hypnotherapy for incontinence - bmj · urinary incontinence is one ofthe most socially...

1

Mouseembryonicstemcellsself-organizeintotrunk-likestructureswithneuraltubeandsomitesJesseVVeenvliet1,†*,,AdrianoBolondi2,†,HeleneKretzmer2,%,LeahHaut1,%,ManuelaScholze-Wittler1,DennisSchifferl1,FredericKoch1,MilenaPustet1,SimonHeimann1,ReneBuschow3,LarsWittler1,BerndTimmermann4,AlexanderMeissner2,*,BernhardGHerrmann1,5,*

1Dept.OfDevelopmentalGenetics,MaxPlanckInstituteforMolecularGenetics,Ihnestr.63-73,14195Berlin,Germany.2Dept.OfGenomeRegulation,MaxPlanckInstituteforMolecularGenetics,Ihnestr.63-73,14195Berlin,Germany.3MicroscopyandCryo-ElectronMicroscopy,MaxPlanckInstituteforMolecularGenetics,Ihnestr.63-73,14195Berlin,Germany.4SequencingCoreFacility,MaxPlanckInstituteforMolecularGenetics,Ihnestr.63-73,14195Berlin,Germany.5InstituteforMedicalGenetics,Charité-UniversityMedicineBerlin,CampusBenjaminFranklin,Hindenburgdamm30,12203Berlin,Germany.†Theseauthorscontributedequally%Theseauthorscontributedequally*Correspondingauthors,[email protected],[email protected],[email protected]

One sentence summary: A platform for generating trunk-like-structures with precursors of

spinalcord,boneandmusclefromstemcellsinadish

Abstract:

Post-implantation embryogenesis is a highly dynamic process comprising multiple lineage

decisionsandmorphogeneticchangesinaccessibletodeepanalysisinvivo.Mouseembryonic

stemcells(mESCs)canformaggregatesreflectingthepost-occipitalembryo(gastruloids),but

lacking propermorphogenesis. Herewe show that embedding of aggregates derived from

mESCsinanextracellularmatrixcompoundresultsinTrunk-Like-Structures(TLS)withahigh

leveloforganizationcomprisinganeuraltubeandsomites.Comparativesingle-cellRNA-seq

analysis demonstrates that TLS execute gene-regulatory programs in an embryo-like order,

and generateprimordial germ cell like cells (PGCLCs). TLS lackingTbx6 formectopicneural

tubes, mirroring the embryonic mutant phenotype. ESC-derived trunk-like structures thus

constitute a novel powerful in vitro platform for investigating lineage decisions and

morphogeneticprocessesshapingthepost-implantationembryo.

.CC-BY-NC-ND 4.0 International license(which was not certified by peer review) is the author/funder. It is made available under aThe copyright holder for this preprintthis version posted March 4, 2020. . https://doi.org/10.1101/2020.03.04.974949doi: bioRxiv preprint

https://doi.org/10.1101/2020.03.04.974949

http://creativecommons.org/licenses/by-nc-nd/4.0/

2

Vertebratepost-implantationdevelopment comprisesamultitudeof complexmorphogenetic1

processes,whichresultfromself-organizationofstemcellsandtheirdescendantsshapingthe2

embryonicbodyplan1.Recentlydevelopedstemcellmodelsrepresentpowerfulplatformsfor3

deconstructing the dynamics of these processes that are inaccessible in the embryo1, 2. The4

mostadvancedmodelsintermsofdevelopmentalstageaccomplishedsofararegastruloids2,3,5

aggregatesofmESCsabletoself-organizeintothethreebodyaxes3.However,gastruloidslack6

propermorphogenesis,suchasa failureofneural tubeformationfromneuralcellsorsomite7

condensationfrompresomiticcells3.Invivo,theextracellularmatrix(ECM)hasacriticalrolein8

tissue morphogenesis4. In vitro, matrigel can serve as ECM surrogate, and culture media9

supplemented with a low percentage of matrigel have been shown to induce complex10

morphogenesisinorganoids5.11

We therefore explored if embryo-like morphogenetic features could be induced by12

embedding96hpostaggregationgastruloids in5%matrigel foranadditional24hperiod (Fig.13

1a). We also tested a combination of matrigel together with the WNT signaling activator14

CHIR99021(CHIR)andtheBMPsignalinginhibitorLDN193189(LDN),thathavebeenreported15

to inducea(pre-)somiticmesoderm((P)SM)fate in2Dand3Ddifferentiationprotocols6–8.To16

facilitatehigh-throughputcharacterizationandquantificationofourconditions,wegenerated17

mESCswithT::H2B-mCherry(hereafterTmCH)andSox2::H2B-Venus(hereafterSox2VE)reporters,18

marking the mesodermal (ME) or neural (NE) lineage respectively (Extended Data Fig. 1a).19

Strikingly, embedding in 5%matrigel alone was sufficient for segmentation in the TmCH+ME20

domain and formation of a Sox2VE+ neural tube like structure (Fig. 1b,d; ExtendedData Fig.21

1b,c). The vastmajority of structures (hereafter referred to as TLS forTrunk-Like-Structures)22


https://doi.org/10.1101/2020.03.04.974949


3

elongatedandformedaTmCH+poleattheposteriorend,withsegmentationoccurringinabout23

halftheTLS(Fig.1c).SegmentswereTmCH+demonstratingtheirmesodermalorigin,andwhole-24

mount in situ hybridization for Tcf15 and Uncx confirmed their somite identity (Fig. 1d,e)7.25

AdditionalCHIRtreatment,bothalone(hereafterTLSC)or incombinationwithLDN(hereafter26

TLSCL) improvedthephysicalseparationofneighboringsegmentswithoutaffectingTmCH+pole27

formation or elongation (Fig. 1b,c; Extended Data Fig. 1d-g), and resulted in an excess of28

segmentsattheanteriorend,arrangedlikea“bunchofgrapes”(Fig.1b,d)9.Moreover,theME29

domainexpandedattheexpenseoftheNEcompartment,withapparentdisorganizationofthe30

posterior end and neural tissue (Fig. 1b; Extended Data Fig. 1f,g), as well as reduction of31

TmCH+/Sox2VE+putativeneuromesodermalprogenitors (NMPs) -bipotentcells thatgive rise to32

both post-occipital NE andME -, as confirmed by flow cytometry (Fig. 1f)10. In all three TLS33

protocols,segmentsweresimilarinsizetoembryonicsomites(ExtendedDataFig.1h).34

Ingastruloids,endodermalcellsgenerallyorganize intoa tubular structure resembling35

anembryonic gut3.Whole-mount immunofluorescenceanalysisof characteristic, endodermal36

expressed transcription factorsFOXA2andSOX17also confirmedgut formation inour trunk-37

likestructures(ExtendedDataFig.2a-d).CellsattheposteriorbasewereSOX17-negative,but38

co-expressedFOXA2withhighlevelsofTmCH(ExtendedDataFig.2b).Thus,ourdatashowthat39

embedding inmatrigel is both necessary and sufficient to drive complex, embryo-like tissue40

morphogenesisofthethreeembryonicgermlayers.41

TocharacterizeourstructuresinmoredetailweperformedRNA-seqanalysis(Extended42

DataFig.3a),andfoundthatTLSmodelthepost-occipitalembryo,similartogastruloids,based43


https://doi.org/10.1101/2020.03.04.974949


4

on selectedmarkergenes (Fig.2a; ExtendedDataFig.3b)3. Compared toTLS,bothTLSC and44

TLSCL showed a significant upregulation of genes involved in (pre)somitic development (e.g.45

Tbx6, Msgn1, Hes77, 8, 11) at the expense of NE marker genes (e.g. Sox1, Pax6, Irx311–13),46

corroborating the flowcytometryand imaging results (Fig.2a;ExtendedDataFig.3b-d).The47

analysis of marker gene sets for NMPs, their direct descendants undergoing lineage choice48

(NMP ME & NMP NE), and for committed NE and ME cells substantiated this finding11.49

Compared to TLS, TLSC and TLSCL displayed reduced expression of markers in all clusters50

includingME(Fig.2b).(P)SMspecificmarkers,however,wereonaverageupregulated,whereas51

intermediateME (IM) and lateral plateME (LPM)markerswere downregulated in TLSC, and52

further reduced inTLSCL (Fig.2b;ExtendedDataFig.3c,e)3, 8. Thesedataare in linewith the53

known role of WNT- versus BMP-signaling in conferring PSM versus IM and LPM subtype54

identity(Fig.2a;ExtendedDataFig.3e)7.55

We next searched for gene expression differences that might underlie the improved56

physical separation of somites observed in TLSC and TLSCL. Among the most strongly57

upregulatedgenescomparedtoTLSwasWnt6,whichactsasasomiteepithelializationfactorin58

vivo (Extended Data Fig. 3f)14. In addition, multiple ephrins and their receptors, and other59

factorsinvolvedinsomiteephithelializationwereupregulated(Fig.2c;ExtendedDataFig.3f)15,60

16. Expression changes of selected somite polaritymarkerswas observed concomitantlywith61

changes in their inducers, inagreementwith the role thatWNTs, in concertwithSHH,BMPs62

andtheirantagonistsplayinsomitecompartmentalizationinvivo(ExtendedDataFig.3g,h)7,17.63

Thus, exposure to CHIR or CHIR/LDN improved segment boundary formation, but affected64

somitepatterning.65


https://doi.org/10.1101/2020.03.04.974949


5

PrincipalComponentAnalysis (PCA) indicates ahigh transcriptional similaritybetween66

gastruloidsandTLSdespiteprofoundmorphologicaldifferences(Fig.2d).Thelatterarebetter67

highlighted by Gene-Set-Enrichment-Analysis (GSEA), which shows that matrigel embedding68

promotestissuemorphogenesisandremodeling(Fig.2e;ExtendedDataFig.3i).Zoominginon69

embryonicandtissuemorphogenesisgenesetsshowedthatupregulatedgenesalsocomprise70

markers of blood vessel development suggesting the inductionof capillarymorphogenesis in71

TLS (Fig. 2f; ExtendedData Fig. 3j). Cell-cell and cell-matrix interactions play pivotal roles in72

morphogenesis,with important functions for (proto)cadherins,ephrinsand integrins in tissue73

formationandorganogenesis18–23.GSEAindeedrevealedanenrichmentofcelladhesionterms74

andoverallasignificantupregulationofcorrespondingmarkergenesinTLS(Fig.2e;Extended75

Data Fig. 4a,b). The most pronounced increase was observed for integrins, transmembrane76

receptorsmediatingcelladhesiontotheECMimportantfore.g.neuraltubeformation,blood77

vessel development, and segmentation (Fig. 2g; ExtendedData Fig. 4a,b)20, 21, 24–26. Of note,78

integrin ligands important for somite boundary formation (fibronectin, collagen IV, laminin79

alpha1and lamininalpha5)wereexpressedat the same levels in gastruloids (ExtendedData80

Fig. 4c)24, 25, 27, 28. Taken together, our molecular data confirm activation of morphogenetic81

programsthroughmatrigelembeddingthatmaybelinkedtotheupregulationofcelladhesion82

molecules.83

Basedon the above resultswedecided to focuson theTLS condition for an in-depth84

characterization, as it produced the most in-vivo-like configuration. First, to ensure85

reproducibility29, we evaluated variability between nine independent structures using the86

quantitative expression of a panel of 41 developmental genes. The data demonstrated high87


https://doi.org/10.1101/2020.03.04.974949


6

correlationbetweenTLSsandrevealedthattheymostcloselyresembletheE8.5post-occipital88

stage(ExtendedDataFig.5a).Theanalysisofindividualgenesshowedthatendothelial,NMP,89

NE, and somiticMEgeneswere reproducibly expressed at similar levels across all replicates,90

whereashighervariationwasobservedintheexpressionofPSM,endodermal,andpluripotency91

genes(ExtendedDataFig.5b,c).92

Next we performed scRNA-seq analysis on a total of 20,294 post-processed cells93

sampled from TLS at 96, 108, and 120 hours (Extended Data Fig. 6a). Clustering analysis94

revealed14differentcellstateswiththe largeronescorrespondingtoderivativesofthePSM95

andNEthatflankputativeNMPs.Smallerclusterscomprisedendoderm,endothelialcellsand96

PrimordialGermCellLikeCells(PGCLCs)(ExtendedDataFig.6b).Themainclustersorganized97

into a continuum of states recapitulating spatio-temporal features of the developing post-98

occipitalembryo(Fig.3a).Acrossthethreetimepointssampledprogenitorsubtypesgradually99

decreasedinfavorofmorematureneuralandsomiticcellsasdevelopmentprogresses(Fig.3b;100

