC-Ctx or control LNP-Ctx at varying concentrations for 12 h. Western blot analysis revealed that KRASTAC-Ctx induced significant concentration-dependent EGFR degradation, with over 70% reduction achieved at 2.5 nM. In contrast, LNP-Ctx failed to induce substantial EGFR degradation (Fig. 4c, d and Supplementary Fig. 6a, b). KRASTAC-Ctx also exhibited a time-dependent degradation pattern, achieving >70% EGFR reduction within 9 h (Supplementary Fig. 6c). This rapid and potent degradation outperforms existing lysosome-targeting platforms: for example, GalNac-LYTAC¹² requires 10 nM treatment for 48 h, and NanoSphere-CTX¹⁰ needs 50 nM for 24 h to attain comparable ( > 70%) EGFR degradation. Corroborating our earlier observations, neither cetuximab alone nor non-targeting antibody-conjugated KRASTAC-IgG induced detectable EGFR degradation (Supplementary Fig. 6d, e). Immunofluorescence analysis also confirmed that 24 h KRASTAC-Ctx treatment significantly reduced EGFR levels compared to PBS or cetuximab alone, whereas LNP-Ctx induced only minimal EGFR degradation (Fig. 4e, f).

Consistent with the mutation-dependent specificity that was shown above, KRASTAC-Ctx induced concentration-dependent EGFR degradation in KRAS-MT NCI-H460 lung cancer cells (Fig. 4g, h and Supplementary Fig. 6f, g). In contrast, KRASTAC-Ctx showed minimal effect on EGFR levels in KRAS-WT SW48 colorectal cancer cells or normal colonic epithelial cell lines NCM460 and CCD 841 conn (Fig. 4i–k).

Furthermore, to verify that this efficient TPD is driven by APOE rather than a non-specific protein corona effect, we evaluated LNP-Ctx formulations coated with alternative proteins, namely APOC2 (another common apolipoprotein) and albumin (ALB, the most abundant plasma protein). Unlike the robust EGFR clearance achieved by APOE-coated KRASTAC, treatment with either APOC2-LNP-Ctx or ALB-LNP-Ctx induced only minimal degradation in HCT116 cells, exhibiting efficacies identical to that of the LNP-Ctx control (Supplementary Fig. 6h, i). These data confirm that the targeted degradation of KRASTAC is specific to APOE.

To further validate the platform’s modular versatility, we engineered KRASTAC using panitumumab (KRASTAC-Pani), an alternative anti-EGFR antibody with distinct epitope specificity from cetuximab. KRASTAC-Pani recapitulated the robust degradation phenotype, achieving over 70% reduction at 2.5 nM. In contrast, LNP-Pani resulted in only minimal degradation (Supplementary Fig. 6j). Moreover, we conjugated trastuzumab, an anti-HER2 antibody, to generate KRASTAC-Tras. Similarly, KRASTAC-Tras showed rapid lysosomal localization (Supplementary Fig. 7a) and promoted concentration-dependent HER2 degradation, with over 75% reduction observed at 0.5 nM. By comparison, LNP-Tras induced only moderate degradation under the same conditions (Supplementary Fig. 7b).

We further extended this approach to target extracellular proteins by conjugating bevacizumab, an anti-vascular endothelial growth factor (VEGF) antibody, to produce KRASTAC-Bev. Upon adding DiD-labeled KRASTAC-Bev and tetramethyl rhodamine (TMR)-labeled VEGF to HCT116 cells, confocal microscopy showed punctate co-localization between KRASTAC, VEGF, and lysosome marker LAMP1, indicating effective uptake and lysosomal delivery. A similar pattern was observed with KRASTAC conjugated to siltuximab, an anti-IL-6 antibody (Supplementary Fig. 8a–d). These results indicate that beyond membrane proteins, KRASTAC facilitates the cellular internalization and lysosomal localization of extracellular proteins in KRAS-MT cancer cells.

Mechanism of oncogenic KRAS mutation-specific degradation by KRASTAC

To confirm the specific internalization route of KRASTAC, we first evaluated critical endocytic pathways using specific inhibitors and siRNAs. Recent studies indicate that macropinocytosis, clathrin-mediated endocytosis (CME), and caveolae-mediated endocytosis (CVME) are common pathways for LNP uptake in tumor cells (Supplementary Fig. 9a). Given our prior finding that APOE-bound LNPs are selectively taken up by KRAS-MT tumor cells via macropinocytosis²⁵, we tested whether KRASTAC internalization followed this mechanism. HCT116 cells were pretreated with the macropinocytosis inhibitor EIPA for 2 h before KRASTAC-Ctx administration. Western blot demonstrated that EIPA significantly reversed KRASTAC-Ctx-induced EGFR degradation (Supplementary Fig. 9b). In contrast, neither CME inhibitors (chlorpromazine and Pitstop2) nor siRNA-mediated knockdown of CME-related CLTC or CVME-related CAV1 attenuated EGFR degradation (Supplementary Fig. 9c–f). These results confirm that KRASTAC internalization and following protein degradation depend on macropinocytosis.

To further identify the specific genes governing KRASTAC macropinocytosis, we performed an unbiased genome-wide CRISPR-Cas9 screen. HCT116 cells carrying the human GeCKO sgRNA library³⁰ were incubated with DiI-labeled KRASTAC. Cells exhibiting high and low KRASTAC uptake were isolated via flow cytometry cell sorting and subjected to next-generation sequencing (Fig. 5a). By applying the robust rank aggregation (RRA) algorithm, we found that sgRNAs targeting SLC35B2 and LDLR were the most significantly enriched in the low-uptake population (negative score), indicating their indispensable role in KRASTAC internalization (Fig. 5b). KEGG pathway analysis of the negative hits identified “Glycosaminoglycan biosynthesis” as the most significantly enriched pathway (Fig. 5c and Supplementary Fig. 10a–c). Moreover, GO term “Heparan sulfate proteoglycan biosynthetic process” and the KEGG term “Glycosaminoglycan biosynthesis – heparan sulfate” were also enriched in our GSEA analysis (Supplementary Fig. 10d, e). Biologically, heparan sulfate represents a major class of sulfated glycosaminoglycans (GAGs). These heparan sulfate sugar chains attach to transmembrane core proteins, such as syndecan-1 (SDC1) to form cell-surface heparan sulfate proteoglycans (HSPGs). Given that SLC35B2 encodes a Golgi-resident transporter essential for the sulfation of these GAGs, these screening results strongly suggested that LDLR and cell-surface HSPGs work in concert to mediate KRASTAC entry.

