Capmatinib is an effective treatment for MET-fusion driven pediatric high-grade glioma and synergizes with radiotherapy

Background Pediatric-type diffuse high-grade glioma (pHGG) is the most frequent malignant brain tumor in children and can be subclassified into multiple entities. Fusion genes activating the MET receptor tyrosine kinase often occur in infant-type hemispheric glioma (IHG) but also in other pHGG and are associated with devastating morbidity and mortality. Methods To identify new treatment options, we established and characterized two novel orthotopic mouse models harboring distinct MET fusions. These included an immunocompetent, murine allograft model and patient-derived orthotopic xenografts (PDOX) from a MET-fusion IHG patient who failed conventional therapy and targeted therapy with cabozantinib. With these models, we analyzed the efficacy and pharmacokinetic properties of three MET inhibitors, capmatinib, crizotinib and cabozantinib, alone or combined with radiotherapy. Results Capmatinib showed superior brain pharmacokinetic properties and greater in vitro and in vivo efficacy than cabozantinib or crizotinib in both models. The PDOX models recapitulated the poor efficacy of cabozantinib experienced by the patient. In contrast, capmatinib extended survival and induced long-term progression-free survival when combined with radiotherapy in two complementary mouse models. Capmatinib treatment increased radiation-induced DNA double-strand breaks and delayed their repair. Conclusions We comprehensively investigated the combination of MET inhibition and radiotherapy as a novel treatment option for MET-driven pHGG. Our seminal preclinical data package includes pharmacokinetic characterization, recapitulation of clinical outcomes, coinciding results from multiple complementing in vivo studies, and insights into molecular mechanism underlying increased efficacy. Taken together, we demonstrate the groundbreaking efficacy of capmatinib and radiation as a highly promising concept for future clinical trials. Supplementary Information The online version contains supplementary material available at 10.1186/s12943-024-02027-6.


Background
Brain tumors are the leading cause of cancer-related death in children, with pediatric-type diffuse high-grade gliomas (pHGG) being one of the most aggressive tumor families [1].Patients suffering from pHGG are typically treated with tumor resection followed by chemotherapy and/or radiation (based on age at diagnosis).This therapy is rarely curative and results in a 5-year survival rate of only ~20% [2].Oncogenic fusions with receptor tyrosine kinase (RTK) genes NTRK, ALK, ROS or MET drive a subgroup of pHGG in infants (IHG, Infant-type hemispheric glioma) [3][4][5][6].IHG has better survival than other pHGG [3,4], but poses a significant therapeutic challenge and is associated with devastating long-term sequelae [5].In pHGG patients >3 years old, MET fusions occur in up to 12 % of cases [6][7][8], and have also been identified in up to 15% of secondary glioblastoma in adults [9].Recent advances yielded remarkable responses of NTRK or ALK fusion pHGG to selective inhibitors [10,11], especially in IHG, but there is currently no effective selective therapy demonstrated for MET fusion-positive glioma.
A plethora of studies have explored new treatment options for pHGG, with solely discouraging outcomes [12].Although novel small molecule inhibitors frequently show promising initial responses, a decade of experience has shown that monotherapy of pHGG inevitably results in therapy-resistant relapses [13].The first FDA-approved inhibitor to target MET was crizotinib (Xalkori ® ).In the context of brain tumors, crizotinib displayed initial efficacy in a patient with pHGG [7], unfortunately followed by rapid progression.Capmatinib, another highly specific MET inhibitor, has shown promising intracranial activity [14,15].However, capmatinib has not been investigated as a treatment option against pHGG so far.
Given the limitations of monotherapies, multiple studies have investigated radiosensitization of tumor cells through RTK inhibition [16].These included MET inhibitors, whose radiosensitizing effects were reportedly mediated by downregulation of DNA repair genes including ATM and/ or by anti-apoptotic factors [17][18][19].However, the effect seems to be model-, tumor-and inhibitor-dependent [20].So far, MET inhibition-mediated radiosensitization has not been explored in the context of pediatric brain tumors.

Methods
All methods and materials are described in the Supplementary Methods (Additional File 1).

Clinical presentation
We analyzed MRI scans from MET fusion IHG patients enrolled on the SJYC07 clinical trial (NCT00602667) [5] or standard institutional protocols, which illustrated typical challenges for IHG surgery.The tumors are often very large, vascular and hemorrhagic, and associated with intraoperative bleeding, difficulties achieving gross total resections, and high morbidity (Fig. 1a-d).Fusion events between CLIP2 and MET have been observed in IHG and pHGG before [3,7], whereas, to our knowledge, we are the first to identify NPM1 and HIP1 as alternative fusion partners of MET.Our institutional experience thus confirmed the significant clinical challenges for MET fusion pHGG patients and the need for novel therapeutic concepts.

A novel, immunocompetent mouse model for MET-driven pHGG
To initially develop a genetically defined model of the disease, we performed in utero electroporation to stably induce expression of the HA-tagged, human TFG-MET fusion gene as well as a CRISPR-mediated knockout of Trp53 in the forebrain of E14.5 mice (Fig. 2a).We chose TFG-MET because it is the smallest identified fusion in both, IHG and pHGG [3,7] (Supplementary Fig. 1a; Additional File 2), fostering efficient somatic gene delivery.All electroporated mice developed tumors that stained positive for the HA-tag, pMET and pErk, validating the delivered fusion gene as an oncogenic driver (Fig. 2b,c).Murine tumors (Fig. 2b) showed similar histopathology to human, MET-fusion driven HGG (Fig. 2d), including characteristically round and relatively monotonous morphology as well as cytoplasmic clearing (Fig. 2b,d, higher magnification boxes, Supplementary Fig. 1b; Additional File 2).Our electroporation model was robust and highly aggressive with 9/9 mice developing neurologic symptoms by day 33 after birth (Fig. 2e).We showed that Trp53 knockout was efficient, inducing a 95 base pair deletion in all analyzed clones (Fig. 2f n=6), thereby recapitulating the loss of TP53 function that is frequently observed in patients with MET-activated pHGG [7].The results highlight Large areas of hemosiderin deposition, evidence of prior hemorrhages and hematoma, are noted in the tumor (second from right).Ample amounts of Gelfoam were needed to achieve hemostasis during surgery (far right).Scale bar is 150µm a novel mouse model with short latency and full penetrance that reflects the histopathology of the human counterpart.

