NF1^+/− Hematopoietic Cells Accelerate Malignant Peripheral Nerve Sheath Tumor Development without Altering Chemotherapy Response Free

Soft tissue sarcomas (STS) are tumors of the connective tissue comprised of a diverse array of subtypes. Frequent disruptions in the Ras pathway are a hallmark of many STS (1, 2). Although mutations in Ras itself have been frequently identified in rhabdomyosarcomas in children (3–5), mutations in other Ras pathway gene members are more common in many adult soft tissue sarcomas, including the tumor suppressor neurofibromin (NF1), a negative regulator of Ras protein function (6). Loss-of-function mutations in NF1 have also been identified in a wide spectrum of other sporadic tumors, including lung adenocarcinoma, glioblastoma, melanoma, and breast cancer (7). In addition, neurofibromin plays a central role in the cancer predisposition syndrome neurofibromatosis type I (also abbreviated NF1), a common genetic condition that affects 1 in 3,000 children. Patients with neurofibromatosis type I (NF1) have germline loss of a single NF1 allele, resulting in NF1 haploinsufficiency (NF1+/−) and hyperactive Ras signaling in all cells throughout the body. Loss of the second NF1 allele results in the formation of neurofibromas, benign growths arising from nerves, which occur in >95% of NF1 patients. Neurofibromas can result in severe morbidity, causing disfiguration, pain, and occasionally lethality. Importantly, neurofibromas occur predominantly in neurofibromatosis patients, suggesting that NF1 haploinsufficiency plays a role in neurofibroma development, although sporadic neurofibromas do develop in the absence of NF1 haploinsufficiency. Mouse models have demonstrated that Nf1 haploinsufficient supporting stromal cells (endothelial cells, immune cells, fibroblasts, etc) can accelerate the development and vascularization of neurofibromas (8, 9), although Nf1 haploinsufficient stroma is not a requirement for neurofibroma formation in mice (10, 11). Neurofibromatosis patients are also at increased risk for developing soft-tissue sarcomas, most commonly malignant peripheral nerve sheath tumors (MPNST), with a 5-year survival of less than 50% (12, 13). However, unlike neurofibromas, the role of the NF1 haploinsufficient tumor microenvironment in MPNST biology is not well understood, as MPNSTs also occur in non-NF1 patients (referred to as sporadic MPNSTs).

MPNSTs present with two distinct clinical histories: NF1-associated (in neurofibromatosis patients) and sporadic (in the general, non-NF1 population). Approximately 8%–13% of neurofibromatosis patients will develop MPNSTs in their lifetime, with a mean age of 26 years at diagnosis. In contrast, sporadic MPNSTs have a much lower incidence (<0.1%), with an average age onset of 62 years (14). NF1-associated and sporadic MPNSTs are indistinguishable by gene expression profile, anatomic location, and histopathology, although groups have reported differing levels of p53 staining between the two tumor types (15). However, outcomes for NF1-associated MPSNTs appear worse (12, 16), although survival may be improving (13, 17). In addition, NF1-assoicated MPNSTs may respond less favorably to cytotoxic chemotherapy, including a standard doxorubicin/ifosfamide combination used for many sarcoma patients (18). It is unknown if the disparity in outcome is due to clinical factors, such as neurofibromatosis patients presenting later with MPNST due to the challenge of identifying the sarcoma within a patient with multiple large neurofibromas, or if the disparity in outcome is due to intrinsic biological properties of the tumors, such as worse response to chemotherapy (19). A better understanding of the biological differences driven by the MPNST microenvironment in NF1 patients might have important clinical implications that could impact patient care and reveal new cellular targets for MPNST-directed therapy.

