Introduction

Myxoinflammatory fibroblastic sarcoma (MIFS) is a relatively recently described entity that was first reported in 1998 by three separate groups of investigators under different designations, including inflammatory myxohyaline tumor of the distal extremities with virocyte or Reed–Sternberg-like cells [1], acral MIFS [2], and inflammatory myxoid tumor of soft parts with bizarre giant cells [3]. It was soon appreciated that all three studies were referring to the same condition, and subsequent studies have demonstrated that this tumor can also occur at sites other than the distal extremities; [4] the term “MIFS” has thus currently been adopted for designating these tumors [5].

MIFS is regarded as a distinctive clinicopathologic entity characterized by a superficial mass of the soft tissues located above the fascia and usually involving the dermis and subcutis of the extremities in young to middle-aged adults. Histologically, these tumors display a combination of features, including a low-grade-appearing spindle or epithelioid cell proliferation admixed with areas containing abundant myxoid stroma and prominent inflammatory elements. The most distinctive histologic feature, however, is the presence of scattered large, atypical cells that can exhibit marked nuclear pleomorphism and contain one or more prominent, inclusion-like nucleoli, sometimes resembling Reed–Sternberg cells or lipoblasts [4]. These features are not always present in all cases or may be present only focally and in varying proportions. Due to histologic overlap with many other lesions the recognition of this tumor may pose a problem for diagnosis [1]. Given the diagnostic challenges posed by these tumors, identification of immunohistochemical or molecular markers that may permit a more definitive diagnosis would be of potential benefit.

The studies reported thus far in the literature have not identified any useful or distinctive immunohistochemical markers that can help separate these lesions from other similar soft tissue tumors [1, 2, 5,6,7,8,9]. Several genetic alterations have been identified in MIFS, including an unbalanced t(1;10)(p22;q24) translocation that leads to juxtaposition of the TGFBR3 and OGA (formerly known as MGEA5) genes resulting in upregulation of NPM3 and FGF8, the formation of supernumerary ring chromosomes with an amplified region in chromosome 3 leading to overexpression of VGLL3, BRAF translocations involving ROBO1 and TOM1L2, and BRAF amplification [10,11,12,13,14,15,16,17,18], all of which have been proposed as potential molecular markers for these tumors. To complicate the picture, two other morphologically distinctive conditions have also been found to harbor t(1;10) translocations; hemosiderotic fibrolipomatous tumor (HFLT) and pleomorphic hyalinizing angiectatic tumor (PHAT). In addition, hybrid lesions showing combined features of MIFS/HFLT and PHAT/HFLT have been reported [12, 14,15,16,17]. The latter have sparked an ongoing debate on whether these three tumors are distinct entities or whether they represent different manifestations in the spectrum of a single neoplastic entity.

The present study focuses exclusively on conventional MIFS, without any features of HFLT or PHAT. An expanded panel of immunohistochemical markers was utilized in an effort to identify potential antibodies that may be of aid for diagnosis and help further define the utility of certain antibodies previously reported as positive in these tumors. In addition, molecular studies were used to evaluate which of the genetic alterations described so far in these tumors are most supportive of a diagnosis of pure MIFS and to clarify the incidence of the t(1;10) translocation. The results of our study and review of the literature are presented herein.

Materials and methods

The cases in this study were retrieved from the surgical pathology files of the pathology departments at the Medical College of Wisconsin, Milwaukee, WI, the Cleveland Clinic, Cleveland, OH, and Biopticka Laboratory, Plsen, the Czech Republic, or from the personal consultation files of the authors (MM, BR, and SS). The study was carried out under Institutional Review Board approval. All cases underwent central review at the institution where the testing for the study was performed (Medical College of Wisconsin), and underwent consensus review by the respective teams of pathologists who contributed cases for the study from the other institutions. Cases of conventional MIFS were accepted for inclusion in this study based on previously published criteria by the WHO [4]. Only cases showing the histologic features of MIFS were included in the study; no cases with hybrid features or with areas resembling HFLT or PHAT were included. Briefly, the criteria used for MIFS included a superficial tumor above the fascia with mixed solid and myxoid areas, low mitotic activity, inflammatory infiltrates, and the presence of typical virocyte-like (Reed–Sternberg like) or lipoblast-like cells [4, 19]. All cases corresponded to surgical resection specimens. From 2–12 sections stained with hematoxylin & eosin were studied for each case. Clinical and demographic information was abstracted from the surgical pathology report, patient’s medical record or by contacting the referring physician, and included age, gender, location and size of the lesion, and the clinical presentation. For molecular studies, DNA and RNA were extracted from deparaffinized tissue sections cut from representative formalin fixed and paraffin embedded tissue blocks.

