A single-cell transcriptomic atlas maps cerebellar astrocyte diversity and uncovers the transcriptional code underlying their maturation trajectories

A single-cell transcriptomic atlas maps cerebellar astrocyte diversity and uncovers the transcriptional code underlying their maturation trajectories | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results A single-cell transcriptomic atlas maps cerebellar astrocyte diversity and uncovers the transcriptional code underlying their maturation trajectories View ORCID Profile Valentina Cerrato , View ORCID Profile Giacomo Turrini , View ORCID Profile Ilaria Vitali , View ORCID Profile Bilian Xiong , View ORCID Profile Laura Solanelles-Farré , View ORCID Profile Andrea Lopes , View ORCID Profile Elia Magrinelli , View ORCID Profile Riccardo Bocchi , View ORCID Profile Magdalena Götz , View ORCID Profile Judith Fischer-Sternjak , View ORCID Profile Enrica Boda , View ORCID Profile Annalisa Buffo , View ORCID Profile Ludovic Telley doi: https://doi.org/10.1101/2025.07.17.665323 Valentina Cerrato 1 Dept. of Neuroscience Rita Levi Montalcini, University of Turin , Italy 2 NICO, Neuroscience Institute Cavalieri Ottolenghi, University of Turin , Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Valentina Cerrato For correspondence: valentina.cerrato{at}unito.it Giacomo Turrini 1 Dept. of Neuroscience Rita Levi Montalcini, University of Turin , Italy 2 NICO, Neuroscience Institute Cavalieri Ottolenghi, University of Turin , Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Giacomo Turrini Ilaria Vitali 3 Dept of Fundamental Neuroscience, University of Lausanne , Switzerland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Ilaria Vitali Bilian Xiong 3 Dept of Fundamental Neuroscience, University of Lausanne , Switzerland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Bilian Xiong Laura Solanelles-Farré 3 Dept of Fundamental Neuroscience, University of Lausanne , Switzerland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Laura Solanelles-Farré Andrea Lopes 3 Dept of Fundamental Neuroscience, University of Lausanne , Switzerland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Andrea Lopes Elia Magrinelli 3 Dept of Fundamental Neuroscience, University of Lausanne , Switzerland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Elia Magrinelli Riccardo Bocchi 4 Biomedical Center (BMC), Faculty of Medicine, LMU Munich , Munich, Germany 5 Institute of Stem Cell Research, Helmholtz Center Munich, German Research Center for Environmental Health , Neuherberg, Germany 6 Department of Basic Neurosciences, University of Geneva , Switzerland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Riccardo Bocchi Magdalena Götz 4 Biomedical Center (BMC), Faculty of Medicine, LMU Munich , Munich, Germany 5 Institute of Stem Cell Research, Helmholtz Center Munich, German Research Center for Environmental Health , Neuherberg, Germany 7 Excellence Cluster of Systems Neurology (SyNergy) , Munich, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Magdalena Götz Judith Fischer-Sternjak 4 Biomedical Center (BMC), Faculty of Medicine, LMU Munich , Munich, Germany 5 Institute of Stem Cell Research, Helmholtz Center Munich, German Research Center for Environmental Health , Neuherberg, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Judith Fischer-Sternjak Enrica Boda 1 Dept. of Neuroscience Rita Levi Montalcini, University of Turin , Italy 2 NICO, Neuroscience Institute Cavalieri Ottolenghi, University of Turin , Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Enrica Boda Annalisa Buffo 1 Dept. of Neuroscience Rita Levi Montalcini, University of Turin , Italy 2 NICO, Neuroscience Institute Cavalieri Ottolenghi, University of Turin , Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Annalisa Buffo Ludovic Telley 3 Dept of Fundamental Neuroscience, University of Lausanne , Switzerland 8 University Claude Bernard Lyon 1, MeLiS—UCBL—CNRS UMR 5284—INSERM U1314 , Lyon, France Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Ludovic Telley For correspondence: valentina.cerrato{at}unito.it Abstract Full Text Info/History Metrics Supplementary material Preview PDF Abstract Astrocytes are increasingly recognized as key regulators of neural circuit development and function, with mounting evidence revealing substantial heterogeneity within and across brain regions. Yet, the full extent of this diversity and its developmental mechanisms remain poorly understood. To address this, we leveraged the uniqueness of the mouse cerebellum, which hosts well-defined astrocyte types and established progenitor pools. Through complementary multi-modal omic approaches, including single-cell RNA sequencing, spatial transcriptomics, trajectory inference, clonal lineage reconstruction, and gene expression and regulatory network analyses, we systematically dissected the molecular diversity and ontogenesis of cerebellar astrocytes. We identified known types and uncovered new subtypes with functional specialization, inferring their developmental trajectories from multiple embryonic niches and postnatal progenitor sources with fate divergence, convergence, and restriction. We further predicted a hierarchical transcriptional regulator code governing this diversification, operating at multiple levels: distinct regulatory modules i) reflect embryonic regionalization and lineage; ii) determine broad astroglial identity; specify iii) Bergmann versus non-Bergmann fates; and guide iv) astrocyte type and v) subtype acquisition. Our findings map and temporally organize transcriptional programs that capture key determinants of astrocyte fate, integrating them along defined trajectories toward diverse astrocyte identities. This high-resolution framework for cerebellar glial diversification offers a model to be challenged across other brain regions. Introduction Over recent decades, astrocytes have emerged as central players in neural circuit development, function, and plasticity, shifting the traditional notion of their role from merely supportive elements to active participants in CNS physiology 1 , 2 . Accumulating evidence underscores the substantial molecular, morphological, and functional heterogeneity among astrocytes, which aligns with their distinct regional identities, developmental origins, and specialized functions 3 – 9 . Astrocyte diversity can be remarkably profound, even within a single anatomical region. For instance, in the cerebral cortex, astrocytes demonstrate transcriptional heterogeneity that partially reflects the cortical laminar structure 10 , 11 . Similarly, distinct astrocyte subtypes have been identified within the layered architecture of the hippocampal dentate gyrus 12 . Nevertheless, clearly defining astrocyte types and subtypes within brain regions remains challenging, as subtle morphological and molecular variations frequently blur distinctions between astrocyte populations, complicating precise identity and functional definitions. Thus, despite substantial progress, achieving a comprehensive characterization of astrocyte heterogeneity, particularly within individual brain regions, and elucidating the developmental mechanisms underlying this diversity, remain open goals. To address these challenges, we leveraged the robust framework provided by the mouse cerebellum, distinguished by well-defined astrocyte types, with stereotyped morphologies, spatial allocations 13 ( Fig. 1d ) and established astroglial progenitor lineages 14 – 16 and by spatially segregated functional compartments 17 . This knowledge, integrated with single-cell/single-nucleus RNA sequencing (sc/snRNA-seq), maturational trajectory inference, clonal investigations, differential gene expression, and gene regulatory network analysis, allowed to systematically dissect cerebellar astrocyte heterogeneity and its development with unprecedented efficacy. Download figure Open in new tab Fig. 1 Sc/snRNA-seq resolves main astrocyte types and uncovers new subtypes in the adult mouse cerebellum. a, UMAP plot of the six integrated datasets of the adult mouse cerebellum, with dots colored based on cell types. The inset shows the cells with a high astrocyte score, defined by the combinatorial expression of Slc1a3 , Slc1a2 , Aldh1l1 and Aqp4. b, Flow diagram showing the dataset composition (i.e. sampled cerebellar region, type of sample, name of the original dataset and number of cells/nuclei per dataset) of the astrocyte subset. n= number of astrocytes. c, UMAP plot of the astrocyte subset. The dotted lines surround the two broad groups corresponding essentially to Bergmann glia (Bergman) and the other non-Bergmann glia astrocytes (Astro). d, Schematic representation of the main astrocyte types of the cerebellum and their localization in the cerebellar grey or white matter regions: Bergmann glia (BG) in the Purkinje cell layer (PCL) and molecular layer (ML), grey matter astrocytes (GLA) in the granular layer (GL), white matter astrocytes (WMA) in the white matter (WM), and cerebellar nuclear astrocytes (CNA) in the cerebellar nuclei (CN). e, Mapping of the identified astrocyte clusters in adult mouse Visium (first line) or Merfish (second and third line) ST datasets. The first column shows the layer annotation of the sagittal sections analyzed. Bar plots at the bottom show the average layer enrichment predicted in multiple (n=16) sections of the Merfish dataset (+-SEM). Cells mapped in the ML/PCL, GL, WM or CN were annotated as BG, GLA, WMA or CNA, respectively. f, Expression of marker genes supposed to be enriched or specific for the distinct astrocyte types ( Slc1a3, Gdf10, Gria1, Gria4 for BG; Aqp4, Cyp26b1, Hsd11b1 for GLA; Slc6a11 for CNA; Cd44, Myoc, C4b for WMA; Atp1b2 for all astrocytes except BG) or subtypes ( Mybpc1 for Bergmann_2 cluster in Kozareva et al. 18 ). g, Expression of the top DEGs (avgLog2FC > 1) of each of the identified astrocyte clusters. h, Pairwise cumulative comparisons across the DEGs of each cluster. Note that most overlaps are between clusters belonging either to the Bergmann or to the Astro groups, with the only exception of BG_ Kcnip4 and GLA_WMA_ Pde10a , that share 7 genes (intersection highlighted in the Venn diagram). BG, Bergmann glia; GLA, granular layer astrocytes; WMA, white matter astrocytes; CNA, cerebellar nuclei astrocytes; ML, molecular layer; PCL, Purkinje cell layer. Building upon this knowledge, this work identifies so far unrecognized astrocyte subtypes, revealing unique spatial distributions and functional specializations. We delineate the transcriptional regulator code orchestrating cerebellar astrocyte heterogeneity and show that this diversity arises from both canonical and non-canonical intra-and extra-cerebellar astrogliogenic niches, involving complex patterns of cell fate divergence, convergence, and restricted lineage paths from multiple progenitor sources. Results Transcriptional heterogeneity resolves main adult astrocyte types and uncovers new subtypes To address the intra-regional diversity of cerebellar astrocytes, we first looked at existing sc/snRNA-seq datasets of the adult mouse cerebellum 18 – 21 . However, we found that they provided limited coverage of non-cortical cerebellar astrocytes, restricting a comprehensive investigation of these cell populations (Extended Data Fig. 1a). To overcome this limitation, we generated new sc/snRNA-seq data through manual microdissection of deep cerebellar regions, specifically of the cerebellar nuclei (this work, Supplementary Table 1) or deep white matter 22 . These new data were integrated in silico with the existing published datasets, resulting, after quality control and filtering (see Methods), in a combined total of 493,505 high quality nuclei/cells (hereafter referred to as “cells”) with a total of 22,335 genes detected ( Fig. 1a ). Following unsupervised clustering in Seurat V5 and based on the expression of cell type-specific molecular markers, we assigned these cells to 18 distinct cell types ( Fig. 1a ), comprising 7 classes and 10 subclasses 18 , 21 (Extended Data Fig. 1b). Cells with a high (>1) astrocyte score, determined by the combined expression of established astrocyte-specific markers 7 , 9 ( Slc1a3 , Slc1a2 , Aldh1l1 and Aqp4 ), were then subsetted for further analyses, resulting in an integrated astrocyte dataset comprising a total of 43,395 astrocytes and 19,914 genes detected. This provided comprehensive coverage and high resolution across the entire cerebellum ( Fig. 1b ). Astrocytes were distributed into 9 clusters within the UMAP embedding, forming two segregated groups ( Fig 1c ). Mapping to two publicly available adult mouse spatial transcriptomics (ST) datasets 20 allowed to infer that the two broad groups were composed of distinct astrocyte types of the grey and white matter, matching the canonical classification of cerebellar astrocytes, comprising Bergmann glia (BG) and velate granular layer astrocytes (GLA) in the cortex, white matter astrocytes (WMA), and cerebellar nuclei astrocytes (CNA; Fig. 1d,e ). Specifically, one group was predominantly composed of Bergmann glia (Bergmann), and the second group of non-Bergmann astrocytes (Astro). Astrocyte subtypes were also identified, either confined to discrete regions or spanning multiple regions ( Fig. 1e ). When looking for the genes at the core of molecular diversity across clusters, the sole expression levels of most of the markers known to be enriched or specific for certain cerebellar astrocyte types 23 – 26 were insufficient on their own to annotate all clusters and, rather, only distinguished Bergmann from Astro ( Fig. 1c,f ). Finer cluster discrimination was instead achieved with the most specific differentially expressed genes (DEGs) of each cluster (Fig.1g and Supplementary Table 2). Hence, final cluster annotation ( Fig. 1c ) was obtained by combining regional distribution predictions with the most specific DEGs of each cluster. Two clusters predicted to map to the PCL or GL did not exhibit any enriched DEGs, denoting a reference transcriptomic profile, and were therefore annotated as BG_ arch (i.e. “archetype”) and GLA_ arch . This classification revealed 6 clusters with predominant localization in a single region (BG_ arch , BG_ Mybpc1 , BG_ Kcnip4 , WMA_ Pgm5 , GLA_ arch and CNA_ Gpc5; Fig. 1e ). Among the clusters mapping in the ML/PCL, BG_ arch and BG_ Mybpc1 had a strikingly complementary distribution in the anterior or posterior lobules, respectively ( Fig. 1e , Extended Data Fig. 2d,e). Moreover, these 2 clusters in Merfish predictions showed low spatial and transcriptional overlap within each other and with BG_ Kcnip4 (Extended Data Fig. 2h), in line with these 3 clusters representing distinct Bergmann subtypes. In contrast, 2 clusters (BG_WMA_ Npy and GLA_WMA_ Pde10a ) spanned across multiple regions and may therefore represent new subtypes of astrocytes with similar transcriptional features but diverse spatial allocations. In the Visium dataset, BG_WMA_ Npy appeared as an hybrid cluster, enriched in both the PCL/ML and WM ( Fig. 1e ). This enrichment varied depending on the cell source: cells collected from the whole cerebellum or the cerebellar cortex 18 – 20 mapped to the PCL/ML, whereas cells sourced from the deep cerebellum (this work, Bocchi et al. 22 ) mapped to the WM (Extended Data Fig. 2a,b). The cluster hybrid nature was further confirmed by its overlapping signature with the cerebellar WMA cluster “Cer_2” of Bocchi et al. 22 (Extended Data Fig. 1c) in conjunction with high expression of typical Bergmann genes ( Fig. 1f ). The BG_WMA_ Npy cluster was enriched with gene terms associated with translation and protein folding, processes typically upregulated in neural progenitors and stem cells 27 , 28 , and exhibited specific enrichment of genes expressed in NSCs and radial glial cells ( Fig. 1g , Extended Data Fig. 1d and Supplementary Table 3), in line with the display of an immature profile 22 . GLA_WMA_ Pde10a cells were enriched in both the GL and WM. This cluster shared a few DEGs with BG_ Kcnip4 , a unique case of overlap between Bergmann and Astro groups ( Fig. 1h ). Both these subtypes also displayed a high glutamate release score, defined by the expression of Syt1 , Snap25 and Slc17a7 , and shared many DEGs with the recently uncovered specialized astrocyte subtype that mediates glutamatergic gliotransmission 29 (Extended Data Fig. 1e). Accordingly, GO analysis highlighted for GLA_WMA_ Pde10a an enrichment of terms related to synaptic vesicle exocytosis and synaptic transmission (Extended Data Fig 1f, Supplementary Table 3). These clusters may therefore represent functionally specialized vesicular-releasing subtypes particularly involved in gliotransmission and localized in distinct grey and white matter cerebellar regions. Of note, the reported analysis showed 3 clusters containing WMA, suggesting distinct subtypes with complementary enrichment patterns in the lobular (WMA_ Pgm5 and WMA_ Pde10a ) or deep WM (WMA_ Npy ; Fig. 1e , Extended Data Fig. 2b). Additionally, we identified a small Slit2 -expressing cluster exhibiting a