ExtendedDataFig.6c).Asexpected,putativeNMPsco-expressedTmCH,Sox2VEandCDX2,and101

showedanin-vivo-likeNMPsignatureandlocation(Fig.3c;ExtendedDataFig.6d)10,11,30.RNA102

Velocityanalysisrevealedneuralandsomitictrajectoriesrooted intheNMPs,suggestingthat103

TLSdevelopmentrecapitulatesthedevelopmentaldynamicsobservedinthepost-occipitalmid-104

gestationalembryo(Fig.3d;ExtendedDataFig.7a)31.Invivo,NMPsandtheirPSMdescendants105

arearranged inanorderofprogressivematurityalong theposterior toanterioraxis7. In line106

with this, ordering of cells along a pseudo-temporal trajectory showed that the somitic107

trajectoryreflectsthegeneticcascadeobservedintheembryo(Fig.3e;ExtendedDataFig.8a).108

Forexample,thetrajectoryfromFgf8+NMPsandPSMviathedeterminationfrontmarkedby109


https://doi.org/10.1101/2020.03.04.974949


7

Mesp2toMeox1+somiteswasfaithfullyrecapitulatedandtheembryo-likespatialarrangement110

was confirmed bywhole-mount in situhybridization (Fig. 3f)7. Likewise, the genetic cascade111

fromNMPstoneuralprogenitorsinTLSreflecteditsinvivocounterpartinspaceandtime(Fig.112

3g).Notably,subclusteringoftheneuralcellsdemonstratedthatTLSgeneratebothdorsaland113

ventralneuralsubtypes,withdorsalsubtypesbeingmoreprevalent(ExtendedDataFig.8b)32.114

The analysis of Hox gene expression at consecutive time points demonstrated in-vivo-like115

collinearity, as described for gastruloids (Extended Data Fig. 8c)3. To test if TLS somites116

establish dorsal-ventral (DV) and anterior-posterior (AP) domains, we reclustered all somitic117

cells.At96hwedetectedtwomaingroups,corresponding to theUncx+posteriorandTbx18+118

anterior somite domains, in line with AP compartmentalization established during119

segmentation (ExtendedDataFig. 9a,b)7. At 120h,we founddistinct clusters of Pax3+ dorsal120

andPax1+ventralcells,andasmallclusterofLbx1+/Met+putativemigratorymuscleprecursors121

(ExtendedDataFig.9c-f)7,33. Inaddition,Scx+syndetomecellsweredetected(ExtendedData122

Fig. 9g), and Uncx and Tbx18 expression were anti-correlated (Extended Data Fig. 9h). An123

unexpectedfindingwastheidentificationofPGCLCs,sincetheyhavenotbeenobservedpriorin124

gastruloids3.Intheembryo,nascentPGCscanbeidentifiedatE7.5asagroupofDPPA3+cells,125

whicharegeneratedintheposteriorprimitivestreakandlatermigratealongthehindguttothe126

gonads34. We assigned PGCLC identity using characteristic PGC genes and identified their127

locationinourstructures(Fig.3h;ExtendedDataFig.10).At108hwefindagroupofSox2VE-high128

cellsthatco-expressedDPPA3(Fig.3h;ExtendedDataFig.10a).At120h,Sox2VE-highcellswere129

detected in contact with FOXA2+ cells, and DPPA3+ cells in contact with a TmCH+ gut-like130

epithelial structure (Fig. 3h; Extended Data Fig. 10b). These data show that trunk-like131


https://doi.org/10.1101/2020.03.04.974949


8

structures indeed contain cells displaying characteristics typical for PGCs. To investigate how132

closethecellularstatesidentifiedinTLSresemblethoseinembryos,wemappedoursingle-cell133

transcriptomestoascRNA-seqcompendiumofpost-occipitalembryoniccellularsubtypes(Fig.134

4a)35.ThedatarevealedgloballyahighaccordanceofTLSandembryoniccell types including135

characteristicmarkergenesandpairwisecomparisonofmappedclustersidentifiedonlyasmall136

fractionofdifferentiallyexpressedgenes(Fig.4b,c;ExtendedDataFig.11a-d).Ofnote,PS-and137

early-NMP-likecellsareexclusivelypresentat96handreplacedbylate-NMP-likecellsat108h138

and 120h (Fig. 4d). Taken together, our scRNA-seq analyses demonstrate that the trunk-like139

structures execute gene-regulatory programs in a spatiotemporal order that resembles the140

embryo.141

Finally,toexploretheutilityoftrunk-likestructuresfurther,wenexttestediftheycould142

reproduce an embryonic mutant phenotype caused by gene ablation in a proof of concept143

experiment. In vivo, loss ofTbx6 results in an expansion ofNE and the formation of ectopic144

neuraltubesattheexpenseofPSMandsomites(Fig.4e)36.WedeletedTbx6fromTbx6::H2B-145

Venus(Tbx6Ve)mESC,generatedTLSandfoundaclearfailuretoformsomites,whichcouldalso146

not be rescued by CHIR or CHIR/LDN (Fig. 4e,f; Extended Data Fig. 12a). Quantitative PCR147

analysisonFACS-purifiedTbx6VE+cellsrevealedupregulationofneuralmarkersattheexpense148

of (P)SMmarkers in Tbx6-/- cells, thus recapitulating the in vivo phenotype at themolecular149

level (Extended Data Fig. 12b). Finally, whole-mount immunofluorescence analysis for SOX2150

showedthatTLS-Tbx6-/-generatedectopicTbx6VE+neuraltubes,whereasTLSCandTLSCLformed151

anexcessofmorphologicallyindistinctSOX2+tissue(Fig.4g;ExtendedDataFig.12c,d).152


https://doi.org/10.1101/2020.03.04.974949


9

Here we report the generation of trunk-like structures and demonstrate that they153

faithfully reproduce important features of post-occipital embryogenesis, as confirmed by154

morphogeneticandtranscriptionalcriteria.Importantly,toourbestknowledgeaxialelongation155

togetherwith neural tube, gut and somite formation in combinationwith the generation of156

PGCLCs, solely by in vitro culture of stem cells, has been achieved for the first time. TLS157

therefore provide a powerful in vitro system for investigating the dynamics of embryo158

patterning and morphogenesis at the single cell level and in molecular detail, allowing to159

“understand the whole from the parts”1. Towards this latter goal, comparative analysis of160

gastruloids and TLSwill help to understand how cell-cell and cell-matrix interactions control161

embryonicarchitectures1.Inthisregard,celltypespecificexpressionofcelladhesionmolecules162

revealedbyoursingle-cellTLSatlaswillprovideguidanceforfuturework(Fig.S13).Although163

TLS display the closest in-vivo-like configuration, we envision that our CHIR and CHIR/LDN164

models can become important systems for studying the morphogenetic aspects of165

somitogenesis inhigh-throughput.OurTLSalreadydisplay remarkable reproducibilitybothat166

the morphological and molecular level. Nevertheless, we realize the importance of further167

integrated comparative analysis of single TLS at single-cell resolution, such as recently168

publishedforbrainandkidneyorganoids37,38.Finally,invivosomitesareformedinarhythmic169

process involving an oscillator - the segmentation clock39. We have observed that somite170

formation in TLS can occur sequentially at an embryo-like pace, and specific control genes171

involvedinthesegmentationprocessarecorrectlyexpressedinaPSMasexpected,suggesting172

thattheclockalsoticksinthePSMofTLS(ExtendedDataFig.14;SupplementaryVideo1).173


https://doi.org/10.1101/2020.03.04.974949


10

In sum, our data demonstrate that the morphogenetic potential of mESC-derived174

aggregates is unlocked by providing an ECM surrogate. The resulting trunk-like structures175

provideascalable,tractable,readilyaccessibleplatformforinvestigatinglineagedecisionsand176

morphogenetic processes shaping the mid-gestational embryo at an unprecedented177

spatiotemporalresolution.178


https://doi.org/10.1101/2020.03.04.974949


11

References1. M.N.Shahbazi,E.D.Siggia,M.Zernicka-Goetz,Self-organizationofstemcellsinto

embryos:Awindowonearlymammaliandevelopment.Science.364,948–951(2019).

2. M.N.Shahbazi,M.Zernicka-Goetz,Deconstructingandreconstructingthemouseand

humanearlyembryo.Nat.CellBiol.20,878–887(2018).

3. L.Beccarietal.,Multi-axialself-organizationpropertiesofmouseembryonicstemcells

intogastruloids.Nature.562,272–276(2018).

4. H.K.Kleinman,D.Philp,M.P.Hoffman,Roleoftheextracellularmatrixinmorphogenesis.

Curr.Opin.Biotechnol.14,526–532(2003).

5. J.A.Brassard,M.P.Lutolf,EngineeringStemCellSelf-organizationtoBuildBetter

Organoids.CellStemCell.24,860–876(2019).

6. M.Matsumiya,T.Tomita,K.Yoshioka-Kobayashi,A.Isomura,R.Kageyama,EScell-derived

presomiticmesoderm-liketissuesforanalysisofsynchronizedoscillationsinthe

segmentationclock.Development.145(2018),doi:10.1242/dev.156836.

7. J.Chal,O.Pourquié,Makingmuscle:skeletalmyogenesisinvivoandinvitro.

Development.144,2104–2122(2017).

8. J.Chaletal.,DifferentiationofpluripotentstemcellstomusclefibertomodelDuchenne

musculardystrophy.Nat.Biotechnol.33,962–969(2015).

9. A.S.Dias,I.deAlmeida,J.M.Belmonte,J.A.Glazier,C.D.Stern,Somiteswithoutaclock.

Science.343,791–795(2014).

10. D.Henrique,E.Abranches,L.Verrier,K.G.Storey,Neuromesodermalprogenitorsandthe

makingofthespinalcord.Development.142,2864–2875(2015).


https://doi.org/10.1101/2020.03.04.974949


12

11. F.Kochetal.,AntagonisticActivitiesofSox2andBrachyuryControltheFateChoiceof

Neuro-MesodermalProgenitors.Dev.Cell.42,514–526.e7(2017).

12. M.Gouti,V.Metzis,J.Briscoe,Theroutetospinalcordcelltypes:ataleofsignalsand

switches.TrendsGenet.31,282–289(2015).

13. E.Dessaud,A.P.McMahon,J.Briscoe,Patternformationinthevertebrateneuraltube:a

sonichedgehogmorphogen-regulatedtranscriptionalnetwork.Development.135,2489–

2503(2008).

14. C.Schmidtetal.,Wnt6regulatestheepithelialisationprocessofthesegmentalplate

mesodermleadingtosomiteformation.Dev.Biol.271,198–209(2004).

15. L.Durbinetal.,Ephsignalingisrequiredforsegmentationanddifferentiationofthe

somites.GenesDev.12,3096–3109(1998).

16. M.Rowtonetal.,Regulationofmesenchymal-to-epithelialtransitionbyPARAXISduring

somitogenesis.Dev.Dyn.242,1332–1344(2013).

17. C.Marcelle,M.R.Stark,M.Bronner-Fraser,CoordinateactionsofBMPs,Wnts,Shhand

nogginmediatepatterningofthedorsalsomite.Development.124,3955–3963(1997).

18. J.M.Halbleib,W.J.Nelson,Cadherinsindevelopment:celladhesion,sorting,andtissue

morphogenesis.GenesDev.20,3199–3214(2006).

19. V.Barone,C.-P.Heisenberg,Celladhesioninembryomorphogenesis.Curr.Opin.CellBiol.

24,148–153(2012).

20. D.Jülichetal.,Cross-ScaleIntegrinRegulationOrganizesECMandTissueTopology.Dev.

Cell.34,33–44(2015).


https://doi.org/10.1101/2020.03.04.974949


13

21. D.Jülich,A.P.Mould,E.Koper,S.A.Holley,Controlofextracellularmatrixassemblyalong

tissueboundariesviaIntegrinandEph/Ephrinsignaling.Development.136,2913–2921

(2009).

22. J.H.Shawky,L.A.Davidson,Tissuemechanicsandadhesionduringembryodevelopment.

Dev.Biol.401,152–164(2015).

23. J.Chal,C.Guillot,O.Pourquié,PAPCcouplesthesegmentationclocktosomite

morphogenesisbyregulatingN-cadherin-dependentadhesion.Development.144,664–

676(2017).

24. S.-W.Chong,Y.-J.Jiang,Offlimits--integrinsholdingboundariesinsomitogenesis.Trends

CellBiol.15,453–457(2005).

25. S.Koshidaetal.,Integrinalpha5-dependentfibronectinaccumulationformaintenanceof

somiteboundariesinzebrafishembryos.Dev.Cell.8,587–598(2005).