Tissue microarray

H&E stained slides were reviewed for each case and used to mark suitable, tumor-rich areas for tissue microarray (TMA) construction. 1 mm tissue cores containing tumor were taken from paraffin-embedded MIFS samples and arrayed using the Veridiam Tissue Arrayer (Oceanside, CA). Three cores per case were transferred onto the TMA. After the cores were placed, TMA blocks were heated in an oven at 37 °C for 1 h and cooled at room temperature. Sections were cut and used for subsequent experimental analysis.

Immunohistochemistry

For immunohistochemical studies, representative paraffin blocks were available from all cases. Immunohistochemical studies were performed in 68 of 73 cases using TMAs. The remaining five cases were studied using standard unstained histologic sections cut from representative paraffin blocks or from unstained sections provided by the consulting pathologists. A broad panel of antibodies including bcl-1, CD10, FXIIIa, D2–40, CD34, CD68, MUC4, INI-1, S100, SOX10, MitF, PGP9.5, Glut-1, calretinin, beta-catenin, e-cadherin, calponin, bcl-2, SMA, desmin, ALK1, cytokeratin AE1/AE3, p63, androgen receptor, FLI-1, WT1, TLE-1, CD15, CD21, CD30, CD31, CD63, CD99, CD117, HBME1, MDM2, P53, PAX5, PAX8, and MIB1 were performed in all cases. See the supplemental material and methods and supplemental Table 1 for a detailed description of immunohistochemistry and antibody specifications.

Fluorescence in situ hybridization

All FISH probes were purchased from Empire Genomics, Buffalo, New York. RP11–55P24 (telomeric, red) and RP11–958B14 (centromeric, green) were selected to target the OGA locus on chromosome 10. RP11–1011D5 (red) was selected to target the telomeric side of the TGFBR3 locus on chromosome 1 and together with RP11–958B14 was used as fusion probes for the t(1;10) translocation; BRAF break-apart probes were purchased as a ready to use kit (catalog number BRAFBA-20-ORGR, CHR7: 140433812–140624564). See supplemental material and methods for detailed description of fluorescence in situ hybridization.

DNA microarray

A custom 8 × 60k Agilent Array was designed using a low-density whole genome backbone with 10,000 tiled, high-density probes (1.285 kbp median probe spacing) targeting 14.062 Mbp of MIFS-specific loci on chromosome 1:88816585–95681151 (91.65% coverage with 4531 probes), chromosome 3:85021127–89006253 (92.28% coverage with 2631 probes), and chromosome 10:101422418–105722084 (95.2% coverage with 2838 probes), (coordinates based on 26247H. sapiens, hg19, GRCh37). See the supplemental material and methods for a detailed description of DNA microarray.

RNA sequencing

Fifteen cases were analyzed for the presence of genomic fusion events using the Illumina TruSight RNA Fusion Panel (Illumina, San Diego, CA) following the manufacturer’s instructions. This RNA panel covers 507 fusion‐associated genes by targeting 7690 exonic regions with 21,283 probes. See the supplemental material and methods for a detailed description of the next generation sequencing performed.

Results

Clinical findings

The main clinical features of our patients are detailed in Table 1. There were 73 patients, including 38 women and 35 men aged 20–91 years (mean = 49.8 years). The lesions measured from 1.9–7.5 cm in greatest dimension (mean = 4.2 cm) and 33 (45%) were located in the upper extremities, including arm, forearm, elbow, wrist, hand, and fingers; 26 (35.6%) were located in the lower extremities, including thigh, knee, leg, ankle, foot, and toe; and 14 (19.2%) were located outside of the extremities in the neck, shoulder, axilla, back, buttock, groin, and cheek. The tumors were described as exhibiting slow growth, ranging from several months to a few years. In three patients, pain was a presenting symptom. All tumors were superficial and centered above the fascia, involving the subcutaneous tissue and the dermis, and often infiltrating the superficial dermis. On cut section they were described as nodular with a solid and myxoid cut surface. Three cases showed ulceration of the overlying epidermis. All patients were treated by complete surgical excision. None of the cases in this study represented recurrences but corresponded to primary excisions. Information regarding the status of the margins and patient outcome was incomplete in a large percentage of cases and was not included in this study.

Table 1 Clinical, histological, and molecular features of 73 Cases of myxoinflammatory fibroblastic sarcoma.