To address this question experimentally, we have developed a new mouse model of malignant peripheral nerve sheath tumor (MPNST) to explore the role of NF1 haploinsufficiency in tumor biology. Previously, we used Cre-loxP technology to generate the first mouse model of temporally and spatially restricted MPNST in Nf1^Flox/Flox; Ink4a/Arf^Flox/Flox mice (20) and established an accessible preclinical platform for testing novel therapies. With this approach, MPNSTs develop in Nf1^Flox/Flox; Ink4a/Arf^Flox/Flox mice following injection of an adenovirus expressing Cre recombinase (Ad-Cre) into the sciatic nerve. Cre deletes both Nf1 and the commonly mutated tumor suppressor Ink4a/Arf (known as p16/19 in mice and CDKN2A or p14/p19 in humans) in cells of the surrounding nerve sheath. The tumor microenvironment remains wild-type for both Nf1 and Ink4a/Arf, as the conditional floxed alleles are not deleted in the stroma. These tumors faithfully reflect the human MPNST by histology and IHC. In addition, these tumors have mutant Cdkn2a and wild-type p53, a genetic combination found in the majority of sporadic and NF1-associated of MPNSTs (21, 22). We have used this system to characterize the biological properties of Nf1-deleted sarcoma and as a platform to test MEK inhibitors in vivo (20). To model the unique genetics of NF1-associated MPNSTs, here we developed a complementary model in Nf1^Flox/−; Ink4a/Arf^Flox/Flox mice. Upon Ad-Cre injection, this model generates MPNSTs deleted for both Nf1 and Ink4a/Arf with a surrounding tumor microenvironment comprised of Nf1^Flox/− cells, mimicking NF1 haploinsufficiency (i.e., NF1^+/−) found in neurofibromatosis patients through the loss of a single Nf1 allele. In this study, we compared the two models of MPNST with either Nf1 intact or Nf1 haploinsufficient stroma to investigate the role of the NF1 haploinsufficient tumor microenvironment in MPNST biology.

All mouse work was conducted in accordance with Duke University Institutional Animal Care and Use Committee approved protocols. The Nf1^Flox/Flox and Nf1^+/− mice (23, 24) were obtained from Jackson laboratories, and Ink4a/Arf^F/F mice (25) were provided by Ron Dephino. Tumors were generated by injection of Ad-Cre (University of Iowa, Iowa City, IA; Vector core) into the sciatic nerve of Nf1^Flox/Flox; Ink4a/Arf^Flox/Flox or Nf1^Flox/−; Ink4a/Arf^Flox/Flox compound mutant mice as described previously (20). Briefly, the nerve was exposed by an incision in the surrounding muscle of the upper left flank, followed by injection of 25 μL Ad-Cre mixed with calcium phosphate into the nerve (26).

Antibodies for IHC were used at 1:200 and include: S100 protein (DakoCytomation), pERK (Cell Signaling Technology, clone D13.14.4E), pS6 (Cell Signaling Technology, clone D68F8), Ki67 (BD Pharmagen, clone B56), cleaved caspase-3 (BD Pharmagen, clone CPP32), and KIT (DakoCytomation). All immunostaining was conducted with citrate-based antigen retrieval. The following stains were performed by the Duke Research Histology Lab: CD31, SMA, CD45, CD3, B220, CD11C, CD68, and Iba1. To visualize mast cells, slides were stained with toluidine blue solution (0.02% toluidine blue in 1% NaCl, pH 2.2) for 2 minutes, followed by two washes in distilled water and three washes in 100% ethanol. Quantification of cells staining for an antibody or toluidine blue was performed on 6 fields of a single slide to generate the total number of positively stained cells. 20× fields were used for all analyses except for Ki67 levels, which used 40× fields.

The MPNST mouse cell lines were derived from primary MPNSTs in Nf1^Flox/Flox; Ink4a/Arf^Flox/Flox or Nf1^Flox/−; Ink4a/Arf^Flox/Flox mice in our own laboratory. These cells were authenticated by PCR genotyping (18) within the last 6 months. Doublings of cell populations has been reported previously (27). Three independent cell lines of each genotype were seeded at 10,000 cells per well for 48 hours, and the number of cell doublings was calculated. From this population, another 10,000 cells were plated, and cell doublings were calculated for at least four iterations. The graph shows the average of three independent cell lines per genotype.