26. E.Nikolopoulou,G.L.Galea,A.Rolo,N.D.E.Greene,A.J.Copp,Neuraltubeclosure:

cellular,molecularandbiomechanicalmechanisms.Development.144,552–566(2017).

27. A.-G.Borycki,Themyotomalbasementmembrane:insightintolaminin-111functionand

itscontrolbySonichedgehogsignaling.CellAdhMigr.7,72–81(2013).

28. P.Rifes,S.Thorsteinsdóttir,Extracellularmatrixassemblyand3Dorganizationduring

paraxialmesodermdevelopmentinthechickembryo.Dev.Biol.368,370–381(2012).

29. M.Huch,J.A.Knoblich,M.P.Lutolf,A.Martinez-Arias,Thehopeandthehypeof

organoidresearch.Development.144,938–941(2017).

30. S.Aminetal.,CdxandTBrachyuryCo-activateGrowthSignalingintheEmbryonicAxial

ProgenitorNiche.CellRep.17,3165–3177(2016).


https://doi.org/10.1101/2020.03.04.974949


14

31. G.LaMannoetal.,RNAvelocityofsinglecells.Nature.560,494–498(2018).

32. A.Sagner,J.Briscoe,Establishingneuronaldiversityinthespinalcord:atimeandaplace.

Development.146,dev182154(2019).

33.M.Buckinghametal.,Theformationofskeletalmuscle:fromsomitetolimb.J.Anat.202,

59–68(2003).

34.M.Saitou,M.Yamaji,Primordialgermcellsinmice.ColdSpringHarb.Perspect.Biol.4

(2012),doi:10.1101/cshperspect.a008375.

35. M.M.Chanetal.,Molecularrecordingofmammalianembryogenesis.Nature.570,77–82

(2019).

36. D.L.Chapman,V.E.Papaioannou,Threeneuraltubesinmouseembryoswithmutations

intheT-boxgeneTbx6.Nature.391,695–697(1998).

37. B.Phipsonetal.,Evaluationofvariabilityinhumankidneyorganoids.Nat.Methods.16,

79–87(2019).

38. S.-J.Yoonetal.,Reliabilityofhumancorticalorganoidgeneration.Nat.Methods.16,75–

78(2019).

39. O.Pourquié,Thesegmentationclock:convertingembryonictimeintospatialpattern.

Science.301,328–330(2003).

40. D.A.Stafford,L.J.Brunet,M.K.Khokha,A.N.Economides,R.M.Harland,Cooperative

activityofnogginandgremlin1inaxialskeletondevelopment.Development.138,1005–

1014(2011).


https://doi.org/10.1101/2020.03.04.974949


15

41. X.M.Yang,K.Vogan,P.Gros,M.Park,Expressionofthemetreceptortyrosinekinasein

muscleprogenitorcellsinsomitesandlimbsisabsentinSplotchmice.Development.122,

2163–2171(1996).

42. Y.J.Jiangetal.,Notchsignallingandthesynchronizationofthesomitesegmentation

clock.Nature.408,475–479(2000).

43. T.J.Cunninghametal.,Retinoicacidactivityinundifferentiatedneuralprogenitorsis

sufficienttofulfillitsroleinrestrictingfgf8expressionforsomitogenesis.PLoSOne.10,

e0137894(2015).

44. R.Moritaetal.,ETStranscriptionfactorETV2directlyconvertshumanfibroblastsinto

functionalendothelialcells.Proc.Natl.Acad.Sci.USA.112,160–165(2015).

45. W.W.Sugdenetal.,Endoglincontrolsbloodvesseldiameterthroughendothelialcell

shapechangesinresponsetohaemodynamiccues.Nat.CellBiol.19,653–665(2017).

46. K.Xuetal.,BloodvesseltubulogenesisrequiresRasip1regulationofGTPasesignaling.

Dev.Cell.20,526–539(2011).

47. S.Kannoetal.,RolesoftwoVEGFreceptors,Flt-1andKDR,inthesignaltransductionof

VEGFeffectsinhumanvascularendothelialcells.Oncogene.19,2138–2146(2000).

48. S.Gauvritetal.,HHEXisatranscriptionalregulatoroftheVEGFC/FLT4/PROX1signaling

axisduringvasculardevelopment.Nat.Commun.9,2704(2018).

49. G.S.Eakin,A.-K.Hadjantonakis,Productionofchimerasbyaggregationofembryonic

stemcellswithdiploidortetraploidmouseembryos.Nat.Protoc.1,1145–1153(2006).

50. S.H.L.Georgeetal.,Developmentalandadultphenotypingdirectlyfrommutant

embryonicstemcells.Proc.Natl.Acad.Sci.USA.104,4455–4460(2007).


https://doi.org/10.1101/2020.03.04.974949


16

51. J.P.Muyrers,Y.Zhang,G.Testa,A.F.Stewart,Rapidmodificationofbacterialartificial

chromosomesbyET-recombination.NucleicAcidsRes.27,1555–1557(1999).

52. F.A.Ranetal.,GenomeengineeringusingtheCRISPR-Cas9system.Nat.Protoc.8,2281–

2308(2013).

53. L.Beccarietal.,GeneratingGastruloidsfromMouseEmbryonicStemCells.Protocexch

(2018),doi:10.1038/protex.2018.094.

54. A.Butler,P.Hoffman,P.Smibert,E.Papalexi,R.Satija,Integratingsingle-cell

transcriptomicdataacrossdifferentconditions,technologies,andspecies.Nat.

Biotechnol.36,411–420(2018).

55. F.A.Wolf,P.Angerer,F.J.Theis,SCANPY:large-scalesingle-cellgeneexpressiondata

analysis.GenomeBiol.19,15(2018).

56. V.Bergen,M.Lange,S.Peidli,F.A.Wolf,F.J.Theis,GeneralizingRNAvelocitytotransient

cellstatesthroughdynamicalmodeling.BioRxiv(2019),doi:10.1101/820936.

57. F.A.Wolfetal.,PAGA:graphabstractionreconcilesclusteringwithtrajectoryinference

throughatopologypreservingmapofsinglecells.GenomeBiol.20,59(2019).

58. M.Martin,Cutadaptremovesadaptersequencesfromhigh-throughputsequencingreads.

EMBnetj.17,10(2011).

59. A.Dobinetal.,STAR:ultrafastuniversalRNA-seqaligner.Bioinformatics.29,15–21

(2013).

60. M.Perteaetal.,StringTieenablesimprovedreconstructionofatranscriptomefromRNA-

seqreads.Nat.Biotechnol.33,290–295(2015).


https://doi.org/10.1101/2020.03.04.974949


17

61. M.I.Love,W.Huber,S.Anders,Moderatedestimationoffoldchangeand''dispersionfor

RNA-seqdatawithDESeq2.GenomeBiol.15,550(2014).

62. A.Sergushichev,Analgorithmforfastprerankedgenesetenrichmentanalysisusing

cumulativestatisticcalculation.BioRxiv(2016),doi:10.1101/060012.


https://doi.org/10.1101/2020.03.04.974949


18

Acknowledgments:Wearegrateful for the supportand feedback received frommembersof

theHerrmann&Meissnerlaboratories,inparticularStefanieGrosswendtandAtsuhiroTaguchi.

We thank DijanaMicic & Judith Fiedler for animal care, Norbert Mages for assistance with

(sc)RNA-Seq, Claudi Giesecke-Thiel & Uta Marchfelder for assistance with FACS, Thorsten

Mielke&BeatrixFaulerforhelpwithmicroscopy,FabianToborandPollyBurtonfortechnical

assistance, and Nikolaus Rajewsky (MDC/BIMSB) for providing access to the NanoString.

Funding: J.V.V.was partly funded by an Alexander vonHumboldt Fellowship. Thisworkwas

supportedbyNIHgrantHG006193 (A.M.)and theMaxPlanckSociety.Author contributions:

B.G.H. initiated the study; J.V.V. and B.G.H. conceived the project; J.V.V., A.M. and B.G.H.

supervised the project. J.V.V., A.B. and L.H. designed, performed and quantified most

experiments.H.K.performedbulk&sc-RNA-SeqcomputationalanalysiswithhelpofJ.V.V.and

A.B. M.S-W., D.S., F.K. and M.P. generated mESC reporter lines. L.W. performed tetraploid

complementation. S.H. performed pilot experiments to optimize culture media. R.B. helped

withimageacquisitionandanalysis.B.T.supervisednext-generationsequencing.J.V.V.drafted

thefirstversionofthemanuscript.ThefinalmanuscriptwaswrittenbyJ.V.V,A.B.,H.K.,A.M.

and B.G.H. Competing interests: The authors declare no competing interests. Data and

materialsavailability:Alldataisavailableinthemaintextorthesupplementarymaterials.Bulk

and single-cell RNA-Seq data have been deposited in the Gene Expression Omnibus (GEO)

underaccessioncodeGSE141175.Allcomputationalcodeused inthisstudy isavailableupon

request.


https://doi.org/10.1101/2020.03.04.974949


19

Fig.1.GenerationofTrunk-LikeStructureswithsomitesandaneuraltube.

a,Schematicofthecultureprotocol:200-250mESCswereaggregatedinultra-low-attachment

plates;WntagonistCHIR99021(CHIR)wasaddedbetween48and72hours(3).96aggregates

wereembeddedin5%matrigel(TLS).Optionally,structuresweretreatedwithCHIRalone(TLSC)

or in combinationwith the BMP antagonist LDN (TLSCL).b, 3D volumetric renderings (upper

panel) and confocal sections (bottom panel) of gastruloids, TLS, TLSC and TLSCL. Scale bars

100μm. Each image is representative of at least ten biological replicates with similar

morphology and expression patterns. c, Quantification of morphometric characteristics in


https://doi.org/10.1101/2020.03.04.974949


20

gastruloidsandTLS(seeSupplementalInformationforscoringcriteria).d,SegmentsinTLSare

TmCH+ and positioned adjacent to the neural tube in TLS. In TLSC and TLSCL the segments are

arrangedin“bunchesofgrapes”.Scalebars25μm.Redarrowheadindicatesneuraltube,white

arrowheadsomites.e,SegmentsexpresssomiticmarkersTcf15andUncxasshownbywhole-

mountinsituhybridization.NotethecharacteristicstripyexpressionpatternofUncxinTLSdue

to Uncx restriction to the posterior somite half, whereas Tcf15 is expressed throughout the

segments (as in theembryo).Noteworthy, inTLSCandTLSCLUncx isdetected throughout the

segments,indicatinglossofanterior-posteriorpolarity.Scalebars100μm.f,PercentageofTmCH+

(mesodermal), Sox2VE+ (neural), and TmCH+/Sox2VE+ (neuromesodermal) cells in gastruloids and

TLS asmeasured by flow cytometry. Dots represent individual TLS/gastruloids, line indicates

median.


https://doi.org/10.1101/2020.03.04.974949


21

Fig.2.GeneexpressiondifferencesbetweengastruloidsandTLSmodels.

a, Heatmap of log2(TPM+0.01) expression (TPM, Transcripts Per Million) of selected genes

associated with development of indicated embryonic structures in 96h aggregates, 120h

gastruloids, 120h TLS, 120h TLSC and 120h TLSCL, as measured by RNA-seq. Replicates were

derived from pools of independent biological samples (see Extended Data Fig. 3a for exact

experimental set-up). CE, caudal end, NMP, neuromesodermal progenitors, PSM, presomitic

mesoderm, LPM, lateral platemesoderm, IM, intermediatemesoderm.b,Box plots showing

distributionofmarkergenesforindicatedcelltypes.Boxesindicateinterquartilerange.Endof


https://doi.org/10.1101/2020.03.04.974949


22

whiskersrepresentminimumandmaximum.Blackdotsindicateoutliers.Notchesarecentered

onthemedian.ListofgenesusedforeachcategoryinSupplmentaryTable1.c,Boxplotlikein

b representing average z-score per column (pool of 3 replicates) for somite epithelialization

factors(ExtendedDataFig.3fforgenes).d,PCAanalysisofsamplesfromawithcolorcodingof

individual samples (dots) as ina. PC1 and PC2 represents the two componentswith highest

percentageofexplainedvariance.e,SelectedsignificanttermsofGeneSetEnrichmentAnalysis

(GSEA) enriched in TLS as compared to gastruloids of 120h. Full list of significant (FDR<0.05)

terms is provided inSupplementary Table2. f, Volcano plot of genes involved in tissue and

embryonicmorphogenesis.Dotsizescaleswithlog2ofabsoluteexpression.Reddots,absolute

log2FC>1andpadj<10e-15.Greendots,absolutelog2FC>1.Upperdottedline,padj(FDR)=10e-15;

bottom dotted line, padj (FDR)=0.05. Green label, involved in somitogenesis; orange label,

involvedinbloodvesseldevelopment;yellowlabel,involvedinboth40-48.g,Heatmapofscaled

expression(rowz-score)ofintegrinswithsignificantlydifferentexpression(padj(FDR)<0.05)in

gastruloidsvs120hTLS.Boxplotrepresentsz-scorepercolumn(sample),withboxesindicating

interquartile range, end of whiskers representing minimum and maximum, dots showing

outliersandcentral linerepresentingmedian.Everycolumnrepresentoneof threebiological

replicates.


https://doi.org/10.1101/2020.03.04.974949


23

Fig.3.Single-cellRNA-SequencingofTLS.

