Genomic and functional characterizations ofphosphodiesterase subtype 4D in human cancersDe-Chen Lina,b,1,2, Liang Xub,1, Ling-Wen Dingb, Arjun Sharmab, Li-Zhen Liub, Henry Yangb, Patrick Tanb, Jay Vadgamac,Beth Y. Karland, Jenny Lesterd, Nicole Urbane, Michèl Schummere, Ngan Doanf, Jonathan W. Saidf, Hongmao Sung,Martin Walshg, Craig J. Thomasg, Paresma Patelh, Dong Yini, Daniel Chanj, and H. Phillip Koefflera,b,j

aDivision of Hematology/Oncology, Cedars-Sinai Medical Center, University of California School of Medicine, Los Angeles, CA 90048; bCancer Science Instituteof Singapore, National University of Singapore, Singapore 117599; cCharles R. Drew University of Medicine and Science, Los Angeles, CA 90059; dWomen’sCancer Program at the Samuel Oschin Comprehensive Cancer Institute, Cedars-Sinai Medical Center, Los Angeles, CA 90048; eFred Hutchinson Cancer ResearchCenter, Seattle, WA 98109; fSanta Monica University of California–Los Angeles Medical Center, Los Angeles, CA 90404; gNational Institutes of Health (NIH)Chemical Genomics Center, National Center for Advancing Translational Sciences, NIH, Bethesda, MD 20892; hBasic Science Program, SAIC-Frederick, Inc.,Chemical Biology Laboratory, Frederick National Laboratory for Cancer Research, Frederick, MD 21702; iKey Laboratory of Malignant Tumor Gene Regulationand Target Therapy of Guangdong Higher Education Institutes, Research Center of Medicine, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University,Guangzhou 510120, China; and jNational University Cancer Institute, National University Hospital, Singapore 117599

Edited by Neal G. Copeland, Methodist Hospital Research Institute, Houston, TX, and approved February 25, 2013 (received for review October 18, 2012)

Discovery of cancer genes through interrogation of genomic dosageis one of the major approaches in cancer research. In this study, wereport that phosphodiesterase subtype 4D (PDE4D) genewas homo-zygously deleted in 198 cases of 5,569 primary solid tumors (3.56%),with most being internal microdeletions. Unexpectedly, the micro-deletions did not result in loss of their gene products. ScreeningPDE4D expression in 11 different types of primary tumor samples(n = 165) with immunohistochemistry staining revealed that its pro-tein levels were up-regulated compared with corresponding non-transformed tissues. Importantly, depletion of endogenous PDE4Dwith three independent shRNAs caused apoptosis and growth inhi-bition in multiple types of cancer cells, including breast, lung, ovary,endometrium, gastric, and melanoma, which could be rescued byreexpression of PDE4D. We further showed that antitumor eventstriggered by PDE4D suppression were lineage-dependently associ-ated with Bcl-2 interacting mediator of cell death (BIM) inductionand microphthalmia-associated transcription factor (MITF) down-regulation. Furthermore, ectopic expression of the PDE4D short iso-form, PDE4D2, enhanced the proliferation of cancer cells both in vitroand in vivo.Moreover, treatmentof cancer cellswith a unique specificPDE4D inhibitor, 26B, triggered massive cell death and growth retar-dation. Notably, these antineoplastic effects induced by eithershRNAs or small molecule occurred preferentially in cancer cells butnot in nonmalignant epithelial cells. These results suggest that al-though targeted by genomic homozygous microdeletions, PDE4Dfunctions as a tumor-promoting factor and represents a unique tar-getable enzyme of cancer cells.

The phosphodiesterases (PDEs) are metallohydrolases, whichdegrade the secondary messengers, adenosine and guanosine

3′,5′-cyclic monophosphates (cAMP and cGMP, respectively) (1,2). The PDEs are grouped into 11 families (PDE1–11) accordingto their substrate specificities, subcellular distributions, and aminoacid sequences (3, 4). Multiple types of human tumor cells pre-dominantly express PDE4 as major regulators of cAMP-hydro-lyzing activity (5). Of note, PDE4 inhibitors (PDE4i) werereported to have antineoplastic functions in several human ma-lignancies including leukemia, colon cancer, and glioma (6–10).Phosphodiesterase subtype 4D (PDE4D) belongs to PDE4

subfamilies (A, B, C, and D). PDE4D gene (human locus 5p12.1)encodes a complex of transcript variants generated by both mul-tiple promoter use and alternative mRNA splicing. PDE4D con-sists of at least nine protein isoforms that all possess thephosphodiesterase catalytic domain in the carboxyl terminus andare categorized into one of the so-called “long,” “short,” and“supershort” forms (reviewed in ref. 11). Interestingly, a previousstudy using a sleeping beauty transposon-based screen suggestedthat PDE4D might function as a proliferation promoting factor inprostate cancers (12). Very recently, Pullamsetti et al. found thatcross-talk that occurred between PDE4 and hypoxia inducible-

