Targeting zebrafish and murine pituitary corticotrophtumors with a cyclin-dependent kinase (CDK) inhibitorNing-Ai Liua, Hong Jianga, Anat Ben-Shlomoa, Kolja Wawrowskya, Xue-Mo Fanb, Shuo Linc, and Shlomo Melmeda,1

Departments of aMedicine and bPathology, Pituitary Center, Cedars-Sinai Medical Center, Los Angeles, CA 90048; and cDepartment of Molecular, Cell andDevelopmental Biology, University of California, Los Angeles, Los Angeles, CA 90024

Edited by Wylie W. Vale, The Salk Institute for Biological Studies, La Jolla, CA, and approved April 7, 2011 (received for review December 3, 2010)

Cushing disease caused by adrenocorticotropin (ACTH)-secretingpituitary adenomas leads to hypercortisolemia predisposing todiabetes, hypertension, osteoporosis, central obesity, cardiovascularmorbidity, and increased mortality. There is no effective pituitarytargeted pharmacotherapy for Cushing disease. Here, we generatedgermline transgenic zebrafish with overexpression of pituitarytumor transforming gene (PTTG/securin) targeted to the adenohy-pophyseal proopiomelanocortin (POMC) lineage, which recapitu-lated early features pathognomonic of corticotroph adenomas,including corticotroph expansion and partial glucocorticoid resis-tance. Adult Tg:Pomc-Pttg fish develop neoplastic coticotrophs andpituitary cyclin E up-regulation, as well as metabolic disturbancesmimicking hypercortisolism caused by Cushing disease. Early devel-opment of corticotroph pathologies in Tg:Pomc-Pttg embryos facil-itated drug testing in vivo. We identified a pharmacologic CDK2/cyclin E inhibitor, R-roscovitine (seliciclib; CYC202), which specificallyreversed corticotroph expansion in live Tg:Pomc-Pttg embryos. Wefurther validated that orally administered R-roscovitine suppressesACTH and corticosterone levels, and also restrained tumor growth ina mouse model of ACTH-secreting pituitary adenomas. Molecularanalyses in vitro and in vivo showed that R-roscovitine suppressesACTH expression, induces corticotroph tumor cell senescence andcell cycle exit by up-regulating p27, p21 and p57, and downregulatescyclin E expression. The results suggest that use of selective CDKinhibitors could effectively target corticotroph tumor growth andhormone secretion.

pituitary cell cycle | pituitary hormone | adrenal function

Despite being small (<2 mm) and often undetectable by MRI,pituitary corticotroph tumors are associated with significant

morbidities and mortality as a result of adrenal glucocorticoid(Gc) hypersecretion in response to autonomous tumor adreno-corticotropin (ACTH) production (1). The standard of care forCushing disease consists of transsphenoidal pituitary tumor re-section, pituitary-directed radiation, adrenalectomy, and/or med-ical suppression of adrenal gland cortisol production. Althoughtranssphenoidal ACTH-secreting tumor resection yields 30% to70% surgical cure rates, adenoma recurrence rate is high (2).Efficacies of other therapeutic modalities are limited by factorssuch as slow therapeutic response, development of pituitary in-sufficiency, and uncontrolled pituitary tumor growth in the settingof adrenal gland resection or inhibition (2, 3). Effective pharma-cotherapy directly targeting corticotroph tumor growth and/orACTH production remains a major challenge (4).The pituitary is highly sensitive to cell cycle disruptions (5, 6).

Pituitary tumors acquire oncogene and tumor suppressor geneticandepigenetic alterations,which result inunrestrainedproliferation,aberrant neuroendocrine regulatory signals, and disrupted humoralmilieu, mediated directly or indirectly by dysregulated cyclin-dependent kinases (CDKs) (5, 7). Although CDK gene mutationshave not readily been identified in human pituitary tumors, overex-pression of cyclins and dysregulation of CDK inhibitors are en-countered in pituitary adenomas, indicating the importance of CDKactivation for potential therapeutic targeting (8, 9). However, pre-clinical studies of CDK inhibitors are hampered by the requirementfor largedrugquantities andprolongeddurationof administration toobserve potential efficacy. Furthermore, although the genetic spec-

trum of tumor-associated mutations and/or their cellular contextmay dictate specific CDK dependence, it is difficult to predict whichCDK inhibitor(s) may be effective against particular tumor types invivo (10, 11). Therefore, animal models faithfully recapitulatinghuman pituitary tumors, which enable rapid and efficient in vivotesting, are needed to identify small molecule CDK inhibitors withoptimal potency.Regardless of cell lineage origin, pituitary tumors almost in-

variably overexpress pituitary tumor transforming gene (PTTG),which encodes a securin that binds separase in the APC complex,and governs faithful chromosome segregation during mitosis (12).PTTG was originally isolated from rat pituitary tumor cells (13).Dysequilibrium of intracellular PTTG abundance leads to cellcycle disruption and neoplastic formation, causing chromosomalinstability and aneuploidy, and also aberrant G1/S and G2/Mtransition by transcriptional dysregulation of cyclin expression(12, 14–19). On the contrary, PTTG overexpression also triggersirreversible cell cycle arrest in pituitary growth hormone (GH)-and gonadotropin (e.g., luteinizing hormone, follicle-stimulatinghormone)-expressing tumors by activating lineage-specific se-nescence pathways, contributing to the benign propensity of pi-tuitary tumors (12, 20).Here, we report the generation of a stable transgenic zebrafish

with zPttg overexpression targeted to pituitary proopiomelanocortin(POMC) lineages (corticotrophs andmelanotrophs). Tg:Pomc-Pttglarvae develop early pathologic processes reflective of corticotrophtumors including neoplastic corticotrophswith partialGc resistance,and hypercortisolemia-induced metabolic disturbances in adulttransgenicfish. Taking advantage of the early-observed corticotrophpathology, combined with pituitary POMC lineage-specific expres-sion of a fluorescent reporter in live transparent larvae, we testedsmall-molecule CDK inhibitors, which lead to identification ofR-roscovitine against PTTG-overexpressing corticotrophs. Inhibit-ory effects of R-roscovitine on corticotroph tumor cells were sub-sequently validated in an in vivo and in vitro mouse model,supporting use of selective CDK inhibitors as effective therapy forCushing disease.

