SalvGlandDx - a comprehensive salivary gland neoplasm specific next generation sequencing panel to facilitate diagnosis and identify therapeutic targets.

Diagnosis of salivary gland neoplasms is often challenging due to their high morphological diversity and overlaps. Several recurrent molecular alterations have been described recently, which can serve as powerful diagnostic tools and potential therapeutic targets (e.g. NTRK or RET fusions). However, current sequential molecular testing can be expensive and time consuming. In order to facilitate the diagnosis of salivary gland neoplasms, we designed an all-in-one RNA-based next generation sequencing panel suitable for the detection of mutations, fusions and gene expression levels (including NR4A3) of 27 genes involved in salivary gland neoplasms. Here we present the validation of the "SalvGlandDx" panel on FFPE histological specimen including fine needle aspiration (FNA) cell block material, against the standard methods currently used at our institution. In a second part we describe selected unique cases in which the SalvGlandDx panel allowed proper diagnosis and new insights into special molecular characteristics of selected salivary gland tumors. We characterize a unique salivary gland adenocarcinoma harboring a ZCCHC7-NTRK2 fusion, a highly uncommon spindle cell and pseudoangiomatoid adenoid-cystic carcinoma with MYBL1-NFIB fusion, and a purely oncocytic mucoepidermoid carcinoma, whereas diagnosis could be made by detection of a CRTC3-MAML2 rearrangement on the cell block specimen of the FNA. Further, a rare case of a SS18-ZBTB7A rearranged low-grade adenocarcinoma previously described as potential spectrum of microsecretory adenocarcinoma, is reported. In addition, features of six cases within the spectrum of polymorphous adenocarcinoma / cribriform adenocarcinoma of salivary gland including PRKD1 p.E710D mutations and novel fusions involving PRKAR2A-PRKD1, SNX9-PRKD1 and ATL2-PRKD3, are described.

Introduction

Salivary gland neoplasms often show a broad morphological spectrum including a variety of overlapping morphological patterns between different entities [1]. The diagnosis of these neoplasms can therefore pose significant difficulties. Especially on fine needle aspiration (FNA) or small biopsies, the differentiation between benign and low-grade malignant tumors can sometimes be impossible based on morphology alone [1, 2]. In recent years, the knowledge about genomic landscapes of salivary gland neoplasms has significantly increased. Thereby, the distinct molecular aberrations are often highly specific to certain entities and can be exploited for diagnostic purposes [3]. The typical molecular aberrations encompass mutations, such as recurrent activating HRAS and/or PIK3CA mutations in epithelial-myoepithelial carcinoma [4]. Furthermore, increased expression levels of NR4A3 due to enhancer hijacking have been described recently as a hallmark of acinic cell carcinoma [5]. In addition, certain aberrations, such as NTRK3 fusions, which are highly recurrent in (mammary analogue) secretory carcinoma provide both a diagnostic [6] and therapeutic alteration for novel targeted inhibitors [7]. To further increase the diagnostic accuracy and facilitate diagnosis of salivary gland neoplasms, we here present a custom-designed comprehensive next generation RNA sequencing panel. This approach covers most of the common molecular alterations of salivary gland neoplasms in one test and can be reliably performed on FFPE tissue including biopsy or cell block specimen of FNAs. Therefore, the SalvGlandDx panel was recently implemented in our institutional diagnostic setting.

Material and methods

Panel design

The target region was designed according to the current knowledge of molecular alterations (fusions and mutations) of salivary gland including odontogenic neoplasms (Table 1) and covers known hotspot mutations and fusion genes. Furthermore, NR4A3 was included in the panel to measure mRNA expression levels. Moreover, NTRK1/2/3 fusions have recently gained high importance as therapeutic targets. Therefore, these genes were included as well. The assay was designed using the Archer Assay Designer software.

Table 1

Molecular alterations in salivary gland and odontogenic neoplasms according to current literature.