factor signaling might promote angiogenesis in lung cancer (13).However, the tumor-specific alterations in PDE4D genome andfunction are not well elucidated.Homozygous deletions in cancer genomes occur at recessive

cancer genes, conferring clonal growth advantage. This type ofdeletions is also found within fragile sites and reflects an in-creased local rate of DNA breakage with no cancer-associatedbiological consequences. To date, the implications of most ho-mozygous deletions in cancer genomes remain elusive. In thepresent study, we identified homozygous microdeletions in thehuman PDE4D gene locus in primary tumors, as well as inestablished cancer cells of diverse origins. Unexpectedly, PDE4DmRNA and proteins were not abrogated, i.e., deletions did notresult in inactivation of PDE4D. Moreover, we found thatPDE4D protein expression levels were elevated in multiple typesof cancers. Further investigations revealed that PDE4D con-tributed to the proliferation and prosurvival of tumor cells.

ResultsIdentification of Genomic and Expression Abnormalities in HumanPDE4D Gene. DNA from 55 triple negative breast cancer (TNBC)individuals were examined by SNP-Chip assay for genomic ab-normalities (SI Appendix, Fig. S1). Homozygous microdeletions inPDE4D (Chr5: q11.2–q12.1) were observed in one patient [triplenegative breast tumor (TNT)-18, validation shown in SI Appendix,Fig. S3A]. This prompted us to analyze somatic copy numberalterations of 5,569 solid tumor primary specimens from TheCancer Genome Atlas (TCGA) projects as well as TumorScape(14). We identified homozygous deletions of PDE4D in 198 sam-ples (3.56%). Notably, 81.3% (161/198) of the deletions affectedonly the intragenic regions of PDE4D, without removing the wholegene (Fig. 1). We further interrogated PDE4D homozygousdeletions in established cancer cell lines with in-house SNP-Chipdata (SI Appendix, Fig. S2), as well as those from the WellcomeTrust Sanger Institute Cancer Genome Project. Internal homo-zygous microdeletions of PDE4D were evident in an additionalnine different tumor cell lines, including breast, lung, and gastriccancers (SI Appendix, Fig. S3 A and B).

Author contributions: D-C.L. and H.P.K. designed research; D-C.L. and L.X. performed re-search; L-W.D., A.S., L-Z.L., H.Y., P.T., J.V., B.Y.K., J.L., N.U., M.S., N.D., J.W.S., H.S., M.W., C.J.T.,P.P., D.Y. and D.C. contributed new reagents/analytic tools; D-C.L, and H.P.K. analyzeddata; D-C.L. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1D-C.L. and L.X. contributed equally to this work.2To whom correspondence should be addressed. E-mail: dchlin11@gmail.com.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1218206110/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1218206110 PNAS | April 9, 2013 | vol. 110 | no. 15 | 6109–6114

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

2, 2

020

We then examined the expression of PDE4D, first at thetranscriptional level and found that most of the PDE4D-micro-deleted samples still expressed PDE4D messenger RNAs (Fig.2A). Different use of PDE4D isoforms was noted in different celllines (Fig. 2A). Importantly, analysis of a large number of casesfrom four different cancer datasets revealed that tumors withPDE4D deletions have comparable abundance of PDE4DmRNA expression (Fig. 2C). At the translational level, all of thecell lines harboring PDE4D microdeletions expressed apprecia-ble levels of PDE4D proteins compared with PDE4D-unaffectedcells, albeit in different isoforms as evidenced by reduced sizeson SDS/PAGE (Fig. 2B and SI Appendix, Fig. S3C).We next examined the expression of PDE4D proteins in a wide

spectrum of primary tumor specimens using immunohistochem-istry. Again surprisingly, PDE4D proteins were detectable in mostof the cancer types tested, with elevated expression in tumorscompared with adjacent normal tissues including melanoma,ovarian, endometrial, and gastric cancers (Fig. 3A and SIAppendix,Table S5). Analysis of survival data from TCGA projects showedthat high PDE4D mRNA expression was associated with worseprognosis in both an endometrium cancer cohort and a head andneck squamous cell carcinoma cohort of patients but not in thosewith either ovarian cancers or breast cancers (Fig. 3B and SIAppendix, Fig. S4).

