BCOR-CCNB3-Fusion Positive Sarcomas: A Clinicopathologic and Molecular Analysis of 36 cases with Comparison to Morphologic Spectrum and Clinical Behavior of other Round Cell Sarcomas

Abstract

BCOR-CCNB3 sarcoma (BCS) is a recently defined genetic entity among undifferentiated round cell sarcomas, which was initially classified as and treated similarly to the Ewing sarcoma (ES) family of tumors. In contrast to ES, BCS shows consistent BCOR overexpression, and preliminary evidence suggests that these tumors share morphologic features with other tumors harboring BCOR genetic alterations, including BCOR internal tandem duplication (ITD) and BCOR-MAML3. To further investigate the pathologic features, clinical behavior, and their relationship to other round cell sarcomas, we collected 36 molecularly confirmed BCOR-CCNB3 sarcomas for a detailed histologic and immunohistochemical analysis. Four of the cases were also analyzed by RNA sequencing. An additional case with BCOR overexpression but negative CCNB3 abnormality showed a novel KMT2D-BCOR fusion by targeted RNA sequencing. The patients ranged in age from 2 to 44 years old (mean: 14.8; median: 15), with striking male predominance (M:F=31:5). The tumor locations were slightly more common in bone (n=20) than soft tissue (n=14), with rare visceral (kidney, n=2) involvement. Histologically, BCS showed a spectrum of round to spindle cells with variable cellularity, monomorphic nuclei and fine chromatin pattern, delicate capillary network, and varying amounts of myxoid or collagenous stroma. The morphologic features and immunoprofile showed considerable overlap with other round cell sarcomas with BCOR oncogenic up-regulation, i.e. BCOR-MAML3 and BCOR ITD. Follow-up available in 22 patients showed a 5-year overall survival of 72%, which was relatively similar to ES (79%, p-value=0.738) and significantly better than CIC-DUX4 sarcomas (43%, p-value=0.005) control groups. Local recurrences occurred in 6 patients and distant metastases (lung, soft tissue/bone, pancreas) in 4. Seven of 9 cases treated with an ES chemotherapy regimen with evaluable histologic response showed >60% necrosis in post-therapy resections. Unsupervised clustering by RNA sequencing data revealed that tumors with BCOR genetic alterations, including BCOR-CCNB3, BCOR-MAML3 and BCOR ITD, formed a tight genomic group distinct from ES and CIC-rearranged sarcomas.

INTRODUCTION

BCOR-CCNB3 sarcomas (BCS) were first identified by Pierron et al in 2012 among a large group of undifferentiated round cell sarcomas lacking known genetic alterations through a whole transcriptome sequencing and subsequent RT-PCR screen.¹ The genetic abnormality involves a paracentric inversion on the short arm of chromosome X, resulting in the fusion of 2 nearby genes BCOR and CCNB3 (10 Mb apart) and leading to the overexpression of CCNB3. This initial study showed that BCS occurs with predilection in male adolescents and skeletal locations and has a primitive round cell morphology reminiscent of Ewing sarcomas (ES).¹ Two subsequent smaller series of 10 and 6 cases, respectively, pointed out these tumors harbor a mixed round to spindle cell morphology and are evenly distributed between bone and soft tissue sites.^2,3 Based on these findings, the differential diagnosis of BCS is quite challenging, including not only tumors in the ES family but also undifferentiated spindle cell sarcomas, such as synovial sarcomas (SS) and malignant peripheral nerve sheath tumors (MPNST). Because of the common presentation within bone, BCS can be further confused with small cell osteosarcoma. Thus in most cases a molecular diagnosis demonstrating the presence of BCOR-CCNB3 fusion is required for a more definitive diagnosis. More recently, emerging evidence suggested that tumors with BCOR genetic abnormalities, including BCOR-CCNB3, BCOR-MAML3 and BCOR ITD (internal tandem duplications) share significant overlap not only at the gene expression level, due to BCOR up-regulation, but also morphologically.⁴ This study sets out to define the morphologic and immunohistochemical spectrum of a large cohort of molecularly confirmed BCS and evaluate features that overlap with other members of the so-called ‘BCOR family of tumors’. Furthermore, as these tumors have been managed with similar chemotherapy regimens to those used in ES, we also examined the therapy-related pathologic response in our cohort and compared the survival of the BCS study group to individual cohorts of molecularly defined ES, CIC-DUX4 sarcoma and SS.

MATERIALS AND METHODS

Study cohort

The Surgical Pathology files of MSKCC as well as the consultation files of the senior authors (C.R.A., C.D.M.F., B.C.D.) were searched for primitive round and spindle cell sarcomas that had molecular confirmation of BCOR-CCNB3 fusion by fluorescence in situ hybridization (FISH) or RNA sequencing (RNAseq). Hematoxylin and eosin stained slides were reviewed. Cases without available material for review were excluded. Three cases were reported previously (cases 18, 21, 27).^5,6 The study group was analyzed for demographic information, anatomic site, tumor size, and morphologic features, including cell type (round, spindle, mixed), degree of cellularity, type and amount of stromal component, nuclear features, mitotic activity and presence of necrosis. Available immuno-histochemical stains were reviewed, including BCOR, SATB2, cyclin D1, TLE1, Bcl-2, and CD99. 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)

