NF1 mutation-driven neuronal hyperexcitability sets a threshold for tumorigenesis and therapeutic targeting of murine optic glioma

Abstract

Background

With the recognition that noncancerous cells function as critical regulators of brain tumor growth, we recently demonstrated that neurons drive low-grade glioma initiation and progression. Using mouse models of neurofibromatosis type 1 (NF1)-associated optic pathway glioma (OPG), we showed that Nf1 mutation induces neuronal hyperexcitability and midkine expression, which activates an immune axis to support tumor growth, such that high-dose lamotrigine treatment reduces Nf1-OPG proliferation. Herein, we execute a series of complementary experiments to address several key knowledge gaps relevant to future clinical translation.

Methods

We leverage a collection of Nf1-mutant mice that spontaneously develop OPGs to alter both germline and retinal neuron-specific midkine expression. Nf1-mutant mice harboring several different NF1 patient-derived germline mutations were employed to evaluate neuronal excitability and midkine expression. Two distinct Nf1-OPG preclinical mouse models were used to assess lamotrigine effects on tumor progression and growth in vivo.

Results

We establish that neuronal midkine is both necessary and sufficient for Nf1-OPG growth, demonstrating an obligate relationship between germline Nf1 mutation, neuronal excitability, midkine production, and Nf1-OPG proliferation. We show anti-epileptic drug (lamotrigine) specificity in suppressing neuronal midkine production. Relevant to clinical translation, lamotrigine prevents Nf1-OPG progression and suppresses the growth of existing tumors for months following drug cessation. Importantly, lamotrigine abrogates tumor growth in two Nf1-OPG strains using pediatric epilepsy clinical dosing.

Conclusions

Together, these findings establish midkine and neuronal hyperexcitability as targetable drivers of Nf1-OPG growth and support the use of lamotrigine as a potential chemoprevention or chemotherapy agent for children with NF1-OPG.

lamotrigine, midkine, neuronal excitability, NF1, optic glioma

Key Points

• Neurofibromatosis type 1 (Nf1) mutation-driven neuronal midkine is necessary and sufficient for Nf1-optic pathway glioma (OPG) growth.

• Lamotrigine (LTR) durably prevents Nf1-OPG progression and suppresses established Nf1-OPG growth.

• Pediatric epilepsy LTR dosing inhibits Nf1-OPG growth in two preclinical models.

Importance of the Study

The authors demonstrate that neuronal excitability and midkine secretion dictate murine low-grade optic glioma (OPG) growth in a germline neurofibromatosis type 1 (Nf1) mutation-specific manner, suggesting that both neuronal activity and midkine expression could serve as biomarkers of OPG risk. Nf1-mutant neuronal hyperexcitability-induced midkine expression is selectively suppressed by lamotrigine, such that lamotrigine treatment prevents Nf1-OPG progression and suppresses established Nf1-OPG growth for months following treatment cessation. Importantly, lamotrigine inhibits Nf1-OPG growth at doses routinely administered to children with epilepsy, supporting its use in clinical trials for children with NF1-OPG.

Brain tumors, and specifically gliomas (astrocytomas), account for nearly a quarter of all pediatric cancers.¹ In children, the most common glioma is the low-grade (grade 1) pilocytic astrocytoma (PA), which frequently arises in the setting of the neurofibromatosis type 1 (NF1; OMIM #162202) cancer predisposition syndrome.^2,3 As such, children with NF1 harbor germline NF1 gene mutations and are predisposed to develop PAs within the optic nerves (optic pathway gliomas; OPGs) following somatic loss of the remaining functional NF1 allele in specific cell types. While few children with NF1-OPG die from their tumors, 30% will experience visual decline or loss,^4,5 and some will develop precocious puberty,⁶ both of which significantly impact their long-term quality of life. Since chemotherapy and radiotherapy can have adverse effects on the developing brain,^7,8 there is a pressing need to identify and evaluate additional therapies for NF1-OPG.

The intimate association of these tumors with resident retinal neuron (retinal ganglion cell; RGC) axons coursing through the optic nerve suggests a functional interdependency.⁹ In this regard, we have previously shown that murine Nf1-OPG initiation and growth are dictated by two distinct, nonoverlapping, neuronal activity-dependent mechanisms: OPG initiation is controlled by Nf1 mutation-regulated bioavailability of neuroligin-3 (Nlgn3) in response to visual experience (light exposure¹⁰), while OPG progression and growth are dictated by Nf1 mutation-induced activity-dependent neuronal midkine expression.⁹ Neuronal midkine production results in immune axis support of optic glioma growth, which can be abrogated by treatment with high-dose lamotrigine (LTR; 20 mg/kg/day) treatment in mice.⁹

While the ability of LTR treatment to reduce murine Nf1-OPG growth reveals its translational potential as a therapeutic agent with children with NF1, a few critical issues remain unresolved before this agent can be considered for the treatment of children with NF1-OPG. To address these important knowledge gaps, we performed a series of studies to (1) determine whether midkine is necessary and sufficient to drive Nf1-OPG growth in vivo, (2) assess the relationship between the germline Nf1 mutation, RGC excitability, and neuronal midkine production, (3) examine the ability of antiepileptic drugs to suppress RGC excitability and midkine expression, (4) assess the ability of LTR to durably suppress tumor growth, and (5) determine whether clinically relevant pediatric epilepsy LTR dosing abrogates tumor growth in more than one preclinical mouse strain. These findings will establish the foundations for future clinical translation.

Materials and Methods

Ethics Statement

All mouse experiments performed were ethically and methodologically approved by the Animal Studies Committee at Washington University School of Medicine (Washington University in St Louis Institutional Animal Care and Use Committee).

