ARTICLE OPEN DOI:10.1038/s41467-017-01004-6 Supramammillary glutamate neurons are a key node of the arousal system Nigel P. Pedersen 1, Loris Ferrari2,3, Anne Venner 2,3, Joshua L. Wang 2, Stephen B.G. Abbott2, Nina Vujovic3, Elda Arrigoni2,3, Clifford B. Saper2,3 & Patrick M. Fuller2,3 Basic and clinical observations suggest that thecaudal hypothalamus comprises a keynode 0 123456789 oHfetrheewaescreepnodrintgthaartogulsuatlasmyasttee-mre,lbeaustinthgenceeulrlotnyspeosfuthnedesrulpyirnagmtahmismairlelarnyortefguilolynu(nSdueMrsvtgoluot2d). producesustained behavioral andEEGarousal whenchemogeneticallyactivated.This effect isnearlyabolishedfollowingselectivegeneticdisruptionofglutamatereleasefromSuMvglut2 neurons. Inhibition of SuMvglut2 neurons decreases and fragments wake, also suppressing theta and gamma frequency EEG activity. SuMvglut2 neurons include a subpopulation con- taining both glutamate and GABA (SuMvgat/vglut2) and another also expressing nitric oxide synthase(SuMNos1/Vglut2).ActivationofSuMvgat/vglut2neuronsproducesminimalwakeand optogenetic stimulation of SuMvgat/vglut2 terminals elicits monosynaptic release of both glutamate and GABA onto dentate granule cells. Activation of SuMNos1/Vglut2 neurons potently drives wakefulness, whereas inhibition reduces REM sleep theta activity. These results identify SuMvglut2 neurons as a key node of the wake−sleep regulatory system. 1DepartmentofNeurologyandEpilepsyService,EmoryUniversity,Atlanta,GA30322,USA.2DepartmentofNeurology,BethIsraelDeaconessMedical Center,Bostan,MA02215,USA.3DivisionofSleepMedicine,HarvardMedicalSchool,Bostan,MA02215,USA.CliffordB.SaperandPatrickM.Fuller contributedequallytothiswork. CorrespondenceandrequestsformaterialsshouldbeaddressedtoN.P.P.(email:[email protected]) ortoP.M.F.(email:[email protected]) NATURECOMMUNICATIONS|8: 1405 |DOI:10.1038/s41467-017-01004-6|www.nature.com/naturecommunications 1 ARTICLE NATURECOMMUNICATIONS|DOI:10.1038/s41467-017-01004-6 I n the 1930s, von Economo1 proposed that damage to comparable to brainstem arousal sites, and at a lower threshold diencephalic-mesencephalic junction was responsible for than for the thalamus. This induced arousal was seen even marked somnolence in patients with epidemic encephalitis after destruction of the thalamus4, while caudal hypothalamic lethargica, and Ranson2 was able to reproduce this somnolence lesions or inactivation resulted in marked and often prolonged withlesionsofthecaudalhypothalamusinmonkeys.Supporting somnolence2, 3, 5–10. These experiments and similar thisobservation,MoruzziandMagoun3foundthatstimulationof observations in humans1, 11–16 supported the existence the caudal hypothalamic region could elicit wakefulness of a key node of the arousal system within the caudal and electrocortical low-voltage fast activity at a threshold hypothalamus. a lox2722 mCherry hM3Dq b ITR hSyn loxP WPREhGH-pA vglut2 A A A A V V SuM LM MM LM vglut2-IRES-Cre lox2722 WPRE ITRhSyn hM3Dq mCherry loxP hGH-pA c d Saline CNO Wake NPM002 Lights off Wake NPM002 Lights off NREM NREM REM REM NPM003 NPM003 NPM004 NPM004 NPM005 NPM005 0(10 AM) 6 h 18 24 0(10 AM) 6 h 18 24 e f 18 h 18 h 100 Hz FET 0 Hz g h * i % Wakefulness 15000 *************************** * CNO 2µectral power (V) 1500000 CSaNlOine** * ** 111A000U–––345 CCCNNonOOtr o-- l3900 mmiinn 111000–––357 Hippocampus NaCl Sp CNO - 180 min 0 0 10–6 H6ours from r1e2jection 18 SW Delta LTheta HTheta Beta Gamma 10 Hz 100 Hz j k 30 HzSaline CNO FFT 0 Hz ECoG EMG 1 s 2 NATURECOMMUNICATIONS|8: 1405 |DOI:10.1038/s41467-017-01004-6|www.nature.com/naturecommunications ARTICLE NATURECOMMUNICATIONS|DOI:10.1038/s41467-017-01004-6 While orexin neurons in the lateral hypothalamus and tuber- SuM region are glutamatergic36,37, there is anatomical evidence omammillary histaminergic neurons both promote arousal, that some of these neurons may contain other neuromodula- lesions of these caudal hypothalamic cell groups have not been tors23,38–41includingasubpopulationofSuMneuronstargeting found to cause the level of hypersomnolence produced by larger thehippocampusthatcontainanatomicalmarkersofbothGABA posterior hypothalamic lesions17, 18. Based upon its anatomical and glutamate release42,43. Here we used a combination of tar- connectivity and cellular composition, we hypothesized that the geted genetic and viral vector technologies in behaving mice to supramammillarynucleus(SuM),aregionlyingjustdorsaltothe determinewhetherSuMneuronscontributetoelectrocorticaland mammillarybodyinthecaudalhypothalamus,mightcomprisea behavioral arousal. We then employed in vitro optogenetics and key site for wake-promotion. The SuM provides extensive application of a novel non-parametric mapping technique to