Phosphoproteomic dysregulation promotes tumor proliferation in Cushing’s disease

Significance

Cushing’s disease (CD) causes significant morbidity and mortality despite best medical and surgical treatment. Surgery is the mainstay of treatment, but carries perioperative risks and is sometimes followed by recurrence. There is a paucity of effective medical treatments, due in part to a limited understanding of tumor mechanisms. The majority of CD adenomas are wild-type, with no known causal mutations. Our study identifies kinase-phosphatase imbalance as a mechanism of CD tumorigenesis common to wild-type and mutant CD adenomas. We show that this pathway can be pharmacologically targeted in vitro and in vivo using fingolimod, an US Food and Drug Administration -approved small molecule, potentially a therapeutic strategy for patients with CD.

Abstract

Pituitary adenomas constitute up to 20% of primary brain tumors, yet somatic mutations are only found in 15% of pituitary adenomas. Epigenomic dysregulation has been proposed as a tumorigenic mechanism in pituitary adenomas causing Cushing’s disease (CD). We created paired datasets of human CD adenomas and en-route margin adult human pituitary glands, and assayed their chromatin accessibility, DNA methylation, transcriptomic, proteomic, and phospho-proteomic landscapes. In CD adenomas, we found epigenetic reactivation of a neurodevelopmental phosphoprotein program typically lost in the postnatal pituitary gland. CD cells overexpressed PPP1R17, an endogenous inhibitor of the protein phosphatase PP2A. Mechanistically, PPP1R17 overexpression in normal murine pituitary cells recapitulated the adenoma phenotype, and the resulting hyperproliferation was reversed in a PP2A-dependent manner by fingolimod, an FDA-approved small molecule, both in vitro and in vivo. Our findings highlight kinase-phosphatase imbalance as a targetable mechanism in CD.

Eukaryotic cellular homeostasis is tightly governed by the balance of protein kinases and phosphatases, which control phosphorylation levels and enzymatic activity of key intracellular proteins (1, 2). Disruption of this kinase-phosphatase balance has been reported underlying malignant tumors driven by discrete mutations (3, 4). In nonmalignant tumors, kinase dysregulation has been described as a mechanism of neoplastic tumor progression following the acquisition of mutations (5). Little is known about early aberrant phosphorylation events in nonmalignant tumor formation, or about epigenetic mechanisms underlying such dysregulation.

The protein phosphatase (PP2A) family of serine/threonine phosphatases plays a critical role in the regulation of signal transduction cascades, cell cycle regulation, cell morphology, and development (6). PP2A is regulated by the relative abundance of its regulatory subunits, which control its substrate specificity, subcellular localization, and catalytic activity (7–10). The catalytic subunit (PP2Ac) is further regulated by methylation or phosphorylation. Several endogenous PP2A inhibitors have been identified, which either bind directly to PP2Ac or target specific PP2A holoenzymes and prevent dephosphorylation of PP2A substrates (11). PPP1R17 is known to inhibit protein phosphatase 1 (PP1), and PP2A in cerebellar Purkinje cells (8), however, its role in pituitary tumorigenesis has not been previously investigated.

The pituitary gland is the site of sporadic adenomas (benign tumors) in up to 35% of the general population (12, 13). Despite its small size (~500 mg), adenomas of the pituitary gland comprise 20% of all primary brain tumors (14). Even after successful removal, histologically benign adenomas can recur in up to 30% of patients, causing long-term morbidity (15). Cushing’s disease (CD) is caused by pituitary adenomas that hypersecrete adrenocorticotropin (ACTH). CD adenomas are often small (2 to 10 mm) at diagnosis, minimally distort the pituitary architecture and have a predictable effect on the hormonal axes (biochemical testing confirms the presence of and remission from CD) (16–18). Nonsurgical treatments are limited, therefore, understanding mechanisms of tumorigenesis may improve our ability to target these tumors (19).

A minority of CD adenomas (total ~40%) have somatic mutations in USP8 (~30%), USP48, or BRAF (20–22). Epigenetic dysregulation has been hypothesized to contribute to CD tumorigenesis via promoter hypomethylation of POMC, ESR1, and CASP8 (23, 24), via downregulation of antiproliferative micro-RNAs miR-15a, miR-16, and miR-26a (25, 26), and via increased glucocorticoid resistance triggered by loss of HDAC2 and Swi/Snf chromatin remodeling complex subunit Brg1 (27). However, the chromatin accessibility landscape of CD remains unknown. Previously, we applied advances in surgical sampling and in vitro modeling to identify antiapoptotic mechanisms in wildtype CD adenomas (28). Here, we investigated common epigenetic mechanisms underlying CD adenomas using a multiomics approach.

Results

PPP1R17 Is Overexpressed in CD Adenomas.

We performed single nucleus (sn) multiome analysis on nuclei extracted from surgically annotated (Movie S1) freshly dissociated CD adenomas versus en-route, adjacent cells in the tumor margin (total 37,263 nuclei; Fig. 1A and Dataset S1). We included wild-type and USP8-mutated CD adenomas in this study (Fig. 1B and Dataset S1) and used marker-based cell-type classification of single nuclei (29, 30) that corresponded well with unsupervised clustering (31, 32). Using a validated set of expression markers (28), we identified canonical pituitary cell classes in each sample and generated snRNAseq gene signatures for corticotrophs, lactotrophs, somatotrophs, endothelial cells, gonadotrophs, leukocytes, and folliculostellate cells (FSCs; Fig. 1C, SI Appendix, Fig. S1, and Dataset S2). Importantly, we distinguished between core CD corticotrophs versus adjacent corticotrophs in the tumor margin (Fig. 1C). We found marked similarities between cell-type specific gene signatures derived from snRNAseq compared to single cell RNAseq (28). In the integrated snRNAseq dataset, most cells within the tumor core were corticotrophs, whereas corticotrophs comprised only a small fraction of cells in the tumor margin (Fig. 1 D, Left two panels).

Fig. 1.

PPP1R17 is overexpressed in CD adenomas. (A) Study schema. Pituitary adenomas and en-route adjacent margin tissues (when available) were surgically annotated and processed separately for single-nucleus multiome (ATACseq plus RNAseq) analysis, Mass spectrometry (MS), multiplex immunocytochemistry (mIHC), and DNA methylation. (B) Summary of human surgical samples and investigations included in the current study. (C) Marker gene expression in the integrated multiome dataset and their corresponding cell classes (SI Appendix, Fig. S1 and Dataset S2). (D, Left panel) UMAP projections of distinct nuclei from snRNAseq analysis using patient-derived samples including postnatal pituitary cells in the CD tumor core versus margin. Nuclei were clustered according to their cell-type identity. (Middle panel) CD core specimens are comprised mostly of corticotrophs. (Right panel) PPP1R17 expression is largely restricted to core CD corticotrophs. (E) scRNAseq analysis (28) evaluating PPP1R17 expression (Right column) mapped over pituitary cell types (Left column). Top row: postnatal pituitary cells from the tumor margin. Middle row: CD core adenomas. Bottom row: non-CD adenomas (1 growth hormone and 1 nonfunctioning pituitary adenoma). (F) Volcano plot from snRNAseq data showing differentially expressed genes between CD core versus margin corticotrophs. (G) UpSet plot with pairwise comparisons between syngeneic CD core versus margin corticotrophs from the same patients evaluated by snRNAseq (P1 versus P2) or scRNAseq (P6 versus P7). CD = Cushing’s disease.