Depletion of PDE4D Triggered Apoptosis and Suppressed Cell Growthin a Panel of Human Cancer Cells. Based on the observations that:(i) the homozygous internal deletions in PDE4D gene did notresult in loss of either mRNA or protein expression levels and(ii) PDE4D proteins levels were up-regulated in several differenttypes of tumors, we hypothesized that PDE4D gene products arenecessary for promoting and/or maintaining the transformedphenotype of the cancer cells. To study the cancer-specific functionof PDE4D, cell line models representing tumors showing elevatedexpression level of PDE4D proteins from our immunohisto-chemistry data were selected, namely melanoma, ovarian, endo-metrial, and gastric cancers. We also included breast cell line(MB-231) because they are triple negative breast cancer cells andour initial observation was derived from this type of cancer. Lungcancer cells (A549) were further investigated because PDE4A andPDE4D have been associated with lung cancer (13). We first de-pleted endogenous PDE4D mRNA and protein expression by

shRNA-mediated knockdown. All three shRNAs targeted thecarboxy terminus of PDE4D, leading to the inhibition of all knownhuman PDE4D gene transcripts. Notably, depletion of PDE4Dinduced apoptosis within 72–96 h after infection of lentiviralparticles as measured by both annexin-V exposure and cleavage ofPoly ADP-ribose polymerase 1 (PARP), regardless of the status ofPDE4D genomic microdeletions (Fig. 4 A and B and SI Appendix,Fig. S5A). Cells became round, refractile, and misformed uponPDE4D knockdown (SI Appendix, Fig. S5B). In addition, shRNAscaused growth retardation of all cell lines tested (Fig. 4C and SIAppendix, Fig. S5C). These results strongly suggest that PDE4D isindispensable for maintaining cancer cell viability.To confirm that the apoptosis triggered by shPDE4Ds was spe-

cifically due to the depletion of PDE4D rather than off-targeteffects, restoration of PDE4D expression experiments were per-formed. We took advantage of the fact that one of the threeshPDE4Ds (shPDE4D-a) was designed to target the 3′ un-translated region (3′-UTR) of the PDE4D transcripts, and ourectopic expression vectors did not contain this region. We firststably expressed ectopic PDE4D2 (supershort form of humanPDE4D transcripts) intoMB-231 cells and then infected these cellswith shPDE4D-a lentiviral particles. The shPDE4D-a powerfullysuppressed the expression of endogenous but not exogenousPDE4D (Fig. 4D). Importantly, flow cytometry (FCM) and cellmorphology analysis revealed that PDE4D2 restoration efficientlyrescued MB-231 from apoptosis induced by depletion of PDE4D(Fig. 4E and SI Appendix, Fig. S5F). Similar results were obtainedwith melanoma M368 cells (SI Appendix, Fig. S6). Furthermore, inA2008 ovarian cancer cells, which expressed undetectable endog-enous PDE4D proteins, shPDE4D-a neither caused apoptosis norinhibited cell proliferation (SI Appendix, Fig. S5D).The findings that PDE4D proteins were elevated in several

types of cancers prompted us to ask whether the antiproliferativeeffects induced by PDE4D depletion were specific to cancercells. To this end, nonmalignant mammary epithelial cells (MCF-10A) were transduced with shPDE4D-a. Importantly, no de-tectable effect on either apoptosis or growth rate was caused bysuppression of PDE4D expression (SI Appendix, Fig. S5E).

Bcl-2 interacting mediator of cell death (BIM) Contributes to shPDE4D-Induced Apoptosis.Wenext investigated themolecular mechanismsgoverning shPDE4D-promoted antitumor effects. As expected,

Fig. 1. Intragenic PDE4D microdeletions in hu-man cancers. Integrative Genomics Viewer (IGV)heatmap of PDE4D copy number data for variouscancer samples from TCGA as well as Tumor-Scape. Copy number value for homozygousdeletions (< −0.79) is calculated based on thepresumptions that: (i) loss of both alleles waspresent in >70% cancer cell populations, and (ii)normal cell contamination was <25%. Lung SCC,lung squamous cell carcinoma; HNSCC, head andneck squamous cell carcinoma; ESCC, esophagealsquamous cell carcinoma; HCC, hepatocellularcarcinoma; RCC, renal cell carcinoma; NSCLC,nonsmall cell lung cancer. M, microdeletions; W,deletions affected whole PDE4D.

6110 | www.pnas.org/cgi/doi/10.1073/pnas.1218206110 Lin et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

2, 2

020

phosphorylation of cAMP responsive element binding protein(phospho-CREB) was enhanced in PDE4D-depleted cells, indi-cating induction of cAMP-mediated signaling (Fig. 5A). Westernblotting analysis suggested that the mitochondrial-associatedapoptosis pathway was activated in shPDE4D-infected cells.Examination of Bcl-2 Homology 3 (BH3)-containing familymembers revealed that the proapoptotic protein BIM wasmarkedly elevated upon PDE4D depletion in both A549 andMB-231 cells (Fig. 5A). Increased expression of BIM was pre-viously reported as one of the mechanisms for cAMP-inducedapoptosis of immature T cells (15). To test whether BIM is acritical determinant of shPDE4D-promoted apoptosis, we sta-bly silenced BIM (shBIM) and assessed the impact of BIM-depletion on shPDE4D-induced cell apoptosis. As shown in Fig.5B, stable silencing of BIM inhibited the up-regulation of BIMafter depletion of PDE4D. Critically, compared with nontargeting

control cells, shBIM-expressing cells were significantly (P = 0.029)resistant to shPDE4D-induced apoptosis.

