The impact of histopathology and NAB2-STAT6 fusion subtype in classification and grading of meningeal solitary fibrous tumor/hemangiopericytoma.

Meningeal solitary fibrous tumor (SFT)/hemangiopericytoma (HPC) is a rare tumor with propensity for recurrence and metastasis. Although multiple classification schemes have been proposed, optimal risk stratification remains unclear, and the prognostic impact of fusion status is uncertain. We compared the 2016 WHO CNS tumor grading scheme (CNS-G), a three-tier system based on histopathologic phenotype and mitotic count, to the 2013 WHO soft-tissue counterpart (ST-G), a two-tier system based on mitotic count alone, in a cohort of 133 patients [59 female, 74 male; mean age 54 years (range 20-87)] with meningeal SFT/HPC. Tumors were pathologically confirmed through review of the first tumor resection (n = 97), local recurrence (n = 35), or distant metastasis (n = 1). A STAT6 immunostain showed nuclear expression in 132 cases. NAB2-STAT6 fusion was detected in 99 of 111 successfully tested tumors (89%) including the single STAT6 immunonegative tumor. Tumors were classified by CNS-G as grade 1 (n = 43), 2 (n = 41), or 3 (n = 49), and by ST-G as SFT (n = 84) or malignant SFT (n = 49). Necrosis was present in 16 cases (12%). On follow-up, 42 patients had at least one subsequent recurrence or metastasis (7 metastasis only, 33 recurrence only, 2 patients had both). Twenty-nine patients died. On univariate analysis, necrosis (p = 0.002), CNS-G (p = 0.01), and ST-G (p = 0.004) were associated with recurrence-free (RFS) but not overall survival (OS). NAB2-STAT6 fusion type was not significantly associated with RFS or OS, but was associated with phenotype. A modified ST-G incorporating necrosis showed higher correlation with RFS (p = 0.0006) and remained significant (p = 0.02) when considering only the primary tumors. From our data, mitotic rate and necrosis appear to stratify this family of tumors most accurately and could be incorporated in a future grading scheme.


Introduction
Meningeal solitary fibrous tumor (SFT)/hemangiopericytoma (HPC) is an often aggressive mesenchymal tumor of fibroblastic origin that arises from the cranial or spinal dura [14]. Although SFT and HPC were initially thought to represent distinct entities, the identification of NAB2-STAT6 fusion as a defining molecular alteration in both tumors has led to the unification of these entities at both dural and extra-dural sites. While the NAB2-STAT6 fusion seems to be unique to SFT/HPC, its detection may be difficult unless whole-genome sequencing is applied to detect breakpoints that occur both in exon and intron boundaries [5,16]. STAT6 nuclear expression is accepted as a sensitive surrogate of all fusions, which causes consistent nuclear relocation of STAT6 [9,14,17]. Tumors arising from the meninges tend to exhibit high rates of local recurrence with propensity for metastasis outside the central nervous system (CNS), but prognostication on histopathology alone has been notoriously difficult regardless of site of origin. The current 2016 WHO CNS grading scheme incorporates phenotype and mitotic rate to stratify tumors into three groups (grade 1, 2, and 3), while non-meningeal soft-tissue tumors are classified by mitotic rate alone into SFT and malignant SFT according to the 2013 WHO classification for soft-tissue tumors [10,14]. Our earlier work suggested that tumors harboring the NAB2 exon 4-STAT6 exon 3 fusion variant exhibited morphologic features similar to the conventional solitary fibrous tumor, while there was a trend toward an association with the hemangiopericytoma phenotype and more aggressive behavior in tumors lacking this variant [11]. We studied a large series of patients with SFT/HPC from six tertiary care centers, to determine the best grading scheme for meningeal-based SFT/HPT and characterize the relationship of NAB2-STAT6 fusion status with phenotype and prognosis.

Materials and methods
This study was reviewed and approved by the Institutional Review Board at all participating institutions.