Selected Molecular Aberrations Covered by SalvGlandDx Panel	Entity
Gene fusions
PLAG1* fusions CTNNB1-PLAG1 LIFR-PLAG1 CHCHD7-PLAG1 FGFR1-PLAG1 ND4-PLAG1 NFIB-PLAG1 TGFBR3-PLAG1	PA, CA ex PA, MECA [8, 9] PA, MECA [8, 9] PA, MECA [8, 9] PA,CA ex PA, MECA 9, 10, 11 MECA ex PA [10] PA [12] MECA, MECA ex PA [10]
HMGA2* fusions HMGA2-WIF1 HMGA2-NFIB HMGA2-TMTC2	PA, CA ex PA [13] PA [12] PA (metastasizing) [14]
MYB* and MYBL1* fusions MYB-NFIB MYBL1-NFIB	Adenoid-cystic carcinoma [15, 16] Adenoid-cystic carcinoma [15, 16]
MAML2* fusions MAML2-CRTC1 MAML2-CRTC3	Mucoepidermoid carcinoma [17] Mucoepidermoid carcinoma [18]
PRKD1*-, PRKD2*-, PRKD3*- fusions ARID1A-PRKD1 DDX3X-PRKD1 PRKAR2A-PRKD1 SNX9-PRKD1 ATL2-PRKD3	Polymorphous adenocarcinoma (including cribriform variant / CASG) 19, 20, 21 PAC, cribriform variant / CASG [20] PAC, cribriform variant / CASG [20] PAC, cribriform variant / CASG, current study PAC, cribriform variant / CASG, current study PAC, classical variant, current study
SS18* fusions SS18-MEF2C SS18-ZBTB7A	Microsecretory adenocarcinoma (incl. potential spectrum) [22, 22] and current study
MSANTD3* fusions HTN3-MSANTD3	Acinic cell carcinoma [23]
RET* fusions TRIM27-RET NCOA4-RET TRIM33-RET VIM-RET	Intraductal carcinoma [24] Intraductal carcinoma [24] Intraductal carcinoma [25] Secretory carcinoma [26]
ETV6* fusions ETV6-NTRK3 ETV6-RET ETV6-MET	Secretory carcinoma [6] Secretory carcinoma [27] Secretory carcinoma [28]
EWSR1* fusions EWSR1-ATF1 EWSR1-CREB1 EWSR1-CREM EWSR1-FLI1	Clear cell (odontogenic) carcinoma [29] Clear cell (odontogenic) carcinoma [30] Clear cell (odontogenic) carcinoma [31] Adamantinoma-like Ewing family tumor / sarcoma [32]
NUTM1* fusions NUTM1-BRD4 NUTM1-BRD3 NUTM1-NSD3	NUT carcinoma [33] NUT carcinoma [34] NUT carcinoma [35]
NTRK1*-, NTRK2*-, NTRK3* fusions
ETV6-NTRK3	Secretory carcinoma [6]
NTRK2-ZCCHC7	Adenocarcinoma NOS, current study
Gene mutations
AKT1 p.E17	Epithelial-myoepithelial carcinoma [4] Intraductal papillary mucinous neoplasm [36] Salivary mucinous adenocarcinoma [37]
CTNNB1 p.I35, p.Q28	Basal cell adenoma [38, 39]
SMO p.W535, p.L412	Ameloblastoma [40]
KRAS p.G12, p.G13, p.Q61	Adenomatoid odontogenic tumor [41], mucoepidermoid carcinoma [42]
PRKD1 p.E710	Polymorphous adenocarcinoma [19] (including cribriform variant) [43]
BRAF p.V600	Intraductal carcinoma [25], sialadenoma papilliferum [44], ameloblastoma [40]
HRAS p.G12, p.G13, p.Q61	Epithelial-myoepithelial carcinoma [4], intraductal carcinoma [45], salivary duct carcinoma [46], Sialadenoma papilliferum [47]
PIK3CA p.E542, p.E545, p.H1047	Salivary duct carcinoma [48], epithelial-myoepithelial carcinoma [4], sclerosing polycystic adenosis / adenoma [49], adenoid-cystic carcinoma (current study)
NRAS p.G12, p.G13, p.Q61	Salivary duct carcinoma [46]
Gene expression
NR4A3 upregulation (due to enhancer hijacking)	Acinic cell carcinoma [5, 50]
*For each covered gene, also rare or novel fusion partners can be detected

PA = Pleomorphic adenoma, CA ex PA = Carcinoma ex pleomorphic adenoma, MECA = Myoepithelial carcinoma, PAC = Polymorphous adenocarcinoma, CASG = cribriform adenocarcinoma of salivary gland

Nucleic acid isolation from FFPE specimen

Surplus FFPE material from surgical or cytologic specimen was used for nucleic acid isolation. The area of interest was marked by an experienced head and neck pathologist (NJR) on a representative H&E slide. Three 0.4 mm2 punch biopsies were then taken from the region of interest of the FFPE block. For FNA cell block specimen, four cuts of ten micrometers were cut and pooled. RNA was isolated using the Maxwell 16 LEV RNA FFPE Purification Kit and DNA was isolated using the Maxwell 16 FFPE Tissue LEV DNA Purification Kit (both Promega, Madison, WI, USA) according to the manufacturer's manual. Quantification was done using the Qubit fluorometric assay (ThermoFisher Scientific, Waltham, MA, USA).

Library preparation and next generation sequencing (NGS)

Library preparation for the SalvGlandDx panel and the Archer FusionPlexSarcoma panel was performed according to the protocol of the Archer FusionPlex technology (ArcherDX, Boulder, CO, USA), using 250 ng input RNA. Libraries were sequenced with 150bp paired-end on a NextSeq550. Library preparation for the custom-extended Archer FusionPlexSarcoma panel was performed according to the manufacturer's protocol (ArcherDX, Boulder, CO, USA), using 250 ng input RNA. Libraries were sequenced on the Ion S5 system (ThermoFisher Scientific, Waltham, MA, USA).

Library preparation for the Oncomine Focus Assay and Oncomine Comprehensive Assay v3 was performed according to the manufacturer's manual (ThermoFisher Scientific, Waltham, MA, USA), using 10 ng or 20 ng of input material. Libraries were templated on Ion 540 chips using the Ion Chef Instrument and sequenced on the Ion S5 system (ThermoFisher Scientific, Waltham, MA, USA).

Fluorescence in situ hybridization (FISH)

Two μm thick sections were incubated with dual color break apart (bap) FISH probes for EWSR1 (Abbott Molecular Vysis LSI EWSR1 (22q12) bap, Abbott Park, IL, USA), ETV6 (Abbott Molecular Vysis ETV6 bap, Abbott Park, IL, USA), MAML2 (bap Zytovision, Bremerhaven, Germany), MYB (bap Zytovision, Bremerhaven, Germany), MYBL1 (bap Empire Genomics, Williamsville, NY, USA), NTRK2 (bap Zytovision, Bremerhaven, Germany) and SS18 (bap Abbott molecular, Abbott Park, IL, USA) according to the manufacturer's protocol. 50 non-overlapping nuclei were analyzed using a fluorescence microscope (Zeiss Axioskop) with a 100-fold magnification oil objective. Pictures with Z-stacks of 20 images with 0.5 μm step distance were performed. Positive FISH was defined as at least 15% cells with break-apart and /or split signals.