PDE4D-Depletion Inhibits MITF Expression in Melanoma. The mech-anisms shown above, however, were not responsible for apoptosisin PDE4D-depleted melanoma cells. These PDE4D silenced cellsdid not have induction of BIM, and silencing BIM could notrescue melanoma cells from apoptosis. Notably, silencing PDE4Dinmelanoma cells (A375 andM368) was associated with amarkeddecrease expression of MITF, a well-characterized lineage-specificoncogene in melanoma (Fig. 5C). MITF is crucial for the survivalof melanoma cells, as evidenced by the findings that MITF de-pletion leads to a massive cell death (16, 17). We noticed thatMITF-decreased cells also had increased cleaved PARP andcaspase-9 and decreased level of B-cell lymphoma-extra large(BCL-XL) (Fig. 5C). In agreement with the reports that MITFactivated cyclin-dependent kinase inhibitor 1 (p21Cip1) expression(18), we found that p21Cip1 protein was markedly decreased uponPDE4D knockdown (Fig. 5C). We further found that MITFmRNA level was down-regulated in PDE4D-depleted cells (Fig.5D). Notably, in silico analysis revealed that in two melanomagene expression databases (19, 20), mRNA expression of PDE4Dand MITF was significantly correlated with each other (Fig. 5 Eand F). Furthermore, Fig. 5E also showed that the expression ofboth PDE4D and MITF increased when comparing level in nor-mal skin, benign melanocytes nevus, and cutaneous melanomas.These results indicate that in melanoma, PDE4Dmight positively

Fig. 2. Examination of the expression of PDE4DmRNA and proteins. (A) PCRanalysis of the cDNA expression of each isoform of PDE4D (1–9) from TNT-18,gastric cancer cells (HGC-27 and SCH), H460 (PDE4D homozygous deleted,Hom Del), and A375 cells (PDE4D diploid, 2N). The specificity of the RT-PCRwas confirmed by shRNA-mediated depletion as well as direct sequencing (SIAppendix, Fig. S3D). Arrows indicate where two gels were spliced. (B) Variouscell lines either with (homozygous deleted, HD) or without (2N) PDE4D in-ternal microdeletions were subjected to Western blot (WB) analysis withantibodies against the carboxy terminus of PDE4D, which detect all knownhuman PDE4D isoforms. GAPDH was used as a loading control. Arrows labelthe bands, which disappeared upon shRNA knockdown of PDE4D, indicatingthey were protein products of PDE4D transcripts. The specificity of WB wasconfirmed by shRNA-mediated knockdown experiments (Fig. 4A). (C) Plotsshowing the correlation between PDE4DmRNA level and its DNA copy number.Matched mRNA and DNA copy number data were from TCGA projects (30):breast carcinoma (451 cases), head and neck squamous cell carcinoma (288cases), lung squamous cell carcinoma (121 cases), and lung adenocarcinoma(112 cases).

Fig. 3. Elevated expression level of PDE4D protein in primary human can-cers. (A) Immunohistochemical staining photographs of representative casesfrom ovary, endometrium tumors, and melanoma showing that PDE4D pro-teins were overexpressed in tumor cells compared with corresponding non-transformed tissues (Right column). (B) Kaplan–Meier plots and P valuesbased on a log-rank test are indicated for the comparison of cases with highlevels (defined as: > mean in endometrium; > mean + 2 SD in HNSCC) versuslow levels (defined as: < mean in endometrium; < mean + 2 SD in HNSCC) ofPDE4D mRNA. Survival data were generated from TCGA endometrium car-cinoma and head and neck squamous cell carcinoma (HNSCC) projects (30).

Lin et al. PNAS | April 9, 2013 | vol. 110 | no. 15 | 6111

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

2, 2

020

regulate MITF expression and thus play a vital role in cell survivaland viability.

PDE4D2 Promoted Melanoma Proliferation Both in Vitro and in Vivo.To address further the PDE4D contribution to the neoplasticphenotype, we ectopically expressed and studied two PDE4Dtranscript isoforms, namely PDE4D2 and PDE4D3. The expres-sion of PDE4D3 did not result in appreciable change in either cellproliferation or resistance to apoptosis (SI Appendix, Fig. S7 Aand B). In contrast, ectopic expression of PDE4D2 significantlyenhanced the proliferation of A375 and gastric cancer cells(HGC-27) (Fig. 6 A and B) but not ovarian cancer cell line(Caov2) and Ishikawa cells in vitro (SI Appendix, Fig. S7C andD).More importantly, A375 cells with forced expression of PDE4D2resulted in the development of significantly (P < 0.002) largertumors compared with the GFP-control cells (Fig. 6 C and D).

