REPORTS

rins, is not sufficient to colocalize these proteins at tacts, potentially increasing the local level of 22. Z. Wang et al., Science 304, 1164 (2004). the cell junctions. The PTPs generally have little phosphatase activity. Because of the high affinity 23. J. A. Besco, R. Hooft van Huijsduijnen, A. Frostholm, A. Rotter, Brain Res. 1116, 50 (2006). substrate specificity, and they rely on noncatalytic of the trans interaction, the balance between cell 24. M. Fuchs, T. Muller, M. M. Lerch, A. Ullrich, J. Biol. Chem. domains to control their subcellular distribution adhesion versus mobility can only be shifted by 271, 16712 (1996). and therefore indirectly regulate their activity by the action of the ADAM 10 protease (14). In both 25. G. C. M. Zondag, A. B. Reynolds, W. H. Moolenaar, J. Biol. restricting access to particular substrates at de- CD45 and RPTPm, however, ectodomain size and Chem. 275, 11264 (2000). 6 32 26. S. M. Brady-Kalnay et al., J. Cell Biol. 141, 287 (1998). fined locations ( , ). RPTPs are known to be rigidity appear to provide a mechanism to allow 27. X. F. Sui et al., Am. J. Pathol. 166, 1247 (2005). constitutively active, and ligand-induced inactiva- cell-cell spacings to regulate intercellular multi- 28. K. Miyaguchi, J. Struct. Biol. 132, 169 (2000). tion has been reported for type I and IV subfam- molecular assemblies. 29. T. J. Boggon et al., Science 296, 1308 (2002). ilies. Such a mechanism is unlikely to apply to type 30. L. A. Staehelin, Int. Rev. Cytol. 39, 191 (1974). IIB RPTPs, where an active enzyme would be References and Notes 31. W. He, P. Cowin, D. L. Stokes, Science 302, 109 (2003). 32. I. A. Yudushkin et al., Science 315, 115 (2007). required to maintain cadherin-catenin complexes 1. M. Perez-Moreno, C. Jamora, E. Fuchs, Cell 112, 535 (2003). 33. S. J. Davis, P. A. van der Merwe, Nat. Immunol. 7, 803 in a dephosphorylated state and thus contribute to 2. B. M. Gumbiner, Nat. Rev. Mol. Cell Biol. 6, 622 (2005). (2006). the stability of cell contacts (2). In this context, for 3. J. L. Sallee, E. S. Wittchen, K. Burridge, J. Biol. Chem. 34. K. Choudhuri, D. Wiseman, M. H. Brown, K. Gould, type IIB RPTPs, the ectodomain-mediated trans 281, 16189 (2006). P. A. van der Merwe, Nature 436, 578 (2005). 35. Single-letter abbreviations for the amino acid residues are homophilic interactions appear to represent the 4. R. L. Del Vecchio, N. K. Tonks, J. Biol. Chem. 280, 1603 (2005). as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; driving force for correct localization and function. 5. M. F. Gebbink et al., J. Cell Biol. 131, 251 (1995). H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; Our results on RPTPm suggest how the type 6. N. K. Tonks, Nat. Rev. Mol. Cell Biol. 7, 833 (2006). R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. IIB RPTPs modulate the stability of adherens 7. J. Besco, M. C. Popesco, R. V. Davuluri, A. Frostholm, 36. T. Maretzky et al., Proc. Natl. Acad. Sci. U.S.A. 102, 9182 A. Rotter, BMC Genomics 5, 14 (2004). (2005). junctions (Fig. 4). The ectodomain trans interac- m 8. A. R. Aricescu et al., EMBO J. 25, 701 (2006). 37. Coordinates and structure factors of eRPTP have been tionisswitchedoffatacidpH(8, 18) (i.e., until deposited in the PDB (www.rcsb.org) with the accession m 9. A. R. Aricescu, W. Lu, E. Y. Jones, Acta Crystallogr. D 62, RPTP reaches the cell surface). The rigid, ruler- 1243 (2006). number 2V5Y. We thank the staff of the ID 29 beamline like ectodomain then acts as a sensor of inter- 10. V. T. Chang et al., Structure 15, 267 (2007). at the European Synchrotron Radiation Facility for m cellular distances, matching cadherin-mediated 11. Materials and methods are available as supporting assistance with data collection; M. Gebbink for the RPTP material on Science Online. cDNA; P. Reeves and H. G. Khorana for the human cell contacts, at which point the trans interaction – – 12. V. Soroka et al., Structure 11, 1291 (2003). embryonic kidney 293S GnTI cell line; M. Shaw for serves as a spacer clamp, locking the phosphatase 13. M. Campan et al., Biochemistry 35, 3797 (1996). assistance with EM imaging; and J. Brown, M. Crispin, activity into proximity with the target substrates. 14. L. Anders et al., Mol. Cell. Biol. 26, 3917 (2006). W.-C. Hon, and D. Stuart for discussions. The work was The spacer-clamp action of RPTPm represents the 15. V. B. Cismasiu, S. A. Denes, H. Reilander, H. Michel, funded by Cancer Research UK (CR-UK). E.Y.J. is a CR-UK Principal Research Fellow. inverse strategy to the size-exclusion mechanism S. E. Szedlacsek, J. Biol. Chem. 279, 26922 (2004). 16. S. M. Brady-Kalnay, A. J. Flint, N. K. Tonks, J. Cell Biol. proposed to regulate the cell surface location of 122, 961 (1993). Supporting Online Material www.sciencemag.org/cgi/content/full/317/5842/1217/DC1 another RPTP, CD45; in that case, the mismatch 17. J. Cheng et al., J. Biol. Chem. 272, 7264 (1997). Materials and Methods between the RPTP ectodomain and the inter- 18. M. F. Gebbink et al., J. Biol. Chem. 268, 16101 (1993). 19. G. C. M. Zondag et al., J. Biol. Chem. 270, 14247 (1995). Figs. S1 to S6 cellular spacing is thought to contribute to T cell Table S1 20. J. Sap, Y. P. Jiang, D. Friedlander, M. Grumet, J. Schlessinger, References signaling by expelling the phosphatase activity Mol. Cell. Biol. 14, 1 (1994). from local zones of cell-cell contact (33, 34). 21. M. Fuchs, H. Wang, T. Ciossek, Z. Chen, A. Ullrich, Mech. 4 May 2007; accepted 17 July 2007 Unlike CD45, RPTPm is maintained at cell con- Dev. 70, 91 (1998). 10.1126/science.1144646

