DOI: 10.1126/science.1208049, 1669 (2011);334 Science

et al.Kartik RamamoorthiContextual Memory FormationNpas4 Regulates a Transcriptional Program in CA3 Required for

This copy is for your personal, non-commercial use only.

clicking here.colleagues, clients, or customers by , you can order high-quality copies for yourIf you wish to distribute this article to others

  here.following the guidelines

can be obtained byPermission to republish or repurpose articles or portions of articles

  ): August 19, 2013 www.sciencemag.org (this information is current as of

The following resources related to this article are available online at

http://www.sciencemag.org/content/334/6063/1669.full.htmlversion of this article at:

including high-resolution figures, can be found in the onlineUpdated information and services,

http://www.sciencemag.org/content/suppl/2011/12/21/334.6063.1669.DC1.html can be found at: Supporting Online Material

http://www.sciencemag.org/content/334/6063/1669.full.html#ref-list-1, 15 of which can be accessed free:cites 53 articlesThis article

http://www.sciencemag.org/content/334/6063/1669.full.html#related-urls7 articles hosted by HighWire Press; see:cited by This article has been

http://www.sciencemag.org/cgi/collection/neuroscienceNeuroscience

subject collections:This article appears in the following

registered trademark of AAAS. is aScience2011 by the American Association for the Advancement of Science; all rights reserved. The title

CopyrightAmerican Association for the Advancement of Science, 1200 New York Avenue NW, Washington, DC 20005. (print ISSN 0036-8075; online ISSN 1095-9203) is published weekly, except the last week in December, by theScience

on

Aug

ust 1

9, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

concentrated on the plane of the Galaxy. Theemission is weak enough to be compatible withexisting upper limits. This result provides an es-timate for the Lya escape rate from the MilkyWay for an outside observer. Enhancements arealso detected outside the galactic plane, poten-tially explained by the presence of nearby H Iclouds outside the galactic plane that scatter theLya radiation formed in H II regions within thegalactic plane.

Of wide use for distant galaxies is the escapefraction, defined as the ratio of escaping to locallyproduced Lya photons. For high-redshift gal-axies, it is found to be a few percent (14). Figure5 shows that, as seen from the Sun, the escapefraction from the brightest H II regions is on theorder of 3%, if we assume the classical Lya/Haproduction ratio of 1.61 and neglect the Ha ab-sorption between the source region and the Sun.This percentage, however, increases stronglywhereHa becomes weaker. When integrating over largefractions of the sky, such areas become substan-tial contributors to the global escape fraction, inthe same way that the H I halo of nearby galaxiesis found from recent observations to contribute tothe Lya emission (41).

Observations of distant galaxies also refer tothe equivalent width WLya, which measures thefraction of light in the form of Lya photons withrespect to the adjacent UV continuum (40).WLya

is positive when Lya is seen in emission andnegative when it is seen in absorption. Usually itis integrated over a whole galaxy. However, re-cent spatially resolved observations (41) showthat WLya is globally increasing from negativevalues for central parts to highly positive valuesfrom surrounding haloes. Here, we detect the var-iability ofWLya at a much smaller scale. Negativevalues on the order of 20 to 40 Å are found in thedirections that correspond to bright starlight UVcontinua. They are negative because, after sub-traction of the heliospheric line, the galactic emis-sion line is not large enough to compensate forthe stellar-interstellar absorption trough in the UVcontinuum.Where the continuumbecomes small-er, WLya becomes positive (the line is seen inemission) and increases rapidly. By definition, itcannot be defined (or is infinite) where the con-tinuum becomes zero.

In the case of our Galaxy, a huge amount ofinformation is available on stars, gas, and dust.Milky Way Lya data can be used to test thecomplex radiative transfer models that are de-veloped for distant galaxies or studies of thecoupling between H I 21 cm and Lya radiationfields in very low-density media (42, 43).

References and Notes1. M. Brocklehurst, Mon. Not. R. Acad. Sci. 153,

471 (1971).2. S. A. Wouthuysen, Astron. J. 57, 31 (1952).3. G. B. Field, Astrophys. J. 129, 551 (1959).4. D. A. Neufeld, Astrophys. J. 350, 216 (1990).5. S. Charlot, S. M. Fall, Astrophys. J. 415, 580

(1993).6. S. Richling, Astrophys. Space Sci. 284, 361 (2003).

7. M. Hansen, S. P. Oh, Mon. Not. R. Acad. Sci. 367,979 (2006).

8. A. Verhamme, D. Schaerer, A. Maselli, Astron. Astrophys.460, 397 (2006).

9. J. Lequeux et al., Astron. Astrophys. 301, 18 (1995).10. C. Gronwall et al., Astrophys. J. 667, 79 (2007).11. J.-M. Deharveng et al., Astrophys. J. 680, 1072 (2008).12. G. Östlin et al., Astron. J. 138, 923 (2009).13. K. A. Kornei et al., Astrophys. J. 711, 693 (2010).14. M. Hayes et al., Nature 464, 562 (2010).15. J. L. Bertaux, J. E. Blamont, Astron. Astrophys. 11,

200 (1971).16. G. E. Thomas, R. F. Krassa, Astron. Astrophys. 11,

218 (1971).17. 1 Rayleigh (R) = 106/4 p photons cm−2 s −1 sr−1.18. G. E. Thomas, J. E. Blamont, Astron. Astrophys. 51,

283 (1976).19. B. T. Draine, E. E. Salpeter, Nature 271, 730 (1978).20. W. D. Vacca, C. D. Garmany, J. M. Shull, Astrophys. J.

