Background
The Prostate-specific membrane antigen (PSMA) is a transmembrane glycoprotein of the prostate secretory acinar epithelium that is upregulated in prostate cancer (PC) and known from its use in diagnostics and targeted therapy in metastatic PC [
1‐
4]. Besides tracer accumulation in prostate tissue, PSMA PET/CT depicts physiological uptake in the salivary and lacrimal glands, liver and kidneys, but also in benign and malignant neoplasms, mostly adenomas and (adeno) carcinomas, of glandular or epithelial origin [
5,
6].
In PC, increased intracellular PSMA expression by immunohistochemistry is related to increased pathological grade, and subsequently correlated with disease-related mortality [
1‐
4]. Malignancies other than PC also express PSMA but in endothelial cells of tumours’ neovasculature, which suggests PSMA involvement in tumour angiogenesis. In salivary glands PSMA was identified on the acinar cells in the epithelium [
3,
7‐
9].
Recently, PSMA PET/CT analysis in a series of patients with head and neck adenoid cystic carcinoma (AdCC) showed tracer uptake in areas of locoregional recurrent and distant metastatic AdCC, and expression was confirmed immunohistochemically [
10]. AdCC is the most common malignant secretory gland tumour in the head and neck region. Incidence peaks in the fifth and sixth decade and has a female predominance [
11‐
15]. AdCC originates from ductal (luminal) and basal/myoepithelial (abluminal) cells and typically arises in the major salivary glands, the minor salivary and seromucous glands of the lip and upper aerodigestive tract, but also in the lacrimal and ceruminous glands. The tumour is characterized by an indolent but persistent growth rate, frequent locoregional recurrence and a delayed silent onset of distant metastasis, mainly in the lungs [
11,
15‐
17]. Surgery is the primary treatment, frequently followed by adjuvant radiation therapy because of positive resection margins and typical perineural growth. Although radiotherapy has probably no benefit to survival, it is reported to improve local and regional control [
15,
16]. Disease-specific survival (DSS) is moderate, with five and 10 year survival rates of 68–78% and 54–65% respectively [
18,
19]. Survival is negatively affected by the occurrence of an irresectable locoregional recurrence, which is considered clinically more relevant than the occurrence of slowly growing -often pulmonary and osseous- distant metastases that develop in almost half of the patients within 5 years after diagnosis [
11,
16,
18]. Other negative prognostic factors are advanced tumour stage, inadequate resection margins, skull base involvement and a solid growth pattern on histopathology. Perineural invasion does not directly affect mortality, but is significantly correlated with metastatic disease [
11,
15]. Regular treatment options are limited in advanced recurrent or metastasized disease [
16]. Given the positive results of PSMA-targeted diagnostic and treatment modalities in PC, this study aimed to analyse PSMA expression in a large cohort of primary and corresponding recurrent and metastatic AdCC tissues of the head and neck [
20]. Secondly, we aimed to explore associations with patient- and tumour characteristics and outcome, analogous to PC.
Methods
Patient selection
All patients diagnosed with AdCC in the University Medical Center Utrecht and Netherlands Cancer Institute-Antoni van Leeuwenhoek Hospital between 1990 and 2017 were analysed in a retrospective cohort study. Patients were selected in case of a histology-proven primary AdCC in the head and neck region with available representative formaldehyde-fixed paraffin-embedded tissue blocks of the resection specimen. Tumour samples of available corresponding locoregional recurrences or distant metastases were collected. Patients with previous (non AdCC) salivary gland disease, radiotherapy to the head or neck, or incomplete data were excluded. All data and samples were handled according to the GDPR.
Clinical parameters and tumour characteristics
The following clinical parameters were retrieved from the medical files: patient’s gender, age at diagnosis, tumour site, treatment regimen, (time to) recurrence or metastasis, vital status (cause of death) and date of last follow-up until January 1st 2018. Two dedicated head and neck pathologists (S.W. and L.S.) re-examined all haematoxylin- and eosin-stained slides for the following parameters: type and diameter of the tumour, pathological T- and N-stage, histopathological growth pattern and associated grade according to the differentiation of Perzin et al. [
21], surgical resection margins and the presence of perineural, vascular and bone invasion.