innervation of the cerebral cortex19, wake-promoting basal fore- delimitandcharacterizethreedistinctcellpopulationsintheSuM brain20–25, and the hippocampus19, 20, 26–28, 33. Neurons in the region, each of which differentially contribute to cortical and SuM are known to play a role in driving hippocampal theta hippocampal activation and behavioral wake. Our results show rhythm, displaying theta band field potentials in freely moving that SuMVglut2 neurons are necessary for normal wakefulness, rats29, and unit discharges that correlate with this theta mod- with activation potently promoting theta and gamma EEG ulation30,31.PreviousworkhasexaminedtherelationshipofSuM activity and wakefulness. SuMNos1/Vglut2 neurons can promote activitytohippocampallocalfieldpotentialsandlearning27,32–34, wake when activated and are also necessary for normal REM andsuggestedaroleforSuMneuronsinREMsleep13,35,butthe theta activity. SuMVgat/Vglut2 neurons provide dense innervation SuM has not been studied at all in relation to wakefulness, nor of the dentate gyrus and CA2, monosynaptically releasing both withrespecttothepossiblydistinctrolesofSuMsubpopulations GABA and glutamate on to the proximal dendrites of dentate in REM sleep and wakefulness. Whereas most neurons in the granule cells, but only minimally affecting sleep−wake and, a NPM047 NPM051 NPM052 FF FF SuM SuM SuM LM MM LM awake awake awake 2.6 h 4.2 h 8.3 h b Bregma - 1.8 Heatmap of injection sites (n = 21) Bregma - 3.0 c Correlated with duration of wake, FDR < 0.01 Fig.2Wake-promotingneuronsarewithintheSuMregion.aSmallerinjectionsites(redoutline)showingprolongedwakefulness(hoursofwakeafter CNOatbottomleftofeachpanel)washighestwithinvolvementofthewholeSuM(scalebar500µm).bPlotofallinjectionsitesasa‘heatmap’of injectionoverlap(n=21,darkredminimaloverlap(1case),whitemaximal(18cases)).cNon-parametricpermutationanalysisshowsthattheareaof transductioncorrelatedwiththelongestdurationofwakefulnessafterCNOisthecaudalhypothalamus,mostnotablytheSuM(yellowareafalsediscovery rate<0.01).MM–medialmammillary,LM–lateralmammillary,andFF–fieldsofForel Fig.1Caudalhypothalamicactivationpotentlydriveswakefulness.aCartoonshowingAAV-FLEX-hM3Dq-mCherryvirusrecombininginvglut2neuronsin thepresenceofCre.bExampleinjectionsiteshowingdiaminobenzadine(DAB)labelingofmCherry(red−brownstaining,redoutline)andnickel/cobalt enhancedDAB(black)labelingofcFosinnuclei(seealsoinset,scalebar10µm)ofneuronsactivatedbyCNO(mainpanelscalebar500µm,SuMshown inblackoutline).c,dHypnogramsaftervehicleandCNOshowingprolongedwakefulnessafteractivationofglutamateneurons(W–wake,N–NREMsleep, R–REMsleep;gray=darkperiod).e,fCompressedspectralarray(CSA)basedonfastFouriertransform(FFT)aftervehicleandCNO(0–100Hz,blackbar showsthedarkperiod,shortwhitebarsrepresent20minandarespaced1hapart;timeperiodsinblueandredboxesforNPM005incanddareshownin eandf),withprominentthetaactivityafterCNO.gGroupdata(mean±SEM)showingsignificantlyhigherwakefulnessforninehoursafterCNO(2-way repeatedmeasuresANOVAfortreatmentp<0.0001,Bonferroni***p<0.001;*p<0.05).hWakingEEG(mean±SEM)showsincreasedhightheta (HTheta,7−13Hz,p=0.0148),gamma(30−120Hz,p=0.0005),anddecreasedlowthetacomponentoftheFFTpowerspectrum(LTheta,4−7Hz,p= 0.0027).iCorticalandhippocampal(inset)EEGFFTspectra(logfrequency,arbitraryunits(AU))showingprominenthippocampal ~9Hzactivity.jRaw wakeEEGandEMGdatawithCSA(0−30Hz)over5sforvehicle,andk.CNO,withprominent9−10HzactivityvisibleintheCSAinthelattercase NATURECOMMUNICATIONS|8: 1405 |DOI:10.1038/s41467-017-01004-6|www.nature.com/naturecommunications 3 ARTICLE NATURECOMMUNICATIONS|DOI:10.1038/s41467-017-01004-6 a Cre-recombinase eGFP WPRE b NPM093 ITRhSyn + hGH-pA lox2722 mCherry hM3Dq ITR hSyn loxP WPREhGH-pA vglut2 A A A A V V SuM vglut2lox/lox vglut2 lox2722 WPRE ITR hSyn hM3Dq mCherry loxP hGH-pA c d NPM093 SuM e f Saline CNO NPM090 Lights off NPM090 Lights off Wake Wake NREM NREM REM REM NPM093 NPM093 0(10 AM) 6 h 18 24 0(10 AM) 6 h 18 24 g h NPM090 saline 18 h NPM090 CNO 18 h i j 100 30 * Saline CNO 25 m Wakefulness 50 ake spectru 211050 % CNO % W 5 Saline 0 0 6Hours fro1m2 injection18 24 SW Delta LTheta HTheta Beta Gamma interestingly, do not as potently promote theta EEG activity SuM,theseresultssuggestthatSuMVglut2neuronscompriseakey despitetargetingthehippocampalnetwork.Inconcertwiththese nod of the sleep−wake regulatory system that is positioned to observations and the described anatomical connectivity of the control both the hippocampal system and arousal. 