Case selection
A cohort of 133 patients with meningeal SFT/HPC (74 males; 59 females) was identified from six tertiary care centers. They ranged in age from 17 to 78 years (median 49.2) at the initial diagnosis. Tumors were pathologically confirmed through review of the first tumor resection (n = 97), local recurrence (n = 35), or distant metastasis (n = 1). Data regarding treatment and clinical follow-up were obtained at the respective institutions based on review of the clinical records.

Histologic review
Archived H&E-stained sections were classified phenotypically as 'SFT', 'HPC', or tumors with intermediate morphological features between HPC and SFT (Fig. 1). SFTs were low-to-moderate cellular tumors composed of spindled to ovoid-shaped cells arranged around branching blood vessels with variable stromal and perivascular hyalinization. HPCs harbored ovoid-to-round cells, often in a sheet-like growth pattern, with high N:C ratios and scant, amphophilic to-clear cytoplasm. Although branching blood vessels were typically present, the vasculature of HPC was frequently less conspicuous than the conventional SFT, and stromal hyalinization was generally absent. Tumors showing the features of both SFT and HPC were considered intermediate. Mitotic rate was assessed by scanning all available H&E slides and counting mitoses (at × 400) in ten consecutive fields in the areas of highest mitotic activity. The presence of necrosis was recorded for all tumors.

Classification and grading
Tumors were classified and graded according to the most recent CNS (2016) (CNS-G) and Soft Tissue (2013) (ST-G) WHO classification schemes (Fig. 2). According to the CNS-G, tumors with a classic SFT histopathological phenotype and fewer than five mitoses (× 10 HPF) were considered grade 1; tumors with intermediate or HPC phenotype and fewer than five mitoses (× 10 HPF) were considered grade 2; tumors with five or more mitoses (× 10 HPF) were considered grade 3 irrespective of their histopathological phenotype. According to the ST-G, irrespective of their histopathological phenotype, tumors with fewer than five mitoses (× 10 HPF) were considered SFT and tumors with five or more mitoses (× 10 HPF) were considered malignant SFT.

Immunohistochemistry
A representative tissue block from each tumor was selected and stained for STAT6 (Santa Cruz Biotechnology, CA, mouse monoclonal antibody, 1:50) and CD34 (Novocastra, liquid mouse monoclonal antibody, 1:100). CD34 immunoreactivity was recorded as negative (< 5%), focal (5-50%), or diffuse (> 50%), while nuclear expression of STAT6 was scored as negative (when nuclear STAT6 expression was not present in tumor cells) or positive (when the cells showed definite STAT6 nuclear expression at least focally in the tumor).

Molecular studies
RNA extraction was performed by the Mayo Clinic Pathology Research Core using the Qiagen miRNeasy FFPE kit by methods previously described by Wang et al. [20]. RT-PCR of hypoxanthine phosphoribosyltransferase 1 (HPRT1) was used as control for RNA integrity as described previously [3,11] and samples with degraded RNA were discarded to avoid generation of false-negative cases. In a first step, we screened for the presence of the most common NAB2-STAT6 fusions exon 4-exon 2, exon 4-exon 3, exon 6-exon 16, and exon 6-exon 17 using the well-established single-plex PCR with subsequent agarose gel electrophoresis, identifying NAB2-STAT6 gene fusions in 52 samples. In a next step, samples with good RNA quality but no result in the single-plex PCRs were analyzed by next-generation sequencing using the Archer FusionPlex Sarcoma Kit (ArcherDx, Inc, Boulder, CO, USA) to detect fusions among 26 genes employing the Anchored Multiplex PCR-based enrichment, identifying NAB2-STAT6 gene fusions in an additional 47 samples. Briefly, a total of up to 250 ng RNA were used as starting material. Library preparation was performed according to the manufacturer's protocol (ArcherDx, Inc) and sequencing was done on a NextSeq550 instrument using NextSeq500/550 High Output v2 kit (150 cycles) (Illumina, Inc., San Diego, CA, USA). The resulting raw data were converted to fastq files and were then processed with the automated Archer Analysis Bioinformatics Platform (Version Archer Analysis 5.1, ArcherDx, Inc).

