Panitumumab-Based EGFR Blockade in SMARCB1-Deficient Renal Medullary Carcinoma: Preclinical Basis and Prospective Clinical Activity

Abstract

Purpose:

SMARCB1-deficient renal medullary carcinoma (RMC) is an ultra-rare, lethal malignancy refractory to therapies approved for other renal cell carcinomas, and its rarity makes randomized trials logistically unfeasible. We investigated whether wild-type epidermal growth factor receptor (EGFR) is a therapeutically actionable dependency in RMC and translated this insight into a deployable clinical regimen.

Experimental Design:

EGFR expression was assessed by CLIA-certified immunohistochemistry in RMC tumors and by integrated RNA- and chromatin immunoprecipitation-sequencing of primary tumors and adjacent kidney. Panitumumab was compared with erlotinib in patient- and cell line-derived xenografts, and its mechanism defined by immunoblotting, confocal colocalization, cell-cycle and apoptosis assays, and natural killer (NK) cell co-culture. Panitumumab-based therapy was then evaluated prospectively in 26 patients with RMC across ten centers on two continents.

Results:

RMC showed uniformly high membranous EGFR (median H-score 300), with enhancer and promoter reprogramming at the EGFR locus and a ligand switch from EGF to epiregulin and amphiregulin. EGFR expression was uncoupled from SMARCB1 status in vitro. Panitumumab outperformed erlotinib in RMC xenografts, suppressed AKT and ERK1/2 signaling, and routed EGFR to LAMP1-positive lysosomes in vitro; its effect was predominantly cytostatic, with minimal NK cell–mediated cytotoxicity. Clinically, panitumumab-based therapy achieved a 53.9% objective response rate, including 15.4% complete responses, with median progression-free and overall survival of 5.8 and 9.5 months.

Conclusions:

Wild-type EGFR is a foundational dependency in RMC that is effectively targeted by panitumumab-based therapy. These findings, derived from a non-randomized registry, provide a biological and clinical rationale for further prospective validation.

INTRODUCTION

SMARCB1-deficient renal medullary carcinoma (RMC) is a rare but highly aggressive renal cell carcinoma (RCC) subtype, characterized by loss of the SMARCB1 tumor suppressor and a predilection for young individuals of African descent with sickle hemoglobinopathies such as sickle cell trait (1-3). Most patients with RMC present with metastatic disease at diagnosis necessitating use of upfront systemic therapies (2,3). However, RMC is refractory to the anti-angiogenic agents and immune checkpoint therapies approved for the treatment of other RCCs such as clear cell RCC (ccRCC), and currently available cytotoxic chemotherapy agents produce at best response rates of ~29% and median progression-free survival (PFS) of ≤ 4 months even in the first-line setting (2-9). Beyond first line, no systemic therapy strategy to date has achieved objective responses >20% or median PFS > 3.5 months in patients with RMC (6,10,11). There is clearly a need to develop novel biology-informed treatment strategies for this lethal disease.

We recently showed that the epidermal growth factor receptor (EGFR) may be a therapeutic target in RMC with significant tumor reductions achieved when patient-derived xenograft (PDX) and cell line-derived xenograft (CDX) animal models of RMC were treated with the EGFR tyrosine kinase inhibition (TKI) erlotinib compared to vehicle control (12). Accordingly, repurposing of combination bevacizumab plus erlotinib used for fumarate hydratase (FH)-deficient RCC (4,12,13), yielded partial responses in 2 out of 10 patients with RMC resulting in a median PFS of 3.5 months (10). Subsequent work indicated that, whereas FH-deficient RCC can be sensitive to both erlotinib and bevacizumab, the loss of SMARCB1 in RMC confers resistance to anti-angiogenic agents such as bevacizumab (12,14). Notably, erlotinib and other TKIs targeting the intracellular EGFR kinase domain are effective primarily in tumors such as non-small cell lung cancer harboring activating EGFR kinase domain mutations (15,16). Conversely, wild-type EGFR driven tumors, including colorectal cancer and head and neck squamous cell cancer, respond to extracellular EGFR-blocking monoclonal antibodies such as panitumumab (17-19). Across multiple molecular profiling studies in RMC, recurrent EGFR or RAS-pathway mutations have not been observed, supporting a dependency on wild-type EGFR (1,20-23).

Based on this biological and clinical rationale, we hypothesized that EGFR blockade with panitumumab would outperform erlotinib in RMC. Herein, we tested this hypothesis in preclinical RMC models and then launched a prospective multi-national clinical registry evaluating the efficacy of repurposing panitumumab-based therapy in 26 heavily pretreated patients with RMC. The resulting responses, including durable complete remissions, indicate meaningful clinical activity and suggest that panitumumab-based therapy may represent the most effective systemic approach reported to date in RMC.

METHODS

Prospective Clinical Study Design and Participants

All RMC samples were confirmed to be SMARCB1 negative by IHC using purified mouse anti-BAF47 Clone 25/BAF47 (BD Biosciences, San Jose, CA, USA, RRID:AB_2191717). Sickle cell status was determined by hemoglobin electrophoresis. Clinical data were collected from 26 patients with RMC treated off-label with panitumumab-based therapy at MD Anderson Cancer Center, Memorial Sloan Kettering Cancer Center, Dana Farber Cancer Institute, City of Hope, Stanford Cancer Center, Carbone Cancer Center, Siteman Cancer Center, Charité – Universitätsmedizin Berlin, Juravinski Cancer Centre, and the University of Ottawa. Off-label use of an FDA-approved drug such as panitumumab is a common and legally permissible practice under the physician’s judgment if it is in the patient’s best interest and there is scientific rationale to support its use (24). Panitumumab administration and monitoring were performed in accordance to regulatory guidelines and the reporting of repurposing panitumumab-based therapies in patients with RMC was approved by the institutional review board/independent ethics committee of MDACC (IRB protocols PA11-1045 and PA16-0736) and retrospective data collection was approved by Charité – Universitätsmedizin (EA4/070/24) with all participants providing written informed consent consistent with the Declaration of Helsinki. Baseline demographic and clinical data for each patient were gathered via individual chart reviews from the institution’s electronic medical records. Toxicities were assessed according to the Common Terminology Criteria for Adverse Events (CTCAE v5).

Sex as a biological variable

Findings apply to both sexes. Sex was not considered in study design. Both males and females were eligible for study. Sex was recorded as it was assigned at birth.

Study approval

The study was approved by the MDACC Institutional Review Board (protocols PA11-1045 and PA16-0736). Panitumumab is an FDA-approved therapy for other solid tumors; its administration to individuals with this ultra-rare, life-threatening kidney cancer was conducted under compassionate-use provisions for an orphan indication, in full compliance with institutional and federal regulations.

Quantification and Statistical Analysis

All statistical analysis on preclinical experiments were performed using GraphPad Prism version 10.3.1 (Dotmatics, Boston, MA, USA). For the longitudinal in vivo tumor growth curves in mouse experiments, inter-group comparisons of tumor volume at each treatment day were performed by two-way ANOVA (treatment × time) with Tukey's multiple comparisons test. For single-timepoint endpoint measurements (EGFR H-score, percentage of Ki67 positive cells, percentage of cleaved caspase-3 cells, and percent fibrosis), one-way ANOVA with Dunnett's post-hoc test was used, with the Mann–Whitney test applied to the RMC32X EGFR and Ki67 comparisons owing to the small sample size.

Patient characteristics were described using median, interquartile range, and minimum/maximum values for continuous variables and frequency (%) for categorical variables. Radiographic responses to panitumumab-based therapy were assessed using Response Evaluation Criteria in Solid Tumors (RECIST), version 1.1 (25). A board-certified radiologist (ESR), blinded to patient history and clinical data, assessed radiographic tumor response in 18 cases with available imaging at MD Anderson. Progression free survival (PFS) was defined as the time interval between the date of first panitumumab dose and the date of disease progression or death from any cause, whichever occurred first. Patients who were alive without disease progression at the time of last follow-up were censored at the time of last follow-up. Overall survival (OS) was defined as the time interval between the date of first panitumumab dose and the date of death due to any cause. Patients that were alive were censored at the last follow-up. Survival analyses were performed using the Kaplan-Meier method to estimate median survival time. The 95% confidence intervals (CIs) for the median survival time were calculated using the log-log transformation method, implemented in R (version 4.5.1) with the survival (version 3.8-3) package’s survfit function. The log-log approach was selected to account for the skewed nature of survival distributions and to ensure robust variance estimation, particularly in the presence of censoring. Median follow-up time was estimated by the reverse Kaplan-Meier estimator (26).

Tissue Immunohistochemistry (IHC)

Clinical grade institutional laboratory tests certified by the Clinical Laboratory Improvement Amendments (CLIA) were utilized for IHC in human RMC and epithelioid sarcoma tissues using mouse monoclonal antibody against EGFR (1:20, clone EGFR.113, Leica Biosystems catalog number NCL-L-EGFR, RRID:AB_563696). Positive EGFR staining was defined as membranous staining in tumor cells. Quantification was performed using the H-score method, calculated according to the formula: H-score = (percentage of cells with weak staining × 1) + (percentage of cells with moderate staining × 2) + (percentage of cells with strong staining × 3). The final score ranges from 0 to 300.