salt and pepper pattern (Extended Data Fig. 2c) with apparent enrichment in the posterior lobules of the vermis (IX/X) (Extended Data Fig. 2d,e). Slit2 expression may reflect a reactive-like state 30 in non-Bergmann glial cells. Cross-comparisons showed significant transcriptional overlap within the Bergmann and Astro groups, but not between them ( Fig. 1h ). GO analysis supported this divergence, confirming Bergmann glia main specialization in maintaining cerebellar structural integrity, regulating glutamatergic inputs, and relying on Shh signaling to sustain their molecular identity 24 , 31 . Conversely, this analysis uncovered novel potential roles for non-Bergmann Astro, including involvement in energy homeostasis via regulation of blood supply, glucose and glycogen metabolism, and neuroprotection (Extended Data Fig. 1g, Supplementary Table 4). Overall, these data point to spatially segregated astrocyte subtypes, some of which exhibit features suggestive of functional specializations. They also indicate remarkable functional divergence between Bergmann and non-Bergmann astrocytes. Transcriptional heterogeneity defines postnatal astroglial-like progenitor pools transitioning into fully mature astrocyte subtypes To elucidate the developmental processes underlying the remarkable heterogeneity of cerebellar astrocytes, we generated an integrated multi-modal sc/snRNAseq dataset of the postnatal cerebellum. This dataset encompasses 28 distinct experiments conducted across developmental stages from postnatal day (P) 0 to P30 (P0, P4–P5, P7, P10–P11, P14–P16, P23, P30) and integrates diverse information, partly exploited in this study, including time of birth, clonal relationships, antero-posterior (AP) and medio-lateral (ML) spatial resolution (Extended Data Fig. 3a, Supplementary Table 1). Cells collected at biologically similar time points (P0-P5: early postnatal development; P7-P11: circuit assembly; and P14-P30: final maturation and refinement) were integrated in silico. Following quality control and filtering steps (see Methods), the dataset retained 126,143 high quality cells (P0-P5 = 85,443; P7-P11 = 19,105; P14-P30 = 21,645), with a total of 20,226 genes detected across all stages. Following unsupervised clustering, a deep neural network-based prediction model 29 combined with manual annotation based on known marker genes were used to classify the cells ( Fig. 2a ; Extended Data Fig. 3b). Comparison with publicly available datasets of the postnatal mouse cerebellum 19 , 32 , 33 demonstrated the unique resolution of this dataset, capturing the full repertoire of cerebellar cell types with exceptional detail (Extended Data Fig. 3c). Download figure Open in new tab Fig. 2 Transcriptomic profiling of the postnatal mouse cerebellum to reveal astrocyte subtypes maturation. a, UMAP plots of the integrated P0-P4, P7-P11 and P14-P30 datasets of the postnatal mouse cerebellum, with dots colored by cell cell types and maturational states. b, UMAP plot of the postnatal subset of Slc1a3- expressing cells, with dotted lines outlining clusters corresponding to distinct cell lineages. Astrocytes (Astro), GABAergic interneurons (INs), Purkinje cells (PC), glutamatergic granule cells (GC), oligodendrocytes (Oligo), and ependymal cells (Epend). c, Expression of lineage-specific markers used to classify cells. d, Enrichment scores of distinct clusters across developmental stages, grouped based on their enrichment patterns at early (top), middle (middle) or late (bottom) stages. e, UMAP insets from panel b highlighting clusters 9, 14 and 20 containing the mature astrocyte subtypes. Annotated subtypes within these clusters are shown in separate insets. f, Dot plots of the expression scores of adult astrocyte subtype DEGs. g, Dot plots of the expression levels of adult subtype-specific markers. We used this dataset to subset astroglial-like progenitors and maturing astrocytes based on expression of the pan astrocyte and astroglial-like progenitor marker Slc1a3 14 , 16 . Slc1a3 expression partially overlapped with other lineage-specific markers, including Pax2 (GABAergic interneurons; “Slc1a3_Pax2”), Pax6 (Glutamatergic neurons, “Slc1a3_Pax6”), and Olig1 (Oligodendrocytes, “Slc1a3_Olig1”). Furthermore, Slc1a3 expression was also detected in cells of the ependymal, microglial, and fibroblast lineages (Extended Data Fig. 3b,d). For further analyses, the Slc1a3 -expressing clusters were filtered, excluding microglia and fibroblasts due to their non-neural origin, and subsequently re-integrated (Extended Data Fig. 3d). In addition, the entire oligodendrocyte and interneuron lineages, also comprising Slc1a3 -negative cells, were incorporated into this subset in view to ensure robust internal validation of cell maturation trajectories (see Methods and Extended data Fig.5 ). Unsupervised clustering partitioned the Slc1a3 -expressing subset into 22 distinct clusters, most of which could be grouped into separate lineages based on reference genes ( Fig. 2b,c ). Specifically, clusters 0, 1, 4, 9, 14, 17 and 19 expressed cerebellar astrocyte-typical genes; clusters 5 and 12 belonged to the GABAergic interneuron lineage; clusters 10 and 21 expressed Purkinje cells (PC)-typical genes; cluster 2 displayed granule cell (GC)-specific features; clusters 8, 13, and 16 represented distinct stages of oligodendrocyte differentiation; and cluster 15 corresponded to ependymal cells. These results align with the presence of Slc1a3 + ancestors for astrocytes, interneurons, oligodendrocytes, GCs, and ependymal cells 14 , 16 , 34 and suggest that some immature PCs retain Slc1a3 mRNA up to P5. Within this subset, cells were distributed according to their age of collection, forming a continuum of cell states ranging from progenitors to fully mature cells (Extended Data Fig. 4a). Distinct clusters exhibited diverse enrichment patterns across developmental ages: some clusters, likely representing progenitors and including cycling cells, were enriched at neonatal stages but became underrepresented at later stages ( Fig. 2b , top panel; Extended Data Fig. 5a). Other clusters, likely corresponding to intermediate progenitor states, exhibited increased enrichment beginning from middle stages and typically plateaued thereafter ( Fig. 2b , middle panel). Additional clusters, reflecting maturing cell states, demonstrated a gradual increase in enrichment, peaking at later developmental stages ( Fig. 2c , bottom panel). Inspection of astrocyte-specific markers and their stage-specific distribution ( Fig. 2c ; Extended Data Fig. 4b,c) classified clusters 9, 14, and 20 as the most mature astrocyte states. Gene expression patterns and spatial mapping confirmed the presence of astrocyte subtypes previously defined in the adult dataset also in these postnatal samples ( Fig. 2e,f ,g, Extended Data Fig. 4d). Cluster 9 encompassed most terminal differentiated BG subtypes, including BG_ arch , BG_ Mybpc1 , and BG_ Npy . Cluster 14 included WMA_ Npy , WMA_ Pgm5 , GLA_ arch , and CNA_ Gpc5 . Notably, the analysis of this dataset discriminated between Npy -expressing BG and WMA, which were instead merged into a single cluster in the adult dataset. Conversely, a few cells in cluster 20 exhibited molecular profiles shared between BG_ Knip4 and GLA_WMA_ Pde10a , with a predicted spatial distribution spanning the PCL, GL, and WM. Due to their vesicular-releasing profile, we hereafter refer to this population as “Vesic” (Extended Data Fig. 4d). Collectively, this postnatal dataset offers a comprehensive representation of both astroglial-like progenitor heterogeneity and mature astrocyte subtypes, making it well-suited for investigating astrocyte maturation trajectories. Inferred maturation trajectories identify postnatal multipotent and committed astroglial-like progenitors To reconstruct the maturation trajectories of the astrocyte subtypes extending back to perinatal progenitor states, we combined URD and two CellRank kernels, thereby integrating three distinct computational approaches to infer cellular dynamics from sc/snRNAseq data, based on similarities in transcriptional profiles and their distribution over time 35 , 36 ( Fig. 3a ). Outputs were integrated into a consensus heatmap by averaging each clusterwise logit-transformed probability of transitioning into a defined transcriptional state ( Fig. 3a , Supplementary Table 5). The predictive power of this strategy was validated by using data from the oligodendrocyte and GABAergic interneuron cell classes (Methods and Extended Data Fig. 5). Download figure Open in new tab Fig. 3 Inferring astrocyte maturation trajectories through a combined computational approach. a, Schematic representation of the in silico approach used to predict maturation trajectories, integrating URD 35 (retrograde random walks) with two distinct CellRank 36 kernels (forward random walks). Representative plots display the root and tip cells of the 1st rounds of random walks (i.e. from or toward the most differentiated cells), color-coded based on the algorithm used. The clusterwise logit-transformed probabilities from each of the three simulations are averaged to generate a consensus heatmap. b, Consensus heatmaps of random walks predicting astrocyte maturation trajectories, showing for each cluster (columns) the probability of differentiating into a defined terminal (1st round) or intermediate (2nd round) fate. Solid and dotted black lines indicate the clusters with the highest or second-highest probability in at least two out of three approaches, respectively. Heatmap colors are scaled independently for each row. c, Consensus heatmap of the random walks performed using as tips the cells in cluster 14 (top) or cluster 9 (bottom) with the highest probability of differentiating into GLA_ arch , WMA_ Pgm5 , CNA_ Gpc5 , or BG_ arch , BG_ Mybpc1 , BG_ Npy . Notably, cluster 4 shows the second-highest probability of differentiating into WMA_Pgm5 in both this additional round and the 1st round of random walks (panel b , ninth row of the heatmap). d, Schematic reconstruction of the differentiation trajectories of each mature astrocyte subtype from perinatal clusters, as inferred from the consensus heatmaps. e, Dendrogram summarizing the inferred maturation steps of all astrocyte subtypes from multipotent or fate-restricted perinatal progenitor clusters. This highlights both convergence and divergence of cell fates, reflecting the complex scenario from which cerebellar astrocyte diversity arises postnatally. Note that the length of the connecting lines and the spatial arrangement of clusters in the dendrogram do not reflect the timing or duration of maturation steps, but were manually adjusted for clarity and to accommodate all elements within the figure layout. f, Expression levels of the top 10 DEGs of the five clusters predicted to be at the origin of the maturation trajectories. In situ hybridization data from the Allen Brain Atlas and immunofluorescence analyses suggest distinct spatial localizations of these clusters within the postnatal BG layer, cerebellar nuclei, or prospective white matter. Scale bars: 200 µm. This strategy predicted that all BG subtypes originate from cluster 0 through maturing precursors in cluster 9, that WMA_ Npy derive from cluster 17 via cluster 19 immature cells and that Vesic emerge from cluster 7 through their predecessors in cluster 20 ( Fig. 3B ). WMA_ Pgm5 , GLA_ arch and CNA_ Gpc5 were all traced back to cluster 14 ( Fig. 3B ), suggesting that this cluster contains multipotent progenitors capable of giving rise to multiple astrocyte types. However, further analysis showed that cluster 14 comprises 3 cell subsets already displaying divergent commitments, each with distinct probabilities of differentiating into one of the three terminal subtypes (Extended Data Fig. 5G-I). An additional analysis to infer the origin of these 3 subsets (Extended Data Fig.5J; Fig. 3C ) uncovered distinct progenitor sources: GLA_ arch and WMA_ Pgm5 were predicted to derive from both cluster 1 and 4, while CNA_ Gpc5 originated exclusively from cluster 1. Hence, cluster 1 represents the only source of CNA_ Gpc5 through cluster 14, whereas the other two subtypes can emerge from either cluster 1 or 4. Applying a similar analysis to cluster 9 (Extended Data Fig. 5K), along trajectories of multiple BG subtypes, confirmed that all these subtypes originate with similar probabilities from cluster 0, highlighting homogeneous differentiation potential within these progenitor cells ( Fig. 3C ). Lastly, although hierarchical relationships among perinatal-enriched clusters could not be directly established (see Methods), clusters that ranked second-highest in initial analyses consistently rose to top positions in subsequent rounds ( Fig. 3B , Extended Data Fig. 5D, dotted black lines), suggesting additional lineage transitions. In particular, cluster 4 may serve as a common progenitor for cluster 0—and thus for BG subtypes—as well as for GLA_ arch and WMA_ Pgm5 . To sum up, this analysis points to distinct perinatal progenitor clusters, each with unique differentiation potentials. Cluster 4 appears to contain multipotent progenitors giving rise to all BG subtypes, GLA_ arch and WMA_ Pgm5 . By contrast, cells in cluster 0 are committed to the BG fate. Cluster 1 is the only source for CNA_ Gpc5 , though it can also differentiate into GLA_ arch and WMA_ Pgm5 , but not BG. Clusters 17 and 7 exhibit more segregated trajectories, giving rise exclusively to WMA_ Npy and Vesic, respectively. Of note, DEG analysis across these clusters and spatial mapping 37 revealed diverse transcriptomic profiles and suggested separate spatial allocations in the perinatal cerebellum, well matching the inferred maturation trajectories ( Fig. 3F , Extended Data Fig. 6E and Supplementary Table 6). Cluster 0 (BG source) expressed high levels of Gria1 and Gdf10 which are enriched in expanding BG at P4 23 ( Fig. 3F ). Cluster 1 (CNA source) expressed Slc6a11 , Olig1 and Olig2 , both highly expressed in the CN territory ( Fig. 3F , Extended Data Fig 6F). Cluster 4 (multipotent progenitors of BG, GLA_ arch and WMA_ Pgm5 ) displayed elevated Angpt1 and Aqp4 , both enriched in the prospective WM (PWM; Fig. 3F ); Apoe , shared by clusters 1 and 4, was likewise expressed in both CN and PWM (Extended Data Fig. 6B). Notably, the inferred PWM localization of cluster 4 multipotent progenitors aligns with the previously reported location of BG+GLA+WMA progenitors at birth 15 , confirming this cluster identity. Moreover, cluster 1, which showed elevated Atp1b2 , mirrored the molecular profile of ACSA-2 + /GLAST + progenitors known to produce GLA and WMA but not BG 16 , consistent with cluster 1 inferred differentiation potential (Extended Data Fig. 6C,D). Interestingly, cluster 7, predicted to give rise to “glutamatergic-like” Vesic, showed high Mytl1 38 and Zmat4 , whose expression patterns match those of emerging glutamatergic cells (Extended Data Fig. 6B), suggesting a shared origin with excitatory neurons, distinct from other VZ-derived astrocyte subtypes. Due to limited data on DEG expression patterns, the spatial distribution of cluster 17 remained undetermined. Finally, despite our integrative strategy did not identify any perinatal progenitor cluster common to both interneuron and WM astrocyte lineages 14 , likely due to low yield of astrocytes from this source, we hypothesized that cluster 11, with specific Ptf1a expression (Extended Data Figs. 5E, 6G) could contain such bipotent progenitors. This hypothesis was confirmed by genetic fate mapping using Ptf1a CreERTM mice 39 (Extended Data Figs. 5E, 6G) showing the generation of both ML interneurons and a smaller proportion of WM astrocytes. Taken together, these results define the profile of CNA perinatal progenitors and reveal a complex mechanism underlying the emergence of astrocyte heterogeneity postnatally, driven by divergence and convergence of cell fates as well as by early fate-restricted paths ( Fig. 3E ). Cerebellar astrocyte heterogeneity unfolds from multiple lineages and shares embryonic origins with distinct other cell types Having defined the maturation trajectories and fate potentials of perinatal astrocyte progenitors, we next sought to explore their embryonic origins and lineage relationships with other cerebellar cell types. To this end, we leveraged the clonal information contained in the P4-P5 dataset (Extended Data Fig. 3A), resulting from the injection at embryonic day 11 (E11) of a lipid nanoparticle (LNP)-encapsulated TrackerSeq library 40 . This transposon-based barcoding approach enables the high-throughput tagging of single embryonic progenitors, allowing for the retrospective identification of sister cells through scRNA-seq. Cells were grouped according to their assignment to distinct perinatal astrocyte progenitor clusters (i.e. “Slc1a3_cl*”) or to other cerebellar cell types, and lineage coupling scores were computed for all pairwise combinations (see Methods). As expected, hierarchical clustering of the pairwise correlations between coupling scores ( Fig. 4A , Supplementary Table 7) revealed a lineage relationship between GABAergic interneurons and astrocyte progenitor clusters 0, 4, 17, and 19 ( Fig. 4A-C ), supporting a ventricular origin for BG, WMA_ Npy and part of GLA_ arch and WMA_ Pgm5 14 , 15 . This was in line with spatial mapping onto an E16.5 ST dataset 41 , which localized clusters 0, 4, and 17 to the VZ and its adjacent cerebellar parenchyma ( Fig. 4H ). Intriguingly, these astrocyte progenitor clusters showed low coupling correlations with both clusters 1 and 7, suggesting a distinct developmental origin for CNA_ Gpc5 and Vesic. Download figure Open in new tab Fig. 4 Lineage relationships and embryonic mapping of perinatal astrocyte progenitors. a, Heatmap of lineage coupling scores between pairs of P4 cells belonging to the perinatal astrocyte progenitor clusters (prefix “Slc1a3_”) or to other cerebellar cell lineages. Values range from 1 (red, highly coupled) to-1 (blue, anti-coupled). b, The pie charts show the composition of clones containing cells from cluster 0 (left) or cluster 4 (right). In both cases, more than 50% of the clones also contain interneurons. Notably, while 11% of cluster 0-containing clones are composed exclusively of cluster 0 cells, cluster 4-containing clones are always shared with other lineages, and 19% of them are shared with cluster 0, further supporting the multipotent nature of cluster 4 progenitors. c, Representative examples of clones shared between astrocyte progenitor clusters 0, 4 and 19 and cells along the interneuron lineage. Clone 8_62 (right) also contains cells in the Purkinje cell lineage (“Slc1a3_Purkinje” comprises cells in both “Slc1a3_cl10” and “Slc1a3_cl21” of the Slc1a3 subset). d, The pie chart illustrates the composition of clones containing cells from cluster 1, highlighting the predominance of mixed clones with oligodendrocytes and/or ependymal cells. e, Examples of clones shared between cluster 1 and cells along the oligodendrocyte (left and middle plots) or ependymal (right plot) lineages. f, Pie chart showing the composition of clones containing cells from cluster 7, the vast majority of which also include granule cells (GCs, ∼87%) or glutamatergic CN neurons (∼33%). g, Representative examples of clones shared between cluster 7 and cells along the GC (left and middle plots) or glutamatergic CN neuron (right plot) lineages. h, Spatial mapping of perinatal astrocyte progenitor clusters onto two sections (hemisphere, top; vermis, bottom) from a E16.5 Stereo-seq dataset 41 . The first column shows the manual broad region annotation based on the Allen Brain Atlas, with corresponding color coding. The cerebellum (Cb) and choroid plexi (ChP) are labeled for visual reference. i, Heatmap showing the percentage of cells from each perinatal cluster of the Slc1a3 subset (columns) that co-cluster with each of the distinct cerebellar cell type lineages in the integrated scRNA-seq dataset of the embryonic/early postnatal cerebellum. Relevant percentages are indicated within the corresponding squares. Download figure Open in new tab Figure 5 T h e TR code underlying the emergence of cerebellar astrocyte heterogeneity. a, UpSet plot illustrating the overlap of regulons identified by SCENIC across different astrocyte subtypes. The pie chart summarizes the proportion of regulons that are subtype-specific, partially shared among subtypes, or universally shared across all astrocyte subtypes. Totals on the right indicate the total number of regulons detected per subtype. b, Heatmap summarizing the activity of selected TRs (SCENIC-derived AUC z-scores) across the distinct trajectories in early (top) or late (bottom) maturation stages. TRs are grouped based on their activation pattern: shared across most trajectories, type-specific (BG, GLA-WMA-CNA) or subtype-specific (Npy, Vesic). c, Heatmaps showing the activity of TRs shared through the distinct trajectories. d, Selected TRs specific for Bergmann vs Astro. Heatmaps on the top show the activity of Gli1 and Smad3 (Bergmann-specific; left) or Stat1 and Stat3 (Astro-specific; right). Dot plots present the corresponding TR gene expression levels, consistent with their SCENIC-predicted activity. e, Selected TRs specific for CNA. Both gene expression (left) and predicted activity (middle) of Foxb1 and Ar highlight specificity for CNA. UMAP plot shows the gene expression score of Foxb1 regulon targets, enriched specifically in CNA progenitors within cluster 1. f, Heatmap showing the activity of TRs increasing through BG_ Npy and WMA_ Npy maturation trajectories. Note that the same TRs are active at the onset of the other trajectories and then gradually decrease. g, Heatmap showing the activity of TRs specific for the Vesic subtype(s). h, Target gene expression scores (top) and TR expression (bottom) of Klf3 and Klf5 showing complementarity between lineages stemming from cluster 4 or 1, respectively. i, Target gene expression scores (left and middle) and TR expression (right) of En1/En2 or Dbx2/Gbx2 showing complementarity between lineages stemming from cluster 4 or 1, respectively. j, Proposed model of the TR code orchestrating the emergence of cerebellar astrocyte types and subtypes. On the one hand, cluster 1 appeared related to both ependymal cells and oligodendrocytes in the clonal data ( Fig. 4A, D, E ) and when projected onto an integrated sc/snRNA-seq dataset of the embryonic and perinatal cerebellum (Extended Data Fig. 7A-C), it was clustered with both astrocyte and oligodendrocyte lineages (70% and 25% of its cells, respectively; Fig. 4I ). Moreover, its predicted localization at E16.5 ( Fig. 4H ) and the embryonic expression of GAT3 ( Slc6a11 , Fig. 3F ; Extended Data Fig. 7D) were higher in the hindbrain than within the cerebellum, consistent with the extracerebellar origin of oligodendrocytes 13 . Supporting this developmental relationship, not only cluster 1 retained the expression of the oligodendrocyte-typical transcription factors Olig1 and Olig2 ( Fig. 3F ), but adult CNA_ Gpc5 also displayed a transcriptional profile closely resembling that of spinal cord and cortical astrocytes derived from Olig2 -expressing progenitors 42 , 43 (Extended Data Fig.7E-F), a well-established source of OPCs in the spinal cord 44 . Therefore, several lines of evidence pointed to a relationship between cluster 1 and the oligodendrocyte lineage. On the other hand, cluster 7, the predicted source of Vesic, showed a strong lineage relationship with granule cells (GC) and with glutamatergic CN neurons ( Fig. 4A, F, G ; Extended Data Fig. 7G). In the integrated embryonic sc/snRNA-seq dataset, it clustered together with these excitatory populations ( Fig. 4I ) and it was predicted to localize within the nuclear transitory zone of the embryonic cerebellum ( Fig. 4H ), a transient developmental region where CN glutamatergic neurons differentiate 45 . Altogether, these findings further support a shared developmental origin between Vesic astrocytes and excitatory neurons, as previously discussed. Altogether, these findings reveal diverse embryonic origins of cerebellar astrocyte subtypes. On the one hand, they confirm the VZ as the source of several astrocyte populations, extending to embryonic stages their lineage relationship with GABA interneurons, previously demonstrated only postnatally. On the other hand, they uncover novel associations with oligodendrocytes and excitatory neurons, suggesting that a subset of astrocytes arises from distinct germinative niches beyond the traditionally recognized VZ. The transcriptional logic driving the emergence of cerebellar astrocyte types and subtypes To dissect the molecular mechanisms underlying cerebellar astrocyte maturation, we employed SCENIC 46 , which identified numerous transcriptional regulators (TRs) putatively active along multiple or specific trajectories ( Fig. 5a , Supplementary Table 8). Manual curation and selection of TRs with predicted activity (see Methods) consistent with their expression pattern revealed a precise code orchestrating the emergence of cerebellar astrocyte heterogeneity at multiple levels ( Fig. 5b , 5j, Supplementary Table 9). A core set of regulators were shared through the distinct trajectories, either in the early or late maturation phases or through the whole process, indicating common molecular programs underlying the acquisition of features fundamental to the astroglial subclass ( Fig. 5b,c ). Consistently, many ( Pou3f3, Elk3, Atf3, Foxo1, Nfkb1 and Nfia ) had documented roles in astrocyte differentiation, maturation and function 47 – 50 or converged onto known developmental signaling pathways. The FGF-ERK-ETV axis ( Etv4 , Etv5 ), previously implicated in BG specification from radial glia 51 was active also in other astrocyte types, with differential deployment of Etv5 or Etv4 in Bergmann vs Astro, respectively ( Fig. 5b,c ), pointing to a more general, though type-biased, role in cerebellar astrocyte induction and maintenance. Similarly, Notch signaling components ( Hes1 , Hes5 ) were prominently active at early stages of all trajectories except for Vesic, consistent with their established roles in gliogenic competence and BG monolayer formation 52 , 53 . Retinoic acid signaling ( Rara , Rxra , Rxrg ) also emerged as a shared regulatory network 54 . However, Rara and Rxra were more prominent at later stages of GLA_ arch and WMA_ Pgm5 maturation, implying selective activation of this pathway in separate trajectories. Notably, our analysis also revealed several new TRs, shared across trajectories ( Fig. 5c , Extended Data Fig. 8b), that lacked prior association with astrocyte development. Specifically, Zic3 was enriched at early stages, while Bcl6 , Ikzf2 , Maf and Pdlim5 were found in more mature astrocytes. Their regulated genes are associated with critical astrocyte functions, such as modulation of glutamatergic synaptic transmission, positive regulation of neuron projection development, ion transport, and the regulation of blood-brain barrier permeability (Extended Data Fig. 8c). Hence, these are novel candidates involved in the functional maturation of cerebellar astrocytes. We also observed trajectory-specific TRs, whose combination defines a transcriptional code guiding the differentiation of Bergmann vs Astro and, in turn, of distinct types and subtypes. Along all BG trajectories ( Fig. 5d ), we predicted a progressive increase in Gli1 , consistent with its established role as a downstream effector of SHH signaling essential for BG maturation and maintenance of their molecular identity 24 , 55 . Smad3 , similarly specific, may act downstream of Zeb2 (or SIP1 - Smad-interacting protein 1-) in mediating TGF-β signaling during BG differentiation 56 . Conversely, along the WMA_ Pgm5 , GLA_ arch , and CNA_ Gpc5 Astro trajectories, high Stat1 and Stat3 implicate engagement of the JAK-STAT signaling pathway, consistent with its well-known role in cytokine-induced astrocyte differentiation 57 ( Fig. 5d ). Notably, Irf1 (Interferon regulatory factor 1), promoter of STAT1 DNA binding 58 , was high in WMA_ Pgm5 and WMA_ Npy trajectories, pointing to a role in the acquisition of WMA type identity (Extended Data Fig. 8d). Ar was instead specific for the CNA_ Gpc5 trajectory ( Fig. 5e ), pointing to a role for androgen signaling in CNA maturation and/or maintenance. GO analysis of Ar targets revealed significant enrichment of genes involved in neurotransmitter uptake (Extended Data Fig. 8e), pointing to its contribution to the specialized homeostatic role of CNA within the cerebellar nuclei. Altogether, these TRs appear to guide cell fate decisions of cluster 4 and 1 multipotent progenitors. Consistent with their role in shaping astrocyte type identity, many of the TRs described above were also active in the hybrid Npy -expressing and Vesic subtypes. Etv5 , Gli1 and Smad3 were active in BG_ Npy , while Stat1 , Stat3 , Irf1 in WMA_ Npy ( Fig. 5c,d and Extended Data Fig. 8d), in line with their respective identities. Along the Vesic trajectory, shared regulators ( Foxo1 , Nfia , Etv4 , Etv5 , Rara , Bcl6 , Ikzf2 , Maf and Pdlim5 ; Fig. 5c ), as well as Bergmann-( Gli1 , Smad3 ) or Astro-specific TRs ( Stat1 , Stat3 ; Fig 5d ) were active at late phases of maturation, when the fraction of mature astrocytes increases (Extended Data Fig. 8a). This combination further supports the hybrid identity of Vesic, comprising BG_ Kcnip4 and GLA_WMA_ Pde10a ( Fig 2f,g ). Other regulators were instead subtype-specific. Both BG_ Npy and WMA_ Npy displayed a highly specific predicted activity of several TRs associated with progenitor-state maintenance, including members of the JUN ( Jun, Jund ) and FOS ( Fos ) families, as well as Klf6 , Hey1 , Ybx1 , Atf4 , and Egr1 22 . These regulators, active in the early maturation phases of the other trajectories, increase later only in these two subtypes ( Fig. 5f ), consistent with their progenitor-like features. They also show elevated expression and activity of Acaa1a , a key enzyme in lipid catabolism, suggesting a reliance on alternative metabolic substrates to support their energetic demands (Extended Data Fig. 8f). Notably, BG_ Npy also exhibited both strong expression and activity of Sox1 and Msx2 (Extended Data Fig. 8g). While Sox1 is a known marker of specialized radial astrocytes in the hippocampus with demonstrated neurogenic and gliogenic potential in vivo 59 , Msx2 shows enriched expression in human BG, based on snRNA-seq data 60 . Taken together, SCENIC analysis identified combinations of transcriptional regulators specific for distinct astrocyte types and subtypes and therefore potentially orchestrating divergent, convergent or restricted fate decisions of cerebellar astroglial progenitors. Lineage-specific TRs contribute defining final astrocyte identity We previously identified separate lineages ( Fig. 4 ) contributing to astrocyte heterogeneity. Intriguingly, we found molecular traces of each embryonic origin persisting in the distinct trajectories. Along the Vesic trajectory, several TRs characteristic of the cerebellar rhombic lip (RL) and its derivatives were highly active - including Barhl1 , Barhl2 , Eomes , Etv1 , Otx2 , and Uncx - ( Fig. 5g ). This regulatory signature supports the notion that Vesic represent a separate astrocyte lineage originating from the RL. Along the trajectories originating from cluster 4 or cluster 1, which also correspond to distinct lineages ( Fig. 3d ), the two Krüppel-like factors (KLFs) Klf3 and Klf5 displayed striking complementary activity patterns ( Fig. 5h ). While other KLFs are known to be involved in cerebellar development and function 61 , 62 , the precise role of Klf3 and Klf5 remains to be elucidated. A similar pattern was followed by En1 and En2 vs Dbx2 and Gbx2 ( Fig. 5i ). On the one hand, En1 and En2 - homeobox genes master regulators in cerebellar development and known to persist up to late stages in many cerebellar cell types (including both VZ and RL-derivatives 63 ) - were both expressed and predicted to be active along the trajectories originating from clusters 0, 4, 17 and 7, covering both VZ and RL progenies. On the other hand, Dbx2 and Gbx2 were high in cluster 1 and its derivatives. While Gbx2 is known to be expressed in rhombomere 1 and required for cerebellar development during early embryogenesis but not at later stages 64 , both these genes are typically known for their role in hindbrain patterning 65 , 66 , consistent with the putative extra-cerebellar origin, in common with oligodendrocytes, of this cluster. A further link between cluster 1 and oligodendrocytes was represented by the activity of Foxb1 ( Fig. 5e ), known to mark bipotent progenitors capable of generating both oligodendrocytes and astrocytes in the medulla and thalamus 67 . In parallel, Egfr , implicated in glial lineage plasticity and in the regulation of astrocyte-to-oligodendrocyte conversion and vice versa 68 , was expressed in both cluster 1 and 18, at the onset of the oligodendrocyte lineage. The Fli1 regulon also showed a similar pattern of activity in both clusters (Extended Data Fig. 8h). Given its previously uncharacterized role in glial development, Fli1 may represent a relevant target to explore the potential switch between astrocyte and oligodendrocyte fates from a common progenitor. Overall, SCENIC analysis allowed to uncover a TR code underlying the emergence of cerebellar astrocyte heterogeneity ( Fig. 5j ), which comprises long lasting traces developmentally inherited from their progenitor sources and that contribute to defining their final identity. Discussion This study provides a comprehensive definition of molecularly distinct cerebellar astrocyte types and subtypes, and reconstructs in silico the developmental mechanisms underlying their emergence, revealing multiple maturation trajectories. These trajectories encompass fate divergence, convergence and restriction, involving multiple progenitor sources that share developmental origin with other cerebellar cell subclasses. Additionally, this study infers the TR modules delineating the