Next, we evaluated genes differentially overexpressed in CD core versus margin cells using snRNAseq. PPP1R17 was predominantly expressed in CD core corticotrophs (Fig. 1 D, Right panel), a finding which was observed both in our previously published scRNAseq dataset (28) as well as in the current snRNAseq dataset (Fig. 1 E and F). Using a combination of trajectory inference and RNA velocity using CellRank (33), the fate probability transition from margin to tumor phenotype was strongly linked to PPP1R17 overexpression (SI Appendix, Fig. S2). Furthermore, when we limited our analysis to syngeneic tumor-margin samples from the same patients (P1 and P2 snRNAseq, and P6 and P7 scRNAseq), PPP1R17 was the only gene differentially overexpressed in CD adenomas versus syngeneic margin pituitary glands (Fig. 1G and Dataset S4).

Mechanisms of PPP1R17 Overexpression in CD Adenomas.

To better understand the mechanisms underlying PPP1R17 overexpression, we leveraged our snATACseq data and mapped the corticotroph-restricted fragment coverage (binned to 1 Mb) over the entire genome. We found increased chromatin accessibility in adenoma corticotrophs compared to margin corticotrophs (Fig. 2A). We then mapped fragments to “peaks” specific to corticotrophs, and individual genes. Upon differential analysis, we found increased coverage at peak regions, regions proximal to transcription start sites and distal to transcription end sites in CD adenoma corticotrophs (Fig. 2B).

Fig. 2.

Altered global and PPP1R17 locus-specific chromatin accessibility in CD adenoma cells. (A) Fragment coverage over the entire genome in the single-nucleus multiome dataset. Fragment coverage of corticotroph cells is averaged over four syngeneic adenoma core-margin pairs and two additional adenomas, binned to 1mBase, and mapped at the same scale for adenoma corticotrophs (red) and adjacent margin corticotrophs (blue). (B) Waterfall plots mapping reads per kilobase million (RPKM)-normalized fragment coverage in corticotrophs over peak (Left), and gene (Right) regions. In each panel, the right column is the adenoma corticotroph map and the left is for corticotrophs derived from the adjacent margin. (C) Chromatin accessibility at the PPP1R17 locus in adenoma corticotrophs (red track) versus corticotrophs derived from the adjacent margin (blue track), plotted at the same scale. Chromatin accessibility data were derived from two syngeneic adenoma margin pairs and two additional CD adenomas and averaged across compartments. (D and E) Differentially methylated regions in CD adenomas (n = 6) versus autopsy-derived normal pituitary glands (n = 11). (F) Bisulfite conversion followed by methylation-specific PCR evaluating methylation at the PPP1R17 promoter in CD adenomas (n = 5) versus autopsy-derived normal pituitary glands (n = 5, subset from E). (G, i) Alpha GSU staining in mouse E14.5 embryos highlights pars tuberalis (PT) and anterior lobe (AL). (ii and iii) Pomc staining (brown) is mostly restricted to the AL, and ppp1R17 staining (blue)is found in the PT and AL (blue arrowheads). (iv) By E18.5, ppp1R17 staining (blue) is mostly restricted to the intermediate lobe (IL) and PT, with only few select AL cells retaining ppp1R17 positivity. (v) Pomc (green) staining AL and IL, and ppp1R17 (red) restricted to the IL. (vi and vii) Minimal Ppp1R17 expression in the adult mouse (red) in the AL, with residual staining in the IL. (Scale bar, 50 µM.) (H) Representative mIHC image (P17) of a whole-mounted CD adenoma and its adjacent margin tissues corticotroph-specific POMC and TBX19 expression. PPP1R17, and its canonical substrate PP2Ac coexpression assessed within the identified adenoma. (Scale bar, 50 µM.) (I) Quantitative analysis of mIHC signal intensity across 18 channels (except PPP1R17 and PP2Ac) identifies core/adenoma cells distinct from margin/normal cells. (J) Averaged proportional coexpression of PPP1R17 and PP2Ac in margin/normal (1.68%) and core/adenoma cells (45%) across four samples.

We identified 1,094 differentially accessible peaks (606 up in the tumor core, 488 up in the adjacent margin) and 517 differentially accessible genes (150 up in the tumor core, 367 up in the adjacent margin; SI Appendix, Fig. S3A and Dataset S4). Comparative gene-set enrichment (GSEA) analysis of gene expression and chromatin accessibility profiles revealed a significant overlap in pathways (n = 67 GO pathways) including cytoplasmic translation and oxidative phosphorylation (SI Appendix, Fig. S3 B and C and Dataset S5). Several transcription factor motifs were differentially accessible in CD adenomas, including NF-Y and SP1 (Dataset S6). NF-Y is involved in chromatin remodeling, promotes cell proliferation, and declines during differentiation (34, 35). SP1 is an essential embryonic transcription factor responsible for regulating cell cycle, methylation-free CpG islands, and formation of active chromatin structures (36). Correspondingly, several NF-Y and SP1 targets were differentially overexpressed in adenoma corticotrophs (SI Appendix, Fig. S3 D and E and Dataset S7).

We focused on the PPP1R17 locus and identified increased chromatin accessibility at the PPP1R17 promoter (Fig. 2C). We also found increased chromatin accessibility at an internal enhancer from which an lncRNA PPP1R17-203 is transcribed (SI Appendix, Fig. S4). PPP1R17-203 shares sequence similarity with MAP7 which plays a role in several cancer types (37), and BLASTN revealed interactions between PPP1R17-203 and several other lncRNAs (Dataset S8). The role of PPP1R17-203 in CD tumorigenesis remains to be explored.

Pituitary adenomas can be clustered according to their DNA methylation profiles (23). We asked if promoter DNA hypomethylation contributed to PPP1R17 overexpression. DNA methylation profiling on tissues from six CD adenomas versus 11 autopsy-derived pituitary adenomas revealed widespread DNA methylation changes (Fig. 2 D and E and Dataset S9). Bisulfite conversion and methylation-specific PCR from a subset of DNA methylation samples including five CD adenomas versus five autopsy-derived human pituitaries identified DNA hypomethylation at the PPP1R17 promoter (P = 0.03; Fig. 2F). Analysis of TCGA data similarly found PPP1R17 promoter hypomethylation in several tumor types including hepatocellular carcinoma and head and neck squamous cancer (SI Appendix, Fig. S5A). Importantly, there was no concurrent over- or underexpression of other PP1 or PP2A subunit genes (SI Appendix, Fig. S5B) in CD tumors compared to adjacent normal gland. These data suggest that epigenetic dysregulation contributes to PPP1R17 overexpression in CD adenomas.

Identification of Upstream Drivers of PPP1R17 Overexpression in CD.

