KRAS is required for plexiform neurofibroma formation and represents a targetable vulnerability in established tumors

Abstract

Patients with neurofibromatosis type 1 develop Schwann cell tumors called neurofibromas that arise within peripheral nerves, driven by loss of neurofibromin and consequent increased RAS/RAF/MEK signaling. MEK inhibitors achieve partial responses for benign neurofibromas but are limited by toxicity and incomplete efficacy, necessitating alternative approaches. Using the Dhh-Cre;Nf1^fl/fl neurofibroma mouse model, we found that genetic ablation of Kras, but not Hras, markedly reduced neurofibroma development, inhibited MAPK activation, and rescued disrupted Remak bundles that are a morphologic hallmark of neurofibromas. These findings reveal a RAS paralog–specific requirement for KRAS in NF1-deficient neurofibroma initiation. Pharmacological KRAS inhibition with BI6674, an orally bioavailable KRAS^multi inhibitor, reduced tumor volume and proliferation in established neurofibromas and remodeled the tumor immune microenvironment, decreasing macrophages and dendritic cells. Combining KRAS and MEK inhibition further enhanced tumor regression. These findings demonstrate that KRAS is essential for neurofibroma formation and represents a promising therapeutic target, supporting clinical evaluation of KRAS inhibition for neurofibromas in patients with neurofibromatosis type 1.

INTRODUCTION

Neurofibromatosis type 1 is an autosomal dominant tumor predisposition syndrome affecting one in 2000 to 3000 children. Approximately 50% individuals with a germline NF1 mutation subsequently develop peripheral nerve sheath tumors called neurofibromas (1). These tumors arise through a classic two-hit mechanism: Following rare somatic cell loss of the remaining wild-type (WT) NF1 allele in peripheral nerve Schwann cells (SCs) or their precursors, the resulting biallelic NF1 inactivation drives neurofibroma formation (2–4). Plexiform neurofibromas (PNFs) are neurofibromas that develop in deep nerves and exhibit most rapid growth during childhood; PNFs can cause substantial morbidity due to their size and/or location (5). Beyond the NF1-deficient SCs that initiate tumorigenesis, neurofibromas are complex multicellular lesions containing fibroblasts and abundant immune cells, including macrophages, dendritic cells (DCs), and T cells (6). Most concerning is the potential of PNF for malignant transformation: 8 to 13% of patients with neurofibromatosis type 1 develop malignant peripheral nerve sheath tumors (MPNSTs), often during adolescence or young adulthood (7). This clinical progression from benign PNF to aggressive sarcomas results from additional genetic changes in tumor SCs (8) and underscores the urgent need for effective therapies that can shrink PNFs before malignant transformation occurs.

The NF1 gene product, neurofibromin (NF1), functions as a GTPase-activating protein that accelerates the hydrolysis of active RAS-GTP into inactive RAS-GDP, thereby attenuating RAS signaling (9). NF1 regulates six homologous RAS paralogs (HRAS, NRAS, KRAS, RRAS, TC21/RRAS2, and MRAS) (10), and NF1 loss in cells could results in RAS hyperactivation with elevated intracellular RAS-GTP levels (9, 11, 12). This RAS hyperactivation drives oncogenic signaling through multiple downstream effector pathways, particularly the mitogen-activated protein kinase kinase/extracellular signal–regulated kinase (MAPKK/ERK or MEK/ERK) cascade, which confers survival and proliferation advantages to NF1-deficient cells (13).

Individual RAS paralogs can have distinct biological properties, including cell type–specific RAS expression patterns (14), paralog-specific subcellular localizations (15, 16), posttranslational modifications (17), and downstream effectors (18). This functional specialization among RAS paralogs suggests that different RAS proteins may contribute differentially to neurofibroma pathogenesis. Consequently, identifying which specific RAS protein(s) drive the proliferative and survival abnormalities in NF1-deficient SCs could reveal previously unknown therapeutic targets and enable more precise, paralog-selective treatment strategies. NF1-deficient SCs express all six RAS proteins (19, 20), yet their individual contributions to neurofibroma formation remain incompletely understood. While RRAS2/TC21 mediates the aberrant migration in NF1-deficient SCs (19) and contributes to self-renewal of SC precursors (21), loss of TC21 only slightly delays tumorigenesis of neurofibroma in mice (21). Recent studies demonstrate that in vitro proliferation of NF1^−/− SCs requires KRAS and, to a lesser extent, HRAS (20). Similarly, KRAS mediated aberrantly increased proliferation in mouse NF1-deficient astrocytes (22). The idea that specific activated RAS proteins drive tumor initiation and maintenance is supported by studies using inducible RAS mutant models. In a murine melanoma model with doxycycline-inducible Hras^G12V, activated HRAS drove tumor initiation, and subsequent loss of mutant HRAS following doxycycline withdrawal led to regression of established tumors (23). Similarly, withdrawal of mutant KRAS in doxycycline-inducible pancreatic ductal adenocarcinoma models eradicated established tumors (24).

The Dhh-Cre;Nf1^fl/fl mouse model we developed faithfully recapitulates human PNF biology, with homozygous Nf1 deletion occurring in approximately half of SCs and driving formation of benign neurofibromas adjacent to the spinal cord, predominantly around the cervical plexus (25). This model’s quantifiable tumor burden, measured through volumetric analysis of sequential magnetic resonance imaging (MRI), provides a robust platform for evaluation of treatment effects (5). The translational value of this system was demonstrated when preclinical MEK inhibitor efficacy in Dhh-Cre;Nf1^fl/fl mice successfully predicted clinical benefit (26), ultimately leading to FDA approval of the MEK inhibitors selumetinib and mirdametinib for pediatric and adult patients with PNF, following demonstration of tumor volume reduction in 40 to 70% of treated individuals (27–29). Despite this therapeutic breakthrough, substantial clinical challenges remain: 30% of patients show no response to MEK inhibition, treatment-related toxicities can be dose limiting, tumor shrinkage rate is rarely greater than 20%, and tumor control requires continuous therapy with potential for regrowth upon discontinuation (28, 30). These limitations highlight a critical need for alternative therapeutic strategies or combination approaches that could enhance response rates and durability or reduce toxicity. The recent development of KRAS-selective inhibitors that block nucleotide exchange on both WT and mutant KRAS, but spare HRAS and NRAS, presents an intriguing therapeutic opportunity (31). These KRAS^multi inhibitors, including the tool compounds BI2493 and BI2865, preferentially bind GDP-bound KRAS and are orally bioavailable compounds that demonstrate efficacy in xenografts with WT KRAS amplification (32). A clinical KRAS^multi compound, BI-3706674 (BI6674), has a similar mode of action and entered clinical development in 2023 (NCT06056024). KRAS^multi inhibitors require ongoing GTP-GDP exchange, as they bind GDP-bound KRAS and block nucleotide exchange by preventing interaction with guanine nucleotide exchange factors, effectively trapping KRAS in the inactive GDP-bound conformation. This kinetic trapping mechanism requires ongoing KRAS cycling. Given that RAS paralogs are not mutant in NF1^−/− cells, we hypothesized that this class of compounds might prove effective in neurofibromas. However, whether KRAS is necessary or sufficient for benign neurofibroma initiation and whether targeting KRAS might demonstrate therapeutic efficacy in vivo remain unknown.