Whole bone marrow cells (WBM) were isolated according to methods described previously (28). WBMs from donor mice with the indicated genotype and expressing tdTomato fluorescent protein were transplanted into 8-week-old recipient mice 6 hours after two fractions of 4.75 Gy total-body irradiation (TBI) with an interval of 18 hours. TBI was performed 50 cm from the radiation source with a dose rate of 220 cGy/minutes with 320 kVp X-rays, using 12.5 mA and a filter consisting of 2.5-mm Al and 0.1-mm Cu (X-RAD 320 Biological Irradiator, Precision X-ray). The dose rate was measured with an ion chamber by members of the Radiation Safety Division at Duke University. At 12 weeks after bone marrow transplantation, donor-derived engraftment of bone marrow cells in the peripheral blood was assessed by flow cytometry according to methods previously described (28).

Cells were dissociated from tumors according to methods described previously (29). Total cells dissociated from tumors were blocked with rat anti-mouse CD16/32 IgG (BD Pharmingen) and then stained with FITC-conjugated anti-mouse Ly-6C (clone: HK1.4), phycoerythrin (PE)-conjugated anti-mouse CD11c (clone: N418), PE-Cy5–conjugated anti-mouse CD11b (clone: M1/70), APC-conjugated anti-mouse NK1.1 (clone: PK136; all from eBioscience) and APC-Cy7–conjugated anti-mouse Ly-6G (clone: 1A8; BioLegend). All antibodies were diluted 1:400. Data were collected from at least 200,000 single cells by FACSCanto (BD Pharmingen) and analyzed by Flowjo (Tree Star, Inc). Tumor-associated immune cells were characterized on the basis of the expression of phenotypic markers (30). Dendritic cells: CD11b⁺ CD11c⁺, monocytes: CD11b⁺ CD11c⁻, monocytic myeloid-derived suppressor cells (MDSC): CD11b⁺ CD11c⁻ Ly6G⁻ Ly6C^high, and granulocytic MDSCs: CD11b⁺ CD11c⁻ Ly6G⁺ Ly6C^low.

A human MPNST tissue microarray (TMA) was built using patient samples obtained under approval of the Institutional Review Board of The University of Texas MD Anderson Cancer Center as described previously (13, 31). An anonymized, full clinical database, including the neurofibromatosis status of each patient, was constructed. KIT (DakoCytomation) IHC was performed on TMA slides as described previously (32, 33). In short, slides were deparaffinized using Bond Dewax solution, subjected to citrate-based antigen retrieval, and visualized using a polymer-HRP secondary antibody with DAB as the chromogen and hematoxylin as the counterstain. Each pair of duplicate, 0.6-mm diameter samples was scored by counting KIT-positive mast cells in each of the paired cores and averaged. They were then triple binned as having <5, 5–50, or >50 mast cells for analysis. Primary and recurrent disease sites were tabulated separately for each patient. When a single patient was represented by multiple tumor samples, an average value was recorded.

RNA was isolated from tumors and cells by TRIzol. cDNA synthesis was conducted with iScript (Bio-Rad). qRT-PCR was conducted on an iQ5 instrument (Bio-Rad) using the ΔΔC_t method.

Mice were treated with a single bolus of vehicle or doxorubicin/ifosfamide (15 mg/kg doxoribicin intraperitoneally and 50 mg/kg ifosfamide via tail vein) when they reached a volume of 200–300 mm³. Tumors were measured by calipers 3 times weekly. Tumor volumes were calculated using the formula V = (π × L × W × H)/6, with L, W, and H representing the length, width, and height of the tumor in mm, respectively. Criteria for defining treatment response included both volumetric analysis and a delay in tumor growth kinetics. For waterfall plot analyses, percentage of maximal change is reported as the greatest percentage of loss in tumor volume from baseline for tumors responding to treatment, as reported previously (34). Tumors that did not respond to treatment are shown as greatest percentage of gain in tumor volume from baseline over the course of the treatment. In this presentation of the data, change in tumor volume is capped at 100%.

Graphs and statistics were conducted in GraphPad 6.0. A nonparametric Student t test was conducted to determine differences between treatment groups. Comparison of survival curves was performed using the log-rank (Mantel–Cox) test, with a P value of less than 0.05 considered statistically significant.

