Reorganization of the heterochromatin-associated gene-dense subcompartment in early neuronal development Open Access

The 3D organization of the genome is dynamically regulated alongside cellular development. This reorganization occurs across different genomic scales, spanning up to entire chromosomes (Fujita et al., 2022). Locally, gene promoters interact with regulatory enhancer elements to directly activate transcription. Beyond these short-range interactions, genomic regions organize at the chromosomal level into euchromatic and heterochromatic compartments, which can be further separated into smaller subcompartments (Rao et al., 2014). These subcompartments occupy distinct subnuclear locales and may be associated with nuclear bodies, often across different chromosomes (Quinodoz et al., 2018). The interaction of genomic loci with nuclear bodies, such as nuclear speckles, may enhance transcriptional or co-transcriptional processes, while genomic interaction with the nuclear lamina maintains gene repression in cells (Bhat et al., 2024; Reddy et al., 2008; Sabari et al., 2018). Thus, regulation of long-range genome organization provides additional mechanisms for controlling gene expression during development.

Post-mitotic neurons located throughout the mammalian brain have been found to undergo a set of seemingly common genome architectural changes, such as the radial repositioning of specific chromosomes. During the differentiation of olfactory sensory neurons, genomic regions containing olfactory receptor gene clusters may be relocated to a large heterochromatin cluster in the nuclear interior (Clowney et al., 2012). This is thought to ensure the constitutive repression of most olfactory receptor genes, while a single olfactory receptor gene may be chosen to escape the heterochromatin cluster and be expressed (Monahan et al., 2019). Similarly, large parts of chromosomes 7, 17, and 19 are repositioned toward the nuclear interior in cortical and hippocampal neurons during development (Takei et al., 2021; Tan et al., 2021; Winick-Ng et al., 2021). These regions are enriched for clustered gene families, such as vomeronasal receptors and immunoglobulins, which are constitutively repressed in mature neurons in the forebrain (Tan et al., 2021). These observations have been also extended to cerebellar granule neurons, the most abundant neuronal cell type in the brain (Tan et al., 2023; Zhao et al., 2023 preprint).

In the mouse cerebellum, we recently found that dense stretches of transcriptionally active genes, which are linearly adjacent to heterochromatin-silenced clustered gene families, robustly interacted with each other over long genomic distances to form a distinct genomic subcompartment in granule neurons (Zhao et al., 2023 preprint). This gene-dense subcompartment occupied a distinct microenvironment within the nucleus that is spatially separated from nuclear speckles and their associated genomic loci (Zhao et al., 2023 preprint). Remarkably, regions in the gene-dense subcompartment further strengthened their interactions in granule neurons during cerebellar development. However, we currently lack a detailed understanding of the specific stages of neuronal development during which these nuclear changes occur. Moreover, it remains unknown whether the transcriptionally active gene-dense subcompartment and linearly adjacent heterochromatin regions might be coordinately restructured during development.

Granule neurons offer a powerful system to study the mechanisms that regulate well-defined stages in neuronal development (Alvarez-Saavedra et al., 2014; Frank et al., 2015; Konishi et al., 2004; Okazawa et al., 2009; Yang et al., 2016). Granule neuron precursors expressing the excitatory neuron lineage marker Pax6 and mitotic cell marker Ki67 proliferate in the external granule layer (EGL) of the cerebellar cortex throughout the first to third postnatal weeks (Engelkamp et al., 1999; Shiraishi et al., 2019). During this period, waves of post-mitotic granule neurons are generated and express the microtubule-associated protein doublecortin (DCX), which is required for their migration through the EGL, molecular layer (ML), and Purkinje cell layer (PCL), before reaching their final location within the internal granule layer (IGL) (Consalez et al., 2021; Gleeson et al., 1999; Komuro and Rakic, 1998). There, granule neurons downregulate DCX expression as they integrate in cerebellar circuits by establishing connections with presynaptic and postsynaptic partners (Gleeson et al., 1999; Sepp et al., 2024; Weyer and Schilling, 2003). Thus, each of these developmental stages can be distinguished by the localization of granule neurons within cerebellar cortical layers and the use of validated protein markers, which together can be employed to elucidate developmental changes in genome organization.