4 NATURECOMMUNICATIONS|8: 1405 |DOI:10.1038/s41467-017-01004-6|www.nature.com/naturecommunications ARTICLE NATURECOMMUNICATIONS|DOI:10.1038/s41467-017-01004-6 Results associated with the most prolonged wakefulness (see examples, SuMvglut2 neurons drive wakefulness. In our first experiment, Fig.2a).Toanalyzethesedatainanobjectivefashion,wemadea wetestedwhetherglutamateneuronsofthecaudalhypothalamus threedimensional(3D)mapoftheinjectionsite,plottedas30µm drive wakefulness. In order to selectively activate caudal cubic voxels on to a template derived from a high resolution hypothalamic glutamate neurons, including SuM glutamate mousebrainMRIscan(Fig.2b).Wethenusedanon-parametric (SuMvglut2)neurons, wefirst made large stereotaxic injections of permutation test of voxels containing transduced neurons and an adeno-associated viral vector (AAV) that conditionally correlated this with the duration of wakefulness resulting from expressed an excitatory modified human G-protein coupled SuMactivation(Fig.2c,n=21;Methodssection).Thisproduced muscarinic receptor (hM3Dq)44,45 fused to the mCherry fluor- an anatomical z-score, revealing that the combination of voxels escent protein (AAV-hM3Dq), into the caudal hypothalamus of that was statistically associated with increased wakefulness vglut2-IRES-Cre mice (Fig. 1a). Functional transduction of the corresponds best to the SuM region (false discovery rate<0.01 SuMneuronswasshownbyadministrationofthehM3Dqligand and corresponding to a z≥3.216, n=21). Conversely, caudal clozapine-n-oxide (CNO, 0.3mg/kg in 10µl/g normal saline hypothalamic regions immediately surrounding the SuM region vehicle i.p.), which drove cFos expression in the transduced did not significantly contribute to prolonged wakefulness. neurons (Fig. 1b, upper right). Administration of CNO after Involvement of somewhat further cell groups such as the lateral injection of AAV-hM3Dq in non-Cre-expressing control litter- and perifornical regions may have accounted for the most mates, or injections of vehicle (Fig. 1c) after injection of AAV- prolongeddurationofwakefulnesswiththelargestinjectionsites, hM3DqinCre-expressingmicewaswithouteffectonsleep−wake given the known wake-promoting effect of orexin neurons that or cFos expression. In contrast, CNO-driven activation of also express vglut2 in that region. Smaller injections with hM3Dq-expressing SuMvglut2 neurons potently promoted con- transduction constrained within the SuM resulted in just over tinual wakefulness, in some cases for periods as long as 12h, 5h of wakefulness (saline 29±5min versus CNO 310±40min, during the animals’ normal sleep period (Fig. 1d; administration n=6, p=0.0012), indicating that this caudal hypothalamic cell 3h after lights-on (Zeitgeber (ZT) 3, 10 AM; Latency to sleep: group independently promotes wakefulness. saline 22±4 (SEM), CNO 534±54min, paired t-test, p=0.000014, n=9, see group data Fig. 1g). In these mice, CNO SuM actions depend on glutamate. Given the presence of also caused a significant increase in higher range theta activity other mediators such as nitric oxide, GABA, and several pep- (HTheta 7–13Hz, paired t-test with Bonferroni correction, tides23,38–43 in the SuM, we next sought to determine whether p=0.0148, n=5, spectral peak at about 9Hz), and decrease in glutamate is necessary for the wake-promoting effect of SuM lower frequency theta EEG activity (low range theta (LTheta) neuronal activation. To test this hypothesis, and whether the 4–7Hz, p=0.0027) (Fig. 1e, f, h, i) with a significant increase in action of unopposed GABA release might decrease wakefulness, gamma frequency EEG activity (gamma 30–120Hz, p=0.0005). we placed stereotaxic injections of both AAV-Cre and a condi- Bipolarrecordingfromthehippocampus(n=6)revealedsimilar tional AAV-hM3Dqinto theSuM region ofvglut2lox/lox mice,to EEG spectral changes, confirming concomitant hippocampal concurrently disrupt synaptic glutamate release and express activity with a spectral peak at 8.5–9Hz (Fig. 1i upper right). hM3Dq in transduced neurons (Fig. 3a). Because only neurons Importantly, the repertoire of behavior after CNO in these mice that had been transduced with the AAV-Cre (and therefore was normal (Supplementary Movie 1), and the amount or speed deleted the vglut2 gene) would also express the hM3Dq, CNO oflocomotoractivitydidnotaccountforthetaactivity(micewere would only drive neurons that lacked functional VGLUT2 pro- oftenstationary,e.g.,seerawdatainFig.1j,kwithperiodsoflow tein. Histologically, there was effective co-expression of hM3Dq EMG in both examples, but increased theta power (red) in and cFos after CNO over a large part of the SuM (Fig. 3b), with Fig. 1k; Supplementary Movie 1). Given the known role of the correlated loss of vglut2 in situ hybridization signal (compare SuMinthegenesisofthethetaactivity,wehypothesizedthatthe Fig. 3c, d). Activation of these SuM neurons resulted in a brief wake-promoting cell group was either identical with