Specific PDE4D Inhibitor Treatment Preferentially Suppressed theProliferation of Malignant Cells. The above observations promp-ted us to test the strategy of PDE4D suppression as a therapeuticapproach in cancer treatment. We examined the antineoplasticactivity of a unique specific PDE4D inhibitor (26B) (Fig. 6E). Theinhibitor 26B is a highly selective PDE4D inhibitor with a 75-foldlower activity against the PDE4A, B, and C subtypes. It alsoexhibits 18,000-fold decreased potency against other PDE familymembers (21). We first examined and determined that 26B

significantly suppressed PDE4D activity at around 1 μM (SI Ap-pendix, Fig. S8D). We further found that exposure of a panel ofhuman tumor cells of diverse origins to compound 26B causeddramatic apoptosis and growth inhibition (Fig. 6F). To rule outoff-targets effects, we codepleted other PDE4 family membersthrough shRNA-mediated knockdown. As shown in SI Appendix,Fig. S8 B–G, 26B suppressed PDE4D-dependent malignant cellgrowth through inhibiting PDE4D; on the other hand, in PDE4D-independent cancer cells (e.g., A2008), 26B caused growth re-tardation by targeting other PDE4 family members. Treatmentwith the compound also elevated BIM expression and down-regulated MITF protein levels, which paralleled the shPDE4D-induced molecular changes (Fig. 6G). Importantly, treatmentwith 26B did not result in cell death or growth retardation inMCF-10A nontransformed breast cells (Fig. 6H). These results

Fig. 4. PDE4D knockdown causedapoptosis and growth retardation in varioustypes of cancer cells. Cancer cells were infectedwith lentiviral particles encodingeither scramble shRNA (nontargeting) or three different shRNAs targetinghuman PDE4D (shPDE4D) for 72 to 96 hours. Subsequently, cells were subjectedto (A) WB analysis with indicated antibodies, or (B) flow cytometry (FCM)analysis after staining with Annexin V-FITC/propidium iodide, or (C) cell pro-liferation assays. Beta-Actin was used as a loading control. (D) MB-231 cellswhich stably expressed exogenous PDE4D2 were established and then trans-duced with shRNA encoding either nontargeting or shPDE4D-a for 72 hours.Subsequently, cells were subjected to WB analysis with indicated antibodies orFCM analysis by Annexin V–FITC/propidium iodide staining (E). ExogenousPDE4D2 (Ex) migrates to approximately 69KD and the major endogenousPDE4D (En) inMB-231migrates to about 95KD on SDS-PAGE. **, P < 0.01; *, P <0.05; N.S., not significant.Apoptosis andproliferationdata representmean± SDof three independent experiments. 2N, PDE4D diploid; HD, PDE4D homozygousdeleted.

Fig. 5. BIM andMITF play roles in apoptosis triggered by PDE4D silencing. (A)A549 andMB-231were transducedwith shRNAs encoding either nontargetingor shPDE4D for 72 h. Subsequently, cells were subjected to WB analysis withindicated antibodies. (B) BIM-depleted MB-231 stable cells were establishedand then transduced with shRNAs encoding either nontargeting or shPDE4Dfor 72 h. Subsequently, cells were subjected to WB analysis with indicatedantibodies (Upper) or FCM analysis by staining with annexin V-FITC/propidiumiodide (Lower). (C) Melanoma cells (A375 and M368) were transduced withshRNAs encoding either nontargeting or shPDE4D for 72 h. Subsequently, cellswere subjected to WB analysis with indicated antibodies or (D) q-PCR analysiswith indicated primers. (E and F) Plots of correlation of PDE4D andMITFmRNAexpression in two dataset from Talantov et al. (19) (E, consist of normal skin,benign melanocytes nevus, and cutaneous melanomas) and Eskandarpouret al. (20) (F, consist of melanoma cancer cell lines). Each dot represents theexpression level of PDE4D (x axis) andMITF (y axis) of a given sample. Pearson Pand R values were calculated with r script by analyzing of the raw microarraydata. Data representmean± SDof three independent experiments. **P< 0.01;*P < 0.05. UD, undetectable.

6112 | www.pnas.org/cgi/doi/10.1073/pnas.1218206110 Lin et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

2, 2

020

suggest that compound 26B has potential therapeutic use inneoplastic diseases.