Intotal,20,294cellsweresampledfrom96,108,and120hourTLS(seeExtendedDataFig.5a

forexperimentalset-up).a,UMAP(UniformManifoldApproximationandProjection)coloured

by the fourteen clusters identified. b, Alluvial plot of percentage of neuromesodermal

progenitors(NMPs),posteriorpresomiticmesoderm(pPSM),anteriorPSM,somiticandneural

tube cells over time. c, NMPs co-express Sox2 and T (left panel, blending expression with

blend.threshold=0.4), and reside at the posterior end of TLS (confocal sections, right panel;


https://doi.org/10.1101/2020.03.04.974949


24

white arrowheads,NMPs; red arrowheads, somites). Scale bars 50μm.d, UMAP coloured by

identified clusters with trajectories inferred from RNA Velocity. Grey arrow flows represent

calculated velocity trajectories. e, Heatmap with scaled expression of genes involved in

somitogenesis in9004 cells from120hTLSMG rooted inNMPsandorderedbypseudotime. f,

UMAP coloured by expression of indicated genes (left panel), and whole-mount in situ

hybridizationforthesamegenesinTLSandE9.5embryos(rightpanel).Numbersindicatethe

fractionofTLSwithembryo-likeexpression.ScalebarsTLS100μm,embryo200μm.g,Heatmap

with scaledexpressionof genes involved inneural development in 9004 cells from120hTLS

rootedinNMPsandorderedbypseudotime(upperpanel)andUMAPcolouredbyexpressionof

indicated genes (bottom panel). h, Split violin plots of expression of marker genes for

primordialgermcelllikecells(PGCLC,leftpanel),andconfocalsectionofTLSshowingSOX2VE-

high/DPPA3+PGCLCsat108h,andDPPA3+cellsinclosecontactwiththeTmCH+gut-like-structure.

Scale bars 50μm for 108h left panel and 120h overview, 25μm for 120hmagnifications. Red

arrowheadsindicategut-like-structure,whitearrowheadsindicateDPPA3+PGCLCs.


https://doi.org/10.1101/2020.03.04.974949


25

Fig.4.TLScelltypesareembryo-likeandrecapitulatetheTbx6-/-phenotypeatthemolecular

andmorphologicallevel.

a,Comparative transcriptomeanalysis of TLS andpost-occipital E7.5 andE8.5embryoat the

single-cell level. b, TLS UMAP coloured by assigned embryonic cell type. TLS clusters are

projected as corresponding coloured contours. Blue font, TLS clusters; Red font, embryo

clusters. c, Split heatmap with percentage of assigned cells (dark grey) and certainty score

(orange) for TLS cells from the indicated cluster upon unbiased mapping to the in vivo

counterpart. Font colour code as in b. d, Alluvial plot of percentage of cells assigned to

indicated in vivo cluster in 96, 108 and 120h TLS. e, Schematic of Tbx6-/- phenotype in vivo


https://doi.org/10.1101/2020.03.04.974949


26

(upper panel) and schematic view of Tbx6::H2B-Venus; Tbx6-/- mESC derivation. f,

Quantification of segmentation phenotype in TLS-Tbx6-/- . Data represent 3 different

experiments performed with 2 independent mESC lines of each genotype. g, Formation of

ectopicneuraltubesinTLS-Tbx6-/-.Green,SOX2;Magenta,Tbx6VE.Whitearrowheadsindicate

Tbx6VE+somitesinWT,andTbx6VE+/SOX2+ectopicneuraltubesinTbx6-/-.Scalebars50μm.


https://doi.org/10.1101/2020.03.04.974949


27

SupplementaryInformation

Mouseembryonicstemcellsself-organizeintotrunk-like-structureswithneuraltubeandsomites

JesseVVeenvliet†*,AdrianoBolondi†,HeleneKretzmer,LeahHaut,ManuelaScholze-Wittler,DennisSchifferl,FredericKoch,MilenaPustet,SimonHeimann,ReneBuschow,LarsWittler,BerndTimmermann,AlexanderMeissner*,BernhardGHerrmann*

†Theseauthorscontributedequally*Corresponding authors. Email: [email protected], [email protected],[email protected]

SupplementaryInformationincludes:

MethodsExtendedDataFigs.1-15CaptionforSupplementaryVideo1CaptionsforSupplementaryTables1-10

OtherSupplementaryMaterials:

SupplementaryVideo1SupplementaryTables1-10


https://doi.org/10.1101/2020.03.04.974949


28

Methods

Animalwork

All animal work was approved by the local authorities (LAGeSo Berlin, license numbers

G0247/13 and G0243/18). T::H2B-mCherry/Sox2::H2B-Venus embryos were generated via

tetraploid complementation49. For embryo isolation, mice were sacrificed by cervical

dislocationanduteriweredissectedinPBS.

Mouseembryonicstemcellculture

AllmediumcompositionsarelistedinSupplementaryTable3.Allmouseembryonicstemcells

(mESCs)usedinthisstudyweremaleandfromanF1G4background50.mESCsweremaintained

on 6cm plates (Corning 430166) gelatinized with 0.1% gelatin (1:20 dilution of 2% gelatin

(SigmaG1393)intissue-culturegradeH2O)andcoatedwithmitoticallyinactiveprimarymouse

embryo fibroblasts (3-4x104cells/cm²) with standard mESC medium containing 15% FCS and

1000U/ml leukemia inhibitory factor (LIF,ChemiconESG1107)at37°Cand5% or7.5% CO2.

mESCs were split every second day with a dilution suitable to the proliferation velocity

(between1:5and1:9).ES+LIFmediumwasrefresheddaily.Forsplitting,mediawasaspirated

andcellswerewashedoncewithPBSandtrypsinized(Tryspin-EDTA(0.05%)(Gibco25300054))

for5-10’at37°C.Trypsinwasneutralizedby3mlES+LIFandcellscentrifugedfor5’at1000rpm,

after which the pellet was resuspended in ES+LIF. For freezing of mESCs, cell pellets were

resuspended inESmediumwith20%FCS,andmixed ina1:1 ratiowithES freezingmedium.

Cellswerefrozendowno/ninthe–80°Candtransferredtoliquidnitrogenthenextday.


https://doi.org/10.1101/2020.03.04.974949


29

Generationoftransgenicmouseembryoniccelllines

ThemouseembryoniccelllinesusedfortheexperimentswereallderivedfromF1G4cells.To

generate the fluorescent reporter constructs,wegenome-engineered themouseBACsRP24-

530D23 (T), RP23-249O15 (Sox2) and RP23-421P23 (Tbx6), containing∼180 - 230kb of the

C57BL/6mousegenomesurroundingtherespectiveloci,viaRed/ETrecombineering51.Inshort,

the starting codon (ATG) of the genes was replaced with a reporter gene containing H2B-

mCherry(forT)orH2B-Venus(forSox2/Tbx6),followedbytherabbitb-globinpolyadenlylation

signal and an FRT-flanked selection cassette (hygromycin or neomycin), driven by the Pgk

promotor.Forrandomintegration,5μgofthemodifiedBACswaslinearizedwithPI-SceI(NEB)

andelectroporatedinto3x106mESCs.SelectionwasperformedusingmESCmediumcontaining

250μg/ml neomycin and 150μg/ml hygromycin for the T::H2B-mCherry;Sox2:H2B::Venus

reporter line. For the Tbx6::H2B-Venus reporter line 350μg/ml hygromycin was used for

selection.Afterselection,singlecloneswerepicked,expandedandcheckedforBACintegration

by PCR. Two independent Tbx6::H2B-Venus mESC lines were generated. In both lines, the

endogenous Tbx6 locus was targeted by CRISPR (Supplementary Table 4) to create null

mutants.Tothisend,adoublenickingapproachwithfourguideRNAs52wasused.Thespecific

target sequences contained 5’-GN19NGG-3’’, withN being any arbitrary nucleotide,G being

guanine andNGG the Protospacer adjacentmotif andGN19 the guide RNA. If there was no

guanineat the5’-endof the templatesequence,anextraguaninehad tobeadded5’ to the

other 20 bases. Single strandedoligonucleotides (Supplementary Table 4)with an added 5’-


https://doi.org/10.1101/2020.03.04.974949


30

CACC-3’ at 5’-end and 5’-AAAC-3’ at 5’-end of the complementary strand were annealed in

10xT4 ligation buffer (PromegaM1801) by continuous cooling from 95°C to 25°C. Annealed

oligoswereclonedintotheBbsIsiteofpx335A_hCas9_D10A_G2P(giftfromDr.BorisGreber)

(Supplementary Table 4) containing expression cassettes for hCas9 nickase, guide RNA and

puromycinresitance.

Onedaypriortotransfection,3x105mESCscellswereseededonfibroblastcoated6-wellplates

(Costar3516)with3mlES+LIFperwell.Afterovernightincubation,themediumwasrefreshed

inthemorning.Mixesof110µlOpti-MEMreducedserummedium(Gibco31985062)plus25µl

Lipofectamine2000 (Thermo Fisher 2125239) and 125µl Opti-MEM plus 8µg per vector were

prepared.125µlofeachmixwere combinedand incubated for15minat room temperature

before being transferred to 1.25ml ES+LIF without Penicillin/streptomycin. After 5h of

incubation with the transfection mix, ESCs were split and plated on puromycin resistant

feeders.24haftertransfection,transientselectionwasstartedwithES+LIFcontaining8µg/ml

puromycin (Gibco10130127)atday1andday2andES+LIFcontaining4µg/mlpuromycinat

day3.Afterapproximatelyoneweek,singlecloneswerepickedandexpandedon96-wellplates

(Costar 3596). Clones were genotyped by PCR using primers spanning Exons 1-4. Deletion

breakpointswereanalyzedbySangersequencingofpurifiedPCRfragments.

Generationofgastruloids

Allmediumcompositionsare listed inSupplementaryTable3.Gastruloidsweregeneratedas

describedpreviously53,withsomeminormodifications.First,mESCswerefeederfreed.Tothis


https://doi.org/10.1101/2020.03.04.974949


31

end,mESCsweretrypsinizedonthefeederplateasdescribedabove,washedwithES+LIFand

resuspendedin2mlES+LIF.Onthreegelatinized(0.1%gelatin)wellsofa6-wellplate,cellswere

sequentiallyplated for25’,20’and15’duringwhichcellswerekept in the incubatorat37°C

and 5% or 7.5% CO2. With each transfer, cells were triturated to maintain a single cell

suspension.Feeder-freedmESCswerethenwashedoncein5mlPBScontainingMgCl2andNaCl

(Sigma A8412) and once in 5 ml NDiff227 (Takara Y40002). mESCs were then pelleted by

centrifugation for 5’ at 1000rpm and resuspended in 500µl of NDiff227. 10µl of the cell

suspensionwasmixedwith10µlofTrypanblue(Bio-Rad1450021)forautomatedcellcounting

withLunaAutomatedCellCounter.200-250livecellswerethenplatedinavolumeof30to40µl

NDiff227 into each well of a 96-well round bottom, low attachment plate (Cellstar 96 well

suspension cultureplate (655185)orCostar7007ultra-lowattachment96well plate (7007)).

Cellswere then allowed to aggregate for 48 h. After these 48h cellswere pulsedwith 3µM

CHIR99021 (CHIR,Merck Millipore) in 150µl NDiff227. Between 72 and 120h, medium was

refreshedevery24hbyremoving150µloftheoldmediaandaddingthesamevolumeofnew,

pre-incubated (37°C and 5% or 7.5% CO2) NDiff227. For gastruloids treated with CHIR and

CHIR+LDN,5µMCHIRwithorwithout600nMLDN-193189(LDN,Reprocell)wasaddedfrom96h

to120h.Forcontrols,anequalvolumeofdiluent(DMSO)wasadded.