Fig. 5. Elucidating mechanism of KRAS-MT cell-specific degradation by KRASTAC.

a Schematic of the genome-wide CRISPR-Cas9 screen to identify essential genes for KRASTAC internalization. KRAS-MT HCT116 cells were transduced with the human GeCKO sgRNA library, incubated with DiI-labeled KRASTAC, and sorted by flow cytometry into high- and low-uptake populations for next-generation sequencing. Created in BioRender. Xixi, H. (2026) https://BioRender.com/w086vwg. b Volcano plot analysis of the CRISPR screen. c KEGG pathway enrichment analysis of the top negative hits from the screen. n = 2 independent experiments. p values were calculated using the MAGeCK algorithm. d–g Confocal microscopy images and corresponding cellular uptake quantification demonstrating significantly reduced internalization of DiI-labeled KRASTAC in HCT116 cells following (d,e) genetic knockout or (f,g) pharmacological inhibition of the identified targets. Scale bar, 5 μm. n = 5 independent experiments. (h,i) Western blot analysis evaluating target protein degradation in LDLR and SLC35B2 knockout HCT116 cells using two independent sgRNAs for each target. j,k Western blot analysis demonstrating the blockade of KRASTAC-mediated targeted degradation following treatment with specific pharmacological inhibitors. Data are presented as mean ± s.d, p values were determined by one-way analysis of variance (ANOVA) with Dunnett’s multiple comparisons test. Data are representative of three independent experiments (h–k). Source data are provided as a Source Data file.

To experimentally validate these screening results, we established LDLR and SLC35B2 knockout cell lines using CRISPR-Cas9. Notably, due to the poor specificity of currently available commercial antibodies for SLC35B2, successful genetic disruption was rigorously verified via targeted PCR amplification and subsequent Nanopore sequencing of the sgRNA-targeted genomic loci, which confirmed premature translation termination resulting from base insertions (Supplementary Fig. 11). Subsequent confocal microscopy confirmed that the disruption of these targets profoundly reduced the cellular uptake of DiI-labeled KRASTAC (Fig. 5d–g). Moreover, by pre-treating cells with EIPA to block macropinocytosis, we found that KRASTAC surface attachment was markedly abrogated in both LDLR and SLC35B2 knockout cells (Supplementary Fig. 12). Furthermore, Western blot analysis demonstrated that knockout of LDLR or SLC35B2, as well as pharmacological inhibition, prevented KRASTAC-mediated target protein degradation (Fig. 5h–k). Notably, while scavenger receptor class B type I (SR-BI) has also been reported as a critical receptor for lipoprotein recognition and LNP internalization (Supplementary Fig. 13a), siRNA-mediated knockdown of SR-BI in HCT116 cells did not alter KRASTAC-Ctx-induced EGFR degradation (Supplementary Fig. 13b). Together, these results demonstrate that the efficient cellular uptake of KRASTAC is dependent on both LDLR and cell-surface HSPGs.

Building upon this mechanistic foundation, our bioinformatic analysis revealed that both LDLR and SLC35B2 expression are significantly upregulated in KRAS-MT tumors compared to WT tumors (Supplementary Fig. 14a), providing the necessary receptor profile for robust KRASTAC uptake. Importantly, this coordinated high expression is also absent in normal tissues. For instance, while normal liver hepatocytes express LDLR for physiological cholesterol clearance, they exhibit significantly lower SLC35B2 expression compared to KRAS-MT tumors (Supplementary Fig. 14b–d). Thus, we reasoned that the simultaneous high expression of both LDLR and SLC35B2 is a feature of KRAS-MT tumors and potentially supports the mutation-specific degradation and minimal off-target effects of KRASTAC observed in our study.

We next investigated the degradation pathway of KRASTAC. As protein degradation primarily occurs via lysosomes or proteasomes, HCT116 cells were pretreated for 2 h with lysosome inhibitors chloroquine (CQ) and bafilomycin A1 (BafA1) or the proteasome inhibitor bortezomib (Bort), followed by KRASTAC-Ctx treatment. Western blot showed that only lysosomal inhibitors markedly reversed EGFR degradation (Supplementary Fig. 15a, b), whereas bortezomib had no effect (Supplementary Fig. 15c, d). This confirms that KRASTAC-mediated protein degradation is lysosome-dependent.

In vivo oncogenic KRAS mutation-specific EGFR Degradation by KRASTAC-Ctx

As KRASTAC-Ctx showed oncogenic KRAS mutation-specific degradation efficacy in vitro, we proceeded to assess its performance in vivo using tumor-bearing mouse models. PBS, DiR-labeled KRASTAC-Ctx, or LNP-Ctx were intravenously administered via the tail vein to HCT116 subcutaneous tumor-bearing mice. Biodistribution was monitored using an IVIS imaging system (Fig. 6a). KRASTAC-Ctx exhibited significantly higher DiR fluorescence intensity in tumor compared to LNP-Ctx (Fig. 6b), while showing similar fluorescence intensity in the liver with LNP-Ctx (Supplementary Fig. 16a, b). In addition, western blot and immunohistochemistry (IHC) demonstrated markedly reduced EGFR levels in KRASTAC-Ctx-treated tumors versus PBS controls, whereas LNP-Ctx caused only marginal EGFR degradation (Fig. 6c, d).

