gsp Mutation Is Not a Molecular Biomarker of Long-Term Response to First-Generation Somatostatin Receptor Ligands in Acromegaly

Simple Summary Acromegaly treatment consists of surgical, medical, and radiation therapy. First-generation somatostatin receptor ligands are the mainstay of medical therapy, with approximately 40% disease control rate. Several parameters have been evaluated as predictors of response to these drugs, including mutations in the stimulatory G-protein α subunit (gsp mutation), which is still controversial. In this study, we aimed to evaluate in a large series of patients whether gsp mutation predicts long-term response to medical treatment and to characterize the gsp mutated population. The ability to predict response to medical therapy would help to choose a therapy that presents higher odds of controlling the disease, which ultimately would reduce treatment costs and disease morbi-mortality. Abstract Background: It is still controversial if activating mutations in the stimulatory G-protein α subunit (gsp mutation) are a biomarker of response to first generation somatostatin receptor ligands (fg-SRL) treatment in acromegaly. Thus, we aimed to evaluate whether gsp mutation predicts long-term response to fg-SRL treatment and to characterize the phenotype of patients harboring gsp mutations. Methods: GNAS1 sequencing was performed by Sanger. SST2 and SST5 were analyzed by immunohistochemistry (IHC) and real-time RT-PCR. The cytokeratin granulation pattern was evaluated by IHC. Biochemical control was defined as GH < 1.0 ng/mL and normal age-adjusted IGF-I levels. Results: gsp mutation was found in 54 out of 136 patients evaluated. Biochemical control with fg-SRL treatment was similar in gsp+ and gsp- patients (37% vs. 25%, p = 0.219). Tumors harboring gsp mutation were smaller (p = 0.035) and had a lower chance of invading cavernous sinuses (p = 0.001). SST5 protein (p = 0.047) and mRNA (p = 0.013) expression levels were higher in wild-type tumors. Conclusions: In this largest series available in the literature, we concluded that gsp is not a molecular biomarker of response to fg-SRL treatment in acromegaly. However, the importance of its negative association with cavernous sinus invasion and SST5 expression needs to be further investigated.


Introduction
Acromegaly is a disease caused by growth hormone (GH) hypersecretion, leading to increased production and secretion of insulin-like growth factor type I (IGF-I) due, in most cases, to a GH secreting pituitary adenoma (somatotropinoma). First-line treatment is surgical resection, generally transsphenoidal surgery (TSS) [1]. However, surgical remission rates vary from 42 to 57%, depending on the experience of the institution [2,3]. The remaining patients ought to be submitted to a diverse treatment, which may include a second surgery, medical treatment, or radiotherapy [1]. First-generation somatostatin receptor ligands (fg-SRLs), namely, octreotide, and lanreotide, are the standard first-line medical treatment options for the majority of patients, with biochemical control rates varying from 19 to 60% [4].
The stimulatory G-protein α subunit, encoded by the GNAS1 gene, is associated with growth hormone releasing hormone receptor signaling by the cAMP pathway, which is an important target of SRLs [5]. GNAS1 activating mutations (gsp mutations) are found in approximately 40% of sporadic somatotropinomas [6], commonly found in codons 201 and 227, and are thought to be associated with smaller, densely granulated somatotropinomas that present higher GH secretion and higher somatostatin receptor subtype 2 (SST2) expression; however, this phenotype has not been clearly defined [7][8][9][10][11]. Another characteristic of patients harboring gsp+ tumors is a putative better response to fg-SRLs [11]. This issue has been studied for a long time but is still a matter of debate, since there is no consensus on whether it can predict response to fg-SRL or not [11][12][13]. A recent meta-analysis showed that gsp+ patients present a higher GH reduction during the octreotide suppression test (OST) [6]. However, studies concerning long-term response to fg-SRL treatment have presented conflicting results [11].
Based on data reported to date, the present study aims to evaluate whether gsp mutation predicts long-term response to fg-SRLs and to better characterize the phenotype of patients harboring gsp mutation.

Study Design
This is a retrospective longitudinal multicenter study, including seven Brazilian pituitary disease reference centers.

Patients
Patients with a confirmed diagnosis of sporadic acromegaly who underwent surgical treatment and had either fresh frozen or formalin-fixed paraffin embedded (FFPE) tissue available for analysis were included. SST2 and SST5 were evaluated exclusively in patients not medically treated prior to surgery.

Data Collection
Demographic, laboratory (GH and IGF-I levels), and imaging data at diagnosis (maximum tumor diameter and cavernous sinus invasion) and pre-and post-fg-SRLs treatment (at least six months on the highest approved fg-SRL dose) laboratory data were collected. Biochemical control was defined as the achievement of GH levels < 1.0 ng/mL and normal age adjusted IGF-I levels [1]. GH and IGF-I reduction after treatment was also evaluated. IGF-I was expressed as times the upper limit of normal (xULN). Cavernous sinus invasion was evaluated according to the modified Knosp-Steiner criteria [14]. Tumors classified as Knosp 3 or higher were considered invasive, whereas tumors classified as Knosp 2 or lower were considered noninvasive.