Capmatinib demonstrates a favorable PK profile in mice compared to crizotinib
To evaluate the brain exposure of crizotinib and capmatinib, we analyzed their PK profiles in CD1 nude mice.Capmatinib was rapidly absorbed and cleared from both, brain and plasma, with concentrations below the detection limit at 16 hours post-dose (Fig. 2g, h and Supplementary Table 1; Additional File 3) while crizotinib slowly equilibrated, reaching Cmax at 4 hours post dose.Both drugs, however, reached physiologically relevant concentrations of >1 µM in brain tissues.Previous studies showed that unbound drug concentrations predict target inhibition more robustly than total amounts [21].Therefore, we performed in vitro protein binding assays with standard mouse plasma and naive brain tissue homogenates, finding that capmatinib has an appreciably higher fraction unbound in brain homogenate (Fu,b) versus crizotinib (Table 1).We then used respective Fu,b values to estimate unbound concentrations (Fig. 2h) and found that capmatinib reaches a 9.6-times higher maximal concentration of unbound drug in the brain than crizotinib (~103nM vs ~11nM).

Capmatinib efficiently inhibits TFG-MET in vitro and in vivo
We cultured tumor cells from our electroporation model in vitro and analyzed the impact of crizotinib or capmatinib treatment on phosphorylation of MET and downstream effectors (Fig. 2a).To investigate the intracellular response at relevant in vivo-concentrations, we challenged the cells with Cmax equivalents, based on the identified, free drug concentrations in the murine brain (0.02 µM of crizotinib and 0.15 µM of capmatinib; Fig. 2h).For both compounds 1 µM was used as positive control.Capmatinib readily inhibited the phosphorylation of TFG-MET and downstream targets Erk and Akt at both tested concentrations, while the Cmax equivalent dose of crizotinib displayed minimal effect (Fig. 3a and Supplementary Fig. 1c; Additional File 2).A dose response assay similarly revealed an in vitro potency of capmatinib >10 times higher than that of crizotinib (Fig. 3b and Supplementary Table 2; Additional File 4).Of note, capmatinib readily inhibited MET, Erk and Akt phosphorylation at the observed in vitro IC 50 concentration of only 9 nM (Supplementary Fig. 1d,e; Additional File 2), further emphasizing its potency.Next, we combined both compounds with RT and observed an increased anti-tumoral efficacy compared to single treatments (Extended Data Fig. 1f,g).Both combinations have an additive effect with a trend towards synergy according to the ZIP synergy model [22], with some clearly synergistic dose ranges (Supplementary Fig. 1h; Additional File 2).The capmatinib synergy score peaked at ~100 nM, an achievable free drug concentration in the murine brain (Fig. 2h).
To analyze the efficacy of capmatinib and crizotinib in vivo, we allografted tumor cells from our electroporation model into CD1 mice, providing a standardized mouse model with an immunocompetent background.In order to utilize this model in a combinatorial RT trial, we first determined a radiation dose at which allografted mice developed a partial but not a full response.As an initial study using 20 Gy demonstrated complete tumor remission in two out of five treated mice (Supplementary Fig. 2a,b; Additional File 5), we lowered the total dose to 12 Gy (in typical clinical fractions of 2 Gy per day [23]) for the subsequent, combinatorial trial in which animals received either A) vehicle B) vehicle + RT C) crizotinib D) crizotinib + RT E) capmatinib F) capmatinib + RT (Fig. 3c).All regimens were well tolerated (Supplementary Fig. 2c; Additional File 5).The time point of radiation was chosen to coincide with the Cmax of the respective drug in the brain (Fig. 2h, dashed squares).To analyze pharmacodynamic properties, mice were sacrificed after receiving their second treatment.These animals formed the "PD cohort" whereas the remaining mice represented the "Survival cohort".Treatment with capmatinib led to greatly reduced levels of phosphorylated MET, Erk and Akt in initial neoplasms of PD animals, whereas crizotinib treatment induced a less complete reduction compared to vehicle-treated mice (Fig. 3d,e and f and Supplementary Fig. 3; Additional File 6).These results indicate that capmatinib readily inhibits MET in intracranial tumors at clinically relevant doses.

Combined capmatinib and RT increases survival-rate and -time of murine allografts
We monitored the mice of the Survival cohort for up to 140 days after transplantation.All treatments increased the average survival time compared to the vehicle treated group, which was most prominent for animals treated with capmatinib + RT (Fig. 4a; p-value vehicle vs. capmatinib + RT=0.0388).Besides the prolonged duration of survival, this combination also increased the survival rate 3-fold (Fig. 4a).Additionally, biweekly bioluminescence imaging allowed us to quantify the combinatorial effect of capmatinib + RT (Fig. 4b, Supplementary Table 3; Additional File 7).While all other treatments mostly slowed down tumor growth, 8/10 capmatinib + RT treated animals displayed a reduction of tumor burden by week 3 (Fig. 3b, Supplementary Fig. 2d and Supplementary Table 3; Additional Files 5 and 7).Capmatinib and RT combined was able to eradicate even large initial neoplasms whereas the survivors in other groups were mice with low initial tumor burden (Fig. 4c).
Brains of mice that had to be sacrificed under treatment were histologically analyzed (Fig. 4d).As expected, mice treated with vehicle or radiation alone showed a strong upregulation of pMET, pErk and pAkt.Interestingly, the amounts of pMET and pErk were reduced to  background levels in only 4/6 capmatinib-treated animals.This reflected the time span between last capmatinib administration and tumor isolation, as the 2 mice with stronger pMET/ pErk signal were sacrificed after a 2-day treatment pause, underscoring the observed rapid clearance of capmatinib in the brain (Fig. 2h).

