elevated gene expression of most microglial markers, and ......2015/11/21 ·...
TRANSCRIPT
Elevatedgeneexpressionofmostmicroglialmarkers,andreducedexpressionofmost pyramidal neuron and interneuronmarkers, in postmortem autismcortexRebecaBorges-Monroy,MSc†,°,ChrisP.Ponting,DPhil†,andT.GrantBelgard,DPhil††MRCFunctionalGenomicsUnit,DepartmentofPhysiology,AnatomyandGenetics,UniversityofOxford°Presentaffiliation:ProgramforBioinformaticsandIntegrativeGenomics,GraduateSchoolofArtsandScience,DivisionofMedicalSciences,HarvardUniversityRunningtitle:AlteredcelltypegeneexpressioninASDPleaseaddresscorrespondenceto:T.GrantBelgardorChrisP.PontingMRCFunctionalGenomicsUnitDepartmentofPhysiology,Anatomy&GeneticsUniversityofOxfordSouthParksRoadOxfordOX13PTUnitedKingdomTelephone:+44(0)1865282690Fax:+44(0)1865285862Email:[email protected]@dpag.ox.ac.uk
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
AbstractAutism Spectrum Disorders (ASD) are clinically and genetically heterogeneous.Nevertheless, a characteristic gene expression signature in postmortem cortex issharedacrossmostASDcases.Knowingwhetherthissignaturereflectschangesincells of particular types would help to determine the molecular, cellular, andanatomical etiology ofASD. To investigatewe took advantage of cell typemarkergenesdefinedbyrecentsinglecelltranscriptomesequencingfrommouseorhumancortex. We find that microglial markers showed significantly and substantiallyhighermedianexpressioninpostmortemASDthanincontrolcortexforbothmouseandhumanmarkers(21%higher fororthologousmarkersdefined inmouse;93%higherforhumanmarkers).Incontrast,neuronalmarkersshowedreducedmedianexpression: 12%and 16% lower formouse interneurons and pyramidal neurons,40% lower for human neurons, and 36% lower for a class of human excitatoryprojectionneurons.CelltypedensityalterationsinASDbrainsareindicatedbecausedistributions of these cell type markers are shifted concordantly.Importantly,orthologousmouse neuronalmarker genes encoding proteins localized to diverseneuronalcompartmentsallshowedreducedmedianexpressioninASD.Ourresultsprovide a framework for revealing the basis of transcriptomic differences inASD.Weproposecomparingcelltypedensityalterationsinpost-mortemtissuewithandwithout these distinctive gene expression changes to reconcile results fromstereologicalandtranscriptomicstudies.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
IntroductionAutism Spectrum Disorders (ASD) are heterogeneous and highly heritableneurodevelopmental disorders characterized by verbal and nonverbalcommunicationdeficits,impairedsocialinteractions,andrepetitivebehaviorsfromearlychildhood1-3.Despiteitsstrongheritability,geneticsusceptibilitytoASDinthepopulation is distributed over a thousand genes4, which complicates efforts togeneralizefromdeepfunctionalstudiesofrarebuthighlypenetrantvariants.Giventhispanoplyofgeneticcauses,andawidevarietyofproposedcausalmechanismsfor ASD, it was unexpected thatmost post-mortem brains from people with ASDexhibit similar gene expression profiles5. By studying this characteristic geneexpression signaturewe seek to reveal neurological features that are common tomuchoftheASDspectrum.Weareonlybeginningtounderstandthismolecularsignature.Alternativesplicingandtheattenuationofregionalgeneexpressiondifferences incortexappear tobekey components5,6.Nevertheless, therearealsoearly signs that largerdifferencesexistwithregardtothedensityofcertaincell typesorsubcellularcomponents. Inparticular, a set of co-expressed genes underexpressed in ASD is enriched insynaptic genes5, ASD risk genes5, 7 andmarkers of glutamatergic neurons5, 7, andcontainssomeinterneuronmarkers5.Asetofco-expressedgenesoverexpressedinASD is enriched for immune-related genes, markers for activated microglia, andmarkers for astrocytes5. Another study found an overrepresentation of genesoverexpressed in ASD in immune cell types and astrocytes and anoverrepresentation of genes underexpressed in ASD in neuronal cell types,particularlyininterneurons8.A parsimonious neurological interpretation of cell type marker enrichment,nevertheless, remains unclear. Such enrichments might be driven by a subset ofmarkers of that type and could therefore reflect changes in cell signaling,physiologicalstate, functionalpathways,structure,andmore.Forexample,defectsin synapse formation in ASD9,10may underlie the underexpression of a subset ofneuronal marker genes expressed at the synapse. On the other hand, if mostneuronalmarkersareunderexpressedinASD,thiscouldalsosuggestareductioninneuronaldensity.Hereweusedensitytorefertocellnumberdensity(thenumberof cells in a volume of tissue)11 or other ratios such as volume density (the totalvolume of cells in a volume of tissue)12, 13 or cross-sectional area stained with amarker (e.g.Nissl-stainedarea tounstainedarea ina section)14. Indeed, therearescatteredandsometimesconflictingreportsofchanges inneuronaldensity,whichmay reflect small sample size and the heterogeneity of ASD14-19. Identifying thesubsetofpost-mortembrainsinwhichthesechangesareexpectedtooccurandthecelltypesinvolvedinthesealterationscouldguidefutureconfirmatorystereologicalstudiesinamoretargetedmanner.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