PPP1R17 is expressed in neural progenitor cells during human fetal development but is lost in adulthood except for cerebellar Purkinje cells and occasional cells in the hypothalamus and pons/medulla (38, 39). We found robust Ppp1r17 expression in E14.5 mouse embryonic pituitary glands, which decreased markedly by E18.5 (Fig. 2 G, iv and v) and diminished further in the adult mouse anterior pituitary lobe where we found only sparse and low expression (Fig. 2 G, vi and vii). In contrast, PPP1R17 protein was abundant within the CD adenoma compartment and colocalized with its target protein PP2Ac, a ubiquitous serine-threonine phosphatase (40) and tumor suppressor (41) (Fig. 2H and SI Appendix, Fig. S6). Upon quantitative analysis [Ward method (42)], a significant (P < 0.01) proportion (45%) of core/adenoma cells coexpressed PPP1R17 and PP2Ac compared to margin/normal cells (Fig. 2I) across all samples tested (Fig. 2J and SI Appendix, Fig. S7).

We next assessed proteins associated with the accessible PPP1R17 promoter in CD adenomas by reverse-chromatin immunoprecipitation (rChIP) using CRISPR dCas9-3XFLAG and a gRNA specific to the PPP1R17 promoter to pull down promoter-associated proteins for Mass spectrometry (43) (SI Appendix, Fig. S8). Control experiments excluded the PPP1R17 targeting gRNA. Proteins bound to the PPP1R17 promoter in CD adenomas (Dataset S10) involved mRNA splicing and processing pathways (SI Appendix, Fig. S8). Several histone variants including H1.2, H1.4, H2A.2, H3.2 were also enriched at the PPP1R17 promoter, as well as HP1BP3 which plays a key role in heterochromatin organization. Additionally, several members of the RAS oncogene family were enriched at the PPP1R17 promoter, including RAB10, RAB15, RAB1A, RAB2B, and RAB3B. Our snRNAseq data verified RAB3B overexpression in CD corticotrophs (Dataset S2). Taken together, these findings suggest that increased RAS signaling (44) and chromatin remodeling events contribute to PPP1R17 reactivation in CD corticotrophs.

PPP1R17 Interacts with PP1 and PP2A in CD Adenomas.

PPP1R17 has been identified as an endogenous inhibitor of protein phosphatase 1 (PP1), and of tumor-suppressor protein phosphatase 2A (PP2A) (45). GSEA of differentially expressed genes in tumor corticotrophs (compared to normal corticotrophs) revealed an inhibition of both PP1 and PP2A pathways [LINCS L1000 KO Consensus Sigs geneset (46); SI Appendix, Fig. S9A]. The observed transcriptomic phenotype recapitulated the effect of PP1 and PP2A inhibitors—okadaic acid and calyculin (LINCS L1000 Chem Pert Consensus Sigs geneset) [(47); SI Appendix, Fig. S9B]. A query with Reactome pathway 2024 geneset revealed a signal of PP2A complex inhibition in tumor corticotrophs (SI Appendix, Fig. S9C). Since PP2A is a known, and targetable tumor-suppressor, our subsequent analysis focused on the effect of PPP1R17 on PP2A in tumor corticotrophs. The most common mode of PP2A inhibition in human tumors is via overexpression of other known endogenous PP2A inhibitors including CIP2A, PME1, and SET (48, 49). Other than PPP1R17, we did not find transcriptional overexpression of other endogenous PP2A inhibitors in adenoma corticotrophs (SI Appendix, Fig. S10). Known endogenous inhibitors PME1 and SET interact with PP2Ac at Arg89 and Tyr127. AlphaFold Multimer (50) predicted binding of human PPP1R17 to the PP2Ac similarly (Fig. 3A). We verified PPP1R17 colocalization with PP2Ac in human CD adenomas by proximity ligation assay (PLA) (Fig. 3B) in addition to mIHC (Fig. 2H). Based on the expression of B regulatory genes including PPP12R2A-D (SI Appendix, Fig. S5B), we suspect that the predominant PP2A holoenzyme in human corticotrophs is B55. Taken together, these findings suggest that PPP1R17 overexpression in tumor corticotrophs results in PP1 and PP2A inhibition.

Fig. 3.

PPP1R17 inhibits PP2A phosphatase activity in CD. (A) AlphaFold Multimer modeling predicting binding of human PPP1R17 to PP2Ac subunit. (B) Representative PLA in margin (Left) versus CD adenoma core (Right) using antibodies for PP2Ac and PPP1R17. Red puncta indicate peptides in proximity. (C) MoMo motif analysis of preferentially hyperphosphorylated and dephosphorylated motifs in CD adenoma core versus margin (n = 3/group; Top panel) or in mCort^PPP1R17 versus mCort^GFP cells (n = 5/group; Bottom panel). (D) Representative multiplexed immunohistochemistry staining of a CD adenoma (P17) and its adjacent margin. Inset identifies the region of interest that is magnified to demonstrate pERK and pAKT at the margin of core and margin. (E) Western blots of mCort cells and AtT-20 cells transiently transfected with lentiviral plasmids encoding PPP1R17 versus GFP evaluating phosphorylation of PP2A targets AKT, ERK, and S6. (n = 3/condition; bar graphs with t tests; *P < 0.05; **P < 0.01; blank–not significant).

PPP1R17 Reactivation Induces Tumorigenic Signaling in Adenomas.

We further explored the downstream effects of PP2A inhibition by performing quantitative phosphoproteomic profiling where we labeled syngeneic, paired CD adenomas and their adjacent margin with Tandem Mass Tag (TMT) reagents, enriched phosphopeptides using TiO2 and iMac methods, and acquired LC–MS/MS data on peptides with and without phosphoenrichment. We normalized phosphopeptides to their corresponding peptides. We performed MoMo motif analysis in CD adenomas versus adjacent margin (51) and identified preferential hyperphosphorylation of peptides with motifs xxSPxx and DxxxSxx (Fig. 3 C, Top panel and SI Appendix, Fig. S11 A and B). Correspondingly, several PP2A targets were hyperphosphorylated in CD adenomas including ERK and AKT (Fig. 3D). We identified hyperphosphorylation of peptides involved in RNA splicing and chromosome condensation (Dataset S11).

We next validated the activation of tumorigenic pathways in vitro using mouse anterior pituitary cells (mCort) to serve as baseline normal cells. We stably overexpressed PPP1R17 tagged to GFP in mCort cells (mCort^PPP1R17; SI Appendix, Fig. S12). With PPP1R17 overexpression, there was no significant up or down regulation of PP1 and PP2A peptides.

Motif analysis in mCort^PPP1R17 versus mCort^GFP cells also identified preferential hyperphosphorylation of peptides with the xxSPxx motif (Fig. 3 C, Bottom panel; SI Appendix, Fig. S11 C and D). mCort^PPP1R17 cells exhibited enrichment of phosphoproteins similarly associated with mitotic spindles, RNA splicing, and G2-M checkpoint regulation (Dataset S12). We interrogated phosphoproteomic modulation of known tumorigenic pathways with western blotting. mCort cells transiently transfected using PPP1R17 lentiviral plasmids showed elevated pAKT (P = 0.002), pERK (P = 0.001), and pS6 (P = 0.009) compared to mCort cells transfected with GFP (Fig. 3 E, Left panel). In AtT-20 cells, a model for CD (52), transient transfection using PPP1R17 lentiviral plasmids similarly resulted in elevated pAKT (P = 0.04) and pS6 (P = 0.02) compared to GFP controls (Fig. 3 E, Right panel).