Capmatinib effectively inhibits TRIM24-MET in human pHGG
During the time of this study a seven-month-old infant presented with a large cerebral mass and leptomeningeal metastasis extending from the brainstem through the cervical spine (C1-7; Fig. 5a).After surgical resection of the cerebral tumor, the patient received six months of chemotherapy, as the patient was deemed too young for radiation therapy post-surgery.The patient had no evidence of disease at the end of therapy (Fig. 5a) but relapsed within seven months thereafter.Molecular analysis revealed a TRIM24-MET fusion in both the initial and the recurrent tumors (Fig. 5b), however subsequent treatment with the MET inhibitor cabozantinib was ineffective.Samples of the pre-treatment (TRIM24-MET-i) and recurrent tumor (after chemotherapy but before cabozantinib; TRIM24-MET-r) were obtained for further characterization and disease modelling (Supplementary Fig. 4a-e; Additional File 8).The two samples were used to establish two stably growing cell cultures and expression of the fusion (predicted molecular weight of 116.8 Kd) was validated by immunoprecipitation (Supplementary Fig. 4f-h; Additional File 8).We performed DNA methylation profiling of both primary biopsies and the corresponding established cultures and found that all samples cluster closely with RTK fusion-driven IHG.We also profiled six biological replicates of our murine TFG-MET tumors using MM285k arrays, performed a cross-species implementation and found that the murine tumors clustered closely to human IHG as well (Fig. 5c).
We challenged human tumor cells with brain-specific unbound Cmax equivalents of capmatinib and crizotinib.Both drugs inhibited phosphorylation of TRIM24-MET and ERK within 30 minutes (Supplementary Fig. 5a; Additional File 9).In comparison to murine tumor cells, the Cmax-equivalent dose of crizotinib also resulted in an observable inhibition, albeit to a lesser extent than 1 µM crizotinib or any analyzed capmatinib concentration (Supplementary Fig. 5b; Additional File 9).In doseresponse assays, we found that capmatinib was more potent than crizotinib and cabozantinib (Fig. 5d), similar to our observations in murine tumor cells.To validate capmatinib's potency in additional MET-fusion-driven pHGG models, we also performed dose response assays with SJ-GBM2 cells [24], harboring a CLIP2-MET fusion and with cells isolated from murine tumors, induced by overexpressing TFG-MET alone (without Trp53 knockout; Supplementary Fig. 5c,d; Additional File 9).Capmatinib potently inhibited both models and displayed an IC 50 of only ~1.17nM against SJ-GBM2 cells, further underscoring its effectiveness against pHGG driven by MET-fusions.

Capmatinib treatment leads to long-term progression-free survival of human xenografts
To investigate capmatinib's anti-tumor efficacy on human cells in vivo, we established a novel PDOX model using TRIM24-MET-i cells.Given the observed rapid clearance of capmatinib in mouse tissues and tumors (Fig. 1h and  3d), we chose to administer capmatinib twice per day (bis in die, BID) to PDOX mice, matching the clinical dosing schedule [25].Subsequent western blot and IHC analysis revealed that capmatinib efficiently blocked phosphorylation of TRIM24-MET, ERK and AKT on this schedule (Supplementary Fig. 6a,b; Additional File 10).
To determine how closely our PDOX model would recapitulate the clinical failure of cabozantinib, we directly compared capmatinib vs. cabozantinib treatment (Fig 6a).Cabozantinib treatment resulted in a 15.5 day increase of median survival (p-value cabozantinib vs. cabozantinib vehicle = 0.002).Despite the statistical significance, this slight reduction of tumor growth would likely not have been appreciable clinically and is therefore consistent with the lack of efficacy in the patient.In striking contrast, capmatinib induced a long-term stable disease with all mice surviving the 19-week treatment period (Fig 6b; p-value capmatinib vs. capmatinib vehicle < 0.0005).Regular luciferase imaging underscored the long-term tumor control and even indicated initial regression in two out of eight capmatinib-treated mice (Supplementary Fig. 6c; Additional File 10).Ultimately seven of these animals relapsed after treatment was ceased (Fig. 6b), indicating that capmatinib monotherapy is not sufficient to consistently induce complete remission.
Consequently, we combined capmatinib with RT in human cells and found that radiation increased the response to capmatinib treatment in vitro (Supplementary Fig. 6d; Additional File 10).We then conducted a 4-arm preclinical trial treating the TRIM24-MET-i PDOX model with: 1) vehicle, 2) vehicle+RT, 3) capmatinib, and 4) capmatinib+RT (Fig. 6c).As METfusion-driven tumors are often diagnosed in infants [3] and we aimed to extend the time frame of potential synergy whereby cells were exposed to both capmatinib and RT and chose a very low-dose fractionation [26] of 0.5 Gy per day, with a total dose of 10 Gy over 20 days to recapitulate a clinical scenario balancing risk and benefit in pediatric patients.All treatments were well tolerated (Supplementary Fig. 6e; Additional File 10).RT alone resulted in a slight survival benefit compared to vehicletreated mice (Fig. 6d,e, 7.7 weeks vs 5.7 weeks, Log-rank test p=0.0086).Beside their cranial tumor outgrowth, all mice in these two groups quickly developed spinal metastases (Supplementary Fig. 6f; Additional File 10).Capmatinib monotherapy again induced a stable disease in all treated animals but persistent tumor cells in both,  brains and spines readily grew out once treatment was withdrawn (Fig. 6d,e).Thereby, all mice treated with capmatinib or RT alone eventually reached tumor-induced endpoints.In striking contrast, combined capmatinib + RT profoundly and stably decreased tumor burden (Fig. 6d,e and f ).Importantly, although radiation was focally administered to the head, only one capmatinib + RT treated mouse experienced a spinal metastasis after therapy was withdrawn, while the remaining mice did not show any detectable signs of residual tumor before reaching natural, cancer-independent endpoints (Fig. 6e,f ).Taken together, these results show that also in the context of a human-derived MET-driven pHGG model, only the combination of capmatinib and RT reduces tumor burden and leads to long-term, progression-and metastasis-free survival.