FISH for BCOR-CCNB3 fusions was performed on 4-µm formalin-fixed paraffin-embedded tissue sections in 34 cases. Custom FISH probes were prepared using bacterial artificial chromosomes (BAC) flanking BCOR (telomeric: RP11-21D3, RP11-1105N2; centromeric: RP11-37K20, RP11-973F20) and CCNB3 (telomeric: RP11-58H17; centromeric: RP11-168F22, RP11-96H3). The FISH procedure was performed as previously described.⁷

RNA sequencing (RNAseq)

One and three cases were submitted to whole transcriptome sequencing (case 19) and targeted RNAseq (case 8, 21, 35), respectively. For whole transcriptome sequencing, total RNA was extracted from frozen tissues using RNeasy Plus Mini (Qiagen), followed by mRNA isolation with oligo(dT) magnetic beads and fragmentation by incubation at 94°C in fragmentation buffer (Illumina) for 2.5 minutes. After gel size-selection (350–400bp) and adapter ligation, the library was enriched by PCR for 15 cycles and purified. Paired-end RNAseq at read lengths of 50 or 51 bp was performed with the HiSeq 2000 (Illumina). For targeted RNAseq, RNA was extracted from FFPE tissue using Amsbio’s ExpressArt FFPE Clear RNA Ready kit (Amsbio LLC, Cambridge, MA). Fragment length was assessed with an RNA 6000 chip on an Agilent Bioanalyzer (Agilent Technologies, Santa Clara, CA). RNAseq libraries were prepared using 20–100 ng total RNA with the Trusight RNA Fusion Panel (Illumina, San Diego, CA), which targets a list of 507 genes of interest. Targeted RNAseq was performed on an Illumina MiSeq platform. Reads were independently aligned with STAR(ver 2.3) against the human reference genome (hg19) and analyzed by STAR-Fusion.

Reverse transcription-polymerase chain reaction (RT-PCR)

RT-PCR was performed in case 19 to validate the fusion transcript identified by RNAseq. RNA was extracted from frozen tissues by RNeasy Plus Mini (Qiagen) and reverse transcribed by SuperScript IV First-Strand Synthesis System (Invitrogen). PCR was performed by Advantage 2 PCR kit (Clontech, Mountain View, CA). Forward BCOR primer (5’-GACCTGGAAGCCTTCAACCC) and reverse CCNB3 primer (5’-GAAGAGAGATGCCTCCTCAGTG) were used. The reaction was run at an annealing temperature of 66°C for 35 cycles.

Expression profile analysis by Affymetrix U133A Plus 2 array

To compare the expression profile of BCOR-CCNB3 sarcomas (BCS) with morphologically similar tumors, including ES and SS, we have downloaded publicly available expression array data from GEO, including: 10 BCS (GSE34800) and 4 ES (GSE34800) from the study by Pierron et al,¹ 34 SS (GSE20196),⁸ and a control group of normal tissues (GSE7307) on Affymetrix Human Genome U133A Plus 2 platform. Gene signatures of each tumor entity were obtained by comparing to a control group of normal tissues, using log2-fold change threshold of positive 3 and negative 5, FDR 0.01. A Venn diagram was used to assess the relationship among these gene signatures. Supervised clustering was performed using the gene signature specific for each group.

Due to the overrepresentation of multiple HOX family genes in the signature of BCS, we further examined the expression of HOX genes in BCS, ES, SS (Affymetrix U133A Plus 2 array), BCOR ITD round cell sarcoma, and CIC-rearranged sarcoma (RNAseq data). The expression fold changes between each tumor type compared to the appropriate control groups (normal tissues for the Affymetrix array; various soft tissue tumors for the RNAseq dataset) were evaluated in each dataset.

Survival analysis

Follow-up data were available in 22 BCS patients. Statistical analysis was performed on an SPSS platform (version 24.0; IBM Corp., Armonk, NY). The overall survival (OS) time was measured in months from the date of diagnosis to the date of death. Kaplan-Meier estimate was used to calculate the OS. Survival data of 57 CIC-rearranged round cell sarcomas, 121 ES, and 34 head and neck SS from our previous studies were used for comparison.^9,10 The OS among different tumors were compared by log-rank analysis. A P<0.05 was considered as significant for all statistical analyses.

RESULTS

BCOR-CCNB3 sarcomas affect mainly male adolescents with slight preference for bone location

The study cohort of 36 patients showed a striking male predominance (31/36, 86%). The age at presentation ranged from 2 to 44 years old (mean & median: 15)(Table 1), with 81% of cases occurring between 10–20 years of age. The tumor locations were distributed between bone (n=20) and soft tissue (n=14), with rare visceral involvement (n=2). Axial locations and extremities were equally involved. Among skeletal lesions, femur was the most common single site (n=5), followed by tibia (n=4), while all the axial skeletal tumors involved pelvic bones (3 in sacrum, 2 in ilium, 2 in pubic ramus). Soft tissue lesions occurred within trunk (n=7), extremities (n=4), and head and neck (n=3). Kidney was the only visceral location identified in this cohort. The tumor sizes ranged from 3–27 cm (mean: 11.7; median: 11).

Table 1.