Here, we used complementary genetic and pharmacological approaches in mice to dissect RAS paralog–specific contributions to PNF pathogenesis. Using genetically engineered mouse models with deletion of KRAS or HRAS and testing a selective KRAS inhibitor, we demonstrate that KRAS plays a predominant role in PNF initiation in vivo, when SCs lack NF1. Pharmacological KRAS inhibition reduced neurofibroma burden and modulated specific immune cell populations within the tumor microenvironment. These findings establish KRAS as a critical therapeutic target in PNF, supporting the clinical evaluation of KRAS-selective inhibitors in patients with neurofibromatosis and PNFs, potentially offering an alternative therapeutic option to current MEK inhibition for neurofibroma treatment.

RESULTS

KRAS, but not HRAS, is essential for mouse PNF formation

In the Dhh-Cre;Nf1^fl/fl neurofibroma mouse model, Nf1 is deleted in SC lineage precursors, resulting in paraspinal neurofibroma development in all mice; neurofibromas are characterized by increased cell proliferation, mast cell and macrophage accumulation, and loss of axon-SC interaction in Remak bundles, all pathologic hallmarks of human PNFs (25). Classical RAS paralogs [HRAS, NRAS, KRAS (HRAS/NRAS/KRAS)] are expressed in both WT and NF1-deficient SCs, but only KRAS and HRAS become activated (GTP-bound) upon exposure to serum plus the SC mitogen neuregulin (20). To determine the role of RAS paralogs in neurofibroma formation, we first investigated HRAS by crossing this model with Hras global knockout (Hras^−/−) mice (fig. S1A). HRAS deletion did not extend survival time of tumor-bearing mice (fig. S1B), consistent with the presence of numerous PNFs on gross section of spinal cords (fig. S1C) and disrupted Remak bundles in Dhh-Cre;Nf1^fl/fl and Dhh-Cre;Nf1^fl/fl;Hras^−/− mice (fig. S1, D and E).

Next, we investigated the role of KRAS in neurofibroma formation. We deleted Kras specifically in the Dhh-positive SC lineage (Fig. 1A). Confirming effective deletion of KRAS, dorsal root ganglia (DRG) lysates from Dhh-Cre;Nf1^fl/fl;Kras^fl/fl contained markedly reduced levels of KRAS compared with those from Dhh-Cre;Nf1^fl/fl;Kras^WT (Fig. 1B). Moreover, in tissue sections of DRG from these mice, phospho-ERK (pERK) levels in Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice were substantially lower as compared to sections from Dhh-Cre;Nf1^fl/fl mice (Fig. 1C). Gross examination of DRG dissected from these mice revealed that only one of seven Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice developed a single tumor, substantially fewer neurofibromas than Dhh-Cre;Nf1^fl/fl mice (Fig. 1, D and E), demonstrating that KRAS is critical for neurofibroma initiation. Survival time of Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice was comparable to that of non-neurofibroma–bearing littermates (Dhh-Cre;Nf1^fl/+; P = 0.103), with extended survival compared to neurofibroma-bearing controls (Dhh-Cre;Nf1^fl/fl; P = 0.0015) (Fig. 1F).

Fig. 1. KRAS is critical for neurofibroma formation in Dhh-Cre;Nf1^fl/fl mice.

(A) Schematic showing Dhh-Cre;Nf1^fl/fl and Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice. (B) Immunoblotting analysis of NF1 and RAS paralogs, with β-tubulin as a loading control, in DRG and nerves from mice of the indicated genotype. (C) Representative images of pERK immunofluorescence staining in DRG and nerve sections from Dhh-Cre;Nf1^fl/fl and Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice. Arrowheads indicate TdTOM⁺ cells with pERK1/2 signaling. (D) Representative images of gross dissections of spinal cord and associated DRG and nerves in Dhh-Cre;Nf1^fl/fl and Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice. White arrowheads indicate PNFs. Quantification of tumor number per mouse for each genotype is shown at right. Student’s t test was used. Data are shown as means ± SEM. (E) Fisher’s exact test comparing tumor incidence between Dhh-Cre;Nf1^fl/fl and Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice. (F) Kaplan-Meier survival analysis of mouse cohorts: WT (n = 29), Dhh-Cre;Nf1^fl/fl (n = 30), Dhh-Cre;Kras^fl/fl (n = 16), and Dhh-Cre;Nf1^fl/fl;Kras^fl/fl (n = 22). Log-rank tests was performed for statistical analysis. ns, not significant; **P < 0.01, and ****P < 0.0001.

We next examined the effects of KRAS loss on the neurofibroma tumor microenvironment. KRAS deletion in tumorigenic NF1-deficient SCs reduced the percentage of Ki-67–positive proliferating cells in neurofibromas (Fig. 2, A and B). The percentage of macrophages, but not mast cells, was also markedly reduced in the Dhh-Cre;Nf1^fl/fl;Kras^fl/fl group compared with the Dhh-Cre;Nf1^fl/fl group (Fig. 2, C and D), suggesting a regulatory role of KRAS for SCs in macrophage infiltration. The percentage of disrupted Remak bundle (one to two bundles per axon) in Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice was rescued; it was comparable to nontumor-bearing Dhh-Cre;Nf1^fl/+ control littermates and lower than Dhh-Cre;Nf1^fl/fl mice (Fig. 2, E and F). Thus, KRAS deletion in tumorigenic NF1-deficient SCs rescues neurofibroma pathology.

Fig. 2. Kras loss in NF1-deficient SCs rescues pathologic features of PNF.

(A) Representative image of Ki-67 immunohistochemistry staining of nerve tissue sections from Dhh-Cre;Nf1^fl/fl and Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice. (B) Bar graph shows percent of Ki-67⁺ cells per high power field (HPF; ×400 magnification). (C) Representative images of hematoxylin and eosin (H&E), toluidine blue (Tol blue; mast cells), Iba-1, and F4/80 (macrophages) staining of nerve tissues from Dhh-Cre;Nf1^fl/fl and Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice. (D) Bar graphs show quantification of metachromatic toluidine blue⁺ cells (mast cells) and percentage of Iba-1⁺ and F4/80⁺ macrophages per HPF. (E) Representative electron micrographs showing ultrastructure of saphenous nerves from WT, Dhh-Cre;Nf1^fl/fl, Dhh-Cre;Kras^fl/fl, and Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice. Arrowheads indicate disrupted Remak bundles. (F) Bar graphs show the percentage of grouped axons. Six or more axons per bundle indicates a normal Remak bundle. For (A) to (D): nonparametric Student’s t test; n = 3 to 4 mice per genotype; each dot reflects data from an individual mouse (average of four fields per sample). For (F), two-way ANOVA with multiple comparisons test. Data are means ± SEM of biological replicates. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