Capmatinib induces dysregulation of DNA repair genes as a possible mechanism of radiosensitizaton
To investigate the molecular basis for the combined effect between capmatinib and RT, we performed RNA-sequencing analysis on murine tumors and validated fusion gene expression as well as p53 inactivation through frameshift in analyzed allografts (Supplementary Fig. 7a-d; Additional File 11).When analyzing the expression of Mapk signature genes [27], we found a significant downregulation in capmatinib treated tumors, whereas crizotinib-treated samples displayed a more heterogenous expression (Supplementary Fig. 7e; Additional File 11).To elucidate capmatinib's molecular effect on the cells, we focused on the differentially expressed genes between 4 capmatinib-treated tumors showing a particularly strong Mapk downregulation (Fig. 7a) and the 6 vehicle-treated PD samples.As expected, we observed a downregulation of gene sets pertaining to proliferation pathways in capmatinib-treated mice (Supplementary Fig. 7f; Additional File 11).Capmatinibtreated tumors of the Survival cohort displayed more heterogenous gene expression patterns than the PD cohort, potentially owing to more variable responses to long term drug exposure (Supplementary Fig. 8a; Additional File 12).Importantly, genes involved in the DNA repair machinery were downregulated in capmatinibtreated tumors (Fig. 7b, Supplementary Fig. 8b and Supplementary Table 4; Additional Files 12 and 13), providing a plausible explanation for the radiosensitizing effect of this drug.In tumors of the Survival cohort, genes involved in cell cycle progression were found to be upregulated after radiation (Supplementary Fig. 8c; Additional File 12), potentially as a late consequence to radiation induced DNA damage and tumor cell selection.Consistent with this finding, we observed a strong correlation between upregulation of genes associated with increased proliferation and upregulation of genes associated with DNA repair across the entire cohort (Fig. 7c).Although the connection between proliferation and expression of DNA repair genes is well known, we found a striking correlation also in further analyzed datasets, including human brain tumors and cells of normal brain development (Supplementary Fig. 8d; Additional File 12), potentially indicating DNA repair gene dysregulation by cell cycle inhibition as general radiosensitization option for certain tumor entities.Furthermore, we found genes involved in Trp53 regulation to be specifically (See figure on next page.)Fig. 7 Capmatinib dysregulates expression of DNA repair genes and enhances radiation-induced DNA damage.a Expression of the indicated Mapk pathway signature (MPAS) genes in tumors of the PD cohort treated with vehicle (+/-RT) or capmatinib (+/-RT, focusing on the 4 strongly affected tumors by capmatinib-treatment, the 2 outliers were excluded for this analysis).With the exception of Epha4, expression of all analyzed Mapk pathway signature genes were inhibited in capmatinib-treated mice.Significantly downregulated (adj.p < 0.05) genes are in bold.b Heatmap of genes in the "DNA REPAIR_7" geneset (baderlab pathways 2019) demonstrating that capmatinib treatment leads to a reduced expression of DNA repair genes.c Correlation between total expression scores of the genesets "DNA REPAIR_7" and "CELL CYCLE_7" (baderlab pathways 2019) amongst all murine tumors (treated and untreated) analyzed by RNAseq in this study.Each dot represents one tumor.d Heatmaps showing expression of MAPK Pathway Activity Score (MPAS) genes for in vitro capmatinib treatments in cell lines derived from TRIM24-MET fusion tumors as compared to a DMSO vehicle control.Significantly downregulated (adj.p < 0.05) genes are in bold.e Western blots of RAD51 and β-ACTIN after indicated treatments of TRIM24-MET or TFG-MET cells for 24 hours.Capmatinib and crizotinib both induce downregulation of RAD51.f Western blots of MET and p-MET after indicated treatments of TRIM24-MET or TFG-MET cells for 24 hours, which serve as controls for Western Blots in e. g γH2AX-immunofluorescence staining of TRIM24-MET L97 human glioma cell lines at different recovery timepoints following 4 Gy-irradiation.Capmatinib (Cap)-treated cells display significantly higher levels of γH2AX compared to DMSO-treated (DMSO) cells.h quantification of γH2AX-foci in f.The percentage of cells with ≥20 γH2AX-foci is significantly higher in capmatinib-treated cells (black bar) compared to DMSO-treated cells (white bar) at 1, 2, 3 and 4 hours following irradiation.Error bars display standard error of mean, statistical significance was determined using t-test analysis.(****;p<0.0001,***;p<0.001,**;p<0.01,*;p<0.05).Scale bar is 10µm.i Western blot of phosphorylated and total Kap1 from TFG-MET allograft tumors treated with vehicle (Veh), or capmatinib (Cap) alone or in combination with irradiation (RT).Samples 7-9 and 11-12 were collected 1 hr after RT, lane 10 was collected 3 hrs after RT, and shows time-dependent decrease of the DNA double-strand break signal.j Quantification of luminescence signal of western blots in panel h normalized to the vehicle control.Each dot represents an individual replicate.Error bars display standard error of the mean.Statistical significance was determined using a One-Way ANOVA followed by Tukey's multiple comparisons test (*p<0.05,**p<0.01,***p<0.001,****p<0.0001).Lysate from lane 10 was excluded due to the different timepoint after RT downregulated in capmatinib-treated samples (Supplementary Fig. 8e,f; Additional File 12), which may contribute to the reduced expression of DNA repair genes despite the absence of Trp53 itself in tumors (Supplementary Fig. 7b; Additional File 11).
To validate this finding in human cells, we also performed RNA-sequencing of TRIM24-MET-i and TRIM24-MET-r cells after 4h in vitro treatment with capmatinib, crizotinib, or cabozantinib.At their respective EC90 concentrations (Supplementary Table 5; Additional File 14), all three treatments caused similar transcriptional responses when compared to DMSO vehicle controls (Supplementary Fig. 9a and Supplementary Table 6; Additional Files 15 and 16).Downregulation of MAPK pathway signature genes confirmed successful MET inhibition (Fig. 7d).Next, we performed pre-ranked gene set enrichment analyses (GSEA) and found that cellular responses to capmatinib treatment between human tumor cultures and allografted mouse tumors were highly similar (Supplementary Tables 7, 8 and 9; Additional Files 17-19).Commonly downregulated genesets included MYC target genes, mTORC1 signaling, and unfolded protein response (Supplementary Fig. 9b,c; Additional File 15).In contrast to our murine tumors, we found that TP53 is expressed in human tumor cells (Supplementary Table 6; Additional File 16).However, we also found a dysregulation of genes involved in DNA repair (Supplementary Fig. 10a; Additional File 20), similar to our observation in allograft models (Fig. 7b).A major DNA repair gene is RAD51, which is involved in DNA double strand break repair and frequently upregulated in various human cancers [28].Despite no significant change in RNA levels, RAD51 protein is downregulated in both murine and human tumor cells after short-term in vitro drug treatment (Fig. 7e,f).
To further confirm the impact of capmatinib treatment on DNA repair, we treated cells with capmatinib or DMSO for 24 hours, performed irradiation and quantified γ-H2AX loci after an additional 1 to 24 hours.When treated with radiation alone, the number of γ-H2AX loci steadily declined over time in murine cells, indicating continuous DNA repair.The addition of capmatinib significantly this process and prolonged recovery (Supplementary Fig. 10b,c; Additional File 20).In human tumor cells the effect was even more pronounced, as capmatinib treatment induced a greatly increased number of DNA double strand breaks (Fig. 7g,h).The ataxia telangiectasia mutated (ATM) kinase initiates a signaling cascade including phosphorylation of Kap1 (KRAB-associated protein 1) at serine 824 in response to DNA double strand breaks [29].To further investigate the effects of capmatinib treatment on DNA damage response in vivo, we evaluated Kap1 pS824 in murine TFG-MET tumors.Phosphorylated Kap1 (p-Kap1), was dramatically increased in irradiated tumors treated with vehicle compared to unirradiated controls and showed a significantly greater increase in tumors treated with capmatinib and RT (Fig. 7i,j).Taken together these findings demonstrate that capmatinib treatment induces a dysregulation of DNA repair genes, and a marked potentiation of radiation-induced DNA damage in vitro and in vivo, providing a rational mechanism for the outstanding combinatorial efficacy in our animal models.