Statistical methods
Patient and tumor characteristics were summarized with frequencies and percentages or medians, interquartile ranges (IQR), or ranges, as appropriate. Fisher's exact tests were used to compare categorical variables between selected groups (e.g., CNS-G, ST-G, and molecular cluster type). Our correlative analysis was based on the time of surgery, which represents when the first tumor was evaluated at the  institution. This was the primary tumor resection for 97 patients, resection of a recurrent tumor for 35 patients, and of a metastasis for 1 patient. This approach was made necessary by the fact that, in 36 patients, the primary tumor was not available, and no assumption could be made regarding the grade of the primary tumor. A recent study has shown histological progression of tumor at recurrence in 16% of cases [2]. Recurrence-free survival (RFS) was defined as the time between surgery and the first adverse event (local recurrence or metastasis) following surgery, censoring patients with no adverse event at time of last follow-up. Overall survival (OS) was defined as the time between surgery and death (any cause), censoring those still alive at last followup. OS and RFS were estimated using the Kaplan-Meier method, and were summarized at 5, 10, and 20-year postsurgery, along with the median survival. OS and RFS were compared between selected groups with Cox proportionalhazard regression models, using the likelihood ratio test to assess significance. 95% confidence intervals (CI) were also reported for the survival estimates and hazard ratios (HR). We also investigated RFS from time of original diagnosis to the first adverse event in the patient's history (this also considers the time interval prior to when a recurrent sample was evaluated), as well as OS from time of original diagnosis. For disease-free survival, deaths due to disease were the events, and all non-events (deaths with unknown or other causes, along with those still alive) were censored. All analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC, USA), or R [19]. p values less than 0.05 were considered statistically significant. Table 1 summarizes the characteristics of the patient cohort.

Clinical outcome
Follow-up was available in 129 of 133 patients (range 4 days-22.2 years; median 5 years). At last follow-up, 104 patients were alive (47 without disease, 24 with disease, and 33 alive but their disease status was unknown) and 29 patients were deceased (20 of disease, 6 of other causes, and 3 of unknown cause). The median recurrence-free (RFS) and overall survival (OS) times from the original diagnosis were 9.6 years and 20.9 years, respectively (Fig. 3). When RFS and OS were calculated from the time at which the first surgical specimen was available for pathology review (indicated as time of surgery), the median RFS and OS were 11.3 years and 14.7 years, respectively. The differences in RFS and OS with respect to time of diagnosis versus time of surgery are explained by the influence of the patients in which the first available specimen was a recurrence rather than a primary tumor, as illustrated in suppl. Figure 1 and suppl. Figure 2 (Online Resource 1 and Online Resource 2). When considering only cases in which the primary tumor was available for review, the median RFS for primary tumors was 12.9 years (79.0% at 5 years) compared to 7.8 years (67.6% at 5 years) for recurrent cases (p = 0.14). The median OS from the time of surgery for cases in which the primary tumor was available for review was 14.7 years (89.1% at 5 years) compared to 9.6 years (70.3% at 5 years) for those in which the recurrence was the first available material (p = 0.03). The disease-free survival for all patients at 5 years was 87.2% (95% CI 80.3, 94.1) and at 10 years was 74.2% (95% CI 62.6, 85.7). For patients in which the primary tumor was reviewed, the 5-and 10-year survivals were 91.5% (95% CI 84.9, 98.1) and 85.5% (95% CI 75.2, 95.7), respectively, while the 5-and 10-year survivals for patients in which a recurrence was reviewed was 85.5% (95% CI 75.2, 95.7) and 43.8% (16.3, 71.3), respectively.
Of the entire cohort, 42 patients experienced at least one adverse event after surgery (local recurrence or metastasis): 7 patients had metastases only, 33 had local recurrence only, and 2 patients had both local recurrence and metastases ( Table 2). When looking only at the patients from whom the primary tumor was reviewed (n = 97), 28 patients experienced either recurrence or metastasis: 4 patients had metastases only, 22 patients had recurrence only, and 2 patients had both recurrence and metastases.
Among those in the full cohort who experienced an adverse event, the median time from original diagnosis to the first local recurrence was 5.6 years (range 27 days-36.7 years), while the median time to first metastasis was 8.8 years (range 7.1 months-23.4 years). For patients from whom the primary tumor was reviewed, the median time to first recurrence was 3.5 years (range 27 days-14.9 years) with median time to first metastasis of 4.5 years (range 7.1 months-10.3 years).

