Response to letter to the editor: All models are wrong; some models are useful

摘要

We agree that models of pediatric low-grade glioma (pLGG) are currently suboptimal. To address this gap, we developed new patient- derived pLGG cell lines, including the first-ever human neurofibromatosis 1 (NF1) null pilocytic astrocytoma (PA) cell line.1 We include this line and other short-term pLGG cultures grown under cell reprogramming conditions in our paper. We disagree with the assertion that infiltrative, grade II gliomas make up the bulk of recurrent/refractory pediatric gliomas. In a series of 198 pediatric non-NF1 tumors, grade II tumors comprised only 6/198 cases. Midline location, extent of resection, and underlying biology (BRAF rearrangements or mutation with loss of CDKN2A) were most likely to impact recurrence and need for treatment.2 Inhibitors of mitogen-activated protein kinase kinase (MEK) penetrate poorly into normal brain parenchyma; however, in the case of avidly gadolinium contrast-enhancing PA—which comprises the vast majority of pLGGs—trametinib, selumetinib, and binimetinib have demonstrated efficacy in pediatric patients.3 The results are so clear that the current Children’s Oncology Group phase III pLGG clinical trial randomizes to carboplatin/vincristine or selumetinib in first-line therapy ({"type":"clinical-trial","attrs":{"text":"NCT04166409","term_id":"NCT04166409"}}NCT04166409). Experimental and clinical evidence does not support the letterwriters’ assertion that heterogeneously contrast-enhancing atypical teratoid rhabdoid tumor (AT/RT) has a more leaky blood–tumor barrier than avidly contrast-enhancing PA. In our hands, binimetinib and selumetinib suppress the growth of AT/RT cells in vitro and in flank xenografts (Shahab et al, in press) but have no efficacy against AT/RT orthotopic xenografts, similar to results recently described for trametinib.4 In contrast, the target of rapamycin complex 1 and 2 (TORC1/2) kinase inhibitor TAK228 inhibits the growth of orthotopic AT/RT xenografts as a single agent.5 TAK228 (also known as MLN128, INK128, and sapanisertib) penetrates the brain in animal studies, and is currently being tested to determine drug penetration in human recurrent glioblastoma tumors ({"type":"clinical-trial","attrs":{"text":"NCT02133183","term_id":"NCT02133183"}}NCT02133183). The human BT40 BRAFV600E pLGG flank model is imperfect. However, the sensitivity of BT40 to MEK inhibitors predicted the success of these agents in pLGG.6 The genetically engineered mouse models (GEMMs) of mitogen activated protein kinase-activated glioma have a long latency of tumor formation (in the case of LGG models) making longitudinal treatment studies impractical, or are high-grade glioma models, which have little shared biology with pLGG. GEMMs have an uneven track record of predicting responses in human tumors—notably the GEMM malignant peripheral nerve sheath tumor model demonstrated susceptibility to mammalian target of rapamycin (mTOR) inhibitors and heat shock protein 90 inhibitors, but these findings did not translate into clinical activity in patients.7 We found that trametinib/TAK228 therapy targeted the vascularity of BT40 tumors. Abnormal vasculature is a hallmark of PA, and drugs targeting blood vessels, such as bevacizumab, are effective. Patients with gadolinium enhancing pLGG often show remarkable reduction in enhancement after initiation of MEK inhibitor therapy. Our work in BT40 (with validation in human vascular models) mechanistically explains this curious clinical effect of MEK inhibition in patients and shows the combinatorial activity of mTOR and MEK inhibitors. As the statistician George Box wrote, “All models are wrong; some models are useful.” 8 While BT40 is an imperfect model, we maintain that its history of predicting response in pLGG patients makes it highly useful.