Fig. 6. KRASTAC-Ctx specifically accumulate in KRAS-MT tumor and degrade EGFR in vivo.

a Schematic illustration of evaluating the biodistribution of KRASTAC-Ctx and LNP-Ctx in mice carrying KRAS-MT HCT116 xenograft (5 mg/kg equivalent antibody dose). b Fluorescence imaging and quantification of KRASTAC-Ctx and LNP-Ctx accumulation in HCT116 xenografts from (a). n = 5 mice. c Western blot analysis of EGFR level in HCT116 xenografts from (a). d IHC analysis of EGFR level in HCT116 xenografts from (a). Scale bar, 20 μm. n = 5 mice. e Schematic illustration of evaluating the biodistribution of KRASTAC-Ctx in KRAS-MT and -WT xenografts. f Fluorescence imaging and quantification of KRASTAC-Ctx in KRAS-MT and -WT xenografts. n = 5 mice. g,h western blot and (i–k) IHC analysis of EGFR level in NCI-H460 xenografts and NCI-H226 xenografts. Scale bar, 30 μm. n = 5 mice. Data are presented as mean ± s.d. p values were determined by (b,f,k) unpaired two-tailed Student’s ttest and (d) one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test. Source data are provided as a Source Data file. (a,e) were created in BioRender. Xixi, H. (2026) https://BioRender.com/cmom43s.

To assess oncogenic KRAS mutation specificity in vivo, PBS or DiR-labeled KRASTAC-Ctx was intravenously injected into mice bearing KRAS-MT NCI-H460 or KRAS-WT NCI-H226 subcutaneous tumors (Fig. 6e). IVIS imaging revealed significantly stronger DiR signal in KRAS-MT tumors than in WT counterparts (Fig. 6f). Both western blot and IHC showed robust EGFR degradation in KRAS-MT NCI-H460 tumors treated with KRASTAC-Ctx but minimal EGFR reduction in KRAS-WT NCI-H226 tumors (Fig. 6g–k). To further validate this specificity, we established an isogenic dual-tumor implantation model within the same animal. Specifically, NCI-H226 cells stably expressing either an empty vector or the KRAS^G12D mutation were subcutaneously inoculated into the opposite flanks of the same nude mice (Supplementary Fig. 17a). Consistent with our previous observations, following intravenous administration of KRASTAC, fluorescence imaging at 24 h demonstrated a higher drug accumulation within the NCI-H226 KRAS^G12D tumors than the contralateral NCI-H226 vector tumors (Supplementary Fig. 17b).

Having validated the mutation-selective accumulation of KRASTAC, we next sought to confirm that this enhanced tumor uptake is driven by macropinocytosis. To this end, we designed an in vivo macropinocytosis inhibition experiment using a bilateral HCT116 tumor model. Prior to systemic intravenous administration of KRASTAC, one flank tumor was pre-treated with the specific macropinocytosis inhibitor EIPA via intratumoral injection, while the contralateral tumor received a vehicle (Supplementary Fig. 18a). Fluorescence imaging revealed that EIPA treatment significantly attenuated KRASTAC accumulation at both early (6 h) and late (24 h) phases (Supplementary Fig. 18b, c). These data demonstrated that macropinocytosis is the dominant driver mediating the specific tumor accumulation of KRASTAC.

These findings confirm that the KRASTAC platform enables efficient and oncogenic KRAS mutation-specific tumor accumulation and protein degradation via macropinocytosis in vivo.

KRASTAC-Ave exerts oncogenic Kras mutation-specific anti-tumor effects in vivo

Based on the demonstrated superiority of KRASTAC in achieving oncogenic KRAS mutation-specific tumor accumulation and efficient degradation of target proteins in vivo, we next evaluated the anti-tumor efficacy of KRASTAC-Ave in a syngeneic mouse model bearing Kras^G12D MC38 tumor. Mice were randomized into four groups and injected with PBS, Ave antibody, LNP-Ave, and KRASTAC-Ave via tail vein (Fig. 7a). Notably, although the antibody dose employed in our study (2 mg/kg) was significantly lower than the common dose (10-20 mg/kg) in similar research^29,31–33, the KRASTAC-Ave treatment elicited a potent anti-tumor response. By contrast, both Ave antibody and LNP-Ave showed minimal tumor inhibition (Fig. 7b, c). In addition, such treatments did not reduce mouse body weight (Fig. 7d). Moreover, KRASTAC-Ave treatment induced sustained tumor suppression, which resulted in the long-term survival of five out of six mice beyond the 60-day study endpoint (Fig. 7e–h). Furthermore, we assessed therapeutic efficacy on larger, established tumors. Treatments were initiated when the Kras^G12D MC38 tumor volume reached ~400 mm³. While tumors in other groups continued to grow rapidly, only KRASTAC-Ave treatment decreased the tumor sizes below their initial baseline (Supplementary Fig. 19), confirming its capability to reduce the burden of established tumors.

Fig. 7. Selective in vivo anti-tumor efficacy of KRASTAC-Ave against Kras-MT tumors.

a When tumor volume reaching 100 mm³, mice carrying Kras^G12D MC38 tumor were injected with PBS, Ave antibody, KRASTAC-Ave and LNP-Ave at a dose of 2 mg/kg equivalent antibody for three times. b Tumor volume, c mice survival and d body weight were monitored. e–h Tumor growth curve of Kras^G12D MC38 tumor treated with PBS, Ave antibody, LNP-Ave and KRASTAC-Ave. n = 5 mice for all groups except n = 6 mice for KRASTAC-Ave group. i Mice carrying WT MC38 tumor were injected with PBS, Ave antibody, KRASTAC-Ave and LNP-Ave at a dose of 2 mg/kg equivalent antibody for three times. j Tumor volume, k tumor weight and l body weight were monitored. n = 6 mice. Data are presented as mean ± s.e.m. p values were determined by one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test and log-rank test. Source data are provided as a Source Data file. (a,i) were created in BioRender. Xixi, H. (2026) https://BioRender.com/cmom43s.

To demonstrate that the therapeutic efficacy of KRASTAC-Ave is dependent on the KRAS mutation status, we conducted a parallel study in mice bearing WT MC38 tumors using an identical treatment regimen (Fig. 7i). In this model, KRASTAC-Ave failed to inhibit tumor growth, with tumor volume (Fig. 7j) and final tumor weight (Fig. 7k) showing no significant difference from other groups. This stark contrast in anti-tumor efficacy between the KRAS-MT and KRAS-WT models underscores the mutation-selective action of our platform.

Furthermore, the tolerability of the treatment was confirmed by the absence of significant body weight loss (Fig. 7l) and the subsequent histopathological analysis of major organs (Supplementary Fig. 20), which showed no treatment-related lesions.