For preclinical tumor tissues, formalin-fixed-paraffin embedded (FFPE) tissue blocks were generated at the MD Anderson Cancer Center DVMS Veterinarian Pathology Services. All blocks were used for H&E, Ki67 staining (MIB-1 mouse monoclonal antibody from Dako (Glostrup, Denmark, RRID:AB_2631211) applied at 1:100 dilution for 15 minutes), cleaved caspase-3 (CP229 rabbit polyclonal antibody from BioCare, Pacheco, California, United States, RRID:AB_2737391) applied at 1:100 dilution for 30 minutes), and cleaved PARP (5625 rabbit monoclonal antibody from Cell Signaling, RRID:AB_10699459, applied at 1:50 dilution for 20 minutes) in DVMS Veterinarian Pathology and EGFR staining was performed at the CLIA-certified clinical pathology core (1:20, clone EGFR.113, Leica Biosystems catalog number NCL-L-EGFR, RRID:AB_563696). Slides were digitally scanned by an Aperio AT2 whole slide digital scanner (Leica Biosystems, Vista, CA, USA) at 20x resolution and viewed with Image Scope v.12.4.6 (RRID:SCR_020993). EGFR and Ki-67 expression were quantitatively assessed using HALO software (version 3.6; Indica Labs, Albuquerque, NM, USA, RRID:SCR_018350). A tissue classifier was trained to distinguish malignant cells from stromal tissue. EGFR membranous staining in malignant cells was categorized as weak, moderate, or strong according to staining intensity. EGFR expression was quantified using H-score method by HALO algorithm (Indica Labs - Membrane v1.7). Ki-67 expression was evaluated using HALO algorithm (Indica Labs – CytoNuclear v2.0.9, RRID:SCR_018350) and expressed as the percentage of malignant cells with positive nuclear staining among all malignant cells. For cleaved caspase-3 and cleaved PARP, a pathologist (X.C.) performed a manual semi-quantitative visual assessment of the percentage of positively stained tumor cells across the entire tumor area. For cleaved-PARP, cases showing extremely low levels of staining were recorded as <1%. These values were considered below the reliable quantification range for semi-quantitative assessment and were reported descriptively.

RNA and Chromatin Immunoprecipitation Sequencing

RNA sequencing (RNA-seq) was previously performed on untreated primary RMC tumors (n=11) and adjacent normal kidney samples (n=6) (1) and can be accessed at the NCBI Sequence Read Archive (SRA; RRID:SCR_004891) hosted by the NIH (SRA accession: PRJNA605003). Gene-level counts were modeled with a negative binomial generalized linear model, and contrasts between conditions (RMC tumor versus adjacent normal kidney) were evaluated with Wald tests. Multiple testing was controlled using the Benjamini–Hochberg false discovery rate (FDR). Analyses were performed with DESeq2 (RRID:SCR_000154) (27) in R 3.2.3. Chromatin immunoprecipitation sequencing (ChIP-seq) for histone modifications (H3K27ac and H3K4me3) were previously performed on untreated primary RMC tumors (n=5) and adjacent normal kidney samples (n=5) (28) and can be accessed at the NCBI Gene Expression Omnibus (RRID:SCR_005012) (accession: GSE284396). Integrative Genomics Viewer (IGV) was used to visualize the signal density of RNA-Seq and ChIP-Seq reads at the EGFR locus.

In Vitro Assays

Cell lines UOK360 (RRID:CVCL_B3PR) and UOK353 (RRID:CVCL_B3PQ) were provided by Linehan lab under a material transfer agreement and were grown as previously described in DMEM high glucose media supplemented 2 mM L-glutamine (29). 786-O (RRID:CVCL_1051), VA-ES-BJ (RRID:CVCL_1785) and G401 (RRID:CVCL_0270) were purchased from ATCC (Manassas, VA, USA). 786-O and VA-ES-BJ were cultured in DMEM high glucose media. Epi544 (RRID:CVCL_N843), RMC2C1 (RRID:CVCL_B3PS), and RMC219 (RRID:CVCL_B3PV) were previously characterized (1,30,31). Epi544 cells were cultured in DMEM/F12 media. G401 cells were cultured in McCoy’s 5A media. RMC2C1 cells (RRID:CVCL_B3PS) were cultured in Minimum Essential Media (MEM) supplemented with L-glutamine and 1X MEM nonessential amino acids. All cell lines were grown in their respective media with 10% heat-inactivated fetal bovine serum and 1% penicillin-streptomycin added. STR fingerprinting and mycoplasma testing were performed on all cell lines. Panitumumab was purchased either from the MD Anderson pharmacy (Vectibix, Amgen) or from MedChemExpress (HY-P99041, Monmouth Junction, NJ, USA, RRID:AB_3694448); while carboplatin (S1215) and nab-paclitaxel (E1068) were purchased from Selleckchem, Houston, TX, USA. Carboplatin and nab-paclitaxel were both dissolved in dimethyl sulfoxide (D2438, MilliporeSigma, Burlington, MA, USA) prior to use, while panitumumab was used as is (saline solution). For IC₅₀ assays, CellTiter-Glo^® 2.0 Cell Viability assay (G9241, Promega, Madison, WI, USA) was used to determine IC₅₀ of panitumumab on UOK360 and 786-O cells. For UOK360 experiments, 7 × 10³ cells were seeded in a 96-well plate, allowed to grow overnight, and then treated with varying concentrations of panitumumab from 5 μM to 1 nM with 6 to 8 replicates per concentration. The experiment was run in quadruplicate. For 786-O experiments, 7 × 10³ cells were seeded in a 96-well plate, allowed to grow overnight, and then treated with varying concentrations of panitumumab from 50 μM to 250 nM with 3 to 6 replicates per concentration. In addition, similar assays were run with carboplatin with drug concentrations from 30 μM to 500 nM and nab-paclitaxel with drug concentrations from 4 μM to 0.5 nM on UOK360 cells. Three to 8 replicates of media alone (no cells) and cells with media (no drug) or with 0.1% DMSO (carboplatin and nab-paclitaxel experiments) were run in each 96-well experiment as controls. All cell proliferation assays were run two to three times unless otherwise noted. In all assays, cells were incubated with drug for 72 hours. IC₅₀ values were determined using data from all replicate experiments. To demonstrate that EGF inhibits the binding of panitumumab, two separate experiments were run on UOK360 cells where 100 ng/ml of recombinant human EGF protein (PHG0313, Gibco, Waltham, MA, USA) was added to the media.

For combination drug assays, UOK360 cells were seeded at 7 × 10³ cells per well in a 96-well plate, allowed to grow overnight, and then were treated with 300 nM nab-paclitaxel (NP); 300 nM panitumumab (P); 5 μM of carboplatin (C); 300 nM NP+ 300 nM P; 300 nM NP + 5 μM C; 300 nM P + 5 μM C; 300 nM NP + 300 nM P+ 5 μM C; 0.1% DMSO; or media alone. These fixed concentrations were selected at sub-IC₅₀ levels relative to the single-agent IC₅₀ values to permit detection of enhanced growth inhibition by the combinations. Five to 6 replicates per treatment and cells were treated for 72 hours prior to being assayed with CellTiter-Glo^® 2.0. Three separate experiments were run. Similar experiments were run on UOK353 cells.

For cell cycle experiments, UOK360 (150,000 cells per well) and 786-O (50,000 cells per well) were seeded in a 6-well plate in their respective media with 10% FBS (complete media) and allowed to attach overnight. Media was then removed, and cells were primed with 0.1% FBS to synchronize cells for 24 hours. Media was then removed and either complete media with or without 1 μM of panitumumab was added. After 24 hours, cells were harvested and used with the cell cycle analysis kit (Abcam, ab287852). Cells were strained through a cell strainer (Corning, 352235) prior to analysis on Beckman Coulter Gallios Flow Cytometer. Three separate experiments were run with 3 to 2 technical replicates in each experiment.

SMARCB1 re-expression experiments

RMC (UOK360, RMC219, RMC2C1), epithelioid sarcoma (VA-ES-BJ), and malignant rhabdoid tumor (G401) cell lines were engineered for tetracycline-inducible re-expression of SMARCB1 (also known as SNF5/INI1) by transduction with the tetracycline-inducible pIND20-fSNF5-HA vector (32) kindly provided by Dr. Bernard E. Weissman. Lentivirus was produced in HEK293FT cells (RRID:CVCL_6911) and used to generate stable tetracycline-inducible cell lines as previously described (33). All plasmid vectors were propagated in Escherichia coli strain DH5α (Invitrogen, Cat. #18265017). Unless otherwise noted, SMARCB1 re-expression was induced with doxycycline at 0.5 μg/mL for 3 days in cells carrying the tetracycline-inducible pIND20-fSNF5-HA vector.

Immunofluorescence imaging

UOK360 cells were seeded onto 12-well culture plates with coverslips and incubated overnight at 37°C with 5% CO₂. The following day, cells were incubated either with panitumumab (100 nM) or erlotinib (100 nM) for 2 h or 72 h. Cells were rinsed with PBS, fixed with 4% paraformaldehyde (PFA) for 15 min at room temperature, and permeabilized with 0.5% Triton X-100 in PBS for 20 min. After blocking with 3% BSA for 1 h, cells were incubated overnight at 4 °C with rabbit anti-EGFR (Cell Signaling Technology Cat#4267, RRID:AB_2246311; 1:250 dilution) and mouse anti-LAMP1 (Cell Signaling Technology Cat#15665, RRID:AB_2798750; 1:200 dilution) antibodies. The next day, cells were washed three times with PBS and incubated with corresponding Alexa Fluor–conjugated secondary antibodies (Invitrogen; 1:1000) for 1 h at room temperature. After three PBS washes, cells were post-fixed with 4% PFA for 10 min, washed, and stained with DAPI (ThermoFisher; 1:10,000, 10 min, RRID:SCR_008452). Coverslips were mounted using ProLong^™ Gold or Glass Antifade Mountant (ThermoFisher, RRID:SCR_008452). High-resolution images were acquired using a Leica SP8 laser scanning confocal microscope equipped with an HC PL APO 100×/1.40 CS2 objective, spectral detectors, and Leica Super Z Galvo stage. Exposure settings were kept constant across conditions. Images were processed using Leica LAS X software (RRID:SCR_013673).