Mice and Treatments

Under an approved and active Animal Studies Committee at the Washington University School of Medicine (Washington University Institutional Animal Care and Use Committee), mice were maintained on a 12-h light/ dark cycle in a barrier facility, at 21°C and 55% humidity, and had ad libitum access to food and water. Heterozygous Nf1 mice harboring NF1 patient-derived Nf1 germline mutations (Figure 3A; G848R c.2542G>C, R1809C c.5425C>T, R681X c.2041C>T) were generated as previously reported.^9,11,12Nf1 Tyr2083X, c.6249T>G-and Nf1 Arg816X, c. 2446C>T-mutant mice were similarly generated by CRISPR/Cas9 engineering directly into C57Bl/6J embryos, resulting in mice with one wild-type (WT) Nf1 allele and one nonsense Tyr2083X or Arg816X mutation, which was confirmed by direct sequencing (IDT Technologies). All Nf1-mutant mice were backcrossed to C57Bl/6J mice, and their WT littermates were used as controls. Optic glioma-prone mice were generated with the specific patient-derived Nf1 mutation or a neomycin cassette inserted in exon 31¹³ as the germline Nf1 allele and underwent somatic Nf1 inactivation in neuroglial progenitor cells (Nf1^f/Y2083X; hGFAP-Cre/Nf1^Y2083X-OPG mice, Nf1^f/R681X; hGFAP-Cre/Nf1^R681X-OPG mice, Nf1^f/R816X; hGFAP-Cre/Nf1^R816X-OPG mice, or Nf1^f/neo; hGFAP-Cre/ Nf1^OPG mice).^14,15Nf1^R816X-OPG mice form optic gliomas at 6 months of age, while Nf1^OPG, Nf1^R681X-OPG, and Nf1^Y2083X-OPG mice form gliomas at 3 months of age.

Nf1^OPG mice were intraperitoneally administered vehicle (saline in 1% methylcellulose) or 20 mg/kg/d LTR (Selleckchem) daily (5 days a week) for 4 weeks, beginning at 4 weeks of age or starting at 12 weeks of age. The mice were then aged either to 12 weeks, or 24 weeks, respectively, for optic nerve and RGC analysis. For pediatric clinical dosing experiments, Nf1^Y2083X-OPG or Nf1^OPG mice were administered 2.5, 5.0, or 7.5 mg/kg/d LTR by daily oral gavage (5 days a week) for 4 weeks, beginning at 4 weeks of age. The mice were aged to 12 weeks of age for optic nerve and RGC analysis.

AAV Intravitreal Injections

Six-week-old Nf1^OPG mice were anesthetized by intraperitoneal injection of a mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg), and after a surgical level of anesthesia was achieved, one drop of proparacaine was used to block ocular reflexes. Spring forceps were used to gently expose the eye from the orbital socket and to hold the eyelid to stabilize the eye. Intraocular injections were performed as previously described¹⁶ with the following modifications: A Hamilton syringe with a 33-gauge sharp needle was inserted 1 mm into the sclera in the superior–medial aspect of the eye, above the limbus, to access the vitreous cavity. Two microliters of vitreous humor was drawn into the syringe and the needle was removed. Two microliters of the appropriate AAV2-vectors (Supplementary Table 1) were injected slowly into the vitreous cavity through the same incision. The procedure was repeated on both eyes, and topical erythromycin ointment was applied to avoid infection. Injected mice were aged to 12 weeks of age for optic nerve and RGC analyses.

Retinal Ganglion Cell Primary Culture

Retinae from heterozygous Nf1-mutant and WT littermate control mice were dissected in Hibernate-A (Gibco), dissociated in papain solution (Worthington) and ovomucoid inhibitor (Worthington) before being depleted of microglia by CD11b magnetic bead filtering (Miltenyi Biotech). The remaining cells (retinal ganglion cells) were plated onto poly-D-lysine (Sigma)/laminin-(Corning)-coated plates and incubated in neurobasal media supplemented with N2, T3, transferrin, BSA, progesterone, putrescine, sodium selenite, L-glutamine, insulin, N-acetyl cysteine, and forskolin. RGCs were grown for 7 days prior to subsequent treatments, electrophysiological, imaging, and molecular analyses. A subset of RGCs was treated with vehicle (DMSO) control or neuron activity-modulating drugs (Supplementary Table 2).

Multielectrode Array Recordings

Primary RGC (150 000 cells/well) neurons from each of the strains assayed were plated on AXION Biosystems 48-well MEA plates and grown for 7 days as previously described.⁹ Each well included neurons from a single mouse. A minimum of 4 individual mice were analyzed per strain and a minimum of 4 replicate wells of an MEA plate were seeded for each animal. All neurons were recorded for 3 min at a 4.5 standard deviation threshold level and 5000 Hz as a digital filter using AXION Biosystems integrated studio (AxIS) version 2.5.1 software. Spike firing rates were calculated from the total number of spikes/3 min and are represented as spikes/min, only accounting for active electrodes. Representative traces of spikes were extracted using the AXION Biosystems neural metric tool.

Calcium Imaging Analyses

Primary RGC (150 000 neurons/well) neurons were plated onto poly-D-lysine/laminin-coated 96-well plates for 7 days before being treated with Fluo-8/AM (1345980-40-6, AAT Bioquest), PowerLoad (P10020, ThermoFisher) and Probenecid (P36400; ThermoFisher) for 30 min at 37 °C and for another 30 min at room temperature. The neurons were subsequently washed with HBSS and incubated for a minimum of 10 min in fresh culture medium supplemented with 5% neuro-background suppressor (F10489; ThermoFisher). Neurons were imaged on a Nikon spinning disk upright epi-fluorescence confocal microscope equipped with a 10× dry objective, and a 488-nm-wavelength laser was used for wide-field imaging. Neurons were stimulated using a Ti LAPP DMD (Deformable Mirror Device) LED source for ultrafast photo-stimulation, with 0.1 mW applied during each recording for Fluo-8 excitation. Fluo-8 images were collected at 15 Hz (2048 × 2048 pixels, 1 × 1 mm) and the duration of each region of interest (ROI) was limited to 10 min. The fluorescence intensity and optical response to depolarizing membrane potential transients (ΔF/F) were calculated in MATLAB programming environment to generate single-neuron activity traces. The ΔF/F threshold was set at 4 standard deviations beyond baseline fluorescence. Neurons from each animal were seeded in 6 wells and a minimum of 3 neurons were recorded per well. Data recorded from a minimum of 18 neurons per animal were averaged. Each data point represents a single animal.