460, 914 (1996).21. J. L. Bertaux et al., Astron. Astrophys. 46, 19 (1976).22. E. Quémerais et al., Astron. Astrophys. 308, 279

(1996).23. E. Quémerais et al., J. Geophys. Res. 108 (A10), 8029

(2003).24. R. Lallement et al., Astron. Astrophys. 140, 243 (1984).25. J. T. Clarke et al., Astrophys. J. 499, 482 (1998).26. A. L. Broadfoot et al., Space Sci. Rev. 21, 183 (1977).27. E. Quémerais et al., Astrophys. J. 711, 1257 (2010).28. The field of view is a 0.1° by 0.87° rectangle. The

signal was integrated over the 9 UVS spectral channelsthat define the Lya emission. Lines of sight close tonearby hot stars were avoided. The photon Poissonnoise is always smaller than 2% because of the longexposure time. For a given heliospheric hydrogendistribution, the local glow brightness is proportionalto the total solar output, which varies with time. Themean brightness level of the H glow is thus slowlyvarying with the solar cycle phase. Small variationsmay also exist within a scan due to anisotropicillumination coupled to solar rotation, but they arelimited by the multiple scattering effects and shouldnot exceed a few percent. Further details on theinstrument and the data are given in (22) andin the SOM.

29. V. B. Baranov, Y. G. Malama, J. Geophys. Res. 98 (A9),15157 (1993).

30. V. V. Izmodenov et al., J. Geophys. Res. 104 (A3),4731 (1999).

31. N. V. Pogorelov, S. N. Borovikov, G. P. Zank, T. Ogino,Astrophys. J. 696, 1478 (2009).

32. J. L. Linsky, B. E. Wood, Astrophys. J. 463, 254(1996).

33. E. Quémerais et al., Astron. Astrophys. 299, 249(1995).

34. D. P. Finkbeiner, Astrophys. J. 146, 407 (2003).35. B. Savage, R. Panek, Astrophys. J. 191, 659

(1974).36. R. H. Buss et al., Astrophys. J. 454, L55 (1995).37. The stellar continuum count level must be lower

than approximately 10% of the central Lyachannel count level.

38. E. Quémerais, V. Izmodenov, D. Koutroumpa, Y. Malama,Astron. Astrophys. 488, 351 (2008).

39. Cross calibrations have been performed with a numberof other instruments. Our conservative estimatederror on the calibration factor is 30%.

40. The equivalent width of an emission line is definedas the wavelength interval over which the continuumaround the line must be integrated to produce thesame intensity as the observed line. It correspondsto the surface of the line when the spectrum isnormalized in such a way that the continuum level is1 and has no dimension. Note that in Fig. 2, the linecontains both heliospheric and galactic contributions;thus, it can not be used to obtain a visualrepresentation of the galactic line equivalent width.

41. C. C. Steidel et al., arXiv:1101.2204 (2011).42. H. Liszt, Astron. Astrophys. 371, 698 (2001).43. S. Baek, P. Di Matteo, B. Semelin, F. Combes, Y. Revaz,

Astron. Astrophys. 495, 389 (2009).

Acknowledgments: We thank J. Holberg for kindly providingVoyager spectra of early-type stars.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/science.1197340/DC1SOM TextFigs. S1 to S11References

3 September 2010; accepted 16 November 2011Published online 1 December 2011;10.1126/science.1197340

Npas4 Regulates a TranscriptionalProgram in CA3 Required forContextual Memory FormationKartik Ramamoorthi,1,2,3 Robin Fropf,1,2 Gabriel M. Belfort,1,2 Helen L. Fitzmaurice,1,2

Ross M. McKinney,1,2 Rachael L. Neve,2 Tim Otto,4 Yingxi Lin1,2*

The rapid encoding of contextual memory requires the CA3 region of the hippocampus, but thenecessary genetic pathways remain unclear. We found that the activity-dependent transcriptionfactor Npas4 regulates a transcriptional program in CA3 that is required for contextual memoryformation. Npas4 was specifically expressed in CA3 after contextual learning. Global knockout orselective deletion of Npas4 in CA3 both resulted in impaired contextual memory, and restoration ofNpas4 in CA3 was sufficient to reverse the deficit in global knockout mice. By recruiting RNApolymerase II to promoters and enhancers of target genes, Npas4 regulates a learning-specifictranscriptional program in CA3 that includes many well-known activity-regulated genes, whichsuggests that Npas4 is a master regulator of activity-regulated gene programs and is central tomemory formation.

The ability to form a long-term memoryafter a single experience is essential for thesurvival of higher organisms. In rodents

and humans, memory of places or contexts canbe formed after a single brief exposure to anovel environment, and this process requires

www.sciencemag.org SCIENCE VOL 334 23 DECEMBER 2011 1669

RESEARCH ARTICLES

on

Aug

ust 1

9, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

the hippocampus (1, 2). It has been suggested thathippocampal areaCA3 is required for rapid encodingof contextual memory (3–6). However, CA3-specific molecular pathways underlying contex-tual memory formation remain uncharacterized.

The formation and maintenance of long-termmemories requires de novo mRNA and proteinsynthesis (7, 8). Learning-induced expression ofactivity-regulated genes, especially immediateearly genes (IEGs), provides a link between be-havioral experience and the molecular eventsrequired to encode memory (9, 10). Genetic per-turbations of IEGs or transcription factors thatcontrol activity-regulated gene expression thusoften lead to deficits in neuronal plasticity andmemory (11–15). However, most IEGs can beinduced by a wide range of stimuli and are in-volved in processes essential to normal cellularfunction and survival (16, 17), which suggests thattheir function may not be specific to learning-related neuronal activity. Therefore, identifyingIEGs whose function is selectively correlated withboth synaptic activity and learning may help toreveal the genetic programs required for memoryencoding.

The expression of the activity-dependenttranscription factor Npas4 (neuronal PAS domainprotein 4) was previously shown to be selectivelycoupled to neuronal activity (18). We therefore in-vestigated whether it regulates a learning-specifictranscriptional program underlying the formationof contextual memories.

