lives saved by making AVs utilitarian may beoutnumbered by the deaths caused by delayingthe adoption of AVs altogether. Thus, car-makersand regulators alike should be considering solu-tions to these obstacles.Moral algorithms for AVs will need to tackle

more intricate decisions than those considered inour surveys. For example, our scenarios did notfeature any uncertainty about decision outcomes,but a collective discussion aboutmoral algorithmswill need to encompass the concepts of expectedrisk, expected value, and blame assignment. Is itacceptable for an AV to avoid a motorcycle byswerving into a wall, considering that the proba-bility of survival is greater for the passenger ofthe AV than for the rider of themotorcycle? ShouldAVs account for the ages of passengers and pe-destrians (20)? If a manufacturer offers differentversions of itsmoral algorithm, and a buyer know-ingly chose one of them, is the buyer to blame forthe harmful consequences of the algorithm’s de-cisions? Such liability considerations will need toaccompany existing discussions of regulation (21),and we hope that psychological studies inspiredby our own will be able to inform this discussion.Figuring out how to build ethical autonomous

machines is one of the thorniest challenges in ar-tificial intelligence today (22). As we are about toendowmillions of vehicles with autonomy, a seri-ous consideration of algorithmicmorality has nev-er been more urgent. Our data-driven approachhighlights how the field of experimental ethicscan provide key insights into the moral, cultural,and legal standards that people expect from auto-nomous driving algorithms. For the time being,there seems to beno easyway to design algorithmsthat would reconcile moral values and personalself-interest—let alone account for different cul-tures with various moral attitudes regarding life-life trade-offs (23)—but public opinion and socialpressure may very well shift as this conversationprogresses.

REFERENCES AND NOTES

1. B. Montemerlo et al., J. Field Robot. 25, 569–597 (2008).2. C. Urmson et al., J. Field Robot. 25, 425–466 (2008).3. M. M. Waldrop, Nature 518, 20–23 (2015).4. B. van Arem, C. J. van Driel, R. Visser, IEEE Trans. Intell.

Transp. Syst. 7, 429–436 (2006).5. K. Spieser et al., in Road Vehicle Automation, G. Meyer,

S. Beiker, Eds. (Lecture Notes in Mobility Series, Springer,2014), pp. 229–245.

6. P. Gao, R. Hensley, A. Zielke, “A roadmap to the future for theauto industry,” McKinsey Quarterly (October 2014);www.mckinsey.com/industries/automotive-and-assembly/our-insights/a-road-map-to-the-future-for-the-auto-industry.

7. N. J. Goodall, in Road Vehicle Automation, G. Meyer, S. Beiker,Eds. (Lecture Notes in Mobility Series, Springer, 2014),pp. 93–102.

8. K. Gray, A. Waytz, L. Young, Psychol. Inq. 23, 206–215(2012).

9. J. Haidt, The Righteous Mind: Why Good People Are Divided byPolitics and Religion (Pantheon Books, 2012).

10. W. Wallach, C. Allen, Moral Machines: Teaching Robots Rightfrom Wrong (Oxford University Press, 2008).

11. F. Rosen, Classical Utilitarianism from Hume to Mill (Routledge,2005).

12. J. D. Greene, Moral Tribes: Emotion, Reason, and the GapBetween Us and Them (Atlantic Books, 2014).

13. S. Côté, P. K. Piff, R. Willer, J. Pers. Soc. Psychol. 104,490–503 (2013).

14. J. A. C. Everett, D. A. Pizarro, M. J. Crockett, J. Exp. Psychol.Gen. 145, 772–787 (2016).

15. N. E. Kass, Am. J. Public Health 91, 1776–1782 (2001).16. C. R. Sunstein, A. Vermeule, Stanford Law Rev. 58, 703–750 (2005).17. T. Dietz, E. Ostrom, P. C. Stern, Science 302, 1907–1912 (2003).18. R. M. Dawes, Annu. Rev. Psychol. 31, 169–193 (1980).19. P. A. M. Van Lange, J. Joireman, C. D. Parks, E. Van Dijk,

Organ. Behav. Hum. Decis. Process. 120, 125–141 (2013).20. E. A. Posner, C. R. Sunstein, Univ. Chic. Law Rev. 72, 537–598

(2005).21. D. C. Vladeck, Wash. Law Rev. 89, 117–150 (2014).22. B. Deng, Nature 523, 24–26 (2015).23. N. Gold, A. M. Colman, B. D. Pulford, Judgm. Decis. Mak. 9,

65–76 (2014).