Discussion
Activating alterations in receptor tyrosine kinases are appealing therapeutic targets that are increasingly identified by clinical genomic approaches, and often play important roles in tumor maintenance and survival.Despite a growing armamentarium of available selective RTK inhibitors, choosing the ideal drug and predicting successful tumor response is complicated by diverse factors [30].While RTK inhibition displayed promising responses in multiple pHGG studies [31,32], responsiveness of adult HGG to RTK inhibition proved to be less striking and is currently under investigation [33].This discrepancy could partially result from the fact that pHGG, especially IHG, typically lacks large-scale structural, copy number, or single nucleotide variants [34,35], rendering the tumor exclusively dependent on the oncogenic RTK such as MET.Targeting a MET fusion gene with crizotinib in one pHGG patient resulted in a partial response with rapid tumor relapse [7], yet no lasting response after MET inhibition has been demonstrated for MET-driven pHGGs so far.Neurosurgery for large vascular IHG is associated with high morbidity, such as intraoperative bleeding, hypovolemic shock, mechanical ventilation and permanent neurologic deficits.Attaining a gross total resection (GTR) is difficult, often requiring multiple craniotomies.Therefore, long-term survivors often suffer from permanent neurocognitive impairment, hemiparesis, seizure disorders, dysarthria, and visual deficits.In a recently published NEJM report, a patient was left moribund after two unsuccessful craniotomies to resect a large hemispheric tumor.As molecular analysis revealed an ALK fusion, the child was treated with ALK inhibitor on a palliative basis.Remarkably, the tumor showed rapid shrinking and could be safely surgically resected with good clinical recovery [11].Similar cases have also been reported with NTRK fusion [10] pHGG.However, there is currently no effective selective inhibitor therapy for MET fusion-driven pHGG.
In this study, we established complementary in vitro and in vivo models of MET-driven pHGG including an immunocompetent allograft with TFG-MET fusion and Trp53 deletion.In contrast to a previous RCAS TFG-MET-driven pHGG model [7], the allograft described here is studied in an immunocompetent, wild-type p53 host background and allows robust preclinical evaluation by standardized tumor cell transplantation.Additionally, we generated two patient-derived cell lines and matched xenografts with TRIM24-MET fusion.All of our models closely recapitulated patient primary tumors as demonstrated by histopathology and methylation profiling.They thereby allowed us to faithfully explore the efficacy of three MET inhibitors in combination with RT against MET-driven pHGG.
Detailed pharmacokinetic analyses are critical to identify the optimal MET inhibitor for brain tumor therapy.
Here, we firstly describe capmatinib's CNS penetration in mice and provide an assessment of crizotinib and capmatinib pharmacokinetic properties.For in vivo studies, we used a crizotinib dose previously reported to be tolerated and efficacious in mice [36], which provided a high total plasma AUC of 64,700 hr-ng/mL.Notably, the maximum tolerated dose (280 mg/m 2 BID) for pediatric solid tumors provided a mean steady total plasma AUC of 6,990 hr-ng/mL [37].Thus, the utilized doses in mice far exceeded clinically achievable doses, even when adjusting for the approximately 2.5-fold higher plasma protein binding of crizotinib in mice versus humans [38].In contrast, capmatinib is administered orally at 400 mg BID in humans [15,39], achieving a mean steady total plasma AUC of 17,300 hr-ng/mL [40] -similar to our estimated murine total plasma AUC of 16,400 hr-ng/mL.In this case, comparisons using total AUCs are appropriate, as the plasma protein binding of capmatinib is similar between mice and humans [41].Therefore, our 25 mg/ kg BID regimen of capmatinib was clinically relevant and provided plasma exposures in mice similar to patients at the approved dose.
We also compared the fractions of unbound capmatinib and crizotinib in mouse brain homogenates and found that the unbound fraction of capmatinib was 8.2times higher than crizotinib, which likely contributes to the higher in vivo efficacy of capmatinib.Because of this higher unbound fraction, capmatinib reached a higher effective exposure in the murine brain for up to 8 hours after administration, even though crizotinib achieved much higher plasma AUCs.
The PDOX models allowed us to compare our preclinical results to the presented patient's clinical outcome.After an initial relapse, the patient was treated with cabozantinib, based on previous clinical studies that showed activity against intracranial metastases [42,43].Importantly, our in vivo PDOX response to cabozantinib, while statistically significant, provided a brief extension of survival that would be biologically inadequate when considered as a patient outcome.Thus, our PDOX model recapitulated the clinical failure of cabozantinib, while capmatinib monotherapy induced stable disease in the PDOX model.It is possible that previous brain metastases were more responsive to cabozantinib because of a higher sensitivity to low-level MET inhibition.Alternatively, differences in the blood-brain barrier in brain metastases compared to pHGG may have allowed greater drug availability in the tumor.These examples highlight the utility of evaluating relevant models for specific diseases, even if a drug has proved efficacious in a different tumor type with a common RTK target.