Npas4 expression is selectively induced byneuronal activity and contextual learning. Wefirst characterized the induction of Npas4, to-gether with several other IEGs, in culturedmousehippocampal neurons. Membrane depolarizationresulted in robust protein synthesis–independentexpression ofNpas4mRNA, which suggests thatNpas4 is an IEG (Fig. 1A).Npas4was selective-ly induced by depolarization and Ca2+ influx,but not by activators of several other signalingpathways that induce IEGs, such as c-Fos, Arc(activity-regulated cytoskeleton-associated pro-tein), and Zif268 (Fig. 1, A and B), similar toprevious observations in dissociated rat neu-rons (18).

To examine experience-induced expressionof Npas4, we trained mice in a hippocampus-dependent contextual fear conditioning (CFC)paradigm, which is thought to be dependent onde novo mRNA and protein (8), and examinedNpas4 mRNA expression in dorsal hippocam-pus (DH). We focused on DH on the basis of

1McGovern Institute for Brain Research, Massachusetts Instituteof Technology (MIT), 77 Massachusetts Avenue, Cambridge, MA02139, USA. 2Department of Brain and Cognitive Sciences,Massachusetts Institute of Technology, 77 Massachusetts Avenue,Cambridge, MA 02139, USA. 3Molecular and Cellular Neuro-science Graduate Program, MIT, 77 Massachusetts Avenue,Cambridge, MA 02139, USA. 4Program in Behavioral Neuro-science, Department of Psychology, Rutgers University, NewBrunswick, NJ 08854, USA.

*To whom correspondence should be addressed. E-mail:yingxi@mit.edu

Fig. 1. Npas4 expression is selectively induced by neuronal activity in vitro and by learning in vivo. (Aand B) Npas4 mRNA (A) and protein (B) in cultured hippocampal neurons is selectively induced bymembrane depolarization and dependent on Ca2+ influx. Induction of Npas4mRNA does not require newprotein synthesis. n = 4 cultures. (C) Npas4 is rapidly induced after CFC. Separate groups of mice werekilled 5 min (n = 8), 30 min (n = 9 to 11), 1 hour (n = 6), or 4.5 hours (n = 5) after CFC and compared tonaive home-cage mice (n = 10). Values are plotted relative to peak time point. (D) Experimental schemeand behavioral outcomes of CFC and control conditions. C+S, context + shock; C, context exposure; S,immediate shock; HC, home cage. (E) Npas4mRNA is selectively induced by context learning. C+S, n = 8to 10; C, n = 8; S, n = 8; and HC, n = 10. All groups were sacrificed 30 min after training. *P < 0.001.

Fig. 2. Npas4 global knockout miceexhibit impaired hippocampal-dependent STM and LTM. (A and B)Npas4−/− and Npas4+/+ littermatesexhibit similar freezing during thetraining session (A) and 5 min aftertraining (B). P = 0.879. (C and D)1 hour (C) and 24 hours (D) afterCFC, Npas4−/− mice freeze at a significantly lower level than Npas4+/+ littermates. *P ≤ 0.001. (E) 24hours after auditory delay conditioning, Npas4−/− and Npas4+/+ mice exhibit similar freezing during atone memory test. P = 0.859.

23 DECEMBER 2011 VOL 334 SCIENCE www.sciencemag.org1670

RESEARCH ARTICLES

on

Aug

ust 1

9, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

extensive work showing that DH is required forCFC (19).

Mice were killed at various time points af-ter CFC to measure expression of Npas4, c-Fos,and Arc mRNA using quantitative polymerasechain reaction (qPCR) (Fig. 1C). Npas4 mRNAreached its peak level 5 min after training andreturned to baseline levels 4.5 hours later. c-Fosand Arc reached their peak levels by 30 min andreturned to baseline levels 4.5 hours after training(Fig. 1C).

Next, we trained mice under CFC conditionsthat provided both context learning and shockassociation (C+S), or under conditions that involved

just context learning (C) or shock (S) alone (Fig.1D). Both C+S and C represent learning condi-tions, because the hippocampus forms contextualrepresentations independent of shock delivery(20, 21), but only C+S provides a behavioral read-out of learning (Fig. 1D). Immediate S fails toinduce long-term contextual memories, as the con-text exposure is not long enough for the hippo-campus to form a representation (Fig. 1D) (20, 22).Therefore, this served as a control condition, allow-ing us to distinguish IEG induction specific tocontext learning from induction due to the shock.

Gene expression analysis in mice sacrificed30 min after training indicated that, compared

with naive subjects, Npas4 was induced in theC+S and C groups, but not in the S group. Incontrast, c-Fos and Arc were significantly in-duced in all behavioral conditions (Fig. 1E).

Learning and memory deficits in Npas4global knockout mice. We next determinedwhether CFC was impaired in Npas4 knockout(Npas4−/−) mice. During the training session andthe memory test 5 min later, we observed robustfreezing behavior in both Npas4−/− and wild-type (Npas4+/+) littermates, suggesting that theability to acquire CFC is normal in Npas4−/−

mice (Fig. 2, A and B). Furthermore, locomotoractivity, anxiety levels, footshock sensitivity, andhippocampal morphology were similar acrossgenotypes (figs. S1 and S2). However, despitehaving intact memories 5 min after training,freezing elicited by the context 1 hour and 24 hoursafter trainingwas significantly reduced inNpas4−/−

mice (Fig. 2, C and D), which suggests that bothshort-term memory (STM) and long-term mem-ory (LTM) are impaired.

There is now a general consensus that theamygdala is required for all forms of fear condi-tioning, whereas only a subset of fear-conditioningparadigms (including CFC) rely on hippocampalintegrity (23, 24). We therefore investigatedwhetherNpas4−/−mice were deficient in auditorydelay conditioning, a form of fear conditioningknown to depend on the amygdala but not thehippocampus (24).We saw no difference betweenNpas4−/− and wild-type mice when tone-inducedfreezing was measured 24 hours after training,confirming that sensory detection and fear mem-ory acquisition are normal in Npas4−/− mice andsuggesting that the impairment we observedin CFC was likely due to a deficit in the hippo-campus, not the amygdala (Fig. 2E).