In this study, we combined single-cell imaging with high-throughput genome profiling to identify the developmental periods for genome architectural changes in mouse cerebellar granule neurons. We focused our analyses on our recently identified gene-dense regions, which we collectively name the heterochromatin-associated gene-dense (hGD) subcompartment. We observed that the hGD subcompartment undergoes significant changes in chromosomal organization in post-mitotic neurons. Utilizing DNA fluorescence in situ hybridization combined with immunohistochemistry, we found that the radial repositioning and the strengthening of long-distance interactions between hGD regions occurred alongside the downregulation of DCX expression in granule neurons as they completed their migration within the IGL. We also discovered, using fluorescence-activated nuclei sorting (FANS) followed by Hi-C analyses, that the entire hGD subcompartment underwent chromosomal reorganization during early granule neuron development. The timing of these major changes in the hGD subcompartment coincided with the nuclear repositioning of heterochromatic chromocenters tethered to hGD regions, though these adjacent structures maintained distinct spatial compartments throughout development. Unexpectedly, we also found that the strengthening of hGD subcompartment interactions was independent of developmental changes in gene expression within these regions. Together, our results reveal the developmental window for the structural rearrangements of chromosomes that are decoupled from transcriptional regulation in granule neurons.

We recently identified a gene-dense subcompartment that is distinct from other transcriptionally active nuclear subcompartments in the cerebellum (Zhao et al., 2023 preprint). This subcompartment is composed of several hundred genomic loci enriched on chromosomes 7 and 17, which preferentially associate within the nuclear interior (Zhao et al., 2023 preprint). To understand how this subcompartment is changed during development, we first analyzed 3D models of developing and adult cerebellar granule neurons from published single-cell diploid chromatin conformation capture (Dip-C) datasets (Fig. 1A) (Tan et al., 2023). Among chromosomes, gene-dense regions within chromosome 7 exhibited the most prominent inward radial movements as granule neurons differentiated (Fig. 1B). When grouping granule neurons by their Dip-C structural stage (S1-S5), inferred through unsupervised clustering (Tan et al., 2023), gene-dense subcompartment regions within chromosome 7 showed the most pronounced repositioning toward the nuclear interior during structural stages S1 to S3 (Fig. 1C).

Given the significant reorganization of the gene-dense subcompartment within chromosome 7 during granule neuron development, we examined the chromatin landscape surrounding these genomic regions to identify factors associated with the observed structural changes. As expected, gene-dense subcompartment regions were marked by histone H3 lysine 4 trimethylation (H3K4me3) and showed robust gene expression, suggesting that they are transcriptionally active (Fig. 1D). These regions were also found adjacent to broad heterochromatin domains marked by histone H3 lysine 9 trimethylation (H3K9me3) (Fig. 1D). We therefore reclassified these transcriptionally active genomic loci as hGD regions, based on their surrounding repressive chromatin environment. These hGD regions may be distinguished from nuclear speckle-associated genomic regions, which are also gene-dense but do not neighbor heterochromatin regions (Chen et al., 2018). Additionally, we observed that heterochromatin regions on chromosome 7 and 17 exhibited increased H3K9me3 levels during cerebellar development (Fig. 1E,F), a feature associated with the strengthening of heterochromatin compartmentalization in post-mitotic neurons (Bashkirova et al., 2023). Together, these results suggest that the nuclear reorganization of hGD regions may be associated with developmental changes in their adjacent heterochromatin domains in the cerebellum.