In addition to these loss-of-function studies, we conditionally expressed an activated Kras^G12D in SCs (fig. S2, A and B). Dhh-Cre;Kras^G12D mice were viable and fertile and did not show altered survival time compared to WT littermates (fig. S2C). On examination of spinal cords and associated nerves, no tumors were identified (fig. S2D). Taking advantage of a lineage reporter (Dhh-Cre;LSL-Tomato) in SCs, we verified elevated pERK in the mutant nerves, confirming expression of the Kras^G12D allele (fig. S2E). Cell proliferation remained low in these Kras^G12D mice, as in WT mice (fig. S2F), as did the number of S100b-positive SCs (fig. S2G). No increase in nerve immune cells was detected (fig. S2H). Thus, activation of the Kras^G12D allele, unlike loss of Nf1, does not cause SC tumor formation. We wondered whether this was due to expression of this strong oncogenic allele inducing cell senescence or cell death, as different levels of KRAS activation can trigger distinct cellular responses in different cell types (33). However, we did not detect increased senescent cells at end point, and lineage-traced (Tomato⁺) cells were present at normal levels. We also examined multiple tissues and organs in Dhh-Cre;Kras^G12D mice, including nerve roots, DRG, and peripheral nerves. No hypertrophy, hyperplasia, or tumor formation was observed, even in aged mice (25 to 30 months of age).

Collectively, these findings demonstrate that loss of NF1 in SCs drives neurofibroma formation and that concurrent loss of KRAS, but not HRAS, prevents tumor development. These results establish KRAS as a major contributor to neurofibroma initiation. On the basis of these findings, we wondered whether KRAS might also contribute to tumor maintenance and whether KRAS inhibitors might therefore be effective.

KRAS^multi inhibitors suppress RAS activation, sphere formation, and proliferation of mouse NF1-deficient SCs in vitro

To assess the potential utility of KRAS^multi inhibitors in NF1-deficient SCs, we tested whether the inhibitors BI6674 or BI2865 affect RAS activation in primary Nf1^−/− mouse SCs derived from global Nf1-knockout embryonic day 12.5 (E12.5) embryos and exposed to serum plus the SC mitogen neuregulin for 5 min (Fig. 3A). Activated RAS-GTP levels of HRAS/NRAS/KRAS in Nf1^−/− SCs were monitored by RAF1-RBD (RAS-binding domain of RAF1) pull-down assays, followed by detection with RAS paralog–specific antibodies. BI2865 (500 nM) preferentially reduced KRAS-GTP, consistent with previous findings (32). While BI6674 and BI2865 are both potent pan-KRAS inhibitors, their activity on other isoforms has been observed when used at higher concentrations or in specific cell lines (32). In Nf1^−/− SCs, pretreatment with the clinical candidate BI6674 (500 and 2000 nM) effectively diminished both KRAS-GTP and HRAS-GTP levels. Consistent with RAS inhibition, pERK levels were reduced by both KRAS^multi inhibitors; total HRAS/NRAS/KRAS protein levels and pAKT levels were not substantially changed. We next used sphere formation assays using E12.5 Nf1^−/− SC precursors to assess the effect of KRAS^multi inhibitors on these cells, which show limited self-renewal and form neurofibromas in vivo (34). KRAS^multi inhibitors reduced sphere formation by 60 to 70%, akin to the effect of MEK inhibitor (Fig. 3, B and C). We also tested the effects of KRAS^multi inhibitors on proliferation of Nf1^−/− mouse SCs. Treatment with KRAS^multi inhibitors BI6674 and BI2865 reduced the viability of Nf1^−/− SCs by ∼20% (Fig. 3D), as do MEK inhibitors in this assay system. These results demonstrate that KRAS^multi inhibitors effectively suppress RAS activation in Nf1^−/− SCs, reduce their proliferation, and limit their tumorigenic potential in in vitro assays.

Fig. 3. KRAS inhibitors suppress RAS activation, reduce sphere formation, and inhibit proliferation of NF1-deficient mouse SCs.

(A) RAF1-RBD pull-down assays detecting active RAS paralogs in Nf1^−/− mouse SCs. Cells were pretreated with KRAS inhibitors for 2 hours, stimulated for 5 min, and subjected to RAF1-RBD pull-down assays. Data are representative of two biological replicates. (B) Sphere formation assays using E12.5 DRG-derived SC precursor in the presence of vehicle, KRAS^multi inhibitors, or MEK inhibitor (PD0395201) (n = 4 per condition, biological replicates). (C) Quantification of sphere formation in (B). (D) CCK-8–based cell viability assay of Nf1^−/− SCs treated with vehicle (DMSO), KRAS^multi inhibitors, or MEK inhibitor (PD0395201) (n = 4 per condition, biological replicates). For (B) to (D), two-way ANOVA with multiple comparisons test. Data are means ± SEM. *P < 0.05, ***P < 0.001, and ****P < 0.0001.

Single-agent KRAS^multi inhibitor BI6674 shrinks neurofibromas

Given these potent effects in NF1-deficient cells, we conducted preclinical testing in tumor-bearing Dhh-Cre;Nf1^fl/fl mice. We first assessed the pharmacokinetics of BI6674 in nontumor-bearing adult Nf1^fl/fl mice (Dhh-Cre negative) to guide treatment strategy. Plasma concentrations were dose dependent, reaching peak levels 1 to 2 hours after dosing and maintaining detectable levels at 24 hours after a second dose at 6 hours (Fig. 4A and tables S1 and S2), indicating a relatively long half-life. On the basis of these findings, we administered drug twice daily. To verify on-target effects of KRAS^multi inhibitor BI6674, we evaluated RAS/MEK pathway inhibition using protein and transcriptional readouts. Although pERK levels in vehicle-treated tumors showed interanimal variability, likely reflecting differences in cellular composition among individual neurofibromas and the dynamic regulation of pERK through tonic pathway feedback, pERK levels were reduced in the treatment group (100 mg/kg) compared to vehicle controls (Fig. 4B), consistent with target engagement. To provide a further measure of on-target RAS pathway inhibition, we quantified Dusp6 and Spry4, downstream transcriptional targets of the RAS/MEK pathway in this system (25). Expression of both transcripts was substantially decreased in the BI6674 treatment groups (30 and 100 mg/kg), similar to effects of the MEK inhibitor selumetinib (Fig. 4C), confirming on-target inhibition of the RAS/MEK pathway by BI6674 at both doses. We next evaluated efficacy by treating 7-month-old tumor-bearing Dhh-Cre;Nf1^fl/fl mice with vehicle or BI6674 (30 or 100 mg/kg twice daily) for 2 months (60 days). Stable body weights indicated that these doses were well tolerated (Fig. 4D). Tumor volumes, measured by serial MRI, increased in vehicle-treated mice but decreased in BI6674-treated mice: 6 of 12 mice in the group of 30 mg/kg and 9 of 13 mice in the group of 100 mg/kg showed tumor reduction (Fig. 4, E and F). Mixed-model analysis showed that both doses of BI6674 reduced tumor growth rate (both P < 0.0001), with no notable difference between them (P = 0.1867) (table S3). Thus, KRAS^multi inhibitor alone can achieve tumor regressions in established neurofibromas. Together with the genetic data demonstrating that KRAS deletion prevents neurofibroma formation (Figs. 1 and 2), these pharmacological findings support KRAS as a viable therapeutic target. However, because BI6674 inhibited HRAS-GTP in vitro, the in vivo tumor regression observed may not be attributable exclusively to KRAS inhibition.