Selective deletion of Npas4 from CA3, butnot CA1, impairs long-term contextual memory.We hypothesized that the memory impairmentobserved in the global knockout was due to aloss of learning-induced Npas4 expression in DH,based on its selective expression after contextlearning (Fig. 1E). Because the different sub-regions within DH may play dissociable roles incontextual memory formation (3), we examinedwhether CFC resulted in a regionally selectiveexpression of Npas4. Although Npas4 was ex-pressed broadly in several brain regions afterCFC, including amygdala and entorhinal cortex(fig. S3), within the hippocampus Npas4 expres-sion after CFC was largely restricted to the CA3subregion (Fig. 3A), with higher expression indorsal CA3 than in ventral CA3 (fig. S4). In con-trast, c-Fos was robustly expressed in both CA1and CA3 (Fig. 3A), and similar patterns of in-duction have been reported for Arc and Zif268(25, 26). We also noted that the highest level ofNpas4 was observed 30 min after CFC, 1 hourbefore the peak expression of c-Fos (Fig. 3B).This observation suggests that pathways ac-tivating Npas4 may be distinct from those forother IEGs. Importantly, localized induction ofNpas4 inCA3 appears to be specific to contextual

Fig. 3. Npas4 expression in CA3 is required for contextual fear conditioning. (A) (Top) Npas4 expressionis increased only in CA3 and to a lesser extent in dentate gyrus after CFC. (Bottom) c-Fos expression isinduced in all subregions of DH after CFC. (Right) Seizure induces Npas4 and c-Fos in all subregions ofhippocampus. (B) Western blot analysis of Npas4 and c-Fos expression in DH at various times after CFC.n = 5 mice per condition. Values are plotted relative to peak time point. *P < 0.04. (C) Selectivedeletion of Npas4 in CA3 or CA1 3 days after injecting HSV-Cre. (D) Selective deletion of Npas4 in CA3impairs long-term CFC memory formation. *P < 0.001.

www.sciencemag.org SCIENCE VOL 334 23 DECEMBER 2011 1671

RESEARCH ARTICLES

on

Aug

ust 1

9, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

learning, because Npas4 was induced in all re-gions of hippocampus after kainic acid–stimulatedseizures (Fig. 3A). Although the CA3 region isknown to be required for rapid contextual learn-ing, these data are the first to show the selectiveinduction of a specific IEG within CA3 by thisform of learning.

If induction of Npas4 in CA3 is required forcontextual memory, then deleting Npas4 in CA3should replicate the memory impairments seen inthe global knockout (Fig. 2). We acutely deletedNpas4 by stereotaxically injecting a herpes sim-plex virus (HSV) expressing Cre recombinase(HSV-Cre) into the CA3 region of Npas4 con-ditional knockout (Npas4flx/flx) mice (Fig. 3C).HSV is naturally neurotropic and reaches peakexpression within 3 days of delivery (27, 28). Inanother group of mice, we used an equivalentamount of virus to delete Npas4 from a similarvolume of cells in dorsal CA1, where we see noactivation of Npas4 after CFC (Fig. 3C). To con-trol for any effects of expressing Cre recombinase,we injectedHSV-Cre into CA3 ofwild-typemice.Mice were injected with HSV-Cre 3 days beforeCFC and tested 1 hour and 24 hours after training(Fig. 3D). All animals showed similar freezingduring the 1-hour context test. However, 24 hoursafter training, animals with Npas4 deletions inCA3 had attenuated freezing responses com-pared to animals with Npas4 deletions in CA1or wild-type animals injected with HSV-Cre inCA3 (Fig. 3D).

Npas4 regulates an activity-dependent ge-netic program that includes several IEGs. As anactivity-dependent transcription factor, Npas4likely regulates a genetic program that is re-quired for CA3-dependent encoding of con-

textual memory. Npas4 expression peaks beforethat of several other IEGs (Figs. 1C and 3B),and its acute deletion abolished expression ofc-Fos (Fig. 4A); together, these data suggestthat Npas4 may regulate the activity-dependentexpression of other IEGs. To explore this pos-sibility, we acutely deleted Npas4 in a high

percentage of cultured Npas4flx/flx hippocampalneurons by infecting them with HSV-Cre andassayed the mRNA expression of several IEGsafter membrane depolarization. Compared withuninfected and HSV-GFP (green fluorescentprotein)–infected controls, deletion of Npas4 abol-ished depolarization-induced expression of Arc,

Fig. 4. Npas4 regulates theexpression of several IEGs. (A)Conditional deletion of Npas4in CA3 results in loss of c-Fosexpression. (B) Dramatic lossof depolarization-induced IEGexpression by acute deletion ofNpas4 in cultured mouse neu-rons. n= 3 cultures. *P< 0.001.(C) The activity of PIBDNF reporteris abolished in the absence ofNpas4, whereas PNpas4, CRE, andMRE reporters show similar in-duction in the presence or ab-sence of Npas4. n = 4 cultures.*P < 0.001. (D) ChIP experi-ments showing that under depo-larized conditions Npas4 bindsto PIBDNF and enhancer II of c-Fos(c-Fos E2). No binding is ob-served at the c-Fospromoter, theb-actin promoter, or a negativecontrol region.