ResultsStable Transgenic PTTG Overexpression Targeted to Pituitary POMCCells Rapidly Induces Early Pathologies of Cushing Disease. As aninitial step toward identification of novel targets for Cushingdisease therapy, we created a zebrafish model of pituitary cor-ticotroph tumors. Given the highly conserved zebrafish PTTGprotein sequence (Fig. S1), we hypothesized that zebrafish PTTGexhibits conserved properties involving cell cycle dysregulation inpituitary tumor formation (21). To test this hypothesis, we tar-geted PTTG overexpression to pituitary POMC lineages under

Author contributions: N.-A.L., S.L., and S.M. designed research; N.-A.L., H.J., and K.W.performed research; N.-A.L., A.B.-S., X.-M.F., S.L., and S.M. analyzed data; S.L. contributednew reagents/analytic tools; and N.-A.L. and S.M. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. E-mail: melmed@csmc.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1018091108/-/DCSupplemental.

8414–8419 | PNAS | May 17, 2011 | vol. 108 | no. 20 www.pnas.org/cgi/doi/10.1073/pnas.1018091108

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

5, 2

020

the control of the zPomc promoter. One- to two-cell stage embryoswere coinjected with transposase mRNA and a Tol2 transposoncassette flanking a zPomc proximal promoter fused to a full-lengthzPttg cDNA. Whole-mount in situ RNA analysis in F2 generationembryos confirmed zPttg overexpression, which temporally andspatially coincided with pituitary POMC cell ontogeny (Fig. 1A) (22).To investigate the effect of zPttg overexpression on embryonic

pituitary POMC lineage development, we analyzed highly con-served pituitary transcription factors as markers for both non-POMC (Pit-1) and POMC (Tpit/Tbx-19) pituitary lineages (23–25).At 2 d post fertilization (dpf), Tg:Pomc-Pttg larva demonstratedincreased pituitary Tpit/Tbx-19 expression, but Pit-1 expression wasnot altered (Fig. 1B), indicating early POMC lineage-specific ex-pansion. We also generated double transgenic embryos (Tg:Pomc-Pttg; POMC-eGFP) by breeding Tg:Pomc-Pttg zebrafish intoa previously established transgenic line, POMC-GFP, whereineGFP expression was targeted to pituitary POMC cells by the samezPomc promoter, thus representing a POMC lineage-specificmarker (22). Live double transgenic (Tg:Pomc-Pttg; POMC-eGFP)larvae exhibited POMC lineage expansion as evidenced by in-creased pituitary eGFP expression (Fig. 1C).Pituitary corticotrophs are a critical component of the

hypothalamic-pituitary-adrenal axis that mediates the stress re-sponse via hypothalamic corticotropin-releasing hormone (CRH)-stimulated, and subsequently pituitary ACTH-stimulated, adrenalgland Gc production. Gcs exert negative feedback on CRH andPOMC-derived ACTH expression and secretion to restore hypo-thalamic-pituitary-adrenal homeostasis following stress. In humancorticotroph tumors, ACTH hypersecretion is partially resistantto Gc-negative feedback regulation, further exacerbating uncon-trolled hypercortisolism (26). To investigate the integrity of theGc-negative feedback pathway in Tg:Pomc-Pttg corticotrophs, weexposed live zebrafish embryos to dexamethasone-containing cul-ture medium starting from 10 h post fertilization (hpf). Pituitary

eGFP expression was suppressed in POMC-GFP larvae exposed to10−7 M dexamethasone by 4 dpf but not in double-transgenic (Tg:Pomc-Pttg; POMC-eGFP) larvae, which only exhibited inhibitionof pituitary eGFP expression in response to 10 times higherdexamethasone concentrations (10−6 M; Fig. 1C), suggesting de-creased Gc sensitivity of Tg:Pomc-Pttg corticotrophs. Thus, Tg:Pomc-Pttg corticotrophs rapidly develop the hallmark pathology ofACTH-dependent Cushing disease within 4 d of embryonic de-velopment, i.e., partial Gc resistance.In adult Tg:Pomc-Pttg fish (20 mo of age), immunohistochem-

istry revealed overt neoplastic-appearing pituitary cells with a highnuclear/cytoplasmic ratio, distinct nucleoli, and basophilic cyto-plasm that stained strongly for ACTH in two of six Tg:Pomc-Pttgpituitary glands analyzed, morphologically resembling human pi-tuitary ACTH-secreting adenomas, whereas none of six WT pitu-itary glands showed a similar phenotype (Fig. 1D). WT zebrafishpituitary glands exhibited an overall PCNA index of 2.3 ± 0.9% vs.3.1 ± 1.3% in Tg:Pomc-Pttg (mean ± SE; P = 0.6), whereasACTH-producing cells in the Tg:Pomc-Pttg pituitary exhibitedincreased PCNA index compared with WT (2.8 ± 0.1% vs. 1.8 ±0.2%, mean ± SE; P = 0.05; Fig. 1E), suggesting altered G1/S inneoplastic corticotrophs as a result of zPttg overexpression.