Previousstudieshaveusedabiasedsubsetofcelltypemarkergenesfromabiasedsubset of cell types isolated with pre-selected markers followed by bulk geneexpressionprofiling8,20.However,clusteringcellsbytypewithoutpriorinformationabout their markers is now feasible with single cell transcriptomics21. Objective,comprehensive,andselectivecelltypemarkerscanbedefinedfromtheseclusters22.Inthisstudy,weexaminehoworthologsofmousecelltypemarkers23areshiftedintheirexpressioninpostmortemASDbrains.Furthermore,weobtainsimilarfindingswithhumancelltypemarkersfromasecondsinglecelltranscriptomestudy6.MaterialsandMethodsDatasetsandGeneListsExpressionDataWe used normalized, regressed, log2 transformed post-mortem gene expressionvaluesfromthesuperiortemporalgyrusof12ASDcasesand12matchedcontrols,rangingfrom15to60yearsoldandofbothsexes6. Thecasesandcontrolsinthatstudy were 57% of the individuals chosen from a larger unpublished RNAsequencingcohortonthebasisthattheydisplaythestrongestofthischaracteristicgeneexpressionsignature.CellTypeMarkerGenesMousecelltypemarkergenesweretakenfromTableS1ofZeiseletal.23.TheywerebasedonRNAsequencingofsinglecellsisolatedonaFluidigmC1AutoPrepSystemfrom the somatosensory cortex andCA1ofhippocampus frommiceof both sexesbetween postnatal days 21 and 3123. Human cell type marker genes, as well asmarkergenesofneuronal communities,were taken fromTableS3ofDarmanisetal.24 andwere based onRNA sequencing of single cells isolated on a FluidigmC1from freshly resected temporal cortex of several humans of both sexes aged 21through 63 years24. The numbers of these marker genes are provided inSupplementaryTable1.LocalizationofNeuronalMarkerGenesWeusedS1pyramidalmarker genes froma single cell transcriptomic study frommousebraintissue23having1:1mousetohumanorthologsasdefinedinEnsembl.ThehumanEnsemblidentifiersforthesetranscriptsgavelinkedsubcellularproteinlocalization information for their encoded proteins from the UniProt database25.MarkerswereseparatedintogroupsbasedonwhethertheycontainedthefollowingGene Ontology Cellular Component identifiers: "GO:0005634" for nuclear proteinlocalization,"GO:0030425"forproteinlocalizationtodendrites,and"GO:0030424"forproteinlocalizationtoaxons26.Mosttranscriptsdonothaveonlyonelocalizationidentifier. Since there was a considerable overlap between transcripts with theaxonal localization identifier and the dendritic localization identifier, these were
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
analyzed together as a single group. To improve specificity, transcriptswith bothnuclear localization identifiers and axonal or dendritic identifiers were removedfrom the analysis.We analyzed themedian log2(expression fold change) of thesedifferent subsets of mouse orthologous S1 pyramidal neuron and interneuronmarker genes as described below. Caveats of thismethod include that it assumesthat a change in density of neuronal processes (and only processes) would bereflected in the expression levels of transcripts whose proteins are localized toneuronal processes but would not be reflected to a comparable extent in theexpression levels of transcriptswhoseproteins are localized to the cell soma27,28,that there isconservationofprotein localization for thesemarkers frommouse tohuman, and that the ontology’s cellular localization evidencemay require furthervalidation26.ExaminingCellTypeGeneExpressionChangesinASDFor themousecell typemarkers,wecross-referenced themousegenesymbols toEnsemblMouseGeneidentifiers,andthentookthe1:1mouse-humanorthologsthatwerealsoincludedintheASDsequencingstudy6usingEnsemblBioMart29.Forthehumancell typemarkers,wekeptthosemarkersthatwereintheASDsequencingstudy.Weplottedthedistributionsoftheirlog2foldchangeswithR.Wetestedthetwo-tailed significance of the median log2 fold change for each cell type byconstructing a null distribution by permuting case-control labels 1000 times andthen subtracting the mean log2(FPKM) in controls from the mean log2(FPKM) incases.Thesewerecorrectedformultipletestingusingthe“p.adjust”functioninR30,applyingtheBenjaminiandHochbergFalseDiscoveryRate(FDR)method31withanFDRcut-offof5%.Confidenceintervalswerecalculatedfrom1000bootstraps.ResultsWefirsttestedanullhypothesisthatthemedianexpressionfoldchangeforeachofeighthigh-levelcorticalcelltypesdefinedinmouse23wasnotsignificantlydifferentfromthechangeobservedwhencase-control labelsarerandomlyshuffled(Figure1). To do so we considered first a set of marker genes defined by single cellsequencinginmousecortex,beforethenobtainingsimilarresultsformarkergenesofsixhigh-levelcelltypesdefinedinhumancortex(Figure2;SupplementaryTables2-3).Comparedtocontrols,themedianexpressionofmouse-definedgenemarkersforS1pyramidalneuronswassignificantly lower(by10-21%atanFDR<5%31) inASD post-mortem human cortex tissue; similarly, median expression of mouse-defined interneuron markers was reduced in ASD brains by 8-16%. By contrast,median expression of mouse-defined microglia markers was increased in ASDbrainsby15-28%.Ependymal,endothelial,andmuralmarkergenesalsoexhibitedslightlyincreasedexpression(ataslightlyrelaxedFalseDiscoveryRate31of5.1%).Similar statistically significant results were observed, with even stronger effects,using the smaller set of cell type gene markers obtained from single celltranscriptomics inhumancortex24 (Figure2;SupplementaryTable3).Humancell
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
typemarkermedian gene expression for neuronswas decreased in post-mortemhuman ASD cortex tissue by 30-49%. Conversely, microglia marker medianexpression was increased in ASD by 48-109%, along with oligodendrocyteprecursorcell(OPC)markergenes,whosemedianexpressionwasincreasedinASDby8-22%.Wethenrepeatedthisanalysisforfourpreviouslydefinedexcitatoryorinhibitoryneuronalsubpopulations inhuman24.Genemarkersof subpopulation4,representinga typeofexcitatoryneuron,weredecreased inmedianexpression inASDbrainsby22-40%(Figure2C,SupplementaryTable4).HavingfoundthatamajorityofgenemarkersofseveralcorticalcelltypeschangeinexpressioninASD,wenextaskedwhethertheseobservationsweremoreconsistentwithalterationsinaspecificpathwayorcellstate,orelsewithlargerchangesincelltype density. We would expect that alterations in specific pathways exclusivelywould produce expression changes amongst a minority subset of genes whereexpressionofsomemarkers(thosecontributingtothepathwayorcellstate)changewhileothersdonot.Instead,wefoundthatexpressionofthesemarkergenesshiftedconcordantly(Figure3),suggestingcelltypedensitychangesinASD.Expression changes of neuronalmarker genes could reflect alterations