unfolding of astrocyte lineages, up to the acquisition of type and subtype identities. Overall, these findings indicate that while environmental cues predominantly guide the acquisition of terminal astrocyte molecular fates, lineage specific factors constrain progenitor potency and impart defined traits in astrocyte subtypes. Sc/snRNAseq and ST identified established and novel adult astrocyte profiles, including four BG (BG_ arch , BG_ Mybpc1 , BG_ Npy and BG_ Kcnip4 ), two GLA (GLA_ arch and GLA_ Pde10a ) and three WMA subtypes (WMA_ Pgm5 , WMA_ Npy and WMA_ Pde10a ), along with new molecular markers for their discrimination. Notably, BG_ arch and BG_ Mybpc1 displayed complementary distributions across anterior and posterior lobules, respectively, consistent with prior observations 19 , 32 . This spatial segregation may depend on involvement in distinct circuitries driven by different afferents, potentially leading to molecular divergence 17 . Further on functional compartmentalization, Npy enriched BG subtype (BG_ Npy ) could correspond to the striped Npy-Gfp -labeled BG 69 . However, investigation of ST datasets (not shown) found no evidence of parasagittal BG banding. Thus, although current molecular profiling effectively captures anteroposterior BG heterogeneity reflective of distinct circuitries, it does not identify transcriptional substrates corresponding to biochemical and functional parasagittal neuron topography 70 . Yet, BG_ Npy exhibited distinctive characteristics, and shared with WMA_ Npy traits reminiscent of quiescent neural stem-like cells. These shared features partly recapitulate those of a subset of astrocytes in the neocortical WM/corpus callosum, known to actively support astrogenesis 22 . Moreover, BG_ Kcnip4 and GLA_WMA_ Pde10a exhibited a unique molecular profile denoting glutamatergic gliotransmission. Their true astrocyte identity is supported by expression of canonical Bergmann or Astro markers, activation of TR involved in astrocyte differentiation, and consistent detection across diverse datasets (adult and postnatal snRNA-seq; ST dataset completely devoid of non-astrocytic cells). Of note, these clusters closely resemble glutamatergic astrocytes found in the hippocampus and substantia nigra 29 , aligning with formerly reported gliotransmission in the cerebellum 71 . GO analysis captured a broad functional divergence between the two Bergmann and Astro groups. While Bergmann main functions in maintaining cerebellar structural integrity, regulating glutamatergic inputs, and relying on Shh for their cellular identity 13 were confirmed, Astro roles, much less characterized, were found primarily related to energy homeostasis and neuroprotection. Yet, since BG are known to display high glucose transport and metabolic activity 72 , the enrichment of energetic pathways in Astro may reflect compensatory transcriptional upregulation in response to a lower cellular density compared to BG, rather than indicating an exclusive specialization. In contrast, the stronger neuroprotective signature in Astro could point to a genuine functional enrichment. Oxidative stress and lipid peroxidation are more pronounced in the PCL and ML across multiple disease models 73 – 75 and amyloid plaques in Alzheimer’s disease are predominantly found in the ML 76 . This suggests a higher resilience of other cerebellar regions possibly supported by enhanced astrocyte properties. The in silico reconstruction of the maturation trajectories revealed a surprisingly complex picture of astrocyte generation from multiple perinatal sources with divergent, convergent, or restricted trajectories. Along these trajectories, TRs appeared to operate at different stages according to combinatorial modules coding regional patterning and regulating lineage-specific, subclass-specific and astrocyte type/subtype-specific features ( Fig. 5j , Supplementary Table 9). First components of the TR code were the regional patterning factors En1/En2 or Dbx2 / Gbx2. These factors are active in VZ-and RL-derived cerebellar lineages or within the hindbrain at the embryonic stages when astrogliogenesis is initiated and were associated with distinct trajectories in our data. Further TR modules progressively overlap, delineating VZ vs RL origins and contributing to the acquisition of specific astrocytic traits. Fate divergence towards BG subtypes, GLA_ arch and WMA_ Pgm5 emerged from multipotent cells (cluster 4), well matching formerly identified PWM progenitors 15 , now defined by the expression of Angpt1 and Aqp4 . Clonal data support that these lineages share common embryonic VZ progenitors with GABAergic interneurons, thereby extending former evidence for postnatal bipotent progenitors 14 and confirming former genetic studies 77 , 78 . All of the aforementioned divergent trajectories are delineated by the expression and activity of the cerebellar patterning factors En1 and En2 , consistent with their origin from the cerebellar VZ 15 , and of Klf3 , whose activity pattern, complementary to that of Klf5 (active in Dbx2/Gbx2- associated derivatives), warrants further investigations. Along time, these factors are associated with fundamental astroglial fate determinants (e.g. Hes1, Hes5, Nfia, Foxo1 ), and specific promoters of either Bergmann or Astro fates. Bergmann trajectories exhibited active Shh and TGF-β signaling whereas Astro displayed cytokine responses (Supplementary Table 9). In addition, Irf1 association with WMA fates suggests a role for interferon in the maturation of this type. Notably, as mentioned before, the BG_ Npy trajectory was distinctively enriched for TR particularly active in progenitors (transcription factors of the AP-1 family Jun , Jund , Fos , Atf4 ), suggesting that this subset may possess particularly plastic properties 22 . Similar features were shared by WMA_ Npy which nevertheless showed an independent restricted origin from cluster 17. While clonal tracing supported a common ancestry with other VZ-derived astrocytes (i.e., cluster 4 derivatives), at the onset of the WMA_ Npy trajectory several RL-typical transcription factors such as Barhl1 , Barhl2 , Eomes , Etv1 , Otx2 and Uncx were found activated. This may suggest a specific ontogenic origin, possibly within the “posterior transitory zone”, a VZ region immediately anterior to the morphologically defined RL where Atoh1 lineage progenitors possess an astrogenic potential 79 , 80 . Another restricted trajectory moved from cluster 7 to form glutamatergic Vesic astroglia, comprising the BG_ Kcnip4 and GLA_WMA_ Pde10a subtypes. Sustained expression and activity of the RL-typical transcription factors mentioned above, combined with En1/En2 , together with clonal analysis and spatial mapping all point to their RL origin. Interestingly, astroglia fate induction appears to occur through the rather late activation and expression of general astrocyte or type-specific TRs. The contribution of the RL to astrogenesis has long been debated, with conflicting evidence both supporting and opposing this possibility 81 . Our data now provide compelling support in favor of a RL contribution. This offers a novel perspective on lineage relationships and raises the possibility that similar ontogenetic programs may give rise to analogous astrocyte populations in other brain regions 29 . Further, additional multipotent progenitors were identified in cluster 1. However, within this developmental trajectory, both the fate potency and inferred TRs appeared to diverge from those characterizing cluster 4. Cluster 1 was predicted as unique source of CNA, in parallel with divergence onto GLA_ arch and WMA_ Pgm5, also converging from cluster 4. Whether this reflects full phenotypic convergence, as described for specific neuronal types 82 , remains unclear. One lineage may eventually dominate, or lineage-specific traits may persist in ways not captured by transcriptomic profiling. Supporting the latter hypothesis, GAT-3 immunostaining revealed a few positive GLA and WMA in the deep cerebellum, in addition to its expression in CNA (Extended Data Fig. 6f). Since Slc6a11 (GAT-3 gene) is specifically enriched in cluster 1, this suggests that at least a small subset of GLA and WMA can be traced back to this origin. Notably, and in contrast to the En1/En2 -defined trajectories, cluster 1 and its derivatives exhibited predicted activity of the hindbrain patterning factors Dbx2 and Gbx2 , in association with Klf5 as well as general astroglial and Astro-specific TR, consistent with its fate potency. Ultimately, additional regulation by Foxb1 and Ar seems to operate towards CNA. Further analyses of cluster 1 also suggested formerly unknown origins for CNA and the other diverging trajectories. Spatial mapping onto embryonic ST datasets, expression of hindbrain TRs, its lineage relationship with oligodendrocytes, as well as CNA similarity with brainstem astrocytes 25 all argue against a VZ origin. Indeed, most cerebellar oligodendrocytes originate from an extra-cerebellar source 77 , 83 , 84 and CNA were rarely labeled in VZ-targeted in utero electroporation clonal analyses 15 . Moreover, this supports earlier suggestions of bipotent astrocyte-oligodendrocyte progenitors in the cerebellum, based on indirect in vivo and in vitro evidence of fate shifts between these glial types upon Bmi1 deletion or BMP pathway inhibition 85 . Molecular traces of such bipotency can also be found in the CNA trajectory. Notably, most of the TR predicted to regulate astrocyte type identity are downstream of diffusible extrinsic signals (e.g., SHH, TGF-β, IFN, cytokines, androgen, retinoic acid), highlighting the pivotal role of the local environment in shaping astrocyte identity. This suggests that spatial and/or temporal gradients of these signals across cerebellar layers may guide fate acquisition and maintenance, as shown for SHH 24 . Consistently, several of these signals exhibit dynamic secretion patterns during cerebellar development 86 – 89 and astrocyte types follow distinct temporal differentiation patterns 15 . In contrast, TR downstream of contact-dependent pathways (e.g. Notch) were less prominent, possibly due to the transient or spatially restricted nature of their activation, which may limit their detection by SCENIC. On the other side, lineage-related factors act cell-autonomously to restrict the fate potency, as exemplified by transplantation studies: when grafted into the developing cerebellar cortex, both cortical astrocyte progenitors and ACSA2+ cerebellar astrocyte progenitors (i.e. cluster 1) fail to generate BG, retaining their original fate potency 16 , 90 . These findings underscore that extrinsic signals alone are insufficient to impose astrocyte identities in the absence of a pre-existing intrinsic competence, like the cerebellum-specific patterning regulators En1/En2. By achieving an unprecedented resolution on cerebellar astrocyte heterogeneity and decoding the underlying regulatory landscape, this study provides a conceptual advance in understanding how astrocyte heterogeneity is developmentally achieved. These insights lay the foundation for exploring whether similar regulatory programs contribute to astrocyte specialization throughout the brain, with important implications for understanding their physiological functions and disease responses. Materials and Methods Experimental animals C57BL/6J (Charles River Laboratories), C57BL/6JRj (Janvier), CD1 (Charles River Laboratories), hGFAPeGFP (FVB/N-Tg(HGFAP::EGFP 91 ) and Ptf1a CreERTM 39 crossed with R26R EYFP/EYFP (R26R YFP 92 were housed at two to five animals per cage under a 12 h–12 h light– dark cycle (lights on from 07:00 to 19:00) at a constant temperature (22 ± 2 °C) and humidity (55 ± 10%) with ad libitum access to food and water. Mice were used at 2-3 months of age and at different postnatal (P) times as indicated in the Results. Mice of either sex were included. The day of vaginal plug detection was defined as E0, and the day of birth was considered as P0. All animal protocols in the present study were approved by the Swiss Federal and Cantonal authorities, by the Council Directive of the European Communities (2010/63/EU) or by the Government of Upper Bavaria (Germany) or by the Italian Ministry of Health. Tamoxifen injections in postnatal mice To induce Cre recombination in Ptf1a CreERTM × R26R YFP P0 mouse pups, we administered tamoxifen (Tx; 20 mg/ml solution dissolved in corn oil) via subcutaneous injection (600 μg/pup). LNP-TrackerSeq library preparation The TrackerSeq library, a piggyBac transposon-based system compatible with the 10x Genomics single-cell transcriptomic platform, was prepared following the protocol described in Bandler et al. 40 . This library enables in vivo lineage tracing by integrating synthetic nucleotide barcodes into the mouse genome. Briefly, the piggyBac donor plasmid (Addgene #40973) was modified to incorporate a Read2 partial primer sequence within the 3′ UTR of eGFP to facilitate retrieval in the 10x pipeline, as well as a sucrose selection system to eliminate empty plasmids. The lineage barcode oligo mix was cloned downstream of the Read2 partial primer sequence via Gibson Assembly. The assembled constructs were then transformed into E. coli, expanded, and purified using standard column-based methods. The resulting plasmid library was quality-checked by high-depth sequencing to assess barcode diversity and overrepresentation. For in vivo delivery, the purified TrackerSeq library was incorporated into lipid nanoparticles (LNPs). LNPs were formulated to encapsulate three plasmids: pCAG-Trackerseq, pCAG-GFP-KASH, and pEF1a-PBase, thereby enabling robust and heritable integration into the progenitor genome. Plasmid DNA was combined to a total of 20 µg. LNP encapsulation was performed using an automated microfluidic system following the manufacturer’s recommended protocol, which combines the plasmid mixture with a specific buffer and a proprietary lipid nanoparticle mix conceived for progenitor targeting. The resulting LNP-encapsulated plasmids were stored at 4°C. For quality control, LNP size distribution was determined by Dynamic Light Scattering (DLS) using a ZetaSizer (Malvern Panalytical). Samples were diluted in PBS, and measurements were performed at 25°C. Optimal LNP preparations typically exhibited a single size peak between 100-120 nm. LNP-TrackerSeq injections in mouse embryos Timed-pregnant mice were deeply anesthetized throughout the surgery duration with isoflurane/oxygen and a small abdominal incision was performed to access the uterine horns. Embryos were constantly maintained moist with physiological saline. On embryonic day (E) 11 or E14, 70nl of the LNP-TrackerSeq library were injected into the third ventricle of the embryos. 