Introduction

GLI1 fusions and amplifications are the main oncogenic events that characterize a subset of epithelioid soft tissue (mesenchymal) tumors with a predilection for the head and neck [1,2,3,4]. Most are characterized by monomorphic epithelioid and/or spindled cells and are frequently associated with a rich vascular network. Immunohistochemistry has thus far been unreliable for diagnosing these tumors. Even targets that reside near the GLI1 gene locus on chromosome 12p, including CDK4, MDM2, and STAT6 on chromosome 12q13–15, are not entirely sensitive, especially for tumors harboring a rearrangement [1]. As a result, molecular confirmation of a fusion or amplification involving the GLI1 locus is typically required for definitive diagnosis.

Most GLI1-altered soft tissue tumors reported in the head and neck region have occurred in the tongue or soft tissue of the neck with rare cases in the floor of mouth [2]. The oropharynx is a lesser-known affected site, with only three cases reported to date: two in the base of tongue (one with extension to the right oral tongue and lateral pharyngeal wall) and one soft palate (with extension to the lateral nasopharyngeal wall) [2,3,4].

Here, we describe five additional cases of GLI1-altered soft tissue tumors, including oropharyngeal primaries, and characterize the extent and pattern of p16 expression by immunohistochemistry.

Materials and Methods

Cases

Cases of GLI1-altered soft tissue tumors were retrieved from the consultation and/or institutional files of three authors (JAB, LMR, and LDRT). Hematoxylin and eosin-stained slides from all cases along with previously performed immunohistochemical stains were reviewed. Additional immunohistochemical studies were performed on select cases using a monoclonal antibody directed against p16 utilizing standard techniques. This study was approved by the Institutional Review Boards at our respective institutions. Four cases (cases 1 to 4) are new while case 5 was previously reported [1, 5].

Fluorescence In Situ Hybridization (FISH)

DDIT3 FISH was performed on 3 of 5 cases using either: (1) the Vysis (Abbott, Abbott Park, Illinois, USA) LSI (12q13) dual color, break apart DNA probe set; and/or (2) the CytoCell (Oxford Gene Technology, Cambridge, UK) dual color, break apart DNA probe set. A custom GLI1 locus spanning probe coupled with CEN12 (copy number control) was used to assess for GLI1 specific amplification in cases with available material.

RNA Sequencing

Case 2 was subjected to a custom, clinically validated NanoString fusion assay [6] while Case 3 underwent a Sarcoma Targeted Gene Fusion/Rearrangement Panel (Mayo Clinic). Targeted RNA sequencing (RNA-seq) was performed on Case 4 at the UTSW NGS Clinical Lab at the Once Upon a Time Human Genomics Center. Briefly, whole-slide tissue sections were cut at 10 µm, and Qiagen AllPrep kits (Qiagen, Germantown, MD) were used for RNA isolation. A sequencing library was generated using a modified TruSight RNA Pan-Cancer kit (Illumina, San Diego, CA) with 1425 genes. Sequencing was performed on the NextSeq 550 (Illumina, San Diego, CA) with a minimum of 6 million mapped reads. Fusions were called using the Star-Fusion algorithm [7] and were manually reviewed via the Integrated Genomics Viewer (Broad Institute, Cambridge, MA).

Results

Clinicopathologic Findings

Clinicopathologic information is summarized in Table 1. The patients ranged in age from 35 to 84 years, all males. Tumors ranged from 15 to 60 mm in greatest dimension. Locations included the oropharynx (Figs. 1, 2, and 3: 1 tonsil, 1 base of tongue/vallecula, 1 not otherwise specified) and tongue (Figs. 4 and 5: 1 left dorsal tongue, 1 left ventral tongue). Incisional or excisional biopsies for routine diagnosis were reviewed, with one case obtained by radical tonsillectomy with partial resection of the posterior pharyngeal wall.

Table 1 Clinicopathologic, immunophenotypic, and molecular characteristics of GLI1-altered soft tissue tumors in the current study
Full size table
Fig. 1
figure 1

Right palatine tonsil GLI1-amplified soft tissue tumor (Case 1). The tonsil is effaced by a cellular multinodular spindle cell proliferation arranged in loose fascicles (A, B, C, D) within a variably myxoid to hyalinized stroma, reminiscent of a nerve sheath tumor, solitary fibrous tumor, or monophasic synovial sarcoma. The tumor cells contain abundant eosinophilic cytoplasm and oval to spindled nuclei with mild nuclear pleomorphism. Additional features include dilated vessels (B), a sprinkling of intratumor lymphocytes, and focal epithelioid morphology (E, F) with a prominent capillary vasculature (F)