Fig. 4. KRAS inhibition shrinks neurofibroma in tumor-bearing mice.

(A) Plasma concentrations of KRAS^multi inhibitor (BI6674) in mice following single-oral doses of 10, 30, or 100 mg/kg. The dotted line marks 6 hours after the initial dose, when a second dose was administrated for twice-daily dosing regimen. (B) Representative immunoblots of tumor lysates collected 4 hours after the final dose (after 60 days of twice-daily dosing) showing pERK, total ERK (tERK), pAKT (S473), total AKT (tAKT), and GAPDH (loading control) with densitometric quantification of pERK and pAKT normalized to tERK or tAKT, respectively, in control (vehicle), KRAS^multi BI6674 (30 mg/kg; KRAS30), and KRAS^multi BI6674 (100 mg/kg; KRAS100) treatment groups. (C) Relative expression of the MEK/ERK target gene Dusp6 and Spry4 measured by qRT-PCR in tumor tissue at study end point. (D) Body weight changes over the nine-week treatment period in control, KRAS30, and KRAS100 groups. Values are means ± SEM. No statistical difference was observed compared to control (P > 0.05), and no mice exhibited >20% weight loss. (E) Representative MRI images of neurofibromas in each treatment group (vehicle, KRAS30, KRAS100) before and after treatment. (F) Waterfall plots of individual tumor volume changes from months 7 to 9 in each treatment group (n = 12 for vehicle and KRAS30; n = 13 for KRAS100). Each bar represents a single mouse; bars below the x axis indicate tumor regression relative to baseline.*P < 0.05, **P < 0.01, ***P < 0.001.

Single-agent KRAS inhibition alters the neurofibroma immune microenvironment

The substantial tumor shrinkage with KRAS inhibition prompted investigation of effects on the complex neurofibroma immune microenvironment. We performed high-dimensional flow cytometry to evaluate how BI6674 (30 mg/kg twice daily) reshapes immune populations compared to the MEK inhibitor selumetinib (10 mg/kg twice daily), the current standard of care for neurofibroma (28). Sunburst plots and UMAP analysis revealed distinct immunomodulatory profiles between KRAS and MEK inhibition within the neurofibroma microenvironment (Fig. 5, A and B, and fig. S3A). While total immune cell numbers remained unchanged after 60 days of treatment (Fig. 5C), MEK inhibition decreased total tumor-associated macrophages (TAMs; CD45⁺,TCRβ⁻, CD11b⁺, side-scatter^low, Ly6G⁻, Ly6C⁻, CD14^hi, CD172a^hi, and CD163⁺), whereas KRAS inhibition paradoxically increased macrophage activation markers CD44 and CD86, suggesting enhanced antigen presentation capacity (Fig. 5D). KRAS inhibition also selectively reduced DCs (DC1 and DC2) (CD45⁺, TCRβ⁻, CD11c⁺, and MHCII⁺) (Fig. 5E) and increased neutrophilic myeloid-derived suppressor cells (nMDSCs; CD45⁺, TCRβ⁻, CD11b⁺, side-scatter ^low, Ly6C⁺, and Ly6G⁺) (Fig. 5F). The finding that KRAS and MEK inhibitions exert complementary rather than redundant effects on immune cell populations provided a rationale for combination therapy.

Fig. 5. Immunophenotyping of the tumor microenvironment in neurofibroma treated with selumetinib or KRAS^multi BI6674.

(A) Sunburst plots show the distribution of major immune cell populations among total live cells (40,000 cells per condition) in tumors from vehicle-treated (vehicle), MEK inhibitor–treated (selumetinib), or KRAS^multi inhibitor–treated (BI6674) mice. Populations include T cells (TCRβ⁺), conventional DC subsets (cDC1 and cDC2), CD163⁻ and CD163⁺ macrophages (MΦ), MDSCs, and monocytes. (B) UMAP dimensionality reduction plots illustrating cell clustering based on surface marker expression for each treatment group (vehicle, selumetinib, and BI6674). Clusters are annotated by color and labeled for T cells, cDC1, cDC2, and MΦ, and MDSCs. Right panels show density plots of cells across the UMAP space. (C) Quantification of immune subsets in vehicle-, selumetinib-, and BI6674-treated mice. Bar graphs show the percentage of CD45⁺ cells among live cells and the frequency of TCRβ⁺ T cells, DCs, MΦ, and MDSCs among CD45+ cells. (D and E) Quantification of CD163⁻ MΦ (TAMs) (D), and DCs and MDSCs (E) populations. Bars depict the percentage of each subset among CD45⁺ cells and expression of activation markers (CD44 and CD86). Each dot represents data from an individual mouse. Data are means ± SEM. *P < 0.05 and ***P < 0.001.

Combined KRAS and MEK inhibition yields enhanced neurofibroma regression

To test whether dual inhibition of KRAS and MEK shows improved efficacy, we treated tumor-bearing mice with vehicle, BI6674 (30 mg/kg), selumetinib (10 mg/kg), or the combination (BI6674 plus selumetinib) for 2 months. We also performed an in-group assessment of selumetinib monotherapy efficacy (n = 10) in the current study. For statistical analysis, these data were combined with 14 additional data points from a previous study of selumetinib using the same dose and schedules (35). Combination therapy achieved the highest response rate, with 81% of mice (13 of 16) showing tumor volume reduction compared to 44% with BI6674 alone (7 of 16) and 58% with selumetinib alone (14 of 24) (Fig. 6A). Combination treatment produced greater tumor volume reduction than BI6674 monotherapy (P = 0.013), demonstrating enhanced efficacy with dual pathway inhibition (table S3). Mechanistically, both BI6674 monotherapy and combination therapy decreased cell proliferation (Ki-67⁺ cells) (Fig. 6B) while also reducing apoptosis (cleaved caspase-3⁺ cells) (Fig. 6C), suggesting cytostatic rather than cytotoxic effects.

Fig. 6. Dual inhibition of KRAS and MEK reduces neurofibroma growth and alters the tumor microenvironment.