Tissue Collection, Volume Measurements, Immunohistochemistry, and Immunofluorescence

Mice were transcardially perfused with lactated Ringer’s solution and 4% PFA as previously described. Optic nerves and retinae were post-fixed in 4% PFA prior to paraffin embedding (optic nerves) and OCT embedding (retinae). Optic nerve volumes were calculated as previously described⁹: Four diameter measurements were taken to estimate the thickness of each optic nerve beginning at the chiasm (D₀), at 150 (D₁₅₀), 300 (D₃₀₀), and 450 µm (D₄₅₀) anterior to the chiasm. The following equation was used to calculate the estimated optic nerve volume in each of the three segments between four points of reference (D₀D₁₅₀; D₁₅₀D₃₀₀; D₃₀₀D₄₅₀), the sum of which was ultimately used to calculate the total optic nerve volume: V¹ = 1/12 πh (D₀² + D₀D₁₅₀ + D₁₅₀²). Paraffin-embedded and frozen tissues were serially sectioned (5 μm) and immunostained with Iba1, Ki67, CD3 (optic nerves) and midkine, Brn3a, RBPMS, SMI32, or GFP (retinae; Supplementary Table 3) antibodies. Immunohistochemical optic nerve staining was performed using the Vectastain ABC kit (Vector Laboratories) and appropriate biotinylated secondary antibodies (Vector Laboratories). Hematoxylin and eosin (H&E) staining was performed following the manufacturer’s instructions (StatLab). Frozen retinal staining was completed using appropriate secondary Alexa-fluor-conjugated antibodies (Supplementary Table 3). Images were acquired using Image Studio Lite Version 5.2 software, and LAS AF Lite 3.2.0 software and analyzed using ImageJ 1.53a software, as well as Adobe Photoshop version 21.1.1.

Blood Collection and Midkine ELISA Assays

For serum measurements, blood was retro-orbitally collected with heparinized capillary tubes, and serum separated and collected following centrifugation in EDTA-coated vials. Mouse midkine (LSBio) ELISA was performed on homogenized retinae, optic nerves, or serum following the manufacturer’s instructions. Each assay was performed using a minimum of 4 independently generated biological replicates. Colorimetric assays were analyzed on a Bio-Rad iMark microplate reader and analyzed using MPM6 v6.3 (Bio-Rad Laboratories) software.

Lamotrigine Detection

HPLC-grade acetonitrile and formic acid were obtained from Fisher (Waltham, MA). Lamotrigine and Lamotrigine (13C, 15N4) were obtained from Cerilliant (Round Rock, TX). Lamotrigine calibrators spanning 0.5–400 µM were prepared in fetal bovine serum. Fifty microliters of serum, calibrator, or quality control specimen were combined with 50 µL internal standard and deproteinated with 0.90 mL acetonitrile. Following centrifugation, 50 µL of serum extract was combined with 0.9 mL water prior to analysis. LC-MS/MS analysis was performed using a Waters (Milford, MA) Acquity UPLC—Xevo TqS system with an Acquity BEH C18 column (1.7 µ, 2.1 × 50 mm). Mobile phases consisted of 0.1% formic acid in water (mobile phase A) or acetonitrile (mobile phase B). Column temperature was 40°C. Lamotrigine (1.0 µL injection) was eluted at a flow rate of 0.6 mL/min initially at 95% A for 0.5 min followed by a ramp to 57% A at 2 min, then a rapid change to 5% A at 2.2 min, then held for 1 additional minute. Column was re-equilibrated and 95% A prior to subsequent injections. Lamotrigine was detected in ESI+ mode using m/z transitions of 256→145 (quantitative) and 256→211 (qualitative). Lamotrigine (13C, 15N4) was detected in ESI+ mode using m/z transitions of 261→214 (quantitative) and 261→145 (qualitative). The accuracy of the method was confirmed against a method performed at an outside commercial reference laboratory. Regression statistics of the method comparison were slope 1.0, intercept −0.4, r = 0.987 in 10 specimens ranging from 2.5 to 15 µM. Quality control specimens were analyzed with each run at 1.0 and 100 µM. Imprecision (CV) was 17% and 4% (10 determinations over 4 runs), respectively, at these concentrations.

RNA Extraction and Quantitative RT-PCR

Total RNA was extracted from retinal ganglion cell pellets, whole retinae, or optic nerves following the manufacturer’s instructions (Macherey-Nagel) and reverse-transcribed using a high-capacity cDNA reverse transcription kit (Applied Biosystems). Quantitative RT-PCR was performed using TaqMan gene expression assays (Mdk, Ccl4, and Ccl5; Supplementary Table 4) and a TaqMan Fast Advanced Master Mix (Applied Biosystems) according to the manufacturer’s instructions. All reactions were performed using the Bio-Rad CFX96 Real-Time PCR system equipped with Bio-Rad CFX Manager 3.1 software. Gene expression levels of technical replicates were estimated by ΔΔCt method using Gapdh (Supplementary Table 4) as a reference gene.