PPP1R17 Overexpression Promotes a Transcriptional Upregulation Program in CD.

In addition to its role as a cytoplasmic phosphatase, PP2A can dephosphorylate nuclear RNA polymerase II via the INTAC complex and globally attenuate transcription (53). We therefore hypothesized that chronic PPP1R17 overexpression activates RNA PolII, resulting in increased transcription and differential protein abundance in CD adenomas. mCort cells stably overexpressing PPP1R17 (mCort^PPP1R17) showed increased pRNA PolII (S2, S5, S7) and increased total RNA PolII compared to mCort^GFP cells (b) suggesting transcriptional derepression. Bulk RNAseq analysis of mCort^PPP1R17 compared to mCort^GFP cells revealed upregulation of genes associated with MYC activation, unfolded protein response, and epithelial–mesenchymal transition (Fig. 4B and Dataset S13), pathways previously associated with CD adenomas (19, 28, 54). We observed significant overlap between genes overexpressed in human CD adenomas and mCort ^PPP1R17 cells, especially in genes involved in protein translation and peptide biosynthesis (Dataset S14).

Fig. 4.

PPP1R17 overexpression in vitro recapitulates tumorigenic programs. (A) Representative western blots evaluating phosphorylation of RNA PolII (S2, S5, and S7) as well as total RNA PolII in mCort^PPP1R17 versus mcort^GFP cells. (B) Geneset enrichment (GSEA) analysis of bulk RNAseq comparison of mCort^PPP1R17 (Left) versus mCort^GFP cells (n = 3/group) showing differential enrichment of several pathways. (C) Proteomic analysis of mCort^PPP1R17 versus mCort^GFP cells (n = 5/group). (D) GSEA analysis of peptides differentially abundant in mCort^PPP1R17 (Left) versus mCort^GFP cells (n = 5/group) highlighting enrichment of pathways for hypoxia, protein secretion, and epithelial–mesenchymal transition. (E) Proteomic analysis of CD adenomas versus their adjacent margin (n = 3/group). (F) GSEA analysis of peptides differentially abundant in CD adenomas (Left) versus adjacent margin (n = 3/group) highlighting enrichment of pathways for epithelial–mesenchymal transition, hypoxia, and protein secretion.

Proteomic analysis of mCort^PPP1R17 compared to mCort^GFP cells identified upregulation of peptides involved in epithelial–mesenchymal transition, hypoxia, and protein secretion (Fig. 4 C and D and Dataset S15). Proteomic analysis of CD adenomas versus their adjacent margin tissues similarly revealed enrichment of proteins involved with epithelial–mesenchymal transition, hypoxia, and protein secretion (Fig. 4 E and F and Dataset S16), suggesting that PPP1R17 overexpression potentiates a hormone secreting tumor program.

PPP1R17 Overexpression Is a Targetable Mechanism of Adenoma Pathogenesis.

PP2A agonists interrupt the interaction between PP2A and its endogenous inhibitors, a strategy that may be of clinical benefit (55–57). However, this strategy has not been explored in the context of hormone-secreting adenomas. We tested PP2A agonists DT061, ABL127, and fingolimod (also called FTY270) in mCort^PPP1R17 cells (0.1 to 10 µM) and found that fingolimod suppressed POMC transcription (Fig. 5A). In AtT-20 cells, anti-PPP1R17 and anti-PP2Ac (catalytic subunit for the PPP2A complex) coimmunoprecipitation (IP), confirmed the interaction of PPP1R17-PP2Ac (Fig. 5B). Additionally, we found evidence of minimal PP1 interaction with PPP1R17 in AtT-20. Next, we confirmed that fingolimod treatment prevented PPP1R17 binding to PP2Ac, but not to PP1 (Fig. 5B) in AtT-20 cells. Fingolimod (2.5 µM; 72 h) inhibited cell proliferation in mCort^PPP1R17 cells (Fig. 5C).

Fig. 5.

PPP1R17 upregulation is a targetable mechanism in CD. (A) RT-qPCR of POMC normalized to ACTBin mCort^PPP1R17 treated with fingolimod, DT061, or ABL127 (10 µM each; six replicates; ***P < 0.05). (B) IP in AtT-20 cells using anti-PPP1R17 and anti-PP2A antibodies, with and without fingolimod (5 µM, 6 h). Western blots shown for PPP1R17, PP2Ac, and PP1-alpha (three replicates). (C) Proliferation of mCort^PPP1R17 cells with or without fingolimod (2.5 µM, 72 h, n = 8/group). (D) Melting temperatures of purified recombinant human PPP1R17 and PP2Ac in the presence and absence of fingolimod (0.1 to 10 uM), determined by Thermofluor assay (four replicates per condition; box-and-whisker plots). (E) Melting temperature of PPP1R17 in the presence of increasing PP2Ac concentrations (0 to 2.5 µM), with or without fingolimod (3 µM; n = 4 replicates per condition). Shaded regions represent 95% CI. (F) Bar plot with SE bars for baseline phosphatase activity of PP2A holoenzyme, and following addition of okadaic acid (OA; 5 nM), purified, recombinant PPP1R17 (0.1 µg/assay), and PPP1R17+fingolimod (5 µM) (six replicates, ANOVA P < 0.05; post hoc Tukey *P = 0.01, ****P < 0.001). (G) Western immunoblotting of phosphorylation status of PP2A targets (AKT, ERK, and S6) in mCort^PPP1R17 versus mCort^GFP cells with and without fingolimod treatment (2.5 µM for 24 h; representative of three biological replicates). Quantitation of three biological replicates represented with bar plot (ANOVA; post hoc Tukey test; *P = 0.01, **P = 0.001, ***P = 0.0004). (H) Western immunoblotting for phosphorylation of PP2A targets (AKT, ERK, and S6) in AtT-20^PPP1R17 versus AtT-20^GFP cells with and without fingolimod treatment (2.5 µM for 24 h; representative of three biological replicates). Quantitation of three biological replicates represented with bar plot (ANOVA; post hoc Tukey test; *P = 0.04, **P = 0.001, ***P = 0.0004). (I) Proliferation of AtT-20^PPP1R17 versus AtT-20^GFP cells after treatment with fingolimod (5 µM) or DMSO control (n = 8/group; ***P < 0.001, ****P < 0.0001). (J) Representative photograph of AtT-20 xenograft tumors from athymic NU/NU mice treated with fingolimod (1 mg/kg i.p.) or normal saline for 17 d. (K) Tumor weights following fingolimod or saline treatment (n = 5/group; *P < 0.05). (L) Tumor volumes over 17 d in fingolimod- versus saline-treated mice (n = 5/group; **P < 0.05).