(A) Waterfall plot showing percent change in tumor volume (months 7 to 9) in vehicle-treated mice, mice treated with BI6674 alone (30 mg/kg), selumetinib alone (10 mg/kg), or combination BI6674 (30 mg/kg) plus selumetinib (10 mg/kg). Each bar represents data from a single mouse; bars below the x axis indicate tumor regression from baseline. Statistical analysis is in table S3. (B and C) Quantification of proliferating cells (Ki-67⁺ percentage) (B) and apoptotic cells [cleaved caspase-3⁺ (CC3⁺) percentage] in tumor sections at study end point. (D) Sunburst plots depicting the distribution of major immune cell populations among live cells (56,000 cells per condition) in tumors from vehicle-, BI6674-, or combination (BI6674 + selumetinib)–treated mice. Populations include T cells (TCRβ⁺), conventional DC subsets (cDC1 and cDC2), macrophages (MΦ), and MDSCs. (E) UMAP projections illustrate cell clustering by treatment condition based on surface marker expression. Left panel shows clusters annotated by color and labeled for MΦ (CD163^+/− MΦ), DC subsets (cDC1 and cDC2), T cells, and MDSCs. Right three panels show density maps depicting overall cell distributions for each treatment. (F) Quantification of immune subsets in vehicle-, BI6674-, or combination (BI6674 + selumetinib)–treated mice. Bar graphs show the percentage of CD45⁺ cells among live cells and the frequency of TCRβ⁺ T cells, DCs, MΦ, and MDSCs among CD45⁺ cells. (G and H) Quantification of activation markers and subset frequencies within TAM and DC compartments. (G) Bars show the percentage of CD163- MΦ, and coexpression of CD44 and CD86 in MΦ, and (H) cDC1/cDC2 subsets among CD45⁺ cells that express the activation marker CD86. Data are means ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Flow cytometry analysis revealed that combination therapy induced more profound immune remodeling than BI6675 monotherapy (Fig. 6, D and E, and fig. S3B). Dual inhibition of KRAS and MEK, compared with KRAS inhibition alone, reduced overall immune infiltration while selectively enriching for MDSCs and T cells (Fig. 6F). At the cellular level, combination therapy depleted CD163⁻ macrophages while activating remaining macrophages (increased CD44 and CD86 expression) and diminished DC frequencies with reduced activation markers (Fig. 6G). These results demonstrate that dual KRAS/MEK inhibition creates a fundamentally altered immune landscape that may contribute to enhanced therapeutic efficacy, establishing proof of concept for rational combination approaches.

DISCUSSION

This study establishes KRAS as a critical driver of neurofibroma formation and validates KRAS as a potential therapeutic target in established tumors. Through genetic loss-of-function studies, we demonstrate that KRAS deletion, but not HRAS deletion, extends survival and prevents neurofibroma development in most Dhh-Cre;Nf1^fl/fl mice. Pharmacological treatment using the KRAS^multi inhibitor BI6674 confirmed that KRAS represents a viable therapeutic target, achieving meaningful tumor regression in established neurofibromas. Furthermore, KRAS inhibitors produced profound effects on immune cells in the immune microenvironment. These findings provide compelling preclinical evidence for KRAS inhibition as a previously unidentified treatment strategy for NF1-associated neurofibromas.

These results position KRAS inhibition as a potential alternative strategy to MEK inhibition, the current standard of care for PNFs. While both approaches target the RAS/MEK pathway, KRAS inhibitors may offer advantages. Our genetic data establish KRAS as the functionally dominant RAS paralog in neurofibroma formation, indicating that selective KRAS inhibition can block tumor formation while sparing signaling through other RAS paralogs (HRAS and NRAS) required for normal growth and development. An inhibitor that is KRAS specific may therefore have a wider therapeutic window compared with MEK inhibitors that suppress ERK activation downstream of all RAS paralogs. If so, then KRAS and MEK inhibitors may exhibit different toxicity profiles. MEK inhibitors carry well-characterized class-effect toxicities including ocular manifestations (chorioretinopathy and retinal vein occlusion) and cardiac effects (reduced left ventricular ejection fraction), which can be dose limiting (36). Early clinical data for RAS(ON) inhibitors, including daraxonrasib (RMC-6236), have shown a toxicity profile dominated by rash and gastrointestinal effects without prominent ocular or cardiac signals (37), suggesting that RAS-targeted approaches may be better tolerated in patients who cannot continue MEK inhibition due to these specific toxicities. In addition, the specific immunomodulatory effects of KRAS inhibition, which differ from those of MEK inhibitors, suggest that distinct mechanisms of therapeutic benefit could be leveraged through rational combination approaches. Last, KRAS inhibitors may demonstrate efficacy in cases resistant to MEK inhibition, providing an alternative therapeutic option for nonresponding patients.

The differential requirements for KRAS versus HRAS in neurofibroma initiation highlight the nonredundant roles of RAS paralogs in cancer biology. While both KRAS and HRAS are activated in NF1-deficient SCs, only KRAS proves essential for tumor formation. A caveat of this comparison is that the KRAS and HRAS genetic models used different strategies: a conditional Kras-floxed allele (SC-specific deletion) versus a global Hras knockout, the latter of which could allow developmental compensatory mechanisms that mask a SC-autonomous role for HRAS. However, complementary studies using individual knockdown of KRAS, HRAS, and NRAS in human NF1^−/− SCs (NF1^−/− iHSC1λ) confirmed that only KRAS knockdown impaired anchorage-independent growth in vitro and nearly abolished xenograft tumor formation in vivo (20). Collectively, these genetic studies across species and experimental systems establish KRAS as the functionally dominant RAS paralog in NF1 loss–driven neurofibroma formation.

This paralog-specific function of RAS proteins parallels observations in other cancer types, in which KRAS and HRAS exhibit distinct roles in tumor initiation and progression (38, 39). Similar nonredundant functions have been reported for other RAS/MAPK family members, including RAF1/BRAF/ARAF and ERK1/ERK2, which display paralog-specific roles in cell proliferation and differentiation (40, 41). These findings establish KRAS as a necessary component of NF1-deficient neurofibroma pathogenesis and provide a mechanistic rationale for KRAS-selective therapeutic targeting.

KRAS activation alone in SCs was insufficient to drive neurofibroma formation, despite confirmed downstream ERK activation. This failure of KRAS^G12D to induce neurofibroma contrasts sharply with NF1 loss, which reliably generates tumors in the same cellular context. Several nonmutually exclusive mechanisms may explain this requirement for NF1 loss beyond KRAS activation. First, NF1 loss activates multiple RAS paralogs simultaneously, including HRAS and KRAS, potentially creating a more complex RAS signaling network that single KRAS activation cannot replicate. Second, RAS-independent pathways are activated by NF1 loss, but not by KRAS^G12D, and these pathways may play critical roles in tumorigenesis. NF1 loss decreases cAMP in cell types including SCs (42). Third, the level, mode, and duration of KRAS activation may differ between constitutive loss of the KRAS-GAP NF1 and expression of a single KRAS^G12D mutant. NF1 loss generates stimulus-dependent, physiologically modulated KRAS activation, whereas KRAS^G12D produces strong, constitutive signaling that trigger oncogene-induced senescence or cell death in certain cell types (43, 44), potentially limiting the tumorigenic capacity of this allele in SCs. A simple increase in WT KRAS expression would also be unlikely to drive neurofibroma formation, as intact RAS-GAP regulation, including NF1, RASA1, and other GAPs, would be expected to efficiently attenuate any increase in KRAS-GTP levels. Moreover, NF1 loss activates multiple RAS paralogs simultaneously and engages RAS-independent pathways, creating a broader signaling context that elevated KRAS activity alone cannot recapitulate. Increasing evidence indicates that mutant and WT KRAS alleles exhibit distinct biological properties, including differences in pathway activation signatures and tissue-specific effects (43), underscoring that the consequences of KRAS activation are allele and context dependent. Several cases of SC tumors were reported in individuals harboring KRAS mutations but lacking NF1 mutations (45–48). These reported human cases described KRAS^G12D mosaicism, associated with hypertrophic neuropathy, nerve root enlargement, and dermal neurofibromas and epidermal nevi. These features might result from mutations in SCs and/or other cell types. In addition, species-dependent differences in the response of SCs to oncogenic KRAS activation cannot be excluded. Further studies are needed to address this issue.