not in cellnumberdensity,but in thedensityof subcellularcompartmentssuchas synapses.Toconsider this,wenextexaminedwhetherproteinproductsofneuronalmarkergenes that are localized in different neuronal compartments alter their mRNAs’expressioninASDcoordinately.S1pyramidalandinterneuronmousemarkerswereclassified based onwhether their protein products are localized in the nucleus ofneuronsor indendritesandaxons (Figure4A,SupplementaryTable5).MouseS1pyramidal genemarkerswhoseproteinproducts are localized in thenucleus, andaxonsanddendrites,aresimilarlyandsignificantlylowerintranscriptexpressioninASD brains by 9-32% and 6-32%, respectively. The corresponding changes formouse interneuronmarkerswere 2-18% and 6-24%, respectively. Approximatelythree-quartersofgenes foreachof these localizationcategoriesshowadecreasedexpressionfoldchange(Figure4B).Thesefindingssuggestthatreducedexpressionofneuronalmarkergenesreflectsdecreasedcelldensity.DiscussionIn this study, the majority of neuronal and microglial cell marker genes exhibitcoordinated expression level change in ASD brains. Previous enrichments ofmicroglial markers in overexpressed-in-ASD ‘modules’ of co-expressed genes andenrichments of neuronal markers in underexpressed-in-ASD ‘modules’ of co-expressedgenes5wereconsistentwithalteredcellularprocessesbesideschangesincelldensity.Forexample, increasedmicroglialactivation,observed in livingyoungmenwithASD32,might result in overexpressionofM2microglial activation genesevenintheabsenceofachangeinmicrogliadensity33.Furthermore,theenrichmentof synaptic marker genes in under-expressed modules in ASD5 may be due todysfunctional synaptic formation9, 10. Indeed, gene markers related toneurotransmission are underexpressed in adult ASD post-mortem cortex34.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
Nevertheless,thatstudyfailedtofinddifferencesinexpressionofneuronalandglialmarkergenesinASD,probablyduetothefewmarkersused.Althoughthereissomeoverlap between the cell type marker genes used here and microglia activationgenes33 or neurotransmission genes34, we propose that the parsimoniousinterpretationoftheseresultsisthatASDbrainscontainalterationsincelldensity.Wedosoforthefollowingreasons:(1)markergenesshifttheirexpressiontogetherin the same direction in the ASD versus control comparison, rather than beingdrivenbyaminoritysubsetasexpectedwhenonlyaspecificpathwayisaltered,and(2)themedianfoldchangesfortranscriptsencodingproteinslocalizedindifferentneuronal compartments are all reduced, which is unexpected if only neuronalprocesseschangedintheirdensity.For microglia, the issues of cell state and density may be correlated, becausemicroglial activation can trigger proliferation35,36. Microglial activation, increaseddensity,andincreasedsomalvolumeinASDcortexhaveallbeenreported37,38.Therole of microglial activation and proliferation in ASD is unclear but increasedmicroglia-neuronal spatial clustering in the dorsolateral prefrontal cortex in ASDsuggeststhatactivemicrogliaarerecruitedtoneurons39.Decreases in neuronal density may not necessarily be caused by a reduction inrelative cell number (i.e. the proportion of cell type nuclei versus all nuclei).WecannotexcludethepossibilitythatthesecelltypemarkergeneexpressionchangesinASDmaybeduetodifferencesincellsize,becausegeneexpressionofindividualtranscripts increases inproportion to cell volume40,41.Hence, decreasedneuronaldensitymaybeexplainedbylesscytoplasmicspacebeingtakenupbyneuronsasaproportionoftotalcytoplasmicspace.Studies analyzing neuronal composition and density in ASD usingimmunohistochemistryandstereologyhavebeenconflicting,perhapsduetosmallsample sizes and the disorders’ heterogeneity. Patches of cortical disorganizationandasmallincreaseinneuronaldensityhavebeenreportedinchildrenwithASD16.In the fusiform gyrus, reduced neuronal density, neuronal numbers, and cellvolumes were reported in ASD, but these changes were not detected in wholecortical grey matter or primary visual cortex42. In the dorsolateral and mesialprefrontal cortex of children, an increase in neuronal counts but not in neuronalvolumewasobserved43.However,inBrodmannareas44and45,smallerpyramidalneurons, albeit similar pyramidal neuronal numbers, were reported in ASD44. Aninteresting possibility is that the underlying structure of neurons in the cortex isdifferent in ASD, resulting in decreased density in the cortical section sequenced.Increased center-to-center minicolumn width has been observed in differentcorticalareasinASD,includingatrendforincreasedwidthintheprimaryauditorycortex(Brodmann41,overlappingtheareausedinthisstudy)19.WiderminicolumnspacinginASDmayexplaindecreasedexpressionofneuronalmarkers,asitwouldimplyalowerdensityofneurons.These results are consistentwith previous transcriptomic studies, and expand on
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
them.Thesehave shownan enrichment of interneuron and glutamatergic neuronmarkers in anASDunderexpressed genemodule in the cortex and enrichment ofactivatedmicrogliaandastrocytemarkersinanASDoverexpressedmodule5,7.Laterworkreportedanover-representationofoverexpressedgenesinimmunecellsandastrocytesandanover-representationofunderexpressedgenesininterneuronsandputative corticalprojectionneurons in theVoineaguet al.dataset8.Another studyalso observed overexpression of microglia and astrocyte markers andunderexpression of neuronal markers in the prefrontal cortex, but used fewmarkers45.Here,weusedobjectivelydefinedgenemarkers,allowingustopinpointexcitatoryneuronalmarkersasbeingspecificallydownregulated.Wealsoincludednon-neuronalandglialcelltypes,whichpreviousstudiesdidnotinclude.Insteadofanalyzing the enrichment of cell type markers in underexpressed andoverexpressedgenesinASD,wetestedwhetherthesemarkerschangeexpressionasagroup.Forthefirsttime,toourknowledge,thismethodallowsusmakeinferencesaboutthecompositionofcelltypesintheASDbrain.