Statistical Analyses

All statistical tests were performed using GraphPad Prism software (versions v5, v_8.2.1, and v_9.3.1). All comparisons between 2 groups of samples had similar sample sizes and were analyzed using unpaired 2-tailed Student’s t-tests. For comparisons involving more than 2 groups of samples, 1-way analysis of variance (ANOVA) with Dunnett’s post-test correction was used to compare each experimental group to the designated control group. Multiple independently generated samples (a minimum of 3) were used for all experiments and sample sizes were chosen based on prior power calculations (80% confidence to detect 25% differences) and previously published experiments in our laboratory.¹⁷ The number of replicates is indicated in each figure legend. For all tests, statistical significance was set at P < .05, and individual P-values are indicated within each graphical figure.

Results

Midkine Is Necessary for Murine Nf1-OPG Growth

To determine whether Mdk is necessary for Nf1-OPG growth, we performed a set of three complementary experiments. First, we intercrossed Mdk^−/− mice with Nf1-optic glioma mice that spontaneously develop OPGs by 3 months of age (Nf1^flox/neo; GFAP-Cre; Nf1^OPG). The resulting Nf1^OPG; Mdk^−/− mice had reduced retinal Mdk RNA and protein expression (Figure 1A, B). Consistent with a critical role for midkine in maintaining optic glioma growth, Nf1^OPG; Mdk^−/−optic nerves exhibited reduced proliferation (37% and 62% reduction in %Ki67⁺ cells at 3 (Figure 1C) and 6 months (Figure 1D), respectively), with no change in optic nerve volumes (Figure 1C, D) or tumor-associated monocyte (TAM) or T-cell content (Supplementary Figure 1A, B). Since midkine regulates Nf1-OPG growth through an immune paracrine axis involving T cell-Ccl4 and TAM-Ccl5 production,¹⁸Mdk loss reduced Ccl4 (41% reduction, Figure 1E) and Ccl5 (51% reduction, Figure 1F) expression. Importantly, midkine loss did not alter RAS activity within the optic nerves of Nf1-OPG mice (Figure 1G), demonstrating that midkine-mediated Nf1-OPG growth regulation is RAS independent.

Figure 1.

Midkine is necessary for murine Nf1-OPG growth. (A, B) Immunohistochemical staining (left) and RNA expression analysis (right) of Nf1^OPG; Mdk^−/− retinae reveals reduced Mdk RNA and protein expression at 3 (A) and 6 (B) months of age, respectively, relative to Nf1^OPG controls. Hoechst 33258 is used as a nuclear counterstain. 3 mice, all groups. Scale bars: 100 µm. (C, D) Top: Representative gross images of optic nerves and quantification shows no change in Nf1^OPG; Mdk^−/−- optic nerve volumes relative to Nf1^OPG mice. Scale bars: 1 mm. Bottom: Immunohistochemical staining and quantification reveal reduced tumor proliferation (%Ki67⁺ cells) in (C) 3- and (D) 6-month-old Nf1^OPG; Mdk^−/− optic nerves relative to Nf1^OPG controls. 6 mice, all groups. Scale bars: (C) 100 µm; (D) 50 µm. RNA expression of (E) Ccl4 and (F) Ccl5 is reduced in 3-month-old Nf1^OPG; Mdk^−/− optic nerves relative to Nf1^OPG controls. 3 mice, all groups. (G) RAS activity (RAS-GTP) is unaltered in Nf1^OPG; Mdk^−/− optic nerves relative to Nf1^OPG controls. 3 mice, both groups. (H) Left: Immunofluorescent staining shows expression of the shCTL-GFP and shMdk-GFP constructs in cells within the retinal ganglion cell layer (Brn3a⁺ cells) of Nf1^OPG mice. 3 mice, both groups. Scale bar: 100 µm. Right: Mdk RNA expression is reduced following shMdk, relative to shCTL, viral injection. (I, J) RNA expression of (I) Ccl4 and (J) Ccl5 is reduced in optic nerves from 3-month-old Nf1^OPG injected with shMdk relative to Nf1^OPG shCTL-injected mice. 4 mice, all groups. (K) Top: Representative gross images of optic nerves and quantification show no change in optic nerve volumes in Nf1^OPG mice injected with shMdk relative to those receiving shCTL virus. 6 mice, both groups. Scale bar, 1 mm. Bottom: Immunohistochemical staining and quantification reveal reduced tumor proliferation (%Ki67⁺ cells) in Nf1^OPG; shMdk optic nerves relative to Nf1^OPG shCTL controls. Scale bar, 100 µm. Data are shown as the mean ± SEM. Two-tailed student’s t-test. The P-values are indicated within each graph. ns, not significant; R.E., relative expression.

Second, we used AAV2-shRNA silencing to genetically reduce Mdk expression in the RGCs of Nf1^OPG mice. Similar to Nf1^OPG; Mdk knockout mice, Nf1^OPG mice following intraocular AAV2-shMdk injection (Nf1^OPG; AAV2-shMdk mice) had decreased retinal midkine protein and optic nerve Mdk RNA (42% reduction, Figure 1H), Ccl4 expression (52% reduction, Figure 1I), Ccl5 expression (47% reduction, Figure 1J), and tumor proliferation (48% reduction in %Ki67⁺ cells, Figure 1K), relative to those receiving control scrambled shRNA virus. There was no change in optic nerve volume (Figure 1K) or TAM and T cell content (Supplementary Figure 1C).

Third, to determine whether midkine induces T cell Ccl4 production in a cell-autonomous neuron-specific manner, we reduced Mdk expression in heterozygous mutant Nf1^+/− (Nf1^+/neo) RGCs in vitro using lentivirus-mediated shMdk silencing (Supplementary Figure 1D). Following shMdk treatment, the ability of Nf1^+/− RGC-conditioned medium to induce T cell Ccl4 production was attenuated (49% reduction, Supplementary Figure 1E). Taken together, these data establish that midkine is necessary for neuron-driven Nf1-OPG growth.