ACKNOWLEDGMENTS

J.-F.B. gratefully acknowledges support through the Agence Nationalede la Recherche–Laboratoires d’Excellence Institute for Advanced

Study in Toulouse. This research was supported by internal fundsfrom the University of Oregon to A.S. I.R. is grateful for financialsupport from R. Hoffman. Data files have been uploaded assupplementary materials.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/352/6293/1573/suppl/DC1Materials and MethodsSupplementary TextFig. S1Tables S1 to S8Data Files S1 to S6

15 January 2016; accepted 21 April 201610.1126/science.aaf2654

PROSTATE DEVELOPMENT

Identification of an NKX3.1-G9a-UTYtranscriptional regulatory networkthat controls prostate differentiationAditya Dutta,1* Clémentine Le Magnen,1* Antonina Mitrofanova,2† Xuesong Ouyang,3‡Andrea Califano,4 Cory Abate-Shen5§

The NKX3.1 homeobox gene plays essential roles in prostate differentiation and prostate cancer.We show that loss of function of Nkx3.1 in mouse prostate results in down-regulation ofgenes that are essential for prostate differentiation, as well as up-regulation of genes thatare not normally expressed in prostate. Conversely, gain of function of Nkx3.1 in anotherwise fully differentiated nonprostatic mouse epithelium (seminal vesicle) is sufficientfor respecification to prostate in renal grafts in vivo. In human prostate cells, theseactivities require the interaction of NKX3.1 with the G9a histone methyltransferase via thehomeodomain and are mediated by activation of target genes such as UTY (KDM6c), themale-specific paralog of UTX (KDM6a). We propose that an NKX3.1-G9a-UTYtranscriptional regulatory network is essential for prostate differentiation, and wespeculate that disruption of such a network predisposes to prostate cancer.

Among the tissues of the male urogenitalsystem, the prostate and seminal vesicle aresecretory organs that develop in close prox-imity under the influence of androgens (fig.S1A) (1, 2). However, the prostate develops

from the urogenital sinus, an endodermal deriv-

ative, whereas the seminal vesicle develops fromthe Wolffian duct, a mesodermal derivative. Amonggenes that distinguish prostate and seminal vesicle,the Nkx3.1 homeobox gene is among the earliestexpressed in the presumptive prostatic epitheliumduring development, and its expression in adultsis primarily restricted to prostatic luminal cells(3, 4), which are secretory cells that are the majortarget of prostate neoplasia (4, 5). Accordingly, inmouse models, loss of function of Nkx3.1 resultsin impaired prostate differentiation and defects inluminal stem cells, as well as predisposes to pros-tate cancer (3, 4).Analyses of expression profiles from Nkx3.1

wild-type (Nkx3.1+/+) andNkx3.1mutant (Nkx3.1–/–)prostates revealed down-regulation of genes asso-ciated with prostate differentiation, such as FoxA1(Forkhead Box A1), Pbsn (Probasin), HoxB13, andTmprss2 (Transmembrane Protease, Serine 2), aswell as luminal cells (cytokeratins 8 and 18), andup-regulation of basal cell markers (cytokeratins5 and p63) (Fig. 1A, fig. S2A, and database S1) (6).Surprisingly, Nkx3.1–/– versus Nkx3.1+/+ prostatesdisplay up-regulation of genes that are expressed,albeit not exclusively, in seminal vesicle, namely