Full size image
Fig. 2
figure 2

Base of tongue ACTB::GLI1 fusion-related soft tissue tumor (Case 2). A cellular proliferation of epithelioid cells arranged in nests and anastomosing trabecula (A–C) is seen beneath an ulcerated mucosal surface (A). Hyalinized stroma is variably present (B, C) along with occasional intratumoral lymphocytes (C). There is strong nuclear and cytoplasmic, block-like p16 immunoexpression in > 70% of tumor cells (D)

Full size image
Fig. 3
figure 3

Oropharynx ACTB::GLI1 fusion-related soft tissue tumor (Case 3). Low power shows expanded nodules of monotonous epithelioid cells (A–C). Variably hyalinized stroma is seen between some of the lobules (B–C). Within the nodules, an arborizing thin-walled vasculature show perivascular tumor cell nests (D–E). Focal myxoid change with microcysts is noted (D). High power view demonstrates round to ovoid nuclei with clear to pale eosinophilic cytoplasm (F)

Full size image
Fig. 4
figure 4

Dorsal tongue ACTB::GLI1 fusion-related soft tissue tumor (Case 4). Infiltrating lobules of epithelioid cells can be seen infiltrating beneath the epithelium (A–C) and protruding into a superficial blood vessel (A). Myxoid change with microcyst formation is focally present (B). Tumor cells are arranged in a perivascular distribution around intratumoral blood vessels (D–E). Strong nuclear and cytoplasmic block-like p16 immunoexpression seen in 50% of tumor cells (F). DDIT3 FISH (CytoCell break apart probe set; green 165 kb probe 5′ to the DDIT3 gene locus and red 146 kb probe 3′ to the GLI1 gene locus) was positive for rearrangement of the 12q13.3 region (G, left-hand image, arrows indicate the cells with split red and green signals). Custom FISH probe specifically spanning the GLI1 locus coupled with a CEN12 FISH probe was negative for GLI1 specific amplification (G, right-hand image, arrows indicate cells with two CEN12 signals and three GLI1 signals including split of at least one of the GLI1 signals secondary to rearrangement involving the GLI1 specific locus)

Full size image
Fig. 5
figure 5

Ventral tongue GLI1-amplified soft tissue tumor (Case 5). Tumor cells are round to focally spindled (A–C) and separated by a variably hyalinized fibrovascular stroma. Strong nuclear and cytoplasmic block-like p16 immunoexpression in > 70% of tumor cells (D)

Full size image

Histologically, tumors were predominantly composed of relatively monomorphic oval shaped cells with clear to pale, eosinophilic cytoplasm, delicate speckled chromatin, and small variably conspicuous nucleoli (Figs. 2, 3, 4, and 5) with the exception of one case which demonstrated a cellular multinodular spindle cell proliferation arranged in loose fascicles within a variably myxoid to hyalinized stroma with dilated vessels (Fig. 1). However, focally the tumor cells were more epithelioid and nested with a prominent capillary vasculature (Fig. 1E–F). At low power, tumors appeared to be relatively circumscribed, but at higher power demonstrated varying degrees of infiltration at the periphery. A fine vasculature with peritheliomatous tumor cell arrangement in vague nodules and anastomosing cords was common. Protrusion of tumor nests into vascular spaces was seen in only one case (Fig. 4). Collagenous stroma was present in varying amounts in all cases but was most notable in Case 1 (Fig. 1A–C). Four cases showed overlying mucosal ulceration. Infiltrating lymphocytes were not prominent. Mitotic activity was moderate in one case (Case 1; 5–7 mitoses/2 mm2). Perineural invasion and necrosis was uniformly absent.

All cases tested by immunohistochemistry showed at least focal expression of CDK4 (3/3) and MDM2 (4/4), while STAT6 was variable (negative–2; weak, focal–2; moderate, diffuse–1). Case 1 was additionally positive for bcl-2, GLUT1, vimentin, and CD56, focally positive for CK8/18 and CD34, and negative for EMA, CD21, CD23, AR, PR, chromogranin, synaptophysin, neurofilament, AE1/AE3, CD99, WT1, ALK1, S100 protein, SOX10, GFAP, caldesmon, calponin, SMA, CD31, ERG, CK5/6, CK14, p63, p40, and beta-catenin. Given the anatomic location and epithelioid morphology, p16 immunohistochemistry was performed on two oropharyngeal cases with available material (Case 1 and 2) and showed strong nuclear and cytoplasmic block-like staining in 50–70% of tumor cells; the ventral and dorsal tongue tumor cases demonstrated similar p16 results (Table 1).