or adjacent period of wakefulness that was not statistically significant from to the SuM neurons that promote theta. vehicle injection (vehicle 21±4 (SE)min versus CNO 49± WenextscatteredsmallinjectionsofAAV-hM3Dqthroughout 12min,p=0.06,n=6,Fig.3e−h,summarydataFig.3i),andwas the SuM and adjacent regions to better understand the precise farshorterthancomparableactivationsofneuronswithoutvglut2 location of the wake-promoting neurons (n=15). We studied deletion (Fig.1c, d),suggesting thatglutamate release fromSuM each injection site in detail, noting which nuclei were wholly or neuronsisrequiredforthebulkofthewake-promotingeffectsof partially involved in the injection site and their degree of SuM activation. Furthermore, EEG spectral changes associated transduction. We noted that small injection sites flanking the withactivationofSuMvglut2-onlyneuronswereabolished,with SuM were associated with minimal wakefulness after CNO no changes in individual spectral bands in any state (p>0.99 in injection, while those involving the whole of the SuM were wake, n=6, see wake spectra Fig. 3g, h, j). Fig.3Wake-promotionafterSuMactivationdependsonglutamate.aCartoonshowingthatafterco-injectionofAAV-FLEX-hM3Dq-mCherryandAAV- Creinavglut2floxmouse,thatvglut2isdisruptedinneuronsexpressingCre,withCrepositiveneuronsalsoabletodrivehM3Dqexpression.Thus,the onlyneuronsactivatedbyCNOareneuronsthatlackglutamaterelease.bmCherry-likeimmunoreactivityandcFosexpressionafterCNOadministration (left)showingwherehM3Dqwastransduced,andcFosexpressionconsistentwithcellularactivationafterCNO.cNormalexpressionofvglut2, revealedbyinsituhybridization,anddinsituhybridizationshowingthelossofvglut2-expressionintheinjectionsiteinthesamemouseshowninb(area withnear-completesignallossshowninyellow,areawithdeletioninsomeneuronsisoutlinedinorange).e,fHypnogramsoftwomiceshowingmarked attenuationofwake-promotingeffectsofSuMactivation,consistentwithsubstantialdependenceonglutamaterelease.g,hCSAshowingtheeffectof vehicleandthenear-absenteffectofactivationwithCNOintheperiodshownbytheblueandredboxesincandd,respectively(0−100Hz,blackbar showsthedarkperiod,shortwhitebarsrepresent20minandarespaced1hapart).iPooleddata(mean±SEMin1hbins,n=6)aftersalineand CNOadministration,withposthoccomparisonsshowingthatmoreofonlythefirsthourisspentawakeafterCNO(Bonferronicorrectedcomparison for1hp=0.0153,allotherhoursp>0.99).jSpectralchangesassociatedwithSuMactivationarealsoabolished(wakefrom6hafterinjectionshown, mean±SEM,n=6) NATURECOMMUNICATIONS|8: 1405 |DOI:10.1038/s41467-017-01004-6|www.nature.com/naturecommunications 5 ARTICLE NATURECOMMUNICATIONS|DOI:10.1038/s41467-017-01004-6 a lox2722 mCherry hM4Di b NPM012 ITRhSyn loxP WPREhGH-pA vglut2 A A AV AV SuM vglut2-IRES-Cre lox2722 WPRE ITRhSyn hM4Di mCherry loxP hGH-pA c CNO d CNO 2 min e f g mV Con CNO Wash 10 5 min h Firing frequency (Hz)543210 CNO i C10o ns j CNO 0 2 4 6 8 10 12 14 16 Time (min) 10 s k l Saline CNO NPM010 NPM010 Wake Wake NREM NREM REM REM NPM012 NPM012 Lights off Lights off 0(7 PM) 3 h 9 12 0(7 PM) 3 h 9 12 m n NPM012 saline 12 h NPM012 CNO 12 h 100 Hz FFT 0 Hz o p q 360 Saline 50 * 80 CNO CNO Saline Sleep for first 6 h (min)32110482600000 * Numberr of sleep bouts 43210000 Longest wake (min)642000 * 0 0 0 Saline CNO Saline CNO Saline CNO r s t Wake spectrum REM spectrum NREM spectrum 1000 1500 2500 2)V 800 *** * *** 2000 SCaNlOine µwer ( 600 1000 1500 Spectral po 420000 500 1050000 0 0 0 SW Delta LTheta HTheta Beta Gamma SW Delta LTheta HTheta Beta Gamma SW Delta LTheta HTheta Beta Gamma Chronicvglut2deletionintheSuMhadlittleeffectonamounts experiments never covered the entirety of the SuM, and because of wake, REM, and NREM sleep, all of which remained within of the time it took for the full deletion to occur and for the normal limits for mice (wake 45.2±2.5%, NREM 48.4±2.4%, animals to recover, there was ample time for compensation. We REM6.4±0.2%,n=6).However,liketheibotenicacidlesionsof therefore tested the effect of acute inhibition of the SuMVglut2 the rat SuM by Renouard et al.35, the deletions in these cells on wakefulness. 