1576 24 JUNE 2016 • VOL 352 ISSUE 6293 sciencemag.org SCIENCE

1Departments of Medicine and Urology, Institute of CancerGenetics, Herbert Irving Comprehensive Cancer Center,Columbia University Medical Center, New York, NY 10032,USA. 2Department of Systems Biology, Columbia UniversityMedical Center, New York, NY 10032, USA. 3Department ofUrology, Columbia University Medical Center, New York, NY10032, USA. 4Departments of Systems Biology, BiomedicalInformatics, and Biochemistry and Molecular Biophysics,Center for Computational Biology and Bioinformatics,Institute of Cancer Genetics, Herbert Irving ComprehensiveCancer Center, Columbia University Medical Center, NewYork, NY 10032, USA. 5Departments of Urology, Medicine,Systems Biology, and Pathology and Cell Biology, Institute ofCancer Genetics, Herbert Irving Comprehensive CancerCenter, Columbia University Medical Center, New York, NY10032, USA.*These authors contributed equally to this work. †Present address:Department of Health Informatics, Rutgers, The State University ofNew Jersey, 65 Bergen Street, Room 350B, Newark, NJ 07101,USA. ‡Present address: Crown Bioscience Inc., 6 West BeijingRoad, Taicang, Jiangsu 215400, China. §Corresponding author.Email: cabateshen@columbia.edu

RESEARCH | REPORTSon M

arch 13, 2021

http://science.sciencemag.org/

Dow

nloaded from

Svs6, Sva, Svs4, and Svs5 (Fig. 1A and fig. S2A)(7). These differentially expressed genes were sig-nificantly enriched in a gene signature comparingseminal vesicle versus normal prostate, and thispattern was conserved in mouse and humans(Fig. 1B and figs. S1B and S2B). At the cellularlevel, we observed reduced expression of Proba-sin and a corresponding up-regulation of Svp2 inNkx3.1–/– versus Nkx3.1+/+ prostates (figs. S1, Cto E, and S2C; and tables S1 to S3) (8).Considering that loss of function of Nkx3.1 leads

to aberrant prostate epithelial differentiation, weasked whether its gain of function in a nonprostaticepithelium is sufficient to induce prostate differ-entiation. Toward this end, we performed tissuerecombination assays, in which relevant epithelialand mesenchymal tissues are recombined in vitro,followed by growth under the kidney capsule ofhost mice in vivo (Fig. 1C) (2). It is well estab-lished that prostate formation requires both epi-thelial and mesenchymal tissues, as well as anappropriate source of androgens (fig. S3), and thatnonendodermal epithelium, even those that are

androgen-regulated, such as seminal vesicle, donot form prostate in this assay (2).To explore whether Nkx3.1 expression can in-

duce prostate differentiation in tissue recombi-nants, we infected seminal vesicle epithelium witha lentivirus expressing Nkx3.1 (or the controlvector) (Fig. 1C). As expected, tissue recombinantsmade with prostate epithelium (PE) generateprostate-like grafts, which are distinguished bytheir prostate-like ductal structures, histologicalappearance resembling prostate epithelium, andexpression of markers of prostate differentiation,including Nkx3.1, probasin, FoxA1, and HoxB13(N = 19 recombinants) (Fig. 1, D and E; fig. S4A;and table S4). Also as expected, tissue recombi-nants made with seminal vesicle epithelium(SVE) lacking Nkx3.1 generate seminal vesicle–like grafts, which are distinguished by their lackof discernible ductal morphology, histological ap-pearance resembling seminal vesicle, and theabsence of markers of prostate differentiation(N = 20 recombinants) (Fig. 1, D and E; fig. S4A;and table S4).

In contrast, tissue recombinants made withSVE expressing exogenous Nkx3.1 more closelyresemble prostate than seminal vesicle. In par-ticular, the Nkx3.1-expressing SVE grafts are dis-tinguished by their appearance of prostate-likeductal structures, histological appearance resem-bling prostate, and expression of markers of pros-tate differentiation, including probasin, FoxA1,and HoxB13 (N = 26 recombinants) (Fig. 1, Dand E; fig. S4A; and table S4); this was not thecase for tissue recombinants made with SVEexpressing an unrelated homeobox gene, Msx1(N = 3 recombinants) (Fig. 1, D and E, andtable S4). Moreover, expression profiling anal-yses showed that tissue recombinants madefrom SVE expressing Nkx3.1 were significantlyenriched with a signature of prostate versusseminal vesicle (fig. S4, B and C). Nkx3.1 is thussufficient to respecify a fully differentiated non-prostate epithelium to form prostate in vivo.To study the underlying mechanisms, we es-

tablished a cell-based assay using an immortal-ized human prostate cell line, RWPE1, which