Immunohistochemistry
CD34 was recorded as diffuse in the majority of cases (n = 76; 57%), while 32 cases (24%) were focally positive and 25 were negative (19%). STAT6 immunostain showed nuclear expression in 132 (99%) cases. The single case lacking nuclear expression of STAT6 was a T11 spinal mass from a 40-year-old woman that recurred multiple times. STAT6 immunochemistry was performed on two separate specimens from this patient and repeated at two different centers (Mayo Clinic and the originating institution), and all samples tested were negative. Both specimens of this tumor were tested by PCR and showed an NAB2 exon 6-STAT6 exon 16 fusion.

Grading
Tumors were classified as grade 1 (n = 43), 2 (n = 41), or 3 (n = 49) using the CNS-G, and SFT (n = 84) or malignant SFT (n = 49) using the ST-G scheme (Fig. 2). When comparing the extent of CD34 immunostaining among those with at least 5% staining, tumors with SFT phenotype were  one death due to disease with unknown date of death and that patient is not included in these plots more likely to show diffuse CD34 immunoreactivity (88%) as compared to tumors with HPC and INT phenotype (35% and 63%, respectively), p = 0.0001. Furthermore, CNS-G 1 tumors were more likely to show diffuse CD34 immunoreactivity (85%) as compared to CNS-G 2 (56%) or 3 (68%), p = 0.03. On univariate analysis, both the CNS-G and ST-G were significantly associated with RFS (Table 3, p = 0.014, 0.004, respectively) (Fig. 4a, b) but not overall survival (p = 0.23, p = 0.12, respectively). Five-year RFS rates were, respectively, 82, 88, and 60%, while 10-year rates were 61, 58, and 34% for CNS grades 1, 2, and 3. They were 85 and 60 at 5 years, and 59 and 34% at 10 years for SFT and malignant SFT, respectively. When considering only the primary tumors, ST-G remained significant (p = 0.03), while the CNS-G did not (p = 0.07). When evaluating only the recurrent cases, the ST-G approached significance (p = 0.05), while the CNS-G did not (p = 0.15). Since the ST-G scheme appeared to be a simpler way to stratify these tumors, and necrosis was associated with decreased RFS on univariate analysis, we attempted to incorporate necrosis into the ST-G and stratify tumors into three tiers: (1) tumors with < 5 mitoses/10 HPF, (2) tumors with ≥ 5 mitoses/10 HPF without necrosis, and (3) tumors with ≥ 5 mitoses/10 HPF with necrosis. This modified soft-tissue grading scheme showed a strong association with recurrence-free survival (p = 0.0006) and remained significant (p = 0.02) when considering only the primary tumors and only recurrent tumors (p = 0.04) (Fig. 5a-c). Among 84 tumors with < 5 mitoses per 10 HPF, necrosis was present only in 3 (3.6%). All three tumors had a hemangiopericytomatous-like phenotype. In two patients, necrosis was present at the time of primary surgery. One of these patients had GTR and one STR. Both received radiation therapy. They were both alive without recurrence, respectively, at 67 and 38 months. In the third patient, we reviewed a recurrent sample, 16 years and 3 months from the original diagnosis. Only 1 month postoperative follow-up was available in this patient. Fusion cluster and mitotic rate Tumors with NAB2 exon 4-STAT6 exon 2-3 or "other fusions" were more likely to have mitotic rates < 5/10 HPF, whereas tumors with NAB2-STAT6 5-7-STAT6 exon 16-17 and without a detectable fusion were more likely to have mitotic rates ≥ 5/10 HPF (p = 0.002).