Hypercortisolism andMetabolic Disturbance in Tg:Pomc-Pttg Zebrafish.We tested whether the observed neoplastic corticotroph cellchanges in Tg:Pomc-Pttg zebrafish lead to autonomous ACTHsecretion and subsequent hypercortisolism. Because we are tech-nically hampered frommeasuring plasmaACTH or serum cortisollevels by the very limited amount of blood obtainable from eachadult zebrafish (∼5 μL), we measured total cortisol content in age-and weight-matched Tg:Pomc-Pttg zebrafish and their transgene-negative siblings. At 3 mo of age, adult Tg:Pomc-Pttg fish showed40% increased cortisol content versus WT siblings (1.4 ± 0.2 vs.1.0 ± 0.2 μg/L/mg, n = 12 for each group, mean ± SE; P < 0.01).

Fig. 1. Pituitary pathology of zPttg transgenic zebrafish,Tg:Pomc-zPttg. (A) Top: Schematic representation of Pomc-zPttg transgene. Bottom: Pituitary expression of zPttg in Tg:Pomc-zPttg zebrafish at 72 hpf. F1 Tg:Pomc-zPttg transgeniczebrafish were crossed with WT zebrafish, resulting in F2embryos with 50% of the progeny positive (Left, Tg), and 50%negative (Right, WT) for pituitary zPttg expression assessed bywhole-mount in situ analysis (ventral view, with anterior as-pect to the left). (B) Tg:Pomc-Pttg embryos showed increasedTpit/Tbx19 expression, whereas no significant change of Pit-1expression by whole-mount in situ analysis at 48 hpf. Anti-sense mRNA probes are indicated at right lower corner of eachpanel. Top, Lateral view; Middle and Bottom, ventral view,anterior aspect to the left. (C) Tg:Pomc-Pttg; POMC-eGFPembryos exhibited increased pituitary eGFP signal, and aremore resistant to Gc-negative feedback than Tg:Pomc-Pttg–negative siblings. Double transgenic embryos (Tg:Pomc-Pttg;POMC-eGFP) were generated by breeding Tg:Pomc-Pttg fishwith an established transgenic line POMC-eGFP, wherein eGFPexpression is driven by the same zPomc promoter. Fluores-cence intensity of POMC-GFP–positive cells was measured inlive embryos after dexamethasone treatment at 4 dpf. (D)Pituitary hematoxylin/eosin stain (Top) and ACTH immuno-histochemistry (Bottom) of sections derived from WT and Tg:Pomc-Pttg transgenic fish (Tg) at 20 mo. Red arrows indicateneoplastic mitotic ACTH-expressing cells. (E) Tg:pomc-pttg pi-tuitary exhibited increased number of PCNA and ACTH coex-pressing cells. Representative confocal pituitary images offluorescence immunohistochemistry detecting PCNA (red) andACTH (green) expression in Tg:Pomc-Pttg (a–c) and WT (d–f)zebrafish. Paraffin slides were counterstained with DAPI(blue). Arrow-heads indicate ACTH-producing cells coexpress-ing intranuclear PCNA. PCNA index was calculated in WT andTg:Pomc-Pttg pituitary (g) (mean ± SE, n = 500 cells countedper pituitary, two pituitaries per group; *P = 0.05). AP, anteriorpituitary; IP, pars intermedia. P, pituitary. (Scale bar, 50 μm.)

Liu et al. PNAS | May 17, 2011 | vol. 108 | no. 20 | 8415

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

5, 2

020

We then performed histological sections of zebrafish kidney toidentify zebrafish Gc steroidogenic cells (27). Tg:Pomc-Pttg fishdemonstrated increased intrarenal epithelial cell layers surround-ing the posterior cardinal vein compared with WT, consistent withACTH-stimulated adrenal hyperplasia (Fig. 2A).To determine the metabolic impact of hypercortisolism in Tg:

Pomc-Pttg zebrafish, we subjected adult Tg:Pomc-Pttg andWT fishto 16 h fasting followed by ad libitum feeding of regular diet for 1 h.Tg:Pomc-Pttg zebrafish exhibited consistently higher levels offasting and postprandial blood glucose levels than WT zebrafish(96 ± 9 vs. 65 ± 10 mg/dL, mean ± SD; P < 0.0001; Fig. 2B),demonstrating both attenuated fasting and postprandial glucosetolerance. Because teleost fish are glucose intolerant as a result ofblunted peripheral responses to insulin (28), andGcs induce insulinresistance in mammals, we assessed insulin sensitivity by testingblood glucose responses to intraperitoneally administered insulin.Whereas WT fish exhibited a brisk hypoglycemic response 30 minafter injection of a relatively high insulin dose (0.1 U/100 mg; P <0.01), Tg:Pomc-Pttg fish showed no significant change of bloodglucose levels for up to 90 min after insulin injection (Fig. 2C).Hepatic lipid content as detected by oil red O staining was in-creased in Tg:Pomc-Pttg fish (Fig. 2 D and E), suggesting visceraladiposity resulting from increased insulin resistance. Finally,chronic hypercortisolism exerts specific myocardial effects lead-ing to increased ventricular wall thickness, with subsequent sys-tolic and diastolic dysfunction contributing to high risk of heartfailure in patients with Cushing disease (1, 29). Reflective of thechronic hypercortisolemic status caused by corticotroph PTTGoverexpression, a spectrum of cardiac hypertrophy was observedin late-stage (24 mo) Tg:Pomc-Pttg fish, with increased heart wallthickness involving trabecular and compact zones of the singleventricular chamber (Fig. 2F). Four of 18 Tg:Pomc-Pttg trans-genic fish also exhibited coexisting overt pericardial effusion (Fig.2F, Top). Taken together, corticotroph targeted PTTG over-

expression in Tg:pomc-pttg zebrafish results in ACTH-dependenthypercortisolism and metabolic disruptions mimicking features ofmammalian Cushing disease.