PSMA immunohistochemistry
A Tissue Microarray (TMA) was used to assess PSMA expression. From each tumour, three central 0.6 mm tissue cylinders from vital tumour were incorporated and covered the different aspects of this tumour morphology. Tumour whole-slides were analysed when patient’s tissue was not incorporated in the microarray, as well as whole-slides of all available recurrent and distant tissues. Representative TMA or whole slide paraffin sections 4 μm thick were immunohistochemically stained using fully automated protocols on the Benchmark XT (Ventana Medical Systems, Tucson, AZ, USA), validated for diagnostic purposes. Incorporated as control tissues were prostate cancer, normal salivary gland and duct tissue. For the primary antibody, a mouse antihuman PSMA monoclonal antibody was used (3E6; DAKO, Carpinteria, CA) of the IgG1 isotype directed against the internal domain of the PSMA antigen (DAKO, cat. no. M3620, Carpinteria, CA, dilution 1/80). The tissue sections were deparaffinised with xylene and ethanol followed by Heat Induced Epitope Retrieval in Ventana Cell Conditioning 1 for 24 min and subsequently incubated with the primary antibody for 60 min. Antigen-antibody reactions were visualized using Ventana OptiViewTM Amplification kit, followed by Ventana OptiViewTM Universal DAB Detection Kit (Optiview HQ Linker 8 min, Optiview HRP Multimer 8 min, Optiview Amplifier H2O2/Amplifier 4 min, Optiview Amplifier Multimer 4 min). Finally the slides were counterstained with haematoxylin, dehydrated and mounted.
PSMA expression analysis
Blinded semiquantitative scoring of all selected primary, recurrent and distant AdCC tumour samples was done until consensus was reached by two head and neck pathologists and two researchers (S.W., L.S., T.K.N. and M.V.). Per tumour core or whole slide the localization of PSMA-positive tumour cells was noted, followed by scoring the percentage of positive tumour cells in increments of 5%. Total tumour PSMA expression of the arrayed cores was defined by the mean percentage of PSMA-positive tumour cells of the three tissue cores. A core was considered inadequate when it contained < 5% tumour tissue. In case of a mean PSMA expression below 10% or in case of more than 1 inadequate core on microarray, one representative tumour whole slide was subsequently stained and scored in order to exclude false-negative results.
Statistical evaluation
Continuous and ordinal variables were reported as medians with interquartile ranges (IQR), categorical variables were reported as the number of patients and percentages. Associations between PSMA expression and all clinical parameters and tumour characteristics, except when categorical with more than 3 categories, were estimated with Spearman-ρ correlation coefficient with corresponding
p-values. The Independent Samples Kruskal-Wallis test (KW) was used to compare expression distribution between patients’ primary tumours, which recurred or metastasized and those that did not. Subsequently, the median expression of corresponding primary, recurrent and distant samples was compared. A cut-off level for the prediction of overall survival (OS), DSS, locoregional recurrence-free survival (RFS), and metastatic-free survival (MFS) was determined by dichotomizing PSMA expression and plotting Receiver Operating Characteristic (ROC)-curves. Differences in baseline characteristics of the groups divided by dichotomization were compared using Pearson chi-square test with appropriate Bonferroni correction. Statistics were performed using SPSS Statistics (version 22.0, IBM Corp., Armonk, NY, USA) for Windows. Multivariate OS, DSS, RFS and MFS survival analyses were carried out to calculate Hazard ratios with 95% confidence interval (CI). A Cox-proportional Hazard regression model was created by using SAS software (version 9.4, SAS Institute Inc., Cary, NC, USA) for Windows. Firth’s correction was applied to reduce bias of maximum likelihood estimation, as it deals with the occurrence of monotone likelihood in small-sample studies with time-dependent effects [
22]. Discriminative ability of the model was assessed by computing Harrell’s C-statistic [
23]. A two-tailed
P-value < 0.05 was considered statistically significant for all analyses.
Discussion
This is the first large cohort study describing PSMA expression in primary, recurrent and distant metastatic AdCC of the head and neck. Positive expression was seen in 94% of the primary tumours, 80% of recurrent tumours and 90% of the distant metastases, with PSMA expressed in 31, 60 and 23% of the tumour cells respectively. Primary tumour expression could not indicate disease progression and could not estimate expression levels in a recurrence or metastasis, although a tendency of respectively increase and decrease was observed. PSMA expression was not correlated to pathological stage and grade.
This is in contrast to PC in which high PSMA expression is correlated with prostate-specific antigen (PSA) recurrence and other prognostic factors which negatively affect survival such as tumour grade, pathological stage and castration resistance [
4,
24]. Multiple studies have shown that PSMA activates AKT and MAPK pathways promoting proliferation and survival of cancer cells, which may lead to an aggressive biological and clinical behaviour [
25,
26]. However, the currently presented inverse correlation of low primary tumour PSMA expression ≤10% as independent predictor of shortened RFS (HR 3.0; 95% CI 1.1–8.5;
p = .04) has also been described in other cancer types and might partly be explained by epigenetic silencing of the PSMA gene upon tumour progression [
27].