6 NATURECOMMUNICATIONS|8: 1405 |DOI:10.1038/s41467-017-01004-6|www.nature.com/naturecommunications ARTICLE NATURECOMMUNICATIONS|DOI:10.1038/s41467-017-01004-6 SuMvglut2 neuronal inhibition causes somnolence. To examine understand if they were also capable of driving cortical and the effect of acute inhibition the SuMVglut2 neurons, we used a hippocampal activation and/or behavioral arousal. We first conditionallyexpressedmodifiedinhibitoryhumanmuscarinic 4 sought to verify the exact location of the GABA/glutamate neu- receptor (hM4Di)44, 45. We made large injections of an AAV- rons in the SuM. We performed in situ hybridization for vglut2 FLEX-hM4Di-mCherry (AAV-hM4Di) vector in the posterior mRNAinavgat-IRES-Crereportermouse(L10reporterline,see hypothalamus of vglut2-IRES-Cre mice (Fig. 4a), deliberately Methods), revealing Vglut2 expression in all of the GABAergic including as much as possible of the SuMvglut2 cell population neurons surrounding the mammillothalamic tract in the SuM— (Fig. 4b). We validated that CNO was able to inhibit transduced thusneuronsintheSuMareofeitherSuMvglut2orSuMvgat/vglut2 neurons by recording SuM neurons in vitro in the cell attached phenotype (Fig. 5a). The latter neurons are larger than the mode. CNO (5µM) reduced the firing frequency of SuMvglut2 adjacent ones expressing only vglut2 (20.54±0.64 (SEM)µm neurons that express hM4Di (firing frequency in control 0.89± longestaxisofsoma,versus13.93±0.50µm,unpairedt-test,p< 0.27Hz, CNO 0.09±0.05Hz; p=0.02 paired t-test, 0.0001, n=12,12), and correspond with what has been termed n=6) (Fig. 4c−g). CNO had no effect in SuMvglut2 neurons that thegrandicellularfield(SuMg)40,formingaclearsubsetoflarger only express mCherry (firing frequency in control 1.2±0.6Hz, neurons in the SuM (Fig. 5b). There were no examples noted of CNO 1.35±0.65Hz; p=0.42, paired t-test, n=5) (Fig. 4h−j). neurons that were only vgat positive within the SuM. In freely behaving mice, increased sleep and fragmentation To determine whether these neurons co-release both neuro- of wakefulness and sleep were present after CNO injection transmitters, we conditionally transduced channelrhodopsin-2 (Fig. 4k−n). Increased NREM sleep was significant, elevated by (ChR2)intheSuMofvgat-IRES-Cremice(SuMvgat;Fig.5c),then about20%overthe6hfollowingCNOadministration(injectionat made whole-cell recordings from dentate granule cells in ZT12, typical time of maximal wakefulness, saline 33.8±2.80% hippocampal brain slices (Fig. 5d). Photostimulation of Vgat (SEM) versus CNO 41.95±1.49%, 2-way ANOVA, p=0.0199, positiveaxonsandterminalsoriginatingfromtheSuM(asshown n=6, Fig. 4o). This was at the expense of both wakefulness and in the cartoon in Fig. 3d) evoked fast postsynaptic currents in REM sleep, although neither reduction reached statistical signifi- granule cells (15/18 neurons). Application of bicuculline or cance (2-way ANOVA by state; wake—62.4±2.85 versus 55.8± DNQX reduced the amplitude of, but did not eliminate, photo- 1.94%, p=0.0659, n=6; REM −3.31±0.34 versus 2.31±0.72%, evokedpostsynapticcurrents.Thecombinationofbicucullineand p=0.2614). There was an increased number of sleep bouts in the DNQX completely blocked postsynaptic responses, suggesting sixhoursfollowinginjection,aswellasareductioninthemaximal mediation by release of both GABA and glutamate, respectively time for which wakefulness was maintained (saline 24.5±2.56 activating GABA-A and AMPA receptors (Fig. 5e). Selective (SEM) versus 37.3±2.90 bouts, paired t-test, p=0.0194, n=6, blockade with these two antagonists revealed that GABA-A- Fig.4p),withthelongestwakeboutsaftersalinebeingsignificantly mediated postsynaptic currents had typical slower kinetics than longer than after CNO (44.5±7.83 versus 28.2±4.98min, AMPA-mediatedcurrents46(Fig.5f;decaytimeconstants:34.57 p=0.0152, n=6, Fig. 4q). The spectral power changes were ±5.99ms for the GABA-A-mediated current and 5.15±1.06ms inverse to thoseof activation, i.e., we saw a significant decreasein for the AMPA-mediated currents, fitting with a single exponen- high frequency theta and gamma power during wakefulness tial;n=3).Reversalpotentialswere−49.74±3.65mV(SEM)and (HThetap<0.0001,andGammap=0.0472,n=5,Fig.4r).There +11.94±2.02mV(n=4neuronsfromfourbrainslicesfromfour was also an increase in the power of low-frequency (4−7Hz), but mice; p<0.001, paired t-test), compatible with GABA-A- nothigh-frequency(7−13Hz)thetaduringREMsleep(p<0.0001, mediated chloride and AMPA-mediated sodium currents, n=5, Fig. 4s), with no effect on other REM spectral bands respectively (Fig. 5g, h). Importantly, both the AMPA-mediated (p>than0.16forallbands),norinthepowerspectrumofNREM and the GABA-A-mediated photo-evoked postsynaptic currents sleep (p=0.38 to>0.99 for all bands, Fig. 4t). These changes in hadashortonsetdelaycharacteristicofphotostimulation-evoked power spectrum and sleep architecture are consistent with release (GABA-A-mediated: 5.04±0.37ms and AMPA- sleepinessduringthewakestateandtheinabilitytosustainlonger mediated: 4.33±0.23ms, n=8, 9 neurons (from 3 vglut2-Cre bouts of wakefulness, but normal NREMsleep. and 5 vgat-Cre mice, respectively), p=0.002, paired t-test) supporting monosynaptic release of both GABA and glutamate from SuMvgat/Vglut2 terminals onto dentate granule cells (Fig. 5i, SuMvgat/vglut2 neurons. GABAergic neurons surround the