Molecular Alterations

RNA-seq showed ACTB::GLI1 fusion in 3 of 5 tumors (Case 2–4). DDIT3 break apart FISH was performed on 1 of the cases with GLI1 fusion (Case 4) and was positive for rearrangement of the 12q13.3 region using both the CytoCell and Vysis probe sets while GLI1 specific FISH was negative for GLI1 amplification (Fig. 4G). Cases 1 and 5 were positive for amplification by custom GLI1 specific FISH and Vysis DDIT3 FISH. Case 1 additionally demonstrated increased copy numbers of MDM2 (Zytovision ZytoLight SPEC, IGENZ, Auckland, NZ) and GLI1 using Kreatech Poseidon FISH probes (IGENZ, Auckland, NZ) but no rearrangement of SS18 (Vysis).

Discussion

GLI1-altered soft tissue tumors characteristically harbor either fusions (most commonly involving ACTB, MALAT1, and PTCH1 gene partners) or high-level amplifications of the GLI1 locus, frequently accompanied by co-amplification of other genes mapped to the chromosome 12q13–q15 region (e.g., MDM2, CDK4, DDIT3, and STAT6). Patients across a wide age range (1–88 years) may present with tumors arising in bone and soft tissue [5, 8, 9], skin [10], lung [5], genitourinary system [8, 11, 12], gastrointestinal tract [13, 14], oral cavity [1, 2, 5, 13], or oropharynx [2,3,4]. Morphologically, this tumor family is typically characterized by a monomorphic population of epithelioid and/or spindled cells with a prominent perivascular distribution of growth, reminiscent of pericytic neoplasms. The immunoprofile of GLI1-altered soft tissue tumors, however, is known to be highly variable with immunoreactivity for relatively nonspecific markers (cyclin D1, CD10, BCOR, bcl-2) and inconsistent expression of S100 protein, actins, keratins, and neuroendocrine markers. Reported cases with available follow up data suggest these lesions behave similarly to low-grade sarcomas and have around a 20% rate of local recurrence and metastasis to regional lymph nodes and distant sites [1, 2, 4, 8, 14,15,16]. Tumor-related death was reported in one case arising in the uterus and metastasizing to brain [15].

Although a few cases of oropharyngeal GLI1-altered soft tissue tumors have been described [2,3,4], our series is the first to report one specifically arising in the palatine tonsil (Table 2). The tonsillar regions of the oropharynx (palatine and lingual tonsils) are particularly unique given the rich supply of lymphatics and small blood vessels, “immune privileged” environment, and discontinuous basement membrane of the reticulated epithelium [17, 18]. Tumors arising in these locations are more prone to regional lymph node spread [19]. Although none of the oropharyngeal primaries presented here have any follow up data, Zhong et al., [4] described a GLI1-rearranged tumor of the right base of tongue that metastasized to an ipsilateral level II lymph node and contralateral sacral soft tissue. The overall prognosis of head and neck GLI1-altered tumors, however, seems to be favorable, despite their potential for local and distant metastases, with no tumor-related deaths reported to date [1, 5].

Table 2 Literature review of GLI1-altered soft tissue tumors primary to the oropharynx
Full size table

Our series is the first to report the consistent finding of strong nuclear and cytoplasmic block-like p16 immunoexpression in GLI1-altered soft tissue tumors. p16 (INK4a/CDKN2A) is a cell cycle protein that inhibits CDK4, keeping Rb hypophosphorylated and preventing cell cycle progression. It is typically overexpressed secondary to oncogene-induced senescence or as a mechanism to arrest the uncontrolled proliferation caused by failure of the Rb pathway in tumors. In HPV-related squamous cell carcinoma (SCC), for example, p16 is overexpressed secondary to viral inactivation of Rb leading to the removal of Rb’s negative regulation of p16. The significance of increased expression of p16 in GLI1-altered tumors is unclear but not entirely unexpected as it is seen in differentiated and de-differentiated liposarcomas [20], which harbor amplifications of the 12q13-15 region encompassing MDM2 and CDK4 [5]. Induced expression of p16 inhibits CDK4-associated kinase activity [21] and GLI1 has been shown to directly activate the PI3K/AKT pathway and subsequently up-regulate CDK4/6 protein [22]. GLI1 is also an important Hedgehog pathway transcriptional regulator of genes such as cell cycle regulators CCND1/2 and CCNE1 [23]. Therefore, p16 overexpression in GLI1-altered tumors could be explained by the concept of induced cellular senescence in which cancer cells attempt to suppress the G1/S transition driven by GLI1 induced oncogenic accumulation of cyclins and CDK4. Extended molecular analysis of a MALAT1::GLI1 fusion-related tumor by Prall et al., [14] also detected chromosomal losses involving 13q (harboring BRCA2 and RB1) and 17p (harboring TP53) as well as a truncating TP53 mutation; thus, one could also anticipate inactivation of RB1 and/or loss of p53 function leading to p16 overexpression.

