Do somatic USP8, USP48 and BRAF mutations differ in their genotype-phenotype correlation in Asian Indian patients with Cushing’s disease?

Purpose To estimate the prevalence of USP8, USP48 and BRAF mutations in patients with Cushing’s disease (CD) from the Indian subcontinent, and determine their genotype-phenotype correlation. Methods We prospectively recruited 46 patients with CD who underwent surgery between 2015 and July 2019 at our institute. Fresh frozen tumour tissue was obtained in all patients. Using Sanger sequencing, the presence of somatic USP8 mutations was documented and the frequency of USP48 and BRAF mutations in USP8 wild-type corticotroph adenomas was determined. Clinical, hormonal and surgical data were then compared between USP8-, USP48- and BRAF-variant carriers and patients with wild-type tumours.


Introduction
Cushing's disease (CD) is a type of adrenocorticotrophic hormone (ACTH) dependant Cushing's syndrome (CS) caused by hypersecretion of ACTH by pituitary adenomas. Extensive research has been done to explore the possible molecular mechanisms responsible for it but the exact aetiopathogenesis remains unclear.
Several recent publications have identi ed somatic mutations in the ubiquitin-speci c protease 8(USP8) genes of 21-62% patients with CD. [1][2][3][4][5] These mutations in exon 14 of the USP8 gene, via increased deubiquitinase (DUB) activity, have been shown to increase the expression of epithelial growth factor receptors (EGFR), thereby up-regulating EGFR-induced proopiomelanocortin (POMC) transcription and ACTH secretion. While the link between EGFR signalling and USP8 mutations was established in in-vitro studies, [1,3] other authors failed to demonstrate increased EGFR expression on immunohistochemistry (IHC) in USP8 mutated corticotroph adenomas.
[6] A recent publication [7] addresses this by demonstrating that mutant adenomas also suffer dysregulation of a number of other proteins namely p27/kip1, CABLES1, HSP90, and activated CREB; all of which have been previously found to be associated with either increased ACTH production or increased cell proliferation. [8] Further research revealed that in USP8 wild-type corticotroph adenomas, mutations in USP48 (10-23%) as well as BRAF (1-16%), which are also known to increase POMC transcription, might be the drivers of ACTH over-production. [9,10] What is evident from the literature is that there is a wide variation in the reported prevalence of USP8, USP48 and BRAF mutations from different countries. There is also limited clinical, hormonal and radiological data available on patients harbouring these mutations. Few studies have shown that USP8mutant corticotroph adenomas occur more frequently in women, are smaller in size, are less invasive, and are more likely to lead to disease remission following surgical resection. [2,5,6] A single study showed that USP48-variant adenomas were smaller in size. [10] In this study we aimed to determine the prevalence of signature and novel mutations in these genes in a cohort of South Asian patients for the rst time, and also look for de nite genotype-phenotype correlation in terms of their clinical, hormonal and radiological pro le.