the differentiation of murine embryonic stem A MicroRNA Feedback Circuit (ES) cells into DNs (4, 5). An ES cell line was obtained that expresses enzyme condition- in Midbrain Dopamine Neurons ally [containing LoxP recombinase sites that flank both chromosomal copies of the Dicer Jongpil Kim,1 Keiichi Inoue,1 Jennifer Ishii,1 William B. Vanti,1 Sergey V. Voronov,1 , herein termed floxed Dicer (6)]. Introduc- Elizabeth Murchison,2 Gregory Hannon,2 Asa Abeliovich1* tion of Cre recombinase into these cells by lentiviral transduction leads to the deletion of (miRNAs) are evolutionarily conserved, 18- to 25-nucleotide, non–protein coding Dicer in nearly 100% of cells (fig. S1A). transcripts that posttranscriptionally regulate during development. miRNAs also ES cultures were differentiated to a midbrain occur in postmitotic cells, such as neurons in the mammalian central nervous system, but their DN phenotype using the embryoid body (EB) function is less well characterized. We investigated the role of miRNAs in mammalian midbrain protocol (fig. S1B) (5, 7). Cre-mediated deletion dopaminergic neurons (DNs). We identified a miRNA, miR-133b, that is specifically expressed in of Dicer at a stage when postmitotic DNs first midbrain DNs and is deficient in midbrain tissue from patients with Parkinson’s disease. miR-133b arise led to a nearly complete loss of DN accu- regulates the maturation and function of midbrain DNs within a negative feedback circuit that mulation, as quantified by the expression of includes the paired-like homeodomain transcription factor Pitx3. We propose a role for this markers including tyrosine hydroxylase (TH) feedback circuit in the fine-tuning of dopaminergic behaviors such as locomotion. (Fig. 1A). Other mature neuronal classes, includ- ing GABAergic neurons, were reduced in these icroRNAs (miRNAs) are derived from Midbrain dopaminergic neurons (DNs) play a long primary transcripts through se- central role in complex behaviors such as reward 1Departments of Pathology and Neurology, Center for Mquential processing by the Drosha and addiction, and these cells are lost in Par- Neurobiology and Behavior, and Taub Institute, Columbia ribonuclease and the Dicer enzyme (1). In the kinson’s disease. A number of transcription fac- University, College of Physicians and Surgeons 15-403, 630 West 168th Street, New York, NY 10032, USA. context of an RNA-induced silencing complex, tors have been identified that regulate midbrain 2 3 Watson School of Biological Sciences, Cold Spring Harbor miRNAs guide the cleavage of target mRNAs DN development, function, and survival ( ). Laboratory, 1 Bungtown Road, Cold Spring Harbor, NY and/or inhibit their translation. miRNAs regulate However, the role of posttranscriptional mecha- 11724, USA. developmental cell fate decisions in the nervous nisms is unknown. To establish a function for *To whom correspondence should be addressed. E-mail: system and elsewhere (2). miRNAs, we first used an in vitro model system: [email protected]