Fig. 5. Npas4 is requiredfor the recruitment of RNApolymerase II to enhancerand promoter regions ofactivity-regulated genes.(A) Localization of Pol IIto PIBDNF and c-Fos en-hancer region E2 is depen-dent on Npas4. No changeis observed in Pol II bind-ing at the c-Fos or b-actinpromoter. (B) Depolarization-induced activity of PIBDNFreporter is abolished inthe absence of Npas4, butthe c-Fos promoter re-porter is unaffected. n =4 cultures. *P<0.001. (C)qPCR analysis of ChIP sam-ples from seized Npas4−/−

and Npas4+/+ littermates.Pol II binding to PIBDNF(*P ≤ 0.008, n = 7 pergenotype) and c-Fos E2(*P ≤ 0.048, n = 6 pergenotype) is diminishedin Npas4−/−mice relativeto Npas4+/+ littermates.No change is observed inPol II binding at the c-Fos promoter (P = 0.333, n = 6 per genotype).

23 DECEMBER 2011 VOL 334 SCIENCE www.sciencemag.org1672

RESEARCH ARTICLES

on

Aug

ust 1

9, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

c-Fos, and Zif268 mRNA (Fig. 4B). Expressionof the housekeeping geneGAPDH (glyceraldehyde3-phosphate dehydrogenase) was not altered.

Deletion of Npas4 could affect expressionof activity-regulated IEGs indirectly, for exam-ple, by generally disrupting the cellular responseto neuronal activity. To examine this possibility,we designed a series of luciferase reporter assaysto determine whether other activity-dependenttranscriptional pathways function normally in theabsence of Npas4. We first characterized tran-scription from the promoter ofNpas4 (PNpas4-Luc)to look at pathways upstream of Npas4. Thisreporter was induced in response to KCl de-polarization but not to activators of other sig-naling pathways, similar to endogenous Npas4(fig. S5; compare Fig. 1, A and B). When Npas4was acutely deleted by the expression of Crerecombinase in cultured hippocampal neuronsgenerated from Npas4flx/flx mice, activity ofPNpas4-Luc in response to KCl depolarizationwas unchanged (Fig. 4C). We also examinedthe activity of the transcription factors CREB(cAMP responsive element–binding protein) andMEF2 (myocyte enhancer factor 2). UnlikeNpas4,

these proteins are constitutively expressed andare activated by posttranslational modificationsin response to depolarization (29, 30). Reportersexpressing luciferase under the control of CREBand MEF2 response elements (CRE and MRE)were unaffected by acute deletion of Npas4(Fig. 4C).

We then directly determined whether Npas4binds to the genomicDNAof two activity-regulatedgenes, BDNF (brain-derived neurotrophic factor)and c-fos, using chromatin immunoprecipitation(ChIP). These genes are dependent on Npas4 fortheir expression in response to neuronal activ-ity (Fig. 4, A to C) (18), have well characterizedgenomic structures (31–35), and have been im-plicated in learning and memory (14, 36, 37).We examined one of the activity-regulated pro-moters of BDNF, promoter I (PIBDNF), the prox-imal promoter region of c-Fos and one of itsupstream enhancer regions, E2 (38). After de-polarization, Npas4 bound to PIBDNF and c-FosE2but not to the c-Fos proximal promoter (Fig. 4D).

Npas4 is required for recruitment of RNApolymerase II to regulatory regions of its targetgenes. Genome-wide ChIP sequencing has re-

vealed that Npas4 colocalizes with RNA poly-merase II (Pol II) at enhancer and promoter sitesof many activity-regulated genes, including BDNFand c-fos (38). However, it is not known whetherthis colocalization plays an important role inregulating transcription of these genes. We hy-pothesized that Npas4 is required for activity-dependent recruitment of Pol II to promoterand enhancer regions of its targets, in order toactivate their transcription.

We acutely deleted Npas4 using HSV-Cre ina high percentage of cultured Npas4flx/flx corticalneurons and then performed ChIP for Pol II after2 hours of membrane depolarization. In controlneurons infected with HSV-GFP, Pol II localizedto PIBDNF, the c-Fos enhancer E2, the c-Fos pro-moter region, and the b-actin promoter after de-polarization (Fig. 5A).WhenNpas4was deletedby HSV-Cre, localization of Pol II to PIBDNFand c-Fos E2 was impaired (Fig. 5A). As we de-scribed above, Npas4 binds to both of these re-gions. Pol II binding to the promoter regions ofc-Fos and b-actin, where we did not observeNpas4 binding (Fig. 4D), was not affected bydeletion of Npas4. To confirm that the Npas4-dependent binding of Pol II is important for geneexpression, we compared luciferase reporters driv-en by PIBDNF and the c-Fos promoter and foundthat expression from PIBDNF was abolished bydeletion ofNpas4, whereas expression from thec-Fos promoter was not attenuated (Fig. 5B).

To confirm our findings in vivo, we per-formed ChIP for Pol II from hippocampal tissueextracted from adult Npas4+/+ and Npas4−/−

littermates. Npas4 is expressed only in a sparsepopulation of neurons after CFC (Fig. 3A),making it difficult to detect Pol II binding inthese cells. We therefore used kainic acid–inducedseizures to activate all neurons to determine thegenomic localization of Pol II in vivo. Seizure hasbeen shown to robustly induce activity-regulatedgenes, many of which have been implicated inmemory formation, and under certain conditionscan induce potentiation similar to long-term po-tentiation (39). In line with our in vitro observa-tions, localization of Pol II to PIBDNF and c-FosE2 was impaired in Npas4−/− mice comparedwith Npas4+/+ littermates, whereas Pol II bindingto the promoter regions of c-Fos and b-actin wassimilar across genotypes (Fig. 5C and fig. S6).