Clinicopathologic factors and follow-up information in BCOR-CCNB3 sarcoma patients

Case	Age/Sex	Location	Size (cm)	Mitotic rate (/10HPFs)	Surgery	Chemotherapy	RT	TNR	Follow-up (months)	Recurrence & Metastasis (site)
1	13/F	Soft palate	3	NA	YESΔ	(ES)	N	100 (IV)	NED (14)
2	15/M	Femur	14.5	11	YESΔ	(ES): VAC-IE, IT	N	100 (IV)	NED (43)
3	15/F	Pelvic cavity	10.3	3	YESγ	(ES): VAC-IE, IT	Y	95 (III)	NED (76)
4	9/M	Sacrum	5	10	YESΔ	(ES)	Y	(III)	NED (26)
5	13/M	Femur	15	1	YESΔ	(OS) then (ES)	N	95 (III)	AWD (60)	R
6	14/M	Iliac bone	16.3	4	YESΔ	(ES): VAC-IE, IT	N	60–70 (II)	NED (11)
7	2/M	RP/paraspinal	5	18	YESΔ	(ES): VAC-IE, IT	N	80 (II)	DUC (40)
8	17/M	Pubic ramus	NA	25	YESΔ	(ES) VA-IE	N	40 (I)	AWD (54)	M (pancreas, at diagnosis), R
9	14/M	Foot	NA	3	YESΔ	(ES)	Y	(I)	NA*	Lung nodules
10	15/M	Femur	NA	8	YESΔ	regimen unknown	N	NA	NED (31)
11	17/M	Calcaneus	15.5	22	YESγ	MTX, Cp	Y	NA	DOD(47)	R, M (lung, pelvis, scapula, thigh)
12	12/M	Pubic ramus	12	1	YES	(ES)	N	--	NED (51)
13	5/M	Calcaneus	3.7	4	YES	(ES)	Y	--	NED (10)
14	18/M	Shoulder	11	16	YES	(ES)	Y	--	NED (7)
15	18/F	Paraspinal C2–C6	7	3	YES	I, A, Mesna	Y	--	NED (26)
16	10/M	Femur	8.2	10	YES	I, A	N	--	NED (91)
17	18/M	Sacrum	7	NA	YES	Er, C, Nedaplastin	Y	--	AWD (16)
18	44/M	Thigh	14	14	YES	I, Cp	Y	--	DOD (22)	M (lung)
19	18/F	Sacrum	5	8	N	(ES) VAC-IE	Y	--	AWD (113)	R, M (lung, skull base)
20	14/M	Foot	5	1	YES	N	N	--	NED (44)	R
21	12/M	Kidney	13	4	YES	N	N	--	AWD (15)	R
22	2/M	Posterior neck	25	11	YES	N	N	--	AWD (11)
23	15/M	Chest wall	NA	6	YES	N	N	--	NED (21)
24	19/M	Pelvic cavity	20	9	YES	NA	NA	--	NA
25	21/M	Chest wall	NA	4	YES	NA	NA	--	NA
26	24/M	Tibia	11.7	15	YES	NA	NA	--	NA
27	11/M	Kidney	27	4	YES	NA	NA	--	NA
28	13/M	Tibia	NA	2	NA	NA	NA	--	NA
29	15/M	Leg	10	6	NA	NA	NA	--	NA
30	15/M	Elbow(bone)	NA	1	NA	NA	NA	--	NA
31	16/M	Tibia	NA	2	NA	NA	NA	--	NA
32	10/M	Femur	7.8	NA	NA	NA	NA	--	NA
33	13/M	Tibia	NA	4	Biopsied	NA	NA	--	NA*
34	15/M	Iliac bone	NA	13	Biopsied	NA	NA	--	NA*
35	16/M	Calcaneus	NA	3	Biopsied	A, Cp	N	--	NA*
36	13/F	Back/paraspinal	19.6	2	Biopsied	NA	NA	--	NA*

HPFs, high power fields; RT, radiation therapy; TNR, tumor necrosis rate; ES, Ewing sarcoma; OS, osteosarcoma; NED, no evidence of disease; AWD, alive with disease; DOD, dead of disease; DUC, dead of unknown cause; NA, not available; V, vincristine; A, doxorubicin/adriamycin; C, cyclophosphamide; I, ifosfamide; E, etoposide; MTX, methotrexate; Cp, cisplatin; Er, epirubicin; R, recurrence; M, metastasis;

^Δ, surgery was performed after neoadjuvant chemotherapy;

^γ, surgery performed after chemo-radiation;

^*, recent case

BCOR-CCNB3 sarcomas showed a wide morphologic spectrum ranging from round to spindle cell phenotype