Thetranscriptomicdatausedinthisstudywasobtainedfromthesuperiortemporalgyrus.Since thisbrain region is important forauditoryprocessingandspeech46-48andhasbeenimplicatedinASD49,50,itisimportanttodeterminehowandifcelltypedensity changes here lead to behavioral deficits. Although these gene expressionpatternsmay be regionally specific, there is strong overlap betweendifferentiallyexpressedgenesinthetemporalandfrontalcortex,suggestingthatgeneexpressionchangesinthecortexmightbeglobalandnotrepresentativeofaparticularregion5.In this study, we used a cell type transcriptomic approach to infer cellularcompositiondifferencesinASDcortex.Themediangeneexpressionofinhibitoryorexcitatoryneuronalmarkergeneswas significantly reduced inASD.Thiswasalsothe case when subdividing mouse orthologous neuronal markers based on thesubcellularlocalizationoftheproteinstheyencode.Themediangeneexpressionofmicroglia markers was significantly increased in ASD. These gene expressionpatterns suggest changes in cell type number and/or volume density. Given theconflicting results in published immunohistochemistry and stereology studies,understandingtheunderlyingcausesoftheseglobalgeneexpressionsignatures inASD should be prioritized. Groups with expertise in stereology might nowindependently validate this inference frommarkers to cell typedensity. Since thepatients used in this analysis were a subset displaying strong characteristicdifferentialgeneexpressionchanges inASD, future integratedtranscriptomicsandstereologicalstudiescouldcomparedensitychangesinpost-mortemtissuewithandwithoutthesedistinctivegeneexpressionchanges.Moreover,thisstudyprovidesaproofofconceptforusingsinglecelltranscriptomicanalysesderivedmarkergenestostudycell typecompositionchanges indisease,andcouldbeeasilyextendedtothe study of other diseases such as schizophrenia, where cell type markerenrichmenthasbeenreportedingeneco-expressionmodules51.Acknowledgements
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
RBMwassupportedbytheClarendonFund;RBM,TGB,andCPPweresupportedbytheMedicalResearchCouncil;andTGBwassupportedbytheHumanBrainProject.TheresearchleadingtotheseresultshasreceivedfundingfromtheEuropeanUnionSeventh Framework Programme (FP7/2007-2013) under grant agreement no.604102(HumanBrainProject).ConflictofInterestTheauthorsdeclarenofinancialconflictsofinterest.References1. AmericanPsychiatricAssociation.Diagnosticandstatisticalmanualofmental
disorders:DSM-5.Fifthedition.edn.AmericanPsychiatricPublishing:Washington,DC,2013,xliv,947pagespp.
2. HallmayerJ,ClevelandS,TorresA,PhillipsJ,CohenB,TorigoeTetal.Genetic
heritabilityandsharedenvironmentalfactorsamongtwinpairswithautism.ArchGenPsychiatry2011;68(11):1095-1102.
3. RosenbergRE,LawJK,YenokyanG,McGreadyJ,KaufmannWE,LawPA.
Characteristicsandconcordanceofautismspectrumdisordersamong277twinpairs.ArchPediatrAdolescMed2009;163(10):907-914.
4. RonemusM,IossifovI,LevyD,WiglerM.Theroleofdenovomutationsinthe
geneticsofautismspectrumdisorders.NatRevGenet2014;15(2):133-141.5. VoineaguI,WangX,JohnstonP,LoweJK,TianY,HorvathSetal.
Transcriptomicanalysisofautisticbrainrevealsconvergentmolecularpathology.Nature2011;474(7351):380-384.
6. IrimiaM,WeatherittRJ,EllisJD,ParikshakNN,Gonatopoulos-PournatzisT,
BaborMetal.Ahighlyconservedprogramofneuronalmicroexonsismisregulatedinautisticbrains.Cell2014;159(7):1511-1523.
7. ParikshakNN,LuoR,ZhangA,WonH,LoweJK,ChandranVetal.Integrative
functionalgenomicanalysesimplicatespecificmolecularpathwaysandcircuitsinautism.Cell2013;155(5):1008-1021.