DNA and RNA Extraction
DNA and RNA were extracted from frozen tumor fragments and DNA from FFPE tissues using the AllPrep™ DNA/RNA/miRNA Kit (Qiagen, Hilden, Germany, Cat. No. 80224) and ReliaPrep™ FFPE gDNA Miniprep (Promega, Wisconsin, USA, Cat. No. A2352), according to the manufacturers' protocols. At the end of the process, DNA and RNA were eluted in the provided buffer and stored at −80 • C until use.
For DNA extraction from FFPE tissues, six sections with 5-µm height and 1 cm 2 were obtained, and effective deparaffinization with molecular biology-grade mineral oil was performed before DNA extraction. Incubation conditions to reverse crosslinking without the need for overnight digestion, xylene, or other hazardous or volatile solvents followed the manufacturer's instructions.

DNA Sequencing
Fragments of GNAS1 gene containing codons 201 and 227 were amplified by polymerase chain reaction (PCR) in a 25-µL reaction mixture with 2.5 µL of 10 x PCR buffer, 0.3 µg of DNA sample, 50 mM MgCl 2 , 10 mM dNTP mix, 10 µM of each primer and Platinum™ Taq DNA Polymerase (Invitrogen™, Waltham, MA USA) and nuclease-free water. Reactions were amplified at an initial denaturation temperature of 94 • C for 5 min, followed by 35 cycles of 94 • C for 45 s, 57.5 • C for 45 s, and 72 • C for 30 s. A final extension at 72 • C for 7 min was carried out to allow complete extension of amplified fragments. The amplified products were then visualized on a 2% agarose gel stained with SYBR™ Safe DNA Gel Stain (Invitrogen™, Waltham, MA USA).

qPCR
Quantitative real-time PCR was performed using the same primers and method as previously described to analyze the number of copies of SST2 and SST5 mRNA [15].

Immunohistochemistry
Immunohistochemistry was performed as previously described [16]. Rabbit monoclonal antibodies directed against SST2 (UMB-1, 1:5000, Abcam, Cambridge, UK, Cat. No. SST2 and SST5 were evaluated using the immunoreactive score (IRS), as previously described, and high expression was defined as an IRS > 5 [15]. For the cytokeratin pattern, tumors were classified as sparsely granulated (SG), densely granulated (DG), or intermediate (which were grouped with DG tumors) as previously described [17]. The samples were analyzed by two observers, and discordant results were submitted to an evaluation by a third observer.

Statistical Analysis
Results are expressed as median (min-max). Categorical variables were compared with Fisher's exact test or a chi-squared test, as appropriate. Numerical variables were compared with a Mann-Whitney test. A p-value < 0.05 was considered significant.

Patient and Sample Characteristics
One hundred thirty-six patients were included in this study; 63 (46%) of them were men, and median age at diagnosis was 43 years old . Median GH and IGF-I levels at diagnosis were 44.1 ng/mL (1.1-611) and 3.9 xULN (1.3-11.4), respectively. Out of 111 patients, 5 (5%) had microadenomas. Of these, data with respect to diameter were available in 69 patients, and the median maximum diameter was 20 mm (5-54). A cavernous sinus invasion was evaluated in 54 patients, including 23 (43%) invasive and 31 (57%) noninvasive tumors. Eighty-one (60%) patients were treated with fg-SRLs, and 24 (30%) of them were biochemically controlled.
On the other hand, SST2 protein and mRNA expressions were positively associated with biochemical control (p = 0.021 and 0.031, respectively). Association was also found with a granulation pattern (p = 0.037), but not with SST5 protein or mRNA expression (p = 0.894 and 0.399, respectively).