KRASTAC-Ave remodels the immune microenvironment specifically in Kras-MT tumors

Building upon the therapeutic efficacy observed in vivo, we next tested whether KRASTAC-Ave could remodel the tumor immune microenvironment in a KRAS mutation-dependent manner. Mice bearing MC38 Kras^G12D tumors were injected twice with PBS, Ave antibody, LNP-Ave, and KRASTAC-Ave (2 mg/kg antibody equivalent) and were sacrificed on day 14 for immune profiling by flow cytometry. KRASTAC-Ave treatment significantly enhanced the infiltration of CD8⁺ T cells, increasing their proportion from 3.93% (PBS group) to 11.12%, whereas Ave antibody and LNP-Ave only induced modest increases to 7.25% and 6.98%, respectively (Fig. 8a, b). Further, KRASTAC-Ave treatment significantly enhanced the activation and effector function of CD8⁺ T cells, as evidenced by the increased expression of CD69 and the ratio of IFN- γ⁺ TNF-α⁺ CD8⁺ T cells (Fig. 8c-f). Recent studies have unveiled the heterogeneity and dynamics of CD8⁺ T cells in cancer, including Ly-108⁺ stem-like T cells and TIM-3⁺ effector-like T cells, and immune checkpoint blockade could induce differentiation of Ly-108⁺ stem-like T cells into TIM-3⁺ effector-like T cells^34,35. Accordingly, we found that KRASTAC-Ave treatment significantly increased the proportion of effector-like T cells and reduced the proportion of stem-like T cells (Fig. 8g–i), suggesting facilitated T cell differentiation toward an effector phenotype. To spatially validate this functional immune activation, we performed multiplex immunofluorescence (mIF) staining. Our results showed that KRASTAC-Ave significantly increased the infiltration of both CD11c⁺ dendritic cells and CD3⁺/CD8⁺/GZMB⁺ cytotoxic T cells in MC38 Kras^G12D tumors (Supplementary Fig. 21).

Fig. 8. KRASTAC-Ave selectively remodels the immune microenvironment in Kras-MT tumors.

Flow cytometry was used to evaluate the immune microenvironment of WT and Kras^G12D MC38 tumors after treatment. Treatments were initiated when the tumor volumes reached 100 mm³ a,b CD8⁺ T cell populations in Kras^G12D MC38 tumors at day 14. c,d IFNγ⁺, TNF⁺ CD8⁺ T cells in Kras^G12D MC38 tumors at day 14. e,f CD69 expression on CD8⁺ T cells in Kras^G12D MC38 tumors at day 14. g–i TIM-3⁺, Ly-108^- and TIM-3^-, Ly-108⁺ CD8⁺ T cells in Kras^G12D MC38 tumors at day 14. n = 6 mice. j,k CD8⁺ T cell populations in WT MC38 tumors at day 14. l CD69 expression on CD8⁺ T cells in WT MC38 tumors at day 14. n = 6 mice. Data are presented as mean ± s.e.m. p values were determined by one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test. Source data are provided as a Source Data file.

To further validate the translational relevance of our platform within a native tissue microenvironment, we evaluated KRASTAC-Ave in the experimental Kras^G12D lung colonization model. Mice were intravenously injected with Kras^G12D Lewis lung carcinoma (LLC) cells and allowed 10 days for tumor lung colonization prior to treatment. Following three systemic treatments, evaluation of the gross lung tumor burden on day 21 revealed that KRASTAC-Ave significantly reduced lung tumor weights, including one complete macroscopic tumor clearance (Supplementary Fig. 22a, b). In contrast, the Ave antibody and LNP-Ave groups exhibited no statistically significant therapeutic benefit compared to the PBS control (Supplementary Fig. 22c). Furthermore, multiplex immunofluorescence analysis revealed that KRASTAC-Ave significantly enhanced the infiltration of CD3⁺/CD8⁺/GZMB⁺ T cells and CD11c⁺ dendritic cells (DCs) within the lung tumors (Supplementary Fig. 23).

To demonstrate that immune remodeling by KRASTAC-Ave is selective for Kras-MT tumors, we performed identical analyses in mice bearing WT MC38 tumors. In contrast to the results in the Kras-MT model, KRASTAC-Ave treatment in WT tumors only slightly enhanced CD8⁺ T cell infiltration but did not elevate activation marker CD69 (Fig. 8j–l). Similarly, mIF staining confirmed that KRASTAC-Ave failed to increase the infiltration of CD3⁺/CD8⁺/GZMB⁺ T cells and CD11c⁺ dendritic cells in the Kras-WT MC38 tumors (Supplementary Fig. 24). The absence of significant immunomodulation in Kras-WT tumors underscores the mutation-specific activity of the KRASTAC platform.

Collectively, these data demonstrate that KRASTAC-Ave reprograms the immunosuppressive tumor microenvironment by enhancing CD8⁺ T cell infiltration, activation, and effector differentiation specifically in Kras-MT tumors, thereby underpinning its mutation-selective antitumor efficacy.

Discussion

In summary, we present KRASTAC as a mutation-specific platform that enables selective protein degradation in KRAS-MT tumors and establish macropinocytosis as a lysosomal-targeting route for TPD. In vitro, KRASTAC induces over 70% degradation of EGFR and PD-L1 at only 2.5 nM antibody within 9 hours in KRAS-MT cancer cells, outperforming existing technologies (e.g., GalNac-LYTAC requires 10 nM/48 h; NanoSphere-CTX needs 50 nM/24 h for comparable EGFR degradation)^10,12. This specificity was further evidenced by the minimal EGFR degradation ( < 10%) observed in KRAS-WT cells under identical conditions. The efficacy extended robustly to in vivo settings, where KRASTAC exhibited significantly enhanced accumulation and over 70% EGFR degradation specifically in KRAS-MT tumors, with negligible effects in WT counterparts. Furthermore, KRASTAC-mediated PD-L1 degradation induced sustained tumor suppression and robustly reversed immunosuppression in syngeneic models.