50ul of Temgesic was injected 30 minutes before the surgery and 6 hours after surgery for analgesia. Sample collection for sc/snRNA-seq Nuclei collection from the adult deep cerebellum Single nuclei from the adult deep cerebellum were obtained as follows. Brains from two P56 hGFAPeGFP or C57BL/6JRj mice were freshly dissected and immediately transferred to cold artificial cerebrospinal fluid (ACSF). Subsequently, they were embedded into 4% Agarose Low Melt at 37°C and cut parasagittally into 600-μm-thick sections. Sections where the CN were clearly visible were selected and the deep cerebellum, containing both the CN and the deep WM, was macrodissected with a microsurgical knife. Tissues from both animals were pooled together and further processed for nuclei isolation following the 10X Chromium recommended protocol CG000366 (Rev A) with small adaptations. Nuclei suspensions from hGFAPeGFP mice were incubated for 30min at 4°C with a PE-coniugated anti-NeuN antibody (Milli-Mark® anti-NeuN-PE, clone A60, Sigma, 1:500) to label neuronal nuclei and subsequently subjected to FACS sorting (BD FACSAria III Cell Sorter). For the FACS gating, we used side scatter versus forward scatter and 7-AAD staining (7-Aminoactinomycin D, ThermoFisher) to distinguish single nuclei from debris and doublets or aggregates. NeuN-PE unstained cells were therefore collected to enrich non-neuronal nuclei. After FACS, nuclei were suspended in FACS buffer at a final concentration of 3000 cells/µl and loaded into the 10x Chromium V3 system. Reverse transcription and library generation were performed according to the manufacturer’s protocol. Cell collection from the adult deep cerebellar WM Single cells of the adult deep cerebellar WM were isolated from 8-12 weeks-old C57BL/6J male mice (12 mice in total, in two independent experiments) as previously described 22 . In brief, parasagittal slices with 1mm intervals were obtained by placing the freshly dissected cerebella onto an acrylic adult mouse brain matrix slicer and the deep WM was manually dissected through examination under a stereomicroscope. This was achieved excluding the CN and the GL to avoid contamination from the surrounding GM territories. Single cells were isolated from the dissected tissues using the Papain Dissociation System (Worthington Biochemical Corporation, 60 minutes incubation) followed by the Dead Cell Removal kit (Miltenyi Biotec) and Myelin Removal Beads II (Miltenyi Biotec) according to the manufacturer instructions. Single cells were eluted and suspended in 1X PBS with 0.04% BSA at a final concentration of 1000 cells/µl for subsequent processing using the Single Cell 3’ Reagent Kits v3.1 (10x Genomics) following the manufacturer instructions. Cells and Nuclei collection from the postnatal cerebellum Single nuclei were isolated from postnatal cerebella ranging from P0 to P30. Brains from C57BL/6JRj, CD1 or hGFAPeGFP mice of both sexes (31 in total, in 22 independent experiments, Supplementary Table 1) were freshly dissected and immediately transferred to cold artificial cerebrospinal fluid (ACSF). In 12 experiments, brains were directly processed for nuclei isolation following the 10X Chromium recommended protocol (CG000366 Rev A) with minor adaptations. In the remaining 10 experiments, brains were embedded into 4% low-melting agarose at 37°C and cut parasagittally into 600-μm-thick sections. From each section, the anterior, central, and posterior lobules, as well as the deep cerebellum, were macro-dissected using a microsurgical knife. Adjacent sections containing the same anteroposterior (AP) regions were pooled together, with separate processing for those collected from the vermis and from three consecutive mediolateral (ML) portions of the two hemispheres (the same portions of the two hemispheres were pooled together). This approach ensured that spatial information along both AP and ML axes regarding the nuclei origin was preserved. All samples were subsequently processed for nuclei isolation as described above. To enable multiplexing of samples from distinct regions, ages or biological replicates (see Supplementary Table 1), single nuclei were stained with TotalSeqA™ HashTag Oligonucleotide (HTO) antibodies (BioLegend) following the manufacturer’s protocol with minor adaptations. Briefly, nuclei were resuspended in staining buffer, incubated with a TotalSeqA™ antibody cocktail for 30 min at 4°C, and washed twice with PBS. The stained nuclei from distinct samples were merged into a single multiplexed sample, filtered through a 40 µm strainer, and loaded into the 10x Chromium V3 system. Single cells were isolated from P4, P5 or P23 CD1 mice injected with the LNP TrackerSeq library at E11 or E14 (17 in total in 5 independent experiments, Supplementary Table 1). Briefly, cerebella were collected and maintained in Leibowitz medium supplemented with 5% FBS. Tissue dissociation was performed using the papain-based system (Worthington), following the manufacturer’s protocol. eGFP labeled cells were then isolated via flow cytometry using a BD FACSAria III Cell Sorter. Non-eGFP-expressing brain tissue served as a negative control to define background fluorescence in all FACS experiments. The sorted cells were processed using the Single Cell 3’ Reagent Kits v3.1 (10x Genomics) following the manufacturer instructions. Construction of sc/snRNA-seq, TotalSeqA™ HTO and TrackerSeq libraries Sc/sn barcoding and library construction were performed using the Chromium Single Cell 3’ Reagent Kits (v3 chemistry) and the Chromium Controller instrument (10x Genomics), following the manufacturer’s instructions. The amplification of the TotalSeqA TM HTO library was performed according to manufacturer instructions. Briefly, the cDNA amplification reaction was modified by adding 0.2 µM of an HTO-specific additive primer to enhance the yield of HTO-derived cDNA products. The reaction mix consisted of 50 μl Amplification Master Mix, 7 μl nuclease-free water, 5 μl cDNA additive, 2 μl cDNA primer mix, and 1 μl of the HTO-specific additive primer, for a total volume of 65 μl. Following cDNA amplification, HTO-derived cDNA (∼180 bp) was separated from mRNA-derived cDNA (>300 bp) using a two-step sparQ or SPRI bead-based selection (0.6X for separation, followed by 2X purification of the supernatant containing the HTO fraction). The purified HTO fraction was then amplified using KAPA HiFi HotStart ReadyMix (2X), with a total reaction volume of 100 μl containing 45 μl purified HTO cDNA, 50 μl KAPA HiFi HotStart ReadyMix, 2.5 μl Illumina TruSeq D70*_s i7 index primer (10 µM), and 2.5 μl SI-PCR oligo (10 µM). The amplified HTO library was further purified using a 1.6X sparQ or SPRI bead-based purification. The amplification of the TrackerSeq barcode library from RNA was carried out as previously described 40 , using the Q5 Hot Start High-Fidelity 2X Master Mix (NEB, #M094S) according to the standard NEB protocol. Each 50-μl PCR reaction included 25 μl of Q5 High-Fidelity 2X Master Mix, 2.5 μl of a 10 μM P7-indexed reverse primer, 2.5 μl of a 10 μM i5-indexed forward primer, 10 μl of molecular-grade H₂O, and 10 μl of cDNA as the template. Sequencing and Data Processing I llumina sequencing libraries were sequenced with HiSeq 4,000 or NovaSeq 6,000 after quality assessment with the Bioanalyzer (Agilent) with an average read depth of 20,000-30,000 raw reads per cell. Sequencing reads in FASTQ format were aligned to the mouse reference transcriptome (mm10-2.1.0) and converted into Unique Molecular Identifier (UMI) counts using Cell Ranger software with default settings (10x Genomics, version 6.0.1). Following sequencing, the applied HTO sequences were utilized for bioinformatic demultiplexing of nuclei and the resolution of individual samples. For TrackerSeq, preprocessing of R2 FASTQ reads involved trimming sequences flanking the lineage barcodes (BC). Barcodes shorter than 37 base pairs were discarded. Cell barcodes (Cell) were extracted from the corresponding Seurat dataset to generate a whitelist, which was then used to incorporate cell barcodes and UMIs into the read names of the lineage barcode FASTQ files. These processed FASTQ files were further analyzed to construct a sparse matrix in CSV format, where rows represented individual cells identified by unique cell barcodes, and columns corresponded to lineage barcodes. Only Cell–UMI–BC triplets supported by at least 10 reads and Cell–BC pairs with a minimum of 6 UMIs were retained for further analysis. CloneID assignment to cell barcodes was performed using Jaccard similarity and average linkage clustering, following the approach described by Wagner et al. 40 , 93 . General data processing and quality control Sc/snRNA-seq datasets were pre-processed with CellBender or SoupX to remove technical noise and ambient RNA, improving the overall accuracy and quality of the data 94 , 95 . Subsequent processing was performed using the Seurat pipeline (versions 5.0.1 or 5.0.2). Raw count matrices were imported into R (versions 4.3.2 or 4.3.3) and converted into Seurat objects using the CreateSeuratObject() function. Cells were filtered based on dataset-specific thresholds for gene detection (minimum and maximum number of genes per cell) and mitochondrial read percentage. Publicly available sc/snRNA-seq datasets Count matrices and associated metadata of adult mouse cerebellum sn/scRNA-seq datasets were downloaded from https://singlecell.broadinstitute.org/single_cell/study/SCP795 (Kozareva et al. 18 ), https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE246717 (Yao et al. 20 ), and https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE160471 (Kebschull et al. 21 ). Embryonic and postnatal available datasets were downloaded from https://www.ebi.ac.uk/ena/browser/view/PRJEB23051 (Carter et al. 32 ), https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE118068 (Vladoiu et al. 33 ). The Sepp et al. 19 dataset, used for both our adult and developmental studies, was downloaded from https://cellxgene.cziscience.com/collections/72d37bc9-76cc-442d-9131-da0e273862db Sc/snRNA-seq dataset integration Adult sc/snRNA-seq integrated dataset The distinct datasets were merged together in a single Seurat object, which was then split into individual layers, each corresponding to one dataset of origin (batch). Normalization using the SCTransform() function (regressing out known sources of unwanted variability, like sequencing depth and mitochondrial transcript proportion) and Principal Component Analysis (PCA) were performed on each layer. Then, to enable efficient integration of large-scale datasets, we applied the Sketch integration approach implemented in Seurat v5. From each dataset (layer) a representative subset of cells was selected using leverage score-based sampling. These sketched cells were integrated using a canonical correlation analysis (CCA)-based workflow to define a shared low-dimensional space. The full datasets were then projected onto this space, allowing harmonized analysis across all cells while minimizing memory usage. Following integration, clusters were identified using the FindNeighbors() and FindClusters() functions in Seurat. The resulting clusters were annotated based on the expression of literature-based cell type-specific marker genes as shown in Extended Data Fig. 1b. To analyze adult astrocytes, an “astrocyte score” was calculated using the Seurat function AddModuleScore(), based on the combinatorial expression of the astrocyte-specific marker genes Slc1a3, Slc1a2, Aldh1l1 , and Aqp4 . Cells/nuclei with an astrocyte score greater than 1 were subsetted for further analysis. This subset was then subjected to the same computational workflow described above, including batch-wise integration using CCA in Seurat V5, to generate the final integrated dataset of adult cerebellar astrocytes. This dataset did not comprise any cells from Kebschull et al. 21 , since the authors sorted NeuN+ neuronal nuclei, thereby excluding non-neuronal cell types, including astrocytes. Postnatal sc/snRNAseq integrated dataset For the postnatal dataset, we adopted a similar Seurat v5 integration strategy but without sketching the data. Instead, the full dataset was merged into a single Seurat object and split into layers corresponding to homogeneous experimental cohorts - defined by shared technical parameters such as 10X chemistry, mouse strain, and sample type (e.g., nuclei or cells) - to account for batch effects while preserving biological variation. Normalization was performed independently per layer, followed by CCA-based integration, and clustering. Slc1a3 -expressing cells were subsetted and re-integrated following the same workflow adopted for the whole dataset. The same approach was also used for the integration of our perinatal Slc1a3 -expressing cells with the embryonic-perinatal cells of publicly available datasets 19 , 32 , 33 , each treated as separate experimental cohort. Spatial mapping onto ST datasets To infer the spatial location of the clusters of either the adult or postnatal datasets, identity transfer of the cluster identities was performed on distinct ST datasets. Visium The sketched representative subset of the adult whole cerebellum dataset (see above), the adult astrocyte datasets or the postnatal Slc1a3 subset were integrated with the adult mouse brain sagittal slice “posterior 1” of the Visium ST dataset “stxBrain” available on Seurat (Mouse Brain Serial Section 1 (Sagittal-Posterior), Spatial Gene Expression Dataset by Space Ranger 1.1.0 (2020, June 23)). An “anchor”-based integration workflow in Seurat was applied, which enables the probabilistic transfer of annotations from a reference to a query set 96 . The spatial reference dataset and the adult astrocyte dataset were normalized using the SCTransform function, which builds regularized negative binomial models of gene expression. Dimensionality reduction using the RunPCA function and label transfer using the functions FindTransferAnchors and TransferData were performed. This procedure outputs, for each spatial spot, a probabilistic classification for each of the sc/snRNAseq-derived cell states. These predictions were added as a new assay in the Seurat object for visualization using the function SpatialFeaturePlot. Merfish The adult astrocyte clusters were aligned with 16 sagittal cerebellar slices of the adult mouse cerebellum 20 . Prior to alignment, cells in each slice were subjected to cluster analysis with Seurat with FindClusters at resolution 0.5 96 . Cells in slices were then scored based on the combined expression of astrocytic markers ( Aqp4 , Aldh1l1 , Sox9 ) with the Seurat function AddModuleScore. Clusters containing the majority of cells with high expression of those markers and cells with scores higher than 2 standard deviations from the mean of the astrocytic score were subsetted and then used for alignment with the adult astrocyte dataset. To align spatial cells with the adult astrocyte dataset an approach based on a loss function of cosine similarity between single cell and spatial data was performed with the Python package Tangram 97 . Clusters from the adult astrocyte dataset were fitted on each cerebellar slice with the tangram function map_cells_to_space based on the normalized expression of a set of training genes. Training genes were defined with the Scanpy 98 function tl.rank_genes_groups and the tangram function pp.adatas, this process scored each cell in the slices based on their probability of belonging to the clusters of the adult astrocyte dataset. Cells in slices were assigned to a cluster identity when their score for belonging to the cluster was higher than 2 standard deviations from the mean of the identity score in the slice. Stereo Seq The subset of Slc1a3 -expressing cells from postnatal dataset were aligned with two sagittal mouse brain sections from embryonic day (E) 16.5 41 and one sagittal mouse brain section from postnatal day 7 37 . Prior to alignment cells belonging to the developing cerebellum and nearby structures were subset and then used for alignment. Astroglial-like progenitors and maturing astrocytes were aligned with the spatial slices with the package Tangram 97 as described in the Merfish paragraph above. Layer enrichment and lobule enrichment analyses Merfish For the layer enrichment analysis, each cell in each Merfish cerebellar slice was assigned to one of four layers: cerebellar nuclei (CN), white matter (WM), granular layer (GL) and molecular layer/Purkinje cell layer (ML/PCL). This manual annotation was performed using a custom R script, based on both cell density patterns observed in the Merfish images and, when available, the expression of regionally enriched marker genes ( Sox10 for WM, Calb1 for ML/PCL, Slc17a7 for GL, while for cerebellar nuclei a pre-existing annotation was adopted). For the cerebellar nuclei, a pre-existing annotation was used. Cells located in extracerebellar regions or in the choroid plexus were excluded from the layer assignment and downstream analysis. For the lobule enrichment analysis, a similar manual curation strategy was applied to assign cells in each MERFISH slice to the corresponding cerebellar lobule. In both analyses, enrichment scores for each cluster identity i within each layer or lobule j were computed as: where n ij is the number of cells from cluster i assigned to layer/lobule j , and N j is the total number of cells in layer/lobule j (across all cell types). Kozareva et al. snRNA-seq dataset To compute the lobule enrichment scores for single-nucleus transcriptomic profiles, astrocytes from the dataset by Kozareva and colleagues 18 were subset from the integrated adult sn/scRNA-seq dataset. The original metadata indicating the lobular origin of each nucleus was combined with our cluster identity annotations. Lobule enrichment scores were then calculated using the same formula described above, where the number of nuclei belonging to each cluster and lobule was normalized by the total number of nuclei per lobule. Cell types/subtypes annotation To assign cell type identities across the whole cerebellum datasets, we used a deep neural network classifier, implemented in PyTorch, following the strategy described by De Ceglia et al. 29 . Briefly, a multiclass prediction model was trained using pre-annotated single-cell transcriptomic data. Cell types with fewer than 50 annotated cells were excluded from the training set to ensure robust classification. The dataset was randomly split into training (80%) and testing (20%) subsets. The network architecture consisted of four fully connected layers. Hardtanh was used as an activation function in the first layer, and ReLU for the subsequent layers. Dropout was applied during training (50% at the input of the first layer, 30% at the following layers) to prevent overfitting. Additionally, the weights of the first layer were normalized to unit norm for each node, and input normalization was adjusted accordingly. Training was carried out for 50 epochs using random sampling to correct for class imbalance.To improve interpretability and reduce noise, a second training step was performed after pruning the input features to retain the 100 genes with the highest weights in the first layer. The final model was validated on the held-out test set and applied to the integrated dataset to predict the identity of each single cell. Clusters were assigned a label based