The majority of cases (28, 78%) were composed predominantly of primitive round cells, while 8 (22%) cases showed mainly spindle cell morphology. Two-thirds of BCS with predominant round cell phenotype (19 cases) also had a minor ovoid to spindle cell component, while 3 of the 8 predominantly spindle cell BCS showed a minor round cell component. The round cell component had scant cytoplasm and monomorphic nuclei with fine chromatin, indistinct nucleoli, and smooth nuclear contours (Fig. 1A). The tumor cells were usually arranged in solid hypercellular sheets without a distinct architectural pattern. At the other end of the spectrum were tumors composed predominantly or exclusively of monomorphic spindle cells, commonly arranged in intersecting fascicles (Fig. 1B) and less often in a vague whorling pattern (Fig. 1C). Most of these spindle cell lesions showed medium to high cellularity, monomorphic, ovoid nuclei with similar fine chromatin pattern. Regardless of the round or spindle cell morphology, most of the tumors showed a rich capillary network (81%). One case revealed alternating low and high cellularity with perivascular condensation, reminiscent of the so-called ‘marbled pattern’ commonly seen in malignant peripheral nerve sheath tumor (MPNST)(Fig. 1D). Half of the cases showed some degree of myxoid stroma (n=19, 53%), either loosely separating small nests, cords or individual tumor cells, or forming a microcystic pattern (Fig. 1E–H). In some cases, an abrupt transition between hypercellular and hypocellular myxoid areas were noted (Fig. 1I). A third of cases (n=13, 36%) showed a more collagenous stroma, either as ropey collagen bundles separating individual tumor cells or as a more confluent collagenous stroma in areas of lower cellularity (Fig. 2A,B, D–F). Metaplastic ossification was noted in a metastatic lung lesion in one case (Fig. 2C). Necrosis, either focal or geographic, was present in 16 (44%) cases. Mitotic activity varied significantly, ranging from 1–25/10 high power fields (HPFs)(mean: 8; median: 6).

Figure 1. Histologic spectrum of BCS with round to spindle cells and occasional myxoid stroma.

(A) A predominant round cell morphology with uniform nuclei, fine chromatin pattern, and delicate vascular network;

(B) Monomorphic spindle cell morphology arranged in a fascicular pattern, reminiscent of SS;

(C) Short spindle to ovoid cells with vague whorling pattern;

(D) Alternating hypercellular and hypocellular areas, mimicking MPNST.

(E) Small amount of myxoid stroma between loosely arranged tumor cells is a common finding.

(F–H) Less common features includes: cell clustering (F), cord-like arrangement (G), and microcystic formation (H);

(I) Sharp contrast of hypercellular round cell component and hypocellular myxoid area

Figure 2. Infrequent morphologic patterns of BCS.

(A) Ropey collagen fibers separating spindle tumor cells.

(B) A more homogenous, silky collagenous background with scattered short spindle cells.

(C) Osteoid matrix deposition in a metastatic lung lesion, but not in the primary tumor.

(D–F) A chest wall tumor of a 15 year-old male shows alternating areas of hypocellular collagenous background (D, left upper; E) and hypercellular round cell component (D, right lower; F)(case 23).

(G–I) A foot mass of a 14 year-old male shows a purely spindle cell tumor with low to intermediate cellularity (G) and thick hyalinized collagen (H) in the primary lesion (case 20). Local recurrences 1 year and 3 years later showed a predominant round cell morphology (I). Both the primary and recurrent specimens were positive for BCOR-CCNB3 fusion by FISH. BCOR IHC showed patchy moderate to strong staining in the primary lesion (H, inset).

Two cases showed a deceptively bland and hypocellular spindle cell morphology in the primary lesion, but recurred as a highly cellular round cell sarcomas. One of them, a markedly cystic kidney tumor in a 12 year-old male patient, has been reported previously.⁶ The other case (case # 20) was a 14 year-old male who presented with a progressively enlarging and tender right 5^th toe mass, centered in the soft tissue and showing phalangeal bone destruction. The excised primary lesion was composed of relatively bland-spindle cells with low to medium cellularity, and associated with a collagenous to hyalinized stroma (Fig. 2G–H). Mitotic activity was low (1/10HPFs). BCOR immunostaining showed patchy moderate to strong staining (Fig. 2H, inset). The surgical margin was involved by the tumor and the patient underwent post-operative radiation therapy. One year later, the patient developed a local recurrence involving mainly the phalangeal bone. A wide excision was performed and showed a predominantly round cell sarcoma (Fig. 2I) with only focal spindle cell morphology, similar to the primary lesion. Both the primary and recurrent specimens were positive for BCOR-CCNB3 fusion by FISH.

BCOR-CCNB3 fusions identified by FISH and RNA sequencing

All except 2 cases were tested by FISH using the 3-color BCOR-CCNB3 fusion assay. FISH results typically showed inversions of both BCOR and CCNB3 and fusion of centromeric BCOR (5’) to centromeric CCNB3 (3’) (Fig. 3A,B). In some cases, additional unbalanced alterations were observed, such as deletion of telomeric BCOR (3’) signals, additional regional breaks, or deletion of the other allele.

Figure 3. FISH and RT-PCR testing for BCOR-CCNB3 fusions.

(A) Normal signal pattern: green and yellow signal probes flanking telomeric and centromeric ends of BCOR, respectively, while red signals flank CCNB3 on each side. (B) Abnormal FISH pattern: green and yellow signals show break-apart, while red signals split into two signals, representing centromeric and telomeric CCNB3 signals. A BCOR-CCNB3 fusion is confirmed when the yellow (centromeric BCOR, 5’) comes together with one of the red signals. (C) Sanger sequencing of RT-PCR product showed BCOR exon 15 fused to CCNB3 exon 5.

Four cases evaluated by RNAseq confirmed the BCOR-CCNB3 fusion, with a similar BCOR exon 15 to CCNB3 exon 5 transcript as previously reported (Fig. 4A).¹ The fusions were further confirmed by RT-PCR in 1 case (Fig. 3C). Akin to sarcomas with BCOR-MAML3 and BCOR ITD, the BCOR genetic abnormality consistently involves the end of the last exon of BCOR (exon 15)(Fig. 4A).^4,11 Marked mRNA up-regulation of the entire BCOR coding sequence was noted in both BCOR-CCNB3 and BCOR ITD tumors (Fig. 4B). In contrast, CCNB3 mRNA up-regulation was noted only in BCS, involving the distal part of the gene, after the exon 5 breakpoint (data not shown). By unsupervised clustering, BCS grouped together with BCOR ITD and BCOR-MAML3 tumors and was separated from ES and CIC-rearranged sarcomas (Fig. 4C).