We purified recombinant human PPP1R17 and PP2Ac and performed Thermofluor protein thermal shift assays to confirm the mechanism of fingolimod action. We titrated purified PPP1R17 and obtained reliable melting temperature measurements at 1 μg of protein. Fingolimod treatment decreased the melting temperature of PP2A (>2.5 C; P < 0.01 at 0.1 µM fingolimod) but not PPP1R17 (Fig. 5D). PP2Ac resulted in a concentration-dependent decrease in the PPP1R17 melting temperature, with a half-maximal effect between 1.5 and 2.5 μg of PP2A (Fig. 5E), verifying direct interaction between PPP1R17 and PP2Ac. This thermal shift was attenuated by addition of fingolimod, a competitive PP2A agonist (Fig. 5E). Phosphatase activity of recombinant PP2A holoenzyme (0.1 µg/assay) in buffer was inhibited by okadaic acid (5 nM; P < 0.001), fingolimod (5 µM), and by purified, recombinant PPP1R17 (0.1 µg/assay; P < 0.001). Adding fingolimod (5 µM) rescued PP2A phosphatase activity (P < 0.001) (Fig. 5F).

Fingolimod was originally discovered to antagonize sphingosine 1 phosphate receptors (S1PR1-5), leading to immune cell sequestration in lymph nodes, and notably immunosuppression. To dissect the dominant mode of fingolimod action in tumor corticotrophs (S1PR inhibition vs. PP2A agonism), we first performed a transcriptomic survey for S1PR1-5, and found minimal expression in pituitary core/margin datasets (SI Appendix, Fig. S13 A and B). mCort^PPP1R17 cells treated with fingolimod, or OSU-2S, a nonphosphorylatable derivative lacking S1PR activity but retaining PP2A agonism, showed that both compounds suppressed proliferation, though fingolimod produced a substantially greater effect (SI Appendix, Fig. S13 C and D). Consistent with a PP2A-dependent component, siRNA-mediated knockdown of PP2Ac partially rescued proliferation in fingolimod-treated mCort^PPP1R17 cells and in primary Cushing’s disease tumor cultures (SI Appendix, Fig. S13 E and F). Together with the negligible S1PR expression in tumor corticotrophs, these data indicate that fingolimod’s antiproliferative activity in this context is S1PR-independent and requires PP2A, but that PP2A agonism alone is insufficient to account for its full effect.

Phosphoproteomic analysis showed that treatment of mCort^PPP1R17 cells with fingolimod selectively decreased phosphorylation of peptides with the xxSPxx motif (SI Appendix, Fig. S14A). Dephosphorylated peptides were involved in cell cycle regulation and RNA polymerase II transcription (SI Appendix, Fig. S14B and Dataset S17). We next interrogated the specific peptides that fingolimod reversed the phosphorylation following PPP1R17 overexpression (hypo-hyper-hypo phosphorylation pattern). Fingolimod reversed PPP1R17-mediated hyperphosphorylation in several proteins (SI Appendix, Fig. S14C) involved in MYC signaling and G2-M checkpoint regulation (SI Appendix, Fig. S14D and Dataset S17). We confirmed that fingolimod treatment reversed AKT (pAKTS473; P = 0.0004) and ERK (pERK1/2T202/2024; P = 0.004) hyperphosphorylation in mCort^PP1R17 cells (Fig. 5F) and induced apoptotic markers Bid, Bbc3, and Hrk (SI Appendix, Fig. S14E).

As in mCort cells, stable PPP1R17 overexpression in AtT-20 cells induced tumorigenic signaling (Fig. 5H) and proliferation compared to AtT-20^GFP cells (Fig. 5I), and both were reversed by 5 μM fingolimod treatment. Specifically, we found a statistically significant reduction in phosphorylated AKT (P = 0.001). Since parental AtT-20 cells have a tumor-like phenotype, we found that fingolimod also reduced the viability of AtT-20^GFP cells. We generated AtT-20 xenografts in athymic NU/NU mice which resulted in adrenal hypertrophy (SI Appendix, Fig. S18A). Fingolimod treatment (1 mg/Kg) led to a significant decrease in tumor size (Fig. 5J), tumor weights (100 vs. 249 mg, P = 0.05; Fig. 5K), and measured volumes (final volume 168 vs. 463 mm³, P = 0.05; Fig. 5L and SI Appendix, Fig. S15B). There were no significant changes in skin thickness or body weights in tumor-bearing mice treated with fingolimod versus saline control for 14 d (SI Appendix, Fig. S15C). Taken together, these findings identify PP2A agonism as a therapeutic strategy for CD adenomas.

Discussion

PPP1R17 Is an Endogenous PP2A Inhibitor in Pituitary Adenomas Causing Cushing’s Disease.

Our findings demonstrate that epigenetic dysregulation of the kinase-phosphatase balance can underlie human tumors. In this study, we identified epigenetic reactivation of PPP1R17, an endogenous inhibitor of the protein phosphatase PP2A, as a driver of CD tumorigenesis. PPP1R17 activity was associated with widespread dysregulation of the phosphoproteome. Reversal of this phenotype using the PP2A agonist fingolimod rescued cellular hyperproliferation and induced apoptosis in a tumor model. Kinase-phosphatase imbalance due to somatic gene amplification of PRKAR1A has been implicated in adrenal tumors causing Cushing’s syndrome (58). Here, we find convergent epigenetic dysregulation along a different pathway leading to kinase-phosphatase imbalance in pituitary adenomas causing Cushing’s disease.

Epigenetic modulation of gene expression is a tumorigenic mechanism whereby environmental cues impinge upon the transcriptome without requiring changes in the DNA coding sequence (59). The majority of CD adenomas are not associated with canonical mutations of known targets BRAF, USP8, or USP48 (22, 60), and studies on CD epigenomics are lacking. Our multiomic approach combining single-nucleus RNA and ATACseq with DNA methylation profiling identified epigenetic PPP1R17 reactivation in CD adenomas, highlighting a role for gene–environment interactions in CD. Our reverse-ChIP results suggested histone subunit recruitment and active chromatin remodeling at the PPP1R17 promoter, which agrees with other reports on histone partitioning as a mechanism of tumor progression (61). Further studies are needed to determine precisely how environmental factors modify histone recruitment in CD.

PPP1R17 Reactivation as a Targetable Mechanism of CD.

PPP1R17, also known as G-substrate due to its preferential phosphorylation by cGMP-dependent protein kinase, is found almost exclusively in human cerebellar Purkinje cells where it plays a role in the induction of long-term depression (7, 45). In the rat brain, Ppp1r17 is also restricted almost entirely to Purkinje cells of the cerebellum, although low levels are detectable in the paraventricular region of the hypothalamus and in the pons/medulla (39). Unlike in adults, PPP1R17 is highly expressed in neural progenitor cells in the developing human cortex (62–64). This expression pattern is restricted to mammals and is not seen in ferrets or mice of comparable ages (65).