on the majority class prediction (>60% of cells within a cluster). Final annotations (Extended Data Fig. 1B, Extended Data Fig. 3B) were manually curated by combining the predicted identities with the expression of known marker genes. Annotation of adult astrocyte subtypes was achieved by integrating the most specific DEG of each cluster with spatial localization predictions. Specifically, cells were assigned to one of four major astrocyte classes - Bergmann glia (BG), granular layer astrocytes (GLA), white p>matter astrocytes (WMA), or cerebellar nuclei astrocytes (CNA) - based on their predicted regional distribution within the cerebellum (i.e., localization to the PCL/ML, GL, WM, or CN, respectively). Spatial assignment was informed by three complementary sources: (i) visual inspection of identity transfer results from the Visium ST dataset; (ii) visual inspection of identity transfer on the MERFISH cerebellar slices; and (iii) results from the quantitative layer enrichment analysis. When at least two of these three lines of evidence were concordant, the corresponding regional identity was assigned to the cluster. Annotation of astrocyte subtypes within the postnatal Slc1a3 subset was achieved by subsetting and re-integrating (as described above) cells from the more mature astrocyte clusters (i.e. clusters 9, 14 and 20) and collected at late developmental stages (from P10 onwards). Final subtype identities were assigned using a multi-layered approach that combined: (i) expression scores of differentially expressed genes (DEGs) characteristic of adult astrocyte subtypes (Supplementary Table 1, Fig. 2f ); ii) the expression of adult subtype-specific genes ( Fig. 2g ); and iii) spatial mapping onto the adult Visium mouse cerebellum ST dataset. Differential gene expression (DEG) analysis Differential gene expression analysis was performed on R with the Seurat function FindAllMarkers() using Wilkoxon Rank Sum test, retaining only genes with an average log2 fold change (avgLog2FC) major or equal to 0.5 and expressed by at least 25% of the members of the cluster. Gene ontology (GO) analysis GO enrichment analysis was performed using the ViSEAGO package 99 . DEGs identified from the pairwise comparison between Bergmann and Astrocyte groups were analyzed for enrichment of biological process GO terms using Fisher’s exact test. Only terms with a p-value ≤ 0.05 were considered significantly enriched. Enriched terms were then clustered based on semantic similarity, and manual curation was performed to highlight the most informative terms relevant to adult astrocyte biology. The same procedure and statistical parameters were applied independently to each astrocyte subtype. Pseudotime Computation Pseudotime analysis of the Slc1a3 -expressing postnatal subset was performed using the Monocle3 R package 100 . The Seurat object was converted into a cell_data_set Monocle object. Size factors were estimated, and dimensionality reduction and preprocessing were carried out prior to Leiden clustering on the UMAP embedding previously computed with Seurat using the cluster_cells() function, allowing identification of a single main partition. A trajectory graph process was then computed. Root nodes were selected based on their proximity to cells with high proliferation gene scores and belonging to perinatal clusters (i.e., clusters 0, 1, 2, 3, 4, 5, 6, 7, 10, 11, 15, 17, 18, and 21; see Fig. 2D ). Maturational trajectories and lineage Inference To infer the maturation trajectories within the postnatal Slc1a3 -expressing subset, we employed three distinct algorithms to compute fate probabilities (i.e. the probability of each cell to be part of a maturation trajectory of a defined lineage). Two of these methods were implemented using the CellRank2 package 101 , while the third was based on the URD framework 102 . Given the computational nature of these analyses and the inherent complexity of developmental processes, we reasoned that combining multiple inference strategies would enhance the robustness and consistency of the resulting trajectories. Anterograde random walks with CellRank2 Three distinct transition matrices (TMs) were generated in CellRank2 to model different aspects of cellular progression: one based on pseudotime values previously inferred with Monocle3 (pseudotime kernel), one based on experimental time using the postnatal age of tissue collection (experimental time kernel), and one based on transcriptional similarity among cells (connectivity kernel). These TMs were used to simulate biased anterograde random walks (aRWs), favoring developmental progression. Two separate simulations were performed: the first combined the pseudotime TM with the experimental time TM, while the second combined the pseudotime TM with the connectivity TM. In both cases, starting points were defined as cells collected at P0, belonging to the perinatal clusters (as described above), and displaying pseudotime values below 2.5. The inference of fate probabilities was carried out over multiple iterative rounds of random walk simulations. In the initial round, terminal states were predefined as mature astrocyte subtypes, oligodendrocytes, interneurons, and ependymal cells. In subsequent rounds, terminal states were redefined based on the clusters most enriched in fate probability from the previous round. Each round yielded a fate probability matrix indicating the likelihood of each cell reaching each terminal state. Probabilities were then averaged across cells belonging to the same cluster, generating an estimate for each cluster’s involvement in the maturation trajectories. Finally, these values were normalized such that, for each terminal state, the sum of probabilities across all clusters equaled 1. Retrograde random walks with URD Maturation trajectories were also inferred using the URD R package. A URD object was created from the subset of Slc1a3 -expressing cells. A diffusion map was computed on the integrated UMAP previously generated with Seurat, using a sigma value of 20. URD pseudotime was then calculated by selecting as roots the cells collected at P0, belonging to the perinatal clusters (as described above), and displaying pseudotime values below 2.5. These same cells were also used as roots to simulate retrograde random walks (rRWs). As with CellRank aRWs, multiple rounds of rRWs were performed. In the initial rounds, simulations were biased to start from mature astrocyte subtypes, oligodendrocytes, interneurons, or ependymal cells. In subsequent rounds, the top-ranked clusters from the previous round were used as new starting points. Each round yielded a visitation frequency matrix indicating the likelihood that a walk initiated from a specific starting point would transition through each cell en route to the roots. Visitation frequencies were then averaged across within the same cluster to estimate each cluster’s involvement in the maturation trajectories. These values were finally normalized so that, for each starting point, the sum of probabilities across all clusters equaled 1. Definition of consensus results To generate a consensus estimate of fate probabilities across methods, the matrices obtained from the two CellRank2 analyses and the URD analysis were combined into a single consensus matrix. First, a logit transformation was applied to each individual probability ( P ) that cluster i contributes to the terminal state j as computed by method k : These logit-transformed values ( L ) were then averaged across methods: The resulting averaged values were then transformed back into probabilities using the inverse logit function: Finally, the probabilities were normalized such that for each terminal state j , the sum of probabilities across all clusters i equaled 1: This approach yielded a robust consensus estimate of the likelihood that each cluster contributes to the maturation trajectory toward a given terminal state. Validation of consensus results To validate the predictive accuracy of our trajectory inference strategy, we leveraged the oligodendrocyte lineage, which undergoes well-characterized differentiation steps marked by stage-specific gene expression 103 , 104 ( Fig. 2C ; Extended Data Fig. 5D–F). In the first round of random walks, the model accurately predicted that mature oligodendrocytes (Cluster 13, characterized by high Mog expression) originate from pre-myelinating oligodendrocytes (Cluster 16, marked by Bcas1 expression). In the second round, Cluster 16 - now designated as the new terminal state - was traced back to Cspg4 -expressing oligodendrocyte progenitor cells (OPCs) in Cluster 8. In the third round, Cluster 8 was further predicted to derive from Cluster 18, which expresses multiple pre-OPC markers such as Rfx4 , Sall3 , Ptprz1 , and Ncald . This strategy also successfully reconstructed the interneuron lineage within the Slc1a3 -expressing subset. The predicted trajectory progressed from transient amplifying progenitors (TAPs) in Cluster 11 - co-expressing Ptf1a , Ascl1 , and Mki67 - to Lhx1/Lhx5/Mki67 + TAPs in Cluster 12, and ultimately to Pax2/Gad1/Gad2 + interneuron precursors in Cluster 5 32 , 105 , 106 (Extended Data Fig. 5G). This approach successfully reconstructed differentiation steps up to clusters enriched at intermediate stages ( Fig. 2D , middle panel). However, additional rounds of random walks to establish hierarchical relationships among clusters enriched at perinatal stages ( Fig. 2D , top panel) were not feasible due to intrinsic constraints of the methods, which requires the definition of root cells and tips - two sets that cannot overlap by definition - and the assignment of a pseudotime score, which is uniformly low and similar across cells collected at the earliest stages (Extended Data Fig. 5B). Lineage coupling correlations analysis To investigate lineage relationships among the clusters within the Slc1a3 postnatal subset (designated as “Slc1a3_” followed by the cluster number, e.g., “Slc1a3_cl0”) and the full postnatal dataset from which were removed cells with no neural tube origin (Microglia and Fibroblasts) and with no clear identity, we analyzed clonal overlaps, lineage coupling z-scores, and correlation patterns, adapting methodologies previously described 40 , 42 , 93 . In this analysis, a TrackerSeq clone was considered to be “shared” between two groups if it included at least two cells assigned to each cluster. We counted the number of shared clones between every pair of clusters and then compared these counts to a null distribution generated by randomly permuting the cluster annotations across 10,000 iterations. From this, we computed a z-score for each observed value to quantify the deviation from random expectations. Subsequently, we calculated Pearson correlation coefficients between the z-score profiles of all pairs to assess the similarity in their clonal sharing patterns. These correlation matrices were visualized through hierarchical clustering. For further methodological details regarding the computation of lineage coupling and correlation metrics, please refer to the original TrackerSeq publication 40 . Scenic gene regulatory network analysis Single-cell regulatory network inference and cell-type–specific transcription factor (TR) activity were assessed using the pySCENIC pipeline 107 . The pipeline was executed with the command line interface of pySCENIC 0.12.1, encapsulated in a Podman container to ensure reproducible software dependencies and environment. The input consisted of the gene-expression matrix containing the counts of the subset of Slc1a3 -expressing cells from the postnatal dataset. First, co-expression modules were inferred with the pyscenic grn command. This step identifies sets of genes that exhibit correlated expression patterns across cells. The co-expression analysis was performed using grnboost2, which ranks potential TR–target relationships based on shared expression patterns. The output from this step is a table of TR– target associations that serve as candidates for subsequent refinement. Next, the initial TR–target associations were pruned and refined with the pyscenic ctx command, integrating both cis-regulatory sequence motif analyses and motif annotations from https://resources.aertslab.org/cistarget/ . In this stage, each TR’s candidate targets are evaluated for the presence of known binding motifs in their promoter and enhancer regions. Targets lacking significant motif enrichment are removed, ensuring that putative TR–target relationships better reflect established regulatory interactions. The output from this step is a set of regulons , each composed of a TR and its high-confidence targets. Finally, the activity of each regulon across individual cells was quantified using the pyscenic aucell command. This calculates an “Area Under the Curve” (AUC) score that reflects the extent to which a regulon’s target genes are expressed in a particular cell. After running the SCENIC pipeline, we linked regulon activity with the consensus matrix of maturation scores obtained as described above.For each trajectory in the consensus matrix, we computed Pearson correlation coefficients between its generation probability (consensus logit values) and each transcription factor’s AUC values. Statistical significance of these correlations was assessed using the built-in cor.test function, followed by multiple-test correction with the Benjamini–Hochberg false discovery rate (FDR). Transcription factors meeting defined significance (e.g., FDR ≤ 0.05) and minimum correlation thresholds (e.g., Pearson’s ρ > 0.1) were considered positively correlated with the corresponding lineage’s maturation trajectory. Immunohistochemistry Under anesthesia, animals were transcardially perfused with an appropriate volume of 4% paraformaldehyde (PFA) in 0.12 M phosphate buffer (PB), pH 7.2–7.4. Brains were removed, stored overnight (o/n) in the same fixative at 4 °C, washed in PBS, and finally cryoprotected in 30% sucrose in 0.12 M PB. The cerebella were then embedded and frozen over dry ice in OCT (TissueTEK), sectioned in the parasagittal plane at 30 μm using a cryostat and placed directly onto glass slides (E14.5-P4 cerebella). For immunolabeling, sections were incubated o/n at room temperature with the appropriate primary antibodies dissolved in PBS with 2% bovine serum albumine (BSA, Sigma-Aldrich) and 0.25% Triton X-100 (Sigma-Aldrich): Anti-ACSA-2 (1:250, rat, monoclonal, Miltenyi Biotec), anti-Aquaporin 4 (AQP4, 1:2,500, rabbit, polyclonal, Sigma Prestige Antibodies,), anti-GAT3 (1:100, rabbit, monoclonal, Abcam), anti-GFP (1:700, chicken, polyclonal, IBA Lifesciences), anti-SOX9 (1:600, goat, polyclonal, R&D). Sections were then exposed for 2 h at room temperature to secondary species-specific antibodies conjugated with Alexa Fluor 555 (1:500; Invitrogen; anti-rabbit, anti-rat or anti-goat) or Alexa Fluor 488 (1:500, Invitrogen, anti-chicken). Cell nuclei were visualized using 4′,6-diamidino-2-phenylindole (DAPI; Fluka). After processing, sections were mounted on microscope slides with Tris-glycerol supplemented with 10% Mowiol (Calbiochem).Histological specimens were examined using a Nikon E-800 microscope, and high-resolution images were acquired with the Zeiss Axioscan slide scanner. Author contributions V.C., A.B. and L.T. conceived the project and designed the experiments; V.C., I.V., B.X., L.S.-F, A.L., and E.M. performed the experiments; R.B., M.G, J.F.