Pathologic findings

All the tumors showed vague lobulation on scanning magnification, with pushing borders rather than jagged, infiltrative edges and they all involved the dermis (Fig. 1a). In the majority of cases dermal involvement was extensive and reached the papillary dermis; in three cases the tumors ulcerated the overlying epidermis. The overall low-power examination of the lesions showed a variegated appearance, with areas that were hypercellular alternating with hypocellular areas (Fig. 1b). Infiltration of the surrounding adipose tissue was a common finding, and in five cases the tumor invaded fascia and infiltrated underlying skeletal muscle. Three histologic features were present in all tumors in varying proportions: 1) solid areas embedded in a variably collagenous stroma that were composed of sheets of round epithelioid cells or fascicles of spindle cells; 2) areas containing abundant myxoid stroma, and 3) areas displaying dense inflammatory cell infiltrates. Solid areas predominated in 33 cases; myxoid areas were predominantly seen in 29 cases; 11 cases showed approximately equal admixtures of both components (Table 1). Solid areas were mostly composed of sheets or nodular aggregates of uniform round to polygonal cells with round to oval nuclei surrounded by an ample rim of densely eosinophilic cytoplasm (Fig. 1c). In some cases, the round cells appeared to display a clear halo surrounding the nucleus. In most cases, the round cells displayed sharp cell borders with vague molding of the cell membranes. The nuclei in these areas were round to oval with fine, vesicular chromatin and small nucleoli, and with occasional irregularities and infoldings of the nuclear membranes. The solid areas in most cases imperceptibly blended with short fascicles of relatively bland-appearing fibroblast-like spindle cells (Fig. 1d). Areas containing spindle cells tended to show a greater degree of stromal sclerosis than areas containing round epithelioid cells. Foci in which round cells appeared discohesive due to accumulation of intercellular myxoid material were also noted. Solid nodules merged with larger areas displaying a prominent myxoid matrix. Myxoid areas in general were paucicellular and contained scattered small ovoid to spindled cells admixed with rare cells with inclusion-like nucleoli and inflammatory elements floating in abundant mucinous matrix. In predominantly myxoid areas, inflammatory elements tended to concentrate around small vessels. Inflammatory elements were unevenly distributed and varied from area to area and comprised a wide variety of inflammatory cells, including small lymphocytes, plasma cells, neutrophils, eosinophils, mast cells, and histiocytes. Well-formed lymphoid follicles with reactive germinal centers were seen in 23 cases.

Fig. 1: Histological Features of MIFS.
figure 1

a Scanning magnification of MIFS involving the entire dermis and subcutis. The tumor shows a pseudonodular architecture with areas that are solid alternating with lighter, less cellular areas that extend to the papillary dermis; b Higher magnification showing solid spindle cell areas alternating with myxoid foci; c Higher magnification of solid component showing large round cells with abundant eosinophilic cytoplasm and sharp cell membranes, admixed with scattered larger cells with vesicular nuclei and prominent eosinophilic nucleoli; notice the scattered inflammatory cells; d Focus showing transitions between round, epithelioid cells (center and left), and fascicles of spindle cells (right).

A common finding seen in nearly all cases was emperipolesis, which varied from just a few inflammatory cells lying within the cytoplasm of tumor cells (Fig. 2a) to cells that were massively distended by large numbers of engulfed inflammatory elements (Fig. 2b). In many cases, emperipolesis was difficult to discern because distended cell membranes were no longer visible but could be inferred due to localized clumping of inflammatory cells intimately admixed with scattered larger atypical nuclei. The most striking and consistent histologic feature seen in all tumors was a subpopulation of cells displaying marked nuclear pleomorphism with enlarged nuclei and inclusion-like eosinophilic nucleoli, which were distributed in varying proportions throughout the solid and myxoid components of the lesions. These cells varied in size from small and round to large, binucleated, multinucleated, or multilobated. Some of these cells resembled ganglion cells, while others resembled Reed–Sternberg cells (Fig. 3a). A distinctive appearance observed focally was one in which the tumor cells adopted an unusual configuration resembling “peas in a pod”, with a string of small vesicular nuclei containing large inclusion-like nucleoli lined up in a row (Fig. 3b). In addition to the cells with large inclusion-like nucleoli, another distinctive type of abnormal cell was seen, mostly within the myxoid areas that was characterized by enlarged, hyperchromatic nuclei with abundant multivacuolated cytoplasm resembling lipoblasts (Fig. 3c). Atypical cells with marked nuclear pleomorphism but lacking inclusion-like nucleoli or cytoplasmic vacuolization were also present. Mitotic activity was low, averaging 1–2 per 10 high power fields. Necrosis was rare, except in areas of ulceration, or at previous biopsy sites.