Corticotroph PTTG Overexpression Induces Cyclin E. Our previousstudies indicated that PTTG facilitates G1/S transition by actingcoordinately with Sp1 to up-regulate cyclin D expression in hu-man choriocarcinoma cells (16). To understand the mechanismfor zebrafish corticotroph PTTG overexpression inducing alteredG1/S transition (Fig. 1), we analyzed expression of key G1/S cellcycle regulators by real-time PCR in adult Tg:Pomc-Pttg and WTpituitary glands. Whereas expression of pituitary cyclin D, p21,and p27 were not different betweenWT and Tg:Pomc-Pttg, cyclinE mRNA levels were more than doubled in the Tg:Pomc-Pttgpituitary (Fig. 3A).Cyclin E up-regulation has been associated with poor clinical

outcomes inhumanmalignancies (30). In the adult pituitary, cyclinEis undetectable in normal cells but preferentially up-regulated intumors of corticotroph, but not other, lineage(s) (31). In murinepituitary POMC cells, cyclin E overexpression collaborates withp27kip1-null mutation to increase cell proliferation, centrosome in-stability, and tumor formation (9). Up-regulated cyclin E is also as-sociated with loss of Brg1 observed in approximately one third ofhuman corticotroph adenomas (9). Enhanced pituitary cyclin EmRNA levels observed in Tg:Pomc-Pttg fish may not representprotein expression, but the minute adult zebrafish pituitary size (<1mm) technically hamperedanalysisof proteinexpressionbyWesternblot. We therefore determined whether PTTG regulates cyclin Eexpression in mammalian (murine) AtT20 corticotroph tumor cellsthat express abundant endogenous PTTG and cyclin E proteins.Suppression of endogenous PTTG expression with a PTTG-specificsiRNA resulted in decreased cyclin E expression and enhancedp27kip1 levels (Fig. 3B), whereas p21 expression was not changed(Fig. 3B). These observations suggest that PTTG up-regulation of

Fig. 2. Tg:Pomc-Pttg transgenic zebra-fish develop hypercortisolism, exacer-bated insulin resistance, glucoseintolerance, hepatic steatosis, and car-diomyopathy. (A) Tg:pomc-pttg (Top,Tg) fish showed steroidogenic cell ex-pansion as depicted by H&E stain ofkidney head paraffin slides. Zebrafish Gcsteroidogenic cells are arranged as epi-thelial cell layers (arrowheads) in asso-ciation with the renal head posteriorcardinal vein (27). Tg:pomc-pttg fish alsoshowed blood cell accumulation withinthe posterior cardinal vein, which wasnot observed in WT (Bottom). (B) Glu-cose tolerance tests were performed in72 Tg:pomc-pttg and WT fish. Zebrafishwere given ad libitum feeding of regulardiet for 1 h (gray column) after 16 h offasting. Glucose levels are presentedas mean ± SE (area under the curve;P < 0.0001). (C) After 20 h of fasting,zebrafish were intraperitoneally injec-ted with insulin (0.1 U/100 mg), andblood glucose levels were measured at30 and 60 min after insulin injection(n = 24, mean ± SE; *P < 0.01). (D) Oilred O staining of liver sections revealhepatic lipid accumulation in Tg:Pomc-pttg transgenic fish. (E) Area distribu-tion and intensity of hepatic oil red Ostaining were scored as described in SIMethods (mean ± SE; *P < 0.01). (F) Top:Tg:Pomc-pttg zebrafish (Tg) with grosspericardial fluid accumulation (arrow) compared with WT at 24 mo. Middle: H&E stain of midbody cross section. Bottom: High magnification (20×) showingventricular hypertrophy of the heart. Tg, Tg:Pomc-Pttg; Pc, pericardial space; h, heart. (Scale bar, 50 μM.)

8416 | www.pnas.org/cgi/doi/10.1073/pnas.1018091108 Liu et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

5, 2

020

cyclin E and down-regulation of p27kip1 in pituitary corticotrophtumor cells occurs independently of p21.

In Vivo Testing of CDK/Cyclin Inhibitors in Tg:Pomc-Pttg Zebrafish.Zebrafish pituitary POMC cell differentiation starts at the ante-rior neural ridge by 20 hpf, and is completed within the maturepituitary by 48 hpf (22). Within the first few days of embryonicdevelopment, our transgenic fish shown here recapitulate hall-mark features of Cushing disease, i.e., lineage-specific cortico-troph expansion with partial Gc resistance (Fig. 1). The observedG1/S alteration, cyclin E up-regulation, and neoplastic cortico-

troph changes led us to screen small-molecule CDK inhibitorswith different spectra of inhibitory selectivity, including flavopir-idol (CDK 4/6, 2, 1, 9), R-roscovitine (CDK 2, 1) (32), olomoucine(CDK 2, 1) (32), PD-0332991(CDK 4/6), and CAY10572 (CDK7) (33). Tg:Pomc-Pttg; POMC-eGFP double transgenic embryoswere exposed to each compound added to the embryo culturemedium. Although flavopiridol retarded early embryonic devel-opment before corticotroph ontogeny occurred, in vivo treatmentof zebrafish embryos with R-roscovitine, olomoucine, PD-0332991, and CAY10572 starting at 18 hpf caused no apparentgrowth defect by 40 hpf (Fig. 3C). Strikingly, R-roscovitine-treated embryos exhibited approximately 40% reduction in pitu-itary POMC-eGFP expression compared with controls (1.0 ± 0.08vs. 0.6 ± 0.09, mean ± SE; n= 7 for each group; P < 0.02; Fig. 3 Cand D). A modest, approximately 20%, reduction of POMC-eGFP expression was also observed in the olomoucine-treatedgroup (1.0 ± 0.08 vs. 0.8 ± 0.07, mean ± SE; n= 7 for each group;P = 0.07), whereas PD-0332991 and CAY10572 caused no sig-nificant change in pituitary POMC-eGFP expression comparedwith controls (Fig. 3 C and D).To determine the specificity of R-roscovitine action against