Image Processing and Colocalization Analysis

Acquired confocal immunofluorescence images were preprocessed to ensure accurate quantitative colocalization of EGFR and lysosomal protein, LAMP1. Otsu automatic thresholding was applied to normalize signal detection across image fields and minimize low-intensity background variability. Background subtraction was performed using a 20-pixel rolling ball radius to remove diffuse cytosolic fluorescence and uneven illumination. A Gaussian filter (σ = 1) was used to suppress high-frequency noise while preserving spatial structure of true signals. Colocalization analysis was performed using the Coloc2 plugin (FIJI/ImageJ; RRID:SCR_002285 for FIJI, RRID:SCR_003070 for ImageJ) with Costes’ significance threshold enabled for objective, signal-dependent thresholding. Manders’ coefficient tM1 (fraction of EGFR intensity overlapping LAMP1) and tM2 (fraction of LAMP1 intensity overlapping EGFR) were calculated from manually defined regions of interest (ROIs) for individual cells. Manders values from individual cells were exported for analysis. Pearson’s correlation coefficient (r) was computed from pixel intensities to evaluate linear relationships between channels. Scatter plots, regression lines, and 95% confidence intervals were generated in GraphPad Prism (RRID:SCR_002798). Additionally, cells with tM1 ≥ 0.5 were classified as exhibiting strong colocalization of EGFR in LAMP1, and cells that show tM1≥ 0.5 and tM2≥ 0.5 were classified as demonstrating high bidirectional EGFR-LAMP1 colocalization and the percentage of high-colocalization per cell per condition was plotted. Statistical comparisons across groups were performed using ANOVA with Dunnett's tests.

Protein extraction and western blot analyses

For cellular protein lysates, cells were scraped on ice using cold 1X RIPA lysis buffer (ThermoFisher Scientific; PI89901, RRID:SCR_008452) supplemented with Pierce Protease and Phosphatase inhibitor cocktail (ThermoFisher Scientific; A32961, RRID:SCR_008452), sonicated for five cycles with 30 sec on and 30 sec off per cycle, and centrifuged at maximum speed for 10 min at 4 °C. The pellet was removed, and supernatants was collected and processed further for quantitation of protein concentration using Pierce^™ BCA Protein Assay Kits (ThermoFisher Scientific; 23227, RRID:SCR_008452) following manufacturer’s instructions. All the protein samples were resolved by using SDS–PAGE (4–20%) using PROTEAN TGX Precast Gels (Bio-Rad) under reducing and denaturing conditions. Resolved proteins were transferred to polyvinylidene difluoride (PVDF) membranes at either 300 mA for 2 to 3 hours or at 15V overnight. Membranes were blocked in tris-buffered saline [50 mM tris-HCl (pH 7.6) and 150 mM NaCl] containing 0.5% Tween 20 (TBST) for 1h at room temperature and incubated with primary antibody with gentle rocking overnight at 4°C. The primary antibodies used included: rabbit monoclonal anti-AKT (Cell Signaling Technology Cat# 4691, RRID:AB_915783), rabbit monoclonal anti- phospho AKT (Cell Signaling Technology Cat#4058, RRID:AB_331168), rabbit monoclonal anti-p44/42 MAPK (ERK1/2) (Cell Signaling Technology Cat# 9102, RRID:AB_330744), rabbit monoclonal anti-phospho-p44/42 MAPK (ERK1/2) (Cell Signaling Technology Cat# 4370, RRID:AB_2315112), rabbit monoclonal anti-PARP (Cell Signaling Technology Cat#9542, RRID:AB_2160739), rabbit monoclonal anti-caspase-3 (Cell Signaling Technology Cat# 9662, RRID:AB_331439), rabbit monoclonal anti-cleaved caspase-3 (Cell Signaling Technology Cat# 9661, RRID:AB_2341188), and mouse monoclonal anti- β-Actin (Santa Cruz Biotechnology Cat# sc-47778, RRID:AB_626632). Membranes were washed three times with TBST, incubated with the appropriate horseradish peroxidase (HRP)–conjugated secondary antibodies (Cell Signaling Technology Cat# 7074, RRID:AB_2099233 or Cell Signaling Technology Cat# 7076, RRID:AB_330924) for 1 hour at room temperature, washed three times with TBST, and developed using Pierce ECL substrate (ThermoFisher Scientific; PI35050, RRID:SCR_008452) or Amersham ECL Prime (Cytiva; RPN2236) for 3 and 5 min, respectively. Quantitation of immunoblots was performed via densitometric analysis using ImageQuant TL software (Cytiva, RRID:SCR_014246), with data normalized to respective total protein.

Antibody Dependent Cellular Cytotoxicity (ADCC)

Cell culture and drug treatment

UOK360 or 786-O cells seeded in 96-well plates at a density of 7.5 × 10³ cells per well. At the time of seeding, a subset of wells was treated with panitumumab or erlotinib or vehicle control (0.1% DMSO) and incubated overnight (O/N). On the following day, an additional subset of wells was shortly treated with panitumumab, erlotinib, or vehicle control for 2 hours prior to NK cell co-culture.

NK cells and Target cells co-culture and cytotoxicity assay

Purified Human Cord blood derived NK cells from different donors were cultured overnight under standard culture conditions in NK cell culture media [45% RPMI-1640 (Cellgro, Manassas, VA) and 45% Click’s media (Irvine Scientific, Santa Ana, CA) supplemented with 10% AB human serum (Atlanta Biologicals, Lawrenceville, GA) and 100 IU/mL IL-2 (Proleukin; Chiron, Emeryville, CA)] with or without panitumumab (100 nM) or erlotinib (100 nM). UOK360 target cells were then co-cultured with NK cells at an effector-to-target (E: T) ratio of 2:1 in 96-well plates containing complete medium for 24 hours. To assess cell death, Propidium Iodide (PI; Thermo Fisher Scientific, Cat. No. P1304MP) at a final concentration of 0.5 μM was added to the co-culture immediately before imaging. Real-time live-cell imaging was performed for 24 hours using EVOS microscope equipped with a heated environmental control chamber (37°C, 5% CO₂), with images acquired every 1 hour. For real-time imaging analysis, images were saved in tiff and analyzed using ImageJ (RRID:SCR_003070). The area of PI-positive (dead) cells was quantified over time and limited to the threshold to determine NK cell–mediated cytotoxicity using ImageJ.

In Vivo Therapeutic Studies

We treated xenograft animal models with 3 mg/kg of panitumumab injected twice a week via intraperitoneal injection (IP) under the MD Anderson approved IACUC protocol 1200-RN03. Panitumumab treatment was compared to erlotinib (25 mg/kg, oral gavage once a day) and vehicle treatment (saline, oral gavage once a day) 5 days on 2 days off. Panitumumab (100 mg vial of Vectibix, Amgen Inc. Thousand Oaks, CA, USA) and erlotinib (150 mg tablet, Sun Pharmaceuticals Industries, Mumbai, India) were used. Panitumumab was diluted with saline to a concentration of 1 mg/ml and erlotinib was ground with a mortar and pestle and resuspended in 15 ml of saline and 150 μl of DMSO (D2650, Millipore Sigma). For RMC32X and FH-RCC1X PDX models, 3-5 mm of tissue was engrafted in female and male NSG mice (Strain 005557, Jackson Laboratory, Bar Harbor, ME, RRID:IMSR_JAX:005557) (12). For UOK360 and 786-O models, 3 x 10⁶ cells in a 1:1 phosphate buffered saline (PBS) and Matrigel mixture were injected into male and female NSG mice (100 μl sub-Q) (RRID:IMSR_JAX:005557). Tumors were measured with calipers once a week and tumor volumes were calculated as V = ½ (width*width*length). Mice began treatment when tumors reached 100 to 200 mm³ in volume. Treatment was continued until day 39 for the RMC32X studies or until euthanasia for the FH-RCC1X, UOK360, and 786-O experiments. Mice were euthanized when tumor volumes reached approximately 2000 mm³. Tumors were excised from the mice, weighed, and cut into multiple pieces to allow for cryofrozen tissue and generation of FFPE blocks. For RMC32X, a separate animal experiment was run with vehicle (n=5 mice) and panitumumab (n=3 mice) for two weeks to allow us to obtain tissue for correlative studies. These tissues were used for IHC staining and histology.

The combination of panitumumab and nab-paclitaxel was tested in a CDX mouse model of RMC (UOK360). 5 x 10⁶ UOK360 cells in a 1:1 phosphate buffered saline (PBS) and Matrigel mixture were injected into male and female NSG mice (100 μl sub-Q) (RRID:IMSR_JAX:005557). Mice began treatment when tumors reached 60 to 100 mm³ in volume. The combination of nab-paclitaxel (30 mg/kg) weekly IV injection, panitumumab (1.5 mg/kg) by IP injection twice a week was compared to each monotherapy and vehicle (saline, IP injection twice a week). For nab-paclitaxel, a 100 mg vial of Sandoz nab-paclitaxel (Basel, Switzerland) was reconstituted in normal saline to a concentration of 5 mg/ml. While on treatment, mice were weighed and tumor volume measured by calipers twice a week. Mice were euthanized after 28 days of treatment or earlier due to tumor size (1 vehicle mouse on day 21) or tumor ulceration (1 nab-paclitaxel mouse on day 25, 1 nab-paclitaxel mouse on day 7). Two mice in the combination treatment arm were found dead on Day 3 of treatment. Tumors were excised from all mice that survived to day 21 through day 28 of treatment. Tumors were weighed and cut into multiple pieces to obtain cryofrozen tissue and generation of FFPE blocks.