Generationoftrunk-like-structures

Allmedium compositions are listed inData S3. The gastruloid protocol described abovewas

followed until 96h.Gastruloidswere then embedded in 5%Growth-Factor-ReducedMatrigel

(MG)(Corning356231).Tothisend,freshNDiff227mediumwaspre-incubatedforatleast20’


https://doi.org/10.1101/2020.03.04.974949


32

at37°Cand5%or7.5%CO2.Pre-incubatedmediumwasthenputonicefor5’,afterwhichMG

wasaddedtoachieveafinalconcentrationof5%intheculturewells.Mediumwasthenputon

roomtemperaturefor5’,duringwhich150µlofoldmediumwasremovedfromtheaggregates.

New medium with MG (150µl) was then added, and the cultures were returned to the

incubatorandfurtherculturedat37°Cand5%or7.5%.TLScultureswereallowedtosettlefor

at least 30’beforeproceeding to furtherexperimentation (e.g. live imaging). For TLS treated

withCHIRandCHIR+LDN,5µMCHIRwithorwithout600nMLDNwasaddedfrom96hto120h

priortoaddingtheMG.Forcontrols,anequalvolumeofdiluent(DMSO)wasadded.

Whole-mountImmunofluorescence

CollectedembryoswerewashedtwiceinPBSandthenfixedin4%PFAfor30’underrotationat

4°C,washed3xwithPBS,andstoredinPBSuntil immunofluorescentstainingwasperformed.

Gastruloidsortrunk-like-structureswerepickedusingap200pipettewiththetipcut-offatthe

50µlmark.GastruloidswerewashedtwicewithPBSandthenfixed in4%PFAfor1h inglass

vials(Wheaton224882)at4°Conaroller.Fortrunk-like-structures,individualstructureswere

pickedusingap200pipettewiththetipcut-offatthe50µlmark,andtransferredtoeither96-

wellplates(Costar3596)orIbidi8-wellglass-bottomplates(Ibidi80827).Trunk-like-structures

werewashedtwicewithPBS+MgCl2andNaCl+0.5%BSA(SigmaA8412),oncewithPBS,and

thenfixedin4%PFAfor1hrat4°Conarockingplatform.Subsequently,gastruloidsortrunk-

like-structureswerewashedtwiceinPBSfor5min,permeabilizedby3x20’incubationin0.5%

Triton-X/PBS (PBST) andblocked in 5% fetal calf serum/PBST (blocking solution) overnight at

4°C. For antibody staining, gastruloids/trunk-like-structures were transferred to Ibidi 8-well


https://doi.org/10.1101/2020.03.04.974949


33

glassbottomplates.Primaryantibodyincubationwasperformedinblockingsolutionfor48-72h

at 4°C, after which gastruloids/trunk-like-structures were washed three times with blocking

solution and three times with PBST. After the last washing step, Gastruloids/trunk-like-

structureswereincubatedinblockingsolutiono/nat4°C.Thenextday,secondaryantibodies

diluted in blocking solution were added, and structures were incubated for 24h at 4°C.

Afterwards, gastruloids/trunk-like-structureswerewashed three timeswith blocking solution

and three timeswith PBST. The last PBSTwashing step after secondary antibody incubation

includedDAPI (0.02%,RocheDiagnostics10236276001).DAPIwas incubatedo/nandwashed

offoncewithPBST.AllprimaryandsecondaryantibodiesarelistedinSupplementaryTable3.

Tissueclearing

Prior to imaging, embryos, gastruloids and trunk-like-structures were cleared with RIMS

(refractive indexmatchingsolution).Tothisend,sampleswerewashedtwicewithPBSfor10

min,post-fixedin4%PFAfor20minutesandwashedthreetimeswith0.1Mphosphatebuffer

(PB,0.025MNaH2PO4and0.075MNa2HPO4,pH7.4).Clearingwasperformedbyincubationin

RIMS(133%w/vHistodenz(Sigma-AldrichD2158)in0.02MPB)onarockingplatformat4°Cfor

atleastonetoseveraldays.

Imaging

Gastruloidsandtrunk-like-structuresstainedwithantibodiesorcarryingfluorescentreporters

were imaged with the Zeiss AxioZoom v16 (wide-field), Celldiscoverer7 (wide-field), Zeiss

LSM710(laser-scanningmicroscopewithAiryscan)orZeissLSM880(laser-scanningmicroscope

withAiryscan)withappropriate filters formCherry,Venus,DAPI,AlexaFluor488,AlexaFluor


https://doi.org/10.1101/2020.03.04.974949


34

568,AlexaFluor647,andcombinationsthereof.EmbryoswereimagedwithZeissLightsheetZ1

withappropriatefiltersformCherryandVenus.Whole-mountinsituhybridizationimageswere

acquiredwithZeissAxioZoomv16.Post-acquisitionimageprocessingwasperformedusingZen

and/orArrivis.ForliveimagingofTLSin96-wellplatesweusedtheCelldiscoverer7(Zeiss)with

incubator chamber temperature set at 37°C and CO2 content at 5%. Acquisition intervals

rangedfrom15-30minutes.

Morphometricanalysisofgastruloidsandtrunk-like-structures

ForquantificationofthenumberofstructureswithaT+-pole,weemployedtheT::H2B-mCherry

reporterline.Ifastructuredisplayedmultipleaxesofelongation,itwassufficientforoneend

to have a T::H2B-mCherry+ pole to be scored as having a “T-pole”. We defined

gastruloids/trunk-like-structuresas“segmented”whenatleastfourneighboringsegmentshad

developedalongtheanteroposterioraxisat120h.Finallyifastructuredisplayedtwoormore

axesofelongation,itwasscoredas“multipleaxes”,ifshowingjustoneas“oneaxes”,ifnone

“no elongation”. For all conditions, three independent experiments, eachone including each

treatment,wereanalyzed.Structures thatgrewoutof focus inaway that theycouldnotbe

ratedforoneofthecategorieswereexcludedfortherespectivecategory.

FlowcytometryandFACS

Individual trunk-like-structureswerewashedwith PBS+0.5%BSA 2 times, afterwhich 50μl of

trypsin was added. Trunk-like-structures were then incubated for 10’ at room temperature,

afterwhichsamplesweredissociatedbytriturationfor50timesusingap200pipettetoachieve

ahomogenoussuspension.Thereactionwasthenstoppedwiththeadditionof100μlof5%BSA


https://doi.org/10.1101/2020.03.04.974949


35

in PBS. Before cell counting and/or sorting on a FACS Aria II (Becton Dickinson) or a BD

FACSCelestaFlowCytometer(BectonDickinson,countingonly)thecellsuspensionwasfiltered

througha35μmmesh.Equalnumbersofsamplesforeachconditionwereharvested inthree

independent experiments. For FACS, samples were dissociated as described above, and the

Tbx6VE+andTbx6VE-fractionwassortedin1.5mllow-bindingtubeswith350μlRLT(Qiagen)+1%

v/vB-mercaptoethanol(Sigma).FlowcytometrydatawerelateranalyzedusingFlowJoV10.

RNAisolation,reversetranscriptionandquantitativePCR

RNA isolation was performed using the RNeasy Micro Kit (Sigma 74004) according to the

manufacturer’s instructions with the followingmodifications: genomic DNAwas digested on

columnwith the addition of an extra 1μl of RNase-freeDNase I (Roche) to ensure complete

digestion, and after the 80% ethanol columnwash and column centrifugation the remaining

ethanol was removed with a p10 pipette tip and columns were left to air-dry for 5’.

Quantitative reverse transcriptase PCR (qRT-PCR) was then performed using a two-step

protocol. First, RNA was reverse transcribed using the Quantitect Reverse Transcription Kit

(Qiagen),accordingtothemanufacturer’sinstructions.qPCRwascarriedoutonaStepOnePlus

Real-Time PCR System (Life Technologies) using GoTaq qPCR Master Mix (Promega) with

validatedgene-specificprimers(SupplementaryTable4).FoldchangewascalculatedfromΔCt

usingTbpashousekeepinggene(SupplementaryTable5).

Measuringofsegmentsize

Segmentsizesweremeasured inFiji1.8.0usingthe imagesacquired in theTmChchannel.For

each structure, the four newest neighboring segments were measured counting from the


https://doi.org/10.1101/2020.03.04.974949


36

posteriorend.Incaseofbilateralsegments,onlysegmentsononesideweremeasured.Forthe

‘bunches of grapes’ in TLSC and TLSCL, four segments along the antero-posterior axis were

measured counting from the posterior end. Length was defined as parallel to the anterior-

posterioraxis,widthwasdefinedasperpendiculartotheanterior-posterioraxis.

Whole-mountinsituhybridization

Probesynthesis

For synthesis of DIG labelled RNA probes, plasmids containing the cDNA of interest and

promoters for sense and anti-sense strand synthesis were used from theMAMEP database

(http://mamep.molgen.mpg.de). In order to obtain a sufficient amount of material, some

plasmidswerefirstretransformedintoE.coli(DH5α)andafterwardsisolatedwithamini-prep

kit (Qiagen) according to the manufacturer’s instructions as described below. To verify the

identityofthevectorandinsertofeachplasmid,plasmidsweredigestedbyrestrictionenzymes

andloadedona1.0%agarose/TAEgel.Probesweresynthesizedbyinvitrotranscriptionusinga

PCRproductofthedesiredcDNAorthelinearizedplasmid.

RetransformationinDH5αandmini-prep

DH5αcellswerethawedoniceand1µloftheplasmidwasaddedto100µlofcompetentcells

andincubatedonicefor30’.Subsequently,cellswereheatshockedat42°Cfor45”inawater

bathandthen immediatelycooleddownon ice.500µlofLBmediumwasaddedandthemix

was incubated for 1h at 37°C (heating block) under shaking. Then, 25µl and 250µl of the

mixture were plated on separate Agar-plates containing Ampicillin (Amp) for selection of


https://doi.org/10.1101/2020.03.04.974949


37

transformed bacteria (all vectors carried a gene for Ampicillin resistance). The plates were

incubated overnight at 37°C. The next day, single colonies were picked in 5ml LB+Amp and

incubated overnight at 37°C under shaking. Plasmids were isolated with the QIAprep Spin

MiniprepKit (Qiagen) followingthe instructionsoftheQuick-StartProtocolusingacentrifuge

forprocessing.DNAconcentrationwasmeasuredwithaNanophotometer(Implen).

RestrictionDigest

Afterplasmiddigestion(250ngoftheplasmid,2µl10xbuffer,0.2µl100xBSAand2µlofeach

restrictionenzymeina20µltotalvolume),expectedbandsizedforthevectorsandinsertswas

confirmedona1%Agarose/TAE-gel,stainedwithSybrSafe(1:20.000).

Polymerase-Chain-Reaction(PCR)

ForPCR,5µlPlasmid(1ng/µl)wasusedina50µltotalreactionvolumecontaining5µl10xPCR

buffer (Invitrogen), 1.5µlMgCl2 (Invitrogen), 10µl dNTPMix (10mMeachnucleotide), 0.25µl

forward Primer (100 µM), 0.25µl reverse Primer (100 µM) and 0.2µl Taq DNA Polymerase

(5u/µl)(Invitrogen).FortheplasmidofUncx4.1adifferentPCRstrategywasused,becauseof

itsCGrichness.Here,the50µlreactioncontained5µlPlasmid(1ng/µl),1µldNTPS,0.25µlU5,

0.25µlL2,0.2µlTaqDNAPolymerase(5u/µl)(Qiagen)and10µl5xQ-solution(Qiagen).Primers

used for the respective vectors are listed in Supplementary Table 4. All PCR products were

checkedona1%Agarose/TAE-gel.

Linearizationofplasmidsforinvitrotranscription


https://doi.org/10.1101/2020.03.04.974949


38

Plasmidswerelinearizedwitharestrictionenzyme(1.5µgDNA,2.5µl10xbuffer,filledupwith

DEPC-H2Oto25µl).The reactionwas incubated for1hat37°C.Subsequently,8µlammonium

acetate (10M)and80µl ice-cold100%ethanolwereadded, followedbycentrifugationof the

sample for30’at4°Cand13.200rpm.Thesupernatantwasthenremovedand150µlof70%

ice-cold ethanol was added. Sample was centrifuged again for 10’, the supernatant was

removed,andthepelletwasairdriedanddissolvedin9.5µlDEPC-H2O.