Figure 4. ‘BCOR family of tumor’.

(A) Schematic diagram showing BCOR-CCNB3, BCOR-MAML3 fusions, and BCOR ITD (common duplicated region underlined) involving the last exon (exon 15) of BCOR. Representative sequences from RNAseq data demonstrating the alterations of BCOR (in blue) only a few nucleotides away from each other. (B) At exon level, up-regulations of BCOR in BCOR-CCNB3 (red) and BCOR ITD (orange) are observed up to the last exon, consistent with the genetic change at the end of exon 15. Black dots indicate other soft tissue tumors in the same platform as control. (C) Unsupervised clustering using RNAseq data shows clustering of BCOR-CCNB3 sample (red) with BCOR ITD (orange) and BCOR-MAML3 (purple) into a group, separate from Ewing sarcomas (blue) and CIC-rearranged sarcomas (green).

A novel KMT2D-BCOR identified by targeted RNA sequencing in a recurrent pelvic soft tissue mass

An additional round cell sarcoma, showing diffuse and strong positivity for BCOR and moderate staining for SATB2, while being negative for BCOR-CCNB3 fusion by FISH, was investigated by targeted RNA sequencing. The lesion presented as a large pelvic soft tissue mass (size: 11 cm) in a 10 year-old female. Microscopically, the tumor showed a round to short spindle cell phenotype, arranged in short fascicles and whorling pattern. RNAseq identified KMT2D-BCOR gene fusion with multiple breakpoints, including the reciprocal transcript (BCOR-KMT2D). The most abundant fusion junction read was KMT2D exon 42 to the last 173 bp of BCOR exon 4. BCOR mRNA expression was up-regulated, especially involving the 3’ exons of BCOR gene, distal to the breakpoint, supporting BCOR as the 3’ partner. The patient was treated with initial debulking surgery followed by chemotherapy, and subsequent re-excision showed rare microscopic residual foci of viable tumor (>90% chemotherapy response). The tumor recurred locally two years later.

BCS shares a similar immunoprofile with other tumors harboring BCOR genetic alterations

All BCS tested showed nuclear positivity for BCOR (18/18), typically with a strong and diffuse nuclear pattern (Fig 5A). BCOR expression was retained in the residual viable tumor in a post-chemotherapy resection, showing >90% tumor necrosis/fibrosis (Fig. 5B). In addition, most tumors were positive for SATB2 (10/12, 83%), cyclin D1 (9/10, 90%), TLE1 (8/10, 80%) and Bcl-2 (7/7, 100%)(Fig. 5C–E). The consistent immunoreactivity for these markers matched the mRNA up-regulation at gene expression level (data not shown). CD99 was only positive in about half of the cases (11/26, 42%) with variable intensity and extent. In addition, NTRK3 was identified as one of the differentially expressed genes in BCOR-CCNB3 tumors.

Figure 5. Immunohistochemistry of BCOR-CCNB3 tumors.

(A) BCOR staining is typically diffuse with a strong nuclear pattern. (B) Post-chemotherapy resection of case 7 showing 90% tumor necrosis and foci of residual viable tumor cells (inset) with retained BCOR staining. SATB2 (C), TLE1 (D), and cyclin D1 (E) are also expressed in the majority of cases.

Distinctive HOX gene family up-regulation in sarcomas with BCOR genetic alterations

By Affymetrix U133A array data analysis, BCS showed up-regulation of multiple members of the HOX gene family, including HOXA, HOXB, HOXC, and to a lesser extent HOXD (Supplementary Fig. 1). Some SS also showed HOX gene expression, especially for HOXD genes; while only occasional, non-recurrent HOX gene expression was noted in the 4 ES samples. Using RNAseq data, we found that HOX genes were also up-regulated in BCOR ITD tumors, especially for HOXA and HOXB, but not in CIC-rearranged sarcomas (Supplementary Fig. 1). Overall, both BCS and BCOR ITD tumors showed distinctive overexpression of multiple HOX genes. Similar results were previously shown by our group in round cell sarcomas with BCOR-MAML3 fusion.¹¹

Furthermore, we obtained the individual gene expression signatures of BCS (249 genes), ES (93 genes), and SS (150 genes) from Affymetrix U133A array data, as described above. A Venn diagram showed an overlap of 11 genes between the 3 sarcoma types, including EZH2, WNT5A, and PMP2 (peripheral myelin protein 2), with most of the remaining genes representing distinctive signatures for each subtype (Fig. 6A). Supervised clustering was performed using the BCS gene signature, showing distinctly separate groups of BCS, ES and SS samples (Fig. 6B).

Figure 6.

(A) Venn diagram showed limited transcriptional overlap between BCS, ES, and SS gene (11 genes in common). Additionally, 64 genes were shared between BCS and SS, and 8 genes between BCS and ES. (B) Supervised clustering using the BCS gene signature showed that ES (green), BCS (red), and most SS (blue) samples clustered into distinct individual clusters.