Sanger sequencing (PRKD1)

DNA was amplified using custom primers covering the PRKD1 p.E710 hotspot. The resulting products were purified using the Qiagen MiniElute PCR Purification Kit (Qiagen, Hilden, Germany) and sequenced as described previously [51] using the following primers:

Primer PRKD1_F: 5′-TGGAAGGTGACAAGATGCTACA-3′

Primer PRKD1_R: 5′-ACTGGAATGAAATTTTGGTGATTCT-3′

Immunohistochemistry

Immunohistochemical stainings were performed as described previously [52], using the monoclonal primary anti-NUT antibody (clone C52B1, 1:200, Cell Signaling Technology, Danvers, MA, USA) and the monoclonal anti-NR4A3 (NOR-1) antibody (clone H-7, 1:25, Santa Cruz Biotechnology, Inc, Dallas, TX, USA), applying the automated Leica Bond III staining system (Leica Biosystems, Nussloch, Germany). The Ventana Benchmark staining system (Oro Valley, AZ, USA) was applied for the following primary antibodies: p63 (clone 4A4, prediluted, Ventana, Oro Valley, AZ, USA), p40 (clone BC24, 1:100, Zytomed Systems, Berlin, Germany), panTRK (clone EPR17341, 1:100, Abcam, Cambridge, UK), CD117 (clone YR145, 1:200, Cell Marque Lifescreen Ltd., Rocklin, CA, USA), Ki-67 (clone 30-9, prediluted, Ventana, Oro Valley, AZ, USA), S100 (polyclonal, 1:2000, DAKO A/S, Jena, Germany), DOG-1 (clone SP31, 1:50 Invitrogen, Carlsbad, CA, USA).

Data analysis

For the SalvGlandDx panel, the sequencing output bcl files were converted to fastq format and analyzed by the Archer analysis software (ArcherDX, Boulder, CO, USA). Fusion calls were considered positive when they had at least three unique reads, five unique start sites and 10% reads covering the breakpoint. Mutation calls were considered positive when they had at least five reads covering the mutant allele, three unique start sites and at least 5% mutant allele frequency. For expression analysis, the Archer analysis software normalizes the expression of the panel genes to the average expression of four control genes. Relative gene expression is then visualized in a heat map and displayed as numerical values.

For analysis of the Oncomine Assays, the IonReporter software (ThermoFisher Scientific, Waltham, MA, USA) was used with the generic filter settings.

Sanger sequencing results were analyzed using visual inspection of the DNA sequence.

Validation cohort (Table 2)

The cohort consisted of 32 FFPE samples (including three cell block specimen), mostly from salivary gland tumors. If salivary gland tumors with a certain alteration were not available, tests were applied to material from tumors with similar alterations. Moreover, the Archer SureShot sequencing controls (positive and negative control for ALK-RET-ROS1 fusions, ArcherDX, Boulder, CO, USA), were included.

Statistical analysis

Student's t-test was performed using GraphPad prism 8.0 to compare the NR4A3 normalized expression values between the different groups. A P value < 0.05 was considered statistically significant. Receiver operating characteristic (ROC) curve analysis was performed using IBM SPSS Statistics 26.

Ethics statement

All patients, whose material was collected after 2015 signed a written informed consent for further use of their tissue and data. The current project is covered by approval from the local ethics review board (Kantonale Ethikkommision Zürich (BASEC-Nr. 2020-01663)), including a waiver for further material use collected prior to 2016.

Results

Panel validation

To validate our custom designed SalvGlandDx panel, we used a cohort of 34 samples, including two sequencing controls (Table 2) with alterations previously detected by current standard techniques for the different alterations (FISH, Sanger sequencing, immunohistochemistry, NGS). All samples passed the QC criteria set by the Archer analysis software.

Table 2

Validation cohort.

Sample	Entity
SG-4	Adenocarcinoma NOS / potential spectrum microsecretory adenocarcinoma
SG-14	Polymorphous adenocarcinoma, cribriform variant / CASG
SG-17	Adenoid-cystic carcinoma, high-grade transformation
SG-24	Polymorphous adenocarcinoma, classical variant
SG-25	Clear cell carcinoma
SG-26	Polymorphous adenocarcinoma, classical variant
SG-29	Acinic cell carcinoma
SG-32	Adenoid-cystic carcinoma
SG-35	Epithelial-myoepithelial carcinoma, solid-oncocytic variant (published before) [53]
SG-36	Acinic cell carcinoma
SG-37	Acinic cell carcinoma
SG-38	Acinic cell carcinoma
SG-39	Acinic cell carcinoma
SG-41	Acinic cell carcinoma
SG-42	Secretory carcinoma
SG-44	Pleomorphic adenoma
SG-46	Adenoid-cystic carcinoma
SG-49	Adenocarcinoma NOS
SG-51	Mucoepidermoid carcinoma, oncocytic variant (OMEC)
SG-66	Carcinoma ex pleomorphic adenoma, minimal invasive
SG-68	Acinic cell carcinoma, high-grade transformation
SG-79	Epithelial-myoepithelial carcinoma, high-grade transformation
SG-84	Basal cell adenoma
SG-85	Acinic cell carcinoma, high-grade transformation
VD-1	Melanoma
VD-2	Melanoma
VD-3	Lung adenocarcinoma
VD-4	Lung adenocarcinoma
VD-5	NUT carcinoma
VD-6	NUT carcinoma
VD-7	Basal cell carcinoma of the skin
VD-8	Lipofibromatosis-like neural tumor
NEG	Archer SureShot negative control
POS	Archer SureShot positive control

SG = Salivary gland tumor, CASG = Cribriform adenocarcinoma of salivary gland, VD = Validation sample, NEG = negative control, POS = positive control

Fusion analysis

Tumors with known fusions were available from 14 patients. Known fusions originate from FISH in five cases, NGS in five cases (Oncomine Focus or Comprehensive Assay v3, or Archer FusionPlexSarcoma panel), FISH plus NGS in one case and FISH plus IHC in one case. One case with NTRK1 fusion and one case with a NUTM1 fusion was confirmed by immunohistochemistry. Furthermore, the SureShot sequencing controls (positive and negative control for ALK-RET-ROS1 fusions) were used.

All known fusions were detected using the SalvGlandDx panel (Table 3). Furthermore, fusion partners from genes shown to be rearranged in FISH could be determined.

Table 3

Validation of known fusions.