DiscussionCancer cells typically exhibit loss of genomic materials contain-ing recessive cancer genes. On the other hand, due to the in-stability of the cancer genome, genes not implicated in cancerdevelopment (passengers) also are deleted. In the present study,we report a unique finding that homozygous deletions also occurin a locus involving a tumor-promoting gene, PDE4D. One of themost important data supporting the conclusion that PDE4Dfunctions similar to an oncogene rather than a recessive cancergene were derived from our demonstrations that PDE4D sup-pression triggered apoptotic cell death and growth retardation incancer cells, but not in nonmalignant cells. The prosurvival/growth roles of PDE4D were observed in a panel of cancer cellsfrom different tissue types, indicating its broad tumor-promotingactivity.We deep sequenced all of the PDE4D coding regions from the

PDE4D-microdeleted (including TNT-18) as well as a number of

PDE4D-unaffacted patient samples of diverse origins and couldfind no nonsynonymous somatic mutations. We also resequenced40 different gastric cancer cell lines, consisting of both PDE4D-microdeleted and PDE4D-unaffected ones, and found twosynonymous single nucleotide variations (SNVs) and one non-synonymous SNV (SI Appendix, Fig. S9A). Polyphen (http://genetics.bwh.harvard.edu/pph2/) predicted that Phe to Tyr atninth amino acid heterozygous SNV is 100% “healthy” for thePDE4D protein, indicating that it was unlikely to result in in-activation or loss of function of PDE4D.Another observation suggesting that PDE4D might be a tu-

mor-promoting factor was the up-regulation of its protein intumor cells compared with adjacent normal tissues. We reasonedthat PDE4D protein elevation in cancers was not caused by ge-nome microdeletions but due to other mechanisms based on thefollowing observations: (i) The deletion did not lead to eitherenhanced or reduced mRNA expression (Fig. 2A). In addition,the PDE4D protein levels were comparable between cells ei-ther with or without PDE4D internal deletions (Fig. 2B and SIAppendix, Fig. S3C); (ii) Compared with the frequency of homo-zygous deletions (∼4%), the overexpression of PDE4D protein ismuch more common. For instance, 6 of 10 gastric tumors (60%)expressed an elevated level of PDE4D, whereas the normal tis-sues did not express detectable PDE4D (SI Appendix, Table S5).In-depth analysis showed that internal microdeletions of

PDE4D affected regions in the first eight exons (SI Appendix, Fig.S3A). These exons encode upstream conserved region 1 (UCR1)and UCR2, which are critical for the control of PDE4D cata-lytic activity. Specifically, UCR1 module determines the func-tional outcome of direct ERK phosphorylation. In the presenceof UCR1, ERK phosphorylation leads to suppression of PDE4Dcatalytic activity; whereas in the absence of UCR1, the functionalconsequence of this modification is activation (22–24). On theother hand, UCR2 bears an autoinhibitory property againstPDE4D catalytic activity (25–27). Given that ERK hyperacti-vation is common across human cancers, loss of UCR1/UCR2caused by the genomic microdeletions will conceivably leadto PDE4D activation. Importantly, we constructed crystal 3Dstructural models and showed that UCR1 (fully) and UCR2(partially) cover the ligand binding site of the PDE4D catalyticdomain (SI Appendix, Fig. S9B). By doing so, both UCR1 andUCR2 are likely to form complexes with the PDE4D catalyticdomain before the substrate cAMP enters the active site, thuspreventing the catalyzing process. Furthermore, we ectopicallyexpressed both PDE4D long form (PDE4D3, which containsboth UCR1 and UCR2) and the super short form (PDE4D2,which does not have either a functional UCR1 or UCR2) andshowed that only PDE4D2 promoted cancer cell growth, in-dicating that PDE4D super short form has stronger oncogenicactivity than the long form.Taken together, genomic microdeletions of PDE4D, which

either partially or fully remove UCR1 and/or UCR2 domains,probably lead to elevation of PDE4D catalytic activity. This, inturn, suppresses cellular cAMP concentrations. As shown bymultiple studies, lower cAMP level is one of the favorable fac-tors for the survival and proliferation of cancer cells (6–10, 28,29). Therefore, the truncated PDE4D protein produced by thegenomic deletions might contribute a growth advantage to thecancer cells. Combined with our findings that inhibition ofPDE4D preferentially induced apoptotic cell death in trans-formed cells with a minimal effect on nonmalignant cells, wepropose that targeting this enzyme is a potential therapeuticapproach.

Materials and MethodsTumor Samples and SNP-Chip Analysis. Genomic abnormalities of tumor DNAfrom 55 TNBCs were studied by SNP-Chip. The sources of tumor specimens arelisted in SI Appendix, SI Methods. The clinical–pathological parameters areshown in SI Appendix, Table S7. SNP-Chips for human 250K arrays were used(SNP-Chip, Affymetrix). Partek Genomics Suite 6.6 was used for analysis of