Midkine Is Sufficient for Murine Nf1-OPG Growth

Next, to determine whether midkine expression is sufficient to increase Nf1-OPG proliferation, Mdk was ectopically expressed in the retinae of Nf1^OPG mice by intraocular AAV2-Mdk injection (Figure 2A). For these experiments, we used Nf1^OPG mice, which, unlike mice with either constitutive heterozygous Nf1 loss (Nf1+/− mice) or conditional biallelic Nf1 loss in OPG cells of origin (Nf1^flox/flox; GFAP-Cre mice), harbor T cells, that are necessary for OPG growth. Following AAV2-Mdk injection, Nf1^OPG optic nerves exhibited increased Ccl4 expression (2.2-fold increase, Figure 2B), Ccl5 expression (2.8-fold increase, Figure 2C), and tumor proliferation (1.4-fold increase in %Ki67⁺ cells; Figure 2D), but similar volumes (Figure 2E) and Iba1⁺ TAM and T cell content (Supplementary Figure 1F), relative to control-injected AAV-CTL Nf1^OPG mice, at 3 months of age. These data demonstrate that midkine is sufficient for neuron-induced Nf1-OPG growth.

Figure 2.

Midkine is sufficient for murine Nf1-OPG growth. (A) Left, immunofluorescent staining reveals expression of CTL-GFP and Mdk-GFP AAV in the retinal ganglion cell layer (Brn3a⁺ cells) of Nf1^OPG mice. Right, Mdk RNA expression is increased following AAV2-Mdk relative to AAV2-CTL injection. 3 mice, both groups. (B, C) RNA expression of (B) Ccl4 and (C) Ccl5 expression is increased in 3-month-old Nf1^OPG-AAV-Mdk optic nerves relative to Nf1^OPG AAV-CTL-injected mice. 3 mice, all groups. (D) Immunohistochemical staining and quantification reveal increased tumor proliferation (%Ki67⁺ cells) in Nf1^OPG; AAV-Mdk optic nerves relative to Nf1^OPG AAV-CTL controls. 5 mice, both groups. (E) Representative gross images of optic nerves and quantification show no change in Nf1^OPG; AAV-Mdk optic nerve volumes relative to Nf1^OPG AAV-CTL controls. 5 mice, both groups. Data are shown as the mean ± SEM. Two-tailed student’s t-test. The P-values are indicated within each graph. ns, not significant; R.E., relative expression. Scale bars: (A, D) 100 µm; (E) 1 mm.

The Germline Nf1 Gene Mutation Dictates RGC Excitability and Midkine Production

We previously established that midkine production by RGC neurons is regulated in both an activity- and Nf1 mutation-dependent manner, such that Nf1-mutant RGCs from Nf1^R1809C mice, which do not develop tumors, express midkine at levels similar to WT mice.⁹ To determine whether the specific germline Nf1 gene mutation is responsible for dictating both RGC hyperexcitability and midkine production, we leveraged five genetically engineered mouse strains with different NF1 patient-derived germline Nf1 gene mutations (Figure 3A). These included Nf1-mutant mice that do not develop OPGs (Nf1^G848R, Nf1^R1809C; “no OPG” group), and Nf1-mutant mice that form OPGs (Nf1^R681X, Nf1^R816X, and Nf1^Y2083X; “OPG” group, Supplementary Figure 2A, B), following somatic Nf1 loss in neuroglial progenitors (OPG cells of origin; Figure 3B, Supplementary Figure 3A). RGCs from heterozygous Nf1-mutant mice harboring germline mutations associated with OPG formation had elevated midkine protein (Figure 3C) and Mdk RNA (Figure 3D, Supplementary Figure 3B) expression relative to WT controls, which, in turn, correlated with increased proliferative indices (%Ki67⁺ cells) in the optic nerves (Figure 3E), similar to their respective Nf1^flox/neo; GFAP-Cre counterparts (Supplementary Figure 3A, B). Conversely, RGCs from mice in the “no OPG” group exhibited midkine protein and Mdk RNA levels, as well as proliferative indices (%Ki67⁺ cells), that were indistinguishable from WT controls (Figure 3C–E).

Figure 3.

The germline Nf1 gene mutation dictates RGC excitability and midkine production. (A) Table summarizing the Nf1-mutant mouse strains analyzed. (B) Tumor proliferation (%Ki67⁺ cells) is increased in optic nerves from mice harboring OPGs relative to those without tumors. “No OPG,” 30 mice; “OPG,” 20 mice. (C) Immunofluorescent staining and (D) analysis of Nf1-mutant RGCs reveals increased midkine protein and Mdk RNA expression, respectively, in the “OPG” group (Nf1^R681X, Nf1^R816X, Nf1^Y2083X) relative to the “no OPG” group (Nf1^G848R, Nf1^R1809C). 9 mice, both groups. (E) Graph demonstrating the relationship between tumor proliferation (%Ki67⁺ cells) in Nf1^flox/mut; GFAP mice and Mdk expression in Nf1-mutant mice. 9 mice, both groups. (F, G) Quantification of electrical activity of Nf1-mutant RGCs using (F) calcium imaging and (H) MEA analyses. (H, I) Graphs demonstrating relationship between neuronal activity and Mdk expression in Nf1-mutant RGCs. 9 mice, all groups. (J) Quantification of different mouse strain serum midkine levels. “No OPG,” 9 mice; “OPG,” 12 mice. Data are shown as the mean ± SEM. (B, D, J) Two-tailed student’s t-test; (F, G) 1-way ANOVA with Dunnett’s post-test correction. The P-values are indicated within each graph. ns, not significant; R.E., relative expression. Scale bar: 100 µm.

Consistent with a requirement of neuronal excitability for midkine production, neuronal excitability was increased in Nf1-mutant RGCs from the “OPG” group, as measured by calcium imaging (2–3.7-fold increase, Figure 3F, Supplementary Figure 3C) and multielectrode arrays (MEAs; 1.8–6.2-fold increase; Figure 3H, Supplementary Figure 3D), but not in RGCs from the “no OPG” group, relative to WT controls. Additionally, RGC excitability positively correlated with Mdk expression in Nf1-mutant mouse retinae and optic nerves in vivo (Figure 3H, I) and peripheral blood serum midkine levels were exclusively increased in Nf1-mutant mice from the “OPG” group (Figure 3J). These findings suggest that midkine expression is regulated by neuron activity.