SCIENCE sciencemag.org 24 JUNE 2016 • VOL 352 ISSUE 6293 1577

Fig. 1. Murine Nkx3.1 respecifies a nonprostatic epithelium to formprostate in vivo. (A) Heat map representations of differentially expressedgenes from Nkx3.1+/+ and Nkx3.1−/− prostate (6). (B) Gene set enrichmentanalysis, using as the query gene set differentially expressed genes fromseminal vesicle versus prostate, compared with a reference gene signatureof Nkx3.1−/− versus Nkx3.1+/+ prostate. (C) Diagram of the tissue recombi-nation assay. Dissociated epithelium from seminal vesicle (SVE) or prostate

(PE) is infected with a lentivirus expressing Nkx3.1 (or control). Mesenchymefrom rat embryonic urogenital sinus is combined with the epithelium and grownunder the renal capsule of host mice. (D and E) Representative tissue re-combinants. (D) (Top) Whole-mount images. (Bottom) Hematoxylin and eosin(H&E) images. (E) Confocal images of immunofluorescence using the indicatedantibodies. Scale bars represent 50 mm in (D) and 20 μm in (E). A summary oftissue recombinant data is provided in table S4.

RESEARCH | REPORTSon M

arch 13, 2021

http://science.sciencemag.org/

Dow

nloaded from

1578 24 JUNE 2016 • VOL 352 ISSUE 6293 sciencemag.org SCIENCE

Fig. 3. Induction of prostate differentia-tion by NKX3.1 is mediated through itsinteraction with the G9a histone methyl-transferase. (A and B) Nuclear extractsfromRWPE1cells expressingFlag-HA-NKX3.1or the control were subjected to immuno-precipitation followedbymassspectrometry(A) orWestern blot analysis (B) (see fig. S6).(A)SilverstainshowingG9a interaction.Mark-ers, as indicated. NS, nonspecific bands. (B)Immunoprecipitation followed by Westernblot analysis. Input shows 5% of the totalprotein, and immunoprecipitation (IP) showsproteins recovered after IP usingan antibodyto Flag. Experiments were performed withthree independent biological replicates; representative data are shown.(C) Diagram of experimental design for (D) and (E). Human RWPE1 prostateepithelial cells were infected with an NKX3.1-expressing lentivirus (expressingred fluorescent protein), followed by infection with an shRNA-expressinglentivirus (expressing green fluorescent protein). Coinfected cells were sortedby fluorescence-activated cell sorting, followed by analyses in vitro (D) orgeneration of tissue recombinants in vivo. (D) Western blot analysis. Experi-

ments were performed with three independent biological replicates; repre-sentative data are shown. (E) Representative tissue recombinants showingwhole-mount, H&E, and confocal images of immunofluorescence staining.Indicated is the kidney and the collagen plug (for the recombinants that didnot grow) or the tissue recombinant. The ruler shows cm scale; scale barsrepresent 50 mm in the H&E images and 20 mm in the immunofluorescenceimages. A summary of tissue recombinants is provided in table S4.

Fig. 2. Induction of prostate differentiation by NKX3.1 requires the homeodo-main. (A) Diagram of the experimental design. Human RWPE1 prostate epithelial cellsare infected with a lentivirus expressing human NKX3.1, NKX3.1(T164A), or a control, fol-lowed by analyses in vitro (B to D) or recombined with mesenchyme and grown under therenal capsule of host mice (E). (B) Western blot analyses. Actin is a control for proteinloading. (C) Gel retardation analysis done using nuclear extracts from RWPE1 cells ex-pressing the control vector, NKX3.1, or NKX3.1 (T164A). The arrow indicates the freeDNA probe. The experiments in (B) and (C) were each performed with three independentbiological replicates; representative data are shown. (D) Heat map representations of se-lected differentially expressed genes; a complete list is provided in database S4. (E) Tissuerecombinants showing whole-mount images, H&E, and immunofluorescence staining.Theruler shows cm scale; scale bars represent 50 mm in the H&E images and 20 mm in theimmunofluorescence images. A summary of all tissue recombinants is provided in table S4.