To model the unique stromal genetics of MPNSTs in distinct patient groups, we created a breeding strategy that would generate paired littermate mice with either Nf1 haploinsufficient (Nf1^Flox/−) or wild-type Nf1 (Nf1^Flox/Flox) stroma. By utilizing conditional alleles, we can fully delete Nf1 and Ink4a/Arf in tumor cells of both models to generate Nf1-null, Ink4a/Arf–null MPNSTs, while still maintaining endogenous expression of any floxed alleles in the surrounding stroma (Fig. 1A and B). Thus, injection of Ad-Cre into the sciatic nerve of Nf1^Flox/−; Ink4a/Arf^Flox/Flox (NI^F/−) mice will generate MPNSTs that are deleted for Nf1 and Ink4a/Arf, while retaining expression of Ink4a/Arf and a single copy of Nf1 in the surrounding stroma, modeling NF1 haploinsufficiency in neurofibromatosis patients. In contrast, Ad-Cre delivery into Nf1^Flox/Flox; Ink4a/Arf^Flox/Flox (NI^F/F) mice will generate MPSNTs that are deleted for Nf1 and Ink4a/Arf, but surrounded by wild-type stroma, modeling sporadic MPNSTs in non-NF1 patients.

A cohort of NI^F/F (n = 59) and NI^F/− (n = 37) paired littermate mice were injected with Ad-Cre into the sciatic nerve to generate MPNSTs. Tumors developed more rapidly in NI^F/− mice than in NI^F/F littermates (P < 0.05, 101 days vs. 139 days, Fig. 1C). Review of the histopathology by a sarcoma pathologist demonstrated that all tumors were MPNSTs based on spindle cell morphology and lack of myogenic staining. All tumors were histologically indistinguishable by genotype and showed similar levels of S100B and Ki67 staining (Fig. 1D and E). To study cell-intrinsic properties of the MPNSTs, we generated cell lines from multiple tumors across the two genotypes. The doubling rate of cells from tumors from NI^F/− and NI^F/F mice were indistinguishable (Fig. 1F, showing the average of three independent cell lines per genotype). Taken together, these data suggest the NI^F/− model represents the distinctly unique characteristics of NF1-associated MPNSTs, which are histologically similar to sporadic MPSNTs but develop at an earlier age.

Neurofibromin regulates signaling through the Ras-MEK-ERK and PI3K/mTOR pathways, and NF1 haploinsufficiency can lead to elevated levels of pERK and pS6 that result in altered activity of endothelial cells, fibroblasts, and immune cells (35). Thus, we hypothesized that the tumor microenvironments of NI^F/F and NI^F/− MPSNTs would differ in the number or activity of these cell types. To explore these potential differences, we performed IHC analysis for levels of pERK and pS6, in addition to components of the tumor microenvironment including tumor vasculature, fibroblasts, and immune cells. Levels of pS6 and pERK were high in most MPNSTs evaluated, with no difference between stromal genotypes (Fig. 2A and B). To examine endothelial cells, we quantified the number of CD31⁺ cells per tumor. Vascular density between the two genotypes was virtually identical (Fig. 2C). Fibroblast content and activity were analyzed by SMA+ cells (a marker of fibroblast activation; refs. 36, 37) and levels of Masson Trichrome staining (a measure of collagen content). There was no discernable difference in either metric of fibroblast activity (Fig. 2D and E). To examine immune cell infiltrate, we analyzed CD45+ cells (Fig. 2F). There was a greater than 2-fold increase in the number of CD45⁺ cells in NI^F/− MPNSTs, suggesting a large immune cell infiltration from the Nf1 haploinsufficient microenvironment.

We hypothesized that the increased hematopoietic infiltrate in the NI^F/− MPNSTs might contribute to the accelerated kinetics of tumor development observed in Fig. 1C. Alternatively, the NI^F/− mice could develop MPNSTs more rapidly because tumorigenesis in the NI^F/− mice only requires Cre to delete a single Nf1^Flox allele (a “second hit”), in contrast to deletion of two Nf1^Flox alleles in NI^F/F mice. To test these hypotheses, we performed bone marrow transplants from NI^F/− or NI^F/F donor mice into NI^F/F recipient animals (Fig. 3A). Recipient NI^F/F mice received whole-body irradiation to eliminate the majority of Nf1 wild-type bone marrow. Half of the mice received NI^F/− bone marrow, while the other half received NI^F/F bone marrow. A ubiquitous fluorescent reporter in the donor mice (membrane-bound Tomato Red, mTom) was used to assess engraftment. Four weeks after recovery, the average percentage of mTomato+ peripheral blood cells was 82.6% ± 1.63 (Supplementary Fig. S1). At this time, the mice were injected with Ad-Cre into the sciatic nerve. Similar to the results in Fig. 1C, mice that received NI^F/− bone marrow developed tumors more rapidly than mice that received NI^F/F bone marrow (Fig. 3B). Histologic analysis of the tumors showed increased infiltration of hematopoietic cells in mice receiving NI^F/− bone marrow, illustrated by CD45 staining. (Fig. 3C). Both genotypes of MPNSTs had similar levels of Ki67 staining (Fig. 3C). These data demonstrate that differences in efficiency of deletion of Nf1 by Cre in the tumor-initiating cell do not fully explain the different time to tumor development between MPNSTs in NI^F/− and NI^F/F mice. Instead, these results indicate that factors derived from the bone marrow compartment are responsible for the accelerated tumor development in NI^F/− mice. Thus, the Nf1 haploinsufficient tumor microenvironment directly impacts MPNST biology.

To further characterize the CD45⁺ immune cell infiltrate that is associated with accelerated MPNST development in NI^F/− mice, we used flow cytometry and IHC to analyze the primary MPNSTs generated in Fig. 1. Flow cytometry allowed us to interrogate multiple immune cell types in MPNSTs and showed an increase in CD11b⁺ myeloid cells in tumors from NI^F/− mice (Fig. 4A and B). Within the CD11b⁺ population, there was no difference between the two genotypes in levels of CD11b⁺ CD11c⁺ dendritic cells. However, the percentage of CD11b⁺ CD11c⁻ monocytes was significantly higher in NI^F/− MPNSTs. Recent studies demonstrate a critical role for MDSCs in promoting tumor initiation (38, 39). Therefore, we also compared the percentage of MDSCs between NI^F/− and NI^F/F MPNSTs, but we did not observe a significant difference in either monocytic MDSCs (CD11b⁺ CD11c⁻ Ly6G⁻ Ly6C^high) or granulocytic MDSCs (CD11b⁺ CD11c⁻ Ly6G⁺ Ly6C^low). IHC analysis found no difference in T cells (CD3⁺) or B cells (B220⁺; Fig. 4C), supporting the flow analysis demonstrating that the major difference between the immune infiltrates was a consequence of the number of myeloid cells. IHC results further supported the flow data by confirming that CD11c⁺ dendritic cells were not enriched in NI^F/− MPNSTs. Notably, although we observed an increase in CD11b⁺ cells in NI^F/− MPNSTs by flow cytometry, IHC staining using two macrophage markers (CD68⁺ and Iba1⁺) did not show a significant difference in MPNSTs of either genotype (Fig. 4C).

We also used histologic techniques to examine cell populations that were not captured by our flow analysis, including mast cells stained by toluidine blue. Mast cells are tissue-resident, histamine-rich, cKit⁺ myeloid cells that play a fundamental role in the allergic response and may influence cancer development in some cellular contexts (40). Mast cells were strongly enriched in MPNSTs from NI^F/− mice, with high levels found exclusively in tumors from mice with an Nf1 haploinsufficient tumor microenvironment (Fig. 4C and D). Importantly, there were no differences in mast cell number between NI^F/F and NI^F/− mice in other tissues, including mast cell–rich skin and tongue samples (Supplementary Fig. S2), suggesting their enrichment is tumor specific. This finding is of particular relevance to neurofibromatosis, as Nf1 haploinsufficient mast cells play important roles in the formation of neurofibromas due to their activation by SCF, the c-Kit ligand (9). There was no difference in the number of partially or fully degranulated mast cells between the two genotypes (Supplementary Fig. S3). To further explore the link between elevated mast cell numbers and MPNSTs in NI^F/− mice, we performed real-time PCR for several cytokines involved in mast cell biology and recruitment. Indeed, levels of the mast cell–activating cytokine SCF were elevated in the tumors from NI^F/− mice (Fig. 4E), while levels of the mast cell–repressive cytokines CXCL1 and CXCL2 were decreased (41). Other cytokines whose role in mast cell biology was less well-defined were unchanged, including IL10. Collectively, our results from flow cytometry and IHC staining indicate that CD11b⁺ CD11c⁻ monocytes and mast cells are the two key hematopoietic cells enriched in the stroma of mouse NI^F/− MPNSTs.

To examine the significance of the mast cell enrichment phenotype in human MPNST samples, we analyzed a tissue microarray of primary and locally recurrent MPNSTs with known NF1 status for c-Kit staining, a commonly used clinical marker of mast cells (31, 33). Mast cells were enriched in NF1-associated MPNSTs compared with sporadic MPNSTs (Fig. 4F). Taken together, these data indicate that mast cells are enriched in human and mouse NF1-associated MPNSTs.

Chemotherapy can play an important role in treating advanced MPNST (42, 43), but overall survival remains low. Doxorubicin is the most common chemotherapy for advanced stage sarcoma (44); however, response rates are low (17–27% for single-agent doxorubicin), and there are risks of significant cardiac toxicity associated with high cumulative doses of doxorubicin. In patients with NF1-associated sarcomas, the response rate may be even worse (14, 18). Improved outcomes may be found by optimizing chemotherapy regimens and combination treatments in clinical trials, but these studies are time-consuming and potentially expose patients to additional toxicity without benefit. There are conflicting reports as to the benefit of multi-agent chemotherapy for treating advanced MPNST in NF1 patients (42, 43). A clinical trial SARC006 (NCI-06-C-0043) attempted to determine whether MPNSTs from neurofibromatosis patients would respond similarly as tumors from patients with wild-type NF1 to a combined ifosfamide/doxorubicin regimen. Patients with NF1-associated MPNST and sporadic MPNST were enrolled and treated with doxorubicin and ifosfamide chemotherapy regimens (Fig. 5A). Unfortunately, this clinical trial closed early due to poor accrual, and no clear differences in chemotherapy response were identified in this patient cohort (45). Therefore, we performed a trial in our primary mouse models to help define the role of the Nf1 haploinsufficient tumor microenvironment in the therapeutic response to standard-of-care chemotherapies (Fig. 5B). These mouse models are well-suited for preclinical studies, as the primary tumors develop within a native, immunocompetent tumor microenvironment and are spatially restricted, which facilitates accurate measurement of tumor volumes to assess treatment response. By defining the role of the tumor stroma in chemotherapy response in mouse models, we may inform the design of future clinical trials for neurofibromatosis patients with MPNSTs.

Following injection of Ad-Cre into the sciatic nerve of NI^F/F or NI^F/− mice, MPNSTs of similar volume (200–300 mm³) were treated with either vehicle alone or a single bolus of doxorubicin/ifosfamide (15 mg/kg i.p. doxycycline; 50 mg/kg ifosfamide i.v.; Fig. 5B). Response to treatment was monitored by caliper measurement, and mice were sacrificed when tumor volume reached Institutional Animal Care and Use Committee–approved endpoints. The majority of tumors treated with the doxorubicin/ifosfamide regime showed some initial tumor shrinkage and delay in tumor growth (Fig. 5C and D). MPNSTs from both NI^F/− and NI^F/F mice showed moderate partial response rate (50% vs. 45%, respectively, defined as >30% reduction in volume), and there was no difference in the size of tumors 10 days after treatment (1.20-fold vs. 1.15-fold change in size, P = 0.95). Taken together, these data suggest that the status of Nf1 in the surrounding stroma does not alter the growth response of MPNSTs to conventional chemotherapy.