The predictions from Dip-C 3D models that hGD regions change their nuclear localization in developing granule neurons led us to visualize these regions using 3D DNA fluorescence in situ hybridization (DNA FISH) in the cerebellum of postnatal mice. We generated probes targeting two hGD regions (hGD1 and hGD2) located 5.6Mb apart on chromosome 7 (Fig. 1D). We examined several cerebellar ages during development, including postnatal day 6 (P6), when the cerebellum is primarily composed of both granule neuron precursors in the external granule layer (EGL) and immature post-mitotic granule neurons in the internal granule layer (IGL), as well as postnatal day 22 (P22), when the EGL has disappeared and mature granule neurons in the IGL have integrated into cerebellar cortical circuits (Fig. 2A) (Yamada et al., 2014). When comparing granule neurons in the IGL at these developmental ages, we observed a significant shift of the hGD1 and hGD2 regions toward the nuclear interior during granule neuron differentiation (Fig. 2B-D).

Next, we asked whether these changes in the radial positioning of hGD regions were associated with other alterations in nuclear organization. Using Hi-C analyses, we have previously found that hGD regions showed increases in interaction frequency with cerebellar development (Zhao et al., 2023 preprint). Therefore, we reasoned that the re-localization of these regions closer to the nuclear interior might also enhance their spatial association in granule neurons during development. To visualize these changes, we examined the 3D spatial distance between the hGD1 and hGD2 regions using DNA FISH analyses of granule neurons in the IGL. We found that the 3D distance between the two hGD regions decreased with cerebellar age during development, with significant changes beginning at P14 and continuing through to P22 (Fig. 2E). Together, these results suggest that the inward radial positioning of hGD regions is accompanied by an increase in their spatial proximity during granule neuron differentiation.

The observed changes in both radial positioning and spatial proximity of hGD regions over development led us to determine whether these rearrangements occurred at specific stages of granule neuron differentiation. Given the heterogeneity of granule neurons in the early postnatal cerebellum, we used markers to identify granule neuron subpopulations at distinct developmental stages. Single nuclei (sn)RNA-seq analyses of cerebellar granule neurons in P11 mice (Rosenberg et al., 2018) revealed genes expressed at defined developmental stages, such as Dcx, which encodes a microtubule-associated protein enriched in immature post-mitotic granule neurons undergoing migration, and Nebl, which encodes an actin-binding protein transiently expressed at a later developmental stage in post-migratory granule neurons (Fig. 3A) (Gleeson et al., 1999; Myers et al., 2020). Notably, in the cerebellum of P14 mice, newly-born post-mitotic granule neurons undergoing cell migration from the EGL toward the IGL possessed a distinctive ring of DCX protein fluorescence around their nuclei (Fig. 3B,C). This DCX ring was then lost, likely as neurons completed their migration and settle in their final position in the IGL (Fig. 3B,C) (Gleeson et al., 1999; Huynh et al., 2011; Zimatkin and Karnyushko, 2017).

To examine how the nuclear positioning of hGD regions was changed at various granule neuron developmental stages, we performed DCX immunolabeling combined with DNA FISH analyses. We observed no significant differences in the radial positioning of hGD1 and hGD2 regions in DCX-expressing granule neurons that are likely migrating in the cerebellar cortex, including in post-mitotic neurons in the inner EGL and a subset of neurons within the IGL with high DCX expression (IGL DCX^high) (Fig. 3D, Fig. S1). Strikingly, however, we found a significant shift of both hGD1 and hGD2 regions toward the nuclear interior in IGL granule neurons with little or no DCX expression (IGL DCX^low) (Fig. 3D, Fig. S1). In conjunction with these changes, we also observed a reduction in the 3D distance between the hGD1 and hGD2 regions in post-migratory granule neurons with low DCX expression, compared to earlier stages in migrating granule neurons with high DCX expression (Fig. 3E). This genome reorganization occurred without changes in nuclear area between DCX^high and DCX^low granule neurons, although changes were observed with cerebellar age (Fig. S2). Together, these results suggest that changes in the nuclear organization of hGD regions, including their radial repositioning and spatial proximity, occur as granule neurons switch from a phase of cell migration to later stages when they integrate in cerebellar circuits.

