Prognostic Significance of Immunohistochemical Markers and Genetic Alterations in Malignant Peripheral Nerve Sheath Tumors: A Systematic Review

Background Malignant peripheral nerve sheath tumors (MPNSTs) are aggressive soft tissue sarcomas with dismal prognosis. Pathological and genetic markers may predict more aggressive behavior in MPNSTs but have uncommonly been investigated, and few are used in daily practice. This study reviews the prognostic value of immunohistochemical markers and genetic alterations in MPNST. Methods A systematic search was performed in PubMed and Embase databases according to the PRISMA guidelines. Search terms related to ‘MPNST’ and ‘prognostic’ were used. Studies investigating the association of immunohistochemical markers or genetic alterations with prognosis were included. Qualitative synthesis was performed on all studies. A distinction was made between univariable and multivariable associations. Results Forty-six studies were included after full-text screening. Sixty-seven different immunohistochemical markers were investigated. Absence of S100 and H3K27me3 and high Ki67 and p53 staining was most commonly independently associated with worse survival and disease-free survival. Several genetic alterations were investigated as well with varying association to survival. TP53, CDK4, RASSF1A alterations were independently associated with worse survival, as well as changes in chromosomal length in Xp, 10q, and 16p. Conclusions MPNSTs harbor complex and heterogeneous biology. Immunohistochemical markers and genetic alterations have variable prognostic value. Absence of S100 and H3K27me3 and increased Ki67 can be of prognostic value. Alterations in TP53 or increase in p53 staining may distinguish MPNSTs with worse outcomes. Genetic alterations and staining of other cell cycle regulatory and Ras pathway proteins may also help stratifying patients with worse outcomes. A combination of markers can increase the prognostic value.

Staging of MPNSTs is important to increase accuracy of outcome prediction, but it may also facilitate treatment stratification. However, the clinical American Joint Committee of Cancer (AJCC) STS staging system is less applicable in MPNST (4,5,13). The histologic Fedeŕation Nationale des Centres de Lutte Contre le Cancer (FNCLCC) grading system used in STS is of prognostic value since low grade MPNST (FNCLCC grade 1) has improved survival (2). However, only 10% of MPNSTs are grade 1, and the FNCLCC grading can likely only distinguish prognosis between grades 1 and 3 (2,5). Moreover, the histological distinction between low-grade MPNST and benign neurofibroma with atypia is difficult as objective criteria are lacking, causing interobserver variability. In the context of NF1, the diagnosis of progression to MPNST is even more challenging. Recently, a consensus view has been published defining "atypical neurofibromatous neoplasm of uncertain biologic potential (ANNUBP)" as an intermediate lesion in NF1 patients (14). While driver mutations are increasingly being studied, the transition of neurofibromas to MPNSTs is not yet fully understood. Clinical parameters as predictors of outcome have been studied more commonly, but independent predictors are found inconsistently (3). Although radiation-induced MPNSTs have repeatedly been associated with worse survival, the influence of NF1 disease on survival has been subject of debate (3,13,15). Better classification systems for MPNSTs are therefore urgently needed.
Currently, surgery remains the only proven treatment to improve survival (1)(2)(3). Chemotherapy has limited effect in localized disease, and its use is controversial. Some studies suggest a minor benefit in high-grade, large, and deep MPNST (16)(17)(18). Moreover, 10-20% of patients present with metastatic or unresectable disease and up to 50% of patients will develop metastases over time (1-5, 13, 19). Targeted therapies are warranted, but so far none have been proven effective (20). Immunohistochemical and genetic markers may predict more aggressive behavior in MPNSTs, but their association with oncological outcome has uncommonly been investigated and few are yet used in daily practice for prognostication. For this reason this systematic review set out to summarize current knowledge on the prognostic value of immunohistochemical and genetic markers. Such markers may enhance prognostication and aid in elucidating driver mutations of malignancy.

Literature Search
A systematic search was performed in Embase and PubMed databases according to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines, in order to identify all potentially relevant articles as of March 2020. The string was built with the help of a professional librarian using search terms related to 'MPNST' and 'prognostic'. The exact search syntaxes for PubMed and Embase are shown in Supplementary Table 1. Studies were included that evaluated the association of immunohistochemical markers and genetic alterations to oncological outcomes in MPNST patients. Exclusion criteria included lack of full text or studies without specific analyses fitting our inclusion criteria. The initial review was conducted by two independent authors (EM. and IA). Disagreements were solved through discussion in which one additional author was involved (CV).