Data Availability

ChiP-seq and RNA-seq data sequencing data used in this study are available in the NCBI Sequence Read Archive (SRA) under accession PRJNA605003 and the Gene Expression Omnibus (GEO) under accession GSE284396. Requests to access raw non-clinical data used to generate figures and tables should be forwarded to the corresponding authors at PMsaouel@mdanderson.org and/or NMZacharias@mdanderson.org and/or MKarki@mdanderson.org and/or ntannir@mdanderson.org and will be promptly reviewed to verify whether the request is subject to any intellectual property or confidentiality obligations. Individual-level clinical data are not publicly available because they contain potentially identifying information from participants in this study. De-identified data may be made available upon reasonable request to the corresponding authors, subject to institutional review board approval and a data transfer agreement.

RESULTS

Confirmation of EGFR Expression Levels in RMC

We performed clinical-grade CLIA-certified IHC for EGFR on tumor tissues from 11 patients with RMC and quantified expression using the H-score method (Figure 1A-B). Membranous EGFR staining was present in 100% of tumor cells in every case, with none scored as negative or weak. Staining intensity was moderate (2+) in 3/11 cases and strong (3+) in 8/11 cases, corresponding to H-scores of 200 and 300, respectively (median H-score 300, range 200–300). These data confirm the uniformly high EGFR protein expression in RMC tissues. We then integrated RNA sequencing (RNA-seq) and chromatin immunoprecipitation sequencing (ChIP-seq) from untreated primary RMC tumors and adjacent kidney controls (Figure 1C). EGFR RNA expression was significantly upregulated in RMC tumors versus adjacent kidney controls (Log2 fold change = 1.64, FDR = 0.00004) with concomitant enrichment of H3K27ac and H3K4me3 histone modifications at an EGFR-linked enhancer region, consistent with an active transcriptional state (Figure 1C). Interrogating the same dataset for the seven canonical EGFR ligands (34) revealed a coordinated ligand switch in RMC (Table S1). The predominant renal ligand EGF was markedly downregulated relative to adjacent kidney (log₂ fold change = −8.6, FDR = 8×10⁻⁶⁰), consistent with loss of the normal nephron EGF source, whereas tumors upregulated a repertoire of alternative EGFR ligands — most prominently epiregulin (log₂FC = +6.3, FDR = 4×10⁻⁸) and amphiregulin (log₂FC = +6.2, FDR = 6×10⁻⁷), both within the top ~1% of upregulated genes, together with betacellulin (log₂FC = +1.7, FDR = 4×10⁻⁴) and a trend toward increased TGF-α (Table S1). Extending the ChIP-seq analysis to these loci showed that the ligand switch is accompanied by epigenetic reprogramming mirroring that observed at the EGFR locus itself. At the EGF locus, H3K27ac and H3K4me3 were depleted in RMC tumors relative to adjacent kidney, with differentially enriched peaks called DOWN, whereas both marks were gained at the EREG and AREG loci, with peaks called UP — in each case concordant with the direction of transcriptional change (Figure S1A). EREG and AREG, which are adjacent on chromosome 4, were further linked to a shared landscape of reprogrammed enhancers distributed across a ~1.1-Mb region (chr4:73,684,677–74,785,946), the majority of which (20) were linked to both genes (Figure S1B), providing a structural basis for their coordinated induction. Thus, RMC tumors co-express abundant wild-type EGFR and a switched-on set of autocrine/paracrine EGFR ligands, with the entire ligand–receptor axis coordinately activated through promoter and enhancer reprogramming, providing a ligand source capable of driving receptor activation in the absence of activating mutations or amplification.

Figure 1. EGFR expression and epigenetic activation in RMC tissues.

A. CLIA-certified IHC for EGFR in representative RMC patient tissue sections. Scale bar: 100 μM. B. Quantification of EGFR staining intensity in tumor samples from 11 patients with RMC. Membranous EGFR staining was present in 100% of tumor cells in every positive case. C. Integrative Genome Viewer (IGV) tracks showing signal density for RNA-seq (black), H3K27ac (blue), or H3K4me3 (red) for primary RMC tumor samples or adjacent normal kidney samples. The EGFR locus is indicated, with blue arrow marking the transcription start site (TSS), and the boxed region shows location of enhancers linked to EGFR by Enhancer Atlas. To the left, the boxed enhancer region is magnified to show differences in signal intensity between normal and tumor samples. Grey or black lines underneath the peaks indicate a differentially enriched peak (DEP) called by DiffReps. Log2FC, Log2 fold change; FDR, false discovery rate.

To determine whether high EGFR expression is a generic consequence of SMARCB1 loss or is more specific to RMC, we compared EGFR protein levels by immunoblotting across a panel of SMARCB1-deficient cell lines — four RMC lines (UOK360, UOK353, RMC219, RMC2C1), two epithelioid sarcoma lines (VA-ES-BJ, EPI544), and the malignant rhabdoid tumor line G401 — alongside the SMARCB1-intact clear cell RCC line 786-O (Figure S2A-B). Although EGFR was detectable across the SMARCB1-deficient lines, the four RMC lines consistently expressed the highest EGFR levels, exceeding those of the epithelioid sarcoma and rhabdoid tumor lines (one-way ANOVA adjusted p-values < 0.05). To confirm this finding in patient tissue, we performed EGFR immunohistochemistry on a tissue microarray of 36 epithelioid sarcoma specimens (30) and compared the results with our RMC cohort (Table S2 and Figure S2C). Whereas all 11 RMC cases showed membranous EGFR positivity of moderate-to-strong intensity (median H-score 300, range 200–300), EGFR staining in epithelioid sarcoma was markedly lower and more heterogeneous: 11/36 cases (30.6%) were negative, 13/36 (36.1%) weak, 10/36 (27.8%) moderate, and only 2/36 (5.6%) strong, yielding a median H-score of 20 (range 0–270). EGFR H-scores were significantly higher in RMC than in epithelioid sarcoma (Mann–Whitney P = 1.1 × 10⁻⁶). These tissue-level data confirm our cell-line findings and indicate that, although EGFR can be expressed in epithelioid sarcoma, the uniformly high EGFR expression characteristic of RMC is not a generic consequence of SMARCB1 loss.

To test whether EGFR is directly controlled by SMARCB1, we re-expressed SMARCB1 using a tetracycline-inducible system in three RMC lines (UOK360, RMC219, RMC2C1), one epithelioid sarcoma line (VA-ES-BJ), and the rhabdoid tumor line G401, and quantified EGFR by immunoblotting (Figure S3). Despite robust SMARCB1 re-expression, EGFR protein levels were unchanged in all five lines. Thus, EGFR expression in these SMARCB1-deficient tumors is not maintained by ongoing SMARCB1 loss, indicating that the high EGFR expression characteristic of RMC is established through a stable, SMARCB1-independent mechanism rather than as a direct, reversible consequence of SMARCB1 deficiency.

In Vivo Differential Efficacy of Erlotinib versus Panitumumab

We compared erlotinib and panitumumab in vivo across four RCC xenograft models (Figure 2, Figure S4). We first investigated two established wild-type EGFR expressing PDX models (RMC32X and FH-RCC1X) known to be sensitive to erlotinib (12). RMC32X is a PDX model developed from the nephrectomy tumor of a patient who had previously received paclitaxel plus platinum chemotherapy (12,35). FH-RCC1X expresses significantly lower EGFR protein levels than RMC32X (12) and is a PDX model developed from the nephrectomy tumor of a female patient with FH-deficient RCC, who had previously been treated with first-line erlotinib plus bevacizumab for 14 months (12). In RMC32X, panitumumab achieved significantly higher responses than erlotinib (adjusted p-value = 0.0012 on day 31) (Figure 2A). Within the RMC32X panitumumab cohort, one mouse was euthanized at day 29 and another at day 56 due to declining health, however the remaining 6 mice were alive at day 107 after treatment initiation, whereas all vehicle- and erlotinib-treated mice died by days 37 and 66, respectively (Figure S5). In addition, 4 out of the 8 panitumumab-treated mice had no palpable tumors by day 31. In FH-RCC1X, panitumumab showed only a weak growth-inhibitory signal versus erlotinib (day 30, adjusted p-value = 0.051; Figure 2B, Figure S6, and Table S3)

Figure 2. In Vivo effects of EGFR inhibition across RCC models.