RESEARCH | REPORTSon M

arch 13, 2021

http://science.sciencemag.org/

Dow

nloaded from

expresses low levels of NKX3.1, low levels of lu-minal cytokeratins, high levels of basal cell markers,and low levels of markers of prostate differenti-ation, such as androgen receptor (AR), FOXA1,PSA, TMPRSS2, and HOXB13 (Fig. 2, A to D). In-fection of RWPE1 cells with a lentivirus expressingNKX3.1 resulted in high levels of NKX3.1 proteinand robust DNA binding (Fig. 2, B and C). Ex-pression profiling and Western blot analyses ofthe NKX3.1-expressing versus control RWPE1 cellsrevealed up-regulation of luminal cell markers(cytokeratins 8 and 18), down-regulation of basalcell markers (cytokeratin 5 and p63), and up-regulation of prostate differentiation markers(AR, FOXA1, PSA, TMPRSS2, and HOXB13) (Fig. 2,B and D). Furthermore, when combined with

embryonic mesenchyme and grown under therenal capsule, NKX3.1-expressing RWPE1 cellsgenerate prostate-like grafts that morphologicallyand histologically resemble prostate, includingthe presence of basal and luminal cell layers andexpression of markers of prostate differentiation(N = 15 recombinants). In contrast, the controlRWPE1 cells failed to grow in this assay (N = 12recombinants) (Fig. 2E, fig. S5, and table S4).NKX3.1(T164A), which has a mutation in thehomeodomain that impairs its DNA bindingcapacity (Fig. 2C) (9), did not induce prostatedifferentiation in vitro, nor did it promoteprostate growth in tissue recombinants in vivo(N = 7 recombinants) (Fig. 2, B to E; fig. S5; andtable S4). Therefore, the ability of NKX3.1 to

induce prostate differentiation requires a func-tional DNA binding domain.Many of the functions of homeoproteins are

mediated by protein-protein interactions thatoften occur through the homeodomain (10).Among NKX3.1-interacting proteins identifiedby mass spectrometry (Fig. 3A and fig. S6) (8)was G9a [also called EHMT2 (euchromatic histonelysine N-methyltransferase 2)], a histone methyl-transferase that forms a complex with a relatedhistonemethyltransferase, GLP [also called EHMT1(euchromatic histone lysine N-methyltransferase 1)],to promote dimethylation of lysine 9 on histone3 (H3K9me2) (11). G9a is essential for embryonicdevelopment (12) and interacts with other hom-eoproteins to mediate differentiation (13, 14). In

SCIENCE sciencemag.org 24 JUNE 2016 • VOL 352 ISSUE 6293 1579

Fig. 4. Induction of prostate differentiation by NKX3.1is mediated by UTY, a male-specific paralog of histonedemethylase UTX (KDM6c). (A) Real-time polymerasechain reaction (PCR) showing expression of NKX3.1 tar-get genes, UTY and EDEM2 (left). Chromatin immuno-precipitation quantitative PCR analysis of NKX3.1 binding(center) and G9a binding (right) to NKX3.1 target geneswere performed. Analyses were performed with three in-dependent biological replicates. Statistical analysis wasdone using a two-tailed t test; data are indicated as mean± SD. (B) Diagram of the experimental design for (C) to (E).Human RWPE1 cells (C and D) or mouse tissues (E) wereinfected with an NKX3.1-expressing lentivirus, followed byinfection with an shRNA (or controls). Cells were analyzedin vitro [human (C)] or in tissue recombinants [human (D)andmouse (E)]. (D) Western blot analysis of RPWE1 cells.[(D) and (E)] Representative tissue recombinants ofhuman (D) andmouse (E) showingwhole-mount, H&E, andconfocal images of immunofluorescence staining.The rulershows cm scale; scale bars represent 50 mm in the H&Eimages and 20 mm in the immunofluorescence images. Asummary of tissue recombinants is provided in table S4.