Sample	Standard Diagnostic Method	Alteration Standard Diagnostic Method	Alteration SalvGlandDx
SG-4	FISH (SS18)	SS18 rearrangement	SS18-ZBTB7A
SG-17	Amplicon sequencing (OCA)	MYB-NFIB	MYB-NFIB
SG-25	FISH (EWSR1)	EWSR1 rearrangement	EWSR1-ATF1
SG-32	FISH (MYBL1)	MYBL1 rearrangement	MYBL1-NFIB
SG-42	IHC (panTRK), FISH (ETV6)	panTRK+ETV6 rearrangement	ETV6-NTRK3
SG-44	Amplicon sequencing (OCA)	FGFR1-PLAG1	FGFR1-PLAG1
SG-46	FISH (MYB)	MYB rearrangement	MYB-NFIB
SG-49	Customized Archer® FusionPlex Sarcoma panel FISH (NTRK2)	ZCCHC7-NTRK2NTRK2 rearrangement	ZCCHC7-NTRK2
SG-51	FISH (MAML2)	MAML2 rearrangement	CRTC3-MAML2
SG-66	Archer® FusionPlexSarcoma panel	HMGA2-WIF1	HMGA2-WIF1
VD-4	Amplicon sequencing (OFA)	KIAA1468-RET	KIAA1468-RET
VD-5	Amplicon sequencing (OCA)	WHSC1L1(NSD3)-NUTM1	NSD3-NUTM1
VD-6	IHC (NUT)	NUT+	BRD4-NUTM1
VD-8	IHC (panTRK)	panTRK+	TPM3-NTRK1
NEG	Sequencing Control	-	-
POS	Sequencing Control	CCDC6-RET	CCDC6-RET

SG = Salivary gland neoplasm, VD = Validation sample, NEG = negative control

POS = positive control, OFA = Oncomine Focus Assay, OCA = Oncomine Comprehensive Assay, IHC = Immunohistochemistry

Mutation analysis

To validate the SalvGlandDx panel for mutation analysis, we used eleven tumors with known mutations, previously detected by Sanger sequencing (3 cases) and amplicon-based DNA sequencing with Oncomine panels (8 cases). Furthermore, the SureShot sequencing controls (positive and negative control) were included, as their backbone cell line HTC-116 has known mutations in CTNNB1, KRAS and PIK3CA that are covered by the SalvGlandDx target region. With the SalvGlandDx panel, we were able to detect all mutations (Table 4).

Table 4

Validation of known mutations

Sample	Standard Diagnostic Method	Alteration STANDARD DIAGNOSTIC METHOD	Alteration SalvGlandDx
SG-14	Sanger sequencing	PRKD1 p.E710D	PRKD1 p.E710D
SG-17	Amplicon sequencing (OCA)	PIK3CA p.E545K	PIK3CA p.E545K
SG-24	Sanger sequencing	PRKD1 p.E710D	PRKD1 p.E710D
SG-26	Sanger sequencing	PRKD1 p.E710D	PRKD1 p.E710D
SG-35	Amplicon sequencing (OCA)	HRAS p.Q61R	HRAS p.Q61R
SG-79	Amplicon sequencing (OFA)	AKT1 p.E17K	AKT1 p.E17K
SG-84	Amplicon sequencing (OFA)	CTNNB1 p.I35T	CTNNB1 p.I35T
VD-1	Amplicon sequencing (OFA)	BRAF p.V600E	BRAF p.V600E
VD-2	Amplicon sequencing (OFA)	NRAS p.Q61K	NRAS p.Q61K
VD-3	Amplicon sequencing (OFA)	KRAS p.G12C	KRAS p.G12C
VD-7	Amplicon sequencing (OCA)	SMO p.L412F	SMO p.L412F
NEG	Sequencing Control	CTNNB1 p.S45del,KRAS p.G13D,PIK3CA p.H1047R	CTNNB1 p.S45del,KRAS p.G13D,PIK3CA. p.H1047R
POS	Sequencing Control	CTNNB1 p.S45del,KRAS p.G13D,PIK3CA p.H1047R	CTNNB1 p.S45del,KRAS p.G13D,PIK3CA p.H1047R

SG = Salivary gland neoplasm; VD = Validation sample; NEG = negative control; POS = positive control; OFA = Oncomine™ Focus Assay; OCA = Oncomine™ Comprehensive Assay

Recurring variants in PIK3CA (p.S514Kfs*2) and PRKD3 (p.D304=) that were detected in the majority of all samples were flagged as artifacts. Moreover, several variants in EWSR1, one variant in MYBL1 and one variant in MAML2 occurred in more than one sample, however not in the majority of samples. All of them were variants of unknown significance and were therefore not further investigated.

Gene expression analysis

Eight acinic cell carcinomas (six classical and two with high-grade transformation) and one secretory carcinoma (negative control) were stained with the NR4A3 antibody. All eight acinic cell carcinomas were positive (Fig. 1A), while the secretory carcinoma showed no positive staining (Fig. 1B). Likewise, the visual as well as the numerical normalized expression values (NEV) for all eight cases were significantly increased, while the values for the negative control were low. Moreover, we compared the normalized expression values of all 25 salivary gland neoplasms of the validation cohort and found a clear difference between positive (mean: NEV = 5.72 ± 1.48) and negative (mean: NEV = 0.1 ± 0.14) samples (P < 0.0001, t-test, Fig. 1C). A ROC curve analysis yielded an area under the curve (AUC) = 1 with a sensitivity and specificity of 100% at an optimal discriminating cutoff of NEV = 2.32. Furthermore, visual inspection of the heatmap clearly distinguished positive and negative cases (Fig. 1D).

Fig. 1

High NR4A3 mRNA expression levels are specific to acinic cell carcinoma (ACC). (A) shows a conventional acinic cell carcinoma with blue zymogen granules and strong nuclear expression of NR4A3 immunohistochemistry (inset). (B) illustrates secretory carcinoma as most important differential diagnosis, negative for nuclear NR4A3 (inset). The normalized expression values (NEV) in (C) depict a significant difference between the n=8 cases of acinic cell carcinomas (mean NEV = 5.72 ± 1.48) and the n = 17 remaining cases (mean NEV = 0.1 ± 0.14; P < 0.0001, t-test). The blue dotted line represents the optimal cutoff at NEV = 2.32, yielding an AUC = 1 with 100% sensitivity and specificity in a ROC curve analysis. (D) depicts representative heatmap of one NR4A3 positive case among negative cases. Scale bar 100 µm.