Fig. 6. Targeting PDE4D induced massive cell death preferentially in ma-lignant cells. (A and B) A375 and HGC-27 cells stably expressing either GFP orPDE4D2 were generated, and examined either by cell proliferation assays inthe presence of 1% FBS or by WB analysis with indicated antibodies. (C andD) A375 cells stably expressing either GFP or PDE4D2 were injected s.c. on theupper flanks of nude mice. After 3 wk, the mice were killed and the tumorswere dissected, photographed, and their weights were measured. (E)Chemical structure of the PDE4D inhibitor 26B. (F) Long-term cell pro-liferation assays: various cancer cells were treated with either diluent (DMSO)or 26B (20 μM) for 7–10 d. Cell proliferation was measured by crystal violetstaining and calculated by OD595

26B/OD595DMSO of each cell line and expressed

as a percentage of proliferation in DMSO control. (G) Various cells weretreated with either DMSO or 26B for 48 h and subjected to either WB analysisor (H) FCM analysis by staining with annexin V-FITC/propidium. *P < 0.05;**P < 0.01; ***P < 0.001; NS, not significant. Data represent the mean ± SDof three independent experiments.

Lin et al. PNAS | April 9, 2013 | vol. 110 | no. 15 | 6113

MED

ICALSC

IENCE

S

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

2, 2

020

SNP-chip data from our in-house dataset as well as the ones from WellcomeTrust Sanger Institute Cancer Genome Project.

Cell Culture and Chemicals. All of the human cancer cell lines were maintainedat 37 °C in a 5% (vol/vol) CO2 air humidified incubator in ATCC-suggestedconditions. Melanoma cells (A375 and M368) were generously provided byAntoni Ribas (University of California, Los Angeles). Ovary cancer cells(Caov2 and A2008) and lung cancer cells H838 were kindly provided by RubyHuang and Jizhong Shi (Cancer Science Institute of Singapore). Compound26B was provided by Craig J. Thomas (Chemical Genomics Center, NationalInstitutes of Health, Bethesda, MD). Other chemicals used were of thehighest grade available from Sigma-Aldrich.

Statistical Analysis. The correlation between the expression levels of PDE4Dand MITF was analyzed using the Pearson’s R correlation test. The log-ranktest was used for Kaplan–Meier survival analysis. Other statistical analyses

were conducted using the two-tailed Student t test. Statistical significancewas defined at P < 0.05.

The rest of the descriptions of the methods are provided in SI Appendix,SI Methods.

ACKNOWLEDGMENTS. For triple negative breast cancer samples, the au-thors thank Eng Chon Boon (National University Health System, Singapore)and Kathy O’Briant (Fred Hutchinson Cancer Research Center). The 26B studyis funded in part with federal funds from the Frederick National Labora-tory for Cancer Research, and National Institutes of Health (NIH) underContract HHSN261200800001E. This work is supported by NIH GrantsR01CA026038-32, U54CA143930, and 2P01 HL073104-06; the SingaporeMinistry of Health’s National Medical Research Council (NMRC) under itsSingapore Translational Research (STaR) Investigator Award (to H.P.K.);and a Singapore NMRC New Investigator Grant 2012 (to D.C.). This re-search is supported by the National Research Foundation Singapore andthe Singapore Ministry of Education under the Research Centres of Excellenceinitiative.

1. Bender AT, Beavo JA (2006) Cyclic nucleotide phosphodiesterases: Molecular regula-tion to clinical use. Pharmacol Rev 58(3):488–520.

2. Conti M, Beavo J (2007) Biochemistry and physiology of cyclic nucleotide phosphodies-terases: Essential components in cyclic nucleotide signaling.Annu Rev Biochem 76:481–511.

3. Ke H, Wang H (2007) Crystal structures of phosphodiesterases and implications onsubstrate specificity and inhibitor selectivity. Curr Top Med Chem 7(4):391–403.

4. Lugnier C (2006) Cyclic nucleotide phosphodiesterase (PDE) superfamily: A new targetfor the development of specific therapeutic agents. Pharmacol Ther 109(3):366–398.

5. Marko D, Pahlke G, Merz KH, Eisenbrand G (2000) Cyclic 3′,5′-nucleotide phospho-diesterases: Potential targets for anticancer therapy. Chem Res Toxicol 13(10):944–948.

6. Goldhoff P, et al. (2008) Targeted inhibition of cyclic AMP phosphodiesterase-4 pro-motes brain tumor regression. Clin Cancer Res 14(23):7717–7725.

7. Murata K, et al. (2000) Cyclic AMP specific phosphodiesterase activity and coloncancer cell motility. Clin Exp Metastasis 18(7):599–604.

8. Kim DH, Lerner A (1998) Type 4 cyclic adenosine monophosphate phosphodiesteraseas a therapeutic target in chronic lymphocytic leukemia. Blood 92(7):2484–2494.

9. Ogawa R, Streiff MB, Bugayenko A, Kato GJ (2002) Inhibition of PDE4 phosphodiesteraseactivity induces growth suppression, apoptosis, glucocorticoid sensitivity, p53, and p21(WAF1/CIP1) proteins in human acute lymphoblastic leukemia cells. Blood 99(9):3390–3397.