RESEARCH | REPORTSon M

arch 13, 2021

http://science.sciencemag.org/

Dow

nloaded from

coimmunoprecipitation assays, NKX3.1 inter-acted with endogenous G9a, as well as GLP,which requires the homeodomain and is directlycorrelated with DNA binding by NKX3.1 (Fig.3B and fig. S7, A to C). In contrast, NKX3.1 didnot interact with other histone methyltransfer-ases, including SUV39H1 (suppressor of varie-gation 3-9 homolog 1), which promotes trimethylationof lysine 9 on histone 3 (H3K9me3), and EZH2(enhancer of zeste 2), which promotes trime-thylation of lysine 27 on histone 3 (H3K27me3)(fig. S7, A and B).To study the functional relevance of the inter-

action between NKX3.1 and G9a, we coinfectedRWPE1 cells with an NKX3.1-expressing lentivirustogether with a short hairpin RNA (shRNA) todeplete G9a (shG9a) or SUV39H1 (shSUV39H1)as a control (Fig. 3, C to E, and fig. S5). TheseshRNAs reduced expression of G9a or SUV39H1,as well as their respective histone marks, H3K9me2and H2K9me3, while not affecting expression ofNKX3.1 (Fig. 3D). However, depletion of G9a, butnot SUV39H1, impaired the ability of NKX3.1 toinduce prostate differentiation, as evidenced byWestern blot analysis of cultured cells, as well asprostate growth in the tissue recombinant assayin vivo (N = 9 recombinants) (Fig. 3, D and E;fig. S5; and table S4). These findings demon-strate that the interaction of NKX3.1 with G9a isrequired for induction of prostate differentiation.Considering that induction of differentiation

by NKX3.1 requires its homeodomain and cor-responding DNA binding activity (fig. S7), wereasoned that these functions are likely to bemediated by gene transcription. Among NKX3.1target genes that have predicted functions fordifferentiation and are conserved between miceandhumans (fig. S8) (8), we focused onUTY, usingEDEM2 as a control. In particular, G9a binds tothe promoter of UTY, but not EDEM2, which isdependent on NKX3.1 binding and is requiredfor NKX3.1-mediated up-regulation ofUTY expres-sion (Fig. 4A and fig. S9). Notably, UTY (KDM6c,ubiquitously transcribed tetratricopeptide repeatcontaining, Y-linked) is the male-specific paralogof UTX (KDM6a), a histone demethylase that isessential for viability and is frequently deregulatedin cancer (15–17). Although its functions as ahistone demethylase are uncertain (16), UTY isessential for male fertility as well as the de-velopment and differentiation of male-specificorgans (16), and it has been linked to prostatecancer (18).To study the consequences of UTY depletion

on prostate differentiation, we coinfected NKX3.1-expressing or control RWPE1 cells with an shRNAto UTY (shUTY) or EDEM2 (shEDEM2) as a con-trol (Fig. 4, B to D, and fig. S5). Expression ofthese shRNAs reduced expression of UTY orEDEM2, respectively, while not affecting expres-sion of NKX3.1 (Fig. 4C). Moreover, depletion ofUTY, but not EDEM2, impaired the ability ofNKX3.1 to induce prostate differentiation in vitroas well as in tissue recombinants in vivo (N = 8recombinants) (Fig. 4, C and D, and table S4).We next investigated whether Uty is also re-

quired for prostate specification in vivo by per-

forming analogous experiments using the mousetissue recombinant model. Depletion of Uty(shUty) in Nkx3.1-expressing SVE abrogated theability of Nkx3.1 to generate prostate, as was evi-dent by the resulting tissue recombinants (SVE +Nkx3.1 + shUty), which more closely resembleseminal vesicle than prostate (N = 8 recombi-nants) (Fig. 4E). Moreover, expression profilingof tissue recombinants generated from the Uty-depleted Nkx3.1-SVE revealed their significantenrichment in a gene signature comparing sem-inal vesicle versus prostate (fig. S10).Cumulatively, our findings support a model in

