Spindle cell tumors with RET gene fusions exhibit a morphologic spectrum akin to tumors with NTRK gene fusions

Abstract

A major breakthrough in classification of soft tissue tumors has been the recent identification of NTRK-fusion related neoplasms which are amenable to highly effective targeted therapies. Despite these therapeutic oportunities, diagnostic challenges have emerged in recognizing tumors characterized by protein kinase fusions, as they are associated with a wide morphologic spectrum, variable risk of malignancy and a rather non-specific immunoprofile. As such, NTRK-related fusions may occur in infantile fibrosarcoma, lipofibromatosis-like neural tumors, tumors resembling malignant peripheral nerve sheath tumors, etc. Triggered by an index case resembling lipofibromatosis-like neural tumor but harboring RET gene rearrangement, we investigated our files for cases showing RET gene abnormalities in order to establish their clinicopathologic features. Tumors were tested with a combination of targeted RNA sequencing and FISH methods. Six cases with RET gene rearrrangements were identified, all except one occured in children, including 4 infants. Their morphologic spectrum was quite diverse, but closely reproduced the phenotype of NTRK-fusion positive tumors, including lipofibromatosis-like neural tumors (n=3), infantile fibrosarcoma-like tumor (n=2) and malignant peripheral nerve sheath tumor-like (n=1). Three cases showed co-expression of S100 and CD34, while the remaining 3 had a non-specific immunoprofile. The tumors ranged morphologically and clinically from benign to highly malignant. None of the lipofibromatosis-like neural tumor cases recurred, while 2 patients with a malignant histology had a highly aggressive course with distant metastases to lung and other viscera. By targeted RNA sequencing these tumors harbored RET fusions with an identical break in exon 12, which retains the tyrosine kinase domain in the fusion oncoprotein, and involving various gene partners (CLIP2, CCDC6, SPECC1L, MYH10 and NCOA4). Our results suggest that RET fusion-positive neoplasms share a similar phenotypic spectrum with the NTRK-positive tumors, displaying either fibroblastic or neural-like differentiation, and spanning a wide spectrum of clinical behavior. These findings open new avenues for targeted therapy with RET inhibitors currently available in clinical trials.

INTRODUCTION

The classification of pediatric soft tissue tumors remains challenging due to their rarity, overlapping histologic features and often non-specific immunoprofile. The wide application of next generation sequencing has revealed novel gene fusions at an unprecedented pace, underscoring a wide range of morphologies. Recurrent fusions involving genes encoding receptor tyrosine or cytoplasmic kinases have been described recently in distinct soft tissue tumors with spindle cell morphology and fibroblastic or neural-like differentiation^1–5. The detection of these gene fusions in clinical practice is critical from both diagnostic and therapeutic standpoints. First, these recurrent genetic alterations may be used as reliable molecular diagnostic markers in the classification of challenging spindle cell neoplasms with a non-specific immunoprofile. Second, several of these oncogenic kinases have been shown to be therapeutically targetable, which may possibly translate into better treatment strategies and improved patient outcomes^6–8. However, the emerging data on tumors with NTRK gene fusions has suggested capacity for a remarkably diverse histogenesis, being involved in neoplasms of various cell lineages, including epithelial, melanocytic and mesenchymal, and spanning to benign to highly malignant^9,10. In soft tissue tumors alone, NTRK1 gene fusions have been described in at least 3 different pathologic entities occuring in children and adolescents, such as lipofibromatosis-like neural tumor, infantile fibrosarcoma and tumors resembling low grade malignant peripheral nerve sheath tumors^1–3. In this study we report 6 cases spanning a very similar morphologic spectrum, but showing instead RET gene fusions with various gene partners.

METHODS

Patient Selection and Tumor Pathologic Evaluation.