BCOR-CCNB3 sarcomas have a significantly better outcome than CIC-rearranged sarcomas

Twenty-two BCS patients with available follow-up were analyzed and compared to previously published data sets of localized ES, CIC-rearranged sarcomas, and head and neck SS.^9,10 The average follow-up duration of the study group was 38 months (range: 7–113 months). All except one patient with available follow-up presented with localized disease at diagnosis, while the remaining patient (17/M) with a pubic ramus tumor and soft tissue extension was found to have a pancreatic metastasis at diagnosis. During the follow-up period, 6 patients developed local recurrence and 3 distant metastases (Table 1). Metastatic sites included lung (n=3), skull base, pelvis, scapula, and thigh (n=1 each).

Of the 22 patients, 11 received neoadjuvant chemotherapy/chemoradiation followed by surgery, 7 had surgery followed by adjuvant chemotherapy, 1 chemotherapy, and 4 surgery alone. Most of the patients treated with chemotherapy were based on protocols for ES (Table 1). Seven of 9 patients treated with an ES chemotherapy regimen and evaluable histologic response showed significant treatment-related pathologic response (60–100% necrosis/fibrosis) in the post-therapy resection specimens. However, a recent case, not included in the survival analysis (case # 9), also treated with ES based neoadjuvant chemotherapy, showed no response. The patient underwent partial foot resection/ray amputation and post-operative radiation therapy. Lung nodules were identified radiographically.

At last follow-up, 13 patients had no evidence of disease, 6 were alive with disease, two died of disease, and one died of an unknown cause. The 3-year and 5-year survival rates were 93% and 72%, respectively. No significant prognostic difference was found in patients with mitotic rates ≥10/10 high power fields versus those with < 10/10 high power fields (p-value=1.000). Comparing these rates to the survival of ES, CIC-rearranged sarcomas, and SS (a head and neck cohort), a significant survival difference was noted (Log-rank p-value <0.0001), with CIC-rearranged sarcomas having the worst prognosis (Fig. 7). BCS showed a significantly better prognosis than CIC-rearranged sarcomas (p=0.005), while no statistical difference was noted between BCS and ES (p=0.738) or between BCS and head and neck SS (p=0.802).

Figure 7.

Overall survival of 22 BCS (blue), 121 ES (black), 34 SS (green), and 57 CIC-rearranged sarcomas (red). BCS was associated with a more favorable outcome compared to CIC-rearranged sarcoma (p=0.005), while no significant survival difference was noted between BCS and ES (p=0.738) or BCS and SS (p=0.802). Duration is shown in months.

DISCUSSION

In this study, we investigate a series of 36 BCS for their clinicopathologic features, including the morphologic spectrum and clinical outcome. Our results confirm the male predominance and predilection for older children and adolescents as the most commonly affected population.¹ In contrast to the initial report, our series reveals only a slight preference for skeletal (56%) over soft tissue (39%) locations, with equal distribution between axial sites and extremities.^1,2 Furthermore, although initially described as a prototypical round cell sarcoma, further evidence showed that common histologic features of BCS include a mixed round, ovoid to short spindle cell phenotype, arranged in fascicles or a whorling pattern, with often myxoid stroma, and arborizing capillary network.^2,3,5,12,13 In this series, 78% of cases showed a predominant round cell component, with most displaying, in addition, variable areas of ovoid to spindle tumor cells. The remaining 22% of cases showed predominantly monomorphic spindle cell features arranged in intersecting fascicles, reminiscent of MPNST or synovial sarcoma. Immuno-histochemically, all tumors tested were positive for BCOR, while cyclin D1, SATB2 and TLE1 were also positive in the majority of cases. The morphologic spectrum and immunostaining profile highlight the significant overlap with other tumor entities sharing BCOR genetic alterations.

BCOR was first identified as a co-repressor of BCL6, interacting with BCL6 through its N-terminal BCL6 binding domain.¹⁴ Subsequently, BCOR was found to be a core component of a variant PRC1.1 complex associated with chromatin remodeling and histone modification.¹⁵ As part of the PRC1.1 complex, the C-terminal BCOR PUFD domain binds to the RAWUL domain of PCGF1.¹⁶ Of interest, all BCOR genetic abnormalities implicated in sarcomagenesis, including BCOR-CCNB3, BCOR-MAML3 and BCOR ITD, involve the last exon of BCOR, which encodes the PUFD domain.¹⁶ Thus the formation of gene fusions and ITD may alter the binding affinity between BCOR and PCGF1, affecting the subsequent recruitment by KDM2B, ubiquitinylation of H2AK119, and downstream gene regulation.¹⁷ Additional fusions involving BCOR (i.e. ZC3H7B-BCOR) have also been reported in endometrial stromal sarcomas and ossifying fibromyxoid tumors, where BCOR is often the 3’ partner, contributing its exons 7–15.^18,19 BCOR-related translocations have also been described in hematoproliferative disorders involving different mechanisms, such as recurrent BCOR-RARA fusions (exon 12 of BCOR to exon 3 of RARA) in a small subset of acute promyelocytic leukemia; resulting in chimeric transcripts lack the PUFD domain at the C-terminus.²⁰ BCOR inactivating mutations have also been identified in rhabdomyosarcomas and several hematologic malignancies,^21,22 while germline BCOR mutation results in the oculofaciocardiodental syndrome.²³