Fig. 2: Emperipolesis in MIFS.
figure 2

a Typical focus of emperipolesis showing large cells with a few inflammatory cells floating within the cytoplasm; b Focus displaying massive emperipolesis within myxoid area in MIFS containing dozens of small lymphocytes, plasma cells, and occasional eosinophils.

Fig. 3: Atypical Nuclear Features in MIFS.
figure 3

a Large binucleated atypical cells in MIFS showing two enlarged nuclei with prominent eosinophilic nucleoli resembling Reed–Sternberg cell; b Elongated atypical cell (center) containing three small fused nuclei with prominent eosinophilic nucleoli aligned in a row resembling a “pea in the pod”; c Pseudolipoblasts (center) in MIFS showing atypical nuclei surrounded by abundant multivacuolated cytoplasm.

Immunohistochemical findings

Pertinent positive immunohistochemistry results are summarized in Table 2 and Supplemental Table 2. The most striking finding in our study was strong nuclear (and sometimes cytoplasmic) positivity for bcl-1. The staining pattern in 69/73 cases (94.5%) showed diffuse and widespread positivity in the nuclei of nearly all tumor cells (Fig. 4a). The antibody marked both the small round and spindle cells in solid areas and the larger cells with inclusion-like nucleoli. This antibody also labeled the large, atypical pseudolipoblastic cells (Fig. 4b). In a few cases, the stain highlighted the emperipolesis by singling out the negatively stained engulfed inflammatory cells contained within their cytoplasm. In 65/73 (89%) cases, more than 50% of the tumor cells showed cytoplasmic positivity for FXIIIa. Staining for FXIIIa was observed in both small round to spindle cells as well as in the larger atypical cells with inclusion-like nucleoli and in pseudolipoblastic cells (Fig. 4c). CD10 antibodies also showed striking cytoplasmic positivity in the majority (>80%) of tumor cells in 59/73 cases (80%). CD10 highlighted the cytoplasm and cell membranes in cells in the solid areas as well as in the scattered large atypical cells with inclusion-like nucleoli. In 41/73 cases (56%), tumor cells also displayed cytoplasmic positivity in more than half of tumor cells for D2–40. Stains for CD34 were positive in tumor cells in 13/80 cases (16%). CD68 showed a few scattered positive cells; the pattern of staining was mostly cytoplasmic within scattered histiocytes but involved fewer than 20% of the tumor cells. A MIB-1 proliferation marker showed scattered nuclear positivity in the more cellular areas, averaging 1–2% of the tumor cells. MUC4 was negative, but in a few cases, it showed a faint granular cytoplasmic positivity which was interpreted as non-specific background staining. A stain for INI-1 showed retained nuclear positivity in tumor cells in all cases. Immunohistochemical stains for S100, SOX10, MitF, PGP9.5, Glut-1, calretinin, beta-catenin, e-cadherin, calponin, bcl-2, SMA, desmin, ALK-1, cytokeratin AE1/AE3, p63, androgen receptor, FLI-1, WT1, TLE-1, CD15, CD21, CD30, CD31, CD63, CD99, CD117, HBME1, MDM2, P53, PAX5, and PAX8 were either completely negative in all cases examined or showed only non-specific or patchy staining in a low percentage of cases and were regarded as non-contributory.

Table 2 Pertinent immunohistochemical findings in 73 Cases of myxoinflammatory fibroblastic sarcoma.
Fig. 4: Immunohistochemical Features of MIFS.
figure 4

a Strong nuclear positivity in MIFS in the majority of tumor cells for bcl-1; b Nuclear positivity for bcl-1 in pseudolipoblasts; c Strong cytoplasmic positivity in MIFS for FXIIIa, including in pseudolipoblasts (center).

FISH for OGA, t(1;10), and BRAF

A summary of the molecular findings is presented in Table 1. The unbalanced t(1;10) translocation observed in MIFS results in loss of the telomeric part of chromosome 10 and can be identified by FISH with OGA probes [12] (Fig. 5a) (See supplementary material and methods for verification of FISH for detection of t(1;10). Two of 53 cases (3.7%) analyzed by FISH demonstrated unequivocal loss of one telomeric OGA probe signal (labeled red), consistent with the t(1;10) translocation (Fig. 5b); five additional cases were equivocal for the t(1;10) translocation with loss of the telomeric OGA signal observed in 10–13% of nuclei; intact telomeric ends of chromosome 10 were observed in the remaining 46 cases. Fusion FISH probes confirmed the t(1;10) in both unequivocal cases (Fig. 5c).