Archival material and personal consult files of the senior authors (CRA, CDF) were searched for cases spanning the morphologic spectrum seen in other kinase fusion positive mesenchymal neoplasia, including: congenital/infantile fibrosarcoma (IFS), infantile fibrosarcoma-like^3,11, cellular congenital mesoblastic nephroma¹², lipofibromatosis-like neural tumors (LPF-NT)¹ and malignant peripheral nerve sheath tumor-like². A large collection of these various morphologies were available in our consultation files from prior molecular studies, which were previously screened and lacked abnormalities in NTRK1/2/3, BRAF, RAF1, and MET genes^1–3,11. Collectively, these tumors showed either a non-specific immunoprofile (vimentin only or vimentin and CD34 positivity) or a partial neural phenotype (co-expression of S100 and CD34, but lacking SOX10). Most cases however had a much larger immunostaining work-up, including desmin, myogenin, cytokeratin, epithelial membrane antigen (EMA), which were negative. H3K27me3 expression was retained in all cases tested. The study group was analyzed for demographic information, anatomic site, tumor size, and morphologic features, including cell type, degree of cellularity, type and amount of stromal component, nuclear features, mitotic activity and presence of necrosis. Available immunohistochemical stains were reviewed and additional work-up was performed in retrospect based on the molecular results. The clinical follow-up information was obtained from review of the electronic medical records and from contacting referring pathologists and clinicians. The study was approved by the Institutional IRB.

Fluorescence in situ hybridization (FISH)

Formalin-fixed paraffin-embedded tissues were available in each case for FISH analysis. FISH for RET was performed on 6 all cases. FISH was performed on 4 μm-thick formalin-fixed paraffin-embedded (FFPE) tissue sections. Custom probes were made by bacterial artificial chromosomes (BAC) clones flanking the RET gene, according to UCSC genome browser (http://genome.ucsc.edu) and obtained from BACPAC sources of Children’s Hospital of Oakland Research Institute (Oakland, CA; http://bacpac.chori.org). DNA from each BAC was isolated according to the manufacturer’s instructions. The BAC clones were labeled with fluorochromes by nick translation and validated on normal metaphase chromosomes. The slides were deparaffinized, pretreated, and hybridized with denatured probes. After overnight incubation, the slides were washed, stained with DAPI, mounted with an antifade solution, and then examined on a Zeiss fluorescence microscope (Zeiss Axioplan, Oberkochen, Germany) controlled by Isis 5 software (Metasystems, Newton, MA).

Targeted RNA sequencing (RNA-seq) and analysis.

Five cases were tested on one of the RNAseq platforms: 1 by whole transcriptome, 2 by Targeted Illumina TruSight RNA sequencing, and 2 by ARCHER, using previously published methods^2,3. The RNA extraction was performed from frozen tumor in one case tested by whole transcriptome RNAseq. RNA was extracted from FFPE tissue using Amsbio’s ExpressArt FFPE Clear RNA Ready kit (Amsbio LLC, Cambridge, MA) in 4 cases. Fragment length was assessed with an RNA 6000 chip on an Agilent Bioanalyzer (Agilent Technologies, Santa Clara, CA). RNA-seq libraries were prepared using 20–100 ng total RNA with the TruSight RNA Fusion Panel (Illumina, San Diego, CA). Each sample was subjected to targeted RNA sequencing on an Illumina MiSeq at 8 samples per flow cell (approximately 3 million reads per sample). All reads were independently aligned with STAR(ver 2.3) and BowTie2 against the human reference genome (hg19) for Manta-Fusion and TopHat-Fusion analysis, respectively. The RET mRNA expression level was evaluated in the case examined by whole transcriptome RNAseq and compared to those of other samples analyzed on the same targeted RNA sequencing platform.

RESULTS

Patients and Tumor samples.

From a total of 50 cases screened for various kinase fusions, half were negative for NTRK1/2/3, RAF1, and BRAF fusions. Six tumors (about 25%) of these cases showed RET gene abnormalities, occuring in 4 females and 2 males. All except one were children and 4 of them were infants (Table 1). The only adult patient was a 47 year-old male, who presented with a large buttock mass and showed histologic features in keeping with a high grade malignant periphereal nerve sheath tumor. The tumors presented at a wide range of anatomic sites, including 3 in the lower extremities (2 in the foot, 1 buttock), 2 in the trunk (1 chest wall, 1 abdominal wall) and one visceral (kidney). Three of the pediatric cases had histologic features in keeping with a lipofibromatosis-like neural tumor, and the remaining two had features resembling an infantile fibrosarcoma. A number of other entities, including 5 lipofibromatosis, 5 bonified H3K27me3-retained high grade malignant peripheral nerve sheath tumors and 10 other high grade spindle cell sarcomas, NOS, were tested for RET gene rearrangements but none were identified.