j).Overall,photostimulationoftheSuMvgat/Vglut2→DGpathway mammillothalamictractasitpenetratestheSuM37,43,andspread evoked GABA-A-mediated and AMPA-mediated synaptic medially and laterally in the more posterior SuM. The terminals responses in a large proportion of dentate granule cells (82%) of these neurons in the hippocampus have been reported to whereas it evoked solely GABA-A-mediated or AMPA-mediated containmarkersofbothGABAandglutamaterelease42,43.Given responsesinonly9and9%ofgranulecells(n=18),respectively. thattheseneuronsarewithinourregionofinterest,wesoughtto WealsoconditionallyexpressedChR2inSuMvglut2neurons,with Fig.4InhibitionoftheSuMresultsinsomnolenceandwakefragmentation.aCartoonshowingAAV-FLEX-hM4Di-mCherryrecombininginvglut2neurons thatexpressCre.bConditionaltransductionofinhibitoryhM4Diinvglut2-IRES-Cremice(red−brownstainingoutlinedinred;scalebar500µm).c−gCNO (5µM)inhibitsthefiringofSuMvglut2neuronsthatexpresshM4Di-mCherry(ccurrent-clamprecordings;dcellattachedrecordings;e−gexpandedtraces oftheoutlinedregionfromd,showfiringincontrol,duringCNOapplicationandinwashout).h−jCNO(5µM)hasnoeffectonthefiringofSuMvglut2 neuronsthatonlyexpressmCherry(h;firingfrequency,1-minbin,redarrowsindicatethe1-minbinsrepresentedini,j;i,jcellattachedrecordingsofthe neuroninh).k,lHypnogramsfromtwomiceover12haftervehicleandCNO,showingincreasedsleepandfragmentationofsleepandwakefulnessafter inhibitionoftheSuM.m,nCSAaftervehicleandCNOfromtheperiodsshownbytheblueandredboxes,respectively,incandd(0−100Hz,blackbar showsthedarkperiod,shortwhitebarsrepresent20minandarespaced1hapart).oThereisincreasedoverallNREMsleepinthe6hafterCNO administration(~20%;2-wayANOVAfor6-hourlybinsbetweenvehicleandCNO,graphwithmean±SEM,p=0.0199),pwake-sleepfragmentationwith anincreaseinsleepboutsoverthe6hafterCNOwhencomparedwithvehicle(pairedt-test,mean±SEM,p=0.0194).qThelengthofthelongest episodeofwakefulnessinthesametimeperiodwasalsosignificantlyreduced(pairedt-test,mean±SEM,p=0.0152).rPowerspectrum(bandmean± SEM)withasignificantsuppressionofwakinghighthetaactivity(2-wayrepeatedmeasuresANOVA,Bonferroni***p<0.0001)andgammaactivity(*p= 0.0472),andsanunexpectedincreaseinlowthetapowerduringREMsleep(***p<0.0001),withtnochangeintheNREMspectrum NATURECOMMUNICATIONS|8: 1405 |DOI:10.1038/s41467-017-01004-6|www.nature.com/naturecommunications 7 ARTICLE NATURECOMMUNICATIONS|DOI:10.1038/s41467-017-01004-6 a b PHA PHA EW fr fr IF SuMp pps SuMg SuMp SuMg SuMp pps SuMs SuMs SuMg SuMg SuMp SuMs SuMs SuMg SuMg dPMA dPMA vPMA vPMA LM MM LM LM MM LM LM MM LM Arc Arc TMA c lox2722 YFP ChR2 d ITRhSyn loxP WPREhGH-pA vgat/vglut2 Flash vglut2 light-source 470 nm A A Patch A A V V electrode CA2 DG vglut2-IRES-Cre or vgat-IRES-Cre lox2722 WPRE ITR hSyn ChR2 YFP loxP hGH-pA e f g Control 50 Bicuculine DNQX 10 pA Bicuc 300 ms Vh = +20 mV s) m h ( dt Vh = 0 mV DNQX wi alf- H Vh = –60 mV Bicuc + DNQX 5 pA 0 100 ms Glut GABA h i j +15 8 s) m –80 mV elay ( 4 d +20 et s n O 3 pA 0 –15 pA 30 ms Glut GABA photostimulation of terminals again evoking both GABA-A- neurons adjacent to the mammillothalamic tract19. The mean mediated and AMPA-mediated synaptic responses in 90% of amplitudeandmeanonsetdelayofthephoto-evokedglutamate- dentate granule cells, with AMPA responses alone in 10% of mediated currents were –5.36±1.06pA and 4.55±0.25ms (Vh granule cells (n=9). This is consistent with previous studies =−70mV)andfortheGABA-A-mediatedcurrentswere+7.11± indicating that the SuM hippocampal projection arises from the 0.89pA and 5.84±0.25ms (Vh=0mV; n=8), again consistent 8 NATURECOMMUNICATIONS|8: 1405 |DOI:10.1038/s41467-017-01004-6|www.nature.com/naturecommunications ARTICLE NATURECOMMUNICATIONS|DOI:10.1038/s41467-017-01004-6 withaphoticstimulation-evokedmonosynapticresponse(Fig.5j). Nos1-IRES-Cremice(Fig.8a),withtransductionintheSuMNos1/ Together,theseresultssuggestthattheSuMinputtothedentate Vglut2cellularfield(Fig.8b).Therewasnosignificantdifferencein gyrus predominantly originates from neurons that express both wake, NREM and REM sleep duration when mice were injected vesicular transporters for glutamate and GABA (SuMvgat/vglut2) with saline (vehicle) or CNO (0.3 or 0.9mg/kg) (Fig. 8c−h), no andthattheseSuMvgat/vglut2neuronsco-releasebothtransmitters overall increase in sleep over the first 6h (p=0.31, two-way at their synaptic terminals. ANOVA for dose, n=4, F (2, 6)=1.423), nor fragmentation of We next examined whether SuMvgat/vglut2 neurons, represent- sleep-wakeasmeasuredbythenumberofwakeboutsoverthe6h ing about 15% of SuM neurons (see below), were wake- after injection (p=0.7588, n=4, t=0.3363, df=3) nor changes promoting, by activating these neurons in vgat-IRES-Cre mice inthedurationoflongestwakeinthisperiod(saline26bouts± after