1220 31 AUGUST 2007 VOL 317 SCIENCE www.sciencemag.org REPORTS

Fig. 1. Dicer is essential for the midbrain DN phenotype. (A) Floxed Dicer conditional knockout ES cultures (flx/flx) were differentiated by the EB method, transduced with Cre or control green fluorescent protein (GFP) lentivirus, and analyzed by immunostaining with antibodies specific for TH (red), TujI (green), and GABA (blue). Cultures transduced with a lentiviral Cre vector (vCre) but not control GFP lentivirus (vGFP) were essentially devoid of TH+ neurons, whereas TujI+ and GABA+ cells were re- duced by approximately 40 to 60%. (n =3independent samples per group). Scale bar, 100 mm. Data represent mean ± SEM; analysis loss of 90% of midbrain DNs in the substantia nigra (SN) and ventral of variance (ANOVA) test, *P <0.05.(B) The Dicer deletion phenotype, as in tegmental area (VTA) and their axonal projections to the striatum relative to (A), can be “rescued” by transfection of midbrain-derived small RNAs (<200 control littermates (DATCRE/+:Dicer flox/+)(n = 3 for each genotype). Scale bars, base pairs) but not large RNAs (>200 base pairs). Two independent 200 mm. (D) Locomotor activity of DATCRE/+:Dicerflox/flox mice in the open field. experiments of three sets each were performed, with 10 visual fields per set; The total distance traveled was significantly decreased in DATCRE/+:Dicer flox/flox data represent mean ± SEM; ANOVA test, **P <0.01.(C)Immunostainingof mice (n = 4 for each genotype). Data represent mean ± SEM; Student’s t test, brain sections from 8-week-old DATCRE/+:Dicer flox/flox mice for TH demonstrates *P <0.05,**P <0.01.

Fig. 2. miR-133b is enriched in the midbrain and is deficient in the tissue of Parkinson’s disease patients. (A) Expression analysis of miR-133b in cerebral cortex (CX), midbrain (MB), or cerebellum (CB) of unaffected controls and Parkinson’s disease (PD) brain. RNA protection assays were performed for miR133a1 and miR133b expression, showing specific expression in control midbrain but not PD midbrain (n =3pergroup). Data represent mean ± SEM; ANOVA test, *P <0.05.(B)Expression analysis for miR-133b in cerebral cortex (CX), midbrain (MB), or cerebellum (CB) of control (WT) and Aphakia mutant mice. RNA protection assays were performed for miR133a1 and miR133b expression in control and Aphakia mutant mouse brain (three independent experiments). Data representmean± SEM;ANOVAtest,*P < 0.05. (C) qPCR analysis of murine ES cultures differentiated by the EB method and transduced with lentiviral vectors for pitx3,thetranscription factor nurr1 (as a negative control), both, or GFP vector control. Pitx3 transduction leads to the specific induction of miR-133b precursor expression; miR-133a1and miR-133a2 precursors are not induced by Pitx3 overexpression (three independent experiments were performed). Data represent mean ± SEM; ANOVA test,*P < 0.05.