NAB2-STAT6 fusion
Fusion cluster and necrosis There was no significant association between fusion type or status (absence or presence of a detected fusion) and necrosis (p = 0.16; p = 0.35, respectively).
Fusion cluster and CNS-G There was a statistically significant association between type of fusion and CNS-G Table 2 Recurrence and metastasis since original diagnosis a 17 patients had at least one metastasis (five of these patients' metastasis was in multiple sites). Two of these patients had two metastases (brain followed by brain; bone followed by epidural) (p < 0.0001). The 29 tumors with NAB2 exon 4-STAT6 exon 2-3 were more frequently grade 1 tumors (n = 21) than grade 2 or 3 (n = 3, n = 5, respectively). The 60 tumors with NAB2 exon 5-7-STAT6 exon 16-17 fusion were more frequently grade 2 and 3 tumors (n = 22, n = 31, respectively).

Fusion cluster and ST-G
There was a statistically significant association between type of fusion and ST-G (p = 0.002). The 29 tumors harboring NAB2 exon 4-STAT6 exon 2-3 or the 10 with "other fusions" were more likely to be solitary fibrous tumors (n = 24, n = 8) compared to

Treatment data
Sixty-three patients (56.3%) underwent gross total resection, while the tumor was subtotally resected in 49 patients (43.8%). In the remaining 21 patients, the extent of surgical resection was unknown. Sixty patients (53.1%) received only adjuvant radiation therapy, a single patient (1%) was treated with chemotherapy alone, 3 patients (2.7%) underwent both adjuvant radiation and chemotherapy, and the remaining 49 patients (43.4%) were treated with surgery alone. Neither extent of resection nor adjuvant radiation therapy was significantly associated with outcome (overall survival: p = 0.051, 0.30; recurrence-free survival: p = 0.27, 0.79, respectively).