transduction with AAV-hM3Dq (Fig. 6a). Injection sites 5.1versus CNO0.9mg/kg 23.75bouts±2.7,p=0.6028, n=4, t covering the region of vgat expression in the SuM had a modest =0.5797, df=3), which had all been noted after chemogenetic but significant effect on the duration of wakefulness following inhibition of all SuM glutamate neurons (Fig. 4). However, CNO injection (saline 14±3.6 (SE)min, versus CNO 92± SuMNos1/Vglut2 neurons by our estimate account for no more 10min, paired t-test, n=4, p=0.0027; Fig. 6c−g). Thus, than 20% of the SuMVglut2 population, so inhibition of such a activation of these neurons was significantly less wake- small percentage of SuMVglut2 neurons may not have been suf- promoting than activation of the entire population of SuMvglut2 ficient to affect overall wake-sleep. On the other hand, power- neurons (duration of wakefulness, CNO in vgat-Cre mice 92± spectral analysis revealed a significant decrease in high theta 10min, vglut2-Cre mice 310±40min, n=4, 6, p=0.0028). power (7–13Hz) during REM sleep (Fig. 8i) in these mice when given CNO, but without changes in spectra during wakefulness SuMNos1/Vglut2 neurons drive wake. As previously described in (p=0.1744,two-wayANOVA,salineversusCNO0.9mg/kg,n= cats and rats40,41, the mouse SuM contains a subpopulation of 4,F(1,18)=1.999;Bonferoniposthoccomparisonsbybandp> 0.9999 except for gamma p=0.2335), suggesting a particularly Nos1-containingneurons (Fig. 7a,b; see alsoAllen Mouse Brain Atlas37). Given that cortical Nos1-expressing neurons have been importantroleforSuMNos1/Vglut2neuronsinnormal REMtheta activity (Fig. 8l). previously implicated in behavioral state control, albeit in the regulationofslow-wave-sleep47,wesoughttofurtherunderstand theanatomicdistribution,cellularco-localizationandphysiology Discussion of SuMNos1 neurons. We first examined Nos1 immunoreactivity Prior studies in humans and animals have indicated that a key, in Vglut2-Cre-L10 reporter mice, and found Nos1 immunor- but hitherto uncharacterized, node of the arousal system resides eactiveneuronstobescatteredthroughouttheSuM,representing in the caudal hypothalamus1–16. Prior neuroanatomical tracing about15–20%ofvglut2-positiveneuronsatallrostrocaudallevels andlesionstudieswereconsistentwiththeSuMplayingarolein (Fig. 7a, b). We then used Nos1-IRES-Cre mice and stereotaxic arousal, but the results of the present study for the first time microinjection of AAV-hM3Dq into the SuM (Fig. 7c) to selec- demonstratethattheSuMisamajorsourceofascendingarousal tively activate these neurons with CNO (see cFos in transduced tone.Furthermore,wedemonstratethatreleaseofglutamatefrom neurons, Fig. 7d). Activation of these neurons potently drove SuMneuronsisnecessaryfornormalamountsofwakeandEEG sustainedwakefulness(latencytosleepaftersaline33±4(SEM) spectral activity in the high-theta and gamma bands and for minversusCNO215±27min,n=7,p=0.0007,Fig.7e,f),with consolidation of sleep−wake; and that acute inhibition of SuMv- post hoc comparisons (Bonferroni corrected) showing a sig- glut2 neurons results in an increase in NREM sleep and frag- nificantlyelevatedwakefulnessforfourhoursafterCNOinjection mentation of wakefulness. (Fig. 7i). While activation of SuMNos1/Vglut2 neurons resulted in With acute inhibition in mice, we also found no changes in lesserwakefulnessthanactivationofallSuMvglut2neuronsinthe REM sleep amounts, although we did note an increase in low cases with large injection sites (Fig. 1), the duration of wakeful- frequency REM theta activity when all SuM neurons were ness was more than twice as long as after activating a similarly inhibited. SuMNos1/Vglut2 neurons were capable of promoting numerous population of SuMvgat/Vglut2 neurons, and similar to wakefulness when activated, but inhibition of these neurons did the amount of wakefulness driven by activation of SuMvglut2 notaltersleep-wakeorthewakingpowerspectra,althoughitdid neurons with smaller injection sites constrained within the SuM result in decreased theta activity in REM sleep. While chronic (see above). cell-specificlesionsoftheSuMinratshavebeenreportedtohave littleeffect onsleep-wake35,48 including REM amounts, a recent SuMNos1/Vglut2 neurons contribute to REM theta. To examine paper by Renouard et al.35, reported that SuM lesions reduced the necessity of SuMNos1/Vglut2 neurons for normal wakefulness, 7–10Hz power in the EEG (high theta in the rat) during REM we made stereotaxic injections of AAV-hM4Di into the SuM of sleep, and that dentate gyrus cFos expression during REM sleep Fig.5SuMGABA/glutamateco-releasingneurons.aInsituhybridizationforvglut2(magenta)inavgat-IRES-Cre/L10reportermouse(green,see