Fig. 5: t(1;10) FISH Analysis in MIFS.
figure 5

a Schematic design of FISH assays used to identify t(1;10). The chromosome numbers, genes, FISH probes, and regions of the chromosome that are deleted in the unbalanced t(1;10) are labeled. The bottom chromosome refers to t(1;10) translocation. b Unbalanced translocation involving OGA on chromosome 10. t(1;10) results in the loss of the telomeric portion of chromosome 10 and is demonstrated by a single centromeric (green) FISH probe (open arrow). Inset is a normal, control nuclei showing intact OGA with a pair of adjacent (fused) red and green FISH probe signals. c The fused red and green FISH probe signal (arrow) demonstrates t(1;10). Inset shows two separate red and green signals indicating intact chromosomes 1 and 10.

We successfully interrogated 70 cases for BRAF gene fusions using BRAF break-apart FISH probes. Four of 70 cases (5.7%) showed an unbalanced rearrangement of BRAF with split probe patterns including loss of either the 5′ telomeric green probe (Cases 17, 20, and 47) or the 3′BRAF centromeric red probe signal (Case 42) together with amplification of BRAF (Fig. 6a-d).

Fig. 6: BRAF Alterations in a Subset of MIFS.
figure 6

a Case 17 with BRAF gene rearrangement demonstrated by individual red (arrow) FISH probe signals and two fused FISH probe signals (1R2F pattern). b Case 20 with BRAF gene rearrangement demonstrated by individual red (arrow) FISH probe signals and one fused FISH probe signal (1R1F pattern). c Case 47 with amplification of the 3′ portion of BRAF (red FISH probe signal, arrow), which contains the kinase domain and multiple fused FISH probe signals (>1R > 1F pattern). In all three cases, isolated green FISH probe signal is not observed indicating loss of 5′ BRAF. d Case 42 shows multiple fused FISH probe signals (open arrows) and amplification of the 5′ end of BRAF (green probe signals, arrow) indicating increased copy number of BRAF (>1G > 1F pattern). LOWER DIAGRAM. RNA sequencing of case 17 shown in (a) identified a novel ZNF335-BRAF in-frame fusion transcript joining exons 6 of ZNF335 and exon 10 of BRAF. The top box shows a portion of the sequence of the in-frame fusion transcript that spans the break point and is indicated by the change in font color. The genes and corresponding reference transcripts are indicated. Below the red line, the fusion transcript is predicted to code for chimeric protein with exon 6 of ZNF 335 fused to exon 10 of BRAF and containing the kinase domain of BRAF. The amino terminal portion of BRAF, which contains the autoinhibitory domain encoded by exons 1–8, is predicted to be lost in the chimeric ZNF335/BRAF protein.

aCGH microarray

A summary of aCGH findings are presented in Table 1. We performed aCGH analysis on 20 MIFS cases using a DNA microarray specifically designed to interrogate alterations in TGFBR3, OGA, and VGLL3, located near chromosomal regions 1p22, 10q24, and 3p12, respectively. Cases 37, 51, and 68 (3 of 20 cases; 15%) demonstrated concurrent loss of genomic material at chromosome regions 1p22 and 10q24 (TGFRB3 and OGA) consistent with an unbalanced t(1;10). In two of these cases (Cases 51 and 68), t(1;10) was also identified by FISH, whereas FISH probe signals were uninterpretable in case 37. Gain of chr3p12 (VGLL3) was observed in 8 of 20 cases (40%), including the 3 cases with loss of TGFRB3 and OGA, making this the most common genetic aberration in our cohort. The reported association of VGLL3 amplification and BRAF gene rearrangements [15] were present in our cohort as two cases with split BRAF FISH probe signals (cases 17 and 20) also had VGLL3 amplification. In addition, aCGH provided an unbiased interrogation of the entire genome, allowing identification of other recurrent genetic aberrations present in MIFS. Two cases showed gain of nearly the entire long arm of chromosome 1 consistent with increased copy number, and five cases demonstrated large losses of chromosome 13q. No other recurrent genetic changes were observed.