www.sciencemag.org SCIENCE VOL 317 31 AUGUST 2007 1221 REPORTS

cultures to a lesser extent (by approximately consequence of long periods of immobility, Aphakia mice deficient in the transcription factor 50%), as were cells expressing TujI, an early which is reminiscent of the phenotype of human Pitx3 (9–11) and mice treated with the DN- general neuronal marker that first appears at the patients with Parkinson’s disease. These results specific toxin 6-hydroxydopamine (5). miR- neural precursor stage of EB differentiation. The suggest that miRNAs are essential for the 133b was specifically expressed in the midbrain Dicer deletion phenotype is partially rescued by terminal differentiation and/or maintenance of of normal mice, as in humans, and expression transfection of RNA species with low molecular multiple neuron types, including midbrain DNs. was markedly reduced in both rodent dopamine- weight (but not high molecular weight) derived Because no specific miRNAs had been deficiency models as demonstrated by RNase from embryonic mouse midbrain, consistent with implicated in midbrain DNs, we sought to protection assays, qPCR, and Northern blotting a model in which miRNAs play a role in identify some. We took a subtractive approach (Fig. 2B and fig. S4, B and C). midbrain DN terminal differentiation and surviv- and compared miRNA expression profiles of the The relative deficiency of miR-133b expres- al (Fig. 1B and fig. S1D). normal adult midbrain with the profiles of a sion in Aphakia mouse strain midbrain was To extend these findings to the intact rodent midbrain depleted of DNs. Expression analyses surprising, given that adult Aphakia mice do central nervous system, we generated mice that were performed by quantitative real-time reverse maintain a population of midbrain DNs within were homozygous for the conditional floxed transcription polymerase chain reaction (qPCR) the ventral tegmental area (10); this suggested the Dicer allele and expressed Cre recombinase for a panel of 224 miRNA precursors in mid- possibility that miR-133b is a direct target of under the regulation of dopamine transporter brain, cerebellum, and cerebral cortex samples Pitx3 transcription activation (as pitx3 is mutated regulatory sequences [DATCRE/+:Dicer flox/flox from Parkinson’s disease patients and normal in Aphakia mice). Consistent with this model, (6)], leading to the specific deletion of Dicer in controls (fig. S3 and table S1). Expression of overexpression of Pitx3 in differentiating ES postmitotic midbrain DNs (8). These mice one of these precursor miRNAs, miR-133b, cultures led to up-regulation of miR-133b display a progressive loss of midbrain DNs, as was specifically enriched in the midbrain and precursor expression (Fig. 2C). Furthermore, quantified by TH and dopamine transporter deficient in the context of Parkinson’s disease expression of a luciferase reporter vector that (DAT) immunostaining (Fig. 1C and fig. S1E), patient samples, as determined by ribonuclease harbors 350 base pairs of proximal miR-133b due to apoptosis (fig. S2A). (RNase) protection assays, qPCR, and North- promoter sequences was specifically induced by Behavioral studies of mice that harbor a ern blotting for mature miR-133b (Fig. 2A and overexpression of Pitx3 in COS cells (fig. S4D). midbrain DN-specific deletion of Dicer revealed fig. S4A). Next, we investigated the consequences of markedly reduced locomotion in an open-field We investigated expression of miR-133b in increased miR-133b expression in either ES cell– assay (Fig. 1D and fig. S2B). This is mostly a two additional DN deficiency models: adult derived cultures or in primary embryonic day

Fig. 3. miR-133b suppresses DN maturation and function. (A) Overexpression of miR-133b precur- sor in primary embryonic rat midbrain cultures led to decreased expression of the mature DN marker, DAT. A lentivirus vector was used to overexpress either miR-133b precursor or control (GFP) se- quences. Expression of TH showed a trend toward reduced expression that is not statistically signifi- cant, whereas Nurr1 and Pitx3 mRNA expression appeared unaffected. Data represent mean ± SEM; three independent experiments were performed; Student’s t test, *P <0.05.(B) Depolarization- induced dopamine release was quantified in ES- derived, EB differentiated cultures transduced with miR-133b precursor lentiviral vector or GFP con- trol. miR-133b precursor overexpression reduced dopamine release in murine ES culture-derived DNs. Data represent mean ± SEM; five independent experiments were performed; ANOVA test, *P < 0.05. (C) Overexpression of miR-133b during EB differentiation resulted in a significant decrease in the accumulation of TH-positive cells. Data repre- sent mean ± SEM; three independent experiments were performed; ANOVA test, *P < 0.05. (D) Reduction of miR-133b by penetratin-conjugated antisense miR133b 2′-O-methyl–modified oligo- nucleotide in primary embryonic rat midbrain culture leads to increased expression of DN mRNAs including TH and DAT, whereas Nurr1 and Pitx3 mRNAs are not significantly altered. Data repre- sent mean ± SEM; five independent experiments were performed; Student’s t test, *P < 0.05. (E) Depolarization-induced dopamine release was quantified in murine ES-derived, EB differentiated cultures transduced with miR-133b reduction– (or control-) modified oligonucleotide. miR-133b re- duction induced dopamine release in these cultures. Data represent mean ± SEM; three independent experiments were performed; Student’s t test, *P < 0.05.