8. XuX,WellsAB,O'BrienDR,NehoraiA,DoughertyJD.Celltype-specific
expressionanalysistoidentifyputativecellularmechanismsforneurogeneticdisorders.TheJournalofneuroscience:theofficialjournaloftheSocietyforNeuroscience2014;34(4):1420-1431.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
9. RubensteinJL,MerzenichMM.Modelofautism:increasedratioofexcitation/inhibitioninkeyneuralsystems.GenesBrainBehav2003;2(5):255-267.
10. JamainS,QuachH,BetancurC,RastamM,ColineauxC,GillbergICetal.
MutationsoftheX-linkedgenesencodingneuroliginsNLGN3andNLGN4areassociatedwithautism.NatGenet2003;34(1):27-29.
11. KreczmanskiP,HeinsenH,MantuaV,WoltersdorfF,MassonT,UlfigNetal.
Volume,neurondensityandtotalneuronnumberinfivesubcorticalregionsinschizophrenia.Brain2007;130(Pt3):678-692.
12. RibeiroAA,DavisC,GabellaG.Estimateofsizeandtotalnumberofneurons
insuperiorcervicalganglionofrat,capybaraandhorse.AnatEmbryol(Berl)2004;208(5):367-380.
13. RamosR,RequenaV,DiazF,VillenaA,PerezdeVargasI.Evolutionof
neuronaldensityintheageingthalamicreticularnucleus.MechAgeingDev1995;83(1):21-29.
14. CasanovaMF,BuxhoevedenDP,SwitalaAE,RoyE.Neuronaldensityand
architecture(GrayLevelIndex)inthebrainsofautisticpatients.JChildNeurol2002;17(7):515-521.
15. LawrenceYA,KemperTL,BaumanML,BlattGJ.Parvalbumin-,calbindin-,and
calretinin-immunoreactivehippocampalinterneurondensityinautism.ActaNeurolScand2010;121(2):99-108.
16. StonerR,ChowML,BoyleMP,SunkinSM,MoutonPR,RoySetal.Patchesof
disorganizationintheneocortexofchildrenwithautism.NEnglJMed2014;370(13):1209-1219.
17. BaileyA,LuthertP,DeanA,HardingB,JanotaI,MontgomeryMetal.A
clinicopathologicalstudyofautism.Brain1998;121(Pt5):889-905.18. CasanovaMF,BuxhoevedenDP,SwitalaAE,RoyE.Minicolumnarpathology
inautism.Neurology2002;58(3):428-432.19. McKavanaghR,BuckleyE,ChanceSA.Widerminicolumnsinautism:aneural
basisforalteredprocessing?Brain2015;138(Pt7):2034-2045.20. DoyleJP,DoughertyJD,HeimanM,SchmidtEF,StevensTR,MaGetal.
ApplicationofatranslationalprofilingapproachforthecomparativeanalysisofCNScelltypes.Cell2008;135(4):749-762.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
21. SandbergR.Enteringtheeraofsingle-celltranscriptomicsinbiologyandmedicine.Naturemethods2014;11(1):22-24.
22. UsoskinD,FurlanA,IslamS,AbdoH,LonnerbergP,LouDetal.Unbiased
classificationofsensoryneurontypesbylarge-scalesingle-cellRNAsequencing.Natureneuroscience2015;18(1):145-153.
23. ZeiselA,Munoz-ManchadoAB,CodeluppiS,LonnerbergP,LaMannoG,
JureusAetal.Brainstructure.Celltypesinthemousecortexandhippocampusrevealedbysingle-cellRNA-seq.Science2015;347(6226):1138-1142.
24. DarmanisS,SloanSA,ZhangY,EngeM,CanedaC,ShuerLMetal.Asurveyof
humanbraintranscriptomediversityatthesinglecelllevel.ProceedingsoftheNationalAcademyofSciencesoftheUnitedStatesofAmerica2015;112(23):7285-7290.
25. UniProtC.UniProt:ahubforproteininformation.NucleicAcidsRes2015;
43(Databaseissue):D204-212.26. RoncagliaP,MartoneME,HillDP,BerardiniTZ,FoulgerRE,ImamFTetal.
TheGeneOntology(GO)CellularComponentOntology:integrationwithSAO(SubcellularAnatomyOntology)andotherrecentdevelopments.JBiomedSemantics2013;4(1):20.
27. MartinKC,ZukinRS.RNAtraffickingandlocalproteinsynthesisindendrites:
anoverview.TheJournalofneuroscience:theofficialjournaloftheSocietyforNeuroscience2006;26(27):7131-7134.
28. WillisDE,TwissJL.Regulationofproteinlevelsinsubcellulardomains
throughmRNAtransportandlocalizedtranslation.MolCellProteomics2010;9(5):952-962.
29. KinsellaRJ,KahariA,HaiderS,ZamoraJ,ProctorG,SpudichGetal.Ensembl
BioMarts:ahubfordataretrievalacrosstaxonomicspace.Database(Oxford)2011;2011:bar030.
30. RCoreTeam.R:Alanguageandenvironmentforstatisticalcomputing.R
FoundationforStatisticalComputing2014.31. BenjaminiY,HochbergY.ControllingtheFalseDiscoveryRate:APractical
andPowerfulApproachtoMultipleTesting.JournaloftheRoyalStatisticalSocietySeriesB(Methodological)1995;57(1):289-300.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
32. SuzukiK,SugiharaG,OuchiY,NakamuraK,FutatsubashiM,TakebayashiKetal.Microglialactivationinyoungadultswithautismspectrumdisorder.JAMAPsychiatry2013;70(1):49-58.