RNA sequencing

Given the previously reported association of VGLL3 amplification and BRAF gene rearrangements [15], we prepared RNA sequencing libraries for 15 MIFS cases including all 8 cases with VGLL3 amplification. Due to low quality RNA, we successfully sequenced only two MIFS cases, both of which had VGLL3 amplification by DNA array (cases 17 and 37). While case 37 was negative, a novel in-frame fusion transcript involving exon 6 of ZNF335 (20q13, NM_022095.3) and exon 10 of BRAF (7q34, NM_001354609.1) was detected in multiple independent reads of the library from case 17 (Fig. 6). The predicted chimeric protein fuses the N-terminal 317 amino acids of the zinc-finger containing protein ZNF335 to amino acids 393–767 of BRAF, which contain the BRAF serine/threonine kinase domain. No fusion events were observed in the other MIFS case or in the control cases that were sequenced. Although ZNF335 contains multiple C2H2-type zinc fingers, which are believed to mediate DNA binding, the translocation breakpoint is upstream of their coding region, and none of the C2H2 motifs are expected to be present in the predicted chimeric protein. Of note, since t(1;10) does not result in a functional transcript [18], RNA sequencing will not be a useful method for screening for this molecular aberration. We were not able to appreciate a well-defined correlation between the morphology and the genetic findings in our cases. Two of the t(1;10) positive cases showed a predominantly solid histology (cases 37 and 68) and one was predominantly myxoid (case 51) (see supplementary figs. 13).

Discussion

MIFS is a relatively recently described clinicopathologic entity that is still not widely recognized by general surgical pathologists due to its variegated morphologic appearance and absence of distinctive immunohistochemical features. The exact criteria for their histopathologic diagnosis remain elusive and somewhat controversial, often requiring expert consultation for definitive diagnosis. It is generally accepted that the main features for diagnosis, in the proper clinical context, include solid pseudonodular areas composed of round, epithelioid cells often admixed with oval to spindle cells in a collagenous matrix, areas displaying prominent myxoid stroma, a variable inflammatory cell component, and randomly distributed larger, atypical cells that contain prominent inclusion-like nucleoli resembling Reed–Sternberg cells and their variants, or with prominent cytoplasmic vacuolation resembling lipoblasts (pseudolipoblastic cells) [1,2,3,4]. An added complicating factor has been the recognition that these tumors may be closely related to two other rare soft tissue tumors with similar clinical presentations; HFLT and PHAT. Some authors have postulated that these three entities may be histogenetically closely related, based on morphologic similarities and shared molecular alterations [11,12,13, 16, 17, 19, 20], while others believe that HFLT and PHAT represent a different overall process with the potential to develop into a higher grade sarcoma. [21, 22].

Herein, we have studied a cohort of 73 patients with classical histologic features of pure (non-hybrid) MIFS to identify immunohistochemical and/or molecular markers that may be of aid for the diagnosis of MIFS. The clinical and demographic features of MIFS have been adequately described in previous studies, and the results found in the present study are similar to those previously reported in the literature [1,2,3, 5, 6, 8]. The tumors occurred over a broad age range (20–91 years; mean = 49.8 years) and affected predominantly the extremities, with the most common location being the upper extremities followed by the lower extremities. As has been demonstrated in the recent literature, these tumors are not restricted to acral sites in the extremities and have now been described in more proximal locations, including the head and neck region, scalp, trunk, and buttocks [8, 23,24,25,26,27]. 17 of our cases presented outside of the extremities. The clinical behavior of these tumors has been well documented in a large study of 104 patients by Laskin et al. [8], who showed local recurrences in 51% of their cases and only one patient with distant metastasis. Other studies have also reported the rate of local recurrence for these tumors to be between 22 and 67% [1, 2, 28, 29].