The contrasting outcomes between genetic KRAS deletion and pharmacological KRAS inhibition provide important insights into tumor biology and therapeutic limitations. While genetic KRAS ablation resulted in near-complete blockade of neurofibroma formation, pharmacological KRAS inhibition achieved tumor regression in 50 to 69% of established neurofibromas comparable to the 40 to 70% response rates observed with FDA-approved MEK inhibitors in clinical trials. Several factors may explain this difference. First, genetic deletion achieves complete KRAS loss, whereas small-molecule inhibitors typically achieve about 70 to 90% inhibition of RAS/MEK signaling (49, 50), leaving residual KRAS activity that may sustain the survival of tumor cells. Second, KRAS may play differential roles in tumor initiation versus maintenance of neurofibromas, as suggested by studies in other cancer models (51, 52). The differing effects on macrophages observed between genetic KRAS deletion and pharmacological KRAS inhibition likely reflect differences in methodology, the macrophage populations assessed, and the biological context of each experiment. In the KRAS knockout model, macrophage abundance was assessed using immunohistochemistry for the pan-macrophage markers Iba-1 and F4/80, which broadly label resident and infiltrating macrophage populations; both markers showed reduced macrophage numbers in KRAS-deleted mice (Fig. 2C). In contrast, macrophage subsets in BI6674-treated tumors were characterized by high-dimensional flow cytometry (Fig. 5D), in which the total CD163⁻ macrophage population did not change, but the CD44⁺CD86⁺CD163-activated TAM subset increased, indicating a shift in macrophage polarization. These markers identify distinct populations: Iba-1 and F4/80 detect macrophages broadly, whereas CD44⁺CD86⁺CD163⁻ gating identifies a more specific monocyte-derived TAM subset enriched in tumors (53). Furthermore, complete genetic ablation of KRAS from the SC lineage during development likely reduces the paracrine signals that recruit macrophages to the tumor microenvironment, whereas partial pharmacological inhibition of KRAS in established tumors may preferentially alter macrophage polarization and activation state without eliminating already-resident populations. Third, as well as relevant for future studies, intrinsic resistance to KRAS inhibition can occur upon chronic treatment. In the setting of neurofibroma, although combining MEK and KRAS inhibitors enhanced the percent of tumors showing regression and induced broader changes in the tumor immune microenvironment compared with KRAS inhibition alone (Fig. 6), the magnitude of improvement may not justify the potential increased risk of overlapping toxicities associated with dual pathway inhibition. Moreover, vertical pathway inhibition aimed at overcoming intrinsic resistance may be insufficient if tumor cells rewire through RAS/MEK-independent bypass pathways (54), including receptor tyrosine kinase up-regulation or YAP signaling (52, 55). YAP signaling activation has been implicated in NF1-deficient neurofibromas (56) so that resistance of neurofibroma to KRAS inhibition might involve this pathway. Overall, this intrinsic resistance suggests that combinations of other pathway inhibitors with KRAS inhibitors will be necessary to cure neurofibromas.

The effects of KRAS inhibition extend beyond direct tumor cell targeting to encompass notable remodeling of the neurofibroma immune microenvironment. Both KRAS (data shown here) and MEK inhibitions (53) altered immune cell populations in neurofibromas, consistent with previous findings in other cancers (49, 57). The overall therapeutic efficacy of KRAS^multi in neurofibroma likely reflects the drugs’ combined impact on both tumor cells and stromal components, particularly immune cells within the microenvironment. Our finding that KRAS^multi increases neurofibroma T cells and decreases DCs and myeloid cells is consistent with similar immunomodulatory effects of KRAS inhibition in models of pancreatic cancer (58). Further studies should clarify the relative contributions of tumor cell–intrinsic versus immune-mediated effects to overall therapeutic efficacy of KRAS inhibition in neurofibroma treatment. While combination approaches with immunotherapies, such as checkpoint inhibitors may be inadvisable in the setting of benign neurofibroma, small-molecule immune agonists/antagonists may be useful to enhance treatment responses in neurofibroma.

Several considerations limit this study. We used a mouse model that, while highly predictive of clinical MEK inhibitor efficacy, may not fully recapitulate all aspects of human neurofibroma biology. Second, optimal dosing, schedule, and duration of KRAS inhibition in patients remain to be determined through clinical trials. Third, longer-term studies are needed to assess potential for resistance development. Fourth, the safety and tolerability of KRAS inhibitors in pediatric populations, who comprise the majority of patients with neurofibroma requiring treatment, will require evaluation in age-appropriate clinical studies. Last, the pharmacological inhibitor BI6674 used in our in vivo efficacy studies also reduced HRAS-GTP levels in SCs in vitro, although it showed selectively toward KRAS in prior studies in other cell types (58). Thus, the therapeutic effects observed with BI6674 may reflect combined inhibition of KRAS and HRAS rather than KRAS inhibition alone. Future studies using more selective KRAS inhibitors with improved in vivo properties will be needed to definitively attribute the pharmacological efficacy to KRAS-specific inhibition.

We conclude that a KRAS^multi inhibitor is effective in shrinking neurofibromas, with effects persisting for at least 60 days. Recent advancements in targeting oncogenic KRAS have recently stimulated intense interest in the development of additional types of molecules with the potential to target RAS. Those expected to target the WT RAS present in NF1 mutant cells include KRAS degrader molecules (59, 60) and some RAS(ON) molecules, which preferentially inhibit RAS in the GTP-bound state (61). Notably, a multi-RAS(ON) inhibitor, RMC-7977, showed effect in an NF1 mutant MPNST xenograft model, although the effect was limited and transient (62). It will be of interest to test these multiple agents for efficacy in neurofibroma, given the substantial ability of KRAS^multi to shrink neurofibromas.

MATERIALS AND METHODS

Animal studies

All animal procedures were conducted following the protocol approved by the Institutional Animal Care and Use Committee of CCHMC (protocol nos. 20180103 and 20210075). Mice were housed in a temperature and humidity-controlled vivarium on a 12-hour dark-light cycle with free access to food and water. The Dhh-Cre;Nf1^fl/fl mouse line described previously (25) was used for all experiments with both sexes, and Dhh-Cre was maintained in the male breeder. For Hras loss-of-function studies, fertile and viable Hras-null (Hras^−/−) mice (63) were used; Hras^−/− mice were obtained from Allan Balmain (UCSF) and intercrossed with Dhh-Cre;Nf1^fl/+ and with Nf1^fl/fl mice to generate double Dhh-Cre;Nf1^fl/fl;Hras^−/− mice. For Kras loss-of-function studies, given the embryonic lethality of Kras^−/− mice (64), we used the Kras-flox (Kras^fl/+) line (65), which carries two loxP sites flanking Kras exon 1, enabling Kras-specific deletion in a Cre⁺ lineage. Kras^fl/+ frozen embryos on a mixed 129vJ/C57BL/6 genetic background were obtained from J. Zhang (University of Wisconsin-Madison), and Kras^fl/+ mice were generated. Kras^fl/+ mice were intercrossed to generate homozygous Kras^fl/fl animals. Dhh-Cre;Nf1^fl/fl;Kras^fl/fl mice were generated by crossing Kras^fl/fl females to Dhh-Cre;Nf1^fl/+ male mice, followed by intercrosses of offspring. For genotyping, we used EconoTaq PLUS Master Mix (Sigma-Aldrich, #LGC300331); primers are listed in table S4. To control for genetic background, littermate controls were analyzed in each cross. Mice of both sexes were used in all experiments, and the experiment was done once.