We investigated the efficacy of capmatinib, crizotinib and cabozantinib in vitro and in vivo to assess disrupted signaling of downstream effectors.Although our RNAseq data demonstrated that all three drugs induced a shared cellular response at respective EC90 concentrations, capmatinib displayed a much greater potency in all examined instances, compared to crizotinib and cabozantinib.This is in agreement with previous reports that demonstrated 10 to 100-fold lower IC50 values of in vitro MET inhibition for capmatinib compared to crizotinib or cabozantinib [44][45][46], although different assays have been utilized in these studies.While it was shown that crizotinib and cabozantinib inhibit a broader spectrum of tyrosine kinases [47,48], capmatinib has been demonstrated to selectively target MET with K D values 1000-fold below its second most high-affinity target [49].The plateau at ~20% cell viability/ abundance at higher capmatinib concentrations, which we observed in our dose response assays has been reported before [49] and is likely a result of growth arrest induced by capmatinib's specificity, in contrast to crizotinib and cabozantinib, which also target additional tyrosine kinases at high doses and thereby induce cell death in a non-specific manner.
Dose and safety data for capmatinib treatment in children is not yet available.The FDA approved capmatinib for adults with metastatic non-small cell lung cancer (mNSCLC) with MET exon 14 skipping mutations based on a clinical trial in which capmatinib was permanently discontinued in 16% of mNSCLC patients due to an adverse reaction, most commonly pneumonitis (1.8%), peripheral edema (1.8%) and fatigue (1.5%) [50], providing initial insights into potential toxicities in the pediatric population.
Capmatinib was even more effective when administered concomitantly with radiation, which we initially demonstrated in vitro for all aforementioned models.In the subsequent preclinical allograft trial, capmatinib and RT increased the survival-rate and -time compared to single treatments.In the PDOX study, the combination induced full responses in all but one treated animal, whereas none of the single-treated mice displayed significant tumor regression.This outstanding efficacy of the combination in the PDOX study compared to the allograft trial may be explained in part by the different underlying treatment schedules, which were adjusted in the PDOX study based on capmatinib's PK profile and to match a patient-equivalent dose based on a published clinical trial.Our results firstly and extensively highlight the striking advantage of combining capmatinib and RT against pHGG, and are in accordance with previous reports that demonstrated radiosensitization by MET inhibition [51][52][53].Additional prior studies indicated that this effect is p53-dependent [54].However, here we demonstrated combinatorial efficacy between capmatinib and radiation in both human TP53-expressing cells and in murine Trp53-deficient tumors, although we observed a differential expression of p53 regulating kinases after capmatinib treatment.
When analyzing the underlying mechanisms of radiosensitization we found a significant downregulation of specific DNA repair genes in capmatinib-treated tumor cells.This is in agreement with previous reports that displayed radiosensitization by downregulation of DNA repair genes after inhibition of MET [55][56][57] but also after inhibition of other RTKs [16,58].Many reports identified an involvement of ATM and ATR [17,18,59], which we also noted.However, the broader range of downregulated DNA repair genes together with the tight correlation between cell cycle progression and DNA repair gene expression that was observed in this study, might suggest a more general paradigm of radiosensitization by RTK-inhibition.The sudden downregulation of certain DNA repair genes, within the previously quickly proliferating tumor cells might render these cells generally more susceptible to RT.In agreement with this notion, we observed that capmatinib indeed potentiates radiation induced DNA damage in tumor cells in vitro.We also showed that tumors treated in vivo with combined capmatinib and RT contained increased levels of phosphorylated Kap1 (pS824) compared to tumors treated with vehicle and RT.This ATM-dependent phosphorylation event [29] further demonstrates the elevated DNA double-strand break signaling when combining capmatinib with RT in vivo and shows that effects of the combination are more than additive.This has important implications for the relative timing of drug and radiation delivery.Additional in vivo studies would be needed to comprehensively elucidate all aspects of the underlying signaling cascade and mechanisms driving the cooperative effects of capmatinib with irradiation.The increased efficacy of this combined therapy merits further investigation to comprehensively identify susceptible tumor entities.For example, secondary adult glioblastoma, in which MET-fusions have been identified in up to 15% [9], potentially represent another promising and eligible entity for concomitant capmatinib-radiation treatment in addition to pHGG.
Our preclinical testing in a MET-fusion IHG PDX, showed that low-dose radiation combined with capmatinib reduced tumor burden, leading to long-term progression and metastasis-free survival.To minimize radiotherapy-associated late effects, chemotherapy-based treatment approaches following surgical resection when feasible have historically been used to defer or delay RT until the age of 3-5 years or at relapse [60][61][62][63].For children in this most vulnerable age group, capmatinib alone may provide a useful approach to reduce morbidity by delaying surgery or as a bridging therapy until an age in which combination with radiation becomes more feasible.The low-dose radiation regimen employed in our human xenograft trials and its significant potentiation with MET inhibition highlights a potentially promising approach for older pediatric and young adult populations with MET-fusion driven tumors who will be otherwise treated with only involved field radiation as a standard of care.Clinical evaluation of this regimen should be reserved for those patients old enough for consideration of radiation therapy or in those that have progressed beyond the reach of successful systemic therapy options.The optimal incorporation of capmatinib in frontline treatment for pHGG with MET fusions as neoadjuvant, adjuvant, or radiation-delaying strategy must be tested in controlled and well-monitored clinical trials.