The selective uptake and lysosomal targeting capability of KRASTAC rely on the distinct endocytic dependencies of KRAS-MT cells. Mechanistically, our genome-wide CRISPR-Cas9 screen and subsequent validations established that this specific macropinocytosis requires the concurrent function of both LDLR and cell-surface HSPGs (synthesized via SLC35B2). It has been revealed that in oncogenic KRAS-driven cancers, cell-surface HSPGs, with SDC1 serving as a critical core protein, drive the macropinocytic scavenging of extracellular proteins for lysosomal degradation³⁶. Therefore, by engaging the LDLR and HSPG, KRASTAC effectively hijacks HSPG-mediated macropinocytosis to drive robust cellular uptake and efficient lysosomal trafficking in KRAS-MT cancers. Furthermore, this mechanism accounts for the low off-target uptake in normal tissues. For instance, while normal hepatocytes express LDLR for cholesterol clearance, they exhibit significantly lower SLC35B2 expression than KRAS-MT tumors. Therefore, the simultaneous high expression of LDLR and SLC35B2 is specific to KRAS-MT tumors, explaining the targeted degradation and minimal hepatotoxicity observed with KRASTAC.

These findings establish KRASTAC as a versatile platform that extends TPD beyond cell type-level targeting toward genetically defined cancer therapy. More importantly, the strategy of exploiting mutation-associated endocytic traits and metabolic profiles could be extended to other common oncogenic mutations. This would enable the development of genetic context-specific degraders for a broader range of cancers.

Beyond serving as a standalone degrader, the versatile LNP-based formulation of KRASTAC offers opportunities for synergistic combination therapies³⁷. For instance, co-delivering KRAS^G12C inhibitors alongside the degradation of compensatory resistance mediators (e.g., EGFR and HER2) could effectively disrupt feedback survival pathways in KRAS-MT cancers. Furthermore, recent advances in lipid engineering, particularly through high-throughput screening platforms, have enabled the organ-selective delivery of LNP^38,39. Building on this progress, we reason that combinatorial lipid screening and artificial intelligence (AI)-driven design hold significant potential to discover lipid formulations that achieve mutation selectivity and enhanced lysosomal tropism^40,41. These advances would enable more precise targeting of genetically defined cancers while enhancing lysosomal delivery, thereby improving the efficacy and specificity of protein degradation therapies. Finally, our strategy of exploiting oncogene-associated metabolic and endocytic vulnerabilities extends TPD beyond conventional cell-type targeting, providing an example for developing genetically defined degradation therapies across a broader range of human cancers.

Methods

Ethics statement

All animal experiments were approved by the Animal Experimentation Ethics Committee of Second Affiliated Hospital of Zhejiang University (No. 2024-325). Female Balb/c nude mice and C57BL/6 mice (6-8 weeks old) were purchased from Gempharmatech (Jiangsu, China). The animals were housed in a specific pathogen-free animal facility under 12-h light–dark cycles with ambient temperature between 21 °C and 24 °C and humidity between 40 and 70%. Institutional Review Board approved a maximal allowable tumor burden of 1500 mm³ for this study. Tumors were measured every 2 days, we confirm that once the tumor achieved this maximal tumor volume, mice were humanely euthanized. All endpoints were implemented in accordance with the approved ethics protocol. Only female mice were used to ensure consistency across experimental groups. No sex-based analyses were performed.

Preparation of KRASTAC

Biotin and azide-modified LNPs were prepared by the microfluidic system. The organic phase was prepared by dissolving 10 mg 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC, 850345, Avanti) and 0.5 mg 1,2-distearoyl-sn-glycero-3-phosphorylethanolamine (DSPE)-PEG₁₀₀₀-biotin or DSPE-PEG₁₀₀₀-N3 (R-0040-1K/R-0045-1K, Ruixibio), with or without 0.1 mg DiD or DiR (D7757/D12731, Thermo Scientific) in 1 mL ethanol. Then the organic and aqueous phases (PBS) were rapidly mixed using the microfluidic chip at defined flow rate ratios (FRR) 1:6 and a total flow rate (TFR) 0.8 mL min^-1. Obtained LNPs were dialyzed against PBS in 3500 Da MWCO dialysis tubes (PURX35015, Sigma-Aldrich). Biotin-modified LNPs were incubated with human (350-02, PeproTech) or mouse (CJ05, Novoprotein) APOE recombinant protein at a 5:1 mass ratio overnight at 37 °C to prepare KRASTAC-biotin.

To prepare DBCO-modified antibodies, cetuximab, panitumumab, avelumab, envafolimab, trastuzumab, bevacizumab, siltuximab (HY-P9905, HY-P99041A, HY-108730, HY-P99115, HY-P9907, HY-P9906, HY-P9956, MedChemExpress) were incubated with DBCO-PEG₄-NHS ester (R-CD-004, Ruixibio) at a 1:5 molar ratio for 2 h at room temperature. The unreacted DBCO-PEG₄-NHS ester was removed through 50 or 30 kDa MWCO ultracentrifugal filters (UFC5050/UFC5030, Sigma-Aldrich). Obtained DBCO-modified antibodies were incubated with azide-modified LNPs at a 1:5 molar ratio at 37 °C overnight, followed by dialysis against PBS through 300 kDa MWCO dialysis bags (131456, Repligen). The antibody conjugation amount was determined by Bradford assay (23246, Thermo Scientific). The antibody-conjugated LNPs were incubated with human or mouse APOE recombinant protein at a 5:1 mass ratio overnight at 37 °C to produce antibody-conjugated KRASTAC.

Characterization and quantification of APOE adsorption

To evaluate the controllability of APOE adsorption on the nanoparticle surface, 300 μg LNP-Ctx was mixed with gradient amounts of recombinant APOE protein. The mixture was incubated at 37 °C overnight. To separate the APOE-bound LNP-Ctx from free APOE, 300 μL of Protein A Magnetic Beads (Cat# 88846, Thermo Scientific) were washed with PBS and added to the mixture. The suspension was then incubated at 4 °C overnight with rotation. Subsequently, the bead-LNP complexes were isolated using a magnetic stand, and the supernatant containing unbound APOE was discarded. The beads were washed three times with PBS. Finally, the bead-bound LNP-Ctx were lysed using NP-40 lysis buffer and the extracted APOE was visualized by SDS-PAGE followed by Coomassie blue staining. The absolute amount of adsorbed APOE was quantified using a human APOE ELISA kit (EK1218, Multisciences) according to the manufacturer’s instructions.