Drug treatment

Mice were acclimated to handling and then administered selumetinib (10 mg/kg; AZD6244, Selleck Chemicals, Houston, TX) twice daily by oral gavage, formulated in 0.5% (w/v) methyl cellulose E-50 with 0.2% (v/v) polysorbate-80 (Tween 80). The KRAS inhibitor BI6674 (provided by Boehringer Ingelheim) was administered twice daily at 30 or 100 mg/kg, suspended in 0.5% Natrosol and 5% hydroxyethyl cellulose (250HX). For combination treatment [selumetinib (10 mg/kg) plus BI6674 (30 mg/kg)], the agents were each prepared at 2× concentration and mixed immediately before dosing, so that all groups received the same volume per weight (10 ml/g of mouse weight). Animals were monitored in accordance with IACUC regulations, with euthanasia required if body weight decreased by more than 20% or morbidity was observed. We randomized tumor-bearing mice to different study groups for tumor treatment. Sample size determination was based on prior analysis in this model (66). Volumetric measurements were carried out by investigators blinded to treatment groups. No mice were excluded from analysis. Each therapeutic study was carried out once.

Cardiac puncture/tumor storage and fixation

Four hours after final dose administration, mice were placed into an isoflurane-filled chamber until breathing ceased. Blood was collected into a 1.5-ml EDTA tube after cardiac puncture and then placed on ice for 30 to 60 min. Plasma was extracted after centrifugation at 4°C for 10 min at 15,000g and stored at −80°C until use. Paraspinal tumors were resected after cardiac puncture, and half of the tumors from each mouse were either flash frozen in liquid nitrogen or placed in 4% paraformaldehyde for 1 hour at room temperature for histological processing. Time from cardiac puncture to tumor removal was under 10 min for each mouse.

Pharmacokinetic analysis

BI6674 and selumetinib were quantified in mouse plasma using BIBI1355BS as internal standard and high-performance liquid chromatography coupled to tandem mass spectrometry (SCIEX QTRAP 6500 Triple Quadrupole). Plasma proteins were removed from the samples by protein precipitation with 170 μl of 80% acetonitrile. Chromatography was performed using a XBridge BEH C18 column (2.5 μM, 2.1 mm by 50 mm Column XP, Waters). Mobile phase A consisted of 5 mM ammonium acetate (pH 4) in water and mobile phase B of 0.1% formic acid in acetonitrile, and the following gradient program was run: 5 to 95% B in 1 min, followed by a column washing phase of 0.3 min at 95% B and then re-equilibration at 5% B. The injection volume was 1 μl, the flow rate was 0.7 ml/min, and the column temperature was set to 60°C. The MRM transitions Q1 to Q3 (positive ion mode) for BI6674 were 600.15/112.31, those for selumetinib were 457.0/301.1, and those for the internal standard BIBI1355BS were 467.30/98.20. The lower limit of quantification for BI6674 was 2.97 nM, and that for selumetinib was 5.88 nM. Samples for concentration measurement were taken in a composite design (group 1: 0.25, 2, and 6 hours; group 2: 1, 7, and 24 hours after first compound administration). Pharmacokinetic parameters including area under the curve were estimated (from average data of male animals) using noncompartmental methods with Phoenix WinNonlin software (Certara).

MRI and analysis

We imaged anesthetized mice using the 7T Brucker Biospec System to acquire images in three planes to position the three-dimensional volume (66). We calculated tumor volume from the area of graphic outlines and MRI slice thickness. To derive P values, we conducted a random-effects model analysis on the log-transformed tumor volume data using the SAS-mixed procedure as described (66).

Cell counting kit-8 viability assay

Cell viability was measured with the cell counting kit-8 (CCK-8) (APExBIO, #K1018) and the absorbance assay in the microplate reader (Molecular Devices, #FLEX316G). Cells (500 to 1000 per well) were seeded in 96-well flat-bottom plates (Corning, #353072) with four biological replicates. KRAS or MEK inhibitors were added to cells the following morning. Ninety-six hours later, 10 μl of CCK-8 stock solution was added to cells in 96-well plates, and the absorbance was read at 450 nm/650 nm (650 nm as the background signal) following the manufacturer’s instructions. The results represent the means ± SEM of biological replicates, expressed as a percentage of the control group after normalization.

SC sphere culture

Mouse SC spheres were generated as previously described (34). DRGs isolated from E12.5 mouse embryos were enzymatically dissociated with 0.25% Trypsin (Thermo Fisher Scientific, #25200056) for 3 to 5 min at 37°C, followed by mechanical dissociation using narrow-bore pipettes. The cell suspension was pelleted by centrifuged at 400g for 4 min. The cells were resuspended and cultured in ultralow attachment 24-well plates (Thermo Fisher Scientific, #07-200-602) in serum-free DMEM/F12 medium supplemented with 1× N2 supplement (Gibco, #17502048), EGF (20 ng/ml; R&D Systems, #236-EG-200), and bFGF (20 ng/ml; Peprotech, #450-33). The chemicals were added on day 1, and fresh medium were added every 3 to 4 days. The number of spheres was counted on day 10 posttreatment.

RAS pull-down assays

Active RAS paralog levels were determined using the RAS activation assay kit (Cell Biolabs, #STA-400) according to the manufacturer’s protocol, with modifications described below. Primary mouse embryonic SCs (Nf1^−/−) were cultured in 10-cm dishes and then serum-starved for 8 hours in serum-free DMEM/F12 medium. Where indicated, KRAS inhibitors (BI2865 or BI6647) were added to serum-starved cells 2 hours before stimulation for 5 min, with 15% FBS plus β-heregulin (25 ng/ml; Peprotech, #100-03) and then lysed in lysis buffer. Cell lysates were clarified by centrifuge: An aliquot representing 5% of the total lysate was saved as an input control, and the remainder was incubated with 30 μl of RAF1-RBD agarose beads at 4°C with end-over-end rotation. Following incubation, beads were washed with assay/lysis buffer, and bound proteins were eluted with 2× SDS sample loading dye, denatured at 95°C for 5 min, and analyzed by SDS-PAGE and immunoblotting using RAS paralog–specific antibodies (20).