In the prenatal macaque and human cortex, PPP1R17 is localized to the outer and inner subventricular zones, and colocalizes with markers of intermediate progenitor cells and dividing cells (65). RNA-seq time-course studies also confirmed that PPP1R17 expression is restricted to neural progenitor cells during fetal development in humans (38), therefore its overexpression in CD likely reflects epigenetic reactivation through coordinated increase in chromatin accessibility and DNA hypomethylation. Other studies have recently described similar epigenetic reactivation of developmental genes in the pathogenesis of pulmonary hypertension (66), however this reactivation has not previously been described in pituitary tumors.

The proximate mechanism of PPP1R17 reactivation in CD adenomas is still not fully understood. Whereas our results suggest underlying epigenetic dysregulation via promoter hypomethylation, and increased chromatin accessibility, the underlying index event is still unclear. It is notable that the pituitary gland resides outside the blood–brain barrier adjacent to the median eminence with its leaky vasculature, which allows for prompt homeostatic responses to environmental stressors. This architecture also allows transmission of blood-borne pathogens and other environmental agents which trigger hormone responses from the hypothalamus and pituitary gland. Given that epigenomic changes frequently occur in response to environmental cues, we speculate that cumulative exposure to environmental insults or infectious agents could create an epigenomic milieu favoring tumorigenesis via PPP1R17 dysregulation. Additional studies are needed to investigate this hypothesis.

PPP1R17 Overactivity Induces Phosphoproteomic Changes.

PP2A is a member of the family of serine/threonine phosphoprotein phosphatases (PPPs), which unlike metal-dependent phosphatases (PPMs) or protein tyrosine phosphatases (PTPs), utilize conserved motifs for metal-dependent catalysis. PPPs include PP1, PP2A, PP2B, PP4, PP5, PP6, and PP7, and often consist of holoenzymes with regulatory subunits which control their location and specificity. While PP2A inactivation is linked to uncontrolled growth in several cancers (67), aberrant PP1 signaling also contributes to tumorigenesis (68). Several other PPPs are upregulated in the tumor context, and associated with poor prognoses (69), suggesting context-specific activation. Our data highlight the role of PP2A inhibition in CD; however, a role for other protein phosphatases cannot be excluded. PPP1R17 was originally characterized as a PP1 inhibitor, and we cannot rule out that concurrent PP1 inhibition contributes to the phenotypes we observe. We focused on PP2A here given its established role as a tumor suppressor and the convergence of our phosphoproteomic and pharmacologic data on PP2A-regulated substrates. Further delineating the relative contributions of PP1 and PP2A to PPP1R17-driven tumorigenesis will require phosphatase-selective tools.

Overexpression of PP2A inhibitors has been identified as an oncogenic mechanism in multiple human cancers, including SET overexpression in CML, AML, T cell ALL, B cell CLL, non–small cell lung cancer and Wilms tumors (48, 70–72), and CIP2A overexpression in hepatocellular, breast, colorectal, ovarian, cervical, prostate, lung, head, and neck cancer (49, 73–75). Small molecules that activate PP2A (SMAPs) are an emerging class of antitumor agents with therapeutic potential as combinatorial agents in lung cancer models (76, 77), as well as in multiple myeloma, leukemia, and colorectal cancer (57). Our results show the therapeutic potential of PP2A agonism using fingolimod in CD both in vitro and in vivo. Interestingly, PPP1R17 overexpression in mouse cortical neurospheres inhibits the G1-S transition and results in decreased proliferation (65). In our study, PPP1R17 overexpression in mature corticotrophs (mCort) similarly resulted in prolonged G0/G1 but led to accelerated S phase and increased proliferation in mCort^PPP1R17 cells, indicating cellular context-specific effects.

In addition to its role in protein dephosphorylation, PP2A was recently shown to exert direct transcriptional control by dephosphorylating RNA PolII via an integrator complex (53, 57, 78). We noted phosphorylation of RNA PolII in mCort^PPP1R17 cells, with resultant differential expression of several genes. Furthermore, we noted significant overlap between the gene expression profile of mCort^PPP1R17 cells and CD adenomas, indicating that several transcriptomic changes in CD likely result from PPP1R17-mediated PP2A inhibition. Future studies will investigate whether PPP1R17 may be partly mediating kinase-phosphatase imbalance through its interactions with PRKACA, which plays a role in the formation of adrenal tumors (79).

Limitations.

Our study is limited by the small number of patient samples included, partly due to limited availability of en-route adult human pituitary tissues in the tumor margin. Fingolimod also possesses PP2A-independent effects, including its effects on immune cells (80), which may limit its therapeutic window for patients with CD. Beyond S1PR engagement, unphosphorylated fingolimod inhibits sphingosine kinase 1 and ceramide synthases and downregulates nutrient transporters, among other activities. While our data indicate that PP2Ac is required for the antiproliferative effect of fingolimod in CD and PPP1R17 overexpression context. However, the incomplete phenocopy by the nonphosphorylatable, S1PR-inactive analog OSU-2S suggests that PP2A agonism alone does not fully account for fingolimod‘s activity. We conclude that additional S1PR- and PP2A-independent mechanisms likely contribute to the observed pharmacological effect. Even as these findings do not affect our central conclusion that PPP1R17-mediated PP2A inhibition drives CD tumorigenesis, fingolimod should be regarded as a proof-of-concept agent rather than a ready-for-clinic pharmacologic agent for CD.

Conclusions

Our study highlights the role of kinase-phosphatase disequilibrium in the pathogenesis of benign tumors, specifically in adrenocorticotropin secreting pituitary adenomas that cause CD. We identified epigenetic reactivation of PPP1R17, an endogenous inhibitor of PP2A, as a central driver of CD tumorigenesis irrespective of the somatic mutation status. The therapeutic potential of targeting this imbalance is underscored by the efficacy of PP2A agonism in reversing the tumorigenic effects of PPP1R17 overexpression in vitro and in vivo. This work advances our understanding of pituitary adenoma formation and suggests a promising avenue for therapeutic intervention in benign tumors.

Materials and Methods

Patients.

Surgical samples were obtained from nonconsecutive patients with a biochemical diagnosis of CD undergoing pituitary surgery by a single surgeon (PC) between 2006 and 2023 (Dataset S1). This study was conducted at the NIH Clinical Center in Bethesda, MD, and approved by the combined neuroscience Institutional Review Board of the National Institute of Neurological Disorders and Stroke, Bethesda, MD (clinicaltrials.gov identifier NCT00060541). Each patient gave written informed consent.

Hypercortisolism in CD patients was diagnosed as previously described (81, 82). All patients underwent a sublabial trans-sphenoidal approach for microscope-assisted selective adenectomy. Adenomas were identified preoperatively using high-resolution MRI with Pituitary protocol, and their locations were verified intraoperatively by visual inspection. Adenomas were removed by microsurgical dissection along a circumferential pseudo-capsular plane. Pituitary adenoma was removed en-bloc within its pseudocapsule wherever possible (83). The en-route normal pituitary gland was removed separately and annotated for analysis. Research specimens were labeled separately and fresh-frozen for further study.

Animal Studies.