A-D. Tumor growth curves for RMC PDX (RMC32X), FH-deficient RCC PDX (FH-RCC1X), ccRCC CDX (786-O) and RMC CDX (UOK360) treated with vehicle alone (saline), erlotinib (25 mg/kg oral gavage 5 days on / 2 days off), or panitumumab (3 mg/kg intraperitoneal twice weekly). Tumor volume (y-axis) is plotted over days on treatment (x-axis). Inter-group comparisons of tumor volume at each treatment day were performed by two-way ANOVA with Tukey's multiple comparisons test; all adjusted P values are provided in Supplementary Table S3. E-F. Representative H&E, EGFR, and Ki67 stained UOK360 (E), and RMC32X (F) tumors across treatment cohorts. Erlotinib-treated RMC32X tissue derives from the experiment reported in ref (12); vehicle- and panitumumab-treated tissue derives from a separate two-week correlative cohort. Erlotinib images are shown for qualitative comparison and are not included in the quantification. Scale bar: 300 μm. G-H. Representative cleaved caspase-3 and cleaved PARP stained UOK360 (G) and RMC32X (H) tumors across treatment cohorts. Erlotinib images are shown for qualitative comparison and are not included in the quantification. Scale bar: 300 μm. I-J. Quantification of EGFR H-score (I) and the percentage of Ki67-positive malignant cells in tumor area (J) for vehicle (V), erlotinib (E), and panitumumab (P) treated tissues; H-Score = 1 x (% Weak Positive Cells) + 2 x (% Moderate Positive Cells) + 3 x (% Strong Positive Cells), ranging from 0 to 300. For RMC32X, statistical significance was determined using Mann Whitney, while 1-way ANOVA with Dunnett's tests adjusted for multiple comparisons was used for the other tumor models. Adjusted p-values between treatment cohorts for each model are shown. K. Percent fibrosis in panitumumab-treated UOK360, RMC32X, and FH-RCC1X tumors. L. The percentage of cleaved caspase-3 positive malignant cells in tumor area in UOK360 vehicle (V), erlotinib (E), and panitumumab (P) treated tissues. Statistical comparisons were performed using one-way ANOVA with Dunnett's tests, p-values adjusted for multiple comparisons shown on graph.

We then evaluated CDX mouse models of RMC (UOK360) and ccRCC (786-O). We chose the 786-O CDX as control because it is well-established and frequently used in the literature as a model of ccRCC (36), which is the most common RCC subtype and is known to be refractory to erlotinib clinically (37,38). As expected for ccRCC, neither erlotinib nor panitumumab reduced tumor growth in comparison to vehicle control in 786-O mice (day 42, adjusted p-value = 0.650 and 0.858, respectively; Figure 2C). Conversely, both agents were active in UOK360 versus vehicle control (day 44, both adjusted p-values < 0.0001) with panitumumab significantly outperforming erlotinib (day 44, adjusted p-value < 0.0001; Figure 2D and Figure S6). These results prompted further exploration of panitumumab in RMC. Inter-group comparisons of tumor volume over time were assessed by two-way ANOVA with Tukey's multiple comparisons test, with all pairwise adjusted P values reported in Table S3.

In Vivo Effects of Panitumumab on Tumor Histology, EGFR, and Proliferation

We evaluated the percentage of fibrosis as a morphologic treatment effect, EGFR protein expression on tumor cells, and cell proliferation by Ki67 staining in a subset of tissues after panitumumab treatment across all models - UOK360 (n=4), RMC32X (n=3), FH-RCC1X (n=5), and 786-O (n=5) (Figure 2E, F, Figure S7, Figure S8, and Figure S9). Expression of apoptotic biomarkers cleaved caspase-3 and cleaved PARP staining was determined in UOK360 and RMC32X tissues (Figure 2G, H). In the UOK360 model, panitumumab reduced EGFR H-score (Figure 2I) and the percentage of Ki67 positive cells in the tumor (Figure 2J). In RMC32X, panitumumab again produced a reduction in EGFR H-score with a lower decrease in Ki67 positive cells (Figure 2I, J). Conversely, panitumumab reduced the percentage of Ki67 positive cells but not EGFR H-score in the FH-RCC1X model (Figure 2I, J, Figure S7) and did not affect Ki67 or EGFR expression in 786-O tumor tissues (Figure 2I, J and Figure S8). In both RMC models, lower expression of cleaved caspase-3 was observed with panitumumab treatment compared to vehicle control using visual assessment of percentage of staining in tumor tissues (Figure 2G,H,L, Figure S10). The expression of cleaved PARP was consistently very low (<1%) across all cohorts in UOK360 tumors (Table S4). Panitumumab was also associated with treatment-related morphological changes such as pronounced fibrosis and hyalinization in UOK360 and RMC32X but not 786-O tissues (Figure S8). FH-RCC1X showed calcifications and occasional fibrosis in 3 out of the 5 tumors after panitumumab treatment (Figure S9), but the extent of fibrosis was significantly lower in FH-RCC1X than in UOK360 or RMC32X (Figure 2K). Collectively, decreased EGFR and Ki67 staining with accompanying fibrosis/hyalinization, and low expression of cleaved caspase-3 and cleaved PARP characterized panitumumab treatment effects in RMC tissues in vivo. This data suggests that panitumumab treatment induces cell arrest and not apoptosis.

Panitumumab Effect on EGFR Signaling and NK cell–mediated cytotoxicity

We next examined how panitumumab modulates canonical EGFR downstream signaling in RMC. EGFR proliferative signaling is transduced by two main pathways mediated by AKT and ERK1/2 respectively (39,40). In UOK360 RMC cells, a 2-hour exposure to panitumumab, but not erlotinib, significantly reduced phosphorylation of AKT (S473) and ERK1/2 (T202/Y204) relative to total AKT and ERK1/2, respectively (Figure 3A-B). In contrast, neither panitumumab nor erlotinib altered pAKT or pERK1/2 levels in 786-O ccRCC cells under the same conditions (Figure 3A-B). Thus, acute inhibition of EGFR downstream proliferative signaling by panitumumab is selective for RMC cells in vitro.

Figure 3. Panitumumab Effect on EGFR Signaling and NK cell–mediated cytotoxicity.

A. Representative Western blots of phosphorylated AKT (pAKT), total AKT, phosphorylated ERK1/2 (pERK1/2), and total ERK1/2 in UOK360 (RMC) and 786-O (ccRCC) cells treated with vehicle (V), erlotinib (E), or panitumumab (P) for 2 hours. β-actin serves as a loading control. B. Densitometric quantification of pAKT/AKT and pERK1/2/ERK1/2 ratios from UOK360 and 786-O cells treated as in (A). Individual data points (yellow), median (solid line), and Interquartile range (IQR) (dashed line) are indicated from four independent biological replicates. Statistical comparisons were performed by one-way ANOVA with Dunnett's tests adjusted for multiple comparisons. C. Representative confocal immunofluorescence images of UOK360 cells treated with vehicle (V), erlotinib (E), or panitumumab (P) for 72 hours and stained for EGFR and the lysosomal marker LAMP1. Merged images show increased EGFR–LAMP1 overlap in panitumumab-treated cells, consistent with enhanced lysosomal routing of EGFR. Scale bar: 25 μm. D. Proportion of cells with Manders coefficients for EGFR signal overlapping LAMP1 (tM1) ≥ 0.5, indicating robust EGFR–LAMP1 colocalization. Individual data points (yellow), median (solid line), and IQR (dashed line) are indicated from independent biological replicates. Statistical comparisons were performed by one-way ANOVA with Dunnett's tests adjusted for multiple comparisons. E. Representative live-cell images from the antibody-dependent cellular cytotoxicity (ADCC) assay. UOK360 or 786-O target cells were pretreated for 2 hours with vehicle (V), erlotinib (E), or panitumumab (P) and then co-cultured with purified human NK cells (effector: target ratio 2:1) for 24 hours in the presence of propidium iodide (PI) to label dying cells. PI-positive (dead) cells (red) are overlaid on the corresponding bright-field (BF) images. Scale bar: 100 μm. F. Quantification of NK cell–mediated cytotoxicity in UOK360 or 786-O cells treated as in (E). Violin plots show the distribution of PI-positive areas for each condition across three independent experiments performed with three different NK cell donors (CB-NK 1- cyan blue symbol; CB-NK 2- yellow symbol; CB-NK 3- purple symbol); median (solid line), and IQR (dashed line) are indicated. Differences among treatment groups were assessed by one-way ANOVA with Dunnett's tests adjusted for multiple comparisons.

To link these signaling changes to receptor trafficking, we performed a comprehensive evaluation using high-resolution confocal immunofluorescence and multiple quantitative colocalization analyses of EGFR and the lysosomal marker LAMP1 in UOK360 cells (Figure 3C-D and Figure S11). Prolonged panitumumab treatment (72 hours) induced a marked redistribution of EGFR from the plasma membrane into prominent intracellular aggregates that strongly colocalized with LAMP1, whereas erlotinib and vehicle controls preserved predominantly membranous EGFR staining (Figure 3C and Figure S11). This spatial convergence was confirmed through extensive pixel-by-pixel intensity histograms and linear regression of Manders’ coefficients, showing that panitumumab-treated cells exhibit a compact, positively sloped pixel cloud and highly significant correlation (r = +0.82, P < 0.0001) indicative of robust receptor-lysosome trafficking (Figure S11A-B). Panitumumab significantly increased the fraction of cells with high Manders coefficients (tM1 ≥ 0.5 for EGFR overlapping LAMP1) and greatly increased cell population exhibiting reciprocal/ bidirectional EGFR ↔ LAMP1 colocalization (showing Mander’s coefficient, tM1 ≥ 0.5 and tM2 ≥ 0.5) compared to erlotinib or vehicle (Figure 3D and Figure S11C). Increased EGFR ↔ LAMP1 colocalization was also observed with panitumumab treatment in RMC UOK353 (Figure S12) and RMC219 (Figure S13) cells but not in ccRCC 786-O cells (Figure S14). Together, these data indicate that panitumumab in RMC both acutely suppresses AKT/ERK1/2 signaling and promotes lysosomal routing of EGFR in RMC cells.