Three samples from our validation cohort originated from cytological specimen, showing that the panel is suitable to process not only tissue biopsies but also material from cell block specimen of fine needle aspirations (SG-51: CRTC3-MAML2 fusion, SG-84: CTNNB1 mutation, SG-85: NR4A3 expression)

Practical application of SalvGlandDx panel on four unique cases

Case 1: Purely oncocytic mucoepidermoid carcinoma (OMEC)

The first case was a 33-years old male with a lump in the right parotid gland. Fine needle aspiration in our cytopathology department revealed a bland oncocytic neoplasm, rendering a broad differential diagnosis (Fig. 2A, inset). The SalvGlandDx panel was performed on the cellblock specimen and detected a rare CRTC3-MAML2 fusion, highly specific to mucoepidermoid carcinoma. A MAML2 FISH showed >15% split signals corroborating the detected MAML2 rearrangement (Fig. 2A, inset). After signing out the final diagnosis of OMEC, the preoperative consenting and surgical plan was adapted. The following resection specimen showed an infiltrative, purely oncocytic neoplasm with intersecting broad fibrous bands and focal eosinophilic secretions (Fig. 2A-B). Only miniscule areas of glandular differentiation with likewise intraluminal eosinophilic secretions could be observed (Fig. 2C), while mucocytes were completely absent (Fig. 2D). Furthermore, tumor manifestations were visible in close vicinity to peripheral nerve branches, however, without obvious perineural invasion. Immunhistochemical expression of p63 was positive in a subset of cells (Fig. 2E).

Fig. 2

Case 1:CRTC3-MAML2 fused oncocytic mucoepidermoid carcinoma (OMEC). (A) shows bland oncocytic cells in the fine needle aspiration cellblock specimen (upper inset). MAML2 FISH split signals can be appreciated in the corresponding inset. Resection specimen (below) depicts an infiltrating purely oncocytic neoplasia, intersecting the distinct fibrous bands. In (B) a magnification of the bland solid-oncocytic cell complexes is visualized with focal eosinophilic secretions (arrowheads). (C) shows very focal glandular differentiation with intraluminal secretions, however lacking the classical triphasic differentiation. In (D) the lack of typical alcianblue-positive mucocytes is illustrated, whereby the distinct spheroid violet reactivity was spatially correlating to the secretions (arrowheads). Immuohistochemical p63 expression was noted in a subset of cells (E). Scale bar 2.5 mm (overview), 100 µm (magnified insets).

Case 2: Adenocarcinoma NOS with NTRK2 fusion and multiphasic growth pattern

The second case was a 65 years-old male with a recurrent and infiltrating left parapharyngeal salivary gland tumor of exceptional morphology (Fig. 3A). The lobulated cellular mass showed dense intersecting and circumferential fibrous bands. The tumor cells were quite monomorphic and of one type, medium to large in size with clear chromatin, distinct nucleoli, eosinophilic and partially prominent clear cytoplasm with somewhat oncocytic aspect. Growth pattern encompassed a multiphasic microcystic, solid and cribriform morphology, with striking slate-grey extracellular deposits and stroma (Fig. 3B). Focal regions were reminiscent of epithelial-myoepithelial carcinoma with clearing of the more abluminal cells (arrowhead; Fig. 3C); however, no true biphasic cell type differentiation could be noted by immunohistochemistry. Furthermore, a segmental morphology with dissecting epithelioid complexes into the described background was observed (Fig. 3D). Another growth pattern showed marked pseudorosettes (arrowheads; Fig. 3E). Occasional mitotic figures were present. On immunohistochemistry SOX10 and S100 were diffusely expressed (Fig. 3F), corroborating a salivary gland origin. Mammaglobin, CD117, NR4A3, DOG1, p63, p40, PLAG1 and androgen receptor were negative (data not shown). GATA3 showed very weak single cell reactivity. Pan-cytokeratin, and CK7 were diffusely positive. Interestingly, Ki-67 showed a distinct pattern with confluent garland-like peripheral proliferation fronts, leaving an impression of a punched-out proliferation-negative geographical center (supplementary Fig. 1). PanTRK immunohistochemistry illustrated a very faint cytoplasmic and membranous reactivity, whereas Alcianblue-PAS staining depicted mucin depositions accentuated around the vascular spaces of the pseudorosettes (Fig. 3F). SalvGlandDx panel detected a ZCCHC7-NTRK2 fusion, which has only been mentioned in an abstract, however without corresponding neoplasm type [54]. NTRK2 FISH confirmed this finding (Fig. 3A, inset). Although features were reminiscent of pleomorphic adenoma, a true chondro-myxoid matrix could not be detected. Furthermore, no change in cell type or size was observed.

Fig. 3

Case 2: ZCCHC7-NTRK2 fused adenocarcinoma NOS with multiphasic growth pattern. In (A) a parapharyngeal lobulated, cellular neoplasm is shown infiltrating into adjacent skeletal muscle. NTRK2 FISH split signal can be observed (inset). Growth pattern encompassed microcystic (B), solid, epithel-myoepithelial carcinoma-like (arrowhead; C) and dissection of epithelioid cells in the distinct slate-gray background (D) pattern. In (E), tumoral pseudorosettes around vascular spaces are seen (arrowheads). Relevant immunohistochemical stains and accentuated alcianblue-PAS reactivity around the vascular spaces of the pseudorosettes is depicted in (F). Scale bar 2.5 mm (overview), 100 µm (magnified insets).