All procedures were approved by National Institute of Neurological Disorders and Stroke, Animal Care and Use Committee, and performed in accordance with NIH guidelines. Adult pregnant NIH Swiss female mice were euthanized at two gestation ages, embryonic day (E) 14.5 and E18.5 according to the procedure approved by the NIH Animal Care and Use Committee. Embryos and pituitaries of the adult dams were collected, fixed in 4% formaldehyde (Sigma)/0.1 M phosphate buffer saline (PBS, 24 h), transferred to 30% sucrose (48 h, 4 °C), embedded in OCT (Sakura Tissue-Tek) and stored at −80 °C until sectioning. Sections (pituitary, 12 µm; E14.5 and E18.5, 16 µm) were cut on a cryostat (Leica Biosystem CM 3050S).

For in vivo tumor injections, 4- to 5-wk-old female athymic nude mice (NU/NU) were purchased from Charles River Laboratories and housed under standard conditions in a pathogen-free facility with ad libitum access to food and water. Mice were grouped into untreated mice, mice injected with cells, and mice injected with cells and treated with fingolimod. Number of mice per group were determined based on power calculations assuming effect size of 30% based on in vitro data. Cell suspensions of 7 × 10⁶ cells in 100 mL PBS were injected into the flank subcutaneously using 25G needles.

After allowing 10 to 12 d for tumor establishment, tumor-bearing mice were randomized into DMSO versus fingolimod treatment groups. Fingolimod was dissolved in normal saline and injected intraperitoneally at a daily concentration of 1 mg/Kg body weight. Mice were monitored daily for clinical signs of distress by dedicated animal handlers at the NIH Animal Research Facility in keeping with NIH IACUC guidelines. A licensed veterinarian from the NIH was on call 24 h for any animal emergencies. At the end of the treatment period, surviving mice were euthanized humanely using carbon dioxide according to AVMA Guidelines for the Euthanasia of Animals and NIH IACUC guidelines.

Experimental Methods (Additional Details in SI Appendix, Methods).

Corticotroph cells and constructs.

Corticotroph cells (mCort) were harvested as previously described (28) from female BALB/c mice aged 6 to 8 wk (Taconic Biosciences, USA) (33) Lentiviral vectors encoding PPP1R17-GFP or GFP-only were obtained from Origene (Cat# RC207007L4V and PS100093). 10 IFUs/5 × 10⁶ cells were applied in normal media to mCORT or AtT-20 cells for 48 h, followed by antibiotic selection for 2 wk using 1 mM Puromycin (Thermo Fisher #A1113802). GFP positivity was confirmed in surviving cells, and PPP1R17 expression was verified using RT-qPCR and Western immunoblotting.

Single nucleus multiome analysis.

CD adenomas and en route normal pituitary glands were embedded in OCT and stored at −80 °C. Nuclei were isolated from the frozen tissues using an adaptation of a previously established protocol (84). Nuclei were then immediately processed using a Single Cell Multiome ATAC+Gene Expression Assay (10× Genomics # 1000285) following manufacturer recommendations. Nuclei were loaded onto a Next GEM Chip J (10× Genomics # 1000230), targeting a yield of 3,000 to 6,000 nuclei. Sequencing reads were demultiplexed and aligned to the hg38 reference genome using the 10× Genomics CellRanger software (arc v2.0.0) mkfastq function with default settings, and counts were generated using the CellRanger-arc count function. Please see SI Appendix, Methods for the bioinformatics approaches to data analysis and interpretation.

DNA methylation.

5 μm tissue sections were cut from formalin-fixed paraffin-embedded surgically derived specimens from Cushing’s disease adenomas. Autopsy-derived human pituitary gland sections were used as controls as previously published (28). DNA was isolated and purified (Qiagen DNeasy #69556). After sodium bisulfite conversion using the Zymogen EZ DNA Methylation kit (#D5001), 1 μg each of methylated and unmethylated DNA was analyzed using the Illumina Infinium MethylationEPIC chip (#20087706). Data were processed and analyzed using the MissMethyl R package. For the PPP1R17 promoter, 2 μL of eluted DNA was used for each PCR reaction, with methylated and unmethylated primers. RT-qPCR was performed using the Sso Universal IT SYBR green supermix (BioRad #1725271). Beta values were quantified using the ratio of methylated/(methylated+unmethylated) expression.

Bulk RNAseq.

Total RNA was extracted from stable cell lines using the RNeasy Mini Kit (Qiagen, #74004). PolyA+ mRNA isolation, size fragmentation, cDNA synthesis, size selection, and next gen sequencing were performed at the National Intramural Sequencing Center, Bethesda, MD, according to their standard protocols, as described previously (28). The Illumina TruSeq RNA Sample Prep V2 Kit was used according to the manufacturer’s instructions. Analysis of bulk RNA sequencing data was performed using the EdgeR package on total exon counts per gene. A paired analysis was performed with the GLM-based approach with design matrix including sample type.

rChIP.

rChIP was performed by modifying a previously published protocol (85). Negative controls included the dCas9-3XFLAG-Biotin Protein, however the synthetic guide RNAs for promoter-specific targeting were excluded. Purification and reversal of cross linking was performed using an EZ-Magna ChIP kit (Millipore Sigma #17-10086). To validate locus-specific pulldown, an aliquot of the eluent was treated with proteinase K (Millipore #20-298) and PCR was run using primers specific to the PPP1R17 promoter. Protein fragments were then purified and treated with RNAse A (Millipore #20-297) and quantified using Mass spectrometry (SimulTOF 300).

Proteomic and phospho-proteomic profiling.

LC–MS/MS data acquisition was performed on a Thermo Scientific Orbitrap Lumos mass spectrometer which was coupled to a Thermo Scientific Ultimate 3000 HPLC. Pairwise analysis of total proteins and phosphopeptide/total protein ratios were computed for each tumor-normal pair and compared using two-sample tests of proportions with FDR corrections for multiple testing as appropriate. Phosphoprotein motif analysis was performed using the MEME suite motif-based sequence analysis tool (86).

Multiplex immunocytochemistry.

5 µm-thick paraffin-embedded formalin-fixed human sections were imaged using an Axio Imager.Z2 slide scanning epifluorescence microscope (Zeiss) equipped with a 20×/0.8 Plan-Apochromat (Phase-2) nonimmersion objective (Zeiss). Pseudo-colored stitched images were exported to Adobe Photoshop and overlaid as individual layers to create multicolored merged composites.

Immunohistochemistry.

Paraffin embedded tissue blocks were sectioned (5 μm) and stained with an automated immunostainer (Bond-Max, Leica). Sections were imaged on a Nikon Eclipse Ni microscope (Nikon) equipped with a Retiga EXi Fast1394 camera (QImaging) or with a Nikon Eclipse TE200 spinning disk microscope (Nikon) equipped with an EMCCD ImageM digital camera (Hamamatsu), using the iVision software (BioVision).

Spatial quantification of mIHC coexpression.