zPttg-overexpressing POMC cells, we generated another doubletransgenic line (Tg:Pomc-Pttg;Prl-RFP) by breeding Tg:Pomc-Pttg fish with a previously generated PRL-RFP transgenic line, inwhich RFP was targeted to pituitary lactotrophs by a zebrafishProlactin promoter (34). In vivo treatment between 18 and 48 hpfof Tg:Pomc-Pttg;Prl-RFP and Tg:Pomc-Pttg;POMC-eGFP em-bryos with R-roscovitine revealed no effect on Prl-RFP expression(1.0 ± 0.08 vs. 1.0 ± 0.09, mean ± SE; n = 9 for each group; P =0.3), but a greater than 50% reduction of POMC-eGFP expres-sion (1.0 ± 0.07 vs. 0.5 ± 0.05, mean ± SE; n= 10 for each group;P < 0.000005) compared with control groups (Fig. 3 E and F).

R-Roscovitine Action in Mouse Corticotroph Tumor Cells. Olomou-cine and roscovitine are structurally related 2,6,9-trisubstitutedpurines, which cause G1/S or G2/M arrest by competing for ATPbinding sites on CDK1 and CDK2. The R-isomer of roscovitine(R-roscovitine, CYC202) is a more potent and selective inhibitorof CDK2/cyclin E, and murine corticotrophs are highly sensitiveto disrupted CDK2/cyclin E-mediated cell cycle pathways (9).Cyclin E up-regulation leads to cell cycle reentry of differenti-ated POMC cells and also inactivates p27kip1, further enhancingcell cycle progression (9). In addition, p27kip1 protects differen-tiated pituitary POMC cells from reentering the cell cycle,whereas p57Kip2 is required for cell cycle exit of pituitary pre-cursor cells (35). Given the in vivo potency of R-roscovitineagainst zebrafish Pttg-overexpressing corticotrophs (Fig. 3), westudied its effect on CDK2/cyclin E-mediated cell cycle pathwaysin mouse ACTH-secreting pituitary tumor cells (Fig. 4).Treatment with R-roscovitine (1–2 × 10−5 M) led to decreased

cell number by 24 h (Fig. 4A). Western blot analysis of proteinextracts derived from R-roscovitine–treated cells revealed evi-dence for cell cycle arrest, including decreased cyclin E, increasedp27Kip1, p57Kip2, and p21Cip1 expression, as well as reducedThr821 phosphorylation of Rb (Fig. 4B). R-roscovitine treatmentalso induced senescent features by 48 h as evidenced by increasedβ-gal expression (Fig. 4C).Consistent with decreased cell viability, we detected decreased

ACTH concentrations in culture medium derived from R-roscovi-tine–treated AtT20 cells (Fig. 4D). Western blot analysis of proteinextracts derived from R-roscovitine–treated AtT20 cells showedsuppressed ACTH expression (Fig. 4E). These results indicate thatR-roscovitine targets cdk2/cyclin E-mediated cell cycle progression,and also inhibits corticotroph ACTH protein expression.

R-Roscovitine Inhibits in Vivo Corticotroph Tumor Growth and ACTHExpression. To further establish R-roscovitine action on cortico-troph tumors in vivo, we injected athymic nude mice (approxi-mately 6–8 wk old) s.c. with AtT20 corticotroph tumor cells(1 × 105 cells). Three days after tumor cell injection, 29 of 30mice had developed small (approximately 2–3 mm3) but visible s.

Fig. 3. In vivo drug testing in Tg:Pomc-Pttg zebrafish. (A) Pttg over-expression directed by zebrafish Pomc promoter induced cyclin E up-regu-lation in Tg:Pomc-Pttg transgenic pituitary at 3 mo. mRNA levels wereassayed by quantitative real-time PCR (mean ± SE of relative expression; n =30 pituitaries for each group). (B) Western blot of mouse corticotroph tumorAtT20 cells transfected with a control or PTTG siRNA. (C) In vivo treatment ofTg:Pomc-Pttg;Pomc-eGFP embryos with small-molecule CDK inhibitors (50μM) or 0.2% DMSO as control from 18 to 40 hpf. One hundred to onehundred fifty embryos were treated with each compound. Representativeimages of live embryos are shown with gross morphology (Right) and pitu-itary Pomc-GFP–positive cells at higher magnification (Left) at 40 hpf. Em-bryos exposed to flavopiridol developed early developmental defect beforepituitary POMC cell ontogeny occurs. (D) Relative expression of pituitaryPomc-eGFP fluorescence analyzed using Volocity 5.2 software (Improvision;mean ± SE of relative expression, n = 7). (E) R-roscovitine specifically sup-presses expansion of pituitary POMC cells overexpressing zPttg from 18 to 48hpf. Double transgenic Tg:Pomc-Pttg;Prl-RFP embryos were generated bybreeding Tg:Pomc-Pttg fish with a previously generated PRL-RFP transgenicline, in which RFP was targeted to pituitary lactotrophs by a zebrafish Pro-lactin promoter (34). Representative fluorescent microscopy of pituitaryPOMC-eGFP (a and b) and PRL-RFP (c and d) expression in live Tg:Pomc-Pttg;Pomc-eGFP and Tg:Pomc-Pttg;Prl-RFP embryos treated with 0.2% DMSO (aand c) or 50 μM R-roscovitine (b and d). (F) Relative expression of pituitaryPOMC-eGFP or PRL-RFP fluorescence were analyzed (mean ± SE of relativeexpression; n = 10). Results represent one of three similar experiments;*P < 0.02 and **P < 0.000005. (Scale bar, 50 μm.)