Discussion
Precision medicine has been increasingly used for disease management in different medical specialties, particularly oncology [18]. In acromegaly, prediction of response to SRL treatment has been the focus of several studies [19][20][21][22]. Numerous clinical, biochemical, immuno-histopathological, molecular, and imaging parameters have been studied with this purpose, including gsp mutation [23]. The present study is the largest series evaluating gsp status with respect to long-term biochemical response to SRLs published so far, including 136 patients, in which gsp mutation was found in 40% of cases. This high prevalence underscores the need to establish its significance in relation to the management of disease following surgical failure. Therefore, we evaluated its association with biochemical control in 81 patients, which was not found.
Attempts to correlate gsp mutation with clinical response to SRLs date from the 1990s, when Yang et al. [12] first described this association. They found that none of the five poor responders in an octreotide suppression test (OST) were gsp+, while three out of five good responders were gsp+. Since then, several studies have evaluated gsp mutations and their correlation with both acute and long-term responses to SRLs [6][7][8][9][10]13,[24][25][26][27][28]. Recently, a meta-analysis encompassing 310 patients found 40% gsp+ tumors, similar to our results [6]. In this meta-analysis, patients harboring gsp+ tumors showed a greater GH level reduction during OST. It is important to emphasize that significant acute GH suppression does not necessarily correspond to a sustained long-term response [29]. If we take into consideration only studies evaluating long-term biochemical control, earlier reports showed a correlation with gsp mutation status [13,26], but these findings have not been confirmed in more recent studies [7,9,24,28]. Our data reinforce these later studies, particularly the study of Fougner et al. [24], who evaluated 38 patients and found no correlation of gsp mutation and GH or IGF-I reduction during OST or long-term fg-SRL treatment. There may be a concern that population heterogeneity could play a role in the differences found among studies. However, our study included patients from all over country, which presents a mixed population. Therefore, we consider that it is unlikely that these results may have been influenced by differences in the population, although it cannot be excluded that some genetic background has been underrepresented.
The phenotype of patients harboring gsp mutations is still controversial in the literature. We did not find any difference between patients with wild-type and gsp mutated tumors with respect to sex, age, GH, and IGF-I levels at diagnosis. However, tumors harboring gsp mutations are thought to be smaller, more frequently densely granulated, and more secretory [30]. Indeed, we found gsp mutated tumors to be smaller and less frequently invade cavernous sinuses, even when only macroadenomas were considered (data not shown). Of note, such invasion, characterized by the Knosp criteria, had never been evaluated in comparison to gsp mutation. Freda et al. [28] evaluated dura mater or bone invasion and found no difference between wild-type or gsp-mutated tumors, despite the latter being slightly smaller. Kim et al. [31] also did not find a difference in invasiveness between wild-type and gsp-mutated tumors classified according to the Hardy and Vezina grading system. Again, gsp-mutated tumors were significantly smaller.
With respect to granulation pattern, similar to Fougner et al. [24], we did not find a difference in the frequency of sparsely or densely granulated tumors between gsp+ and gsptumors. It is important to note that ours and Fougner's series are the largest described series of somatotropinomas.
A hypothesis that could favor a better response of gsp+ tumors to SRLs is the demonstration that SST may be differentially expressed between gspand gsp+ tumors, in particular, with a higher expression of SST2 mRNA levels in the latter tumors [7][8][9]. However, these data have not been confirmed by other authors [13,32,33]. In our study, we did not find a significant correlation between gsp+ tumors and SST2 protein or mRNA levels.
In contrast, we did find that wild-type tumors express higher SST5 levels, both at the protein and mRNA levels. A previous finding of higher SST5 mRNA levels in gsptumors had been described with borderline significance (p = 0.06) by our group [9]. The present series, aggregating a large number of patients from many Brazilian centers, allowed us to confirm the higher SST5 protein and gene expression, by two different very sensitive methods. In addition, we also confirmed a high correlation between SST5 mRNA and protein, as previously demonstrated [34]. The clinical importance of the different SST2 and SST5 expression in the presence or absence of gsp mutation has not yet been investigated. Since SST5 expression has been shown to predict response to pasireotide in patients who are resistant to fg-SRLs [35], studies evaluating gsp mutations and the response to pasireotide treatment should be further examined in acromegaly.
We confirmed previous findings that showed association between SST2 expression and granulation pattern with response to fg-SRL [11,15,36,37]. Several other biomarkers have been shown to be associated with medical or surgical therapy response, but none with optimal predictive power [16,[38][39][40][41]. Thus, there is still a need for better biomarkers or a combination of them. We recently developed a machine learning based prediction model, including clinical (age and sex), biochemical (GH and IGF-I levels), and immunohistochemical (SST2 and SST5 expression and cytokeratin granulation pattern) features, that predicted response to fg-SRL with an accuracy of 86.3%, positive predictive value of 83.3% and negative predictive value of 87.5% [42]. Another recent study evaluated clinical, biochemical, and imaging parameters, and found that age was negatively correlated with IGF-I reduction, whereas IGF-I levels at diagnosis and tumor T2 hypointensity at MRI were positively correlated [43].

Conclusions
The present study confirmed the frequency of gsp+ mutations in approximately 40% of sporadic somatotropinomas. It also reinforced that, whereas gsp+ tumors were significantly smaller than gsptumors, the phenotype did not differ with respect to sex, age, GH, and IGF-I levels at diagnosis or the association with biochemical control. This is the largest series evaluating gsp mutation status with respect to long-term medical treatment with fg-SRLs, and we concluded that the gsp+ mutation cannot be used as a molecular biomarker of long-term response to fg-SRLs. However, the higher expression of SST5 in gsp+ tumors raises an important question regarding gsp mutation and the response to pasireotide treatment in acromegaly.