Selective Inhibition of Nf1-Mutant RGC Activity-Dependent Midkine Production

Prior studies from our laboratory employed LTR to restore Nf1+/− RGC hyperexcitability and midkine production to WT levels in vitro, as well as to attenuate Nf1-OPG growth in vivo.⁹ To explore the specificity of LTR as an inhibitor of Nf1+/− RGC excitability and midkine production, we performed an unbiased screen employing a library of drugs known to reduce neuronal activity through various mechanisms. Besides LTR, only rufinamide reduced Nf1-mutant RGC Mdk expression (Figure 4A, B), and RGC hyperexcitability to WT levels (Figure 4C). In contrast, commonly used antiepileptic drugs (AEDs), like phenytoin, vinpocetine, picrotoxin, levetiracetam, and zonisamide, actually increased Mdk expression (Figure 4A).

Figure 4.

Selective inhibition of Nf1-mutant RGC activity-dependent midkine production. (A, B) Nf1-mutant RGC Mdk RNA expression is reduced following lamotrigine and rufinamide treatment, but not after the treatment with other antiepileptic drugs. A minimum of 3 independently generated replicates were used for each treatment. CTL 19 mice; Rufinamide, 15 mice. (C) Rufinamide treatment reduces RGC excitability. 6 mice, both groups. Two-tailed student’s t-test. The P-values are indicated within each graph. R.E., relative expression.

Lamotrigine Treatment Durably Reduces Nf1-OPG Growth

As rufinamide is typically not used as seizure monotherapy in children, we focused on LTR for subsequent preclinical studies. We previously demonstrated that intraperitoneal LTR treatment (20mg/kg/day) suppressed Nf1-OPG growth in vivo. However, to determine whether LTR treatment had long-lasting effects on optic glioma growth, two different experiments were performed.

First, Nf1^OPG mice were intraperitoneally injected with 20 mg/kg/day LTR for 4 weeks during the period of tumor evolution (4–8 weeks of age) and assessed at 24 weeks of age (Figure 5A), 3 months after the formation of optic gliomas. At 24 weeks of age, Nf1^OPG mice exhibited reduced Mdk expression (52% reduction; Figure 5B), optic nerve proliferation (58% reduction in %Ki67⁺ cells; Figure 5C), and TAM content (23% reduction in %Iba1⁺ cells; Supplementary Figure 4A) compared to vehicle-treated controls, but had optic nerve volumes (Figure 5B) and T cell content (Supplementary Figure 4A) indistinguishable from vehicle-treated controls.

Figure 5.

Lamotrigine treatment durably reduces Nf1-OPG growth. (A, D) Schematic representation of mouse treatment and analysis timelines. (B, E) Left: Representative gross images of optic nerves and quantification show no change in LTR-treated Nf1^OPG optic nerve volumes relative to vehicle-treated (Veh) mice. Right: Mdk RNA expression is reduced following LTR treatment relative to vehicle-treated mice. 6 mice, all groups. (C, F) Immunohistochemical staining and quantification reveal reduced tumor proliferation (%Ki67⁺ cells) in 6-month-old Nf1^OPG mice treated with LTR relative to Veh mice. 6 mice, all groups. Data are represented as the mean ± SEM. Two-tailed student’s t-test. The P-values are indicated within each graph. ns, not significant, R.E., relative expression. Scale bars: (B, E) 1 mm; (C, F) 50 µm.

Second, to evaluate whether LTR treatment has durable effects on optic glioma proliferation after tumors have formed, 12-week-old Nf1^OPG mice were intraperitoneally treated with 20 mg/kg/day of LTR for 4 weeks and harvested at 24 weeks of age for analysis (Figure 5D). Similarly, LTR-treated Nf1^OPG mice exhibited reduced Mdk expression (55% reduction; Figure 5E), optic nerve proliferation (48% reduction in Ki67⁺ cells; Figure 5F), and TAM content (32 reduction in %Iba1⁺ cells; Supplementary Figure 4B) at 6 months of age relative to vehicle-treated mice, but had optic nerve volumes (Figure 5E) and T cell content (Supplementary Figure 4B) indistinguishable from vehicle-treated mice.

Clinically Relevant LTR Pediatric Epilepsy Dosing Suppresses Tumor Growth in Two Different Nf1-OPG Strains

While LTR reduced both midkine expression and Nf1-OPG proliferation, there are a number of other factors to consider prior to clinical translation. These include the use of pediatric dosing and mode of administration, serum LTR detection, and the ability to suppress tumor growth in additional patient-relevant Nf1-OPG mouse strains. To this end, we first assessed the impact of LTR treatment on Nf1-OPG growth using oral doses routinely administered to children with epilepsy (2.5–7.5mg/kg/day^19,20). Nf1^OPG mice were given oral LTR at 2.5, 5.0, and 7.5mg/kg/day from 4 to 8 weeks of age and analyzed at 12 weeks of age (Figure 6A). All LTR-treated Nf1^OPG cohorts exhibited reduced retinal midkine (Figure 6B) and optic nerve Mdk (37%–50% reduction; Figure 6C) expression relative to vehicle-treated Nf1^OPG mice.

Figure 6.