which NKX3.1 regulates the expression of a geneprogram associated with prostate differentiation,while simultaneously inhibiting the inappropriateexpression of nonprostatic genes (fig. S11A). Inparticular, Nkx3.1 can respecify a fully differ-entiated mouse tissue to form prostate in vivo,and its expression in basal-like human prostateepithelial cells can promote their differentiationto luminal-like cells that form prostate in vivo.These functions of NKX3.1 are mediated by thecoordinate actions of G9a and UTY (fig. S11, Aand B). Notably, we find that G9a functions as acoregulator of NKX3.1, which is associated withactivation, as well as repression, of NKX3.1 targetgenes. This further underscores the complexity ofG9a function in transcriptional control. Indeed,although G9a is a histone methyltransferase fora “repressive”mark, it is active on euchromatin(11), and it has been shown that G9a can repressor activate transcription depending on the con-text (19, 20). This includes the glucocorticoidreceptor (21, 22), which also interacts withNKX3.1, as assessed by mass spectrometry (fig.S6), and has been implicated in drug resistancein prostate cancer (23). Thus, we envision thatour findings presage a role for G9a in both pros-tate differentiation and cancer.Our findings also shed new light on UTY as

an essential downstream mediator of NKX3.1 inprostate differentiation. Given the importance ofUTY for male fertility, as well as for differentia-tion of male-specific organs (16, 24), our currentstudy provides an example of how, in additionto the well-known role of androgen signaling, thepromotion of “maleness” may be essential forprostate differentiation.Last, there are still relatively few examples in

which a single gene can respecify an otherwisefully differentiated epithelium to a new fate, aswe have observed for NKX3.1. Notably, our find-ings regarding NKX3.1 in prostate are strikinglyconcordant with the functions of NKX2.1 in lungdifferentiation and lung cancer (25–28). In par-ticular, loss of function of NKX2.1 leads to im-paired lung differentiation, which is associatedwith the derepression of an aberrant gene ex-pression program (28), while, conversely NKX2.1expression is associated with inhibition of lungmetastases (26, 27). Consistent with the action ofNKX3.1 in prostate, these actions of NKX2.1 inlung are dependent upon its level of expression.Thus, these observations suggest that NK-classhomeobox genes function as key regulators oftissue-specific differentiation, as well as key gate-

keepers whose diminution of expression in spe-cific tissue contexts may enhance susceptibilityto cancer.

REFERENCES AND NOTES

1. C. Abate-Shen, M. M. Shen, Genes Dev. 14, 2410–2434(2000).

2. G. R. Cunha et al., Endocr. Rev. 8, 338–362 (1987).3. R. Bhatia-Gaur et al., Genes Dev. 13, 966–977

(1999).4. X. Wang et al., Nature 461, 495–500 (2009).5. Z. A. Wang, R. Toivanen, S. K. Bergren, P. Chambon,

M. M. Shen, Cell Reports 8, 1339–1346 (2014).6. X. Ouyang, T. L. DeWeese, W. G. Nelson, C. Abate-Shen,

Cancer Res. 65, 6773–6779 (2005).7. M. C. Ostrowski, M. K. Kistler, W. S. Kistler, J. Biol. Chem. 254,

383–390 (1979).8. Materials and methods are available as supplementary

materials on Science Online.9. S. L. Zheng et al., Cancer Res. 66, 69–77 (2006).10. T. R. Bürglin, M. Affolter, Chromosoma 2015, 1–25

(2015).11. Y. Shinkai, M. Tachibana, Genes Dev. 25, 781–788

(2011).12. M. Tachibana et al., Genes Dev. 16, 1779–1791

(2002).13. J. Wang, C. Abate-Shen, PLOS ONE 7, e37647 (2012).14. B. Lehnertz et al., Genes Dev. 28, 317–327 (2014).15. C. Wang et al., Proc. Natl. Acad. Sci. U.S.A. 109, 15324–15329

(2012).16. K. B. Shpargel, T. Sengoku, S. Yokoyama, T. Magnuson, PLOS

Genet. 8, e1002964 (2012).17. G. van Haaften et al., Nat. Genet. 41, 521–523 (2009).18. Y. F. Lau, J. Zhang, Mol. Carcinog. 27, 308–321 (2000).19. C. P. Chaturvedi et al., Proc. Natl. Acad. Sci. U.S.A. 106,

18303–18308 (2009).20. B. Lehnertz et al., J. Exp. Med. 207, 915–922 (2010).21. D. Bittencourt et al., Proc. Natl. Acad. Sci. U.S.A. 109,

19673–19678 (2012).22. D. Y. Lee, J. P. Northrop, M. H. Kuo, M. R. Stallcup, J. Biol.