Cell lines

HCT116, SW48, COLO320, BxPC-3, NCI-H358, NCI-H226, NCI-H661, MIA PaCa-2, LLC and CCD 841 conn cell lines were obtained from the American Type Culture Collection (ATCC). NCM460, WT and Kras^G12D MC38 cell lines were obtained from Ubigene. Cells were cultured in RPMI 1640 or DMEM medium (11875093/11965092, Gibco) supplemented with 10% (v/v) FBS (10099141 C, Gibco) and 1% (v/v) penicillin-streptomycin (15140122, Gibco). All cell lines were authenticated using short tandem repeat (STR) profiling. Each cell line was maintained separately and stocked in early passages to minimize contamination and to preserve cell identity. Cell lines were routinely tested for mycoplasma contamination.

To enhance PD-L1 expression in tumor cells during in vitro culture, all PD-L1 degradation assays were performed with cells pretreated overnight with 100 ng/mL interferon-gamma (IFN-γ) in the culture medium.

For mRNA knockdown experiments, siRNA transfection was performed using Lipofectamine 3000 (L3000015, Thermo Scientific) according to the manufacturer’s protocol. Following 24 h incubation, knockdown efficiency was assessed by Western blotting and subsequent functional assays were performed. All siRNA sequences (Supplementary Table. S1) were synthesized by Genema Biotech.

To establish stable overexpression cell lines, sequence encoding KRAS^G12D mutant protein was inserted into the pLVX-IRES-Puro vector. The sequence encoding human LAMP1 was inserted into the pLVX-ZsGreen1-N1 and pLVX-mCherry-N1 vectors. Similarly, the sequence encoding human PD-L1 was cloned into the pLVX-EGFP-N1 vector. For specific gene knockouts, sgRNA sequences targeting LDLR and SLC35B2 (Supplementary Table. S2) were cloned into the lentiCRISPR v2 vector.

For lentivirus production, the respective constructed plasmids were co-transfected with lentivirus packaging plasmids into HEK293T cells using Lipofectamine 3000 and viral supernatants were collected 2 days after infection. Cells were infected by addition of viral supernatant and 10 µg mL^-1 polybrene for 24 h followed by puromycin selection.

SDS-PAGE and Coomassie staining

Protein samples were prepared by adding protein loading buffer to the KRASTAC and heated at 95 °C for 5 min. Protein samples were then loaded onto the 10% Tris-Glycine precast mini gels (MA0283, Meilunbio). Protein electrophoresis was performed at 160 V for 60 mins and the gel was subjected to Coomassie stain (E901, Vazyme) according to the standard protocol.

SA uptake experiments

Cells were plated onto a glass-bottom dish (D35-14-1.5-N, Cellvis) and allowed to grow until approximately 50% confluence. Cells were incubated with 30 μg/mL Alexa fluor 647 labeled SA (S21374, Thermo Scientific) and 10 μg/mL KRASTAC-biotin or LNP-biotin for 2 h. Cells were rinsed three times with ice-cold PBS, followed by 4% polyoxymethylene fixation and nuclear staining. Images were captured at 63X magnification using LSM 900 confocal microscope system (Zeiss) and analyzed using Fiji software (v2.1.0).

Cell surface protein degradation experiments

Cells were plated onto 12-well plates and allowed to grow until approximately 50% confluence. Then cells were treated with antibodies, LNP-antibody or KRASTAC-antibody at different concentrations for the indicated time.

For degradation reversion experiments, HCT116 cells were plated onto 12-well plates and subjected to either: (a) siRNA-mediated gene silencing for 24 h or (b) pretreatment with 25 μM EIPA, 5 μM Chlorpromazine, 5 μM Pitstop 2, 1 mg/mL heparin or 1 mg/mL heparan sulfate (HY-115604, HY-12708, HY-101840, HY-17567, HY-101916, MedChemExpress) for 2 h. Then cells were incubated with KRASTAC-Ctx (2.5 nM cetuximab-equivalent concentration) for 12 h. Cells were washed three times and lysed as detailed in ‘Western blot analysis’.

Western blot analysis

Cells or tumor samples (homogenized using a tissue grinder) were lysed in the RIPA buffer (9806, Cell Signaling Technology) supplemented with protease and phosphatase inhibitor cocktails (78440, Thermo Scientific), The protein concentration was measured by BCA protein assay kit (23227, Thermo scientific). resolved in 10% Tris-Glycine precast mini gels or 4–12% NuPAGE Bis-Tris Mini Gels (NP0322, Thermo Scientific) and transferred to PVDF membrane (ISEQ00010, Millipore). Membranes were blocked using 5% non-fat milk for 1 h and then incubated with primary antibodies overnight. After washing three times with TBST, the membranes were incubated with 1:10,000-diluted HRP-conjugated anti-mouse/rabbit IgG (HA1001/1006, Huabio) for 1 h. The following primary antibodies were used: anti-EGFR mAb (4267), anti-human PD-L1 mAb (13684), anti-CLTC mAb (4796) (all purchased from Cell Signaling Technology and were used at 1:1,000 dilution); anti-mouse PD-L1 mAb (ab213480, Abcam, 1:2000); anti-SRBI mAb (HY-P81415), anti-CAV1 mAb (HY-P82037) (all purchased from MedChemExpress and were used at 1:1,000 dilution); anti-LDLR mAb (A20828, Abclonal, 1:2000); anti-RAS (G12D mutant) mAb (MA5-36256, Thermo Scientific, 1:2000); anti-ubiquitin (10201-2-AP, 1:5000 dilution) anti-β-actin mAb (66009-1-Ig, 1:10,000 dilution) and anti-GAPDH mAb (60004-1-Ig, 1:10,000 dilution) (all purchased from Proteintech). Uncropped and unprocessed scans of the Western blot have been presented in the source data file and Supplementary Information.

Immunofluorescence

Cells were plated onto glass-bottom dishes and allowed to grow until approximately 50% confluence. After incubation with free antibody, LNP-antibody and KRASTAC-antibody for 24 h, cells were rinsed three times with ice-cold PBS and were fixed in 4% paraformaldehyde for 15 min. Cells were then permeabilized and blocked with PBS containing 0.2% Triton X-100 and 5% goat serum at room temperature for 1 h. Cells were incubated with anti-EGFR mAb (4267) (Cell Signaling Technology, 1:100 dilution) in PBS containing 0.2% Triton X-100 and 1% BSA at 4 °C overnight. Cells were then rinsed and incubated with Coralite 555-conjugated anti-rabbit IgG (1:500; RGAR003, Proteintech) at room temperature for 1 h. After nuclei staining with DAPI, images were captured at 63X magnification using LSM 900 confocal microscope system (Zeiss) and analyzed using Fiji software (v2.1.0).

