Predictive and prognostic significance of tumour subtype, SSTR1‐5 and e‐cadherin expression in a well‐defined cohort of patients with acromegaly

Abstract In somatotroph pituitary tumours, somatostatin analogue (SSA) therapy outcomes vary throughout the studies. We performed an analysis of cohort of patients with acromegaly from the Czech registry to identify new prognostic and predictive factors. Clinical data of patients were collected, and complex immunohistochemical assessment of tumour samples was performed (SSTR1‐5, dopamine D2 receptor, E‐cadherin, AIP). The study included 110 patients. In 31, SSA treatment outcome was evaluated. Sparsely granulated tumours (SGST) differed from the other subtypes in expression of SSTR2A, SSTR3, SSTR5 and E‐cadherin and occurred more often in young. No other clinical differences were observed. Trouillas grading system showed association with age, tumour size and SSTR2A expression. Factors significantly associated with SSA treatment outcome included age, IGF1 levels, tumour size and expression of E‐cadherin and SSTR2A. In the group of SGST, poor SSA response was observed in younger patients with larger tumours, lower levels of SSTR2A and higher Ki67. We observed no relationship with expression of other proteins including AIP. No predictive value of E‐cadherin was observed when tumour subtype was considered. Multiple additional factors apart from SSTR2A expression can predict treatment outcome in patients with acromegaly.


| INTRODUC TI ON
Pituitary neuroendocrine tumours (PitNETs) 1  the studies. [6][7][8][9] The importance of other somatostatin receptors (SSTRs) is not well understood. Among other factors that may modulate SSA treatment response, E-cadherin and aryl hydrocarbon receptor-interacting protein (AIP) have been studied repeatedly. [10][11][12][13] The prediction of SSA response and clinical behaviour of the tumour is important for the further clinical management of patients with acromegaly, and brings essential economic consequences as well. Thus, we decided to analyse the clinicopathological features of a large cohort of somatotroph PitNETs and to assess possible predictive factors in a subgroup of patients treated with SSA. reduction > 50%, and poor response -IGF1 reduction < 50%) and a 4-tiered system (IGF1 reduction < 20%; IGF1 reduction 20%-50%; IGF1 reduction > 50% and attainment of normal IGF1 levels for age).

| Patients -Cohort characteristics
The cut-offs were defined arbitrarily. All the patients included in the RESET database had initially signed an informed consent agreeing with a future research use of the tissue. The design of the study was approved by the ethical committee of the first author's institution, where the experimental work was performed.

| Immunohistochemistry
Archived formalin-fixed paraffin-embedded (FFPE) tissue blocks of tumours were collected from six participating institutions.

| SSTRs, E-cadherin and D2DR immunohistochemistry evaluation
Evaluation of SSTR1-5 and D2DR expression was performed by three observers (JS, MM and LP) independently after an initial consensus meeting at the multi-head microscope. The percentage of positive tumour cells was established and the intensity of staining in the cellular subgroups was assessed on a scale from 1 to 3.
The histoscore (H-score) was then calculated for each tumour as the percentage of cells multiplied by their respective staining intensity, thus ranging from 0 to 300. Ie for a tumour with 30% of weakly positive cells, 20% of moderately positive cells and 10% of strongly positive cells, the H-score was 100 (1*30 + 2*20 + 3*10).  F I G U R E 1 Sparsely granulated somatotroph tumours (SGST) showed statistically significant lower age and a tendency for bigger size compared to all non-SGST histological subtypes of tumours

| Cases with preoperative treatment
Nine out of 110 patients received preoperative treatment with a firstgeneration SSA. The median treatment duration before the surgery was 10 months. Compared to the untreated group, a significantly lower level of Ki67 index was observed in the pre-treated group (median 0.79% vs 2.91%, P = .02, Mann-Whitney test), while no difference in other parameters was found. In further analyses of Ki67, and Trouillas scoring system, we thus excluded pre-treated cases.

| Trouillas grade
Concerning the proliferative and non-proliferative tumour subsets  Figure 4.
The relationship between response and other parameters did not achieve statistical significance.

| Analysis of the significance of the histological subtype for the response
The treated group included one cytokeratin-negative tumour: we decided to exclude the patient from further analysis related to the tumour subtype due to the lack of literary data and consensus about subclassification of these tumours. Patients with SGST showed a trend towards poorer treatment response compared to patients from the non-SGST group, but the finding did not achieve statistical significance (P = .058, Fisher's exact test).