Case 3: Adenoid-cystic carcinoma with spindle cell and pseudoangiomatoid pattern

The third case describes an adenoid-cystic carcinoma with very uncommon morphology. This tumor was excised due to a submucosal mass on the left cheek of a 55 years-old male patient. Histology showed a quite well circumscribed cellular neoplasm (Fig. 4A) with mixed morphology. The main part of the neoplasm consisted of bland spindle cells embedded in a chondro-myxoid like stroma intermingled with mature fat (Fig. 4B). Furthermore, a pseudoangiomatoid pattern could be noticed, besides myxoid nodules with fibrosis (Fig. 4C-D). Limited foci showed an adenoid-cystic like morphology (Fig. 4E) and focal neural invasion and infiltration of the peripheral soft tissue. Immunohistochemistry for p40 was diffusely positive without noticeable biphasic differentiation, reminiscent of a myoepithelial overgrowth; whereas CD117 was virtually absent. Ki-67 proliferation index was very low (<5%; Fig. 4F). Although the tumor was resembling pleomorphic adenoma, the infiltrative pattern exceeded the classical observed features and rendered the differential diagnosis of carcinoma ex pleomorphic adenoma or unusual de novo adenoid-cystic carcinoma. Panel sequencing revealed a MYBL1-NFIB fusion, which was confirmed by corresponding split signals in MYBL1 FISH (Fig. 3A, inset). No rearrangement of PLAG1 or HMGA2 was detected. Final diagnosis was adenoid-cystic carcinoma with spindle cell and pseudoangiomatoid differentiation, mimicking pleomorphic adenoma.

Fig. 4

Case 3: MYBL1-NFIB fused adenoid-cystic carcinoma with largely unusual morphology. (A) shows a quite well circumscribed cellular neoplasm. MYBL1 FISH shows split signals (inset). In (B) the bland spindle cells mixed with mature (probably infiltrated) fat and the prominent myxoid background is depicted. The pseudoangiomatoid pattern is represented in (C) with capillary-like structures. The fibro-myxoid nodules can be appreciated in (D) and (E) shows very limited foci of more conventional cribriform morphology. The diffuse p40 expression with monophasic pattern is visualized in (F), with CD117 negativity and very low Ki-67 proliferation index. Scale bar 2.5 mm (overview), 100 µm (magnified insets).

Case 4: SS18-ZBTB7A rearranged low-grade adenocarcinoma

This case derives from the parotid gland of a 60 years-old male, who underwent surgery due to a squamous cell carcinoma in the parotid gland. Besides, the resection specimen revealed a 7 mm diffusely infiltrating adenocarcinoma (Fig. 5A) with bland solid, cribriform, tubular and trabecular epithelial complexes (Fig. 5B-D). Focal oncocytic differentiation was observed, including rhabdoid- / histiocytoid-like single cells dissecting into the stroma (Fig. 5E). A prominent sclerotic stroma was evident in the background, while also distinct intraluminal bubbly secretions were noted (Fig. 5E). Immunohistochemistry showed diffuse expression of SOX10, while DOG-1 showed an apical-membranous pattern, indicative of an intercalated duct differentiation. S100 was diffusely expressed in a heterogeneous pattern. A p40 staining was largely negative, including very limited areas of a biphasic pattern with single positive abluminal cells (Fig. 5F). SalvGlandDx panel showed a SS18-ZBTB7A fusion, corroborated by positive SS18 FISH (Fig. 5A, inset). The morphology and molecular profile matched with the single other published case, which has been reported as potential spectrum of the recently described microsecretory adenocarcinoma [22].

Fig. 5

Case 4: SS18-ZBTB7A fused low-grade adenocarcinoma. (A) shows diffusely infiltrating adenocarcinoma with split signals in SS18 FISH (inset). The magnification in (B) illustrates the solid growth of medium-sized atypical cells including subtle nucleoli and slightly oncocytic differentiation. In (C) the cribriform ductal differentiation is depicted, whereas (D) pictures the small irregular tubular complexes within a sclerotic stroma, reminiscent of tubular breast cancer. (E) outlines the more oncocytic single cells with a histiocytoid / rhabdoid differentiation (left arrowhead). The right arrowhead points to the intraglandular bubbly secretions. The relevant immunhistochemical stainings are shown in (F), whereas the lower p40 staining describes the focal biphasic differentiation in contrast to the negative upper staining representative for most of the tumor. Scale bar 2.5 mm (overview), 100 µm (magnified insets).

Identification of novel fusions involving the PRKD genes

To identify fusion partners of the investigated PRKD genes, six cases of polymorphous adenocarcinoma including cribriform variant / cribriform adenocarcinoma of salivary gland (CASG) were identified from our archive including consult cases and subjected to sequencing with the SalvGlandDx panel. Clinicopathological overview is depicted in Table 5.

Table 5

Clinicopathological features of investigated polymorphous adenocarcinomas (PAC) including cribriform variant / cribriform adenocarcinoma of salivary gland (CASG).

Case	1	2	3	4	5	6
Age	33	54	74	34	84	67
Sex	f	m	f	m	f	m
Localization	Buccal	Hard / soft palate	Buccal, maxillary sinus	Soft palate	Parotid gland	Soft palate
Size	1.3 cm	0.9 cm	5.5 cm	1.1 cm	2.4 cm	0.5 cm
Morphology	PAC, classical variant	PAC, classical variant	PAC, cribriform variant / CASG with basaloid features	PAC, cribriform variant / CASG	PAC, cribriform variant / CASG	PAC, classical variant
Molecular aberration	PRKD1 p.E710D	PRKD1 p.E710D	PRKD1 p.E710D	PRKAR2A-PRKD1	SNX9-PRKD1	ATL2-PRKD3
Perineural invasion	yes	yes	yes	yes	yes	yes
Metastases	no	no	no	yes, Neck, Level II-III, 14 years after primary	no	no
Therapy	Resection	Resection	Resection, adjuvant radiotherapy	Resection, adjuvant radiotherapy	Resection, adjuvant radiotherapy	Resection
Follow-up	NED, 74 months	NED, 39 months	NED, 12 months	NED, 243 months	NED, 16 months	NED, 3 months