Conclusions
In conclusion, we generated novel, MET-fusion-driven pHGG mouse models to identify the optimal selective inhibitor for this devastating disease.Capmatinib showed greater potency and superior pharmacokinetic properties, including a greater proportion of unbound drug in the brain, when compared with crizotinib and cabozantinib.Combination of capmatinib with low-dose radiation potentiated RT-induced DNA damage and induced robust tumor regression in vivo, while treatment with cabozantinib recapitulated the lack of efficacy seen in the patient.Our consistent results of preclinical data using two independent and complementary mouse models provide a strong rationale for combining capmatinib and RT as novel treatment against MET-activated pHGG.

Fig. 1
Fig. 1 MET fusion IHG are large vascular tumors posing significant surgical challenges.a MRI images of IHG with CLIP2-MET fusion (right panel).Left panel: Left: T2 weighted image shows a large solid cystic tumor encompassing the entire right cerebral hemisphere, Middle: Subtraction weighted Image sequences (SWI).The yellow arrows indicate intra tumoral hemorrhagic regions.Right: T2 weighted image shows large tumor resection cavity after surgery.b MRI images of IHG with NPM1-MET fusion (right panel).Left panel: Left: T2 weighted MRI Image shows a large solid cystic tumor encompassing the entire temporal lobe of the left hemisphere.Right: Image post first attempt neuro-surgical resection.Due to massive bleeding and hemorrhage during surgery only a fraction of tumor could be resected.The yellow arrows show the large cysts within the tumor.c Images of IHG with HIP1-MET fusion (right panel).Left panel: Left: An emergent CT scan performed in the ER on a 4-week-old baby who presented with irritability and bulging anterior fontanelle.Shows a massive right hemispheric hemorrhagic tumor.The yellow arrow points toward the large hemorrhagic focus.Right: Diffusion Restricted images (DWI) of MRI.The restricted water diffusion (dark/black area noted by yellow arrow) represents high cellular density and proliferating tumor.d Histologic sections of a human MET-fusion tumor (TRIM24::MET) show large and abnormal thin-walled vessels invaded by the tumor cells (both upper panels), with mural thrombi (two left panels) and acute hemorrhages (second from left).Large areas of hemosiderin deposition, evidence of prior hemorrhages and hematoma, are noted in the tumor (second from right).Ample amounts of Gelfoam were needed to achieve hemostasis during surgery (far right).Scale bar is 150µm

Fig. 2
Fig. 2 TFG-MET-driven mouse model and pharmacokinetic profiles of MET inhibitors.a Schematic illustrating the method and utilized vectors to induce CRISPR/Cas9-mediated Trp53 deletion and TFG-MET overexpression following in utero electroporations.b H&E staining and Immunohistochemical analysis of a tumor generated by in utero electroporation, visualized by the HA-tag of TFG-MET.In contrast to normal tissue (bottom right corners) tumors display elevated levels of pMET and pErk.Scale bars are 100 µm in large panel and 25 µm in high magnification inset.c, H&E staining showing a large and invasive HGG in the mouse brain.Red rectangle indicates the region shown in b. d H&E staining of a human MET-driven pHGG demonstrating similar features as murine neoplasms.Scale bars are 100 µm in large panel and 25 µm in high magnification inset.e Survival curve indicating penetrance and latency of tumors induces by in utero electroporation.f, Sanger sequencing of PCR products of the targeted Trp53 locus in a tumor revealed a 95bp deletion in all analyzed sequences (n=6).g, h Plasma(G)-and brain(H)-concentrations of capmatinib and crizotinib at the indicated time points after administration of CD-1 nude mice with the respective compounds.Three mice were analyzed per compound and time point.Error bars indicate the standard deviation.Dashed rectangles indicate time windows of radiation in the following preclinical allograft study