-S. contributed data, conceptualization; E.B., contributed methodology; V.C., G.T. and L.T. analyzed the data; V.C., G.T., A.B. and L.T. wrote the manuscript with input from all authors. Funding This work was funded by the ERC starting grant (to L.T.; CERDEV_759112), the SNSF grant (to L.T.; 31003A_182676/1), the European Union’s funded Consortium NSC-Reconstruct: Novel Strategies for Cell based Neural Reconstruction (to A.B.; H2020, GA no. 875758), the Banca d’Italia grant (to A.B. and V.C.; Spacer, 2022), the European Advanced infraStructure for Innovative Genomics, H2020 EASI-Genomics grant PID15185 (to A.B. and V.C.), the Italian Ministry of Research and Education (MIUR) national project “Dipartimenti di Eccellenza 2018–2022 and 2023-2027” awarded to the Department of Neuroscience “Rita Levi Montalcini” (University of Turin). Valentina Cerrato was supported by a Post-doctoral fellowship by Fondazione Umberto Veronesi (“Post-doctoral Fellowship Travel Grant-anno 2020” to V.C.), by salaries on the Italian MUR grant PRIN-PNRR 2022 (ID: P20225Z3J5 to E.B.) and on the NSC-Reconstruct grant. Extended Data Figures Download figure Open in new tab Extended Data Fig. 1 Molecular and functional annotation of the distinct clusters. a, Schematic representation of the astrocyte sampling in publicly available datasets shows the low enrichment of non-cortical astrocytes. Circle size is proportionate to the number of cells sampled. Names correspond to the annotations used by the authors of the datasets. b, Expression of marker genes used for cell type annotation.The bottom panel shows the classification of cell types into classes (Neurons, Glia, Immune, Vascular, Fibroblast, Pericytes, Choroid) and subclasses (Ihhibitory and Excitatory neurons, Astrocytes, Oligodendrocytes, Ependymal cells, Immune, Vascular, Fibroblast, Pericytes, Choroid). Insets on the right show the spatial mapping of selected cell types in the Visium dataset. c, Expression of selected DEGs of WM astrocytes belonging to the Cer_1 and Cer_2 clusters in Bocchi et al. 22 highlights correspondence with the WMA_ Pgm5 or BG_WMA_ Npy respectively. d, GO analysis (left) and gene expression (right) point to a NSC-like signature of BG_WMA_ Npy cells. e, Expression of Syt1 , Snap25 and Slc17a7 - either individually or combined in a score - show an enrichment in the GLA_WMA_ Pde10a and BG_ Kcnip4 clusters. The dotplot on the right shows the enrichment of selected DEGs of hippocampal “glutamatergic” astrocytes 29 in the same clusters. f, GO analysis performed on the DEGs of the GLA_WMA_ Pde10a cluster shows an enrichment of terms associated with synaptic transmission and vesicular exocytosis. g, Expression of genes associated with selected enriched GO terms in Bergmann vs Astro. Download figure Open in new tab Extended Data Fig. 2 Mapping adult cerebellar astrocyte clusters through ST. a, Spatial mapping of BG_WMA_ Npy (highlighted in the UMAP plot, top left) in the Visium (top right) or Merfish (bottom) datasets. b, Spatial mapping of the cells belonging to the BG_WMA_ Npy cluster sampled from the whole cerebellum and cerebellar cortex (Yao 20 and Kozareva 18 datasets, top) or from the deep cerebellum (our and Bocchi 22 datasets, bottom) highlights an enrichment in the ML/PCL or WM, respectively. The bar plots show the average layer enrichment calculated from Merfish predictions. c, Spatial mapping of Astro_ Slit2 in the Visium or Merfish 20 datasets show a broad distribution of these cells across all regions. d, Lobule enrichment in the vermis (top) or hemisphere (bottom) calculated from Merfish predictions for the clusters that populate the cortical layers. Bars indicate the average enrichment across multiple sections 土 SEM (n=7, vermis; n=9, hemisphere). e, Lobule enrichment in the vermis (top) or hemisphere (bottom) calculated for the astrocyte clusters that populate the cortical layers based on tissue sampling of the snRNA-seq Kozareva18. The spatial distribution of the BG_Mybpc1 cluster is consistent with that of the Mybpc1-expressing Bergmann_2 cluster in the same study. f,g, Clusters that show a lobule enrichment pattern in Merfish predictions ( f ) or Kozareva18 dataset ( g ) which is significantly different from the cerebellar astrocyte population as a whole (Pearson’s chi-squared); x axis shows −log10-transformed q values, with q value < 0.001 indicated by dashed vertical line; y axis shows the maximum lobule enrichment for each type/subtype, with lobule enrichment = 1 (comparable to a random distribution) indicated by dashed horizontal line. h, Venn diagram showing the intersections across the cells localized in the ML/PCL predicted as BG_ arch , BG_ Mybpc1 or BG_ Kcnip4 in the Merfish 20 dataset highlights a low level of overlap across the three populations (n= predicted cells). Download figure Open in new tab Extended Data Fig. 3 Overview of the multi-modal sc/snRNA-seq dataset of the whole postnatal mouse cerebellum. a, Schematic representation of the information contained within each experimental set of the newly generated postnatal dataset, sampling nuclei or cells across several developmental stages from P0 to P30. Biological and technical replicates are not shown. b, Expression levels of key cell type-specific markers used for cell annotation. c, Comparison of our postnatal dataset with publicly available datasets 19 , 32 , 33 , highlighting the improved coverage of the full repertoire of cerebellar cell lineages. UMAP plots display the integrated dataset across all developmental stages, split by dataset and colored by cell type. d, UMAP plots of the P0-P4, P7-P11 or P14-P30 integrated newly generated datasets showing Slc1a3 expression (top panels) and the Slc1a3 -expressing clusters that were subsequently subset for further analyses (bottom panels). Download figure Open in new tab Extended Data Fig. 4 Temporal dynamics and profiling of Slc1a3-expressing cells during postnatal cerebellar development. a, UMAP plot of the postnatal subset, colored by stage of sampling. b, Barplot showing cluster distribution across distinct stages. Clusters 9, 14 and 20, which contain differentiated astrocytes, become predominant after P10-P11. c, Bar plot illustrating the fraction of cells collected at each developmental stage after P10-P11 for each annotated astrocyte subtype. d, Spatial mapping of the astrocyte subtypes identified within the postnatal subset onto the adult mouse sagittal section of the Visium dataset. Download figure Open in new tab Extended Data Fig. 5 Details and validation of the combined computational approach to predict the maturation trajectories. a, UMAP plot of the postnatal subset of Slc1a3 -expressing cells, with cells colored according to their computed proliferation score. This score was used as a starting point to calculate the pseudotime using Monocle3 100 ( b, see Methods). c, UMAP highlighting the cells used as roots for random walk simulations, colored by cluster. d, Validation of the predictive power of the approach using the oligodendrocyte lineage (see Methods). e, Validation of the predictive power of the approach using the interneuron lineage (see Methods). f, Ependymal cells, corresponding to cluster 15, were predicted to originate from cluster 3, both of which consistently exhibit high levels of Rfx3 , a key TR required for ependymal cell differentiation 108 . g-i , UMAP plots showing visitation probability values obtained in the first round of random walks simulated with URD ( g ) or CellRank ( h,i ) from or toward each of the mature astrocyte subtypes defined as tips. Green arrows point to cells in cluster 14 with a high probability of differentiating into WMA_ Pgm5 or GLA_ arch . Red arrows indicate cells in cluster 14 with a high probability of differentiating into CNA_ Gpc5. This is consistent with the results obtained in the additional round of random walks shown in Fig. 3c . j,k, UMAP insets highlighting the cells in cluster 14 (j) or 9 (k) with the highest probability of differentiating into WMA_ Pgm5 , GLA_ arch , CNA_ Gpc5 or BG_ arch , BG_ Mybpc1 , BG_ Npy , respectively. Notably, cells in cluster 14 show minimal overlap, indicating distinct pools of committed precursors for WMA_Pgm5, GLA_ arch , and CNA_Gpc5. Download figure Open in new tab Extended Data Fig. 6 Profiling of perinatal progenitor clusters. a, Expression levels of key genes differentially expressed in cluster 1, cluster 4 or cluster 7 that also show distinct spatial expression patterns as indicated by ISH data (Allen Brain Atlas) or immunofluorescence ( b ). c, Genes enriched in the ACSA2/GLAST+ progenitor pool 16 are also highly expressed in cluster 1. d, Genes differentially expressed by the ACSA2-/GLAST+ progenitor pool 16 show low expression in cluster 1 but high expression in other clusters of the subset, namely cluster 0, 4 and 11, consistent with the multipotency of this progenitor pool along both interneuron and glial lineages. e, Spatial mapping of the perinatal astrocyte progenitor clusters onto a P7 Stereo-seq dataset 37 . Although the cerebellum at P7 likely hosts few remaining cells belonging to these perinatal clusters (mostly composed of cells collected between P0 and P5, Extended data Fig. 4a), the spatial predictions align well with their inferred maturation trajectories. Cluster 0 predominantly maps to the BG layer, cluster 1 to the CN, cluster 4 to the WM and GL, cluster 17 to the WM. The predicted localization of cluster 7 within the CN and outer GL, regions populated by emerging glutamatergic neurons, is consistent with the expression pattern of its DEGs ( b ) and likely reflects their transcriptional similarity with developing excitatory cells. f, Immunofluorescence for GAT3 (red) in hGFAP:GFP mice at P30 highlights its expression in the CN and in few astrocytes of the deep WM ( f’ ) and GL ( f’’ ), further confirming the inferred fate potency of cluster 1. g, Ptf1a expression is restricted to cluster 11 early postnatally (top panel). Genetic fate mapping performed at P10 after Tx administration at P0 in Ptf1a CreERTM × R26R YFP mice (Ptf1a-YFP) shows that Ptf1a-expressing progenitors at P0 give rise mostly to ML interneurons and to few astrocytes localized in the WM. Scale bars: 50 µm. Download figure Open in new tab Extended Data Fig. 7 Spatial and transcriptional evidence for diverse embryonic origins of cerebellar astrocytes. a, UMAP of the dataset obtained by integrating our Slc1a3 -expressing perinatal clusters ( Figure 2D , top panel) and publicly available embryonic and early postnatal cerebellar sc/snRNA-seq datasets 19 , 32 , 33 , with dots colored by cell type/lineage. b, UMAP highlighting, within the integrated dataset, the cells belonging to the Slc1a3 -expressing clusters from our dataset, colored by cluster identity. c, UMAP plots of the integrated dataset, split by study and colored by developmental stage. d, Immunofluorescence for GAT3 (red) at E14.5 or E17.5, showing its initial high expression in the pontine hindbrain at E14.5, followed by broader hindbrain expression and selective cerebellar localization in the CN region by E17.5; note the absence of signal in the vermal cerebellum at E17.5. e, Dot plots showing expression scores of DEGs from spinal cord Olig2-expressing (Ast_sc_Olig2) and Olig2-negative (Ast_sc_Olig2_depleted) astrocyte populations 43 across the adult cerebellar astrocyte subtypes. f, Dot plots displaying the expression scores of DEGs from cortical astrocyte clusters derived from either Olig2-progenitor or S100a1-progenitor lineages 42 across the adult cerebellar astrocyte subtypes. g, UMAP highlighting developing CN neurons within the integrated P0-P4 dataset, annotated as GABAergic (CN_GABA) or glutamatergic (CN_Glut) based on the expression scores of DEG signatures identified in the corresponding adult CN neuron populations ( Fig. 1a ). Download figure Open in new tab Extended Data Fig. 8 Additional characterization of SCENIC-inferred TRs. a, Cluster distribution along maturation trajectories. Lines represent the percentage of cells belonging to distinct clusters across bins (total of 50 bins per trajectory, y-axis maximum = 100%). Each subpanel illustrates the cluster dynamics specific to a single maturation trajectory. Trajectories are presented in the same order as in TRs activity heatmaps. b, Expression of TRs whose activity is shown in Fig. 5c . c, Expression scores for target genes of novel candidate regulators of cerebellar astrocyte maturation. Networks in each panel show the protein-protein interactions inferred using the STRING database ( http://string-db.org ). Relevant biological processes (GO analysis performed on STRING) are highlighted in distinct colors. Genes lacking interactions or involvement in relevant biological processes are excluded. d, Predicted activity (left) and gene expression (right) of Irf1, enriched in WMA trajectories. e, STRING-inferred protein-protein interactions among Ar target genes, indicating enrichment in neurotransmitter uptake. f, Predicted activity (left) and gene expression (right) of Acaa1a, specifically enriched in Npy trajectories. g, Predicted activity (left) and gene expression (right) of Sox1 and Msx2, enriched in BG_ Npy . h, Egfr expression and Fli1 target genes expression score, both high in cluster 1 and 18, at the onset of the oligodendrocyte lineage. Acknowledgments We thank professor Paola Berchialla for constructive help with the analyses. Footnotes ↵ * co-last authors References 1. ↵ Nagai , J. et al. Behaviorally consequential astrocytic regulation of neural circuits . Neuron 109 , 576 – 596 ( 2021 ). OpenUrl CrossRef PubMed 2. ↵ Lee , H.-G. , Wheeler , M. A. & Quintana , F. J . Function and therapeutic value of astrocytes in neurological diseases . Nat. Rev. Drug Discov . 21 , 339 – 358 ( 2022 ). OpenUrl CrossRef PubMed 3. ↵ Tsai , H.-H. et al. Regional astrocyte allocation regulates CNS synaptogenesis and repair . Science 337 , 358 – 62 ( 2012 ). OpenUrl Abstract / FREE Full Text 4. Boisvert , M. M. , Erikson , G. A. , Shokhirev , M. N. & Allen , N. J . The Aging Astrocyte Transcriptome from Multiple Regions of the Mouse Brain . Cell Rep . 22 , 269 – 285 ( 2018 ). OpenUrl CrossRef PubMed 5. Zeisel , A. et al. Molecular Architecture of the Mouse Nervous System . Cell 174 , 999 – 1014 .e22 ( 2018 ). OpenUrl CrossRef PubMed 6. Lozzi , B. , Huang , T.-W. , Sardar , D. , Huang , A. Y.-S. & Deneen , B . Regionally Distinct Astrocytes Display Unique Transcription Factor Profiles in the Adult Brain . Front. Neurosci . Volume 14 , ( 2020 ). 7. ↵ Batiuk , M. Y. et al. Identi fi cation of region-speci fi c astrocyte subtypes at single cell resolution. 1 – 15 ( 2020 ). doi: 10.1038/s41467-019-14198-8 OpenUrl CrossRef PubMed 8. Herrero-Navarro , A. et al. Astrocytes and neurons share brain region-specific transcriptional signatures. 1 – 23 ( 2020 ). doi: 10.1101/2020.04.21.038737 OpenUrl Abstract / FREE Full Text 9. ↵ Endo , F. et al. Molecular basis of astrocyte diversity and morphology across the CNS in health and disease . Science (80-.) . 378 , eadc9020 ( 2022 ). 10. ↵ Lanjakornsiripan , D. et al. Layer-specific morphological and molecular differences in neocortical astrocytes and their dependence on neuronal layers . Nat. Commun . ( 2018 ). doi: 10.1038/s41467-018-03940-3 OpenUrl CrossRef PubMed 11. ↵ Bayraktar , O. A. et al. Astrocyte layers in the mammalian cerebral cortex revealed by a single-cell in situ transcriptomic map . Nat. Neurosci . 1 – 10 ( 2020 ). doi: 10.1038/s41593-020-0602-1 OpenUrl CrossRef PubMed 12. ↵ Karpf , J. et al. Dentate gyrus astrocytes exhibit layer-specific molecular, morphological and physiological features . Nat. Neurosci . 25 , 1626 – 1638 ( 2022 ). OpenUrl CrossRef PubMed 13. ↵ Cerrato , V. & Buffo , A. Gliogenesis . Handb. Cerebellum Cerebellar Disord . 1–17 ( 2019 ). doi: 10.1007/978-3-319-97911-3_108-1 OpenUrl CrossRef 14. ↵ Parmigiani , E. et al. Heterogeneity and Bipotency of Astroglial-Like Cerebellar Progenitors along the Interneuron and Glial Lineages . J. Neurosci . 35 , 7388 – 402 ( 2015 ). OpenUrl Abstract / FREE Full Text 15. ↵ Cerrato , V. et al. Multiple origins and modularity in the spatiotemporal emergence of cerebellar astrocyte heterogeneity . PLoS Biol . 16 , ( 2018 ). 16. ↵ Kantzer , C. G. et al. ACSA-2 and GLAST classify subpopulations of multipotent and glial-restricted cerebellar precursors . J. Neurosci. Res . 99 , 2228 – 2249 ( 2021 ). OpenUrl CrossRef PubMed 17. ↵ Witter , L. & De Zeeuw , C. I . Regional functionality of the cerebellum . Curr. Opin. Neurobiol . 33 , 150 – 155 ( 2015 ). OpenUrl CrossRef PubMed 18. ↵ Kozareva , V. et al. A transcriptomic atlas of mouse cerebellar cortex comprehensively defines cell types . Nature 598 , 214 – 219 ( 2021 ). OpenUrl CrossRef PubMed 19. ↵ Sepp , M. et al. Cellular development and evolution of the mammalian cerebellum . Nature 625 , 788 – 796 ( 2024 ). OpenUrl CrossRef PubMed 20. ↵ Yao , Z. et al. A high-resolution transcriptomic and spatial atlas of cell types in the whole mouse brain . Nature ( 2023 ). doi: 10.1038/s41586-023-06812-z OpenUrl CrossRef PubMed 21. ↵ Kebschul , J. M. et al. Cerebellar nuclei evolved by repeatedly duplicating a conserved cell-type set . Science (80-.) . 370 , ( 2020 ). 22. ↵ Bocchi , R. et al. Astrocyte heterogeneity reveals region-specific astrogenesis in the white matter . Nat. Neurosci . 28 , 457 – 469 ( 2025 ). OpenUrl PubMed 23. ↵ Mecklenburg , N. et al. Growth and differentiation factor 10 (Gdf10) is involved in Bergmann glial cell development under Shh regulation . Glia 62 , 1713 – 23 ( 2014 ). OpenUrl CrossRef PubMed 24. ↵ Farmer , W. T. et al. Neurons diversify astrocytes in the adult brain through sonic hedgehog signaling . Science 351 , 849 – 54 ( 2016 ). OpenUrl Abstract / FREE Full Text 25. ↵ Tatsumi , K. et al. Olig2-Lineage Astrocytes: A Distinct Subtype of Astrocytes That Differs from GFAP Astrocytes . Front. Neuroanat . 12 , 8 ( 2018 ). 26. ↵ Mockenhaupt , K. et al. Yin Yang 1 controls cerebellar astrocyte maturation . Glia 71 , 2437 – 2455 ( 2023 ). OpenUrl CrossRef PubMed 27. ↵ Llorens-Bobadilla , E. et al. Single-Cell Transcriptomics Reveals a Population of Dormant Neural Stem Cells that Become Activated upon Brain Injury . Cell Stem Cell 17 , 329 – 340 ( 2015 ). OpenUrl CrossRef PubMed 28. ↵ Signer , R. A. J. , Magee , J. A. , Salic , A. & Morrison , S. J . Haematopoietic stem cells require a highly regulated protein synthesis rate . Nature 509 , 49 – 54 ( 2014 ). OpenUrl CrossRef PubMed 29. ↵ de Ceglia , R. et al. Specialized astrocytes mediate glutamatergic gliotransmission in the CNS . Nature 622 , 120 – 129 ( 2023 ). OpenUrl CrossRef PubMed 30. ↵ Park , J.-H. , Pak , H.-J. , Riew , T.-R. , Shin , Y.-J. & Lee , M.