Table 1.

Clinical, Pathologic and Molecular Features of RET Fusion Positive Tumors

#	Age/Sex	Diagnosis	Location/Size/Clinical information	IHC (only positive results)	RET Fusion/Platform
1	1/F	LPF-NT	Ankle (1.4 cm, superficial), Incomplete resection AWD stable, 3 years FU	CD34 diffuse, S100 focal	CCDC6-RET (ARCHER)
2	0/F	LPF-NT	Foot (3 cm, deep) Present at birth (biopsy only) AWD stable, 2 years FU	S100 focal, CD34 focal	NCOA4-RET (Foundation Med)
3	13/F	LPF-NT	Abdominal wall (2.9 cm, superficial) marginal resection; no FU	CD34	RET gene rearrangement (FISH)
4	2mo/M	IFS-like	Chest wall (deep-seated, bone), marginal resection NED, 9.5 years	SMA patchy	CLIP2-RET (TruSight RNA-seq)
5	1/F	IFS/cellular mesoblastic nephroma	Kidney (10.5 cm) Bilateral lung & brain metastases Recent case		SPECC1L-RET (ARCHER)
6	47/M	HG MPNST-like	Buttock (11 cm, deep) Neoadjuvant chemotherapy, adjuvant radiation Lung metastases, DOD 24 mo	Focal S100, H3K27me3 retained	MYH10-RET (Whole transcriptome)

Mo, months; M, male; F, female; LPF-NT, lipofibromatosis-like neural tumor; IFS, infantile fibrosarcoma; HG MPNST, high grade malignant peripheral nerve sheath tumor; NED, no evidence of disease; AWD, alive with disease; DOD, dead of disease; FU, follow-up.

Lipofibromatosis-like neural tumors show recurrent RET gene fusions and follow a benign course.

Three cases, all presenting in children, including 2 infants, showed microscopic features of lipofibromatosis-like neural tumors, displaying a highly infiltrative growth pattern. Both infantile cases (one present at birth) occurred in the foot, one as a 1.4 superficial ankle lesion and the other as a 3.0 cm superfical and deep seated lesion within the plantar aspect of the foot, infiltrating the metatarsal bones. The first case was marginally resected with positive margins; the residual tumor remained stable and the patient is alive with disease 4 years later. The latter case was only biopsied, as it was deemed unresectable short of an amputation; the patient appears stable without any additional therapy 2 years later. The third case occurred in a 13 year-old female with a superficial 2.9 cm abdominal wall mass which was resected with positive margins; no follow-up was available. Microscopically, all 3 cases showed similar findings of an infiltrative, low to intermediate cellularity spindle cell proliferation arranged in streaming patterns and vague fascicles, forming delicate septa to confluent solid sheets within the adipocytic component (Fig 1). The spindle cells had monomorphic cytomorphology with scant amphophilic cytoplam and ovoid nuclei with fine chromatin. The tumors lacked nuclear pleomorphism, increased mitotic activity or necrosis. Two of the cases showed co-expression of S100 and CD34, albeit in one case the degree of staining was quite focal for both markers (Fig. 1). A Pan-NTRK stain performed in these 2 cases was negative. The abdominal mass was only positive for CD34, while S100, SOX10, desmin, CK, were negative and an H3K27me3 expression was retained.

Figure 1. RET fusions characterize a subset of infantile lipofibromatosis-like neural tumors.

A-D (case 1) Superficial ankle soft tissue lesion centered within the subcutis and removed with positive margins in a one-year-old female (A). Higher power shows a highly infiltrative lesion composed of confluent sheets of bland spindle cells arranged in a patternless or vague storiform growth with variable degree of stromal hyalinization (B). Tumor was focally positive for S100 protein (C) and diffusely strongly positive for CD34 (D). E-F (case 2) A deeply infiltrative spindle cell tumor within fat and skeletal muscle showing variable degrees of cellularity and fibrotic stroma (E); tumor was focally positive for both S100 (F) and CD34.