Immunohistochemistry has been amply studied in MIFS without demonstrating a distinctive immunophenotype. Variable reactivity has been observed in previous studies for a variety of markers, including CD34, D2–40, CD68, SMA, EGFR, and CD163 [1,2,3, 5,6,7,8]. In the present study we found consistent and strong immunoreactivity in more than 50% of our cases for several antibodies, including bcl-1, FXIIIa, CD10, and D2–40. In our cohort of 73 cases, 69/73 (94.5%) cases showed positivity for bcl-1 in the majority of the tumor cells. Bcl-1 is a cell-cycle regulatory protein located on the long arm of chromosome 11 that is encoded by the human CCND1 gene and is essential for G1-S transition [30, 31]. In addition to overexpression in a variety of epithelial malignancies, mantle zone lymphoma, and neuroblastoma, a few types of sarcoma have been shown to overexpress bcl-1, including leiomyosarcoma, endometrial stromal sarcoma, epithelioid sarcoma, and synovial sarcoma [32,33,34,35]. However, bcl-1 expression has not been demonstrated in other types of sarcomas arising in superficial soft tissues that may enter the differential diagnosis of MIFS, thus making it a potentially useful marker in this context. Another consistent finding in our study was positivity for FXIIIa in 65/73 (89%) cases. One previous study also reported a similar finding in a limited number of cases (4/5 cases) [20], but this feature has not been extensively investigated in other studies. FXIIIa is expressed in dermal dendrocytes and is a useful marker for the classification of dermatofibromas [36, 37]. To the best of our knowledge, this marker has not been identified in superficial soft tissue sarcomas that may enter in the differential diagnosis of MIFS, and may therefore represent a potentially useful marker to support the diagnosis in equivocal cases in superficial soft tissue tumors involving the dermis. Other markers that showed significant expression in our cases were CD10 and D2–40. CD10 has been studied in a variety of soft tissue sarcomas and is positive in a high percentage of various types of tumors, calling into question its specificity [38]. However, it has been reported as a useful adjunct in the diagnosis of atypical fibroxanthoma (AFX) [39], a primary dermal neoplasm characterized by striking nuclear pleomorphism that has the potential for being confused with MIFS on small or limited biopsy samples, particularly in cases of MIFS with extensive involvement of the superficial dermis. The differential diagnosis in such cases can be facilitated by strong immunoreactivity for bcl-1 and FXIIIa in MIFS, which would not be expected to occur in AFX. Finally, a significant number of cases in our study showed positivity in over 50% of the tumor cells for D2–40. This marker was found to be expressed in 86% of cases in the study by Laskin et al. [8]. D2–40, also known as podoplanin, is a small membrane glycoprotein found primarily on the surface of lymphatic vessels, type I pneumocytes and glomerular podocytes, but which has also been identified in many types of normal cells originating from various organs [40]. Because of its broad distribution in many different tumor types, we feel this represents a less reliable marker. Although it is acknowledged that the specificity of bcl-1 and FXIIIa is generally low in soft tissue sarcomas, in the appropriate clinicopathologic context, the high sensitivity of these markers in MIFS may be of aid in difficult or equivocal cases.

Recent studies have identified several genetic alterations in MIFS. Lambert et al. [10] reported a unique t(1;10) translocation in one case that showed a complex karyotype containing a reciprocal translocation t(1;10)(p22;q24) in addition to loss of chromosomes 3 and 13. Subsequent molecular studies by Hallor et al. [11] demonstrated t(1;10) in 4/5 MIFS cases using cytogenetic analysis. FISH analysis of 3/4 cases showed a rearrangement involving 1p22 with 10q24 leading to juxtaposition of TGFB3 and OGA in two cases; the third case also showed t(1;10), however only OGA (10q24) was involved. Gene expression analysis of three t(1;10)-positive MIFS cases showed upregulation of NPM3 and FGF8; two genes located close to OGA in the 5 Mb region proximal to the breakpoint of chromosome 10, suggesting that they are targets for transcriptional upregulation. Amplification of 3p11–12 was initially identified using cytogenetics (as a supernumerary ring chromosome 3) and/or aCGH in two cases of MIFS [11]. Both global and targeted gene expression analysis, using microarray and qPCR respectively, demonstrated increased expression of VGLL3 and CHMP2B in the 2 MIFS cases with amplification of 3p11–12. Baumhoer et al. also reported the presence of 3p11–12 amplification via aCGH in 1/2 cases [41]. A study by Antonescu et al. [12] also showed rearrangements of TGFBR3 and OGA by FISH in 5/7 cases of MIFS, as well as 12/14 HFLT and 3/3 mixed MIFS/HFLT. Carter et al. [13] subsequently studied a large cohort of patients by FISH and found that the TGFBR3/OGA rearrangement was more frequent in PHAT (6/10 cases) and in mixed MIFS/HFLT (3/3) than in MIFS (0/6). The largest genetic series of MIFS prior to this study was by Zreik et al. [14] who studied 31 MIFS cases using break-apart FISH probes targeting TGFRB3 or OGA. In their study, 2/31 (6.4%) MIFS cases showed rearrangement of OGA only; rearrangements involving TGFRB3 were not observed in any of their MIFS cases. Hallor et al. [11] also described a t(1;10) MIFS case involving only OGA with upregulation of NPM1 and FGF8. Therefore, it is possible that the two cases described by Zreik et al. [14] harbor a t(1;10) involving a loci other than TGFRB3, leading to upregulation of these target genes. Furthermore, they identified rearrangements involving TGFRB3 and/or OGA in 6/8 hybrid MIFS/HFLT (75%), 2/2 PHAT and 1/1 pure HFLT cases. They concluded that aberrations in TGFBR3/OGA are much more common in hybrid HFLT/MIFS than in classical MIFS. A more recent study by Arbajian et al. [18] utilized whole genome sequencing and aCGH analysis in seven cases of MIFS to better map chromosomal breakpoints and concluded that t(1;10) translocation, despite being recurrent in MIFS, does not appear to transcribe a functional fusion gene nor lead to consistent genomic rearrangements. This study also identified a novel ROBO1-BRAF gene fusion in a t(1;10) negative case. Similarly, Kao et al. [15] reported alterations of BRAF in 6/27 (22%) cases of pure t(1;10)-negative MIFS using FISH, and further characterized a TOM1L2-BRAF fusion in one of their cases.