Liu et al. PNAS | May 17, 2011 | vol. 108 | no. 20 | 8417

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

5, 2

020

c. tumors, and were randomized to receive R-roscovitine (150 mg/kg) or vehicle via oral gavage twice daily for 5 d each week. After 3wk, R-roscovitine caused approximately 50% weight reduction ofdissected tumor xenografts (40.0± 4.7mg vs. 21.0± 2.6mg,mean±SE; n = 13–14 for each group; P < 0.02; Fig. 5A).Consistent with the in vitro observations, Western blot and

immunohistochemistry analysis of tumor specimens showed sup-pressed ACTH and PCNA protein expression by R-roscovitine(Fig. 5 B and C). R-roscovitine–treated mice exhibited more than50% reduction in plasmaACTH levels (1256± 596 pg/mL vs. 596±103 pg/mL, mean ± SE; n= 13–14 for each group; P < 0.01), andapproximately 50% reduction in serum corticosterone levels[1,046± 109 ng/mL vs. 561± 72 ng/mL, mean± SE; n= 13–14 foreach group (P < 0.005); linear regression between ACTH andcorticosterone, r = 0.9425 (P < 0.0001); Fig. 5D]. The highbaseline plasma ACTH levels may represent tumor secretion aswell as stress-induced responses during CO2 euthanasia.

DiscussionTumor-targeted drug development for Cushing disease is a majorchallenge, as the pathogenesis of corticotroph adenomas remainsenigmatic. Recently, protein kinases, i.e., epidermal growth factorreceptor family (e.g., HER) and CDKs, have been suggested astherapeutic targets for pituitary tumors (8).* Although tumorresponses to protein kinase inhibitors is selective, and may bedictated by specific mutations and/or tumor cellular context,preclinical testing is hampered by poor predictabilities with re-spect to molecular pathophysiology of the tumors being assessed.Here, we report generation of germline transgenic zebrafish

overexpressing zPttg targeted to pituitary POMC cells, as a smallvertebrate animal model of Cushing disease. Although the pheno-type of hypercortisolism was observed in adult Tg:Pomc-Pttgzebrafish by 3 mo of age, pituitary corticotroph expansion withpartial resistance to Gc-negative feedback was already detectedwithin the first 2 d of embryonic development of stable transgeniczebrafish. Furthermore, the Tg:Pomc-Pttg pituitary demonstratesa characteristic feature of human corticotroph adenomas, i.e., cyclinE up-regulation andG1/S phase disruption. Themolecular featuresand early pathologies of corticotroph tumors in Tg:Pomc-Pttgtransgenicfishallowedus togain insight intomechanismsunderlyingthe disease pathogenesis, and also to test drug efficacy in vivo.Cyclin E overexpression is associated with disrupted G1/S

transition contributing to development and progression of breastcarcinomas, leukemia, and lymphomas (30). In the pituitary,cyclin E expression is preferentially up-regulated in corticotrophadenomas compared with tumors arising from other lineages,the mechanisms of which remain to be fully defined (31, 36, 37).In a subgroup of corticotroph adenomas, cyclin E up-regulation

was associated with loss of Brg1 expression, suggesting thepresence of additional cyclin E regulators in corticotrophs (26).Our results show that corticotroph zPttg overexpression inducescyclin E, whereas PTTG siRNA suppresses cyclin E expressionin murine corticotroph tumor cells (Fig. 3). PTTG is overex-pressed in more than 90% of pituitary tumors, including corti-cotroph adenomas (12). In addition to inducing aberrant G1/Sand G2/M transition via transcriptional dysregulation of cyclinexpression (12, 14–17), causing chromosomal instability and an-euploidy, pituitary PTTG overexpression activates lineage-spe-cific senescence pathways triggering irreversible cell cycle arrestin GH- and gonadotropin-expressing tumors (12, 19, 20). Corti-cotroph cyclin E up-regulation may represent another pathwayfor PTTG-induced pituitary lineage-specific effects, although it isyet unclear whether PTTG regulates cyclin E expression directlyor indirectly.Corticotroph cyclin E up-regulation contributes to cell cycle

reentry of differentiated corticotrophs and centrosome instability(9). To investigate the clinical significance of cyclin E dysregula-tion in corticotroph adenomas, we performed in vivo drug testingon Tg:Pomc-Pttg embryos using known small molecule com-pounds with different spectra of CDK/cyclin inhibitory selectivity.Our results indicated inhibition of PTTG-overexpressing corti-cotrophs by the 2,6,9-substituted purine analogues, olomoucineand R-roscovitine, with the latter demonstrating a higher efficacyin vivo (Fig. 3). Corticotroph inhibitory effects of R-roscovitinewere further validated inmouse corticotroph tumors (Figs. 4 and 5).R-roscovitine arrests G1/S or G2/M phases via CDK1/2 inhibitionby competing for ATP binding sites (11), inhibition of RNApolymerase II-dependent transcription, and selective actionagainst CDK2/cyclin E (38, 39). The molecule is currently un-dergoing clinical trials for several malignancies, and the oraldosing route and relatively mild side effects of R-roscovitine makedaily long-term treatment of Cushing disease feasible.Our results suggest that R-roscovitine inhibits corticotroph