Nuclear segmentation of multiplex immunohistochemistry TIFF files was accomplished using “Cell Detection” in Qupath on the DAPI channel. Exported single-cell measurements were imported into Python using pandas and structured using anndata. Tumor regions were defined computationally by calculating a per-cell TumorScore as the mean expression of CIP2A, CYCLIN D1, EGFR, HK2, Kip1, LDHA, NOXA, PCSK1, PFKFB3, PME1, POMC, RAB3B, S100B, SET, SOX2, SOX9, TBX19, cMyc, and RBC, excluding PP2AC and PPP1R17. An adaptive intensity threshold was determined using Otsu’s method from sklearn and cells exceeding this threshold were designated as tumor candidates. Data visualization, including threshold histograms and grouped bar plots with significance annotation, was performed using matplotlib and seaborn. Figures were constructed using Adobe Illustrator.

RNA extraction and quantitative real time PCR (RT-qPCR).

RNA was extracted from both human and mouse cells using an RNeasy Mini Kit (Qiagen #74004) according to the manufacturer’s instructions. Complementary DNA (cDNA) libraries were constructed using SuperScript III for RT-qPCR (Invitrogen Life Technologies). RT-qPCR was performed using the Sso Universal IT SYBR green supermix (BioRad #1725271). Relative gene expression was calculated using the DDCt method with beta actin or GAPDH as housekeeping genes.

Cell cycle analysis.

Cell cycle was analyzed using a Click-iT Plus EdU Alexa Fluor 647 Flow Cytometry assay kit (Thermo Fisher #C10635) following the manufacturer’s instructions. Proportions of cells in the S-phase were compared between samples using two-sample tests of proportions, with P < 0.05 considered statistically significant.

Cell viability assays.

Cell proliferation assays were performed using CellTiterGlo (Promega #G7571) according to the manufacturer’s instructions. Plates were incubated for 30 min at room temperature prior to recording luminescence on a Synergy Neo2 microplate reader (BioTek). Cells were treated with pharmacologic agents and/or siRNA as described in figure legends and in SI Appendix, Methods.

PLA.

Proximity between PPP1R17 and PP2Ac was determined using a Duolink PLA Red Mouse/Rabbit kit (Sigma #DUO92101) following the manufacturer’s instructions. 5 μm human CD adenoma sections were processed and mounted using Duolink in-situ mounting media with DAPI and imaged.

Recombinant PPP1R17.

Human PPP1R17 was cloned into pET FLAG vectors, transformed using BL21 gold competent bacterial cells (Sigma) and amplified. After amplification, FLAG-tagged PPP1R17 was purified using a HiTrap nickel column (Sigma).

IP.

IP was performed according to the Cell Signaling Technology Immunoprecipitation Protocol using antibodies against PPP1R17 (#40165), PP2Ac (#2259), or normal IgG (#3900). Antibodies were incubated overnight with the lysates at 4 °C under continuous rotation to allow formation of immune complexes. IgG immunoprecipitation and bead-only controls were included as negative controls. A recommended secondary antibody (#S5127) was used to minimize detection of IgG heavy and light chains and prevent interference with target protein detection.

Thermal shift assay.

Protein thermal shift assays were performed on 384-well plate or a Quantabio qPCR instrument. The plate was subjected to thermal shift by ramping the temperature from 25 °C to 99 °C at 0.05 °C increments per second. The relative fluorescence emitted by the thermal shift dye was recorded during the temperature ramp phase and plotted versus temperature. The Tm values were plotted as a function of TDP43 concentration in GraphPad Prism, and the inflection point of the curve determined the binding affinity of ligand to binding protein. The experiments were completed with n = 4 replicate samples per treatment. Each experiment was repeated three times.

Phosphatase assay.

Recombinant human PPP2R1A/PP2A-alpha complex (Abcam Cat # ab271717; 0.1 µg/assay) was assayed in manufacturer’s recommended phosphatase buffer, using recommended techniques. For inhibitor-treated samples, the enzyme complex was preincubated with okadaic acid (5 nM), purified, recombinant PPP1R17, (0.1 µg/assay) or fingolimod (5 µM) prior to initiation of the reaction.

AlphaFold multimer.

AlphaFold multimer was designed specifically to predict the structures of protein complexes more accurately than AlphaFold and AlphaFold2 (50). We generated protein structure and interaction inference using the AlphaFold2 algorithm (82). Briefly, FASTA files with the canonical protein sequence (PP2A, PPP1R17, and PME1) were analyzed on the NIH HPC cluster (BioWulf). The AlphaFold2 algorithm preferentially used existing experimental PDB structures where available. Multiple sequence alignment and model predictions were then generated for protein interaction inference and multimer models. Multimer model prediction alignment error rates, and pLDDT values were further analyzed with the alphapickle package (M. J. Arnold. 2021. AlphaPickle. doi.org/10.5281/zenodo.5708709). The top ranked multimer models were used to analyze protein interaction sites, and to generate the figure panels.

Statistical analysis.

P values < 0.05 two-tailed were considered statistically significant. Statistical analyses were performed using STATA 14/IC (StataCorp LP, College Station, TX) or GraphPad Prism 9.0 (GraphPad Software, La Jolla, CA) or Python using statsmodels or scipy packages.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Dataset S03 (XLSX)

Dataset S04 (XLSX)

Dataset S05 (XLSX)

Dataset S06 (XLSX)

Dataset S07 (XLSX)

Dataset S08 (XLSX)

Dataset S09 (XLSX)

Dataset S10 (XLSX)

Dataset S11 (XLSX)

Dataset S12 (XLSX)

Dataset S13 (XLSX)

Dataset S14 (XLSX)

Dataset S15 (XLSX)

Dataset S16 (XLSX)

Dataset S17 (XLSX)

Movie S1.

Annotation and removal of enroute normal pituitary gland during transsphenoidal surgery. A. (Timestamp 00:07s) Clinical presentation with coronal 3D T1 weighted MRI image with gadolinium contrast. B. (Timestamp 00:12s) Sagittal image as in A. C. (Timestamp 00:18s) Schematic representation of macroadenoma approach as viewed with surgical microscope. D. (Timestamp 00:19s) Start of surgical video. Removal of bone overlying sella turcica. E. (Timestamp 00:28s) Schematic representation of dural opening. F. (Timestamp 00:30) Dural incision and elevating dural flap. G. (Timestamp 00:47s) Schematic of pituitary capsule incision in schematic. H. (Timestamp 00:49s) Incision of anterior pituitary capsule and removal of anterior normal en-route pituitary gland. I. (Timestamp 01:03s) Schematic of tumor debulking technique. J. (Timestamp 01:05s) Debulking and removal of main tumor mass. K. (Timestamp 01:14s) Schematic representation of technique for removal of the tumor with a pseudo-capsular plane. L. (Timestamp 01:17s) Removal of tumor by dissecting along the pseudocapsular plane. M. (Timestamp 01:26s) Schematic representation of removal of main tumor mass. Video ends at 01:30s.

Data, Materials, and Software Availability

Single cell RNAseq single nucleus multiome (RNAseq + ATACseq) data have been deposited in GitHub GEO Gene Expression Omnibus GEO GSE208112 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE208112) (87) and GEO accession: GSE325789 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE325789) (88). Patient-related data generated in clinical protocols will be available as anonymized data upon reasonable request to the Principal Investigator. Custom analysis toolkits and code developed for this study has been uploaded to Github (https://github.com/chittiboina-lab/PPP1R17_pituitary_adenoma) (89).