Because panitumumab is an IgG2 monoclonal antibody, we next tested whether natural killer (NK) cell–mediated antibody-dependent cellular cytotoxicity (ADCC) contributes to its antitumor activity in RMC. UOK360 target cells were co-cultured with cord blood–derived human NK cells from three independent donors in the presence of vehicle, erlotinib, or panitumumab. Our ADCC evaluation yielded robust results across multiple experimental conditions, including varying pre-incubation periods and the use of NK cells from different donors (Figure 3E-F and Figures S15 and S16). Assay performance was confirmed using 786-O cells as a positive control for NK effector function and assay performance, in which NK cell–mediated killing was robust and reproducible across donors (Figure 3E-F and Figure S15). In UOK360 cultures, however, NK cell–mediated cytotoxicity after 24 hours of co-culture did not differ significantly among vehicle, erlotinib, or panitumumab treatment arms (Figure 3F). Notably, no consistent enhancement of killing was observed regardless of donor-specific variability or pre-incubation conditions (Figure S16). Similarly, panitumumab did not increase NK cell–mediated killing in UOK353 and RMC219 cells (Figure S17). These findings suggest that, under the tested conditions, NK cell–dependent ADCC is unlikely to be a major contributor to the antitumor effects of panitumumab in RMC, which instead appear to be primarily mediated through direct disruption of EGFR signaling and receptor trafficking.

Prospective Evaluation of Panitumumab-based Therapy in Patients with RMC

Between May 2022 and May 2025, we prospectively evaluated panitumumab-based therapy in 26 patients with RMC (Tables S5 and S6). The first and last patients-initiated treatment on May 13th, 2022, and May 31st, 2025, respectively. At the data cutoff (September 28th, 2025), median follow-up time was 19.8 months, and 3/26 patients remained on therapy without evidence of disease progression. As is representative of RMC (2,9,41), patients were young (median age 33.5 years), predominantly Black (61.5%) and male (76.9%), with stage 4 disease at diagnosis (88.5%) and right-sided kidney primaries (57.7%). Most patients had sickle cell trait (92.3%), and 50% had undergone nephrectomy before starting panitumumab-based chemotherapy. Nearly all patients (96.2%) previously received platinum-based chemotherapy, and 84.6% had received ≥2 prior systemic therapies (Table S5). Starting doses and treatment schedules of the panitumumab-based therapy used for each patient are detailed in Table S7. Twenty patients began the triplet of panitumumab + nab-paclitaxel + carboplatin (a regimen adapted from HER2-negative EGFR wild-type inflammatory breast cancer(42)), two started doublet therapy with panitumumab + nab-paclitaxel, and four received panitumumab monotherapy.

Clinical Activity

All 26 patients treated with a panitumumab-based regimen were evaluable for efficacy. Clinical response outcomes are summarized in Table S8. Figure 4A shows the waterfall plot of the best-measured response compared with baseline imaging. Four patients (15.4%) achieved confirmed complete responses (CR), 10 (38.5%) had partial responses (PR) and 7 patients (26.9%) had stable disease (SD), yielding an objective response rate of 53.9%. The remaining 5 patients had progressive disease as best response per RECIST v1.1. Treatment duration is shown in the swimmer plot (Figure 4B). Median PFS from panitumumab-based therapy initiation was 5.8 months (95% CI, 4.0 to 8.2; Figure 4C) and median OS was 9.5 months (95% CI 7.6 to non-estimable; Figure 4D).

Figure 4. Clinical activity of panitumumab-based therapy in RMC.

A. Waterfall plot showing the maximum change in the sum of the longest dimensions in each of the 26 evaluable patients treated with panitumumab-based therapy. Individual starting doses are provided in Table S7. B. Swimmer plot showing the time on panitumumab-based therapy until progression or last follow-up. C. Progression-free survival (PFS) from panitumumab-based therapy initiation. The shaded areas indicate the 95% confidence bands. CI = Confidence interval. D. Overall survival (OS) from panitumumab-based therapy initiation. The shaded areas indicate the 95% confidence bands. NE, non-estimable.

Representative images illustrate a radiologic PR in a heavily pretreated patient (four prior systemic therapies) with rapidly progressing disease at baseline (Figure 5A). Another patient (RMC89), previously treated with three systemic regimens, achieved radiologic CR allowing for cytoreductive radical nephrectomy per consensus recommendations (3). Pathology showed ypT1a with a largest residual tumor focus of 0.1 cm, marked fibrosis and hyalinization, lymphocyte infiltration, macrophage aggregates, and hemosiderin deposition, with only rare residual tumor cells (Figure 5B-C). This histomorphologic pattern, which mirrors the stromal remodeling observed in RMC32X and UOK360 models after panitumumab (Figure S9), stands in stark contrast to the dense, high-grade malignant epithelial structures characteristic of untreated primary RMC (Figure S18). We additionally performed EGFR immunohistochemistry to assess target expression in the residual tumor bed of RMC89. The tumor component on the block selected for IHC was extremely limited, and on a deeper stained section only rare EGFR-positive epithelial structures were identified (Figure S19). On expert genitourinary pathology review (P.R.), these foci could not be reliably classified as residual RMC versus benign renal structures, so EGFR expression in this specimen was deemed indeterminate.

Figure 5. Treatment effects of panitumumab-based therapy in RMC.

A. Axial computed tomography (CT) imaging of RMC lung metastasis at baseline (1 day before initiation of panitumumab-based therapy) and after 8 weeks of panitumumab + nab-paclitaxel + carboplatin (patient RMC51). B. Nephrectomy specimen from patient RMC89 after panitumumab + nab-paclitaxel + carboplatin showing extensive fibrosis/hyalinization with scant residual tumor. Scale bar: 600 μm. C. Higher-magnification views of numbered boxed regions in panel B. Scale bars: 200 μm.

Adverse events

Treatment-related adverse events (TRAEs) and their grade distribution are listed in Table S9. No unexpected safety signals were observed and no treatment-related deaths occurred. The most common TRAEs of any grade were acneiform rash (80.8%), peripheral sensory neuropathy (57.7%), thrombocytopenia (50%), nausea (47%), anemia (40%) and anorexia (40%). The most common grade ≥3 adverse event was grade 3 asymptomatic decrease in neutrophil count (7.7%) whereas grade 3 febrile neutropenia occurred in one patient (3.8%). Infusion related reaction due to carboplatin occurred in 6/20 (30%) patients treated with the triplet therapy of panitumumab + nab-paclitaxel + carboplatin.

In Vitro Activity of Panitumumab-Based Combination Therapies

Using 786-O cells as a negative control (half maximal inhibitory concentration [IC₅₀] value = 131 μM), UOK360 RMC cell line was sensitive to panitumumab monotherapy with IC₅₀ of 841 nM (Figure 6A). The efficacy of panitumumab in UOK360 cells could be abrogated with the addition of 100 ng/ml of EGF in the media (Figure S20). Panitumumab inhibited UOK360 cells from leaving the G0/G1 phase which was not observed in 786-O cells (Figure S21). Panitumumab did not increase cleaved PARP or cleaved caspase-3 in either cell line at 100 nM or 1 μM for 3 days, the latter exceeding its UOK360 IC₅₀ of 841 nM (Figure S22). Furthermore, nab-paclitaxel was active (IC₅₀ = 1.51 μM), while carboplatin showed minimal effect (30 μM viability 69.2 ± 18.9%) in UOK360 cells (Figure 6B). Fixed-dose combination assays (300 nM panitumumab, 300 nM nab-paclitaxel, 5 μM carboplatin) in UOK360 and UOK353 RMC cells demonstrated that panitumumab + nab-paclitaxel produced significantly greater growth inhibition than either agent alone (Figure 6C). These fixed doses were selected as sub-IC₅₀ concentrations to better evaluate the potency of combination regimens. Both RMC lines were carboplatin-resistant, and adding carboplatin did not enhance the activity of panitumumab or nab-paclitaxel (Figure 6C). Triplet therapy with panitumumab + nab-paclitaxel + carboplatin did not outperform the panitumumab + nab-paclitaxel doublet in either cell line (Figure 6C). Together with the high rate of carboplatin-related infusion reactions observed clinically with triplet panitumumab-based therapy, these data support dropping carboplatin and favor the panitumumab + nab-paclitaxel doublet in RMC.

Figure 6. Preclinical panitumumab combination studies.