NED = no evidence of disease

Three cases (50%) with an origin in the oral mucosa showed an activating PRKD1 p.E710D mutation, of which two cases showed a classical polymorphous (low-grade) morphology (Fig. 6A). One case was a large polypoid tumor of 5.5 cm in the maxillary sinus and showed largely solid differentiation including prominent nuclear clearing, compatible with cribriform variant /CASG. Of note, besides focal typical glomeruloid growth, this case showed an uncommon basaloid morphology with extracellular basal membrane-like material reminiscent of basal cell adenocarcinoma (Fig. 6B, inset). The other three cases showed PRKD gene fusions. One case showed a PRKAR2A-PRKD1 fusion and was metastasized to the neck lymph nodes 14 years after the primary tumor was resected from the soft palate. Morphology encompassed typical features of cribriform variant / CASG with solid, glomeruloid and focal papillary growth (case not shown). Another case harbored a SNX9-PRKD1 rearrangement and likewise represented a typical solid, glomeruloid and focal papillary architecture with calcifications as described in cribriform variant / CASG [55] (Fig. 6C). Remarkably, this tumor was located in the parotid gland without clinical evidence for any other manifestation. The remaining case showed an ATL2-PRKD3 fusion, however, was depicting a classical monophasic low-grade appearance similar to the first two cases with focal perineural invasion in the oral mucosa (Fig. 6D).

Fig. 6

Morphological comparison of PRKD1 p.E710D mutated and PRKD1/PRKD3 rearranged polymorphous adenocarcinoma (PAC) / cribriform adenocarcinoma of salivary gland (CASG). (A) shows classical monomorphic tubular and trabecular PAC complexes with diffuse infiltrative pattern including targetoid perineural invasion (arrowhead), a slate-grey stroma and an underlying PRKD1 p.E710D mutation. In (B) another case with PRKD1 p.E710D mutation derived from a maxillary sinus polypoid lesion is depicted, however illustrating a solid differentiation with prominent nuclear clearing reminiscent of papillary thyroid carcinoma, compatible with cribriform variant / CASG. In the insets focal perineural invasion (arrowhead) is shown and different growth patterns encompassing cribriform, glomeruloid, basaloid jigsaw-like and ductal-like. (C) illustrates the typical morphological spectrum of a (parotid gland) cribriform variant / CASG harboring a SNX9-PRKD1 fusion, including nuclear clearing of solid tumor complexes, perineural invasion (arrowhead), glomeruloid and papillary growth pattern as well as focal calcifications (insets). Of note, the stroma is sclerotic. (D) depicts a morphological similar case to (A), including targetoid perineural invasion (arrowhead) and slate-grey stroma, however harboring ATL2-PRKD3 fusion. Scale bar 100 µm.

Discussion

With the validation of the known alterations in our cohort, we were able to design a custom NGS panel that facilitates the diagnosis and classification of salivary gland neoplasms with one single test. Well-known fusions as well as hotspot mutations and elevated gene expression of NR4A3 could reliably be detected not only in histological FFPE tissue, but also in cell block specimen of fine needle aspirations. The latter in particular may allow a more precise diagnosis and alleviate the need for core needle biopsy and/or incisional biopsy. To our knowledge this is the first validated panel that includes all relevant genes to detect specific alterations, including mutations, fusions and gene expressions on RNA level. While usually only single alterations can be tested at the time (e.g. FISH, IHC, Sanger sequencing), some commercially available NGS panels cover certain genes, however, not all of them within one assay.

Besides the detection of relevant alterations for salivary gland neoplasm diagnosis and classification, our panel can reliably detect therapeutic targets, such as NTRK fusions. Here, NTRK inhibitors (e.g. larotrectinib, entrectinib) were recently shown to be effective in patients with NTRK fusions [56, 57]. According to the ESMO guidelines, testing with an RNA-based NGS assay is preferred for these alterations [58].

High expression levels of NR4A3 were specifically associated with acinic cell carcinoma in our study, as described by Haller et al. [5]. In particular on FNA specimen, the diagnosis of acinic cell carcinoma can be hampered by high similarity to normal acinic cells [59]. Moreover, cases with high-grade morphology might not show the typical features. Our two cases with high-grade transformation could correctly be assigned to acinic cell carcinoma by the panel results, corroborated by positive NR4A3 immunohistochemistry. One case could already be diagnosed on the FNA specimen guided by high NR4A3 expression levels detected by the SalvGlandDx panel. The (high-grade) morphology on the FNA specimen was not specific; however, the diagnosis was confirmed on the histological specimen showing partial obvious classical (low-grade) differentiation.

Case 1 showed an intriguing oncocytic mucoepidermoid carcinoma (OMEC), lacking the typical mucocytes and triphasic differentiation as recently described by Skalova et al. [60] and one similar case by Todorovic et al. [61]. The CRTC3-MAML2 fusion is highly specific to mucoepidermoid carcinoma [18], whereas the detection on the cellblock specimen combined with the oncocytic morphology was sufficient for this diagnosis. Current data suggest that (purely) oncocytic and sclerosing variant of MEC are most likely low-grade neoplasms [62, 63], which seem to show a more indolent biological behavior. However, one has to be aware that current grading schemes (e.g. AFIP and Brandwein-Gensler) might lead to a distinct appraisal [60]. Future data will be needed to clarify this issue and potentially grade these subentities separately.