33. GuptaS,EllisSE,AsharFN,MoesA,BaderJS,ZhanJetal.Transcriptome
analysisrevealsdysregulationofinnateimmuneresponsegenesandneuronalactivity-dependentgenesinautism.Naturecommunications2014;5:5748.
34. vandeLagemaatLN,NijhofB,BoschDG,Kohansal-NodehiM,Keerthikumar
S,HeimelJA.Age-relateddecreasedinhibitoryvs.excitatorygeneexpressionintheadultautisticbrain.FrontNeurosci2014;8:394.
35. KettenmannH,HanischUK,NodaM,VerkhratskyA.Physiologyofmicroglia.
PhysiolRev2011;91(2):461-553.36. GraeberMB,TetzlaffW,StreitWJ,KreutzbergGW.Microglialcellsbutnot
astrocytesundergomitosisfollowingratfacialnerveaxotomy.Neuroscienceletters1988;85(3):317-321.
37. MorganJT,ChanaG,PardoCA,AchimC,SemendeferiK,BuckwalterJetal.
Microglialactivationandincreasedmicroglialdensityobservedinthedorsolateralprefrontalcortexinautism.BiolPsychiatry2010;68(4):368-376.
38. TetreaultNA,HakeemAY,JiangS,WilliamsBA,AllmanE,WoldBJetal.
Microgliainthecerebralcortexinautism.JAutismDevDisord2012;42(12):2569-2584.
39. MorganJT,ChanaG,AbramsonI,SemendeferiK,CourchesneE,EverallIP.
Abnormalmicroglial-neuronalspatialorganizationinthedorsolateralprefrontalcortexinautism.BrainRes2012;1456:72-81.
40. WuCY,RolfePA,GiffordDK,FinkGR.Controloftranscriptionbycellsize.
PLoSBiol2010;8(11):e1000523.41. ZhurinskyJ,LeonhardK,WattS,MargueratS,BahlerJ,NurseP.A
coordinatedglobalcontrolovercellulartranscription.CurrBiol2010;20(22):2010-2015.
42. vanKootenIA,PalmenSJ,vonCappelnP,SteinbuschHW,KorrH,HeinsenH
etal.Neuronsinthefusiformgyrusarefewerandsmallerinautism.Brain2008;131(Pt4):987-999.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
43. CourchesneE,MoutonPR,CalhounME,SemendeferiK,Ahrens-BarbeauC,HalletMJetal.Neuronnumberandsizeinprefrontalcortexofchildrenwithautism.JAMA2011;306(18):2001-2010.
44. Jacot-DescombesS,UppalN,WicinskiB,SantosM,SchmeidlerJ,
GiannakopoulosPetal.DecreasedpyramidalneuronsizeinBrodmannareas44and45inpatientswithautism.ActaNeuropathol2012;124(1):67-79.
45. EdmonsonC,ZiatsMN,RennertOM.Alteredglialmarkerexpressionin
autisticpost-mortemprefrontalcortexandcerebellum.MolAutism2014;5(1):3.
46. HillisAE,WitykRJ,TuffiashE,BeauchampNJ,JacobsMA,BarkerPBetal.
HypoperfusionofWernicke'sareapredictsseverityofsemanticdeficitinacutestroke.AnnNeurol2001;50(5):561-566.
47. MesgaraniN,CheungC,JohnsonK,ChangEF.Phoneticfeatureencodingin
humansuperiortemporalgyrus.Science2014;343(6174):1006-1010.48. PicklesJO.Anintroductiontothephysiologyofhearing.4thedn.Emerald:
Bingley,2012,xxiii,430p.,434p.ofplatespp.49. GendryMeresseI,ZilboviciusM,BoddaertN,RobelL,PhilippeA,SfaelloIet
al.Autismseverityandtemporallobefunctionalabnormalities.AnnNeurol2005;58(3):466-469.
50. BiglerED,MortensenS,NeeleyES,OzonoffS,KrasnyL,JohnsonMetal.
Superiortemporalgyrus,languagefunction,andautism.DevNeuropsychol2007;31(2):217-238.
51. TorkamaniA,DeanB,SchorkNJ,ThomasEA.Coexpressionnetworkanalysis
ofneuraltissuerevealsperturbationsindevelopmentalprocessesinschizophrenia.GenomeRes2010;20(4):403-412.
52. KrzywinskiM,AltmanN.Visualizingsampleswithboxplots.Naturemethods
2014;11(2):119-120.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
Figures
Figure 1. Schematicdescribing the analyses.A)Cell typemarkerswereobtainedfrommouse23andhuman24singlecellsequencingstudies.B)Themedianexpressionlog2(FoldChange[FC])inASDwascalculatedforallmarkergenesineachcelltype.Genes underexpressed in ASD are indicated in green. C) Case-control labelswerepermuted 1000 times to compute a null distribution of median log2(FC). D) Anempiricalp-valuefortheobservedmedianlog2(FC)wasobtainedfromitspositionin this null distribution. E) 95% Confidence intervals were obtained bybootstrappinglog2(FCs)ofeachcelltypeandobtainingamedian1000times.
Mouse and HumanCell Type Markers
(Zeisel et al. 2015)1:1 mouse to human
orthologs
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
Figure 2. Plots of medianexpression fold change for celltype marker genes withbootstrapped 95% confidenceintervals in the ASD versuscontrol comparison. (A) Fororthologous mouse markergenes. (B) For human markergenes. (C) For human neuronalcommunity marker genes.Community 4 is an excitatoryneuronal community andcommunities 3, 5, and 6 areinhibitory neuronalcommunities24. Significant p-values at an FDR<5% aremarked by an asterisk (*) andmarginally significant p-valuesatanFDR<5.1%aremarkedbyaplus symbol (+).We also foundconsistentsignificantdifferencesin analytical distributional testsbut not in permutation-baseddistributional tests (data notshown).Thiscouldbeduetothenon-independenceofexpressionofcelltypemarkers(i.e.celltypegene markers will tend to becorrelated in expression),ordinary interindividualvariation in the relativeabundance of these cell types,andtherelativelyunderpowerednature of a 12 versus 12permutationtest.