Expression of Npas4 in CA3 rescues tran-scription and memory formation in global knock-outs. We next investigated whether reexpressingNpas4 in CA3 of Npas4−/− mice leads to expres-sion of its genetic program and consequentlyrescues memory formation. The CA3 region ofNpas4−/−micewas infected with HSVexpressingNpas4 (HSV-Npas4) (Fig. 6A), and activationof Npas4 gene targets was examined using im-munostaining. HSV-Npas4 induced the expres-sion of c-Fos (Fig. 6B), but a transcriptionallyinactive version of Npas4 (∆Npas4) did not, con-firming that the transcription activation abilityof Npas4 is required.We also investigatedwheth-er expression of Npas4 is sufficient to induce

Fig. 6. Acute expression of Npas4 in CA3 reverses short-term and long-term memory deficits ob-served in Npas4−/− mice. (A) Expression of Npas4 in CA3 of Npas4−/− mice using HSV-Npas4. (B)Npas4, but not DNpas4, restores expression of c-Fos in vivo. (C) Expression of Npas4, but not DNpas4,drives activity of PIBDNF independent of KCl depolarization. n = 4 cultures. *P ≤ 0.001. (D) Npas4−/−micewith Npas4 injected into CA3 freeze at similar levels to Npas4+/+ mice injected with GFP 1 hour and24 hours after training. CA1 injection of Npas4 or CA3 injection of DNpas4 did not reverse the memorydeficit. *P < 0.001.

www.sciencemag.org SCIENCE VOL 334 23 DECEMBER 2011 1673

RESEARCH ARTICLES

on

Aug

ust 1

9, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

BDNF by measuring the activity of a PIBDNF re-porter construct in vitro. We transfected Cre intoNpas4flx/flx neurons and found that activity of thePIBDNF reporter was abolished. CotransfectingNpas4, but not ∆Npas4, rescued the activity ofthe PIBDNF reporter (Fig. 6C).

We then determined whether expressingNpas4 in CA3 of the global knockout mice issufficient to restore the ability to form long-termcontextual memories. The use of HSVallowed usto acutely express Npas4, with a peak expression3 days after injection (27, 28).Mice were injectedwith virus, trained 3 days after injection, andtested 1 hour and 24 hours after training (Fig.6D). Expressing Npas4 in CA3 completely re-versed both the short-term and long-term contex-tual memory deficits observed in the globalknockouts, because Npas4 knockout mice withHSV-Npas4 injected into CA3 showed similarfreezing behavior to wild-type control animalsinjected with GFP. Global knockouts with HSV-Npas4 delivered to CA1 showed no such recov-ery (Fig. 6D). Expressing ∆Npas4 in CA3 failedto overcome the memory deficits in Npas4−/−

mice, confirming that activation of the geneticprogram regulated by Npas4 is required for res-cue of memory formation.

Discussion. We have identified a geneticpathway in CA3 required for rapid encoding ofhippocampal-dependent contextual memory. Al-though several studies have identified CA3 func-tion and output as essential to the encoding ofcontextual information (3–6), very little is knownabout the molecular mechanisms underlying thisprocess. We found that acute deletion of Npas4fromCA3 resulted in a dramatic reduction in IEGexpression and impaired contextual memory for-mation, and that expression of transcriptionallyactive Npas4 in CA3 was sufficient to restoreboth IEG expression and memory formation inthe global knockout. Additionally, we found thatexpression of Npas4 in CA1 is neither necessarynor sufficient for contextual memory formation.Although our viral strategy cannot target all ofCA1, these findings are in line with other studiesusing transgenic mouse lines targeting CREB inCA1 [(40, 41), but see (42)]. Our data indicatethat regulation of a transcriptional program byNpas4 is a mechanism through which CA3 sup-ports the rapid acquisition and consolidation ofcontextual information.

Activity-dependent gene expression is thoughtto be required for LTM, but not for STM (8). Weobserved a STM deficit in the Npas4 globalknockout mice, but not in the conditional CA3knockout (Fig. 3C). Although the STM impair-ment could be due to a developmental deficitcaused by germline deletion of Npas4 in theglobal knockout, the rescue by acute expressionof Npas4 argues against this explanation. It ispossible that basal levels of neuronal activitymaintain a low level of Npas4, which in turnprovides a moderate level of the downstreammolecules required for STM. Then, perhaps, al-though acute deletion of Npas4 does not re-

duce the level of those genes below thatrequired for STM, chronic deletion in the glob-al knockout results in insufficient levels to sup-port STM.

It is intriguing that Npas4 global knockoutmice function normally in auditory delay condi-tioning, which is hippocampus-independent butamygdala-dependent, because long-termmemoryformation in the amygdala is thought to be de-pendent on activity-regulated gene expression.We observed that the expression of Npas4 genetargets is attenuated in the Npas4 global knock-out, but not to the degree that was observed inthe conditional deletion (fig. S7), which suggeststhat compensatory pathways may result in someexpression of target IEGs. Conceivably these path-ways are sufficient to support memory forma-tion in the amygdala, but IEG expression fails toreach a level sufficient to support the hippocam-pal learning required for CFC. Alternatively, oradditionally, the activity-regulated genetic pro-gram induced through compensating pathwaysindependent of Npas4, although including certainIEGs such as c-Fos and BDNF, may not containall the components necessary for CFC. It seemslikely that acute deletion of Npas4 in the amyg-dala will result in impairment of auditory delayconditioning.

Our findings suggest the possibility of ahierarchical genetic program in which Npas4 isupstream of several activity-regulated genes. How-ever, Npas4 itself is regulated by activity at themRNA level, and although it reaches peak ex-pression slightly earlier than other rapidly re-sponding IEGs (Fig. 1C and Fig. 3B), it is unclearwhether Npas4 protein is synthesized quicklyenough to initiate the first wave of IEG expres-sion. It seemsmore likely that Npas4, through therecruitment of Pol II, only enhances and sustainsIEG expression at later time points, as suggestedrecently for Npas4-dependent regulation of BDNFtranscripts (43).

The mechanism by which Npas4 affects PolII recruitment to its target genes is not immedi-ately obvious. It could directly recruit Pol II togenomic regions in a manner similar to CREB-binding protein, or it could be indirectly involvedthrough interactions with other proteins, such asCREB (44, 45).