Invitrotranscription

Forinvitrotranscription,9µlofPCRproductor1.5µgoflinearizedplasmidwasincubatedwith

3µl10xTranscriptionbuffer,3µlACGnucleotides (4mMeachnucleotide),0.75µldigUTP-UTP

Mix (4mM),1.5µlDTT,1µlRNase inhibitorand60U of the respectiveRNApolymerase, ina

totalreactionvolume(filledupwithDEPC-H2O)of30µl.Thereactionwas incubatedfor2hat

37°C(T7,T3)or40°C(Sp6).Subsequently,3µlofRNase-freeDNaseI(10u/µl)wasaddedandthe

reactionwasincubatedfor15’at37°C.TheRNAprobewasthenpurifiedwithProbeQuantG50

Sephadex columns (Pharmacia).Adding20µlDEPC-H2O to theprobe, spinning it through the

resinofthecolumnandaddingagain30µlDEPC-H2Oincreaseditsvolumeto100µl.Aliquotsof

30µlwereimmediatelyputondryiceandstoredat-80°C.TheRNAprobeswerecheckedona

1%Agarose/TAE-gel.

Fixationandmethanolseries

Collectedembryos, gastruloidsand trunk-like-structureswerewashed twice inDEPC-PBSand

then fixed in 4% PFA overnight at 4°C. The next day, samples were transferred into 100%

methanolviaamethanolseries, includingtwowashingstepsinDEPC-PBSandatransferfrom


https://doi.org/10.1101/2020.03.04.974949


39

25%to50%to375%to100%methanol/DEPC-PBS(10’each).Upontransferto100%methanol,

sampleswerewashedtwicein100%methanolandstoredin100%methanolat-20°C.Allsteps

wereperformedforatleast10’at4°C.

Pre-hybridization&Hybridization

Unless stated differently, all steps were performed for 10’ at 4 °C under rocking. For the

composition of the solutions used for in situ (Pre-hybridization, Hybridization, Post-

hybridization washes and antibody incubation, Post-antibody washes, Staining) we refer to

SupplementaryTable3.First,sampleswerepre-hybridizedbytransferringtoDEPC-PBSTviaa

reverse methanol series (100%, 75%, 50%, 25% for 10’ each). Subsequently, samples were

incubated in6%H2O2/DEPC-PBSTat4°C, trunk-like-structures/gastruloid for10’andembryos

for 20’, followed by three washes with DEPC-PBST. ProteinaseK/DEPC-PBST (10µg/ml)

treatment was performed at 4°C, (7’ for trunk-like-structures/gastruloids, 10’ for embryos).

ProteinaseK activity was quenched with Glycine/DEPC-PBST (2mg/ml) and two washes with

DEPC-PBST. Subsequently, sampleswerepost-fixed in0.2%Glutamine/4%PFA/DEPC-PBS for

30’ at room temperature and then washed twice with DEPC-PBST. Samples were then

incubated with 68°C pre-warmed Hyb for 15’ at room temperature after which Hyb was

refreshedandstructuresincubatedfor2morehoursat68°C.IfnotimmediatelyusedforRNA

probehybridization,thesampleswerepre-cooledfor15’atroomtemperatureandstoredat-

20 °C. Prior to hybridization with the RNA probe, the samples and Hyb solution were pre-

warmedat 68°Cand incubated in freshHyb for 15’ at 68°C.Meanwhile, theRNAprobewas

dilutedinHyb(200ng/ml)andpre-heatedfor13’at80°Cinaheatingblock.Thesampleswere


https://doi.org/10.1101/2020.03.04.974949


40

thenincubatedwiththeRNAprobeat68°Covernight.TheHybsolutionofthe15’ incubation

stepwasstoredat-20°Cforthefirstwashingsteponthenextday.Allstepswereperformed

underrocking.

Post-hybridizationwashes,antibodyincubationandpost-antibodywashes

Thenextday, sampleswerewashedoncewithHyb (stored fromdaybefore) for30’at68°C,

twicewithSolution1for30’at68°C,twicewithSolution3Tfor30’at68°C,twicewithSolution

3Tfor1hat68°CandthreetimeswithTBSTfor15’atroomtemperature.Duringthe1hwashes

withSolution3T,antibodysolutionwasprepared.TBSTwithonegrainofembryopowderwas

pre-heatedfor30’at70°Cinawaterbathandcooleddownonice.Subsequently,1%v/vlamb

serumand0.2%v/vAnti-DIGantibody(Roche)wereaddedandincubatedfor1hat4°Cinthe

darkwhilerocking.Themixwascentrifugedfor10’at4000rpmat4°Cand1%lambserum/TBST

wasaddedtothesupernatant(finalantibodyconcentration1:2000).Afterfinishingthewashing

steps,thesampleswereblockedwith10%lambserum/TBSTfor2.5hatroomtemperatureand

incubatedwiththeantibodysolutionovernightat4°C.Allstepswereperformedunderrocking.

Thenextday,sampleswerewashedtwicewithTBSTfor15’,twicefor30’andsixtimesfor1h,

allatroomtemperature.ThefinalwashingstepwasperformedovernightinTBSTat4°C.

Staining

SampleswerewashedfourtimesinfreshlypreparedNTMTfor15’atroomtemperatureunder

rocking. In themeantime, the staining solutionBMPurple (Roche)was pre-warmed at room

temperatureandcentrifugedfor1’at13.200rpm.Thesupernatantwasthenusedforstaining

thesamplesand incubateduntila clearandspecific signalappeared.The first15’of staining


https://doi.org/10.1101/2020.03.04.974949


41

were performed under rocking, afterwards without. For stopping the staining reaction, the

samples were washed once with NTT and twice with PBST, for at least 10’ each at room

temperatureunder rocking. The sampleswere stored in4%PFA/PBSat4°C. Embryos, trunk-

like-structuresandgastruloidswereimagedwiththeAxioZoom.V16(Zeiss).

Single-celltranscriptomeprofilingofTLSs

96h, 108hTLS and120hTLSwere generatedasdescribedabove. For 96h, 6 structureswere

selected based on the presence of a TmCh+ pole and absence ofmultiple axes formation; for

108h TLS, 5 structures were selected based on the presence of a TmCh+ pole, one axis of

elongation,andinitiationofneuraltubeformation,butnosegmentationintheTmCh+domain;

for 120h TLS, 3 structureswere selected based on the presence of a TmCh+ pole, one axis of

elongation, clear formation of a neural tube Sox2Ve+ domain and segmentation in the TmCh+

domain.TLSswerepickedwithap200withthepipettetipcut-offatthe50µlmark,andserially

washedthroughpipettetransferring(cut200µltip)inwellsfilledwith200µlof1xPBS/0.4%BSA

(5transfers)togetridoftheMatrigel.TLSsofthesametimepointwerethenpooledtogether

anddissociated in200µl TrypLEExpress (Gibco) for 15minutes (96h), 20minutes (108hTLS)

and25minutes(120hTLS)at37ºC,withpipettingevery5minintervals.Thecellsuspensionwas

filtered using Scienceware Flowmi Cell Strainers, 40µm. Cells were washed twice with 1ml

1xPBS/0.4%BSAwithcentrifugationstepsperformedfor5minutesat1200rpminlowDNAbind

Eppendorf tubes. The cell concentration was determined using a hemocytometer and cells

weresubjectedtosingle-cellRNAsequencing(10xGenomic,Chromium™SingleCell3’v3;one

reaction per timepoint has been used) aiming for a target cell recovery of up to 10,000


https://doi.org/10.1101/2020.03.04.974949


42

sequenced cells per sequencing library (timepoint). Single-cell libraries were generated

according to the manual, with one modification: fewer PCR cycles (n=8) were ran than

recommended during cDNA amplification or library generation/sample indexing to increase

librarycomplexity.Librariesweresequencedwithaminimumof230millionpairedendreads

accordingtoparametersdescribedinthemanual.

BulkRNA-seqofgastruloidsandtrunk-like-structures

96h aggregates, gastruloids, TLS, TLSC and TLSCL were generated as described above. We

sequenced 3 biological replicates per condition (96h aggregates, Gastruloids, TLS, TLSC and

TLSCL). For 96h, 10 structures per replicatewere selected (see previous section for selection

criteria)andpooled; forGastruloids,6 structuresper replicatehavebeenselected (basedon

thepresenceofaTmCh+poleandoneaxisofelongation)andpooled;forTLS,TLSCandTLSCL,6

structuresperreplicatewereselected(seeprevioussectionforTLSselectioncriteria;TLSCand

TLSCL were selected based on the presence of of a TmCh+ pole, elongation along one axis,

segmentationinbunchesofgrapesattheanteriorTmCh+domain)andpooled.Allsampleswere

washed twice with 1xPBS/0.4%BSA. Then 350µl of RLT Plus buffer containing 1% β-

mercaptoethanol (Thermo) was added to dissociate the structures and lyse the cells. After

pipette dissociation and vortexing, samples were frozen at -80C. The next day, RNA was

extracted using RNeasy Plus Micro Kit (Qiagen) and RNA concentration and quality was

measured using the Agilent RNA 6000 Pico kit on an Agilent 2100 Bioanalyzer. All samples

analyzedhadaRINevaluehigherthan8.0,andweresubsequentlyusedforlibrarypreparation.

mRNAlibrarieswerepreparedusingKAPAStrandedRNA-SeqKit(KapaBiosystem)accordingto


https://doi.org/10.1101/2020.03.04.974949


43

themanufacturer’s instructions. 500ng of total RNAwas used for each sample to enter the

librarypreparationprotocol.Foradapter ligationdual indexeswereused (NEXTFLEX®Unique

Dual IndexBarcodesNOVA-514150)ataworkingconcentrationof71nM(5µlof1uMstock in

each70µlligationreaction).Qualityandconcentrationoftheobtainedlibrarieswasmeasured

usingAgilentHigh SensitivityD5000 ScreenTapeon anAgilent 4150TapeStation.All libraries

were sequenced using 75bp-paired end sequencing (150 cycles kit; FC-410-1002) on a

HiSeq4000platformataminimumof23,7millionfragmentspersample.

ExpressionprofilingofTLSvariabilityusingNanoString

mESCs,mouseembryosandTLSweregenerated,selectedanddissociatedasdescribedabove.

FormESCs,atotalof20,000cellswereharvestedandfrozenin350µlRLTPlusbuffercontaining

1%β-mercaptoethanol(Thermo).Threeembryosforeachdevelopmentalstage(E8.5andE9.5)

weredecapitatedinordertoobtainthepost-occipitalportion,pooled,dissociated,andfrozen

in350µlRLTPlusbuffercontaining1%β-mercaptoethanol.For120hTLS,9individualstructures

havebeenselected(basedontheselectioncriteriaoutlinedbefore),dissociatedindependently

andfrozenin350µlRLTPlusbuffercontaining1%β-mercaptoethanol.RNAwasisolatedfrom

the 12 samples in parallel using RNeasy PlusMicro Kit (Qiagen) and RNA concentrationwas

measuredusingQubit™RNAHSAssayKit.

Toprofiletheexpressionof41genesand4housekeepinggenes(Polr1b,Hprt,Abcf1,Gusb),a

rangeof 36-161ng total RNA/samplewere used in aNanoString nCounter Element assay to

profilethe12 individualsamples.Probehybridizationwassetupaccordingtomanufacturer’s

instructions and performed for 24 hours (MAN-10040-05). Reactions were ran on the


https://doi.org/10.1101/2020.03.04.974949


44

NanoStringnCounterSPRINTInstrument.Falsenegativeprobesdetectedupto5counts,which

informedthemagnitudeofpotentialfalsenegativesignal.Thus,6countswereconservatively

removed from all measurements. Positive control probes normalization step was applied

(geometric mean) and finally the combination of the four housekeeping genes was used to

obtain the final normalized counts table. To assess variability in the expression of individual

marker genes, the log2(MAX(all_samples)/MIN(all_samples))was used as a proxywith a high

valuerepresentinghighvariabilityandviceversa.SeeSupplementaryTable6forprobesdesign

andsequencesandDataS7fornormalizedgenecounts.

Computationalanalysis

Ifnotstatedotherwise:AllstatisticsandplotsaregeneratedusingRversion3.6.0“Plantingofa

Tree”andSeuratversion3.054.

Single-celltranscriptomeprofilingofTLSs

Preprocessing

TheCellRangerpipelineversion3(10xGenomicsInc.)wasusedforeachscRNA-seqdatasetto

de-multiplextherawbasecallfiles,generatethefastqfiles,performthealignmentagainstthe

mousereferencegenomemm10, filter thealignmentandcountbarcodesandUMIs.Outputs

frommultiplesequencingrunswerealsocombinedusingCellRangerfunctions.

Qualitycontrol


https://doi.org/10.1101/2020.03.04.974949


45

The initialquality controlwasperformedusing scanpy55.Cellswith less than10,000ormore

than40,000counts,amitochondrial-fractionabove0.1andlessthan3,000geneswereflagged

asinsufficient.