Methodssection)revealsvglut2/vgatco-expression(white).bTheseneuronsconstituteasubpopulationofthelargerneuronsofthegrandicellularSuM (partiallyopaque).cCartoonshowingAAVx-FLEX-ChR2-mCherryrecombininginthepresenceofCreineithervglut2-Creorvgat-Cremice.dCartoonof optogeneticallyevokedstimulationofSuMterminalsondentategyrusgrandulecells,superimposedonahistologicalimageshowingdenseanterograde labeling(AAV-FLEX-GFPinvgat-Cremouse)ofSuMterminalsinthedentategyrussupragranularlayerandCA2(darkbrown)withalightNissl counterstain(faintblue−purple;scalebar500µm).eRepresentativebrain-slicevoltage-clamprecordingadentategyrusgranulecellduring photostimulationofSuMvgatterminalsexpressingChR2-YFPinvgat-IRES-Cremice(Vh=−70mV;KCl-basedpipettesolution;three5mslightpulses: bluebars).Photo-evokedpostsynapticcurrents(averageof30trials)recordedincontrolACSF(Con),inbicuculline-methiodide(Bic,10µM),inDNQX (20nM)andDNQX+Bic.fHalf-widthofGABA-A-mediatedandAMPA-mediatedphoto-evokedpostsynapticcurrents(30trials,3neurons;mean±SEM inred;p<0.001unpairedt-test).gPhoto-evoked(SuMVgat→DG)GABA-A-mediated(left)andAMPA-mediated(right)postsynapticcurrentsrecorded atdifferentmembranepotential(Cs-basedpipettesolutions)andh.I−VcurvesoftheGABAA-(blackcircle)andAMPA-(whitecircle)postsynaptic currents(mean±SEM,n=4).iGABAA-(toptrace)andAMPA-(bottomtrace)postsynapticcurrentsevokedbyphotostimulationofSuMVglut2→DG inputinvglut2-IRES-Cremice.jOnsetdelayofGABAA-andAMPA-mediatedpostsynapticcurrents(blackcircle:vgat-IRES-Cremiceandwhitecircle: vglut2-IRES-Cremice;mean±SEMinred;p=0.002pairedt-test) NATURECOMMUNICATIONS|8: 1405 |DOI:10.1038/s41467-017-01004-6|www.nature.com/naturecommunications 9 ARTICLE NATURECOMMUNICATIONS|DOI:10.1038/s41467-017-01004-6 was reduced. Based on our data, the loss of SuMNos1/Vglut2 neu- promotes wakefulness, but destruction of these neurons only ronsmayhavebeenresponsiblefordecreasedREMthetapower. prevents wake maintenance in a novel environment, without While the effect of inhibition may appear modest, it was changing baseline sleep−wake51. actually striking in relation to the effect of acute inhibition of The role of grandicellular SuMvgat/vglut2 neurons remains to be other components of the arousal system with increased NREM determined. They are apparently less involved than the SuMNos1 sleepandfragmentationofwakefulness.Inhibitionordestruction cells in the control of wakefulness, but given the extraordinary of other accepted components of the ascending arousal system density of SuMvgat/vlgut2 terminals on the proximal dendrites of such as the locus coeruleus, tuberomammillary histamine neu- dentate granule cells42–53 in the supragranular layer, an effect on rons, midline thalamus, or lateral hypothalamic orexin neurons temporal precision of “dentate gating” and perhaps mnemonic haslittleornoeffectofsleepwake18,21,35,48–51whilestimulation processes appears likely. Furthermore, they may play a role in the in some of these studies can show potent arousal responses52. plasticchangesinthedentategyrusthatoccurafterstatusepilepticus While it is tempting to conclude that the ascending arousal net- andduringepileptogenesis,astheSuMterminalssproutandtakeon work is highly redundant, it may instead indicate a specificity of an aberrant distribution in the dentate gyrus in a rat pilocarpine functionforindividualcomponentsofthearousalnetworkandit modelofpost-status epilepticus medialtemporallobeepilepsy53. doesnotmeanthatthesenodesarenotplayinganimportantrole Based on the anatomical connectivity of the SuM, as well as in intact animals. For example, locus ceruleus acute activation these physiological effects, our findings strongly suggest that the a lox2722 mCherry hM3Dq b NPM126 ITR hSyn loxP WPRE hGH-pA vgat/vglut2 A A A A V V SuM vgat-IRES-Cre lox2722 WPRE awake 1.3 h ITR hSyn hM3Dq mCherry loxP hGH-pA c d Saline CNO NPM126 Lights off NPM126 Lights off Wake Wake NREM NREM REM REM NPM128 NPM128 0 (10 AM) 6 h 18 24 0 (10 AM) 6 h 18 24 e 100 Hz NPM126 saline 18 h g * * 100 * CNO FFT * NaCl 0 Hz ess n f eful 50 100 Hz NPM126 CNO 18 h Wak % FFT 0 0 Hz 6 12 18 Hours from injection Fig.6ActivationofSuMvgatneuronsresultsinbriefwakefulness.aCartoonshowingAAVx-FLEX-hM3Dq-mCherryrecombininginvgatneuronsthat expressCre.bAnexampleinjectionsiteshowinghM3Dqtransduction(redoutline)intheregioncontainingGABA-glutamateneuronsintheSuMg(lower left—durationofwakefulnessafterCNOinthismouse,scalebar500µm).c,dHypnogramsshowingsleep−wakeintwomicewithsaline,thenCNO,witha transientincreaseinwakefulness.e,fExampleCSA’scorrespondingtotheperiodsshownbytheblueandredboxesincandd,respectively(0−100Hz, blackbarshowsthedarkperiod,shortwhitebarsrepresent20minandarespaced1hapart) 10 NATURECOMMUNICATIONS|8: 1405 |DOI:10.1038/s41467-017-01004-6|www.nature.com/naturecommunications
Description: