Prognostic roles for fibroblast growth factor receptor family members in malignant peripheral nerve sheath tumor

Background Malignant peripheral nerve sheath tumors (MPNST) are rare, highly malignant, and poorly understood sarcomas. The often poor outcome of MPNST highlights the necessity of identifying prognostic predictors for this aggressive sarcoma. Here, we investigate the role of fibroblast growth factor receptor (FGFR) family members in human MPNSTs. Results aCGH and bioinformatics analysis identified frequent amplification of the FGFR1 gene. FISH analysis revealed that 26.9% MPNST samples had amplification of FGFR1, with both focal and polysomy patterns observed. IHC identified that FGFR1 protein expression was positively correlated with FGFR1 gene amplification. High expression of FGFR1 protein was associated with better overall survival (OS) and was an independent prognostic predictor for OS of MPNST patients. Additionally, combined expression of FGFR1 and FGFR2 protein characterized a subtype of MPNST with better OS. FGFR4 protein was expressed 82.3% of MPNST samples, and was associated with poor disease-free survival. Materials and Methods We performed microarray-based comparative genomic hybridization (aCGH) profiling of two cohorts of primary MPNST tissue samples including 25 patients treated at The University of Texas MD Anderson Cancer Center and 26 patients from Tianjin Medical University Cancer Institute and Hospital. Fluorescence in situ hybridization (FISH) was used to validate the gene amplification detected by aCGH analysis. Another cohort of 63 formalin-fixed paraffin-embedded MPNST samples (including 52 samples for FISH assay) was obtained to explore FGFR1, 2, 3, and 4 protein expression by immunohistochemical (IHC) analysis. Conclusions Our integrated genomic and molecular studies provide evidence that FGFRs play different prognostic roles in MPNST.


INTRODUCTION
Malignant peripheral nerve sheath tumors (MPNSTs) are highly malignant sarcomas derived from the neural crest and account for approximately 5-10% of all soft tissue sarcomas [1,2]. MPNSTs occur either sporadically or in association with neurofibromatosis type 1 (NF1). The overall incidence of MPNST in the general population is 1/100000, of which 5-42% are associated with NF1 [2]. The relative rarity of MPNST and the lack of any specific diagnostic, radiological, or pathological signature pose considerable management challenges for this disease. Even with multidisciplinary treatment, the prognosis for patients with MPNST remains very poor [3]. Therefore, identification of novel prognostic features and therapeutic strategies are required to benefit patients with this aggressive sarcoma.
The fibroblast growth factor (FGF) receptor (FGFR) signaling pathway regulates multiple biological processes, including normal cell growth, survival, differentiation, and angiogenesis. FGFRs are also implicated in tumor development [4,5]. FGFRs possess an extracellular ligand domain that comprises three immunoglobulin-like domains (I-III), a transmembrane domain, and an intracellular tyrosine kinase domain that transmits signals induced by ligand binding to the interior of the cell [6]. There are four cell-surface FGFRs (FGFR-1-4) and seven FGFR isoforms (1b, 1c, 2b, 2c, 3b, 3c, and 4). These isoforms result from alternative splicing of exons coding for Ig-III-likes domains and account for different ligand-binding specificities [7,8].
Identification of the roles and relationships within the FGF/FGFR family and of their links with tumor growth and progression is critical for the design of novel drug therapies to target FGFR pathways. FGFR1 inhibitors are considered potential therapeutic agents in FGFR1-amplified lung squamous cell carcinoma, and early-stage clinical trials have been conducted [9][10][11]. Additionally, high expression levels of FGFR1, FGFR2, or FGFR4 are associated with tumor progression and poor survival in patients with gastric cancer [12]. FGFR1 expression in cancer-associated fibroblasts also has prognostic value in head and neck squamous cell carcinoma [13].
We evaluated the prognostic value of FGFR expression in MPNST using a microarray-based comparative genomic hybridization (aCGH) method, fluorescence in situ hybridization (FISH), and immunohistochemical (IHC) methods to evaluate the gene status and protein expression levels of FGFR1-4 in MPNST samples. Contrasting with the role of FGFRs in epithelial cancers, high expression of FGFR1 predicted better overall survival (OS) for MPNST patients. Furthermore, combined high expression of FGFR1 and FGFR2 protein characterized a subtype of MPNST with better OS, while increased FGFR4 protein expression predicted worse disease-free survival (DFS).

RESULTS aCGH and FISH detected and validated alterations to FGFR genes in MPNST
Integration of copy number profiles of 51 individual MPNST samples revealed frequent gene deletions and amplifications ( Figure 1A; Table 1). Bioinformatics analysis revealed that the amplification rate of the FGFR1 gene was 37% in MPNST samples ( Figure 1B). The deletion rate of the FGFR2 gene was 41%, while that of FGFR3 was 27%. There were no significant alterations to FGFR4.
We identified FGFR1 gene copy amplification in 26.9% (14/52) samples. NF1-positive cases had a higher frequency of FGFR1 gene amplification (χ 2 = 5.091, p = 0.024). FGFR1 amplification was not correlated with prognosis or any other clinical variables, including gender, age, tumor site, American Joint Committee on Cancer (AJCC) staging, tumor recurrence, or metastasis (Table 2). Furthermore, survival analysis demonstrated that FGFR1 amplification had no significant impact on DFS (Supplementary Figure S1A
Meanwhile, recurrence and FGFR1 protein expression were independent prognostic predictors of OS ( An MPNST subtype is characterized by FGFR1 gene amplification, high expression of FGFR1 protein, and improved OS High expression of FGFR1 protein indicates a poor prognosis for multiple tumor types, including lung cancer, triple-negative breast cancer, and gastric cancer [12,14,15]. Contrastingly, we found that high expression of FGFR1 indicated a favorable prognosis for MPNST patients. We further investigated the prognostic role of FGFR1 in MPNST by separating the 52 cases of MPNST samples based upon FGFR1 gene amplification, forming an amplified group (n = 14) and a normal group (n = 38). FGFR1 protein expression significantly influenced OS in the FGFR1-amplified group (χ 2 = 5.311, p = 0.021; Figure 2D), but had no relationship with DFS (Supplementary Figure S1D). Expression of FGFR1 had no significant effect on DFS or OS in the normal group (Supplementary Figure S1E, S1F). These findings reveal that FGFR1 gene amplification characterizes a special subtype of MPNST patients with higher expression of FGFR1 protein and improved OS.

Combined expression of FGFR1 and FGFR2 characterizes a subtype of MPNST patients with better prognosis
Our initial aCGH study revealed a deletion rate of 41% for FGFR2. We next used IHC to evaluate expression of FGFR2 protein ( Figure 3A, 3B). Staining for FGFR2   Figure S2A, S2B), or any other clinical variables assessed. We next regrouped patients based on FGFR1 and FGFR2 protein expression as follows: both positive, both negative, or single-positive for either. Patients expressing both FGFR1 and FGFR2 exhibited the best OS (χ 2 = 6.215, p = 0.045; Figure 3C). However, DFS was not significantly different among the three groups (Supplementary Figure S2C). Combined expression of FGFR1 and FGFR2 enhanced the prognostic value of FGFR1 and FGFR2 in MPNST.

Increased FGFR4 protein expression predicted worse DFS
Bioinformatics analysis revealed a deletion rate of 27% for FGFR3 in MPNST. We used IHC to assess expression of FGFR3 protein in 63 MPNST samples. However, compared with the positive FGFR3 protein expression in KMS-11 cells (Supplementary Figure S2D), we did not identify FGFR3 protein expression in any of the 63 MPNST cases (Supplementary Figure S2E).
No significant aberrations in the FGFR4 gene were detected by aCGH analysis of 51 MPNST tissue samples. However, high levels of FGFR4 protein were present in 82.3% (51/62) of MPNST cases ( Figure 3D, 3E). Survival analysis revealed that high expression of FGFR4 conferred worse DFS compared with the low expression group (χ 2 = 4.546, p = 0.033; Figure 3F; Table 4). However, high expression of FGFR4 in MPNST was not correlated with OS or any other clinical variables assessed (Supplementary Figure S2F).

DISCUSSION
MPNST has been previously known as malignant schwannoma, malignant neurilemmoma, neurogenic sarcoma, and neurofibrosarcoma [16]. MPNST is a rare but highly malignant sarcoma of soft tissues that occurs either sporadically or in association with NF1 [17]. The incredibly poor prognosis for MPNST highlights the need for new prognostic markers in this disease. Additionally, an improved understanding of the key genetic and molecular events involved in MPNST development and progression are critical for the development of effective therapeutics. We conducted genomic and molecular studies of human tumor samples to characterize the roles played by FGFRs in MPNST. Our findings provide evidence of different genetic aberrations among FGFRs in MPNST. Additionally, we demonstrate that FGFRs are potentially useful markers for prognosis of MPNST.
The expression of FGFRs in neurogenic tumors has been poorly studied. Therefore, a major contribution of this study is identification of the genetic status of FGFR family member expression in MPNST by aCGH and validation of FGFR1 gene amplification by FISH. Gene dosage is an important regulator of FGFR protein expression [18,19]. Consistently, we found that FGFR1 gene dosage played an important role in increasing FGFR1 protein expression. We also identified deletion of the FGFR2 and FGFR3 genes, consistent with the low or negative expression their corresponding proteins. Meanwhile, no significant alterations to the FGFR4 gene were identified, though FGFR4 protein was expressed at a high level. These findings indicate that overexpression of FGFRs is dependent on both the gene copy number and other mechanisms such as aberrant transcriptional regulation.
Additionally, internalization and degradation of the activated FGF-FGFR complex is an important consideration [5]. The four FGFRs are ubiquitinated to different extents, and this ubiquitination appears to dictate whether these receptors are transported back to the plasma membrane or degraded. FGFR1 generally undergoes extensive ubiquitination and is transported to lysosomes for degradation. Meanwhile, FGFR4 is only lightly ubiquitinated, inefficiently degraded, and is preferentially recycled to the cell surface. Therefore, further evaluation of the mechanisms underlying the differential regulation of FGFR family members is important for a more complete understanding of the precise roles of FGFRs in tumor development.
We identified improved prognosis for MPNST patients with higher expression of FGFR1. Additionally, FGFR1 protein expression is an independent prognostic predictor of improved OS. However, FGFR1 expression is associated with unfavorable outcomes in other tumors, including lung cancer, triple-negative breast cancer, and gastric cancer, and FGFR1 may be a useful therapeutic target in these diseases [12,14,15]. Shin et al. identified upregulation of FGFR1, FGFR2, and FGFR4-but not FGFR3-in gastric cancer tissues [20]. Another study identified high expression of all four FGFRs in gastric cancer tissues [12]. Additionally, a reciprocal relationship between FGFR1 and FGFR3 in colorectal tissue has   been characterized [21]. These apparent discrepancies in FGFR expression among studies may be attributable to differences in disease stage or the techniques used for protein expression analysis. Importantly, although FGF signaling can promote tumorigenesis, in certain contexts FGFRs can also mediate anti-tumor functions. FGFR expression can be down-regulated though autocrine or paracrine FGF-FGFR signaling loops out-of-context, indicating that FGFRs may act as tumor suppressors in certain contexts [5,21]. For example, if an FGFR-expressing cell also overexpresses the corresponding ligand, an autocrine loop can be established and the cancer cell becomes self-sufficient in growth signals.
FGFRs play important roles in tumor progression by activating downstream signaling pathways. Two classes of regulator modulate signaling output from activated FGFRs. Negative regulators include Sprouty proteins, mitogen-activated protein kinase phosphatase 3(MKP3), and similar expression to FGF protein [22,25] Positive regulators of FGFR signaling include the fibronectin-leucine-rich transmembrane proteins (FLRT)-1-3 [26,27]. Therefore, high expression of FGFR1 may simply serve as a marker of favorable prognosis rather than being the actual mechanism underlying the favorable prognostic outcome for these MPNST patients.
Our data reveal that the different FGFRs have different prognostic roles for MPNST. FGFR1 expression was an independent predictor of favorable OS, while FGFR4 protein expression predicted poor prognosis. These differences may reflect the differing physiological functions  of FGFR1 and FGFR4. Consistently, the different FGFRs play an assortment of roles in numerous tumors and tissues, and co-expression of FGFRs in various combinations may cause subtle changes in the progression of cancer [28,29]. Contrasting with the role of FGFRs in epithelial cancers, FGFR1 and FGFR2 protein expression characterized a subtype of MPNST with better OS. Additionally, overexpression of FGFR4 is associated with advanced stage cancer and poor survival in rhabdomyosarcoma (RMS). FGFR4 protein is expressed in the two main variants of RMS-embryonal RMS and alveolar RMS (aRMS). Interestingly, only certain subgroups of aRMS cells are rescued by FGFR4 signaling following induction of apoptosis by compounds targeting the IGF1R-PI3K-mTOR pathway [30][31][32]. Therefore, expression patterns of FGFRs could facilitate selection of patients for adjuvant systemic therapy. Importantly, as this effect is seen at the protein level, the biomarker panel can be readily implemented in routine clinical testing using IHC.
We used integrated genetic and molecular profiling to confirm genetic alterations of the FGFR1-4 genes and measure expression of the corresponding proteins in MPNST tissues. High expression of FGFR1 was an independent prognostic predictor of OS, and tumors with higher expression of FGFR1 had better prognosis than those with lower expression. Additionally, patients with tumors expressing high levels of FGFR4 had worse DFS. However, the molecular details underlying the observed effects are largely unknown. Additional thorough investigations and clinical trials are needed to enhance our understanding of the different prognostic roles for FGFR family members in MPNST.

Patients and primary tumors
Patient information collected included age, sex, NF1 status, tumor location, largest diameter of the tumor, clinical AJCC stage, time to recurrence, metastatic status, treatments, and outcome. The presence of the NF1 syndrome was determined by the NIH criteria. Collection of tissue and information for this retrospective study was approved by the Institutional Review Boards at Tianjin Medical University Cancer Institute and Hospital and The University of Texas, MD Anderson Cancer Center, and occurred with patient consent [33,34].
A cohort of 51 patients with histologically confirmed MPNST and matching patient records was included in this study for genome-wide copy number measurements using the aCGH method [33,34]. Of these samples, 25 FFPE tumor specimens were acquired from The University of Texas MD Anderson Cancer Center, and another 26 fresh tumor samples were from the Tianjin Cancer Hospital of China [33,34]. IHC validation was performed using an independent cohort of 63 FFPE tumor samples acquired from TMUCIH, with 52 of these FFPE tumor samples subjected to FISH for validation of specific copy number aberrations.

Array CGH hybridization and bioinformatic analysis
Genome-wide copy number levels for the 51 primary tumor samples were mapped using aCGH with commercially available normal genomic DNA as a reference (Clontech Laboratories, Inc., Mountain View, CA) [33]. Genomic DNA was isolated from tumors according to standard procedures and the labeled genomic DNAs were hybridized using the Agilent 4 × 44 k Human Genome CGH Microarray kit (Agilent Technologies, Santa Clara, CA). Processing of aCGH data and frequency analyses were performed as described previously [33,34]. Briefly, the ratios of intensity values from tumor and normal tissues were transformed to log2-space. Log ratio data were then subjected to a circular binary segmentation algorithm to reduce the effect of noise. After that, the CGHcall algorithm was used to give each segment an aberration label: normal, deletion, or amplification. An aberration frequency for each probe was established by combining the labels from individual samples.

FISH analysis
The ZytoLight SPEC FGFR1/CEN 8 Dual Color Probe (ZytoVision, Bremerhaven, Germany) was used for detection of FGFR1 gene amplification. The CEN8 probe exhibits an orange signal and indicates the centromere of chromosome 8, while the FGFR1 probe exhibits a green signal. FISH was performed as previously described [35]. Briefly, deparaffinized sections were pretreated by incubation with deionized water at 95°C for 20 min followed by digestion with pepsin at 37°C for 50 min. Tissue sections and FGFR1/CEN8 FISH probes were then denatured at 78°C for 10 min and incubated overnight at 37°C to allow hybridization. Slides were then washed and samples counterstained with 4′, 6-diamidino-2-phenylindole and mounted with coverslips.
Alterations of FGFR1 gene copy number were evaluated by two pathologists in a blinded fashion according to established methods [36][37][38]. An increase in FGFR1 gene copy number was defined as an FGFR1/ CEN8 signal ratio ≥ 1.5 and when two or more FGFR1 gene copies were present per cell in > 90% of MPNST cells [33].

Statistical analysis
Correlations between FGFR expression and clinicopathological variables were analyzed using the Chi-square test. Patient survival curves were plotted according to the Kaplan-Meier method and a log-rank test. Multivariate Cox regression analysis was used to identify significant independent prognostic factors. OS was defined as the time period from the date of diagnosis to death or the last follow-up. For DFS analysis, the duration was defined as the time from diagnosis until the occurrence of metastasis or recurrence. A two-sided p-value < 0.05 was considered statistically significant. All statistical analyses were performed using SPSS version 19.0 statistical software (SPSS, Chicago, IL).