1222 31 AUGUST 2007 VOL 317 SCIENCE www.sciencemag.org REPORTS

14.5 (E14.5) midbrain cultures. miR-133b pre- markers including DAT and TH (quantified by transcriptionally, and, in turn, Pitx3 activates cursor overexpression (using a lentiviral vector; qPCR analyses; Fig. 3, D and E). In ES-derived midbrain DN gene expression (5, 16) and induces fig.S3A)ledtoarelativereductionintranscrip- cultures, transduction of the miR-133b inhibitory transcription of miR-133b. Fluorescence-activated tion of the late midbrain DN maturation marker oligonucleotide potentiated potassium-stimulated cell sorter (FACS) analysis of permeablized pri- DAT, although transcription of early midbrain dopamine release. Taken together, these data im- mary rat midbrain cells with an antibody that DN markers, such as Pitx3 and the transcription plicate miR-133b in the regulation of midbrain recognizes Pitx3 protein revealed that miR-133b factor Nurr1, appeared unaltered or increased DN maturation and function. overexpression induced a reduction in Pitx3 (Fig. 3, A and B). Consistent with this, dopamine Individual miRNAs appear to regulate the protein levels in TH+ cells (Fig. 4A), whereas release in the context of potassium-induced de- expression of numerous targets posttranslation- miR-133b reduction led to an increase in Pitx3 polarization was markedly reduced with miR-133b ally (13). To identify potential physiological tar- protein in TH+ cells (Fig. 4B). overexpression. Overexpression of miR-133b at gets for miR-133b activity, we used available If Pitx3 is a direct target of miR-133b, one the neural precursor stage of EB-differentiated miRNA target prediction programs based on prediction is that miR-133b inhibition by modi- ES cultures led to a significant reduction in the 3′ untranslated sequence homology to miR- fied oligonucleotide transduction would fail to number of TH-positive cells (but not TujI-positive 133b (14, 15). The Pitx3 3′-untranslated region induce TH and DATexpression in Pitx3-deficient, cells; Fig. 3C). (3′ UTR) was identified as a potential target of Aphakia primary neuron cultures, and this was The activity of miR-133b can be inhibited miR-133b activity, and consistent with this, Pitx3 observed(Fig.4C).Finally,FACSanalysison using a 2′-O-methyl–modified RNA oligo- 3′ UTR sequences were subject to suppression by acutely dissociated midbrain DNs from young nucleotide homologous to the miR-133b miR-133b when placed downstream of a lucifer- Dicer mutant mice (10 days old, derived from sequence and linked to a short peptide derived ase reporter gene (fig. S6A). Thus, a hypothetical DAT CRE/+:Dicer flox/flox mice) revealed that Pitx3 from the Drosophila Antennapedia protein that model for the observed phenotypes associated protein expression is up-regulated in TH-positive mediates cell transduction (12) (fig. S5A). with altered miR-133b expression is that miR- neurons (relative to control DAT CRE/+:Dicer flox/+ Suppression of miR-133b in ES cell EB differ- 133b functions within a negative feedback circuit cells; Fig. 4D and fig. S6C), consistent with a role entiation cultures induced expression of DN that normally suppresses Pitx3 expression post- for miRNA in Pitx3 regulation.

Fig. 4. Pitx3 is a target of miR-133b activity. (A) Primary midbrain cultures (at day 1 in vitro) transduced with miR-133b precursor or control lentiviral vectors were analyzed (after 7 days) by FACS using antibodies for Pitx3 and TH. Expression of TH and Pitx3 protein is reduced in cells transduced with miR-133b precursor relative to control vector (GFP) or miR-18 precursor. Data represent mean ± SEM; three independent experiments were performed; ANOVA test, *P <0.05.(B) Reduction of miR- 133b in primary midbrain cultures (at day 7 in vitro) with the use of 2′-O- methyl–modified oligonucleotide but not control oligonucleotide leads to a significant induction in Pitx3 and TH protein as quantified by FACS analysis (after 7 days). Data represent mean ± SEM; three independent experiments were performed; Student’s t test, *P < 0.05. (C) miR-133b inhibition by 2′- O-methyl–modified oligonucleotide in Pitx3-deficient Aphakia primary neuron cultures fails to induce TH or DAT transcription. Data represent mean ± to control DAT CRE/+:Dicer flox/+ mice. FACS analyses were performed on SEM; three independent experiments were performed; Student’s t test, *P < acutely dissociated, permeablized midbrain cells with the use of TH- and 0.05. (D) Pitx3 protein expression was significantly increased in TH-positive Pitx3-specific antibodies. Data represent mean ± SEM; three independent cells from 10-day-old miRNA-deficient DAT CRE/+:Dicer flox/flox mice relative experiments were performed; Student’s t test, *P < 0.05.