In the oropharynx, p16 is used as a surrogate marker for transcriptionally active high-risk HPV in primary oropharyngeal SCCs. Although GLI1-altered tumors seem to be morphologically distinctive relative to HPV-related SCCs, if only limited biopsy material is available, one could easily misdiagnose a GLI1-altered tumor as a nonkeratinizing HPV-related SCC, or, if there is immunoreactivity for neuroendocrine markers, an HPV-related neuroendocrine carcinoma, which is typically treated with platinum-based chemotherapy. Follicular dendritic cell sarcoma can also occur in the oropharynx and overexpress p16 but would typically stain with dendritic cell markers such as CD21, CD23, and CD35 [24, 25]. Further, given that GLI1-altered soft tissue tumors may show hyalinized stroma with monomorphic, low-grade clear cells, hyalinizing clear cell carcinoma, which can also express p16 and arise in the oropharynx [26], could enter the differential diagnosis. The spindled morphology of a subset of GLI1-altered soft tissue tumors also overlaps with that of solitary fibrous tumor, a fibroblastic neoplasm with variable clinical behavior that rarely occurs in the oropharynx [27, 28]. Solitary fibrous tumors are distinctly characterized by a recurrent intrachromosomal NAB2::STAT6 fusion with the resultant fusion protein acting as a transcriptional activator through early growth response (EGR)-mediated pathways [29]. Not only do solitary fibrous tumors show increased nuclear immunoexpression of the C terminus of STAT6, a subset of cases may also overexpress p16 [30]. The small number of cases in our series limits the generalizability of p16 overexpression in GLI1-altered tumors but highlights the potential pitfall for misdiagnosis in the oropharynx.

Cases reported in the literature thus far have used a variety of methodologies including targeted exome sequencing [5, 11, 15], RNA sequencing [1,2,3, 5, 10, 11, 15, 16, 31], and FISH [1, 2, 4, 5, 11, 15, 31]. FISH is frequently the assay of choice when compared to the time and resources required for targeted exome or RNA sequencing; however, it is currently unclear whether a specific GLI1 targeted probe set is required for detection. FISH analysis thus far has largely been limited to custom bacterial artificial chromosome (BAC) clones flanking the DDIT3 gene locus with or without GLI1 (Table 3). The GLI1 and DDIT3 gene loci are located near each other on 12q13.3; the distance between the 3′ GLI1 gene locus (ENST00000228682.7; chr12:57459785-57472268; 12,484 bp) and 3′ DDIT3 gene locus (ENST00000623876.2; chr12:57516588-57521737; 5,150 bp) is 44,320 bp (44.320 kb). Commercially available FISH break apart probes for DDIT3 are commonly used to identify translocations involving DDIT3 associated with myxoid/round cell liposarcoma. The Vysis DDIT3 break apart FISH probes flank DDIT3 (approximately 134 kb gap) with a 700 kb probe lying proximal to DDIT3 (spanning STAT6 and GLI1 gene loci) and a 663 kb probe extending distally from DDIT3 (spanning the CDK4 gene locus), while the CytoCell DDIT3 break apart FISH probes flank both GLI1 and DDIT3 (100 kb gap) with a 146 kb probe lying proximal to GLI1 and a 165 kb probe extending distally from DDIT3. Our experience using these two commercially available DDIT3 break apart FISH probe sets suggests that it is an effective surrogate for GLI1, and that GLI1 specific FISH probes and custom BAC clones are not needed to diagnose GLI1-altered (rearranged or amplified) tumors in routine practice. One limitation of utilizing FISH assays is the possibility of cryptic intrachromosomal inversion events resulting in false negative FISH results, but these are not yet reported for GLI1. For GLI1-amplified tumors, DDIT3 break apart FISH also appears to be sufficient for diagnostic purposes regardless of co-amplification status of neighboring genes (e.g., STAT6, MDM2, and/or CDK4).

Table 3 FISH probes used in previous studies for GLI1-altered soft tissue tumors
Full size table

In summary, GLI1-altered soft tissue tumors can occasionally arise in the oropharynx, show strong nuclear and cytoplasmic block-like p16 immunoexpression, and can be diagnosed using commercially available DDIT3 break apart FISH. Like many other immunohistochemical markers, p16 expression is nonspecific and should be interpreted with caution in small biopsy material where characteristic histomorphologic features may be obscured or limited.