Multiplex immunofluorescence was performed according to standard protocol of tyramide signal amplification (TSA) technology. Briefly, formalin-fixed, paraffin-embedded tissues were mounted on charged slides and baked at 60 °C for 1 h, deparaffinized and dehydrated. Slides were then washed with deionized water for 2 min followed by antigen retrieval at 96 °C with 10 mM sodium citrate (pH 6) for 8 min. Once the slides cooled down to room temperature, they were washed three times (2 min per wash) with PBS and endogenous peroxides were quenched with 3% hydrogen peroxide for 15 min. Slides then underwent multiple cycles of staining, with each cycle consisting of primary antibody incubation, horseradish peroxidase (HRP)-conjugated secondary antibody application, and tyramide signal amplification. The primary antibodies used included CD3ε (78588), CD8α (98941), granzyme B (46890), and CD11c (97585) (all purchased from Cell Signaling Technology and used at a 1:500 dilution). For signal development, the following specific TSA fluorophores were sequentially applied: iFluor 430 tyramide (45096), iFluor 488 tyramide (11060), Cy3 tyramide (11065), and Cy5 tyramide (11066) (all purchased from AAT Bioquest and used at a 1:600 dilution). Following the final multiplex cycle, slides were counterstained with DAPI. Whole slides were scanned using a Pannoramic MIDI digital scanner (3D HISTECH). QuPath (0.7.0) was used for image quantification according to the official tutorials.

Immunohistochemistry

Tumors were fixed in 4% paraformaldehyde for 48 h at 4 °C, followed by embedding in paraffin wax. Tissue sections were then cut and mounted on glass slides. Antigen retrieval was performed by boiling the sections in a Tris-EDTA buffer (10 mM Tris base, 1 mM EDTA, 0.05% Tween 20, pH 9.0) for 10 min. After cooling, tissue sections were permeabilized with 0.2% Tween 20 in PBS for 10 min and blocked with 1.5% BSA and 10% goat serum in TBS for 1 h at room temperature. Tissue sections were incubated with anti-EGFR mAb (1:100, 4267, Cell Signaling Technology) at 4 °C overnight in TBS containing 1.5% BSA and 0.1% Tween 20. The sections were then incubated with the secondary antibody and stained with diaminobenzidine (DAB) and hematoxylin. Images were captured by a slide scanner (VS200, Olympus). Quantitative analysis of the immunohistochemical staining was performed using QuPath (version 0.7.0). Briefly, regions of interest (ROIs) were annotated, and the built-in positive cell detection algorithm was applied to determine the DAB optical density and calculate H-score across the tissue sections.

Genome-wide CRISPR-Cas9 screen in HCT116 cells

The human genome-wide CRISPR knockout pooled library (#52957, Addgene) was amplified and packaged into lentiviruses following standard protocols. HCT116 cells were transduced with the lentiviral library at a low multiplicity of infection (MOI) of 0.3 to ensure that the majority of infected cells received only a single single-guide RNA (sgRNA) integration. To maintain library representation and minimize genetic drift, a minimum coverage of 300× was strictly maintained throughout all cell culture, passaging, sorting, and sequencing steps. Following transduction, successfully infected cells were selected using puromycin (2 μg/mL) for 5 days.

For the CRISPR screen based on KRASTAC uptake, the pooled cells were incubated with DiI-labeled KRASTAC-Ctx at a cetuximab-equivalent concentration of 2.5 nM for 12 h at 37 °C. After rinsing three times with ice-cold PBS, the cells were harvested and resuspended in ice-cold FACS buffer. Fluorescence-activated cell sorting (FACS) was performed using a BD FACSAria III flow cytometer. Based on the DiI fluorescence intensity, cells falling into the top 10% and bottom 10% were strictly sorted and collected, representing the high-uptake and low-uptake populations, respectively.

Genomic DNA (gDNA) from both sorted cell populations, as well as an unsorted input cell pool, was extracted using the QIAamp DNA Blood Maxi Kit (51192, Qiagen) according to the manufacturer’s instructions. The integrated sgRNA sequences were amplified from the gDNA via a standard two-step PCR strategy to attach necessary sequencing adapters and sample-specific barcodes. The resulting PCR amplicons were purified, quantified, and subjected to next-generation sequencing (NGS) on an Illumina platform. Finally, the sequencing data were processed and thoroughly analyzed using the MAGeCKFlute (v3.8) bioinformatics package to evaluate the enrichment and depletion of sgRNAs across different populations and to perform downstream pathway enrichment analysis.

Transcriptomic analysis of LDLR and SLC35B2 expression

Normalized transcript data were obtained from the UCSC Xena dataset (https://xenabrowser.net/datapages/). Expression values were analyzed as log₂(TPM + 1). KRAS-MT and KRAS-WT tumor samples from colorectal cancer (CRC), lung adenocarcinoma (LUAD), and pancreatic adenocarcinoma (PAAD) were compared with the normal tissues and normal liver tissue. Pairwise comparisons were performed between the indicated groups.

Flow cytometry

For analysis of surface markers, cells were first stained with Fixable Viability Stain 510 (564406, BD Biosciences) in PBS at room temperature for 15 min and then incubated with TruStain FcX (101319, Biolegend) for 10 min in cell staining buffer (420201, Biolegend) and then stained with the appropriate antibodies on ice for 30 min. For intracellular cytokine detection, cells were stimulated for 4 h with Cell Activation Cocktail (423304, Biolegend) and stained with a fixation/permeabilization kit (554714, BD Biosciences) according to the manufacturer’s instructions. Flow cytometry data were acquired on LSR Fortessa (BD Biosciences) using FACSDiva software (v8, BD Biosciences) and analyzed using FlowJo software (10.8.1, BD Biosciences). The following fluorescent conjugate-labeled antibodies were used: APC/Cyanine7 anti-mouse CD45 (30-F11, 103116, 1:400), PerCP/Cyanine5.5 anti-mouse TCRβ (H57-597, 109227, 1:200), BV650 anti-mouse CD8α (53-6.7, 100741, 1:400), BV421 anti-mouse CD69 (H1.2F3, 104527, 1:400), FITC anti-mouse TNFα (MP6-XT22, 506304, 1:200), PE/Cyanine7 anti-mouse IFNγ (XMG1.2, 505826, 1:200), PE anti-mouse CD366 (TIM-3) (RMT3-23, 119704, 1:400), PE anti-mouse CD274 (10 F.9G2, 124308, 1:200) (all from Biolegend), R718 anti-mouse Ly-108 (13G3, 752059, 1:100, BD Biosciences). The gating strategy used for flow cytometry analysis is detailed in Supplementary Fig. 25.