Our previous work identified a role for Npas4in the activity-dependent regulation of inhibitorysynapse development (18). Thus, the genetic pro-gram controlled by Npas4 may be involved incontextual memory formation, at least in part,through the modulation of inhibitory synapsesin the hippocampal circuit. Consistent with thisidea, learning-induced increases in inhibitory syn-aptic transmission have recently been reported inthe hippocampus (46, 47).

We have focused here on the role of Npas4 inhippocampus-dependent contextual learning, butthe genetic program regulated by this transcrip-tion factor likely contributes to several otherexperience-dependent processes. We hope toleverage the function of Npas4 in order to dis-

sect specific neural circuits actively engaged ininformation processing to better understand themolecular and cellular mechanisms underlyinglearning and memory.

References and Notes1. G. Neves, S. F. Cooke, T. V. Bliss, Nat. Rev. Neurosci. 9,

65 (2008).2. W. B. Scoville, B. Milner, J. Neuropsychiatry Clin. Neurosci.

12, 103 (2000).3. R. P. Kesner, Learn. Mem. 14, 771 (2007).4. T. Nakashiba, J. Z. Young, T. J. McHugh, D. L. Buhl,

S. Tonegawa, Science 319, 1260 (2008).5. K. Nakazawa et al., Neuron 38, 305 (2003).6. I. Lee, R. P. Kesner, Hippocampus 14, 301 (2004).7. H. P. Davis, L. R. Squire, Psychol. Bull. 96, 518

(1984).8. C. M. Alberini, Physiol. Rev. 89, 121 (2009).9. W. Tischmeyer, R. Grimm, Cell. Mol. Life Sci. 55,

564 (1999).10. S. Kubik, T. Miyashita, J. F. Guzowski, Learn. Mem.

14, 758 (2007).11. J. F. Guzowski, Hippocampus 12, 86 (2002).12. M. W. Jones et al., Nat. Neurosci. 4, 289 (2001).13. N. Plath et al., Neuron 52, 437 (2006).14. A. Fleischmann et al., J. Neurosci. 23, 9116

(2003).15. S. Kida et al., Nat. Neurosci. 5, 348 (2002).16. B. E. Lonze, A. Riccio, S. Cohen, D. D. Ginty, Neuron 34,

371 (2002).17. N. Ramanan et al., Nat. Neurosci. 8, 759

(2005).18. Y. Lin et al., Nature 455, 1198 (2008).19. M. S. Fanselow, H. W. Dong, Neuron 65, 7

(2010).20. J. W. Rudy, R. C. O’Reilly, Behav. Neurosci. 113, 867

(1999).21. M. S. Fanselow, Behav. Brain Res. 110, 73

(2000).22. J. Landeira-Fernandez, J. P. DeCola, J. J. Kim,

M. S. Fanselow, Behav. Neurosci. 120, 873(2006).

23. S. Maren, Eur. J. Neurosci. 28, 1661 (2008).24. S. G. Anagnostaras, G. D. Gale, M. S. Fanselow,

Hippocampus 11, 8 (2001).25. J. F. Guzowski, B. L. McNaughton, C. A. Barnes,

P. F. Worley, Nat. Neurosci. 2, 1120 (1999).26. M. E. Lonergan, G. M. Gafford, T. J. Jarome,

F. J. Helmstetter, Neural Plast. 2010, 139891(2010).

27. M. Barrot et al., Proc. Natl. Acad. Sci. U.S.A. 99,11435 (2002).

28. J. H. Han et al., Science 316, 457 (2007).29. J. M. Kornhauser et al., Neuron 34, 221 (2002).30. Z. Mao, A. Bonni, F. Xia, M. Nadal-Vicens,

M. E. Greenberg, Science 286, 785 (1999).31. V. Coulon, K. Chebli, P. Cavelier, J. M. Blanchard, PLoS

ONE 5, e11235 (2010).32. M. Sheng, S. T. Dougan, G. McFadden, M. E. Greenberg,

Mol. Cell. Biol. 8, 2787 (1988).33. R. Treisman, Cell 42, 889 (1985).34. R. Treisman, Cell 46, 567 (1986).35. B. J. Wagner, T. E. Hayes, C. J. Hoban, B. H. Cochran,

EMBO J. 9, 4477 (1990).36. L. Minichiello, Nat. Rev. Neurosci. 10, 850

(2009).37. Y. Lu, K. Christian, B. Lu, Neurobiol. Learn. Mem. 89,

312 (2008).38. T. K. Kim et al., Nature 465, 182 (2010).39. Y. Ben-Ari, A. Represa, Trends Neurosci. 13, 312

(1990).40. D. Balschun et al., J. Neurosci. 23, 6304

(2003).41. C. Pittenger et al., Neuron 34, 447 (2002).42. J. Athos, S. Impey, V. V. Pineda, X. Chen, D. R. Storm,

Nat. Neurosci. 5, 1119 (2002).43. P. Pruunsild, M. Sepp, E. Orav, I. Koppel, T. Timmusk,

J. Neurosci. 31, 3295 (2011).

23 DECEMBER 2011 VOL 334 SCIENCE www.sciencemag.org1674

RESEARCH ARTICLES

on

Aug

ust 1

9, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

44. T. Nakajima et al., Cell 90, 1107 (1997).45. R. P. Kwok et al., Nature 370, 223 (1994).46. Y. Cui et al., Cell 135, 549 (2008).47. S. Ruediger et al., Nature 473, 514 (2011).