The second case revealed a novel ZCCHC7-NTRK2 fusion, which could be verified independently in other laboratories using NGS panel and a break-apart FISH assay. That approach shows, that directed molecular testing can, albeit in rare instances, reveal targetable alterations. The uncommon mixed morphology was not assignable to a known entity. The immunohistochemical phenotype, encompassing SOX10 and S100 co-expression, corroborated the diagnosis of a salivary gland neoplasm. Although the tumor showed areas reminiscent of pleomorphic adenoma (PA), including the variety of growth patterns and distinct background, no true chondro-myxoid matrix or fibrous (pseudo)capsule could be observed. Furthermore, the cell type was overall too monomorphic, large and atypical in comparison to a conventional PA. The observed infiltrative growth pattern on the resection specimen, without a typical change of cell type or cell overgrowth rather represents a distinct clonal entity than a carcinoma ex pleomorphic adenoma in our view. Despite similarities to a myoepithelial-like immunophenotype [64], the presence of abundant ductal structures did not admit diagnosis of a (secretory) myoepithelial carcinoma [65, 66]. An uncommon variant of secretory carcinoma, involving commonly ETV6-NTRK3 fusions, could not be diagnosed due to a distinct morphology, molecular and immunohistochemical profile (e.g. mammaglobin negative) [6]. Altogether, the observed cell type shows similarities to striated ducts in normal salivary gland tissue, however the immunophenotypes cannot be matched [67]. Besides, the distinct background and single cell type showing different growth patterns could lead to the differential diagnosis of polymorphous adenocarcinoma; however, the comprehensive profile does not fit in this category [68]. NTRK2 fusions are extremely rare and even less common than NTRK1 and NTRK3 fusions. In two recently published clinical trials investigating the NTRK inhibitors entrectinib and larotrectinib in solid tumors, only one out of 54 and 55 participants, respectively, harbored a NTRK2 fusion without mentioning the tumor entity [56, 57]. We were not able to find a specific report of a NTRK2 fused salivary gland neoplasm in the literature at the current time. Morphology encompassed a single cell type with multiphasic and infiltrative growth pattern, whereas further cases have to elucidate a potential recurrent nature of this outstanding type of neoplasm.

The third case showed a very uncommon morphology, which could lead to a misinterpretation of pleomorphic adenoma. Parallel sequencing of the fusion genes with the described panel could thereby help to further corroborate a carcinoma ex pleomorphic adenoma, which typically harbors PLAG1 or HMGA2 fusions [69, 70]. MYBL1-NFIB rearrangements can be detected in up to 24% of adenoid-cystic carcinoma and are highly specific for this tumor entity [15, 16]. Overall, this fusion corroborates the diagnosis of adenoid-cystic carcinoma although the morphology was largely unusual. The molecular finding is in line with a recently published similar case, however obscuring its origin by illustrating squamous differentiation [61].

The fourth case was initially classified as adenocarcinoma NOS, as it did not fit to a known category. Sequencing revealed a SS18-ZBTB7A fusion, which has been recently described in a potential spectrum of microsecretory adenocarcinoma (MSA), typically harboring MEF2C-SS18 fusions and occurring in the oral cavity [22]. Interestingly, it was similarly small (< 1 cm) and located in the parotid gland likewise as the only case with this fusion described by Bishop et al. [22]. Our case showed the identical fusion (SS18 exon 10, ZBTB7A exon 2), a similar morphological and immunohistochemical profile (largely negative for p63 and p40). In addition, we describe a DOG1 expression, indicative of a potential intercalated duct differentiation [71]; however, no information about DOG1 expression in bona fide MSA is available. Due to an identical morphological, immunohistochemical and molecular profile, our case corroborates that this tumor type is a recurrent finding. However, further cases have to elucidate whether this is a genuine spectrum of MSA or a separate entity sharing some features with MSA.

PRKD1 mutations were specific to the spectrum of polymorphous adenocarcinoma (PAC). Most of the classical PAC did show a recurrent PRKD1 p.E710D mutation, whereas one (metastasized) cribriform variant / CASG did show a previously undescribed PRKAR2A-PRKD1 fusion. However, one case of a classical PAC did show an ATL2-PRKD3 fusion and one case of a non-metastasized cribriform variant / CASG with uncommon basaloid features harbored a PRKD1 p.E710D mutation, corroborating the findings from Sebastiao et al. that the type of aberration is not entirely specific for classical PAC or cribriform variant / CASG [43]. Moreover, the case harboring the SNX9-PRDK1 fusion was located in the parotid gland without evidence of other manifestations, underscoring the previous finding that this entity can also occur rarely in the major salivary glands [21]. Polymorphous adenocarcinoma is rarely observed in the sinonasal tract [72], whereas cribriform variant /CASG has to the best of our knowledge not been described to involve the maxillary sinus, yet. Morphological classification of these neoplasms has been difficult among a large multicenter study [21]. However, molecular analysis can help to assign a certain tumor with uncommon morphology or on a small biopsy /FNA to the described spectrum. Mutations in the other investigated genes are generally less specific. However, there are only exceedingly rare reports of a benign salivary gland neoplasm harboring an activating HRAS mutation (in this case: sialadenoma papilliferum) [47]. With a compatible morphology, HRAS mutation is strongly underscoring epithelial-myoepithelial carcinoma, even in uncommon variants [4].

We envision the SalvGlandDx panel as a useful all-in-one tool to diagnose and classify salivary gland neoplasms. With this panel, we can corroborate the findings from Todorovic et al., who showed the importance of next generation sequencing in salivary gland neoplasms from different aspects [61]. The great advantage is thereby the simultaneous capturing of aberrant expression, gene mutations and fusions with the only need of RNA extraction, also reliably working on FFPE cell block specimen. One drawback might be uncovered individual fusions, such as ALK rearrangements, recently described in a small subset of intraductal carcinoma [73]. However, due to the design of the Archer FusionPlex technology, the panel can easily be amended in the future to include relevant emerging genes; however, keeping it as small as possible to optimize utilization on the sequencer. Nevertheless, this assay can significantly help to proper classify most of the salivary gland neoplasms, as it includes most of the recurrent gene aberrations known to date. Furthermore, it is able to simultaneously reveal potential molecular therapeutic targets such as NTRK and RET fusions, and can be applied on cellblock specimen of minimal invasive fine needle aspiration.

Acknowledgments

The authors thank Susanne Dettwiler and Fabiola Prutek from the institutional biobank, the laboratory for in situ techniques, the IT support, Marion Bawohl and the laboratory for Sanger sequencing, all from the Department of Pathology and Molecular Pathology at the University Hospital Zurich for excellent technical assistance.