A. Dose-response curves and calculated IC₅₀ and R² values for panitumumab in 786-O (clear cell RCC) and UOK360 (RMC) cells. The x-axis is log10-scaled (nM). B. Dose-response curves and IC₅₀ values for carboplatin and nab-paclitaxel in UOK360 cells. C. Fixed-dose assays in UOK360 and UOK353 testing nab-paclitaxel (NP, 300 nM), panitumumab (P, 300 nM), and carboplatin (C, 5 μM) as mono-, doublet-, and triplet combinations. Points reflect all replicates across three independent experiments. Statistical significance is observed between NP vs NP+ P and P versus NP + P (one-way ANOVA with Dunnett's tests, multiple comparisons adjusted p-values shown on graphs). D. Tumor growth curves between the combination of panitumumab (1.5 mg/kg, intraperitoneal twice weekly) and nab-paclitaxel (30 mg/kg, intravenous weekly) shown in purple (P+NP) versus panitumumab as a monotherapy (blue), nab-paclitaxel as a monotherapy (red), and vehicle (saline, intraperitoneal twice weekly, black) in a CDX mouse model of RMC (UOK360). Inter-group comparisons of tumor volume at each treatment day were performed by two-way ANOVA with Tukey's multiple comparisons test; all adjusted P values are provided in Supplementary Table S10. E-H. UOK360 tumor bearing mice that survived early mortality within the initial cohort were euthanized 21-28 days post-treatment with vehicle (V), nab-paclitaxel (NP), panitumumab (P), or combination therapy (P + NP) to analyze the remaining tumor tissue by histology. (E) EGFR H-score, (F) the percentage of Ki67-positive tumor cells, (G) percentage of fibrosis, and (H) the percentage of cleaved-caspase 3 positive tumor cells within tissues shown in violin plots. Inter-cohort statistical differences were evaluated using a one-way ANOVA followed by Dunnett’s multiple comparisons test for EGFR H-score, % of Ki67-positive cells, and % of cleaved caspase-3 positive cells. Mann-Whitney was used for fibrosis. Graphs display the adjusted p-values.

In Vivo Activity of Panitumumab-Based Combination Therapies

We subsequently compared tumor growth between the combination of panitumumab + nab-paclitaxel versus each as a monotherapy in the UOK360 RMC CDX mouse model. To evaluate possible drug cooperativity, we administered a reduced dose of panitumumab (1.5 mg/kg vs. 3 mg/kg) with or without a 30 mg/kg weekly dose of nab-paclitaxel. All three treatment cohorts on day 21 had significant reduction in tumor volume compared to the vehicle control (Figure 6D, Table S10). Individual tumor growth in each cohort are shown in Figure S23A and adjusted p-values between cohorts are shown in Table S10. We did observe some possible toxicity with the dual therapy as two mice died after three days of treatment. All mice survived to day 28 in the panitumumab cohort, one vehicle mouse was euthanized early based on tumor burden (day 21), and two mice were euthanized in the nab-paclitaxel cohort for ulceration of the tumor (day 7 and day 25). The mass of the mice stayed relatively consistent in all four cohorts (Figure S23B).

UOK360 tumor-bearing mice that survived early mortality within the initial cohort were euthanized 21-28 days post-treatment to analyze the remaining tissue by histology. We evaluated EGFR protein expression on tumor cells (Figure 6E, Figure S24) and cell proliferation by Ki67 (Figure 6F, Figure S24) across all treatment cohorts (vehicle n=6, nab-paclitaxel n=4, panitumumab n=6, and combination n = 3). Significantly reduced EGFR H-score was observed in both panitumumab and panitumumab + nab-paclitaxel treated tissues compared to vehicle and nab-paclitaxel (Figure 6E, Figure S24). EGFR H score was similar between panitumumab monotherapy and the panitumumab + nab-paclitaxel combination treatment (Figure 6E, Figure S24). Both panitumumab alone and combined with nab-paclitaxel caused a similar decrease in percentage of Ki67 positive tumor cells compared to vehicle treated tissues (Figure 6F, Figure S24). In addition, fibrosis, as a histologic feature consistent with treatment effect, was prominent in the panitumumab treated and panitumumab + nab-paclitaxel treated cohorts (Figure 6G, Figure S24 and Figure S25). Apoptosis markers cleaved caspase-3 and cleaved PARP expression were assessed across all treatment cohorts (Figure S24, Table S11). Cleaved caspase-3 expression (Figure 6H, Figure S24, Table S11) was highest in nab-paclitaxel treated tissues (6.0 ± 1.8%, n = 4) with the lowest expression found in panitumumab monotherapy (0.58 ± 0.37%, n=6) and the combination therapy tissues (0.67 ± 0.29%, n=3). Similar results were found with cleaved PARP (Figure S24, Table S11), with highest expression in the nab-paclitaxel tissue (2.0 + 1.0%, n=4) and < 1% staining in both panitumumab treated cohorts.

DISCUSSION

In this translational study, we demonstrate that wild-type EGFR is a therapeutically exploitable vulnerability in RMC. At the biological level, RMC tumors showed uniformly high EGFR protein expression by CLIA-certified immunohistochemistry, accompanied by increased EGFR transcription and active chromatin marks at an EGFR-linked enhancer, consistent with epigenetic activation of the locus. Notably, EGFR expression was uncoupled from SMARCB1 status: RMC expressed substantially higher EGFR than epithelioid sarcoma or rhabdoid tumor lines despite shared SMARCB1 loss, and re-expression of SMARCB1 did not reduce EGFR in any of these lineages. Together, these findings indicate that SMARCB1 loss is neither sufficient to establish nor required to maintain the high EGFR expression characteristic of RMC, which instead reflects a stable, RMC-specific lineage and enhancer program. In preclinical models, panitumumab, an extracellular EGFR-blocking monoclonal antibody, produced potent tumor regressions and stromal remodeling in RMC xenografts but not in clear cell RCC, and was more effective than the EGFR tyrosine kinase inhibitor erlotinib despite similar activity of the latter in fumarate hydratase–deficient RCC (12, 13). Together with the absence of recurrent EGFR kinase domain mutations in RMC (1,20-23), these data support a model in which RMC is dependent on signaling from wild-type EGFR that is preferentially disrupted by extracellular antibody blockade rather than by kinase inhibition. Consistent with this model, RMC tumors express a repertoire of EGFR ligands, most notably epiregulin and amphiregulin (1). Epiregulin and amphiregulin are the same ligands whose expression best predicts benefit from anti-EGFR antibodies in RAS wild-type colorectal cancer (43-45), and their induction provides a plausible autocrine/paracrine source of ligand-dependent wild-type EGFR activation. In keeping with a ligand-competitive mechanism, saturating exogenous EGF fully abrogated panitumumab activity in UOK360 cells. Because these expression data derive from bulk tumor tissue, the relative autocrine versus paracrine contribution of individual ligands and their functional hierarchy remain to be defined in future single-cell and functional studies.

Mechanistically, panitumumab rapidly attenuated canonical EGFR downstream signaling in RMC cells, with acute suppression of AKT and ERK1/2 phosphorylation and subsequent redistribution of EGFR into LAMP1-positive lysosomes, consistent with ligand-independent internalization and lysosomal routing of antibody–receptor complexes (39,40,46-50). The depth of this evaluation is underscored by our use of both unidirectional and bidirectional Manders’ coefficient analyses, which confirmed a true spatial convergence of EGFR within the lysosomal compartment rather than mere random overlap. These findings parallel observations in colorectal cancer, where reductions in tumor EGFR staining after anti-EGFR antibody therapy correlate with clinical benefit even when baseline EGFR expression does not (51), and extend this paradigm to a rare renal cancer driven by SMARCB1 loss. Furthermore, the lack of enhanced NK cell–mediated ADCC in UOK360 and UOK353 despite robust killing in control 786-O cells, coupled with the IgG2 isotype of panitumumab and prior evidence that IgG1 antibodies such as cetuximab induce substantially stronger ADCC (46,52), suggests that immune effector engagement is not occurring with panitumumab. Indeed, mechanistically the lack of NK cell-mediated ADCC is consistent with the IgG2 isotype of panitumumab which engages the NK-cell activating receptor FcγRIIIa (CD16) poorly (52). This limited effector engagement may be reinforced by the rapid panitumumab-induced internalization and lysosomal routing of EGFR, which depletes surface antibody-antigen complexes.

Clinically, the panitumumab-based regimens evaluated in our prospective, multi-national registry produced responses and survival outcomes that appear unprecedented in RMC. Historically, first-line platinum-based chemotherapy yields objective response rates around 29% and median progression-free survival (PFS) of ≤4 months, with even shorter PFS beyond the first line and limited activity of immune checkpoint inhibitors or targeted agents (2,5,6,9,10). In our cohort of 26 patients—most of whom were heavily pretreated and 96% of whom had already received platinum-based chemotherapy—we observed an objective response rate of 53.9%, including 15.4% complete responses, a median PFS of 5.8 months, and median overall survival of 9.5 months from initiation of panitumumab-based therapy. Several complete responses were durable and one enabled surgical consolidation with nephrectomy, with pathologic specimens showing extensive fibrosis and only rare residual tumor cells, recapitulating the fibrosis and hyalinization observed in our RMC xenografts. These outcomes, achieved in a real-world population enriched for poor-risk features and multiple prior therapies, support further prospective evaluation of panitumumab-based therapy in RMC.

Our integrated preclinical and clinical data also refine the optimal chemotherapy backbone for EGFR blockade in RMC. In vitro, UOK360 and UOK353 RMC cells were sensitive to panitumumab and nab-paclitaxel, whereas carboplatin had minimal single-agent activity. Fixed-dose combination assays showed that panitumumab plus nab-paclitaxel produced greater growth inhibition than either agent alone, while adding carboplatin did not increase growth inhibition and the triplet did not outperform the doublet in either RMC line. Consistent with this, combination in vivo mouse studies suggest that panitumumab and nab-paclitaxel operate independently; the combination treatment failed to reduce tumor volume beyond panitumumab monotherapy, despite a clear overall tumor response driven by nab-paclitaxel alone. Concordantly, in the clinic we observed a high incidence of infusion reactions attributable to carboplatin and no signal of superior efficacy for triplet panitumumab + nab-paclitaxel + carboplatin compared with doublet panitumumab + nab-paclitaxel. Taken together, these data suggest that carboplatin provides little added value and that a carboplatin-sparing doublet of panitumumab plus nab-paclitaxel should be preferred in most patients with RMC.