Cell Types
S1 Pyramidal
Interneuron
Microglia
Astrocytes
Oligodendrocytes
Ependymal
Endothelial
Mural
0.7 0.8 0.9 1 1.1 1.2 1.3 1.4Median Fold Change
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
Figure 3. Boxplots representing the expression log2(fold change) distributions of cell type markers in ASD. (A) Fororthologous mouse marker genes. (B) For human marker genes. (C) For human neuronal community marker genes.Community 4 is an excitatory neuronal communitywhile communities 3, 5, and 6 are inhibitory neuronal communities24.Community 3, n=5 genes; community 4, n=23 genes; community 5, n=13 genes; community 6, n=8 genes. Interpreting theshapesofboxplotsforsamplesizes<10shouldbeapproachedwithcaution52.Themiddlelinerepresentsthemedian,theleftline is the25thpercentile (Q1), the right line is the75thpercentile (Q3),whiskers represent theminimumandmaximumvaluesexcludingoutliers,andoutliersaredepictedasindividualpointsforvalueslessthanQ1-1.5*(Q3-Q1)andgreaterthanQ3+1.5*(Q3-Q1).Theasterisks (*) represent cell types significant at anFDR<5%and theplus symbols (+) represent thosesignificantatanFDR<5.1%.
� 2 � 1 0 1 2
Log2 Fold Change
Community 3Inhibitory
Community 4Excitatory
Community 5Inhibitory
Community 6Inhibitory
Log2 Fold ChangeHuman Neuronal Communities
Cell Type Marker Genes
Endothelial
OPCs
Oligodendrocytes
Astrocytes
Microglia
Neurons
�2 �1 0 1 2
Log2 Fold Change HumanCell Type Marker Genes
Log2 Fold Change
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
Figure4.(A)Plotsindicatingthe95%confidenceintervalsofmedianfoldchangeinASD for orthologous mouse S1 pyramidal and interneuron marker genes withdifferentproteinproductlocalizations.(B)Boxplotrepresentingthedistributionofexpression log2(fold changes) in ASD of orthologous mouse S1 pyramidal andinterneuron marker genes with different protein product localizations. AlldifferencesaresignificantatanFDR<5%.
S1 Pyramidal NeuronsNucleus
S1 Pyramidal NeuronsDendrites and Axons
Interneurons Nucleus
InterneuronsDendrites and Axons
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3
Neuronal Localization
Log2 Fold Change Mouse OrthologousNeuronal Cell Type Marker Genes with
Particular Protein Localizations
Median Fold Change of Gene Expressionfor Neuronal Genes with Particular Protein Localizations
InterneuronsDendritesand Axons
InterneuronsNucleus
S1 PyramidalNeuronsDendritesand Axons
S1 PyramidalNeuronsNucleus
�2 �1 0 1 2Log2 Fold ChangeMedian Fold Change
A) B)
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
SupplementaryinformationforElevatedgeneexpressionofmostmicroglialmarkersandreducedexpressionof most pyramidal neuron and interneuronmarkers in postmortem autismcortexRebecaBorges-Monroy,MSc†,°,ChrisP.Ponting,DPhil†,T.GrantBelgard,DPhil††MRCFunctionalGenomicsUnitattheUniversityofOxford°Presentaffiliation:ProgramforBioinformaticsandIntegrativeGenomics,GraduateSchoolofArtsandScience,DivisionofMedicalSciences,HarvardUniversityRunningtitle:AlteredcelltypegeneexpressioninASDPleaseaddresscorrespondenceto:T.GrantBelgardorChrisP.PontingMRCFunctionalGenomicsUnitDepartmentofPhysiology,Anatomy&GeneticsUniversityofOxfordSouthParksRoadOxfordOX13PTUnitedKingdomTelephone:+44(0)1865282690Fax:+44(0)1865285862Email:[email protected]@dpag.ox.ac.uk
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
Supplementary Table 1. Cell Types and Number of Genes per Cell Type forMouseOrthologousandHumanMarkerGenesCellType(frommouse
singlecelltranscriptomics)
NumberofGenesinPublication(Zeiselet
al.)
NumberofGenesinAnalysis
S1PyramidalNeurons 294 180Interneurons 365 264Microglia 436 234Astrocytes 240 185Oligodendrocytes 453 348EpendymalCells 484 266EndothelialCells 353 237MuralCells 155 95CellType(fromhuman
singlecelltranscriptomics)
NumberofGenesinPublication(Darmanis
etal.)
NumberofGenesinAnalysis
Neurons 21 21Microglia 21 13Astrocytes 21 19Oligodendrocytes 21 19OligodendrocytePrecursorCells(OPCs)
21 15
Endothelial 21 16Supplementary Table 1. The cell type marker genes from the mouse single celltranscriptomicsstudy23were filteredbasedonwhether theyhadmouseEnsemblIDs,wereone-to-onemousetohumanorthologs,hadhumanEnsemblIDsandwerepresentintheASDRNA-seqstudy6.Thecelltypemarkergenesobtainedfromthehumansinglecelltranscriptomicsstudy24werefilteredbasedonwhethertheyhadhumanEnsemblIDsandwereintheASDRNA-seqstudy.