Although there were no differences in tumor growth response between MPNSTs in NI^F/F and NI^F/− mice, we wondered whether there could be differences in the molecular response to cytotoxic agents. Therefore, we analyzed vehicle-treated and doxorubicin/ifosfamide–responding MPNSTs for markers of cell proliferation (Ki67), apoptosis (cleaved caspase-3), and mast cell infiltration (toluidine blue, Fig. 5F; Supplementary Fig. S4). In addition, we analyzed samples 48 hours after doxorubicin/ifosfamide treatment, as this time point has been reported to show maximum tumor cell death following doxorubicin exposure in mice (46). As expected, Ki67 levels were reduced in both genotypes at 48 hours, but were similar between vehicle-treated and responding tumors at terminal harvest. Cleaved caspase-3 levels were elevated at 48 hours post-doxorubicin/ifosfamide, suggesting cell death is increased directly following chemotherapy, but levels decrease by the time of terminal harvest. Importantly, there were no differences in cell death or proliferation markers between MPNSTs in NI^F/F and NI^F/− mice 48 hours post-doxorubicin/ifosfamide. This is similar to the tumor growth data in Fig. 5C and D; and supports the conclusion that these tumors respond similarly to doxorubicin/ifosfamide treatment, even at early time points. The number of mast cells was not altered by doxorubicin/ifosfamide treatment, and the enrichment of mast cells in NI^F/− MPNSTs was still observed across all time points. This suggests that mast cells may not play a role in the response to conventional chemotherapy, which could explain why there is no difference in response between the two genotypes. Taken together, these data using primary mouse models show that MPNSTs in mice respond similarly to a combination treatment with doxorubicin and ifosfamide, irrespective of stromal genotype.

Although NF1-associated MPNSTs are surrounded by a tumor microenvironment of NF1 haploinsufficient cells, it is unknown whether elevated NF1-dependent signaling in these stromal cells influences MPSNT biology. Both anecdotal reports and clinical studies have suggested that NF1-associated MPNSTs have a worse outcome than sporadic MPSNTs, but to our knowledge this has never been addressed in an experimental model, nor has there been any mechanistic explorations of potential biological differences. Here, we used paired littermate mice to model the distinct genetics of the tumor microenvironment in NF1-associated and sporadic MPSNTs. NI^F/− mice, which model Nf1 haploinsufficiency, developed MPNSTs at an accelerated rate with an elevated immune cell infiltrate that was characterized by increased numbers of CD11b⁺CD11c⁻ myeloid cells and mast cells. Analysis of patient samples showed that MPNSTs in NF1 patients are also enriched for the presence of mast cells. Moreover, the Nf1 haploinsufficient hematopoietic compartment was sufficient to accelerate tumor development, further underscoring the importance of the immune system in modifying MPNST biology. While it is possible that Adeno-Cre injection into the nerve of NI^F/− mice may not only delete Nf1 and Ink4a/Arf in Schwann cells, but may also delete these tumor suppressors from surrounding stromal cells, the bone marrow transplant experiment indicates that deletion of Nf1 and Ink4a/Arf in stromal cells does not contribute to the accelerated formation of MPNST in these mice. Instead, the bone marrow transplant experiment demonstrates a fundamental role for the Nf1+/− hematopoietic compartment, which is independent of any effect of Adenoviral infection of stromal cells.

Given the hyperactive Ras signaling that has been reported in several NF1-deficient cells, including endothelial cells, fibroblasts, and immune cells, we initially hypothesized that the number of these cell types would be altered in NI^F/− tumors. The Nf1 haploinsufficient stroma plays a well-known role in the etiology of the neurofibromatosis disease spectrum. Nf1+/− endothelial cells are hyper-responsive to VEGF signaling, which results in neovascularization and altered vascular morphogenesis (47). Fibroblasts haploinsufficient for neurofibromin proliferate at accelerated rates and secrete excess levels of collagen (48). NF1-associated myeloid cells, including mast cells and macrophages, are involved in neurofibroma development and resistance to targeted therapies, respectively (9, 49). However, the role of these cells in the MPNST tumor microenvironment has not been previously explored.