tumor cell growth via CDK2/cyclin E and Rb-mediated pathways,independent of p53 (Fig. 4). Both in vitro and in vivo results showthat R-roscovitine also suppresses ACTH expression/production(Figs. 4 and 5), suggesting other regulatory mechanisms in addi-tion to CDK2/cyclin E-mediated cell growth. One of the possiblemechanisms may involve inhibition of CRH receptor signalingpathways in corticotroph tumor cells, as other 2,6,9-trisubstitutedpurine analogues have been developed as CRH receptor antag-onists exhibiting potential anxiolytic and antidepressant activity(40). Further in vivo screening of small molecule libraries withTg:Pomc-Pttg transgenic fish may lead to identification of com-pounds with more potent dual effects targeting both corticotrophtumor growth and ACTH production.

Fig. 4. In vitro inhibition of mouse corticotroph tumorcells by R-roscovitine. (A) Treatment of ACTH-secretingAtT20 cells with R-roscovitine (1–2 × 10−5 M) led to de-creased number of viable cells at 24 and 48 h, as depictedby Wst-1 proliferation assay (mean ± SE; **P < 0.01). (B)Western blot of protein extracts derived from AtT20 cellstreated with vehicle or R-roscovitine. (C) R-roscovitinetreatment (10 μM) for 48 h induced senescence as in-dicated by increased β-gal expression. (D) ACTH concen-tration by radioimmunoassays of culture medium fromAtT20 cells treated with vehicle or R-roscovitine (mean ±SE; **P < 0.01 and ***P < 0.001). (E) Western blot ofprotein extracts derived from AtT20 cells treated withR-roscovitine. Vehicle is 0.2% DMSO.

*Fukuoka H, et al. 92nd Annual Meeting of the Endocrine Society, June 19–22, 2010, SanDiego, CA.

8418 | www.pnas.org/cgi/doi/10.1073/pnas.1018091108 Liu et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

5, 2

020

MethodsWeusedatol2 transposoncassette togenerate theTg:pomc-pttg transgene,andtransgenic founder fish were generated as described previously (22). Furtherdetails on experimental procedures can be found in SI Methods.

ACKNOWLEDGMENTS. We thank Yunguang Tong, Song-Guang Ren, LihuaXia, Cuiqi Zhou, Meina Ren, and Svetlana Zonis for technical assistance, andVera Chesnokova for helpful discussions. Thisworkwas supported byNationalInstitutes of Health Grants KO8 DK 064806 (to N.L.), CA75979 (to S.M.), andRR13227 (to S.L.) and the Doris Factor Molecular Endocrinology Laboratory.

1. Nieman LK, et al. (2008) The diagnosis of Cushing’s syndrome: An Endocrine SocietyClinical Practice Guideline. J Clin Endocrinol Metab 93:1526–1540.

2. Biller BM, et al. (2008) Treatment of adrenocorticotropin-dependent Cushing’ssyndrome: A consensus statement. J Clin Endocrinol Metab 93:2454–2462.

3. Assié G, et al. (2007) Corticotroph tumor progression after adrenalectomy in Cushing’sdisease: A reappraisal of Nelson’s syndrome. J Clin Endocrinol Metab 92:172–179.

4. Boscaro M, et al. (2009) Treatment of pituitary-dependent Cushing’s disease with themultireceptor ligand somatostatin analog pasireotide (SOM230): A multicenter, phaseII trial. J Clin Endocrinol Metab 94:115–122.

5. Melmed S (2003) Mechanisms for pituitary tumorigenesis: The plastic pituitary. J ClinInvest 112:1603–1618.

6. Moons DS, et al. (2002) Pituitary hypoplasia and lactotroph dysfunction in micedeficient for cyclin-dependent kinase-4. Endocrinology 143:3001–3008.

7. Melmed S (2009) Acromegaly pathogenesis and treatment. J Clin Invest 119:3189–3202.

8. Sotillo R, et al. (2005) Cooperation between Cdk4 and p27kip1 in tumor development:A preclinical model to evaluate cell cycle inhibitors with therapeutic activity. CancerRes 65:3846–3852.

9. Roussel-Gervais A, et al. (2010) Cooperation between cyclin E and p27Kip1 in pituitarytumorigenesis. Mol Endocrinol 24:1835–1845.

10. Malumbres M, Barbacid M (2009) Cell cycle, CDKs and cancer: A changing paradigm.Nat Rev Cancer 9:153–166.

11. Fabian MA, et al. (2005) A small molecule-kinase interaction map for clinical kinaseinhibitors. Nat Biotechnol 23:329–336.

12. Vlotides G, Eigler T, Melmed S (2007) Pituitary tumor-transforming gene: Physiologyand implications for tumorigenesis. Endocr Rev 28:165–186.

13. Pei L, Melmed S (1997) Isolation and characterization of a pituitary tumor-transforming gene (PTTG). Mol Endocrinol 11:433–441.

14. Marangos P, Carroll J (2008) Securin regulates entry into M-phase by modulating thestability of cyclin B. Nat Cell Biol 10:445–451.

15. Romero F, et al. (2001) Human securin, hPTTG, is associated with Ku heterodimer, theregulatory subunit of the DNA-dependent protein kinase. Nucleic Acids Res 29:1300–1307.

16. Tong Y, Tan Y, Zhou C, Melmed S (2007) Pituitary tumor transforming gene interactswith Sp1 to modulate G1/S cell phase transition. Oncogene 26:5596–5605.