Clinical data
Between September 2015 and July 2019, all patients diagnosed to have CD using standard clinical and biochemical criteria and who had discernible tumours on dedicated magnetic resonance imaging (MRI) sequences of the pituitary gland were enrolled in the study after obtaining informed consent in the case of adults, and assent in the case of minors. All patients underwent endoscopic transsphenoidal resection of their corticotroph adenomas. If patients were on ketoconazole, the drug was stopped 2 weeks prior to surgery. Using the surgical strategy previously published by us, [11] either selective adenomectomy, enlarged adenomectomy, hemihypophysectomy or subtotal tumour resection was done, depending on whether tumour margins were clearly de ned intraoperatively. Perioperative steroids were not administered in order to enable early evaluation of remission. Postoperatively, patients were started on steroids only if their 8 AM cortisol level dropped < 5 µg/dl or if they demonstrated symptoms of hypocortisolaemia or hyponatraemia. Patients were asked to review 3 months following surgery and thereafter every year. Remission was de ned by either the presence of a basal cortisol of < 5 µg/dl with requirement for steroid replacement therapy, or a basal cortisol > 5 µg/dl but with suppression to < 1.8 µg/dl with the 1 mg overnight dexamethasone suppression test Pathological evaluation and mutational analysis Each biopsy specimen obtained for a patient was equally divided for histopathological examination and genetic sequencing. Tissue for histopathological examination was xed in 10% buffered formalin and embedded in para n. Standard haematoxylin-eosin and reticulin staining was employed. Disruption of the normal acinar architecture along with the presence of monomorphic cells was considered as con rmation of tumour. If tumour could not be identi ed on initial sections, serial sectioning and examination with both haematoxylin-eosin and reticulin staining was done. Periodic acid Schiff (PAS) staining and IHC for ACTH (1:150, Biogenex, San Ramon, CA), other anterior pituitary hormones, and MIB-1 labelling index were performed on all surgical specimens using 5-µm sections. PAS and cytokeratin immunostainings were also used to con rm Crooke's hyaline change in normal adenohypophysis.
DNA was extracted from the fresh frozen samples of tissue that was con rmed to be tumorous on histopathology. The samples, which were stored at -70˚C, were rst thawed and a standardised protocol for DNA extraction from tissues was followed. The QIAamp DNA Mini Kit (Qiagen) was used for all extractions. The DNA was quantitated using a NanoDrop™ microvolume spectrophotometer (NanoDrop Technologies) and the 260/280 ratio was determined. Samples with a poor ratio or quantity were ampli ed with primers targeting a 133 bp region of an endogenous control, before they were subjected to further analysis.
The polymerase chain reaction (PCR) for the two primer sets of exon 14 (USP8) were performed using previous published primers. [2] For USP48 and BRAF, the primer sequences used were based on that employed by Chen et al. [9] and Ferchichi et al. respectively. [12] Details of primer sequences and thermal cycling conditions are elaborated in Supplementary Table 1. All reactions were carried out in 25 µl volume. The following thermal cycling pro le was employed for all PCRs: 95 0 C for 8 minutes, 95 0 C for 30 seconds, optimized anneal for 30 seconds, 72 0 C for 1 minute and nal extension of 72 0 C for 10 minutes. The PCR product was detected using a 1.5% agarose gel. Sanger sequencing of both the sense and antisense strands of products for both primer sets were performed with an automated DNA sequencer (ABI PRISM 310 genetic analyser) using the ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems). Mutational analysis was performed by comparing the sequence with the wild type and by looking for the presence of all known mutations in this exon.

Data collection and statistical analysis
A comprehensive proforma was used to record the clinical, biochemical and radiological parameters of each patient. Details of the surgery, histopathological examination, mutational analysis and postoperative complications were also documented. Follow-up data was regularly updated as and when patients reviewed in the outpatient clinic. Mean with standard deviation (SD) / median with inter-quartile range (IQR) were used for continuous variables as applicable. Frequency and percentage were used for categorical variables. Prevalence was expressed as a proportion with 95% CI. Categorical variables were compared using Chi-square testing. Continuous variables were compared using Student's t-test if normal assumptions were satis ed; otherwise the Mann-Whitney U test was used. Multivariate penalised logistic regression was done on variables which had p-values < 0.2 on bivariate analysis. Statistical analysis was

Results
During the study period, a total of 49 patients were recruited for the study, however 3 patients were subsequently excluded because adequate fresh frozen tumour tissue was not available for DNA extraction.

Remission and recurrence rates
At 3 months follow-up, 37 (80.4%) out of 46 patients were in remission. Only 4 out of 8 (50%) males entered remission, while 33 (86.8%) females achieved the same. Amongst those who went into remission, long term follow-up was available in 31 patients with a mean duration of 25.3 ± 13.6 (range 6-50) months. Seven (22.6 %) of these 31 patients (6 female and 1 male) subsequently had disease recurrence with a mean time to recurrence of 20.1 ± 7.4 (range 8-28) months.