-Y . Increased expression of Slit2 and its receptors Robo1 and Robo4 in reactive astrocytes of the rat hippocampus after transient forebrain ischemia . Brain Res . 1634, 45 – 56 ( 2016 ). 31. ↵ De Zeeuw , C. I. & Hoogland , T. M . Reappraisal of Bergmann glial cells as modulators of cerebellar circuit function . Front. Cell. Neurosci . 9 , 1 – 8 ( 2015 ). OpenUrl CrossRef PubMed 32. ↵ Carter , R. A. et al. A Single-Cell Transcriptional Atlas of the Developing Murine Cerebellum . Curr. Biol . 28 , 2910 – 2920 .e2 ( 2018 ). OpenUrl CrossRef PubMed 33. ↵ Vladoiu , M. C. et al. Childhood cerebellar tumours mirror conserved fetal transcriptional programs . Nature ( 2019 ). doi: 10.1038/s41586-019-1158-7 OpenUrl CrossRef PubMed 34. ↵ Redmond , S. A. et al. Development of Ependymal and Postnatal Neural Stem Cells and Their Origin from a Common Embryonic Progenitor . Cell Rep . 27 , 429 – 441 .e3 ( 2019 ). OpenUrl CrossRef PubMed 35. ↵ Farrell , J. A. et al. Single-cell reconstruction of developmental trajectories during zebrafish embryogenesis . Science 360 , ( 2018 ). 36. ↵ Lange , M. et al. CellRank for directed single-cell fate mapping . Nat. Methods 19 , 159 – 170 ( 2022 ). OpenUrl CrossRef PubMed 37. ↵ Cheng , M. et al. A Cellular Resolution Spatial Transcriptomic Landscape of the Medial Structures in Postnatal Mouse Brain . Front. Cell Dev. Biol. Volume 10 , ( 2022 ). 38. ↵ Yen , A. et al. MYT1L deficiency impairs excitatory neuron trajectory during cortical development . Nat. Commun . 15 , 10308 ( 2024 ). 39. ↵ Pan , F. C. et al. Spatiotemporal patterns of multipotentiality in Ptf1a-expressing cells during pancreas organogenesis and injury-induced facultative restoration . Development 140 , 751 – 764 ( 2013 ). OpenUrl Abstract / FREE Full Text 40. ↵ Bandler , R. C. et al. Single-cell delineation of lineage and genetic identity in the mouse brain . Nature 601 , 404 – 409 ( 2022 ). OpenUrl CrossRef PubMed 41. ↵ Chen , A. et al. Spatiotemporal transcriptomic atlas of mouse organogenesis using DNA nanoball-patterned arrays . Cell 185 , 1777 – 1792 .e21 ( 2022 ). OpenUrl CrossRef PubMed 42. ↵ Zhou , J. et al. Dual lineage origins of neocortical astrocytes . bioRxiv 2023.09.12.557313 ( 2023 ). doi: 10.1101/2023.09.12.557313 OpenUrl Abstract / FREE Full Text 43. ↵ Ohayon , D. , Aguirrebengoa , M. , Escalas , N. , Jungas , T. & Soula , C . Transcriptome profiling of the Olig2-expressing astrocyte subtype reveals their unique molecular signature . iScience 24 , 102806 ( 2021 ). 44. ↵ Ohayon , D. et al. Sulfatase 2 promotes generation of a spinal cord astrocyte subtype that stands out through the expression of Olig2 . Glia 67 , 1478 – 1495 ( 2019 ). OpenUrl CrossRef PubMed 45. ↵ Altman , J. & Bayer , S. A . Embryonic development of the rat cerebellum. II. Translocation and regional distribution of the deep neurons . J. Comp. Neurol . 231 , 27 – 41 ( 1985 ). OpenUrl CrossRef PubMed Web of Science 46. ↵ Aibar , S. et al. SCENIC: single-cell regulatory network inference and clustering . Nat. Methods 14 , 1083 – 1086 ( 2017 ). OpenUrl CrossRef PubMed 47. ↵ Tiwari , N. et al. Stage-Specific Transcription Factors Drive Astrogliogenesis by Remodeling Gene Regulatory Landscapes . Cell Stem Cell 23 , 557 – 571 .e8 ( 2018 ). OpenUrl CrossRef PubMed 48. Methot , L. et al. Nuclear factor-kappaB regulates multiple steps of gliogenesis in the developing murine cerebral cortex . Glia 66 , 2659 – 2672 ( 2018 ). OpenUrl CrossRef PubMed 49. Lattke , M. et al. Extensive transcriptional and chromatin changes underlie astrocyte maturation in vivo and in culture . Nat. Commun . 12 , 4335 ( 2021 ). 50. ↵ Viana , J. F. et al. Astrocytic Foxo1 regulates hippocampal spinogenesis and synaptic plasticity to enhance fear memory . bioRxiv 2023.05.01.538923 ( 2023 ). doi: 10.1101/2023.05.01.538923 OpenUrl Abstract / FREE Full Text 51. ↵ Leung , A. W. & Li , J. Y. H . The Molecular Pathway Regulating Bergmann Glia and Folia Generation in the Cerebellum . Cerebellum 17 , 42 – 48 ( 2018 ). OpenUrl CrossRef PubMed 52. ↵ Sloan , S. A. & Barres , B. A . Mechanisms of astrocyte development and their contributions to neurodevelopmental disorders . Curr. Opin. Neurobiol . 27 , 75 – 81 ( 2014 ). OpenUrl CrossRef PubMed 53. ↵ Hiraoka , Y. et al. Delta-like 1 regulates Bergmann glial monolayer formation during cerebellar development . Mol. Brain 6 , 25 ( 2013 ). 54. ↵ Asano , H. et al. Astrocyte Differentiation of Neural Precursor Cells is Enhanced by Retinoic Acid Through a Change in Epigenetic Modification . Stem Cells 27 , 2744 – 2752 ( 2009 ). OpenUrl CrossRef PubMed 55. ↵ Cheng , F. Y. , Fleming , J. T. & Chiang , C . Bergmann glial Sonic hedgehog signaling activity is required for proper cerebellar cortical expansion and architecture . Dev. Biol . ( 2018 ). doi: 10.1016/j.ydbio.2018.05.015 OpenUrl CrossRef PubMed 56. ↵ He , L. et al. Transcriptional Regulator Zeb2 is Essential for Bergmann Glia Development . J. Neurosci . 2674–17 ( 2018 ). doi: 10.1523/JNEUROSCI.2674-17.2018 OpenUrl Abstract / FREE Full Text 57. ↵ He , F. et al. A positive autoregulatory loop of Jak-STAT signaling controls the onset of astrogliogenesis . Nat. Neurosci . 8 , 616 – 625 ( 2005 ). OpenUrl CrossRef PubMed Web of Science 58. ↵ Zenke , K. , Muroi , M. & Tanamoto , K . IRF1 supports DNA binding of STAT1 by promoting its phosphorylation . Immunol. \& Cell Biol . 96 , 1095 – 1103 ( 2018 ). OpenUrl 59. ↵ Venere , M. et al. Sox1 marks an activated neural stem/progenitor cell in the hippocampus . Development 139 , 3938 – 3949 ( 2012 ). OpenUrl Abstract / FREE Full Text 60. ↵ Siletti , K. et al. Transcriptomic diversity of cell types across the adult human brain . Science (80-.). 382, eadd7046 ( 2023 ). 61. ↵ Zhang , P. , Ha , T. , Larouche , M. , Swanson , D. & Goldowitz , D . Kruppel-Like Factor 4 Regulates Granule Cell Pax6 Expression and Cell Proliferation in Early Cerebellar Development . PLoS One 10 , e0134390 ( 2015 ). OpenUrl CrossRef PubMed 62. ↵ Li , D. et al. Distribution and functional significance of KLF15 in mouse cerebellum . Mol. Brain 18 , 3 ( 2025 ). 63. ↵ Wilson , S. L. , Kalinovsky , A. , Orvis , G. D. & Joyner , A. L . Spatially Restricted and Developmentally Dynamic Expression of Engrailed Genes in Multiple Cerebellar Cell Types . The Cerebellum 10 , 356 – 372 ( 2011 ). OpenUrl PubMed 64. ↵ Li , J. Y. H. , Lao , Z. & Joyner , A. L . Changing Requirements for Gbx2 in Development of the Cerebellum and Maintenance of the Mid/Hindbrain Organizer . Neuron 36 , 31 – 43 ( 2002 ). OpenUrl CrossRef PubMed Web of Science 65. ↵ Takahashi , M. & Osumi , N . Pax6 regulates specification of ventral neurone subtypes in the hindbrain by establishing progenitor domains . Development 129 , 1327 – 1338 ( 2002 ). OpenUrl PubMed Web of Science 66. ↵ Wassarman , K. M. et al. Specification of the anterior hindbrain and establishment of a normal mid/hindbrain organizer is dependent on Gbx2 gene function . Development 124 , 2923 – 2934 ( 1997 ). OpenUrl Abstract 67. ↵ Zhang , Y. , Hoxha , E. , Zhao , T. , Zhou , X. & Alvarez-Bolado , G . Foxb1 Regulates Negatively the Proliferation of Oligodendrocyte Progenitors . Front. Neuroanat . 11 , ( 2017 ). 68. ↵ Fu , Y. et al. Heterogeneity of glial progenitor cells during the neurogenesis-to-gliogenesis switch in the developing human cerebral cortex . Cell Rep . 34 , 108788 ( 2021 ). 69. ↵ Reeber , S. L. , Arancillo , M. & Sillitoe , R. V . Bergmann Glia are Patterned into Topographic Molecular Zones in the Developing and Adult Mouse Cerebellum . Cerebellum ( 2014 ). doi: 10.1007/s12311-014-0571-6 OpenUrl CrossRef 70. ↵ Apps , R. & Hawkes , R . Cerebellar cortical organization: a one-map hypothesis . Nat. Rev. Neurosci . 10 , 670 – 681 ( 2009 ). OpenUrl CrossRef PubMed Web of Science 71. ↵ Rudolph , R. et al. The inhibitory input to mouse cerebellar Purkinje cells is reciprocally modulated by Bergmann glial P2Y1 and AMPA receptor signaling . Glia 64 , 1265 – 1280 ( 2016 ). OpenUrl PubMed 72. ↵ Barros , L. F. et al. Preferential transport and metabolism of glucose in Bergmann glia over Purkinje cells: a multiphoton study of cerebellar slices . Glia 57 , 962 – 970 ( 2009 ). OpenUrl CrossRef PubMed Web of Science 73. ↵ Lopez , I. A. et al. Evidence for oxidative stress in the developing cerebellum of the rat after chronic mild carbon monoxide exposure (0.0025% in air) . BMC Neurosci . 10 , 53 ( 2009 ). 74. Yamashita , T. et al. Oxidative injury is present in Purkinje cells in patients with olivopontocerebellar atrophy . J. Neurol. Sci . 175 , 107 – 110 ( 2000 ). OpenUrl CrossRef PubMed 75. ↵ Leung , T. C. S. et al. Mitochondrial damage and impaired mitophagy contribute to disease progression in SCA6 . Acta Neuropathol . 147 , 26 ( 2024 ). 76. ↵ Braak , H. , Braak , E. , Bohl , J. & Lang , W . Alzheimer’s disease: amyloid plaques in the cerebellum . J. Neurol. Sci . 93 , 277 – 287 ( 1989 ). OpenUrl CrossRef PubMed Web of Science 77. ↵ Grimaldi , P. , Parras , C. , Guillemot , F. , Rossi , F. & Wassef , M . Origins and control of the differentiation of inhibitory interneurons and glia in the cerebellum . Dev. Biol . 328 , 422 – 433 ( 2009 ). OpenUrl CrossRef PubMed Web of Science 78. ↵ Sudarov , A. et al. Ascl1 genetics reveals insights into cerebellum local circuit assembly . J. Neurosci . 31 , 11055 – 69 ( 2011 ). OpenUrl Abstract / FREE Full Text 79. ↵ Wizeman , J. W. , Guo , Q. , Wilion , E. M. & Li , J. Y . Specification of diverse cell types during early neurogenesis of the mouse cerebellum . Elife 8 , ( 2019 ). 80. ↵ Khouri-Farah , N. , Guo , Q. , Morgan , K. , Shin , J. & Li , J. Y. H . Integrated single-cell transcriptomic and epigenetic study of cell state transition and lineage commitment in embryonic mouse cerebellum . Sci. Adv . 8 , eabl9156 ( 2022 ). 81. ↵ Buffo , A. & Rossi , F . Origin, lineage and function of cerebellar glia . Prog. Neurobiol . 109 , 42 – 63 ( 2013 ). OpenUrl CrossRef PubMed 82. ↵ Konstantinides , N. & Desplan , C . Neuronal differentiation strategies: insights from single-cell sequencing and machine learning . Development 147 , dev193631 ( 2020 ). 83. ↵ Hashimoto , R. et al. Origins of oligodendrocytes in the cerebellum, whose development is controlled by the transcription factor, Sox9 . Mech. Dev . 140 , 25 – 40 ( 2016 ). OpenUrl PubMed 84. ↵ Mecklenburg , N. , Garcia-López , R. , Puelles , E. , Sotelo , C. & Martinez , S . Cerebellar oligodendroglial cells have a mesencephalic origin . Glia 59 , 1946 – 57 ( 2011 ). OpenUrl CrossRef PubMed 85. ↵ Zhang , X. , Santuccione , A. , Leung , C. & Marino , S . Differentiation of postnatal cerebellar glial progenitors is controlled by Bmi1 through BMP pathway inhibition . Glia 59 , 1118 – 31 ( 2011 ). OpenUrl PubMed 86. ↵ De Luca , A. et al. Sonic hedgehog patterning during cerebellar development . Cell. Mol. Life Sci . 73 , 291 – 303 ( 2016 ). OpenUrl CrossRef PubMed 87. Araujo , A. P. B. et al. Effects of Transforming Growth Factor Beta 1 in Cerebellar Development: Role in Synapse Formation . Front. Cell. Neurosci. Volume 10 , ( 2016 ). 88. Koibuchi , N. & Ikeda , Y . Hormones and Cerebellar Development . in Handbook of the Cerebellum and Cerebellar Disorders (eds. Manto , M. , Schmahmann , J. D. , Rossi , F. , Gruol , D. L. & Koibuchi , N. ) 319 – 339 ( Springer Netherlands , 2013 ). doi: 10.1007/978-94-007-1333-8_16 OpenUrl CrossRef 89. ↵ Yamamoto , M. , McCaffery , P. & Dräger , U. C . Influence of the choroid plexus on cerebellar development: analysis of retinoic acid synthesis . Dev. Brain Res . 93 , 182 – 190 ( 1996 ). OpenUrl CrossRef PubMed 90. ↵ Chierzi , S. et al. Astrocytes Transplanted during Early Postnatal Development Integrate, Mature, and Survive Long Term in Mouse Cortex . J. Neurosci . 43 , 1509 – 1529 ( 2023 ). OpenUrl Abstract / FREE Full Text 91. ↵ Nolte , C. et al. GFAP promoter-controlled EGFP-expressing transgenic mice: a tool to visualize astrocytes and astrogliosis in living brain tissue . Glia 33 , 72 – 86 ( 2001 ). OpenUrl CrossRef PubMed Web of Science 92. Srinivas , S. et al. Cre reporter strains produced by targeted insertion of EYFP and ECFP into the ROSA26 locus . BMC Developmental Biology 1 , 4 ( 2001 ). 93. ↵ Wagner , D. E. et al. Single-cell mapping of gene expression landscapes and lineage in the zebrafish embryo . Science 360 , 981 – 987 ( 2018 ). OpenUrl Abstract / FREE Full Text 94. ↵ Fleming , S. J. et al. Unsupervised removal of systematic background noise from droplet-based single-cell experiments using CellBender . Nat. Methods 20 , 1323 – 1335 ( 2023 ). OpenUrl CrossRef PubMed 95. ↵ Young , M. D. & Behjati , S . SoupX removes ambient RNA contamination from droplet-based single-cell RNA sequencing data . Gigascience 9 , giaa151 ( 2020 ). 96. ↵ Hao , Y. et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis . Nat. Biotechnol . 42 , 293 – 304 ( 2024 ). OpenUrl CrossRef PubMed 97. ↵ Biancalani , T. et al. Deep learning and alignment of spatially resolved single-cell transcriptomes with Tangram . Nat. Methods 18 , 1352 – 1362 ( 2021 ). OpenUrl CrossRef PubMed 98. ↵ Wolf , F. A. , Angerer , P. & Theis , F. J . SCANPY: large-scale single-cell gene expression data analysis . Genome Biol . 19 , 15 ( 2018 ). 99. ↵ Brionne , A. , Juanchich , A. & Hennequet-Antier , C . ViSEAGO: a Bioconductor package for clustering biological functions using Gene Ontology and semantic similarity . BioData Min . 12 , 16 ( 2019 ). 100. ↵ Trapnell , C. et al. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells . Nat. Biotechnol . 32 , 381 – 386 ( 2014 ). OpenUrl CrossRef PubMed 101. ↵ Weiler , P. , Lange , M. , Klein , M. , Pe’er , D. & Theis , F . CellRank 2: unified fate mapping in multiview single-cell data . Nat. Methods 21 , 1196 – 1205 ( 2024 ). OpenUrl CrossRef PubMed 102. ↵ Siebert , S. et al. Stem cell differentiation trajectories in Hydra resolved at single-cell resolution . Science (80-.) . 365 , eaav9314 ( 2019 ). OpenUrl Abstract / FREE Full Text 103. ↵ Marques , S. et al. Transcriptional Convergence of Oligodendrocyte Lineage Progenitors during Development . Dev. Cell 46 , 504 – 517 .e7 ( 2018 ). OpenUrl CrossRef PubMed 104. ↵ van Bruggen , D. et al. Developmental landscape of human forebrain at a single-cell level identifies early waves of oligodendrogenesis . Dev. Cell 57 , 1421 – 1436 .e5 ( 2022 ). OpenUrl CrossRef PubMed 105. ↵ Fleming , J. T. et al. The Purkinje Neuron Acts as a Central Regulator of Spatially and Functionally Distinct Cerebellar Precursors . Dev. Cell 27 , 278 – 292 ( 2013 ). OpenUrl CrossRef PubMed Web of Science 106. ↵ Schilling , K . Revisiting the development of cerebellar inhibitory interneurons in the light of single-cell genetic analyses . Histochem. Cell Biol . 161 , 5 – 27 ( 2024 ). OpenUrl PubMed 107. ↵ Van de Sande , B. et al. A scalable SCENIC workflow for single-cell gene regulatory network analysis . Nat. Protoc . 15 , 2247 – 2276 ( 2020 ). OpenUrl CrossRef PubMed 108. ↵ Baas , D. et al. A deficiency in RFX3 causes hydrocephalus associated with abnormal differentiation of ependymal cells . Eur. J. Neurosci . 24 , 1020 – 1030 ( 2006 ). OpenUrl CrossRef PubMed Web of Science View the discussion thread. Back to top Previous Next Posted July 22, 2025. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following A single-cell transcriptomic atlas maps cerebellar astrocyte diversity and uncovers the transcriptional code underlying their maturation trajectories Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share A single-cell transcriptomic atlas maps cerebellar astrocyte diversity and uncovers the transcriptional code underlying their maturation trajectories Valentina Cerrato , Giacomo Turrini , Ilaria Vitali , Bilian Xiong , Laura Solanelles-Farré , Andrea Lopes , Elia Magrinelli , Riccardo Bocchi , Magdalena Götz , Judith Fischer-Sternjak , Enrica Boda , Annalisa Buffo , Ludovic Telley bioRxiv 2025.07.17.665323; doi: https://doi.org/10.1101/2025.07.17.665323 Share This Article: Copy Citation Tools A single-cell transcriptomic atlas maps cerebellar astrocyte diversity and uncovers the transcriptional code underlying their maturation trajectories Valentina Cerrato , Giacomo Turrini , Ilaria Vitali , Bilian Xiong , Laura Solanelles-Farré , Andrea Lopes , Elia Magrinelli , Riccardo Bocchi , Magdalena Götz , Judith Fischer-Sternjak , Enrica Boda , Annalisa Buffo , Ludovic Telley bioRxiv 2025.07.17.665323; doi: https://doi.org/10.1101/2025.07.17.665323 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Neuroscience Subject Areas All Articles Animal Behavior and Cognition (7633) Biochemistry (17681) Bioengineering (13890) Bioinformatics (41930) Biophysics (21446) Cancer Biology (18586) Cell Biology (25493) Clinical Trials (138) Developmental Biology (13374) Ecology (19897) Epidemiology (2067) Evolutionary Biology (24308) Genetics (15607) Genomics (22498) Immunology (17736) Microbiology (40385) Molecular Biology (17175) Neuroscience (88584) Paleontology (666) Pathology (2831) Pharmacology and Toxicology (4823) Physiology (7641) Plant Biology (15149) Scientific Communication and Education (2045) Synthetic Biology (4293) Systems Biology (9823) Zoology (2271)