Supplementary Table 2. Median expression log2(FC) of Mouse Orthologous Marker Genes CellType Number
ofGenesMedianlog2(FC)
MedianFC
Empiricalp-value
log2(CI)2.5%
log2(CI)97.5%
CI2.5%
CI97.5%
S1PyramidalNeurons
180 -0.25
0.84
0.0060*
-0.33
-0.15
0.79
0.90
Interneurons 264 -0.18 0.88 0.010* -0.26 -0.12 0.84 0.92Microglia 234 0.27 1.21 0.0060* 0.20 0.36 1.15 1.28Astrocytes 185 0.10 1.07 0.058 0.06 0.15 1.04 1.11Oligodendrocytes 348 -0.04 0.97 0.72 -0.08 0.03 0.95 1.02EpendymalCells 266 0.07 1.05 0.024✚ 0.01 0.12 1.01 1.08EndothelialCells 237 0.13 1.10 0.034✚ 0.09 0.17 1.07 1.13MuralCells 95 0.17 1.13 0.034✚ 0.08 0.27 1.05 1.21
SupplementaryTable2.Medianexpressionlog2FoldChange(FC)resultsformouseorthologous marker genes per cell type. Empirical two-tailed p-values werecalculated by comparing the observedmedian log2(FC)with a null distribution of
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
1000 median log2(FC) obtained from 1000 case-control label permutations.Significant p-values at an FDR<5% aremarked by an asterisk (*) andmarginallysignificantp-valuesatanFDR<5.1%aremarkedbyaplussymbol(�).ThethresholdforsignificancewassetatFDR<5%.ConfidenceIntervals(CIs)weredeterminedbybootstrapping log2(FC) from cell type marker genes 1000 times and obtaining amedianeachtime.
Supplementary Table 3. Median expression log2(FC) of Human Marker Genes CellType Number
ofGenesMedianLog2(FC)
MedianFC
Empiricalp-value
log2(CI)2.5%
log2(CI)97.5%
CI2.5%
CI97.5%
Neurons 21 -0.73 0.60 0.0040* -0.98 -0.51 0.51 0.70Microglia 13 0.95 1.93 0.0020* 0.56 1.07 1.48 2.09Astrocytes 19 0.18 1.13 0.40 -0.01 0.41 0.99 1.33Oligodendrocytes 19 -0.82 0.57 0.064 -0.91 -0.57 0.53 0.67OPCs 15 0.20 1.15 0.024* 0.12 0.29 1.08 1.22EndothelialCells 16 0.23 1.17 0.30 0.13 0.30 1.09 1.23
SupplementaryTable3.Medianexpressionlog2FoldChange(FC)resultsforhumanmarker genes per cell type. Two-tailed empirical p-values were calculated bycomparing theobservedmedian log2(FC)withanulldistribution from1000case-control label permutations. Significant p-values at an FDR<5% aremarked by anasterisk(*).ConfidenceIntervals(CIs)weredeterminedbybootstrappinglog2(FCs)1000timesandobtainingamedianeachtime.
Supplementary Table 4. Median expression log2(FC) of Human Neuronal Communities Marker Genes NeuronalCommunity
NumberofGenes
MedianLog2(FC)
MedianFC
Empiricalp-value
log2(CI)2.5%
log2(CI)97.5%
CI2.5%
CI97.5%
3 5 0.05 1.03 0.82 -0.46 0.53 0.73 1.444 23 -0.64 0.64 0.0040* -0.75 -0.37 0.60 0.785 13 -0.27 0.83 0.056 -0.52 0.14 0.70 1.106 8 -0.02 0.98 0.86 -0.53 0.31 0.69 1.24
SupplementaryTable4.Medianexpressionlog2FoldChange(FC)resultsforhumanneuronalcommunitiesmarkergenes.Empiricaltwo-tailedp-valueswerecalculatedbycomparingtheobservedmedianlog2(FC)withanulldistributionfrom1000case-control label permutations. Significant p-values at an FDR<5% aremarked by anasterisk(*).ConfidenceIntervals(CIs)weredeterminedbybootstrappinglog2(FC)1000timesandobtainingamedianeachtime.
Supplementary Table 5. Median expression log2(FC) of Mouse Orthologous Neuronal Marker Genes Subdivided by Protein Product Localization Neuronal Class and Protein Product Localization
Number of Genes
Median Log2(FC)
Median FC
Empirical p-value
log2(CI) 2.5%
log2(CI) 97.5%
CI 2.5%
CI 97.5%
S1 Pyramidal Nucleus 23 -0.27 0.83 0.012 * -0.56 -0.14 0.68 0.91
S1 Pyramidal Axon and Dendrites
16 -0.27 0.83 0.018* -0.55 -0.09 0.68 0.94
Interneurons Nucleus 53 -0.14 0.91 0.010* -0.29 -0.03 0.82 0.98
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint
Interneurons Axon and Dendrites
23 -0.29 0.82 0.0040* -0.39 -0.09 0.76 0.94
SupplementaryTable5.Medianexpressionlog2FoldChange(FC)resultsformouseorthologous S1 pyramidal and interneuron marker genes with protein productslocalizedinthenucleusoraxonsanddendrites.Empiricaltwo-tailedp-valueswerecalculatedbycomparingtheobservedmedianlog2(FC)withanulldistributionfrom1000 case-control label permutations. Significant p-values at an FDR<5% aremarked by an asterisk (*). Confidence Intervals (CIs) were determined bybootstrappinglog2(FC)1000timesandobtainingamedianeachtime.
not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted November 21, 2015. ; https://doi.org/10.1101/032557doi: bioRxiv preprint