www.sciencemag.org SCIENCE VOL 317 31 AUGUST 2007 1223 REPORTS

Our data support a model in which miR-133b 3. A. Wallen, T. Perlmann, Ann. N. Y. Acad. Sci. 991,48 13. D. P. Bartel, Cell 116, 281 (2004). functions within a feedback loop, as Pitx3 spe- (2003). 14. B. John et al., PLoS Biol. 2, e363 (2004). 4. J. H. Kim et al., Nature 418, 50 (2002). 15. B. P. Lewis, I. H. Shih, M. W. Jones-Rhoades, D. P. Bartel, cifically induces transcription of miR-133b, and 5. C. Martinat et al., Proc. Natl. Acad. Sci. U.S.A. 103, 2874 C. B. Burge, Cell 115, 787 (2003). Pitx3 activity is down-regulated by miR-133b (2006). 16. M. P. Smidt, S. M. Smits, J. P. Burbach, Cell Tissue Res. posttranscriptionally (fig. S6D). Midbrain DN 6. E. P. Murchison, J. F. Partridge, O. H. Tam, S. Cheloufi, 318, 35 (2004). function is dynamic, and such feedback circuitry G. J. Hannon, Proc. Natl. Acad. Sci. U.S.A. 102, 12135 17. A. Becskei, L. Serrano, Nature 405, 590 (2000). has been shown to increase the robustness and (2005). 18. We thank O. Hobert and R. Liem for comments and 7. Materials and methods are available as supporting F. Beal for technical assistance. This work was funded by speed response time and stability in the context of material on Science Online. NINDS and Spitzer Funds. dynamic changes (17). Furthermore, we present 8. N. Chuhma et al., J. Neurosci. 24, 972 (2004). 9. D. Y. Hwang, P. Ardayfio, U. J. Kang, E. V. Semina, evidence that Dicer deletion leads to the progres- Supporting Online Material sive loss of midbrain DNs, suggesting that K. S. Kim, Brain Res. Mol. Brain Res. 114,123 (2003). www.sciencemag.org/cgi/content/full/317/5842/1220/DC1 miRNAs in addition to miR-133b function in 10. I. Nunes, L. T. Tovmasian, R. M. Silva, R. E. Burke, Materials and Methods these cells. S. P. Goff, Proc. Natl. Acad. Sci. U.S.A. 100, 4245 Figs. S1 to S6 (2003). Tables S1 to S3 References and Notes 11. P. van den Munckhof et al., Development 130, 2535 References 1. L. He, G. J. Hannon, Nat. Rev. Genet. 5, 522 (2004). (2003). 29 January 2007; accepted 26 July 2007 2. V. Ambros, Cell 113, 673 (2003). 12. T. J. Davidson et al., J. Neurosci. 24, 10040 (2004). 10.1126/science.1140481