RET fusions may function as alternative genetic abnormalities in infantile fibrosarcoma cellular mesoblastic nephroma spectrum.

Two infantile spindle cell sarcomas with features within the spectrum of infantile fibrosarcoma revealed RET gene rearrangements. The first case involved the chest wall subcutis, skeletal muscle and bone of a 2 month old baby boy. The tumor was marginally resected with positive margins and the patient remains without evidence of disease 9.5 years after diagnosis. The tumor was composed of intersecting fascicles of uniform spindle cells with scant fibrillary eosinophilic cytoplasm, imparting at least focally a myofibroblastic appearance. Immunostaining showed only SMA reactivity, while desmin, myogenin, S100, and CD34 were negative.

The second case was that of 1 year-old girl with a 10 cm renal mass, who presented at diagnosis with disseminated disease to lungs and brain (recent case). Histologically the tumor showed diffuse cellularity, arranged in intersecting long fascicles or solid sheets (Fig. 2 A–C). The tumor was associated with a staghorn-type vasculature and showed geographic areas of necrosis. The lesional cells had uniform cytomorphology, having fusiform nuclei with fine chromatin and scant pale-eosinophilic cytoplasm. The tumor showed a high mitotic count of >20 MF/10 HPFs. The tumor was only positive for vimentin, being negative for S100, SOX10, CD34, NTRK-pan, and showed retained H3K27me3 expression.

Figure 2. Morphologic spectrum of tumors with RET gene rearrangements.

A-C (case 5). Infantile fibrosarcoma/cellular congenital mesoblastic nephroma morphology in a one-year-old female showing at low power large areas of necrosis and rich vascular network with a staghorn appearance (A); at higher power the cells were arranged either in solid sheets (B) or intersecting fascicular pattern (C); the lesional cells showing monomorphic cytomorphology with scant fibrillary eosinophilic cytoplasm and fusiform nuclei with fine chromatin. The tumor was associated with a high mitotic activity (B). D-E (case 6). High grade malignant peripheral nerve sheath tumor-like morphology in a 47 year-old male with a buttock lesion. The tumor was associated throughout with geographic areas of necrosis (D,E), variable cellularity and delicate collagenous stroma (B). At higher power the tumor showed relatively uniform cytology, with ovoid nuclei showing coarse chromatin, conspicuous nucleoli, and a high mitotic activity (C).

Rare spindle cell sarcomas resembling high grade malignant peripheral nerve sheath tumor in adults may harbor RET gene fusions.

This 47 year-old male presented with a large, deep-seated, 11 cm buttock mass, which was resected with close margins. The tumor was highly cellular and arranged in solid sheets and long intersecting fascicles, being associated with a staghorn-type vasculature and geographic areas of necrosis (Fig. 2 D–F). The lesional cells showed mainly a primitive monomorphic phenotype, with plump ovoid nuclei with coarse chromatin and scant pale-eosinophilic cytoplasm. Scattered cells with enlarged and pleomorphic nuclei were also seen. The tumor showed brisk mitotic activity of >20 MF/10 HPFs. The patient followed a highy aggressive clinical course, developing lung and retroperitoneal soft tissue metastases, within 6 months from diagnosis and succombed of disease shortly after.

RET fusions encompass a wide variety of gene partners and result in RET gene upregulation.

By targeted RNA sequencing these tumors harbored RET fusions with various gene partners (CLIP2, CCDC6, SPECC1L, MYH10 and NCOA4). (Fig. 3) All 5 cases tested by targeted or whole transcriptome RNAseq showed an identical break in exon 12, which included the tyrosine kinase domain within the projected fusion oncoprotein (Fig. 3). The tumor tested by whole-transcriptome fusion also showed upregulation of RET mRNA in contrast to other >100 soft tissue tumors available on the same platform (Fig. 3). FISH analysis confirmed the presence of RET gene rearrangements in all 5 cases with available material (Fig. 3).

Figure 3. Diagrammatic representation of various RET fusions investigated by RNA sequencing platforms.