Immunostaining

Mice were anesthetized with isoflurane and then received transcardiac perfusion with ice-cold 1× PBS. For staining paraffin sections, following perfusion, tissues were rapidly dissected, fixed in 10% neutral-buffered formalin for 24 hours, and embedded in paraffin. Paraffin blocks were sectioned at 4 to 5 μm, deparaffinized, rehydrated, and subjected to antigen retrieval using standard protocols. To reduce endogenous peroxidase activity, tissues were incubated with 0.3% hydrogen peroxide at room temperature for 10 min. Sections were then blocked with 5% normal goat serum to minimize nonspecific binding for 1 hour at room temperature. Primary antibody incubation occurred overnight at 4°C with antibodies listed in table S5. After incubation with species-specific biotinylated secondary antibodies, immunoreactivity was detected using a streptavidin-biotin peroxidase complex and visualized with 3,3′-diaminobenzidine as the chromogen. Slides were counterstained with hematoxylin before mounting with Histomount.

For frozen sections, after PBS perfusion, animals were perfused with 4% freshly prepared paraformaldehyde in PBS (pH 7.4), dissected, rinsed with PBS, and frozen in OCT. Frozen sections (10 μM) were cut, and OCT was removed by incubation with 1× PBS. We permeabilized sections in ice-cold MeOH for 10 min, followed by incubation in normal goat (005-000-121) or donkey (017-000-121) (Jackson ImmunoResearch) and 0.3% Triton X-100 (Sigma-Aldrich, catalog no. X100). Primary antibody incubation occurred overnight at 4°C, and then secondary antibodies were donkey or goat anti-rat/rabbit/goat from Jackson ImmunoResearch reconstituted in 50% glycerol and used at 1:200. To visualize nuclei, sections were stained with DAPI for 10 min, washed with PBS, and mounted in Fluoromount-G (Electron Microscopy Sciences, Hatfield, PA). Images were acquired with NIS-Elements software using confocal microscopy (Nikon). Images were analyzed using ImageJ. For toluidine blue staining of mast cells, tissues sections were incubated for 5 min in 0.5% toluidine blue solution (toluidine blue dissolved in acetate buffer) and then rinsed in water. Hematoxylin and eosin (H&E) staining was performed using the commercial kit (Thermo Fisher Scientific, 7221 and 7111), according to the manufacturer’s instructions. Four representative fields per tissue section were imaged for analysis. Ki-67–positive nuclei were quantified manually, while pERK staining intensity was assessed using relative semiquantitative values. Statistical analyses were performed to evaluate differences between experimental groups, with appropriate positive and negative controls included throughout.

Electron microscopy

Mice were perfusion fixed with 4% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M phosphate buffer at pH 7.4. Saphenous nerve was dissected and postfixed overnight, then transferred to 0.175 M cacodylate buffer, osmicated, dehydrated, and embedded in Embed 812 (Ladd Research Industries). Ultrathin sections were stained in uranyl acetate and lead citrate and viewed on a Hitachi H-7600 microscope.

Western blots

Tumor and lung tissue lysates were prepared in RIPA buffer with added protease and phosphatase inhibitors (Thermo Fisher Scientific, #A32961). Protein concentrations were determined using the BCA kit from Bio-Rad. Equal amounts of protein were loaded onto on 4 to 20% SDS-PAGE gels transferred onto polyvinylidene difluoride membranes and separated by electrophoresis. Membranes were then blocked with 5% nonfat milk in 1× TBS-Tween 20 (0.1%) and then incubated with primary antibodies overnight at 4°C, as per the manufacturer’s recommended protocol. Primary antibodies are provided in table S5. After washing, membranes were incubated with horseradish peroxidase–conjugated secondary antibodies and ECL detection.

Nerve tumor processing

After euthanasia, mice were briefly perfused intracardially with ice-cold PBS, and tissues were removed and stored in Opti-MEM at 4°C for 3 to 36 hours. For tissue dissociation, we prepared fresh enzyme mix [100 μl of collagenase A (100 mg/ml), 100 μl of collagenase type 4 (100 mg/ml), 40 μl of soybean trypsin inhibitor (100 μg/ml), 10 μl of deoxyribonuclease I (250 U), 5 μl of 1 M CaCl2 (5 mM final concentration), and 745 μl of complete RPMI 1640 (cold); 1 ml per sample per 1.5-ml tube]. Tissue was centrifuged at 1400 rpm for 5 min at 4°C, most media were removed, and tissue was minced into small pieces (∼1 mm³), then placed into the enzyme mix, placed on a prewarmed shaker, and rotated at 180 rpm for 30 min at 37°C. The cell suspension was transferred to a 50-ml conical tube, and enzymatic dissociation halted by adding ice-cold 30 ml of RPMI 1640 + 1% FBS. After gentle mixing and centrifugation at 1400 rpm for 5 min at 4°C, the supernatant was partially removed, leaving 3 to 5 ml. After adding ∼25 ml of RPMI 1640 + 1% FBS, we filtered the cell suspension through 100- and 70-μm cell strainers, centrifuged, removed supernatant, and resuspended the pellet in ∼1 ml for subsequent cell counting without dilution. One million cells per sample were transferred into standard 5-ml flow tubes and washed twice with 1 ml of PBS to eliminate residual serum. Subsequently, a live/dead marker (Thermo Fisher Scientific, catalog no. L34964 or L23105) was used to exclude nonviable cells, and cells were then incubated for 60 min at room temperature in a 100-μl antibody mixture as indicated in table S5. Unbound antibodies were then removed by washing the cells 2× with 1.5 ml of PBS. The cells were then fixed with freshly prepared 2% PFA in 1× PBS for 30 min, washed and resuspended in 350 μl in PBS for immediate analysis or stored in 500 μl of Cyto-Last Buffer (BioLegend, #422501) and then analyzed on a Cytos Aurora flow cytometer with a five-laser configuration (Cytek) within 10 days.

Flow cytometry and analyses

We used UltraComp eBeads^TM compensation beads (Thermo Fisher Scientific, catalog no. 01-2222-41) and Live-Dead Thermo Fisher Scientific ArC amine-reactive compensation bead kits (A10346) for single-color controls. Each tumor or bone marrow sample was run together with its own negative control (unstained) to control for endogenous fluorescence. Tumors from 3 to 11 mice were analyzed, as designated for each experiment. Datasets were downsampled to include the same number of events for each event for each tumor, using FlowJo software. UMAP analysis was performed in FlowJo with default settings of nearest neighbors and minimum distance. Cluster explorer was used to generate expression profiles with a cutoff of 99% ILE (outliers were removed), and means of scaled values were used to generate plots.

Statistical analysis

Statistical parameters, including the type of tests, number of samples (n), descriptive statistics, and significance, are reported in the figures and figure legends. All data were analyzed in GraphPad Prism 10. Fisher’s exact test was used for comparing neurofibroma volumes for each mouse at each time point (5, 7, and 9 months) and PNF incidence in mice. The SAS-mixed procedure (66) was used to compare volumetric changes in neurofibroma and the log-rank test for survival analysis. We performed unpaired two-tailed t tests for two group comparisons and one-way or two-way analysis of variance (ANOVA) for multiple-group comparisons. Data were reported as means ± SD (or means ± SEM) as indicated. P < 0.05 was considered significant.

Supplementary Materials

This PDF file includes:

Figs. S1 to S3

Tables S1 to S5