Discussion
Our study confirms the natural history of meningeal-based SFT/HPC, showing high rates of local recurrence and distant metastasis. Furthermore, patients in our cohort demonstrate a tendency for late adverse events, a subset developing their first recurrences/metastasis over 3 decades after the initial diagnosis. Similarly, patients continue to die of disease almost 40 years after the initial diagnosis. In a subset of our cases, only material from recurrence (not the primary tumor) was available for review. As these patients are at greater risk for adverse events, as shown by our data, these cases influence the OS and RFS of the entire cohort. However, trends for overall survival and RFS of both primary and recurrent cases remain similar. On univariate analysis, mitotic rate and necrosis were each significantly associated with RFS but not OS. Since the CNS-G and ST-G schemes take mitotic rate into account, it is not surprising that both appear to show association with RFS. However, there does not appear to be a significant difference between grade 1 and grade 2 according to CNS-G, and the ST-G scheme, based solely on mitotic rate, appears to stratify tumors in a simpler and more efficient fashion. Furthermore, the ST-G maintains statistical significance when looking only at primary cases. A recent risk stratification model proposed by Demicco et al. for non-meningeal tumors incorporates mitotic rate and necrosis, along with patient age and tumor size, and appears to show prognostic significance in determining propensity for metastatic disease [8]. The class sizes used in soft tissue with cut-offs of 5, 10, and 15 cm would not be easily applicable to intracranial tumors, which grow in a confined space. However, necrosis was associated with decreased RFS on univariate analysis in our series. Furthermore, incorporating necrosis into ST-G, i.e., classifying tumors based on mitotic rate and necrosis, seems to improve the outcome models when compared to mitotic rate alone. In their recent work from 2018, Macagno et al. sought to validate an updated version of the Marseille Grading Scheme which segregated 132 tumors into three groups based on mitotic activity (≥ 5 mitotic figures/10 HPF) and necrosis [4,15]. On univariate analysis, they found that extent of surgery, radiotherapy/chemotherapy, and mitotic rate were significant in predicting RFS, while radiation, mitotic rate, and necrosis were significant for disease-specific survival [15]. Extent of surgery and mitotic rate remained significant prognostic factors for RFS with multivariate analysis, while necrosis and radiotherapy were significant for disease-specific survival [15]. Our data parallel these findings, confirming that a grading scheme incorporating mitotic rate and necrosis best stratifies SFT/HPC, and the combination of high mitotic activity and necrosis portends poor prognosis for this entity. If these findings are corroborated by additional studies, this grading scheme could be utilized to guide treatment decisions regarding adjuvant therapy (e.g., radiotherapy) after surgery. Of note is the fact that, once necrosis is added to the ST-G model, the RFS curve of SFT does not appear to be significantly different from that of malignant SFT without necrosis, which, however, remains distinctly separate and inferior. Concern could be raised about applying the term malignant SFT as per the 2013 WHO classification for soft-tissue tumors to these tumors. Larger studies and longer follow-up are needed to answer this question.
In 2014, Barthelmess et al. examined a series of nonmeningeal SFT/HPC and showed that tumors harboring NAB2 exon 4-STAT6 exon 2/3 fusions had lower recurrence rates than those with NAB2 exon 6-STAT6 exon 16/17 or other fusions [3]. Similarly, our subsequent work with meningeal-based lesions suggested a trend toward more aggressive behavior in tumors lacking NAB2 exon 4-STAT6 exon 3. Even though additional studies of dural-and nondural-based solitary fibrous tumors have shown consistent association between mitotic rate and fusion variant, with tumors harboring NAB2 exon 4-STAT6 exon 2/3 fusions consistently exhibiting lower mitotic rates than those with NAB2 exon 6-STAT6 exon 16/17, no further relationship between fusion variant and disease-free survival has been reported [1,3,6,12,18,21]. Our current study that focused specifically on a large cohort of meningeal-based tumors from multiple large tertiary care centers provides further support that fusion status fails to associate with RFS or overall survival.
Interestingly, several studies, including our earlier work, have suggested that fusion type appears to be associated with phenotype. The NAB2 exon 4-STAT6 exon 2-3 variant seems to be most often found in tumors resembling the conventional 'solitary fibrous tumor,' while tumors with NAB2 exon 6-STAT6 exon 16-17 exhibit morphologic features consistent with what was previously considered 'hemangiopericytoma' [1,3,11]. Our current findings are concordant with these prior reports and suggest that tumor phenotype is related to fusion variant.
The sensitivity of STAT6 for the diagnosis of solitary fibrous tumors has been reported to range from 86 to 98% in large series at both CNS and soft-tissue sites [7,9,13,17]. We identified only a single case of SFT/HPC which harbored an NAB2 exon 6-STAT6 exon 16 fusion but lacked nuclear STAT6 immunoreactivity. The tissue blocks from this patient were from 1997 and 1998, and it is possible that the age of the blocks precluded accurate immunohistochemical analysis. We also considered the possibility that the epitope recognized by our antibody may be lost in this specific fusion. However, the fusion in this case is the most frequently recognized, and, consequently, we would not expect loss of STAT6 staining based on fusion type. Regardless, the absence of STAT6 immunostaining in a tumor with morphologic features consistent with SFT/HPC should lead to the consideration of molecular testing to assess for the presence of NAB2-STAT6 fusion.
In conclusion, meningeal SFT/HPC is an aggressive tumor with propensity for high rates of recurrence and metastasis, which sometimes occur decades after the initial diagnosis. A grading scheme incorporating mitotic rate and necrosis seems to stratify this family of tumors most accurately. Although fusion type does not appear to relate to outcome in dural-based lesions, it does seem to be associated with tumor phenotype.