In vivo accumulation and EGFR degradation of KRASTAC-Ctx

To establish cell-line-derived xenografts in mice, 1 × 10⁶ HCT116 cells, NCI-H226, or 5×10⁵ NCI-H460 cells were mixed at a 1:1 dilution with Matrigel (354234, Corning) in a total volume of 100 μL and injected subcutaneously into the right flank of nude mice, C57BL/6 mice. When tumor size reached ~500 mm³, mice were randomized and assigned to different groups and DiR-labeled KRASTAC-Ctx or LNP-Ctx were injected into mice via tail vein at a dose of 5 mg/kg cetuximab for 3 days. Mice were euthanized 24 h after the final injection and tumor were collected for fluorescent imaging and subsequent Western blot and IHC analysis.

For the establishment of bilateral tumor models, 1×10⁶ HCT116 cells, or an isogenic pair of NCI-H226 empty vector and KRAS^G12D OE cells, were mixed with Matrigel (1:1 ratio, 100 μL) and subcutaneously inoculated into both the left and right flanks of nude mice. To evaluate the role of macropinocytosis in vivo, the HCT116 bilateral model was utilized. Once tumor size achieved ~500 mm³, EIPA was intratumorally injected into the left tumor at a dose of 20 μg per tumor, while the right tumor received a vehicle control. Three hours after EIPA pretreatment, mice were intravenously injected with DiR-labeled KRASTAC at a dose of 5 mg/kg cetuximab via the tail vein. Paired tumors were then harvested at 6 and 24 h post-injection for fluorescence imaging to quantify KRASTAC accumulation. To evaluate mutation-specific accumulation in vivo, mice bearing the NCI-H226 vector/KRAS^G12D OE bilateral tumors were intravenously injected with DiR-labeled KRASTAC via the tail vein, and paired tumors were harvested 24 h later for fluorescence imaging.

In vivo anti-tumor effects of KRASTAC-Ave

For the subcutaneous tumor experiments, 5 × 10⁵ Kras^G12D or WT MC38 cells were mixed at a 1:1 dilution with Matrigel in a total volume of 100 μL and injected subcutaneously into the right flank of the C57BL/6 mice. When tumor size achieved ~100 mm³, mice were randomized and assigned to different groups and PBS, Avelumab, LNP-Ave or KRASTAC-Ave were injected into mice via tail vein at a dose of 2 mg/kg IgG in day 8, 11, 13 after tumor inoculation. In addition, to assess therapeutic efficacy on larger, established tumors, treatments were initiated at the day1, 4, 7 after the Kras^G12D MC38 tumor volume reached ~400 mm³ Tumors were measured every two days with digital calipers and tumor volumes were calculated by the formula: 0.5 × length [mm] × (width [mm])².

For the LLC lung tumor experiments, 5×10⁵ Kras^G12D LLC tumor cells were intravenously injected into mice via tail vein. Mice were allowed 10 days for tumor lung colonization before treatment initiation. The animals then received three treatments (every other day) and were euthanized on day 21 for tumor weight evaluation and mIF.

Statistics and reproducibility

All statistical analyses were performed using GraphPad Prism (10.2.3). Data were shown as mean ± s.e.m. or mean ± s.d. as indicated in each experiment. Comparisons of two groups were assessed using two-tailed Student’s ttest. Comparisons of multiple groups were assessed using one-way ANOVA with Tukey’s or Dunnett’s multiple comparisons test. Survival curves were compared using the log-rank test. p values were calculated using two-tailed tests and p values of less than 0.05 were considered statistically significant. No statistical method was used to predetermine sample size, but our sample sizes are similar to those reported in other publications. No data were excluded from the analyses. Tumor size-matched mice were randomly assigned to control and experimental groups; no other randomization was performed. The investigators were not blinded to allocation during experiments and outcome assessment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

H.Y., Q.X., Q.Z., and K.D. conceived the study. Q.X., Q.Z., and K.D. supervised the project. H.Y., X.Z., J.W., X.R.L., Q.Z., C.G., and K.D. established the methodology. H.Y., J.W., C.Y., Y.J., X.F.L., Q.X., and Q.Z. performed the experiments. H.Y., X.Z., and J.W. contributed to data visualization and figure preparation. H.Y. wrote the original draft. X.Z., Q.X., Q.Z., and K.D. reviewed and edited the manuscript.

Peer review

Peer review information

Nature Communications thanks Green Ahn and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This research was funded by the National Natural Science Foundation of China: 82373449 (to K.D.), 82373234 (to Q.X.), 52473151 (to Q.Z.), 52173141 (to Q.Z.), 52503206 (to J.W.); Noncommunicable Chronic Diseases-National Science and Technology Major Project: 2024ZD0520100 (to K.D.); the Key R&D Program of Zhejiang Province: 2023C03049 (to K.D.), 2025C02065 (to Q.X.); the China Postdoctoral Science Foundation: 2025M770132 (to J.W.), 2025M782261 and GZC20251518 (to H.Y.).

Data availability

The raw sequence data of our CRISPR screen have been deposited in the Genome Sequence Archive in National Genomics Data Center, China National Center for Bioinformation/ Beijing Institute of Genomics, Chinese Academy of Science (GSA-Human: HRA018762) that are accessible at https://ngdc.cncb.ac.cn/gsa-human. All other data supporting this work are available in the main article, supplementary information or source data file. Source data are provided with this paper.

Competing interests

Authors declare that they have no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77697-5.