(A) Chromosomal localization of RET and all 5 various gene partners; red vertical lines depict the genomic breakpoint locus. Arrows show the direction of transcription of each gene. (B) Each of the 5 different RET fusions are depicted in relationship to their exonic composition. In all cases the fusion transcript included the 3’ of RET gene, starting from exon 12, thus retaining the kinase domain in the putative fusion oncoprotein. The protein domains of each of the genes involved is also schematically depicted. (C) FISH showing RET break-apart signals in case 1 (red, centromeric; green telomeric). (D) RET mRNA expression is upregulated in case 6 studied by whole transcriptome sequencing compared to other sarcoma types available on the same platform.

DISCUSSION

The recent discoveries of NTRK fusions in a previously unrecognized heterogeneous basket of mesenchymal lesions of either fibroblastic or neural-like differentiation have drawn focus on their clinicopathologic features. As a number of kinase inhibitors are now available for targeted strategies based on the specific genomic profile of these tumors (e.g. NTRK inhibitor Larotrectinib, LOXO-292 RET selective inhibitor), the quest for a refined pathologic and molecular diagnosis has become critical to match the correct patients with the appropriate targeted therapy. However, the emerging data from NTRK-fusion positive soft tissue tumors has revealed a broad morphologic spectrum, a rather non-specific immunoprofile and a wide clinical spectrum and behavior, which raises significant challenges in diagnosing these lesions. Triggered by an index case resembling lipofibromatosis-like neural tumor but lacking NTRK1 gene rearrangements and instead harboring a CCDC6-RET fusion, we sought to investigate the clinicopathologic features of a series of 6 RET fusion positive cases.

Similar to the NTRK-fusion group of tumors, most RET positive cases occurred in children, and often in infants. There was a wide anatomic distribution, including lower extremity, trunk, and kidney. All three patients with lipofibromatosis-like neural tumor occurred in females, two of them being infants occuring in the foot and one a 13 year-old girl with an abdominal wall soft tissue lesion. Similar to the lipofibromatosis-like neural tumor harboring NTRK1 fusion, the lesions showed highly infiltrative growth, two being superficial, while one of the foot lesions extended into skeletal muscle and encased bone. Microscopically, the two genetically distinct groups showed similar morphologic features, being composed of uniform spindle cells arranged in a streaming pattern or solid sheets within the adipocytic component. Although coexpression of S100 and CD34 was present in the infantile foot lesions, the pan-NTRK immunostain was negative, in contrast to the NTRK1-rearranged cases¹. Also, similar to NTRK1-positive tumors, none of the patients with RET-positive lipofibromatosis-like neural tumors followed an aggressive course, showing no tumor growth/progression despite incomplete resections in all cases. Only one previous lipofibromatosis-like neural tumor with NTRK1 fusion has been described in infants,¹³ which occurred as a 3.5 cm violaceous, hyperpigmented, slow-growing back plaque. It is tempting to speculate that RET fusions might be more permissive for this phenotype than NTRK1 in this particular clinical setting.

Two of the RET-rearranged cases had a morphology within the spectrum of infantile fibrosarcoma, associated with a non-specific immunoprofile. The first case, occuring in an infant (2 month-old boy), showed locally aggressive behavior, involving skeletal muscle and bone, however, no further recurrence was detected despite marginal resection. The 2^nd case occurred in the kidney and thus the synonymous terminology applied includes a cellular congenital mesoblastic nephroma. However, in contrast with the morphology and clinical behavior of most cellular congenital mesoblastic nephromas, the tumor was associated microscopically with high grade histologic features (>20 MF/10 HPFs, geographic necrosis) and the patient presented with widespread distant metastases at diagnosis. Recent studies have expanded the genetic alterations of tumors which fall in the spectrum of both infantile fibrosarcoma and cellular congenital mesoblastic nephroma, outside the cannonical ETV6-NTRK3 fusions, including NTRK1, BRAF and MET fusions as well as complex BRAF insertion/deletions^3,4,6,11,12. Of note, a larger subset of these non-cannonical translocations have been associated with aggressive clinical behavior, but the overall numbers remain too limited to draw more definitive conclusions. Most of the reported cases with alternative gene fusions have a spindle cell fascicular phenotype and non-specific immunprofile, in keeping with the infantile fibrosarcoma-like terminology, albeit a number of these cases present in children beyond the first 2 years of age. To complicate things further, a recent report described 2 cases of adult fibrosarcoma (one arising in the bone, the other in soft tissue) showing STRN-NTRK3 gene fusions⁵, suggesting that recurrent kinase fusions may be associated with a ‘fibrosarcoma phenotype’ occurring in a wide range of clinical presentations, including children or adults, and at soft tissue, visceral or bone locations.