F I G U R E 3 ROC curves for expression of E-cadherin, SSTR2A
and SSTR3 in SGST vs non-SGST group as defined by cytokeratin expression; E-cadherin was the most sensitive and specific feature of SGST (AUC = 0.97, P < .0001) followed by SSTR2A and SSTR3. Cytokeratin-negative tumours were excluded from the analysis A summary of response rates is shown, along with median duration of treatment.

| Response to treatment in patients with non-SGST
There were 14 patients with non-SGST altogether. Treatment response was superior to that in the SGST group: the reduction in

| D ISCUSS I ON
We performed a comprehensive statistical analysis of the biological and clinical parameters of a large somatotroph PitNETs series and included the evaluation of potential predictors for treatment with a first-generation SSA. The distribution of histological subtypes is comparable to the literature. 17  be due to the lack of standardized cut-off values for the proportion of βTSH + cells that define this subgroup throughout the studies.
When comparing clinical features of different pathological subsets, we observed only an association of SGST with younger age and a tendency for larger tumour size, when compared to non-SGST.

| E-cadherin
Low E-cadherin expression in our study was the most sensitive and specific feature of SGST defined by the presence of fibrous bodies: in the non-SGST group, only one case (1.5%) showed complete absence of E-cadherin compared to 30 in the SGST group (71.4%). Three plurihormonal Pit1 + tumours showed absence of E-cadherin as well; it is unknown whether this illustrates the general mechanisms (ie epithelial-mesenchymal transition) responsible for the more aggressive behaviour of these tumours or reflects the possible relationship between the two entities, as suggested recently. 26 The predictive value of E-cadherin for SSA treatment has been suggested in two studies. 11,12 However, based on the results, we do not consider E-cadherin to be an independent predictor, but rather a surrogate marker of the SGST subtype: this is illustrated by the fact that all the SGST in the poor response subgroup (8/10) showed no expression of E-cadherin (H-score 0) while two remaining non-SGST tumours showed E-cadherin expression comparable to the subgroup with good response (H-scores 104 and 155).
Moreover, we did not observe any differences in E-cadherin scores between subsets of good and poor responders in SGST and non-SGST subgroups.

| SSTR3 and D2DR
The significantly higher expression of SSTR3 in non-SGST has not been reported in the literature 13,22,23 and we interpret this finding as a result of more sensitive methodology (use monoclonal antibody UMB5 and H-score instead of semiquantitative scales). We observed a significant association between low expression of D2DR and tumour invasion.
Such finding has not been reported previously. 7,27,28 However, due to the small difference in D2DR H-score between the invasive and noninvasive group, the clinical significance remains to be validated.

| Trouillas grade
The relationship between Trouillas grade and the clinicopathological features of somatotrophs at the time of diagnosis has not been reported in the literature. In our study, tumours classified as 2b were significantly larger, showed lower expression of SSTR2A and were more common in younger patients.. A more aggressive phenotype of the disease has been previously reported in a subset of younger patients with SGSTs that showed higher Ki67 index. 29 Although no association of Trouillas grade with histological subtype was observed in our study, we speculate that the agressive subset previously identified by the cluster analysis 29 and grade 2b tumours may at least partially overlap, given the phenotypic similarities. The reason for these findings is currently unknown, but it may reflect different pathogenesis of somatotroph tumours in different age groups.

| Response to somatostatin analogue treatment -role of histological subtype and SSTR
Using the two-tiered system, a good response to SSA was sig-

| Response to somatostatin analogue treatment -role of AIP
Recently, low expression of AIP in sporadic somatotroph tumours was also associated with poor response to the first-generation SSAs, 10,13,31 higher Ki67 index, 10 larger tumour size at presentation, and SGST subtype. 36 In our study, however, we observed no association between AIP H-score and SSA response or any other clinicopathological parameters, including histological subtypes. In our work, we observed levels of H-score for AIP similar to those from the largest available study on the subject. 31 Using the same antibody (clone 35-2) at 5-times greater dilution, the median H-score for AIP was 230 and 240 in our study, compared to 220 and 270 in the poor response and good response groups of the aforementioned paper.
While the lack of association between AIP and SSA response has also been reported previously, 36

| CON CLUS ION
It is important to subclassify somatotroph PitNETS correctly according to their histological features because the individual tumour subtypes may differ in biological and clinical features. Although a poor response to SSA is more common in the SGST subtype, the subgroup is not homogeneous with respect to treatment response, and different predictive factors play a role in this context. Low SSTR2A receptor expression is significantly associated with a low response to treatment with a somatostatin analogue regardless of histological subtype. Although low E-cadherin expression is a predictor of poor outcome, we were unable to prove its value independent of the histological subtype of tumour, and in our opinion, it is merely a characteristic SGSTs, which in general tend to show more unsatisfactory response. We were unable to show a predictive value for AIP expression, possibly due to the lack of standardization of the immunohistochemical reaction, and this may limit the routine use of the antibody in this context. At least in non-SGSTs, expression of SSTR5 might play an advantageous role in attaining a biochemical response.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.