10. Koeffler HP, Golde DW (1980) Humoral modulation of human acute myelogenousleukemia cell growth in vitro. Cancer Res 40(6):1858–1862.

11. Houslay MD (2001) PDE4 cAMP-specific phosphodiesterases. Prog Nucleic Acid ResMol Biol 69:249–315.

12. Rahrmann EP, et al. (2009) Identification of PDE4D as a proliferation promoting factorin prostate cancer using a Sleeping Beauty transposon-based somatic mutagenesisscreen. Cancer Res 69(10):4388–4397.

13. Pullamsetti SS, et al. (2012) Phosphodiesterase-4 promotes proliferation and angio-genesis of lung cancer by crosstalk with HIF. Oncogene 32(9):1121–1134.

14. Beroukhim R, et al. (2010) The landscape of somatic copy-number alteration acrosshuman cancers. Nature 463(7283):899–905.

15. Zambon AC, Wilderman A, Ho A, Insel PA (2011) Increased expression of the pro-apoptotic protein BIM, a mechanism for cAMP/protein kinase A (PKA)-inducedapoptosis of immature T cells. J Biol Chem 286(38):33260–33267.

16. Garraway LA, et al. (2005) Integrative genomic analyses identify MITF as a lineagesurvival oncogene amplified in malignant melanoma. Nature 436(7047):117–122.

17. Yi X, et al. (2011) MITF-siRNA formulation is a safe and effective therapy for humanmelasma. Mol Ther 19(2):362–371.

18. Carreira S, et al. (2005) Mitf cooperates with Rb1 and activates p21Cip1 expression toregulate cell cycle progression. Nature 433(7027):764–769.

19. Talantov D, et al. (2005) Novel genes associated with malignant melanoma but notbenign melanocytic lesions. Clin Cancer Res 11(20):7234–7242.

20. Eskandarpour M, Huang F, Reeves KA, Clark E, Hansson J (2009) Oncogenic NRAS hasmultiple effects on the malignant phenotype of human melanoma cells cultured invitro. Int J Cancer 124(1):16–26.

21. Aspiotis R, et al. (2010) The discovery and synthesis of highly potent subtype selectivephosphodiesterase 4D inhibitors. Bioorg Med Chem Lett 20(18):5502–5505.

22. Baillie G, MacKenzie SJ, Houslay MD (2001) Phorbol 12-myristate 13-acetate triggersthe protein kinase A-mediated phosphorylation and activation of the PDE4D5 cAMPphosphodiesterase in human aortic smooth muscle cells through a route involvingextracellular signal regulated kinase (ERK). Mol Pharmacol 60(5):1100–1111.

23. Baillie GS, MacKenzie SJ, McPhee I, Houslay MD (2000) Sub-family selective actions inthe ability of Erk2 MAP kinase to phosphorylate and regulate the activity of PDE4cyclic AMP-specific phosphodiesterases. Br J Pharmacol 131(4):811–819.

24. MacKenzie SJ, Baillie GS, McPhee I, Bolger GB, Houslay MD (2000) ERK2 mitogen-activated protein kinase binding, phosphorylation, and regulation of the PDE4DcAMP-specific phosphodiesterases. The involvement of COOH-terminal docking sitesand NH2-terminal UCR regions. J Biol Chem 275(22):16609–16617.

25. Jin SL, Swinnen JV, Conti M (1992) Characterization of the structure of a low Km,rolipram-sensitive cAMP phosphodiesterase. Mapping of the catalytic domain. J BiolChem 267(26):18929–18939.

26. Kovala T, Sanwal BD, Ball EH (1997) Recombinant expression of a type IV, cAMP-specific phosphodiesterase: Characterization and structure-function studies ofdeletion mutants. Biochemistry 36(10):2968–2976.

27. Lim J, Pahlke G, Conti M (1999) Activation of the cAMP-specific phosphodiesterasePDE4D3 by phosphorylation. Identification and function of an inhibitory domain. JBiol Chem 274(28):19677–19685.

28. Bang YJ, et al. (1992) Cyclic AMP induces transforming growth factor beta 2 geneexpression and growth arrest in the human androgen-independent prostate carci-noma cell line PC-3. Proc Natl Acad Sci USA 89(8):3556–3560.

29. Bang YJ, et al. (1994) Terminal neuroendocrine differentiation of human prostatecarcinoma cells in response to increased intracellular cyclic AMP. Proc Natl Acad SciUSA 91(12):5330–5334.

30. Cerami E, et al. (2012) The cBio cancer genomics portal: An open platform for ex-ploring multidimensional cancer genomics data. Cancer Discov 2(5):401–404.

6114 | www.pnas.org/cgi/doi/10.1073/pnas.1218206110 Lin et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

2, 2

020