Acknowledgments: We thank C. M. Fletcher, A. Heynen,C. Alberini, C. Jennings, R. Desimone, M. Baratta,J. Biedenkapp, R. Froemke, J. Czerniawski, and D. Bambah-Mukkufor critical reading of the manuscript. We also thank G. Haleand M. Wilson for help characterizing the Npas4−/− mice

and S. Ramirez for help with the open field assay. Y.L.acknowledges the generous support of the McGovernInstitute for Brain Research at MIT. This work wassupported by the MIT Presidential Marcus Fellowship toHonor Norman B. Leventhal (K.R.), a postdoctoral fellowshipfrom the MIT Simons Initiative on Autism and the Brain(G.M.B), NSF grant IOS 0919159 (T.O.), a WhitehallFoundation research grant, an Anonymous Foundationresearch grant, the John Merck Scholar Program, andNIH grant MH091220-01 (Y.L.).

Supporting Online Materialwww.sciencemag.org/cgi/content/full/334/6063/1669/DC1Materials and MethodsSOM TextFigs. S1 to S7References (48–53)

9 May 2011; accepted 21 October 201110.1126/science.1208049

How a DNA Polymerase Clamp LoaderOpens a Sliding ClampBrian A. Kelch,1* Debora L. Makino,1*† Mike O’Donnell,2 John Kuriyan1,3,4,5‡

Processive chromosomal replication relies on sliding DNA clamps, which are loaded ontoDNA by pentameric clamp loader complexes belonging to the AAA+ family of adenosinetriphosphatases (ATPases). We present structures for the ATP-bound state of the clamp loadercomplex from bacteriophage T4, bound to an open clamp and primer-template DNA. Theclamp loader traps a spiral conformation of the open clamp so that both the loader and theclamp match the helical symmetry of DNA. One structure reveals that ATP has been hydrolyzedin one subunit and suggests that clamp closure and ejection of the loader involvesdisruption of the ATP-dependent match in symmetry. The structures explain how synergyamong the loader, the clamp, and DNA can trigger ATP hydrolysis and release of theclosed clamp on DNA.

Chromosomal DNA replication relies onmultiprotein replicases that copy DNAwith high speed and processivity (1, 2).

The polymerase subunits of the replicase aretethered to ring-shaped sliding clamps that en-circle DNA, allowing the polymerase to bindand release DNA repeatedly without dissociatingfrom the progressing replication fork. All repli-cases use a conserved sliding clamp mechanismfor processivity (3–6), even though the bacterialand eukaryotic replicative polymerases haveevolved independently (7, 8). Sliding clamps arealso used for scanning DNA in several DNA re-pair processes (9).

Sliding clamps cannot load onto DNA sponta-neously because they are closed circles (5, 10, 11)(Fig. 1A). Instead, adenosine triphosphate (ATP)–dependent complexes known as clamp loadersopen the sliding clamps and load them ontoprimed DNA in the correct orientation for pro-ductive engagement of the polymerase [the clamploaders are the g/t complex in bacteria, replicationfactor–C (RFC) in eukaryotes and archaea, and

gp44/62 in T4 bacteriophage (Fig. 1B)]. Clamploaders are members of the AAA+ superfamilyof adenosine triphosphatases (ATPases), a diversegroup of oligomeric ATPases whose functions

include motor and helicase activity and the abil-ity to disassemble protein complexes (12, 13).In contrast to typical AAA+ATPases, all clamploader complexes are pentameric rather than hex-americ. The lack of the sixth subunit in the clamploader creates a gap in the assembly that is es-sential for the specific recognition of primer-template junctions (14, 15). The five subunits ofthe clamp loader are designated A, B, C, D, and Eand are identified in Fig. 1B.

Each clamp loader subunit consists of threedomains that are conserved in structure (14, 16–19).The first two of these domains form a AAA+ATPase module, and five of these modules arebrought together in intact clamp loaders suchthat ATP can be bound at interfacial sites (14)(Fig. 1B). The third conserved domain in eachsubunit is integrated into a circular collar that holdsthe assembly together in the absence of ATP.

A key role for ATP in the mechanism ofclamp loaders is to trigger the formation of aspiral arrangement of AAA+modules, leading to

1Department of Molecular and Cell Biology and CaliforniaInstitute for Quantitative Biosciences, University of California,Berkeley, CA 94720, USA. 2Howard Hughes Medical Institute,Rockefeller University, New York, NY 10021, USA. 3Depart-ment of Chemistry, University of California, Berkeley, CA94720, USA. 4Howard Hughes Medical Institute, University ofCalifornia, Berkeley, CA 94720, USA. 5Physical BiosciencesDivision, Lawrence Berkeley National Laboratory, Berkeley, CA94720, USA.

*These authors contributed equally to this work.†Present address: Max Planck Institute of Biochemistry, De-partment of Structural Cell Biology, D-82152 Martinsried,Germany.‡To whom correspondence should be addressed. E-mail:kuriyan@berkeley.edu

A

B

clamp loader

I

II

III5 2

143

6

ADP

ATP ATP

ADP

primer-templatejunction

3

5

3

5

bacterial eukaryotic T4 bacteriophage

collar

AAA+module

domain IIdomain I

E DC

B

A

E

D

CB

AE

D

C B

AA

RFC1RFC5

A

ATP

ATP

gp44

gp62

ATP

gp44

Fig. 1. Clamp loaders and sliding clamps. (A) Clamp-loading reaction. The clamp loader has low affinityfor both clamp and primer-template DNA in the absence of ATP. Upon binding ATP, the clamp loader canbind the clamp and open it. The binding of primer-template DNA activates ATP hydrolysis, leading toejection of the clamp loader. (B) Three classes of clamp loaders. Bacterial clamp loaders are pentamersconsisting of three proteins: d (A position), g (B, C, and D positions), and d´ (E position). Eukaryotic clamploaders (RFCs) consist of five different proteins, with the A subunit containing an A´ domain that bridgesthe gap between the A and E AAA+ modules. The T4 bacteriophage clamp loader consists of two proteins:gp44 (the B, C, D, and E subunits) and gp62 (the A subunit).

www.sciencemag.org SCIENCE VOL 334 23 DECEMBER 2011 1675

RESEARCH ARTICLES

on

Aug

ust 1

9, 2

013

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om