In keeping with the above observation, one of our cases occurred in the soft tissue of an adult with morphology indistinguishable from a high grade malignant peripheral nerve sheath tumor, expressing focal S100 positivity, while SOX10 was negative and H3K27me3 was retained. No family history of NF1 or prior radiation was noted. Our group has recently reported recurrent kinase fusions (RAF1, BRAF, NTRK1) in a distinct group of low to intermediate grade spindle cell neoplasms with morphology and an immunoprofile reminiscent of malignant peripheral nerve sheath tumors². The current case showing a MYH10-RET fusion showed overtly high grade features and was associated with an aggressive clinical course. Further studies are needed to evaluate the incidence of RET or other kinase fusions among the high grade malignant peripheral nerve sheath tumors in patients lacking NF1 and showing H3K27me3 retained expression.

RET gene, located at 10q11.21 locus, encodes a transmembrane receptor and member of the tyrosine protein kinase family of proteins. Binding of ligands such as GDNF (glial cell-line derived neurotrophic factor) and other related proteins to the encoded receptor stimulates receptor dimerization and activation of downstream signaling pathways that play a role in cell differentiation, growth, migration and survival. The encoded receptor is important in development of the nervous system, and the development of organs and tissues derived from the neural crest. This proto-oncogene can undergo oncogenic activation through both cytogenetic rearrangement and activating point mutations. RET gene fusions have been described in papillary thyroid carcinoma¹⁴, salivary gland intraductal carcinoma^15,16 and small subsets of breast, colorectal and and non small cell lung cancer^17–19. Interestingly, papillary thyroid carcinomas from children exposed to radiation harbor either RET or NTRK fusions¹⁴. In mesenchymal lesions, RET gene rearrangement has only been reported in an isolated case of a pulmonary inflammatory myofibroblastic tumor²⁰; this tumor type being a prototypical example of a kinase fusion-driven mesenchymal spindle cell neoplasm. As seen with most other kinase fusions (i.e. ALK, ROS1, NTRK1, BRAF), RET fusions involved a wide range of gene partners, with a non-recurrent pattern in the 5 cases investigated, 2 resulting from an intra-chromosomal translocation, while the remaining 3 were interchromosomal. By RNAseq the different partners included: CLIP2 (7q11.23, cytoplasmic linker protein), CCDC6 (10q21.2, coiled-coil domain protein), SPECC1L (22q11.23, coiled-coil domain protein), MYH10 (17p13.1, myosin superfamily) and NCOA4 (10q11.22, androgen receptor coactivator) genes. Only CCDC6 and NCOA4 have been described as recurrent gene partners with RET in the setting of thyroid, lung and breast cancer, both resulting from intrachromosomal fusions. The latter NCOA4-RET fusion was shown to be constitutively activated when transfected in vitro and was inhibited in vivo by the RET inhibitor cabozantinib in a patient with metastatic breast carcinoma¹⁹.

In summary, we report 6 spindle cell neoplasms which share similar RET gene rearrangements. The morphologic spectrum, immunoprofile and clinical outcome were diverse. However, they appeared overall quite similar to the range of pathologic entities and clinical presentations described in spindle cell tumors with other kinase fusions, particularly those with NTRK1 rearrangements. Half of the cases showed features in keeping with a lipofibromatosis-like neural tumor, commonly occuring in children or infants, displaying S100 and CD34 co-expression, and following a benign clinical course. The remaining three cases showed malignant histology, resembling either infantile fibrosarcoma or malignant peripheral nerve sheath tumor, two of which followed a highly aggressive clinical course. As their immunoprofile is non-specific, molecular studies, such as FISH or targeted RNAseq, may be useful to confirm the diagnosis and to select patients for targeted strategies with RET inhibitors when appropriate.