GPR1, BOI1, FLO8, NCE102, MSN1,orGIC1, Cap-Independent Translation Is resulted in efficient translation of ApppG- capped hairpin mRNAs compared with length- Required for Starvation-Induced matched negative control constructs containing reverse-complement 5′UTR sequences. In addi- tion, the 5′UTRs from TPK2, HMS2, and Differentiation in Yeast YEL033w lacked IRES activity (Fig. 1C). The invasive growth cellular IRESs’ activities ranged Wendy V. Gilbert, Kaihong Zhou, Tamira K. Butler, Jennifer A. Doudna* from 8 to 33% of that of the control m7GpppG- Cellular internal ribosome entry sites (IRESs) are untranslated segments of mRNA transcripts capped mRNA (table S2). We tested the IRES- thought to initiate protein synthesis in response to environmental stresses that prevent canonical 5′ containing 5′UTRsforactivityinvivobyelec- cap–dependent translation. Although numerous cellular mRNAs are proposed to have IRESs, none troporating the reporter mRNAs into yeast cells has a demonstrated physiological function or molecular mechanism. Here we show that seven yeast to avoid any possibility of mistaking cryptic required for invasive growth, a developmental pathway induced by nutrient limitation, promoter or splicing activity for IRES activity. contain potent IRESs that require the initiation factor eIF4G for cap-independent translation. In All seven 5′UTRs promoted efficient cap- contrast to the RNA structure-based activity of viral IRESs, we show that an unstructured A-rich independent translation in vivo (Fig. 1D). Ex- element mediates internal initiation via recruitment of the poly(A) binding protein (Pab1) to the 5′ periments with bicistronic reporters, in which the untranslated region (UTR) of invasive growth messages. A 5′UTR mutation that impairs IRES activity 5′ ORF is translated via cap-dependent initiation compromises invasive growth, which indicates that cap-independent translation is required for and the 3′ ORF is efficiently translated only when physiological adaptation to stress. an active IRES is inserted between the two ORFs, corroborated our finding that the 5′UTRs ranslation initiation is a crucial point of downstream open reading frame (ORF) of a of seven invasive growth genes mediate internal regulation of eukaryotic gene expression, naturally occurring bicistronic cellular message translation initiation (fig. S1). Tallowing cells to adapt rapidly to changing (6), suggesting that invasive growth genes might A subset of viral IRESs recruits the translation environmental conditions. In response to glucose be translated by a mechanism that depends on machinery by providing high-affinity internal deprivation, haploid Saccharomyces cerevisiae an internal ribosome entry site (IRES). To test binding sites for the translation initiation factor cells dramatically down-regulate translation of whether the 5′UTRs of invasive growth genes are eIF4G (9). To test whether invasive growth IRESs most cellular messages, while also exhibiting strik- capable of internal translation initiation, we require eIF4G, we prepared extracts genetically ing morphological changes leading to invasive inserted these sequences into a firefly luciferase depleted of eIF4G (8). Reducing eIF4G levels to growth (1, 2). Whereas the global translational reporter (F-luc) containing a stable stem-loop 37% decreased translation from invasive growth repression requires the mRNA 5′ decapping structure [change in Gibbs free energy (DG°) = IRES reporters to ~25% activity (Fig. 2A), machinery (3), the developmental switch requires –58 kcal/mol] at the 5′ end to inhibit scanning (7) whereas translation from the eIF4G-independent new protein synthesis, which suggests that pro- and capped with a nonphysiological ApppG cap cricket paralysis virus (CrPV) IRES was only teins required for invasive growth might be trans- that reduces binding of the cap-binding initiation slightly affected. Further depletion abolished lated by a cap-independent mechanism. factor (eIF4E) by three orders of magnitude (7). activity (fig. S2). In cell extracts containing 9- Many invasive growth genes have unusually With no IRES inserted, the ApppG-capped hair- fold overexpressed eIF4G, invasive growth IRES long 5′ untranslated regions (5′UTRs) with the pin RNA is poorly translated, yielding 0.4% in activity increased 10- to 20-fold (Fig. 2B), which potential to form stable RNA secondary struc- vitro and 0.04% in vivo compared with an indicated that eIF4G is limiting for IRES activity. tures (table S1) (4). Furthermore, one gene re- m7GpppG-capped mRNA (at 100%), with an eIF4G is the least abundant initiation factor in quired for invasive growth, YMR181c (5), is the unstructured 18-nucleotide (nt) 5′UTR (Fig. 1A) yeast (10) and becomes unstable in nutrient- (8). We confirmed by Northern blots that the limited cells (11), which suggests a need for reporter mRNAs were stable in both extracts and continued synthesis of eIF4G protein in glucose- Department of Molecular and Cell Biology, Department of Chemistry, and Howard Hughes Medical Institute, Uni- cells (Fig. 1B), which ruled out differential RNA starved cells in order to maintain invasive versity of California, Berkeley, CA 94720, USA. stability as a source of differences in luciferase growth IRES activity. We therefore tested wheth- *To whom correspondence should be addressed. E-mail: activity. Insertion of the 5′UTR sequences from er the 5′UTRs from either yeast eIF4G gene [email protected] YMR181c or from the invasive growth genes are themselves capable of internal initiation

1224 31 AUGUST 2007 VOL 317 SCIENCE www.sciencemag.org