Chem. 281, 8476–8485 (2006).23. V. K. Arora et al., Cell 155, 1309–1322 (2013).24. H. J. Cooke, P. T. Saunders, Nat. Rev. Genet. 3, 790–801

(2002).25. B. A. Weir et al., Nature 450, 893–898 (2007).26. M. M. Winslow et al., Nature 473, 101–104 (2011).27. C. M. Li et al., Genes Dev. 29, 1850–1862 (2015).28. E. L. Snyder et al., Mol. Cell 50, 185–199 (2013).

ACKNOWLEDGMENTS

We are grateful to A. Aytes, F. Constantini, E. Gelmann,D. Reinberg, and M. Shen for comments on the manuscript.We thank C. Bieberich (University of Maryland) for Nkx3.1antisera. We acknowledge support from the JP SulzbergerColumbia Genome Center and the Proteomics Shared Resource,which are shared resources of the Herbert Irving ComprehensiveCancer Center at Columbia University, supported in part byNIH/NCI grant P30 CA013696. This research is supported byfunding to C.A.-S. from the National Cancer Institute (CA154293).A.D. was supported in part by the National Center forAdvancing Translational Sciences, National Institutes of Health,grant UL1 TR000040. C.L.M. was supported by the Swiss NationalScience Foundation (PBBSP3_146959 and P300P3_151158).A.M. is a recipient of a Prostate Cancer Foundation YoungInvestigator Award. C.A.-S. is an American Cancer SocietyResearch Professor supported in part by a generous gift from theF.M. Kirby Foundation.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/352/6293/1576/suppl/DC1Materials and MethodsSupplementary TextFigs. S1 to S11Tables S1 to S6Databases S1 to S4References (29–44)

29 November 2015; accepted 27 May 201610.1126/science.aad9512

1580 24 JUNE 2016 • VOL 352 ISSUE 6293 sciencemag.org SCIENCE

RESEARCH | REPORTSon M

arch 13, 2021

http://science.sciencemag.org/

Dow

nloaded from

differentiationIdentification of an NKX3.1-G9a-UTY transcriptional regulatory network that controls prostate

Aditya Dutta, Clémentine Le Magnen, Antonina Mitrofanova, Xuesong Ouyang, Andrea Califano and Cory Abate-Shen

DOI: 10.1126/science.aad9512 (6293), 1576-1580.352Science 

, this issue p. 1576Scienceprobably contributes to prostate cancer development.with prostate differentiation by interacting with the G9a histone methyltransferase. Disruption of this regulatory network

regulates the expression of a gene program associatedNKX3.1seminal vesicle epithelium to differentiate into prostate. , causesNKX3.1 show that forced expression of a single gene, the homeobox gene et al.and mouse models, Dutta

tissues could help solve why cancer arises frequently in the prostate but only rarely in seminal vesicles. Working with cellandrogenic hormones. A better understanding of the molecular mechanisms controlling the development of the two

The prostate and seminal vesicle have closely related developmental histories and both are regulated by the sameClues to cancer from an identity change

ARTICLE TOOLS http://science.sciencemag.org/content/352/6293/1576

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2016/06/22/352.6293.1576.DC1

CONTENTRELATED

http://stm.sciencemag.org/content/scitransmed/7/292/292ra101.fullhttp://stm.sciencemag.org/content/scitransmed/7/312/312re10.fullhttp://stm.sciencemag.org/content/scitransmed/7/312/312re11.fullhttp://stm.sciencemag.org/content/scitransmed/8/333/333ra47.full

REFERENCES

http://science.sciencemag.org/content/352/6293/1576#BIBLThis article cites 43 articles, 20 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Copyright © 2016, American Association for the Advancement of Science

on March 13, 2021

http://science.sciencem

ag.org/D

ownloaded from