Clinically relevant pediatric LTR dosing suppresses tumor growth in two different Nf1-OPG strains. (A) Schematic representation of mouse treatment and analysis timeline of Nf1^OPG mice. (B, C) Immunohistochemical staining (B) and RNA expression analysis (C) of Nf1^OPG retinae following LTR treatment reveals reduced midkine protein and Mdk RNA expression in all LTR-treated mice relative to vehicle-treated controls. 3 mice, all groups. (D) Quantification of serum peak LTR levels 4 h after 0 (4 mice), 2.5 (5 mice), 5.0 (5 mice), and 7.5 (4 mice) mg/kg/day oral LTR administration. (E, F) Representative immunohistochemical staining (left) and quantification (right) reveal (E) reduced tumor proliferation (%Ki67⁺ cells; Veh, 10 mice; 2.5, 9 mice; 5.0, 5 mice; 7.5, 5 mice), and (F) increased retinal nerve fiber layer thickness (SMI-32⁺ fibers; Veh, 6 mice; 2.5, 6 mice; 5.0, 3 mice; 7.5, 6 mice) in 3-month-old Nf1^OPG mice treated with LTR relative to vehicle-treated (Veh) mice. (G) Schematic representation of mouse treatment and analysis timeline of Nf1^Y2083X-OPG mice. (H, I) Immunofluorescent staining (H) and RNA expression analysis (I) of Nf1^Y2083X-OPG retinae following LTR treatment reveals reduced midkine protein and Mdk RNA expression in all LTR-treated mice relative to Veh controls. 3 mice, all groups. Hoechst is used as a nuclear counterstain. (J, K) Representative immunohistochemical staining (left) and quantification (right) reveal (J) reduced tumor proliferation (%Ki67⁺ cells; Veh, 13 mice; 2.5, 8 mice; 5.0, 7 mice; 7.5, 6 mice), and (K) increased retinal nerve fiber layer thickness (SMI-32⁺ fibers; Veh, 3 mice; 2.5, 5 mice; 5.0. 5 mice; 7.5, 3 mice) in 3-month-old Nf1^Y2083X-OPG mice treated with LTR relative to Veh mice. Hoechst is used as a nuclear counterstain. Data are represented as the mean ± SEM. One-way ANOVA with Dunnett’s post-test correction. P-values are indicated within each graph. ns, not significant, R.E., relative expression. Scale bars: (B, F, H, K) 100 µm, (E, J) 50 µm.

Second, we measured serum LTR levels following treatment, which showed a dose-dependent increase in peak LTR levels 4 h after 2.5–7.5 mg/kg/day LTR treatments (1–9 µM LTR; Figure 6D), within the range of human serum peak levels (<20 µM).²¹

Third, all LTR-treated Nf1^OPG mouse optic nerves had reduced proliferation (%Ki67⁺ cells; 68%–84% reduction; Figure 6E), as well as reduced TAM (%Iba1⁺ cells; 52%–64% reduction; Supplementary Figure 3C) and T lymphocyte (CD3⁺ T cells; 71%–75% reduction; Supplementary Figure 4C) content, without any change in volume (Supplementary Figure 4C), relative to vehicle-treated Nf1^OPG mice. Since visual acuity is an important medical consequence of NF1-OPG progression,^22,23 we also measured retinal nerve fiber layer (RNFL) thickness in Nf1-OPG mice following treatment. Only females were used for these analyses, since the effect of OPG on vision loss is sexually dimorphic in mice.^24,25 We observed increased retinal nerve fiber layer thickness (SMI-32⁺ fibers, 1.7–1.8-fold increase; Figure 6F), to levels seen in WT mice. It should be noted that mouse Nf1-OPG RNFL thickness assessed using histology accurately reflects measurements obtained using optical coherence tomography,²⁶ which is often used as a surrogate for vision loss in children with NF1-OPGs in whom accurate visual acuities cannot be obtained.²⁷

Fourth, we analyzed the effect of LTR treatment on an additional strain of Nf1^OPG mice harboring the NF1 patient-derived Y2083X germline Nf1 gene mutation (Figure 6G–K, Supplementary Figure 4D) using the same range of LTR dosing. As seen with the workhorse Nf1^OPG mouse strain, LTR treatment of the Nf1^Y2083X-OPG strain resulted in reduced retinal midkine (Figure 6H) and optic nerve Mdk (28%–51% reduction; Figure 6I) expression, reduced proliferation (%Ki67⁺ cells; 69%–81% reduction; Figure 6J), TAM (%Iba1⁺ cells; 53%–65% reduction; Supplementary Figure 4D), and T cell content (%CD3⁺ cells; 67%–75% reduction; Supplementary Figure 4D), with no change in optic nerve volume (Supplementary Figure 4D). As before, LTR treatment increased RNFL thickness (SMI-32⁺ fibers, 1.4–1.7-fold increase; Figure 6K) to WT levels. These findings strongly support the use of LTR as a chemotherapeutic agent for NF1-OPG.

Discussion

Current therapy for NF1-OPG relies on conventional chemotherapy (carboplatin/vincristine^28,29) and investigational drugs that target the MEK/ERK signaling pathway (selumetinib³⁰) activated in NF1-deficient tumor cells. While these treatments demonstrate clinical efficacy and good patient tolerability, there is a pressing need to identify additional treatments for these common tumors in young children with NF1 due to problems with clinical progression following treatment discontinuation.^31–33 For this reason, we explored the possibility that repurposing LTR, a well-tolerated drug used for the treatment of epilepsy and mood disorders in children, might durably block Nf1-OPG progression and growth. Herein, using a variety of mice harboring different germline Nf1 mutations, we establish that RGC activity-regulated midkine production is necessary and sufficient for Nf1-OPG growth and that pediatric epilepsy doses of LTR suppresses RGC excitability, midkine expression, and tumor proliferation months after the cessation of treatment. These findings raise three important points.