The current study represents the largest cohort evaluated thus far for genetic abnormalities in cases of conventional MIFS. We were able to demonstrate t(1;10) in only 3/54 (5.5%) cases using a combination of FISH and aCGH. aCGH demonstrated an unusually high percentage of amplification for VGLL3 on chromosome 3 in 8/20 (40%) of the cases studied, consistent with the 52% reported in the literature (Table 3). The gain of chromosome 3p12 (which includes VGLL3) in four cases without the t(1;10) translocation also suggests that amplification of VGLL3 is a relatively common occurrence in MIFS and not specifically associated with t(1;10). Hélias-Rodzewicz et al. identified VGLL3 amplification in ~10% of 404 soft tissue sarcomas examined, including dedifferentiated liposarcoma, undifferentiated pleomorphic sarcoma, and myxofibrosarcoma [42]. While the sensitivity of VGLL3 as a diagnostic marker is low, it appears to occur at a significantly higher rate in MIFS than in sarcomas in general. Furthermore, both of our cases with BRAF gene rearrangements demonstrated amplification of chromosome 3p12 (region of VGLL3). A similar correlation was observed by Kao et al. [15] who reported VGLL3 amplification in 4 of 6 MIFS cases with BRAF fusion events. Alterations of BRAF were identified in 4/70 (5.7%) cases tested in our study; a rate lower than the 22% (6 of 27 cases) reported by Kao et al. [15]. It is unclear why we observed significantly fewer BRAF gene abnormalities. One possible explanation is that the FISH assay utilized in this study may have a lower analytical sensitivity; however, the FISH probe hybridization signal in all cases was clear and bright with low background. Furthermore, two fused probe signals were clearly observed in nearly 95% of our cohort consistent with intact BRAF loci. Study of further cases may help clarify some of these discrepancies.

Table 3 Literature review of t(1;10) (p22;q24) TGFRB3/OGA, VGLL3 amplification, and BRAF gene abnormalities in MIFS.

The novel ZNF335-BRAF fusion transcript identified in our study represents the third separate partner gene described in these tumors (the others being TOM1L2 and ROBO1) suggesting that BRAF likely shares multiple partner genes in MIFS. To date, all the reported cases of MIFS with BRAF alterations, including the cases described here, are negative for t(1;10) supporting the notion that BRAF abnormalities and t(1;10) are mutually exclusive molecular alterations in this tumor typeA summary of the literature (Table 3) shows that out of a total of 112 conventional MIFS cases analyzed to date, the t(1;10) translocation has been identified in only 15 (13.4%) cases, while BRAF alterations have been identified in only 11/104 (10.6%) cases tested.

In summary, we have presented the largest immunohistochemical and molecular analysis of MIFS to date in an effort to identify adjunct methods of value for supporting the diagnosis. Although no specific immunohistochemical or molecular markers can be applied to these tumors, our results support that judicious use of a panel of immunohistochemical markers that includes bcl-1 and FXIIIa may be potentially helpful in supporting the diagnosis of MIFS in equivocal cases. Although these markers are not specific and should not be used in isolation, they may be helpful in combination for the differential diagnosis in the appropriate clinical context. In addition to the classically described histologic features, and despite the inclusion of the term “fibroblastic” in their designation, the spectrum of morphologic features for these tumors continues to expand, including the recognition of a prominent epithelioid cell component and striking emperipolesis. Our study also underscores that recurrent molecular alterations, with the exception of VGLL3 amplification, are relatively uncommon in these tumors. Given the low rate of t(1;10) and BRAF gene rearrangements in our study, we believe neither of these is cost-effective or sensitive enough to serve as routine diagnostic markers for MIFS. While VGLL3 amplification has been identified in various soft tissue neoplasms, it appears to be present quite commonly in MIFS, occurring in approximately half of all MIFS cases studied to date. Thus, demonstration of VGLL3 amplification may be more helpful and sensitive than t(1;10) or BRAF alterations to support the diagnosis in the appropriate clinical and immunohistochemical context. For the time being, however, the diagnosis of MIFS remains primarily a clinicopathologic diagnosis that requires a combination of the appropriate clinical setting along with the characteristic histopathologic findings.