The tolerability profile of panitumumab-based therapy in this ultra-rare population has important practical implications. Expected toxicities such as acneiform rash, hypomagnesemia, and neuropathy were generally low-grade and manageable; grade ≥3 hematologic adverse events were infrequent, and no treatment-related deaths occurred. Notably, some patients initiated panitumumab with Eastern Cooperative Oncology Group performance status 3, and panitumumab monotherapy was well tolerated even in this frail subgroup. Notably, the four patients treated with panitumumab monotherapy showed only limited single-agent activity, mirroring RAS wild-type colorectal cancer, in which anti-EGFR antibody monotherapy yields modest response rates (~17%) and clinical benefit improves when it is combined with chemotherapy (17,18,53) — an improvement consistent with the independent, additive action of active agents rather than synergy (54-56). This concordance reinforces the wild-type EGFR dependency of RMC and, in keeping with our finding that panitumumab and nab-paclitaxel act independently, supports pairing these two active agents in most patients, while panitumumab monotherapy remains a reasonable option for frail individuals or those with limited marrow reserve. Our data therefore supports a pragmatic, stepwise strategy in which panitumumab monotherapy can be started in very sick patients or those with limited marrow reserve, with subsequent addition of nab-paclitaxel once performance status and counts improve. In fitter patients, panitumumab plus nab-paclitaxel can be deployed in either the first-line or salvage setting. The high response rates and depth of response observed even after multiple prior regimens suggest that prior exposure to platinum does not preclude benefit. We emphasize that this sequencing strategy reflects pragmatic clinical judgment rather than evidence from controlled comparison. Because the rarity of RMC precludes randomized evaluation of treatment sequencing, the optimal ordering and relative contributions of panitumumab and nab-paclitaxel remain undetermined, and the approach we describe should be regarded as a reasonable default rather than a data-based recommendation.

Beyond its therapeutic implications, this work highlights a scalable model for rapidly disseminating effective regimens in rare cancers. Using a prospective registry anchored at high-volume centers but open to global collaborators, we translated preclinical data into off-label panitumumab-based treatment across two continents within a relatively short time frame. The registry design enabled standardized collection of baseline characteristics, treatment details, toxicity, and centrally reviewed imaging in a disease where traditional randomized trials are logistically challenging. Similar biology-driven registries could accelerate the evaluation and dissemination of promising regimens in other variant histology RCCs and rare SMARCB1-deficient malignancies (4,57-59). Our experience offers additional lessons for other ultra-rare cancers: the decades-long delay in recognizing EGFR as an RMC dependency stemmed not from a lack of tools — IHC and anti-EGFR antibodies have long been available — but from the absence of aggregated samples, preclinical models, and functional validation, compounded by an EGFR paradigm historically centered on mutation and amplification rather than wild-type dependency. Concentrating rare cancer cases within collaborative referral networks, building models and biobanks, and applying systematic molecular and functional profiling (1,23,60) — translated through pragmatic registry and basket-trial platforms — can close such gaps far faster.

Future work should focus on elucidating mechanisms of primary and acquired resistance to panitumumab in RMC and on designing rational next-generation combinations. Systematic profiling of pre-treatment and progression biopsies from panitumumab-treated patients—integrating genomics, epigenomics, transcriptomics, phospho-proteomics, and spatial immune analyses—will be critical to identify resistance circuits and predictive biomarkers, as well as define how EGFR expression and chromatin accessibility evolve on therapy and at resistance (1,7,12,23). Such insights could inform combination strategies pairing panitumumab with emerging targeted agents against other RMC dependencies (14,23,28). Ultimately, embedding these correlative efforts within prospective therapeutic registries or basket trials may provide the statistical efficiency needed to iteratively refine RMC treatment in real time (57,61-63). Furthermore, because RMC signaling proceeds through wild-type EGFR→ ERK/AKT without activating RAS mutations, node-level inhibitors of this axis — in particular emerging agents active against wild-type RAS, such as pan-KRAS and RAS(ON) multi-selective inhibitors (64,65), alongside MEK, ERK, or PI3K/AKT inhibitors — represent rational future combinations with panitumumab or strategies to counter downstream pathway reactivation, and their activity in RMC models remains to be defined.

The predominance of cytostasis over apoptosis, accompanied by marked fibrosis and hyalinization, is reminiscent of the response of gastrointestinal stromal tumors (GIST) to imatinib, in which effective therapy frequently produces myxohyaline degeneration and metabolic quiescence without early tumor shrinkage (66-68). In this setting, size-based RECIST may underestimate activity, and density-based or metabolic response criteria may be more informative (68-70). Because RMC is highly FDG-avid (71), prospective incorporation of FDG PET/CT to capture metabolic response — together with correlative assessment of therapy-induced stromal remodeling — represents an important future direction for evaluating panitumumab-based therapy. We emphasize, however, that the biological basis of this stromal remodeling in RMC remains incompletely defined. Whether the fibrosis and hyalinization reflect a direct consequence of EGFR blockade on tumor–stroma signaling, a secondary response to tumor cell dropout, or recruitment of host stromal and immune populations cannot be resolved from our data and warrants dedicated investigation. Furthermore, the predominantly cytostatic effect of naked panitumumab, together with the efficient lysosomal routing of EGFR and the uniformly high wild-type EGFR expression of RMC, provide a strong rationale for EGFR-directed antibody-drug conjugates (ADCs) as a next-generation strategy (23,60): lysosomal delivery of a cytotoxic payload could convert growth arrest into tumor-cell killing and enhance durability. ADCs that bind wild-type EGFR, such as becotatug vedotin (MRG003) — an anti-EGFR–MMAE conjugate — deliver this payload selectively to EGFR-expressing cells and warrant evaluation in RMC (72). These agents have not been evaluated in RMC, and their activity in this disease remains entirely hypothetical pending clinical investigation. More broadly, RMC uniformly overexpresses several actionable surface antigens — EGFR and TROP2, EPCAM, CLDN6, and CDH6 (23,60) — making it particularly amenable to ADC-based precision therapy, while its immune refractoriness (5,8) and characteristic CA125 expression (28) nominate epigenetic-immune combinations (e.g., p300/CBP suppression with checkpoint blockade) and MUC16-directed strategies as complementary avenues (60). Given the rarity of RMC, we envision panitumumab-based therapy as a systemic backbone onto which these biologically grounded strategies are prioritized and combined within adaptive registry- and basket-trial platforms.

This study has limitations. It is a single-arm, non-randomized registry with a modest sample size, and comparisons to historical controls are inherently subject to selection and lead-time biases. Patients were treated off-label across multiple centers with some heterogeneity in dosing schedules and chemotherapy backbones, and a subset lacked standardized radiographic follow-up or central imaging review. Although our preclinical data strongly implicate panitumumab as the key driver of activity and argue against a major contribution from carboplatin, the relative contributions of panitumumab versus nab-paclitaxel cannot be definitively disentangled without randomized comparisons. Correlative tissue was limited at progression, constraining our ability to systematically study resistance mechanisms. Consequently, the mechanisms underlying acquired resistance to EGFR blockade in RMC remain undefined, and their elucidation will require paired pre-treatment and progression samples analyzed by integrated genomic, transcriptomic, and epigenomic profiling. Nonetheless, the concordant preclinical and clinical signals, the depth and durability of responses in heavily pretreated patients, and the favorable safety profile collectively provide a strong rationale for further prospective validation of panitumumab-based therapy in RMC and a platform on which to build the next generation of RMC therapeutics.

STATEMENT OF TRANSLATIONAL RELEVANCE.

RMC is a lethal malignancy predominantly affecting young individuals of African descent. This study identifies wild-type EGFR as a foundational RMC dependency effectively targeted via panitumumab-induced lysosomal degradation. A prospective registry demonstrates a 53.9% objective response rate in heavily pretreated patients. These non-randomized findings provide a clinical rationale for further prospective validation of panitumumab-based therapy in RMC.

Abbreviations list:

ADC

Antibody–drug conjugate

ADCC

Antibody-dependent cellular cytotoxicity

ANOVA

Analysis of variance

AREG

Amphiregulin

ccRCC

Clear cell renal cell carcinoma

CDX

Cell line–derived xenograft

ChIP-seq

Chromatin immunoprecipitation sequencing

CI

Confidence interval

CLIA

Clinical laboratory improvement amendments

CR

Complete response

CT

Computed tomography

CTCAE

Common terminology criteria for adverse events

EGF

Epidermal growth factor

EGFR

Epidermal growth factor receptor

EREG

Epiregulin

ERK1/2

Extracellular signal–regulated kinase 1/2

ES

Epithelioid sarcoma

FDR

False discovery rate

FFPE

Formalin-fixed paraffin-embedded

FH

Fumarate hydratase

GIST

Gastrointestinal stromal tumor

H&E

Hematoxylin and eosin

IC₅₀

Half-maximal inhibitory concentration

IGV

Integrative Genomics Viewer (software)

IHC

Immunohistochemistry

LAMP1

Lysosomal-associated membrane protein 1

NK

Natural killer (cell)

OS

Overall survival

PARP

Poly(ADP-ribose) polymerase

PD

Progressive disease

PDX

Patient-derived xenograft

PFS

Progression-free survival

PR

Partial response

RCC

Renal cell carcinoma

RECIST

Response evaluation criteria in solid tumors

RMC

Renal medullary carcinoma

RNA-seq

RNA sequencing

SD

Stable disease

TKI

Tyrosine kinase inhibitor

TRAE

Treatment-related adverse event