17. Zou H, McGarry TJ, Bernal T, Kirschner MW (1999) Identification of a vertebrate sister-chromatid separation inhibitor involved in transformation and tumorigenesis. Science285:418–422.

18. Bernal JA, et al. (2008) Proliferative potential after DNA damage and non-homologous end joining are affected by loss of securin. Cell Death Differ 15:202–212.

19. Kim D, et al. (2005) Pituitary tumour transforming gene (PTTG) induces geneticinstability in thyroid cells. Oncogene 24:4861–4866.

20. Chesnokova V, et al. (2008) p21(Cip1) restrains pituitary tumor growth. Proc Natl AcadSci USA 105:17498–17503.

21. Donangelo I, et al. (2006) Pituitary tumor transforming gene overexpressionfacilitates pituitary tumor development. Endocrinology 147:4781–4791.

22. Liu NA, et al. (2003) Pituitary corticotroph ontogeny and regulation in transgeniczebrafish. Mol Endocrinol 17:959–966.

23. Camper SA, Saunders TL, Katz RW, Reeves RH (1990) The Pit-1 transcription factorgene is a candidate for the murine Snell dwarf mutation. Genomics 8:586–590.

24. Lamolet B, et al. (2001) A pituitary cell-restricted T box factor, Tpit, activates POMCtranscription in cooperation with Pitx homeoproteins. Cell 104:849–859.

25. Liu J, et al. (2001) Tbx19, a tissue-selective regulator of POMC gene expression. ProcNatl Acad Sci USA 98:8674–8679.

26. Bilodeau S, et al. (2006) Role of Brg1 and HDAC2 in GR trans-repression of thepituitary POMC gene and misexpression in Cushing disease. Genes Dev 20:2871–2886.

27. Hsu HJ, Lin G, Chung BC (2003) Parallel early development of zebrafish interrenal glandsand pronephros: Differential control by wt1 and ff1b. Development 130:2107–2116.

28. Alexander EL, Dooley KC, Pohajdak B, Xu BY, Wright JR, Jr. (2006) Things we havelearned from tilapia islet xenotransplantation. Gen Comp Endocrinol 148:125–131.

29. Muiesan ML, et al. (2003) Left ventricular structural and functional characteristics inCushing’s syndrome. J Am Coll Cardiol 41:2275–2279.

30. Iida H, et al. (1997) Overexpression of cyclin E in acute myelogenous leukemia. Blood90:3707–3713.

31. Jordan S, Lidhar K, Korbonits M, Lowe DG, Grossman AB (2000) Cyclin D and cyclin Eexpression in normal and adenomatous pituitary. Eur J Endocrinol 143:R1–R6.

32. De Azevedo WF, et al. (1997) Inhibition of cyclin-dependent kinases by purineanalogues: Crystal structure of human cdk2 complexed with roscovitine. Eur JBiochem 243:518–526.

33. Montagnoli A, et al. (2008) A Cdc7 kinase inhibitor restricts initiation of DNAreplication and has antitumor activity. Nat Chem Biol 4:357–365.

34. Liu NA, et al. (2006) Prolactin receptor signaling mediates the osmotic response ofembryonic zebrafish lactotrophs. Mol Endocrinol 20:871–880.

35. Bilodeau S, Roussel-Gervais A, Drouin J (2009) Distinct developmental roles of cellcycle inhibitors p57Kip2 and p27Kip1 distinguish pituitary progenitor cell cycle exitfrom cell cycle reentry of differentiated cells. Mol Cell Biol 29:1895–1908.

36. De Martino I, et al. (2009) HMGA proteins up-regulate CCNB2 gene in mouse andhuman pituitary adenomas. Cancer Res 69:1844–1850.

37. Quereda V, Malumbres M (2009) Cell cycle control of pituitary development anddisease. J Mol Endocrinol 42:75–86.

38. Whittaker SR, Walton MI, Garrett MD, Workman P (2004) The cyclin-dependent kinaseinhibitor CYC202 (R-roscovitine) inhibits retinoblastoma protein phosphorylation, causesloss of Cyclin D1, and activates themitogen-activated protein kinase pathway. Cancer Res64:262–272.

39. McClue SJ, et al. (2002) In vitro and in vivo antitumor properties of the cyclindependent kinase inhibitor CYC202 (R-roscovitine). Int J Cancer 102:463–468.

40. Legraverend M, Grierson DS (2006) The purines: Potent and versatile small moleculeinhibitors and modulators of key biological targets. Bioorg Med Chem 14:3987–4006.

Fig. 5. In vivoactionofR-roscovitine inmousecorticotroph adenomas. Athymic nude micewere s.c. inoculated with corticotroph tumorAtT20 cells (1 × 105 cells). Three days after in-jection, mice were randomized to receive R-roscovitine (150 mg/kg) or vehicle by oral ga-vage twice daily, 5 d/wk. After 3 wk, tumorxenografts were dissected and (A) tumor vol-umes were decreased in R-roscovitine–treatedanimals. (B) Western blot of representativetumor specimens showed decreased ACTHand PCNA expression in R-roscovitine–treatedtumors. (C) R-roscovitine–treated corticotrophtumors exhibited decreased PCNA and ACTHcoexpressing cells. Fluorescence microscopyimage of immunohistochemistry detectingPCNA (red) and ACTH (green) expression incontrol (a–c) and R-roscovitine–treated tumors(d–f). Cryosection slides were counterstainedwith DAPI (blue). (D) Blood was collected fromeach animal for measurement of plasma ACTHand serum corticosterone levels (mean ± SE;n = 13–14 mice for each group; **P < 0.01).

Liu et al. PNAS | May 17, 2011 | vol. 108 | no. 20 | 8419

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

5, 2

020