Mutational analysis
In all 46 patients, the diagnosis of corticotroph adenoma was con rmed by histopathological examination and immunopositivity for ACTH. On DNA sequencing, 17 (37%) patients had somatic mutations of the USP8 gene in the mutational hotspot of exon 14 at the 14-3-3 binding motif. Of these, 11 were missense mutations while the remaining were frameshift mutations. All mutations were located in the amino acids (AA) 718-720 with the most common being p.Pro720Arg and p.Ser719del fs. The full list of mutations is found in Supplementary Table 2. Figure 1(a) shows the DNA sequencing chromatogram of a USP8 wild-type tumour, Figure 1 In the 29 USP8 wild-type adenomas, USP48 mutations were present in 4 (13.8%), with three missense mutations (p.Met415Ile) and1 splice-site mutation (p.Pro433IVS+2 T>A). There were no BRAF mutations in any of our patients. Figure 2(a) depicts the DNA sequencing chromatogram of patient with no USP48 mutation, Figure 2(b) is the chromatogram of a patient with a p.Met415Ile mutation in exon 10 of the USP48 gene and Figure 2(c) shows a USP48 splice-site mutation (p.Pro433IVS+2 T>A). Tables 2 and 3 compare the clinical, radiological, biochemical and surgical characteristics of the USP8mutants and USP8 wild-type tumours. The only signi cant difference between both groups was an increased presence of Crooke's hyaline change in the adjoining hypophysis of the mutated tumours (70.6% vs. 37.9%, p=0.032). There was larger number of macroadenomas in the USP8-mutant group but this fell just short of statistical signi cance (52.9% vs. 27.6%, p=0.085). Only 1 out of 8 (12.5%) males harboured the mutation while 16 out of 38 (42.1%) females had mutated tumours although this was not statistically signi cant. Also not statistically signi cant but important to note, was a higher rate of disease recurrence in patients with USP8-mutant tumours (33.3% vs. 18.2%, p=0.38). On multivariate analysis, Crooke's hyaline change remained signi cantly higher in patients with mutated tumours (OR 5.05, 95% CI 1.02 -25; p=0.047). All other variables were similar in both groups.

Genotype-phenotype correlation of patients with USP8 and USP48 mutations
The comparison between patients with USP48-mutants and USP48 wild-type tumours is made using bivariate analysis in Tables 4 and 5. The only signi cant difference between both groups was a higher rate of cavernous sinus invasion in the USP48-mutants (50% vs. 4%, p=0.042). Tumours with USP48 mutations recurred more frequently than their wild-type counterparts (33.3% vs. 15.8%; p=0.470), although this was also not statistically signi cant like in the case with the USP8-mutants. On multivariate analysis, there was no variable that was signi cantly different between the 2 groups.
Somatic mutations of both USP48 and BRAF genes in patients with CD were reported only in 2018. [9,10] The pooled incidence of USP48 and BRAF mutations in these two studies was 41/308 (13.3%) and 16/227 (7%). [15] In our study, the prevalence of USP48 mutations in USP8 wild-type adenomas was 13.8%, almost identical to that in the literature. Of interest is that almost all USP48and BRAF-variants were noted only in patients with USP8 wild-type tumours, although they themselves were not mutually exclusive.
Another important nding in our study was that not a single BRAF mutation was detected in the 29 patients with USP8 wild-type status. The absence of BRAF mutations in our patients with CD is in stark contrast to the 16.5% prevalence reported from China [9] but is similar to 1% prevalence amongst Caucasians. [10] Ethnicity may also have a bearing on the presence of USP8-variants, [1,2,5,7,10,14] as is evident from the higher prevalence of USP8 mutations (62%) in the Chinese population [3] as compared to that in the European population (21-48%).