Clusterdetermination

Single cell data created for the threedevelopmental timepoints (96h, 108h and120h)were

loaded to Seurat54,with aminimum requirementof 200 features and3 cells and filtered for

previously flagged barcodes. Subsequently, the expression data were independently

normalized,variablefeaturesweredetectedandlog-normalizedandscaledto10,000(default

settings).Next,fordownstreamintegrationofthethreetimepointssequenced,aPCAwasrun

foreachtimepointpriortointegrationanchorsetdetection(reduction="rpca",dims=1:30).

Finally, these integrationanchorswereusedto integratethedatapointsusingthepreviously

calculated anchor sets. A list of cell cyclemarkers loadedwith Seuratwas used to cell cycle

scoreallcellsandsubsequentlyrunthedefaultworkflowforscalingwithvars.to.regresssetto

cellcyclescoresforSandG2Mphase(ExtendedDataFig.15a).Fordownstreamanalysisand

visualizationofthe integrateddataset,aPCAwasruntothencalculatea jointUMAP(dims=

1:30, n.neighbors = 10). Standard workflow steps were applied for cluster generation

(FindNeighbors, dims = 1:20 and FindClusters, resolution = 0.5), resulting in a total of 15

clusters.Finally,twosmallclusters(Seurat_10andSeurat_11)wereremovedduetopresence

of stressed cells (high mitochondrial RNA counts and low total RNA counts) (Seurat_10;


https://doi.org/10.1101/2020.03.04.974949


46

ExtendedDataFig.15b)anddoublets(shownbyalmostdoubleamountoftotalRNAscounts

and UMIs) (Seurat_11; Extended Data Fig. 15c). All remaining clusters show a similar

distribution of average UMIs and genes detected per cell (Extended Data Fig. 15d). See

SupplementaryTable8formarkergeneslistforeachidentifiedcluster.

Subclusteringofsomiticandneuralcells

For subclustering of somitic cells, data were first split by sampled timepoints and somitic

clusters (“Somite (0)”, “Somite”, “Somite (Dermo)”, “Somite (Sclero)”) were extracted.

Subclusteringwas thenperformed inScanpy (resolution=0.3). For subclusteringof theneural

tubecluster,cellsassignedto“NeuralTube2”fromTLSwereextracted.Subclusteringwasthen

performedinScanpy(resolution=0.65).

Comparisonwithmouseembryo

Apreviouslyestablishedreferencedatasetofmousepost-implantationdevelopment(E6.5to

E8.5)35 was utilized for gene expression comparison to in vivo data as well as cell type and

proportion comparisons. The in vivo data were filtered to include only the relevant embryo

time points (E.7.5, E8.0 and E8.5) all extra-embryonic and occipital cell-types (that are not

generatedingastruloids3andTLS(thisstudy))wereremoved.SeeSupplementaryTable8for

markergeneslistforeachidentifiedcluster.Both,themouseinvivoandtheTLSsinvitro,data

sets were normalized in parallel (SCTransform, default settings). To adjust for different

resolutionsincelltypedetectionbetweentheembryoandTLSdatasets,the invivo“somites”


https://doi.org/10.1101/2020.03.04.974949


47

and “PSM” cell states were subclustered (resolution=0.15) to identify a cell cluster with an

anterior PSM (aPSM) signature (Extended Data Fig. 15e,f). These cells were relabeled

accordingly and subsequently treated as an individual cell type. The subclustering step also

generated two clusters from the in vivo PSM original cell state, but due to marker genes

similarity this subclustering was not taken into account in the further downstream analysis

(ExtendedDataFig. 15e,f). Prediction scoreswereusedasmeasurement for certainty in cell

typematching("certaintyscores").Scoresshownaremeansacrossallcellsevenifnotassigned

to the respective cluster. Finally, the in vivomouseand in vitro TLS cell typeswerematched

using an integrated reference, by first finding of transfer anchors, followed by data transfer

usingtheanchorsandfinallythepredictedcelltypeswereusedasinvivo/invitromatchedcell

typecounterparts.Fordifferentialexpressionandconservedmarkercalculationbothdatasets

were merged and genes located on the sex chromosomes were excluded to avoid biases

resultingfromcomparisonofonlymaleTLSswithinvivocellsfrommaleandfemaleembryos.

Analysisofconservedmarkergenes

TheconservedmarkergeneswascalculatedontheintegrationoftheTLSandembryodataas

describedaboveusing theFindConservedMarkers functionwithdefault setting.Cut-offsused

weresettoaminimumpercentagedifferenceof25%(forpercentofcellsexpressingthegene

ofinterest(GOI)),andanabsoluteminimumlog2FCof1.SeeSupplementaryTable9forlistof

top25conservedmarkergenes.

AnalysisofdifferentiallyexpressedgenesbetweenTLSandembryo


https://doi.org/10.1101/2020.03.04.974949


48

TLS and embryo data were integrated as described above. For the analysis, the in vivo cell

stateswereusedasareference,andcomparedtotheTLScellsattributedtothatcluster.Tothis

end,thepercentofcellsexpressingageneaswellastheaverageexpressionforthatgene in

theinvivocellstatesaswellastheinvitrocellsattributedtothatclusterwascomputedusing

the FindMarkers function with default settings. Differential expression between TLS and

embryowastested foreitherall thegenesexpressed in theclusterof interest,or the20top

markergenesofeach invivocellstates.Cut-offsusedwereaminimumpercentagedifference

of25%(forpercentofcellsexpressingthegeneof interest (GOI)),andanabsoluteminimum

log2FCof1.SeeSupplementaryTable9forlistofgenomewidedifferentiallyexpressedmarker

genes.

RNAvelocityandpseudo-timeanalysis

RNAvelocitywascalculatedusingthevelocytotool31,56andvisualizedusingscanpy55.Trajectory

inference and pseudo-time analysis were done using the scanpy package PAGA57. The

previously calculatedUMAPand cell cluster assignmentwasused for velocity projection and

trajectoriesforpseudo-time.Forpseudo-timeanalysis,somiticclusters(“Somite(0)”,“Somite”,

“Somite (Dermo)”,“Somite (Sclero)”)weremerged, theputativeNMPclusterwassetas root

and108hand120hcellswereanalyzedseparately.

BulkRNA-seqofgastruloidsandtrunk-like-structures

Processingandexpressionlevels


https://doi.org/10.1101/2020.03.04.974949


49

RNAseqtriplicatesofthedifferentgrowthconditions96h,Gastruloids120h,TLS,TLSCandTLSCL

werepre-processedusingcutadapt58toremoveadapterandtrimlowqualitybases.Readswere

subsequently aligned against the reference genome mm10 using STAR59 (parameter:

outSAMtype BAM SortedByCoordinate --outSAMattributes Standard --outSAMstrandField

intronMotif --outSAMunmapped Within --quantMode GeneCounts). Finally, Stringtie60 was

used for transcript assembly, e.g. calculation of strand-specific TPMs. The three biological

replicateshighlycorrelatebetweeneachother(ExtendedDataFig.15g).

Differentialgeneexpression

Differential expression analysis was done independently per group comparison using the R

packageDESeq261usingtherawexpressioncountsfromSTAR'sreadspergeneoutput.z-scores

were calculated according to the formula ((VALUE(sample)-

AVERAGE(all_samples))/STDEV(all_samples). TPM values for all genes can be found in

SupplementaryTable10.

GeneSetEnrichmentAnalysis(GSEA)

GSEA was performed using the R package “fgsea”62 (minSize=5, maxSize=500, number of

permutations=10.000) takingall genes statisticallydifferentiallyexpressedbetweenTLSand

gastruloids(FDR<0.05,nolog2FCcut-off).

Dataavailability


https://doi.org/10.1101/2020.03.04.974949


50

Bulkandsingle-cellRNA-SeqdatahavebeendepositedintheGeneExpressionOmnibus(GEO)

underaccessioncodeGSE141175.

Codeavailability

Allcomputationalcodeusedinthisstudyisavailableuponrequest.


https://doi.org/10.1101/2020.03.04.974949


51

ExtendedDataFig.1Inductionofembryo-likemorphologyintrunk-like-structures

a,SchematicofT::H2B-mCherry;Sox2::H2B-VenusdoublereportermESCs(upperpanel)and invivovalidationby

tetraploid complementation (bottom panel). Optical sections of light-sheet imaging of a E9.0 mouse embryo

caudalend(bottom-leftpanel)andofsomitesandneuraltube(bottom-rightpanel).Scalebar50μm.b,Wide-field

fluorescentandbright-fieldimagesof120hgastruloids.Scalebar200μm.c,Wide-fieldfluorescentandbright-field

images of 120h TLS.d,Wide-field fluorescent and bright-field images of 120h gastruloids treatedwith CHIR or

CHIR+LDNfrom96to120hshowingnosignsofsegmentationintheabsenceof5%matrigel.e,Quantificationsof

gastruloidsfromconditionsind.TreatmentofgastruloidswithCHIRandLDNdoesnotsignificantlyaltertheaxis

formation and induction of a TmCH+ pole, and is not able to induce segmentation. f,Wide-field fluorescent and


https://doi.org/10.1101/2020.03.04.974949


52

bright-field images of 120h TLSC.g,Wide-field fluorescent andbright-field images of 120h TLSCL. Inb-dand f,g,

scalebaris200μm,exceptformagnifications,wherescalebaris50μm.h,QuantificationofsomitesizeinTLSand

embryosusingFiji (seeSupplemental Information)revealsthatTLSandTLSCsomiteshaveasimilar lengthbuta

slightlysmallerwidthascomparedtotheembryo,resultinginanoverallslightlyreducedsegmentarea. InTLSCL,

both somite length and width are smaller than in vivo. Boxes indicate interquartile range. End of whiskers

representminimumandmaximum.Symbolsindicateindividualsomite.Centrallinerepresentthemedian.


https://doi.org/10.1101/2020.03.04.974949


53

ExtendedDataFig.2Formationofgut-like-structuresintrunk-like-structures

a, 3Dmaximum intensityprojectionofTLS (leftpanel) stained for FOXA2 (blue) to reveal thegut-like-structure.

Scale bar 50μm. b, Confocal sections showing the FOXA2+ (upper panel) and SOX17+ (bottom panel) gut-like

structure inTLS.NoteFOXA2+/TmCH-highcellsat thebaseof thegut (whitearrowheads).Redarrowheads indicate

somites. Scale bars 25μm. c, Confocal section showing gut-like structure in TLSC. d, 3D maximum intensity

projectionshowinggut-likestructureinTLSCL.Inthemagnifications(rightpanels)inc,dwhitearrowheadsindicate

FOXA2+/TmCH-highcellsatthebaseofthegut.Redarrowheadsindicateguttubularcavity.


https://doi.org/10.1101/2020.03.04.974949


54

ExtendedDataFig.3BulkRNA-Sequencingofgastruloidsandtrunk-like-structures

a, Schematic of experimental set-up. N=number of pooled structures per replicate, n=number of replicates. b,

Heatmap of expression for an extended panel of selected genes associated with development of indicated


https://doi.org/10.1101/2020.03.04.974949


55

embryonic structures in 96h aggregates, gastruloids, TLS, TLSC and TLSCL, as measured by RNA-seq. TPM,

Transcripts PerMillion. CE, caudal end,NMP, neuromesodermal progenitors, PSM, presomiticmesoderm, LPM,

lateralplatemesoderm, IM, intermediatemesoderm.c,Heatmapwithscaledexpression(rowz-score)ofknown

markergenesofposteriorpresomiticmesoderm(pPSM)8.Scoresareanaverageofthe3 independentreplicates

percondition.d,Heatmapwithscaledexpression(rowz-score)ofneuraltubeknownmarkergenes.Scoresarean

averageof the3 independent replicatesper condition. Boxplot showsdistributionof z-scores (per column) for

differentconditions.Dotsindicateoutliers.e,Boxplotsshowingdistributionofz-scoresforknownmarkergenesof

indicated tissue layers. Endo, Endoderm; IM, IntermediateMesoderm; LPM, Lateral PlateMesoderm;PSM,Pre-

SomiticMesoderm.Boxes indicate interquartile range.Endofwhiskers representminimumandmaximum.Dots

indicateoutliers.Centrallinerepresentthemedian.ListofgenesusedforeachcategoryinSupplementaryTable

1. f,Heatmapof scaledexpression (rowz-score)of somiteepithelialization factors ingastruloids,TLS,TLSCand

TLSCL. Scores are an average of the 3 independent replicates per condition.g, Schematic overview of signalling

factors involved in somitecompartmentalization invivo.M,medial; L, lateral;D,dorsal;V,ventral. h,Heatmap

with scaled expression (row z-score) of signalling factors depicted in g (top panel) and heatmap with scaled

expression of marker genes for different somite compartments (bottom panel). A, anterior; P, posterior; M,

medial; L, lateral;D, dorsal; V, ventral. Scores are an averageof the 3 independent replicates per condition. i,

GeneSetEnrichmentAnalysisenrichmentplotsofselectedsignificantpathways.j,Up-anddownregulatedgenes

involved inembryo&tissuemorphogenesis (relatedtoFig.2e).Dotsizescaleswith log2ofabsoluteexpression.