Given the many cell types and functions that are altered by Nf1 loss, we were surprised that only levels of CD45⁺ immune cells were altered in NI^F/− MPNSTs. Initially, we suspected macrophages were the key immune cell population altered in NI^F/− tumors, as multiple studies have shown a key role for macrophages in NF1 disease biology, but IHC showed no difference in macrophage markers. Instead, we found that the increased CD45⁺ population was comprised primarily of CD11b⁺ CD11c⁻ monocytes and mast cells. Further experiments, such as using lineage-specific cell depletion (50), are required to determine whether monocytes, mast cells, or other cell types, are necessary for accelerating formation of NI^F/− MPNSTs. Indeed, a combination of signaling networks in these cell types may be necessary to optimally accelerate tumor formation by the Nf1 haploinsufficient hematopoietic compartment. Several hypotheses for how these cells may alter tumor initiation include influencing angiogenesis or producing a cytokine storm that alters immune cell signaling at early time points.

Mast cells have long been recognized as important contributors to neurofibromatosis disease phenotypes, and in vitro studies have shown that Nf1 haploinsufficient mast cells are hyper-responsive to SCF in the neurofibroma environment. Activated mast cells secrete elevated cytokines, which may accelerate MPNST formation. Mast cells are enriched in NF1-associated neurofibromas and MPNSTs when compared with control skin and nerve (51). Of note, the first clinical trial for neurofibromas in neurofibromatosis patients aimed to limit mast cell numbers via ketotifen administration (7, 52). Experiments in mice showed that neurofibroma formation was influenced by cKit-positive cells in the stroma (9), although it is important to note that the cKit-knockout model is not exclusive to mast cells (53, 54). Treatment of several patients with imatinib, a Kit inhibitor, decreased the growth of head and neck neurofibromas; however, less efficacy was observed for neurofibromas at other locations (55). Taken together, these data support a central role of the mast cell in NF1 biology. Future studies will determine the contribution and mechanism of mast cells and/or monocytes to the Nf1 haploinsufficient tumor microenvironment and MPNST biology.

Generating sarcomas in a temporally restricted manner at a spatially restricted site facilitates measuring tumor volumes and allows accurate assessment of treatment response. Because the primary tumors develop within a native, immunocompetent, tumor microenvironment, we were able to model the effects of an Nf1 haploinsufficient immune system on conventional chemotherapy. In the future, we will be able to use these models to study the impact of an Nf1 haploinsufficient immune system on immunotherapy. The spatially restricted nature of this model also facilitates the analysis of changes in tumor histology at distinct points following chemotherapy treatment. This analysis allowed us to examine changes in proliferation and cell death within the tumor across different time points after treatment. Our data shows that MPNSTs in NI^F/F and NI^F/− mice respond similarly to conventional chemotherapy, helping to clarify the impact of Nf1 haploinsufficient stroma on tumor response to multiagent chemotherapy. Thus, our results may inform future clinical trials because they suggest that the NF1-associated and sporadic MPNSTs can be grouped together in study arms for conventional chemotherapy. Although Nf1 haploinsufficient stroma did not impact response to chemotherapy in our mouse models, our finding of a differential immune environment between NF1-associated and sporadic MPNSTs suggests that response to immunotherapy or other targeted therapies may be impacted by NF1 haploinsufficiency. Future studies using these and other models will help clarify the impact of the distinct NF1-assoicated immune microenvironment on these therapies.

No potential conflicts of interest were disclosed.

Conception and design: R.D. Dodd, C.-L. Lee, W.C. Eward, D.G. Kirsch

Development of methodology: C.-L. Lee, W.C. Eward, L. Luo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.D. Dodd, C.-L. Lee, T. Overton, W. Huang, W.C. Eward, D.R. Ingram, K.E. Torres, A. Lazar

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R.D. Dodd, C.-L. Lee, T. Overton, D.M. Cardona, A. Lazar, D.G. Kirsch

Writing, review, and/or revision of the manuscript: R.D. Dodd, C.-L. Lee, D.R. Ingram, D.M. Cardona, A. Lazar, D.G. Kirsch

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Luo, Y. Ma, D.R. Ingram

Study supervision: W.C. Eward, D.G. Kirsch

We thank Ron Dephino for providing the Ink4a/Arf^Flox/Flox mice. We thank Rich Riedel for helpful suggestions regarding the chemotherapy regimen for the coclinical trial and members of the Kirsch lab for useful discussion and suggestions.

This work was supported by W81XWH-14-1-0067 grant from the Department of Defense (D.G. Kirsch).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.