Genotype-phenotype correlation of USP8-variants
Several studies have shown a signi cantly higher, if not exclusive occurrence of USP8-variants in women. [2,3,6,10,13,16] This was however not the case in the series by Albani et al., [14] who found that there was no signi cant gender association (USP8 mutations in 39% of females vs. 30% of males). In our series, only 1 male out of 8 (12.5%) had a USP8 mutation while 16/38 (42.1%) females carried USP8-variants. No male patient out of the 4 with USP8 wild-type CD harboured USP48 mutations. While this difference did not attain statistical signi cance due to the relatively small sample size, there seems to be a clear tendency towards female preponderance and might explain, at least partially, the well-known female predilection in CD which so far has not been adequately understood. Perez-Rivas et al. [2] rst proposed that oestrogens could demonstrate a growth-stimulating effect on USP8-mutated corticotrophs, a hypothesis supported by the fact while a balanced sex-ratio exists amongst paediatric patients with CD [17] in adults there is a distinct female predominance of the disease. 11 In addition, the demonstration of oestrogen receptors on corticotroph cells along with the stimulatory effect of oestradiol on murine corticotroph cell proliferation via EGFR signalling might be a reason for a higher incidence of CD in females. [18,19] Some reports indicate that USP8-mutated tumours occur in younger adults [2,7,14] and older children, [13] however, we and others [3][4][5][6] found no signi cant age difference between carriers of USP8-mutant and wild-type adenomas. BMI was found to be signi cantly higher in patients harbouring USP8-mutant tumours in two studies, [2,10] while another found that BMI was lower amongst patients with USP8mutant tumours. [13] We could not demonstrate any statistically signi cant difference in BMI between the 2 groups.
We found that the USP8-mutants in our series tended to be larger than their wild-type counterparts, although this did not achieve statistical signi cance (p=0.09). Findings on the issue of tumour size vary considerably, with variant-carrying tumours being reportedly smaller and less invasive in some studies, [3,6] while others have observed a greater size in USP8-mutant microadenomas compared with wild-type microadenomas. [2,5] This difference may be at least partially attributed to a selection bias in some studies. For example, Hayashi and colleagues[6] oversampled Crooke's cell adenomas, while the Chinese series [3] included a relatively large proportion of invasive tumours (>20%) and giant adenomas. Also worth noting is that none of the series reported so far mirror the ratio between micro-and macroadenomas normally found in CD, i.e. less than 10-20% macroadenomas, probably a consequence of the need for adequate pathological specimens to perform DNA or RNA sequencing.
With regards to the biochemical characteristics of these patients, one study [2] found that patients with USP8 mutations demonstrate increased suppression of cortisol after 8 mg dexamethasone while another found signi cantly lower plasma ACTH levels in these patients.
[6] Ma et al. [3] found that ACTH secretion was higher in patients carrying USP8 mutations. We failed to establish any signi cant difference between wild-type and mutated tumours in the biochemical parameters studied, a nding corroborated by others. [2,5,13,14] However, despite revealing no signi cant difference between the ACTH, UFC and serum cortisol in USP8variant and USP8 wild-type we did nd that the proportion of patients with Crooke's hyaline change in the adjoining adenohypophysis was signi cantly higher in the USP8-variant group. Crooke's hyaline change is typically seen with higher levels of hypercortisolism, [20] so it may be conjectured that patients with USP8mutant adenomas may have experienced severe hypercortisolism at some point of their illness.
The remission rates of CD was similar in the USP8-variant and wild type tumors as reported by others. [2,3,14] Some authors[5,6] note a higher remission rate in patients with USP8 mutations while others [13,14] report higher recurrence rates in mutated tumours. Another study found parallel recurrence rates, but a shorter mean time to recurrence in the USP8-variant adenomas. [3] We too found a higher recurrence rate amongst our patients with a USP8 mutation (33% vs. 18%), although it was not statistically signi cant due to the small sample size.

Genotype-phenotype correlation of USP48-variants
The study by Chen et al. [9] could not demonstrate any signi cant clinical, biochemical or radiological differences between patients with USP48-variants and USP48 wild-type corticotroph adenomas. However a subsequent study reported that USP48-mutant adenomas were smaller than wild-type adenomas. [10] In our series, the occurrence of cavernous sinus invasion was signi cantly higher in the 4 patients with USP48-mutant adenomas (50% vs. 4%, p=0.042). This difference has not been reported by any other study, however further research is required on these less common mutations before any meaningful conclusions can be drawn. The literature available on the genotype-phenotype correlation of CD patients with USP8 and USP48 mutations is summarised in Table 6.

Strengths and limitations of this study
This single-centre, prospective study from the Indian subcontinent, determines the prevalence of USP8, USP48 and BRAF mutations in patients with CD using fresh frozen tumour tissue that provided uniformity in the quality of tissue used for mutational analysis. The main limitation of the study was its relatively small sample size that precluded the establishment of statistically signi cant phenotypic differences between patients with USP8-/ USP48-variant tumours and their wild-type counterparts. Moreover, only mutations in the previously described mutational hotspots of the USP8, USP48 and BRAF genes were tested for. Next-generation sequencing (NGS) may have identi ed additional mutations in our series, as was seen in the study by Ballmann and colleagues, [4] where NGS detected a few USP8 mutations that were missed on Sanger sequencing.

Conclusions
The prevalence of somatic USP8 mutations in corticotroph adenomas was 37%. The frequency of USP48 mutations in patients with USP8 wild-type adenomas was 13.8% while BRAF mutations were not found in any of them. We report for the rst time, a novel USP48 splice-site mutation (p.Pro433IVS + 2 T > A).USP8mutated tumours had an increased presence of Crooke's hyaline change in the normal adenohypophysis. Tumours with USP48 mutations had signi cantly more cavernous sinus invasion than USP48 wild-type adenomas.