Analogous to PC, AdCC demonstrates expression of PSMA in the epithelial tumour cells, while expression in different other tumours mainly concentrates in endothelial cells of tumour-associated neovasculature [
3,
6,
28].
Of all primary PCs, 95% show heterogeneous PSMA expression with on average 53 ± 32% (mean ± SD) positive tumour cells. Mean expression in regional lymph nodes and distant metastasis is more extensive (72 ± 36% and 92 ± 10% respectively). Normal prostate tissue shows high PSMA expression in 100% of the samples (77 ± 32% positive cells), but with significantly less staining intensity than tumour tissue [
3,
4].
In contrast, staining intensity in AdCC is relatively constant. Although it is known that major and minor salivary glands depict high tracer uptake on PSMA PET/CT, a comparison between expression intensity in normal salivary gland tissue and AdCC tumour tissue could not be made due to the lack of data on PSMA expression in non-pathologic salivary glands [
29]. Comparing published PC data and our AdCC data, the present study concludes that similar to PC, 94% of primary AdCC expresses PSMA, but AdCC expression is more homogenous in a lower percentage of positive tumour cells.
Some points need to be addressed. Multiple Spearman-ρ correlation analyses and KW nonparametric tests were carried out to analyse possible differences between a large amount of clinical parameters and tumour characteristics. False-positive findings might have been introduced, as adjustments to correct for multiple comparisons are not desirable in explorative studies and were therefore not applied. The results of these analyses should therefore be interpreted carefully [
30].
By comparing the dichotomized groups of PSMA expression, it is noticed that a relatively large number of tumours in the low ≤10% PSMA expression group are located in the nasal cavity, nasopharynx or maxillary sinus, sites that are known to have a worse prognosis when compared to tumours originating from other subsites [
11]. Notwithstanding the small sample sizes in these study subgroups, specifically the aforementioned tumour sites are, besides poor RFS, concordantly associated with poor DSS and OS (data not shown). Tumour stage T4b and bone invasion were also over-represented in the low PSMA expression (≤10%) group. Although these parameters themselves are no independent predictors of RFS, tumour stage is strongly correlated with DSS and OS. Furthermore, tumour stage and bone invasion are mutually highly correlated (Pearson chi-square test
p < .01) as all except one T4b tumour showed bone invasion. Moreover, bone invasion is significantly associated with tumour localization in the nasal cavity, nasopharynx or maxillary sinus. Tumours at these locations recur more often. These collinearities could be explained by delayed presentation of tumours from this subsite (that often involves the skull base), but may have confounded the results.
Another factor of debate is the limited discriminative strength of the 10% cut-off point. The above mentioned considerations are supported by the minimal increase of Harrell’s C-statistic of the multivariate Cox-proportional Hazard model by 0.02 and therefore the additive value of PSMA to the prediction of RFS remains questionable.
A deep locoregional recurrence or growing distant metastases, for which conventional treatment options are no longer applicable due to functional irresectability or exceeded radiation limits, is a relevant problem in the management of AdCC. The limited experience with palliative chemotherapeutic agents and new initiatives with different targeted agents in this setting was recently reviewed [
31]. The present results of high PSMA expression in primary, recurrent and metastatic tumour cells with limited spatial and temporal variability, as well as the uptake of PSMA ligand in recurrent and distant metastatic AdCC on PET/CT, could suggest a potential role for palliative targeted treatment with Lutetium-177-PSMA [
10,
32]. First large trials of this radionuclide treatment of metastatic castration-resistant PC show high response rates with low toxicity, improved quality of life and even prolonged OS [
20,
33]. Regarding AdCC, the high amount of associated grade I xerostomia (87%) must be taken into account, as these patients usually have already been exposed to radiotherapy to the head and neck before [
20].
Conclusions
This study shows unambiguous PSMA expression in a large cohort of primary, recurrent and metastatic AdCC of the head and neck. Expression was seen in 94% of the primary tumours, which is analogous to PC except for the median lower number of positive tumour cells. In general, there was no relation between upregulated PSMA expression and pathological stage, tumour grade (growth pattern), the occurrence of locoregional recurrence or metastasis, and survival. Low primary tumour expression ≤10% is significantly associated with worsened RFS although its predictive value is limited. Of the recurrent and distant samples, respectively 87 and 90% were PSMA-positive, but staining could not be estimated based on primary tumour expression. This study provides encouraging supporting results that when other palliative systemic treatment options fail, PSMA-targeted imaging followed by experimental Lutetium-177-PSMA radionuclide therapy in AdCC, might be an alternative.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.