Reddots, log2FC>2andpadj<10e-15.Greendots, log2FC>2.Upperdotted line,padj=10e

-15; bottomdotted line,

padj(FDR)=0.05. Green label, involved in somitogenesis; orange label, involved in blood vessel development;

yellowlabel,involvedinboth.


https://doi.org/10.1101/2020.03.04.974949


56

ExtendedDataFig.4ExpressionofcelladhesionmoleculesingastruloidsandTLS

a,Heatmapwithscaledexpression(rowz-score)ofcadherins,protocadherins,andephrinsingastruloidsandTLS.

Boxplotsshowdistributionofz-scores(percolumn)fordifferentsamples.Boxesindicateinterquartilerange.End

ofwhiskers representminimumandmaximum.Dots indicate outliers. Central line represent themedian. Every

columnineachheatmaprepresentoneofthethreebiologicalreplicates.b,Heatmapofscaledexpressionofgenes


https://doi.org/10.1101/2020.03.04.974949


57

in a with significantly differential expression (FDR<0.05). c, Bar graph showing expression of integrin ligands

fibronectin (Fn1), lamininalpha-1 (Lama1), lamininalpha-5 (Lama5)andcollagen4a1 (Col4a1) ingastruloidsand

TLS.Bargraphsrepresentaverageof3samples.Errorbarsrepresentstandarddeviation.


https://doi.org/10.1101/2020.03.04.974949


58

ExtendedDataFig.5NanoStringanalysisofindividualTLS

a,HeatmapcorrelationplotandPCAanalysisofmESCs,E8.5andE9.5post-occipitalembryo,and9individualTLS

basedonexpressionof41developmentalgenesmeasuredbyNanoStringanalysis.PC1andPC2representsthetwo

components with highest percentage of explained variance. b, Dot plot for 45 genes (41 test genes and 4

housekeepinggenes(redfont)).Dotsizescaleswithlog2ofexpression.Colorindicateslog2(max/min)asaproxyfor

rangeofexpression.Genesarerankedfromlowestrange(upperleft)tohighestrange(bottomright).c,Heatmap

withnormalizedcountsofindicatedinvivomarkergenesforembryoniccelltypesinmouseembryonicstemcells

(mESCs), the E8.5 and E9.5 post-occipital embryo, and 9 individual TLS. Pluri, pluripotency; Neuro, neuro-

ectoderm;NMPs,neuromesodermalprogenitors;PSM,presomiticmesoderm;Endo,endoderm;Hox,Hoxgenes;

HK,housekeepinggenes.


https://doi.org/10.1101/2020.03.04.974949


59

ExtendedDataFig.6Single-CellRNA-sequencingofTLS

a,Experimentalset-up.CriteriaforfilteringandremovalofdoubletsaredescribedinMethods.b,Heatmapwith

scaled expression of top marker genes of the fourteen identified clusters. c, UMAPs coloured by sampled

timepoints.d,UMAPscolouredbyexpressionofindicatedinvivomarkergenesforneuromesodermalprogenitors

(NMPs).pPSM,posteriorpre-somiticmesoderm;aPSM,anteriorpre-somiticmesoderm;PGCLCs,primordialgerm

celllikecells.


https://doi.org/10.1101/2020.03.04.974949


60

ExtendedDataFig.7RNAVelocityofinvivocelltypemarkergenesinTLS

a,Velocityandexpressionof invivomarkergenesof the somiticandneural trajectoryprojectedon thepooled

(96+108+120hTLS)UMAP.CellsarecolouredfollowingtheclustercoloursinFig.3a.


https://doi.org/10.1101/2020.03.04.974949


61

ExtendedDataFig.8SpatiotemporalprogressionofgeneexpressioninTLS

a,Heatmapwithscaledexpressionofgenesinvolvedinsomitogenesisin4914cellsfrom108hTLSrootedinNMPs

andorderedbypseudotime. aPSM,anteriorPSM,pPSM,posteriorPSM.b,Heatmapwith scaledexpression for

neural lineagemarkergenesofeachclusterthatconferprogenitor identity alongtheDV-axis invivo (leftpanel)

andcolumnplotshowingsubclusterdistributions.D,Dorsal;V,Ventral.c,HeatmapwithscaledexpressionofHox

genesinNMPsandtheirdirectdescendants(NeuralTube1,pPSM),orderedbypseudotimeafterrootingin96h.


https://doi.org/10.1101/2020.03.04.974949


62

ExtendedDataFig.9SomitecompartmentalizationinTLS

a,UMAPrepresentationofreclusteringof96hTLSsomiticcells.b,UMAPcolouredbyexpressionof indicated in

vivomarkergenesforanteriorandposteriorsomitecompartments.c, Invivo, thedorsalsomitecompartment is

markedbyPax3expressionandgivesrisetodermamyotomeandsubsequentlymigratorymuscleprecursorsand

myotome. The ventral somite compartment is marked by Pax1 and gives rise to the sclerotome, which

subsequentlydifferentiatesintochondrocytesandsyndetome.d,UMAPrepresentationofreclusteringof120hTLS

somitic cells. e, UMAP coloured by expression of indicated marker genes for dorsal and ventral somite

compartments. Asterisk refers to Pax1+/Scx+ population. f, Heatmap with scaled expression for top 10 marker

genesofeachcluster.g,UMAPcolouredbyexpressionofinvivosyndetomemarkerScx.NotethattheScxdomain

partially overlaps with a separate cluster of Pax1+ cells (indicatedwith an asterisk in e).h, UMAP coloured by

expressionofposterior(upperpanel)andanterior(bottompanel)markersUncxandTbx18.


https://doi.org/10.1101/2020.03.04.974949


63

ExtendedDataFig.10PrimordialGermCellLikeCells(PGCLCs)inTLS

a,Confocalsectionof108hTLSshowingPGCLCsthatco-expressSox2VEandDPPA3.b,Confocalsectionof120hTLS

stained forFOXA2andSox2VE.NotehowSox2VE-high cells contact theFOXA2+ cells in120hTLS. Scalebars50μm,

25μmformagnifications.


https://doi.org/10.1101/2020.03.04.974949


64


https://doi.org/10.1101/2020.03.04.974949


65

ExtendedDataFig.11ComparisonofTLScell-typeswithpost-occipitalcell-typesofthemouseembryo

a, Heatmap showing correlation of cell types identified in TLS (blue font) and cell types identified in the post-

occipitalE7.5andE8.5embryo(redfont)baseontop20markergenesforinvivoclusters.b,Dotplotsshowingthe

tenmost conserved cell typemarkers for indicated cell types between TLS and the embryo. c, Bar graphwith

percentage of expressed genes (average expression in cluster > 0) that are differentially expressed in TLS as

comparedtotheembryo(min.diff.pct>0.25andlog2FC>1).d,Bargraphwithpercentageoftop20invivocluster

markergenesthataredifferentiallyexpressedinTLSascomparedtotheembryo(min.diff.pct>0.25andlog2FC>

1).


https://doi.org/10.1101/2020.03.04.974949


66

ExtendedDataFig.12LossofsegmentationandmolecularphenotypeinTbx6-/-TLSmodels

a,SchematicandSangersequencingvalidationoftwoindependentTbx6::H2B-Venus(Tbx6VE);Tbx6-/-mESClines.

b, Segmentation phenotype of TLSC-WT, TLSC-Tbx6-/-,TLSCL-WT and TLSCL-Tbx6-/-. Data represent 3 different

experimentsperformedwith2independentmESClinesofeachgenotype.c,qRT-PCRshowingexpressionof(P)SM,

NMPs and NE genes in sorted Tbx6VE+ cells from TLS-WT and TLS-Tbx6-/-. Log10FCwas calculated from the fold

change ratio between TLS-Tbx6-/- and TLS-WT. Bars represent the mean of two biological replicates. d, 3D

maximumintensityprojectionsofTLSC-WTandTLSC-Tbx6-/-immunostainedforTbx6VE(magenta)andSOX2(green).

e,3DmaximumintensityprojectionsofTLSCL-WTandTLSCL-Tbx6-/-immunostainedforTbx6VE(magenta)andSOX2

(green).Scalebars100μm.


https://doi.org/10.1101/2020.03.04.974949


67

ExtendedDataFig.13CelltypespecificexpressionofcelladhesionmoleculesinTLS

a,UMAPof108hTLScolouredbyexpressionofindicatedgenes.b,UMAPof120hTLScolouredbyexpressionof

indicatedgenes.


https://doi.org/10.1101/2020.03.04.974949


68

ExtendedDataFig.14SequentialsegmentationinTLSsomites

a,Stills frombrightfield live imaging showing sequential generationof somites along the anterior-posterior axis

withapproximatelyatwohourintervalinTLS.Arrowheadsindicateindividualsomitesthroughouttime.Scalebar

100μm.A,anterior;P,posterior.


https://doi.org/10.1101/2020.03.04.974949


69

ExtendedDataFig.15Qualitycontrol

a,UMAPof96h,108hand120hTLScolouredbypredictedcellcycle,showingthatcellsdonotclusterbycellcycle.

To thisend, first foreachgene the cell cyclewasestimatedand subsequently this sourceofheterogeneitywas

regressedoutsothatthescaleddatacouldbeusedforUMAPcalculationwithoutacellcyclebias.b,Statisticsfor

eachoftheclusterbeforecorrectionfornumberofUMIs,genesandmitochondrialfraction.Cluster10isaclear

outlier showingastrongbias towardsanextremelyhighmitochondrial fractionandcluster11wasmarkedbya

highpercentageofcellswithanunusualnumberofUMIsandgenessuchthatafterremovalofthesecells,onlya

smallfractionofcellswasleft.Both,cluster10and11,wereexcludedfromfurtheranalysis.c,Tabledisplayingthe

averagenumberofUMIsandgenesdetectedforallcellsseparatedbycluster.TheaveragenumberofUMIsand

genesishomogeneousbetweenclustersanddoesnotshowanyoutliers.d,UMAPofinvivocellsfromSomiteand

PSMcell states.UMAPcolouredby initialcell state (left)comparedtomorestringentclusterassignment (right).

Thestringentclusterassignmentresults infoursub-clusters,thatsplitPSMintotwoequallysizedclusters,while

the Somites split into one larger cluster and a second group of cells that are located at the transition of the

previouslyidentifiedSomitesandPSMcluster.e,Heatmapofnormalizedexpressionofmarkergenesforthenewly

assignedfourclusterscomingfrominvivoPSMandSomitecells.ThetwoPSMclusterbothshowstrongexpression


https://doi.org/10.1101/2020.03.04.974949


70

ofthesamesubsetofmarkergenes,however,thenewsmallsubsettransitioningSomitecellshasaclearanterior

PSM(aPSM)signature.f,CorrelationanalysisofbulkRNA-seq,showingveryhighcorrelationbetweenreplicatesof

thesametreatment.Onlygenescoveredbyallsamplesandexpressedinatleastonesamplewereconsidered(n=

29,963)andthepairwisePearsoncorrelationcoefficientwascalculatedonthelog2(TPM+1)values.


https://doi.org/10.1101/2020.03.04.974949


71

SupplementaryVideo1.Wide-fieldliveimagingofTLSfrom96-120hshowingelongation,somitesegmentation,andneuraltubeformation.


ListsofgenesusedtogenerateboxplotsfromFigure2bandExtendedDataFig.3e


SignificantlyenrichedprocessesinTLSversusGastruloidsasrevealedbyGeneSetEnrichmentAnalysis


Overviewofbuffers,culturemedia,kits,andantibodiesusedinthisstudy


Sequencesofoligo’sandinsituhybridizationprobesusedinthisstudySupplementaryTable5.RawCtvaluesfromquantitativePCRanalysis


DetailsofNanostringprobesusedinthisstudy


Nanostringnormalizedgenecounts


MarkergenesforTLSclusters(single-cellRNA-Seq)&Markergenesforembryoclusters(single-cellRNA-Seq)


Top25conservedclustermarkergenesbetweenTLSandembryo&GenesdifferentiallyexpressedbetweenTLScelltypesandtheircorrespondinginvivocelltype


ExpressionvaluesinTPM(TranscriptsPerMillion)forallgenes(bulkRNA-Seq)


https://doi.org/10.1101/2020.03.04.974949


72


https://doi.org/10.1101/2020.03.04.974949


hypnotherapy for incontinence - bmj · urinary incontinence is one ofthe most socially...

Documents