CCNB3 is a relatively new member of the cyclin B family involved in cell cycle control,²⁴ being mainly expressed in the developing germ cells in the testis. The predicted BCOR-CCNB3 fusion oncoprotein retains the CCNB3 cyclin box, while lacking the N-terminal destruction box motif associated with CCNB3 degradation.²⁴ In adult human tissues, both BCOR and CCNB3 are expressed in the testis,^2,25 indicating that they may have important function in germ cell survival or development. In BCS, CCNB3 expression is consistently up-regulated.¹

Recent data from our group have suggested that undifferentiated sarcomas characterized by BCOR genetic alterations, spanning both gene fusions and ITD, share similarities at the morphologic and immunohistochemical level, despite the wide spectrum of clinical presentations and pathologic entities involved.^4,25 Although our preliminary results suggest that infantile soft tissue undifferentiated round cell tumors (URCS) with BCOR ITD, primitive myxoid mesenchymal tumor of infancy (PMMTI) and clear cell sarcoma of the kidney (CCSK) represent a morphologic continuum of a single pathologic entity based on the shared genetic alterations, it is less certain if sarcomas with BCOR-related fusions, such as BCOR-CCNB3, also fit within this spectrum. The present study investigating the morphologic and immunohistochemical profile of a large group of BCS further emphasizes the similarities to the BCOR ITD genetic group, such as URCS, PMMTI and CCSK. Despite their predilection for infants and soft tissue, infantile URCS with BCOR ITD or YWHAE fusions show striking morphologic overlap with BCS, including round to spindle cell morphology, fine chromatin pattern, rich capillary background, and myxoid stroma.⁴ Additionally, BCS and BCOR ITD tumors show an identical immunoprofile, with strong/diffuse reactivity for BCOR, SATB2, and cyclin D1 in the majority of cases.^6,25 However, despite the close genomic grouping by RNAseq unsupervised clustering, BCOR ITD spectrum of tumors do not show CCNB3 mRNA expression, a common finding in BCS. Likewise, the 2 renal BCS reported by our group recently resemble CCSK, both morphologically and immunohistochemically.⁶ Being the second most common pediatric renal tumor following Wilms’ tumor, CCSK typically shows uniform round to ovoid cells with fine chromatin separated by a rich vascular network and is characterized by BCOR ITD and BCOR up-regulation in the majority of cases.²⁶ Nevertheless, CCSK mainly affects children between 2 to 4 years old, in contrast to the teenage group of BCS.

As the majority of BCS show some evidence of spindling, with 12% of BCS harboring a predominantly spindle cell fascicular growth, the differential diagnosis of BCS also includes spindle cell sarcomas, specifically monophasic SS, fibrosarcoma and MPNST. In addition, half of SS are positive for BCOR, and both SS and BCS are immunopositive for TLE1.^25,27 SATB2 reactivity, often expressed in BCS (80%), can also be detected in a small subset of SS (12%).²⁵ CCNB3 staining has been reported to be negative in SS, except nonspecific cytoplasmic staining in some cases.⁵ A further pitfall was recently revealed in a SS18-negative spindle cell sarcoma showing diffuse BCOR expression and morphologically suspected to represent a BCOR-alteration sarcoma, however, RNA sequencing found a rare SS18L1-SSX variant fusion, diagnostic of SS.⁷

Despite their similar clinical and demographic presentation, BCS shows histologic features distinctive from ES, including the more variable cytomorphology, from round, ovoid to spindle, common myxoid background and a rich capillary network. In contrast, ES is usually characterized by solid sheets of perfectly monomorphic round cells, ill-defined cell borders and fine chromatin, and the lack of significant stromal component. Immunohistochemically, ES shows consistent strong and diffuse CD99 membranous staining, which is much more variable in BCS. Cyclin D1 is expressed in both ES and BCS.²⁸ In contrast, strong and diffuse BCOR expression is the defining feature for BCS, while being only infrequently expressed in ES (1/9, 11%).²⁵

Another important differential diagnosis is with CIC-DUX4 fusion positive undifferentiated round cell sarcoma. In contrast to BCS, CIC-rearranged sarcomas typically affect young adults rather than children (mean age of 32) with no gender predilection. The overwhelming majority occur in soft tissue locations with infrequent intra-osseous presentation.⁹ Morphologically, CIC-rearranged sarcomas show a higher degree of nuclear variability, with prominent nucleoli and consistent immunoreactivity for ETV4 and WT1.^9,29

As BCS often presents as an intra-osseous lesion with a round cell/undifferentiated phenotype, expression of SATB2 might suggest an osteoblastic line of differentiation and be misinterpreted as a small cell osteosarcoma. Indeed 2 cases in this series were initially misdiagnosed as small cell OS involving femur and tibia, respectively, both locations being common for both diagnoses.³⁰ Osteoid or metaplastic bone formation in BCS is rare and so far only observed in metastatic lung lesions, including the case reported in our series.² Thus, our results point out that SATB2 expression is not specific for bone-forming tumors, and the immunowork-up of a skeletal round cell malignancy should include, in addition to SATB2, other markers such as BCOR and cyclin D1 to exclude a BCS diagnosis.

The RNA sequencing data from the 4 BCS cases and the publicly available expression array data from the Pierron study¹ revealed significant mRNA up-regulation of NTRK3 in BCS, which was also confirmed in a subset of our cases at the protein level (data not shown). The mechanism of NTRK3 mRNA up-regulation remains unclear. Similar NTRK3 overexpression is also observed in the infantile URCS with BCOR ITD (data not shown).