Fig. 3
Fig. 3 Capmatinib is effective against TFG-MET-driven tumor cells.a Western blot of phosphorylated and total MET and the downstream effector Erk in cultured, murine tumor cells after different time points of crizotinib (cri) or capmatinib (cap) addition at the indicated concentrations.b Dose-response curves of murine tumor cells after treatment with capmatinib or crizotinib.Each dot represents one replicate of triplicates.Viable cells were analyzed 72 hours after compound addition using the CellTiter-Glo Assay.The vertical dotted lines indicate EC50 values.c Overview schematic depicting the various treatments and the two different cohorts of our preclinical allograft study.d Immunohistochemical stainings of phosphoproteins in tumors of the PD cohort, which were treated with the indicated therapies.Levels of pMET, pErk and pAkt were significantly reduced after capmatinib treatment.Scale bar is 50 µm.e Western blot of phosphorylated and total MET, Akt and Erk from allograft tumors treated with vehicle (veh), crizotinib or capmatinib alone or in combination with irradiation.f Quantification of luminescence signal of western blots in panel C normalized to the respective vehicle control.Each dot represents an individual replicate.Error bars display standard error of the mean.Statistical significance was determined using a One-Way ANOVA followed by Tukey's multiple comparisons test (*p<0.05,**p<0.01,***p<0.001,****p<0.0001)

Fig. 4
Fig.4 Combining capmatinib and RT increases survival-rate and -time in vivo.a Kaplan-Meier curve of mice enrolled in the "Survival cohort".All treatments started 1 week after transplantation.Radio therapy was administered for 6 days, delivering 12 Gy total.Compound treatment was continued for 84 days.After an additional 49 days of monitoring (140 days after transplantation) the trial ended and none of the remaining mice showed any hints of residual tumor.Three mice that were treated with capmatinib + RT reached this time point, whereas each of the other groups contained only 1 "survivor".N = 8 (vehicle arms) or n = 10 (compound-treated arms), respectively.P-values for groups that displayed statistically significant survival differences are indicated.b Bioluminescence-imaging pictures from four representative mice of the vehicle arm (middle ranks according to initial luciferase intensity) and from all mice of the capmatinib + RT arm.First row is depicted in another intensity scale to visualize tumors in all mice.The depicted scale bar indicates the range from 5x10^5-1x10^7 photons/sec/cm 2 /sr.The combinatorial treatment induced tumor regression in 8/10 animals around day 21 on treatment.c Tumor burdens according to BLI of all enrolled mice before treatment are depicted as area of circles (left panel).The right panel shows the initial tumor sizes of mice that survived for 140 days without residual tumor.While the surviving animals of the vehicle groups displayed the smallest initial tumors, neoplasms of all sizes could be cured with combinatorial therapy of capmatinib and radiation.d Immunohistochemical analyzes of phosphoproteins in tumors of the Survival cohort, which were treated with the indicated therapies until onset of neurological symptoms.Phospho-MET, pErk and pAkt levels were significantly reduced in capmatinib-treated mice collected on days of treatment (Mo.-Fr.), however elevated levels reappeared in tissue collected during treatment pauses on weekends.Scale bar is 100µm

Fig. 5 .
Fig. 5. Sensitivity of human tumor samples to MET inhibition.a MRI images from an IHG patient with TRIM24-MET fusion.Left panel: Image at diagnosis showing a large solid cystic tumor filling the entire temporal lobe of the left hemisphere.Middle panel: Image at the end of resection and chemotherapy.Right panel: MRI image at recurrence.b The fusion encompassed TRIM24 exons 1-12 and exon 15 of c-MET, encoding a chimeric protein that contains the N-terminal moiety of TRIM24 and the c-MET kinase domain.c, tSNE projection of a combined methylation dataset comprised of a reference set of glioma subtypes (n=1128, circles from Capper, et al.Nature 2018, triangles from Clarke, et al.Cancer Discov 2020).The TRIM24-MET and TFG-MET tumor samples and cell lines from this study (squares, TRIM24-MET-i primary n = 4, cell culture n = 1; TRIM24-MET-r primary n = 3, cell culture n = 1; TFG-MET Models n = 6) group together with infant HGG with RTK fusion genes (IHG).d Dose-response curves of TRIM24-MET-i and TRIM24-MET-r cells after treatment with capmatinib, crizotinib or cabozantinib for 72 h.Data from three independent experiments.The vertical dotted lines indicate EC50 values

Fig. 6
Fig. 6 Combination of capmatinib and RT eradicates human tumor cells in vivo.a Overview schematic depicting the four treatment arms of the preclinical study comparing in vivo response to capmatinib and cabozantinib.b Kaplan-Meier curve of mice enrolled in the study depicted in A. All treatments started 13 days after transplantation.Compound treatment was continued for 133 days.Within the subsequent 6 months of monitoring, 7 of 8 mice in the capmatinib-treated group experienced tumor relapse.c Overview schematic depicting the four treatment arms of the preclinical study comparing the combination treatment of capmatinib and RT vs either treatment alone.d Kaplan-Meier curve of mice enrolled in the study depicted in C. All treatments started 18 days after transplantation.Radiotherapy was administered at 0.5 Gy per day, delivering 10 Gy total.Compound treatment was continued for 301 days.After an additional 147 of monitoring, the trial ended with all mice having reached their tumor-induced or natural endpoint.e Trend of total flux (photons/sec/cm2/sr) at the cranial and spinal cord region of capmatinib-treated mice enrolled in the 4-arm preclinical trial depicted in C. f Bioluminescence-imaging pictures from mice of the vehicle + RT arm at the time closest to the humane endpoint and from capmatinib + RT treated mice at that time.Color scale range: 1.19x10^6-2.08x10^7photons/sec/cm 2 /sr

Table 1
In vitro ADME (adsorption, distribution, metabolism, and excretion) profiling of crizotinib, capmatinib and cabozantinib.AVG = average, STD = standard deviation.Values indicate the unbound drug fractions in the depicted environment