First, we demonstrate that midkine expression underlies neuronal activity-dependent regulation of Nf1-OPG growth. This finding argues that neuronal midkine is a critical determinant of glioma proliferation, operating through an immune circuit involving T lymphocyte-Ccl4 and TAM-Ccl5 expression.^18,34,35 This dependency offers opportunities to disrupt the obligate stromal support circuit necessary for tumor maintenance. Importantly, we demonstrate that neuronal activity-dependent control of Mdk expression is highly selective, as it is only suppressed by rufinamide and lamotrigine.³⁶ While the current study is focused on LTR suppression of midkine expression, it is possible that other intracellular signaling pathways within neurons (or other cells within the OPG microenvironment) directly or indirectly impact glioma growth. Ongoing studies beyond the scope of this report are focused on (1) analyzing single-cell RNA sequencing of Nf1-OPG tumors following LTR treatment, (2) defining the direct effects of LTR on optic glioma tumor cell growth in vitro, (3) determining whether ectopic midkine expression in OPG-bearing mice can override the tumor suppressive effects of LTR, and (4) elucidating the mechanism by which NF1 mutation results in increased neuronal excitability.

While anti-epileptic drugs have been previously used to attenuate RGC loss in the setting of ischemia³⁷ and to prevent hyperactivity-induced neurodegeneration in dementia,³⁸ Alzheimer’s disease³⁹ and brain injury,⁴⁰ the use of LTR to suppress brain tumor growth is unique and expands our repurposing of neuronal activity modulators to treat cancer (cancer neuroscience^9,41). It should be noted that LTR is well tolerated in children, where the most commonly occurring side effects are dizziness, nausea, and headaches (8%–20% of children) and skin rashes (12%).⁴² The potentially fatal Stevens–Johnson syndrome arises in 0.1%–1% of treated children, which can be avoided by prescreening and premedication.^43,44 To date, malignant glioma growth has only been inhibited by dampening glutamatergic^45,46 or GABA⁴⁷ signaling, with ongoing studies focused on delineating the specific molecular mechanism(s) underlying the differential effects of these neuron activity regulators on brain tumor signaling and growth. Future studies will be required to explore a cause-and-effect relationship between AED use and OPG growth in children with NF1.

Second, we show that LTR might be used either preventatively or therapeutically. In this regard, treatment during the period of tumor development (4–8 weeks of age) and after the formation of an optic glioma (12 weeks of age) both durably suppressed Nf1-OPG proliferation, as well as attenuated further RNFL thinning. The ability of LTR to inhibit tumor growth during optic glioma formation suggests that LTR could be used as a chemoprevention agent to delay the need for definitive treatment in future clinical trials. Moreover, the preclinical experiments reported herein establish that LTR is effective at attenuating optic glioma growth in two different Nf1-OPG strains, using doses routinely administered to young children for the chronic management of epilepsy.^19,20 It is important to note that the long-term suppression of Nf1-OPG progression and growth seen with LTR treatment stands in stark contrast with preclinical studies using RAS/RAS pathway inhibitors where continued drug treatment is required,^17,48 as well as clinical experience with MEK inhibitors.^31–33,49 Preclinical studies are currently underway to evaluate the efficacy of combined oral LTR and MEK inhibitor (PD0325901) treatment on Nf1-OPG growth.

Third, the use of an allelic series of NF1 patient-derived Nf1-mutant mouse strains establishes that the Nf1 germline mutation acts at the level of neurons (RGCs) to dictate thresholds for glioma penetrance and growth. Both baseline excitability and midkine production are differentially regulated by specific germline Nf1 mutations, similar to PIK3CA mutations that differentially enhance neuronal activity and malignant glioma progression.⁵⁰ This mutational specificity raises the intriguing possibility that neuronal activity, as well as midkine expression, used as a surrogate readout of baseline neuronal excitability, could serve as biomarkers of disease risk. As such, midkine has been previously reported to be a biomarker of lung,⁵¹ gastric,⁵² thyroid,⁵³ and high-grade brain tumors.^54–56 Future studies will be required to validate midkine levels as a clinically relevant indicator of neuronal activity and optic glioma formation in children with NF1, as well as to determine whether antiepileptic drug use in children with NF1 affects OPG development or progression. Integration of midkine levels into future artificial intelligence-generated predictive risk algorithms may accelerate the development of personalized approaches to managing children with NF1.

In conclusion, we determine that clinically relevant pediatric epilepsy doses of LTR effectively suppress RGC excitability, midkine expression, and optic glioma growth, as well as improve retinal pathology, more than 3 months after the cessation of treatment. With adequate prescreening of children at risk for adverse drug reactions,^57,58 LTR is a compelling candidate for clinical trials in pediatric NF1-OPGs and represents a promising drug to add to our therapeutic armamentarium.

Supplementary material

Supplementary material is available online at Neuro-Oncology (https://academic.oup.com/neuro-oncology).

Funding

This work was funded by a VRI grant from the Giorgio Foundation (to D.H.G.), the National Cancer Institute (1R01CA261939 to D.H.G.), the National Institutes of Health (R35NS097211 to D.H.G., and R50CA233164 to C.A.), and the Gilbert Family Foundation (to D.H.G. and R.A.K). The Washington University Ophthalmology Core facility is supported by funding from the National Eye Institute (P30EY002687), while the Washington University Genome Engineering and iPSC Core Center is subsidized by funding from an NCI Cancer Center Support Grant (P30-CA091842).

Acknowledgments

We thank Michelle D. Lambert for her critical assistance in the treatment and processing of mice.

Conflict of interest statement

The authors have no financial conflicts of interest to disclose.

Authorship statement

C.A., J.C., and D.H.G. designed and analyzed the experiments. C.A., J.C, J.P.K., Y. G., S. B., C. M. K., J-K.C., C. E. K., L.I.M.M., C. J. B., and D. D. conducted and/or interpreted the experiments. R.A.K. provided the Nf1^+/Arg816X mouse strain. The manuscript was assembled by C.A. and D.H.G. and further edited by J.C. and J.P.K. D.H.G. was responsible for the final production of the manuscript.

Data Availability

No large datasets were generated as part of this study. The data used and/or analyzed during the current study will be made available upon reasonable request.