Besides the morphologic differences between BCS and other undifferentiated round to spindle cell sarcomas, such as ES, CIC-rearranged sarcomas, and SS, our study also addressed their clinical behavior. Previous studies suggested a 5-year overall survival rate of 75% for BCS, with axial location, local recurrence, and metastasis being poor prognostic factors.^2,12 In patients with localized disease, induction therapy with ES-based protocols was associated with a better outcome.¹² The overall survival of BCS reported in a small series of 10 patients (6 with localized stage at presentation) was not statistically different from an ES cohort.² However, no prior studies have compared the outcome of BCS with CIC-rearranged sarcomas or SS, two pathologic entities commonly included in the differential diagnoses. Our study analyzed the clinical outcome of 22 BCS patients (mean follow-up: 38 months) showing that the overall survival of BCS (3-year: 93%, 5-year: 72%), is superior to CIC-rearranged sarcomas, while not statistically different from ES and SS. Given that these retrospective cohorts were composed of patients treated non-uniformly in different institutions, further investigation of patients with matched risk factors, disease stages, and uniform treatment protocols would be needed to confirm this finding. An additional point of investigation was assessing the pathologic response of BCS tumors to the applied neoadjuvant chemotherapy regimens. As BCS was originally classified among ES family of tumors, most patients have been managed so far with ES-related chemotherapy protocols. Two previous studies have suggested that BCS are chemo-responsive;^2,12 one study showing a good histologic response (>90% necrosis) in 77% (10 of 13) of the evaluable patients treated mainly with ES chemotherapy but also other regimens.¹² In another study, 4 of 6 post-chemotherapy resections showed complete response, while the remaining 2 had scattered residual tumor cells.² Our study has further investigated the extent of tumor response in 9 resection specimens treated with ES regimens, showing 5/9 patients with >90% necrosis and 2/9 patients with >60% necrosis. Of note, other tumors with oncogenic BCOR up-regulation, e.g. CCSK, have been treated with somewhat different chemotherapy regimens, which usually do not include ifosfamide as in most ES regimens.^31,32 Given the overlapping morphology and transcriptional profile of BCS and CCSK and given that CCSK has shown sensitivity to doxorubicin-based chemotherapy,^32,33 it remains to be determined if BCS patients might also benefit from an overall less toxic protocol compared to the presently applied ES-based regimens (doxorubicin and ifosfamide combination). Another consideration is that CCSK are well-known to relapse many (over ten) years after diagnosis, mandating long term follow-up for these children. Given the genetic relatedness of BCS and CCSK, it is possible that BCS patients may relapse late too, though there is little data to address this concern at this time.

As indicated by previous microarray data analysis¹ and our RNAseq hierarchical clustering, BCS showed a distinct transcriptional profile compared to ES. This study further demonstrated the BCS expression signature is similar to tumors harboring BCOR ITD and BCOR-MAML3, with a significant up-regulation of HOX (homeobox) family genes.^1,11 HOX genes encode a group of transcription factors responsible for embryo development, including anterior-posterior body axis, and adult fracture healing process.^34,35 HOX abnormalities have also been reported in various cancer types, including carcinomas of breast, colon, prostate, and lung.³⁶ In sarcomas, overexpression of HOX genes has been reported in ES,³⁴ but restricted to HOXD genes, which was not observed in our 4 ES samples. In addition, WNT and Hedgehog signaling pathways are similarly activated in both BCS and BCOR ITD tumors of central nervous system.^1,37,38

In conclusion, we report a large series of 36 BCS, highlighting the significant morphologic, immunohistochemical and transcriptional overlap with other primitive round/spindle cell sarcomas with BCOR gene alterations, despite differences in their demographics and clinical presentations. Our results also show that BCS have histologic features, immunoprofile and expression signatures which are distinct from other common round cell sarcomas, such as ES and CIC-DUX4 tumors. Emerging evidence suggests that most BCS tumors are chemosensitive to ES-based therapy protocols, although other less toxic regimens currently applied in CCSK might also be investigated. Lastly, the overall survival rate of BCS is significantly more favorable than CIC-rearranged sarcomas, but appears similar to ES and SS though long term follow up may be needed.

Supplementary Material

Supplemental Data File _.doc_ .tif_ pdf_ etc._

Supplementary Fig. 1. Heatmap of HOX genes expression in undifferentiated round to spindle cell sarcomas. BCS (red), BCOR ITD sarcoma (orange), and BCOR-MAML3 round cell sarcoma (yellow) showed up-regulation of multiple HOX genes, with BCS and BCOR-MAML3 round cell sarcoma spanning HOXA, HOXB, HOXC genes, while BCOR ITD URCS mainly HOXA and HOXB genes. SS (gray) also showed scattered HOX gene expression, mainly focused on HOXD. No consistent HOX gene up-regulation was noted in ES (blue) and CIC-rearranged sarcoma (green). BCS, SS, and ES were analyzed on Affymetrix U133APlus2 array data and BCOR ITD URCS, BCOR-MAML3 round cell sarcoma, and CIC-rearranged sarcoma on RNAseq data. The fold changes of expressions in comparison to the control groups in each dataset are demonstrated in the color scale.