Data Extraction and Synthesis
Data extracted from studies included: study period, total number of patients, mean age and range, percentage NF1 patients, markers and genetic alterations investigated for prognostic value, and analyses used to identify prognosticators. For all markers and genetic alterations investigated additional information was extracted: number of patients with survival data, population with 'positive' test, oncological outcome analyzed, and whether its prognostic value was corrected for common clinical prognostic factors. Whenever the marker was independently associated with outcome, the hazard ratio was noted. Common factors for which could have been adjusted in multivariable models included: age, presence of NF1, tumor size, tumor site, metastasis at diagnosis, tumor depth, tumor grade, and surgical margin (3). All results of the predictive value of markers were presented or re-calculated to represent the marker cut-off as a negative predictor of survival. Qualitative synthesis was performed for all studies, summarizing results based on type of analysis. Immunohistochemical markers were further stratified into markers of differentiation, receptors and their ligands, Ras pathway, cell cycle regulation, p53 pathway, vascularization, and others. For each immunohistochemical marker cumulative incidence of univariable and multivariable association to survival (disease-specific or overall) or disease-free survival (recurrence, metastasis, or both) were calculated. screening resulted in 55 potentially relevant articles, of which 46 were selected for qualitative synthesis after full-text screening. Mean age differed between 11 and 50 years old (range of all patients 1-94). Prevalence of NF1 patients in study populations ranged from 0 to 100% (mean: 48.0%). Immunohistochemical markers were studied exclusively in 36 studies, genetic alterations in seven studies, and both in three studies ( Table 1). A total of 67 different immunohistochemical markers and numerous genetic alterations were evaluated (Table 2, Figure 2).

Differentiation
Seven mesenchymal and neuronal differentiation markers were evaluated ( Table 2), most commonly S100 (4,45,46,56,58,62). In univariable analysis complete absence of S100 was found negatively associated with survival in four/six studies. Two studies showed the absence of S100 to be an independent predictor of worse survival with HR 4.5 (95% CI: 2.0-12.1) and HR 6.6 (95% CI: 1.8-23.8) (4,62). All seven markers were also evaluated for association with disease-free survival (DFS). Negative S100 staining was associated with worse DFS in two/ four studies, of which one study showed an independent association (HR 4.2, 95% CI: 1.5-12.3) (62). Negative smooth muscle actin (SMA) and CD57 staining were also found associated with worse DFS in univariable analysis in one study, but not in multivariable analysis (58).

Vascularization
Microvascular density (MVD) and vascular epithelial growth factor (VEGF) staining were evaluated as vascularization markers ( Table 2) (4,28,30,57,59). High MVD was associated with worse survival in one/four studies. This association was also significant in multivariable analyses (HR 7.3, 95% CI: 1.4-38.5) (57). High VEGF staining was associated with worse survival in one/two studies, but this was not studied in a multivariable model (59). No markers were studied for association with DFS.

Extracellular Matrix
Twelve extracellular matrix markers were studied, of which none was evaluated more than once (

Ras Pathway
Ten different Ras pathway proteins were stained, but only phosphorylated MAPK kinase (MEK) was evaluated more than once (

DISCUSSION
he underlying biology of MPNSTs remains complex as is highlighted by the diverse findings of studies included in this review. Many markers and genetic alterations have been proposed to be of prognostic value, yet outcomes are infrequently repeated. Alterations in TP53 or its resulting increased p53 staining were commonly found associated with survival and DFS as were several other proteins and genes involved in cell cycle regulation. Epigenetic modulatory proteins, especially loss of H3K27me3, and more general markers as absence of S100 and increased Ki67 were commonly found to be of prognostic value too.

Prognostication in MPNST
The predictive value of clinical parameters including patient and tumor characteristics has been studied more commonly than immunohistochemical or genetic biomarkers in MPNST.
Increasing age, large tumor size, metastatic disease at diagnosis, and tumors not amenable to complete resection are the most commonly found predictors of worse survival in MPNST (2,3,5,13,45,65). This emphasizes the importance of early diagnosis of MPNST in order to completely resect tumors, along with finding new systemic therapies to improve the prognosis of irresectable and metastatic disease. Non-extremity tumor sites have also been shown to have a negative impact on survival; however, this may be more true for those arising in retroperitoneal or pelvic sites (1,3,5,13,66). Tumor depth used to be incorporated for prognostication in the AJCC staging system for STS, but has varyingly been shown to be of prognostic value in MPNST (2,3,5,13,45,65). The importance of NF1 disease has also been subject of debate. A meta-analysis in 2012 showed no difference in survival for patients in papers published after 2000 (15). However, recent large cohorts did find an independent association with worse survival for NF1 patients (3,13,67,68). Altogether, clinical parameters seem to be able to predict some part of a patient's course of disease. The addition of tumor biology to clinical parameters may further increase our ability to stratify subgroups of patients based on prognosis. TP53 is one of the few recurrently mutated genes found in MPNST. TP53 mutations and high p53 staining were independently associated with survival or DFS in five different studies (4,24,34,43,56). This may indicate that aberrations in this gene may indeed be of clinical importance.
Other genes involved in cell cycle regulation such as CDKN2A and downstream proteins are commonly altered and may not only contribute to tumorigenesis but also be of clinical significance, supporting a belief that dysregulations in this cellular pathway are of overall importance. Loss of polycomb regressive complex 2 (PRC2) complex has recently been shown to be common in MPNSTs due to mutations in EED and SUZ12 (9,69). This results in loss of H3K27me3 which can reliably distinguish high-grade MPNSTs from their benign counterparts by immunohistochemistry (26,70). MPNSTs without loss of H3K27me3 staining may also be associated with less aggressive behavior as many low-grade MPNSTs are known to retain this expression (14,26). Preclinical research on targeted therapies has most frequently shown promising results targeting proteins in the Ras pathway, especially when combined with other target drugs, but unfortunately no clinical trial has proven benefit to date (20). Activated proteins in the Ras pathway, including p-mTOR, p-4E-BP1, p-S6RP, COX-2, and Myc as well as methylation of RASSF1A may however predict worse survival (6,7,27,32). Targeting vascular pathways in MPNSTs may be beneficial, but unfortunately few studies have focused on this. Studies included in this review also showed that increased vascularity, as evidenced by increased microvascular density as well as increased expression of VEGF, may be associated with more aggressive biological behavior (57,59). It seems that many other targets may be of prognostic value as well emphasizing the need for further research into MPNST tumor biology. Survivin markers may for instance stratify a subgroup of patients and survivin has been shown a viable target in a xenograft mouse model (71). Seeing as MPNSTs are heterogenic and markers such as p53 are not be MPNST specific, combined scores of different markers and genetic alterations may be of most clinical importance. Four studies in this review highlight this phenomenon demonstrating increased prognostic value when markers are combined (25,28,34,41).

Strengths and Limitations
Unfortunately, due to the large heterogeneity of published studies meta-analyses were not presumed feasible. All studies included in this review were retrospective of nature inherently harboring bias. None of the markers and genetic alterations found in these studies were prospectively validated. Moreover, many did not evaluate the prognostic value of their markers in a multivariable model nor on their discriminative ability. Studies that evaluated the prognostic value of markers in a multivariable model were nonetheless not always capable to correct for all common clinical variables. MPNSTs are rare sarcomas, which in combination with their complex biology, make it difficult to obtain enough cases to create valuable models. But as shown in this review, several markers and genetic alterations may already be of clinical importance as they have shown an independent association with survival in addition to clinical parameters. Future research should therefore be encouraged to replicate these results using larger datasets obtained by large-scale international collaborations. Important immunohistochemical staining may include Ki67, S100, p53, and H3K27me3 in all patients, and possibly further staining of proteins associated with cell cycle regulation. In turn individual prediction models for MPNST patients specifically may arise taking their significant heterogeneity into account. Such models may better elucidate patient selection for (neo)adjuvant treatment and targeted therapies, which should then be validated in a prospective database. But as MPNSTs remain rare entities one may also turn to exploratory analyses using machine learning techniques on large STS genetic databases to identify attractive genes as biomarkers or prognostic markers in subtypes of STS (72).

CONCLUSION
MPNSTs harbor complex and heterogenic biology and currently lack adequate staging systems. Immunohistochemical markers and genetic alterations are varyingly of prognostic value. Absence of S100 and H3K27me3 and increased Ki67 staining were commonly found to be of independent prognostic value alongside of clinical parameters. Alterations in TP53 or its consequential increase in p53 staining seems to distinguish a subgroup of MPNSTs with worse outcomes. Immunohistochemical staining and associated genetic alterations of proteins involved in cell cycle regulation and the Ras pathway may also help stratifying patients with worse outcomes. Other markers will likely need further evaluation for validation. A combination of markers may increase the prognostic value.

DATA AVAILABILITY STATEMENT
All datasets presented in this study are included in the article/ Supplementary Material.