We next asked whether these genome architectural changes might be linked to changes in association with nuclear bodies such as chromocenters, nucleoli, or nuclear speckles, which have roles in organizing chromosomal hubs in cells (Quinodoz et al., 2018). Since hGD regions are linearly adjacent to large heterochromatin domains on the genome, we reasoned that they might also be physically positioned nearby heterochromatic chromocenters in the nucleus_­. We first examined the localization of chromocenters, which may be visualized using the Hoechst DNA dye, in granule neurons. We found that initially in P6 mice, chromocenters were positioned nearby the nuclear lamina at the nuclear periphery in immature granule neurons (Fig. 4A,B). However, in mature granule neurons in P22 mice, heterochromatic chromocenters shifted modestly away from the periphery, possibly as a result of heterochromatin regions undergoing radial repositioning (Fig. 4A,B). Indeed, we observed that H3K9me3-marked heterochromatin regions (H1 and H2) adjacent to hGD1/2 regions within chromosome 7 were robustly repositioned toward the nuclear interior during neuronal differentiation (Fig. S3A-C). Moreover, similar to hGD regions, H1 and H2 also exhibited a significant decrease in 3D distance over cerebellar development (Fig. S3D). To determine the spatial relationship between transcriptionally active hGD regions and heterochromatic chromocenters, we next measured the enrichment of chromocenters within a 350 nm radius circle surrounding the hGD probes. We found that throughout granule neuron development, there was a weaker overlap between chromocenters and the hGD1 and hGD2 regions compared to chromocenters and the hGD-adjacent heterochromatin regions H1 and H2 (Fig. 4C, Fig. S3E). Together, these data suggest that hGD regions and their physically tethered heterochromatic chromocenters are both repositioned away from the nuclear periphery during granule neuron differentiation, but these structures retained 3D spatial separation throughout development.

In addition to the coordinated radial repositioning of chromocenters and hGD regions, we investigated whether hGD regions might be associated with other nuclear bodies. Nucleolin-marked nucleoli were found at the periphery of chromocenters (Fig. 4D), where they may establish a nucleolar-associated inter-chromosomal hub (Quinodoz et al., 2018). As observed for hGD regions, nucleoli were positioned at an intermediate radial location between the nuclear periphery and nuclear interior (Fig. 4D,E). However, unlike hGD regions, nucleoli positioning remained largely unchanged throughout cerebellar development (Fig. 4E). We also found that the association between hGD regions and nucleoli was substantially lower compared to the association of hGD regions with chromocenters during development (Fig. 4C,F). These findings suggest that changes in the organization of hGD regions with granule neuron differentiation is largely independent of nucleolar association in granule neurons.

Besides heterochromatin-associated structures, we also characterized nuclear speckles, which are hubs for splicing factors that boost the mRNA processing of proximal gene loci (Bhat et al., 2024; Sung et al., 2023). Like hGD regions, nuclear speckle-associated genomic regions contain dense stretches of genes and are sites of high levels of mRNA transcription (Quinodoz et al., 2018). We observed that nuclear speckles, marked by the splicing factor SF3a66, were primarily located in the nuclear interior throughout granule neuron development and were well separated from heterochromatic chromocenters (Fig. 4G,H, Fig. S4A). Similar observations on nuclear positioning were made for a genomic region (NS1) that is robustly associated with nuclear speckles in a conserved manner from embryonic stem cells to granule neurons (Fig. S4B,C) (Quinodoz et al., 2018; Zhao et al., 2023 preprint). When examining the association of hGD regions with nuclear speckles, we observed a modest enrichment of the nuclear speckle marker SF3a66 around the hGD2 region compared to the hGD1 region in immature granule neurons at P6 (Fig. 4I). With development, the hGD1 region increased its association with nuclear speckles, matching that of the hGD2 region in mature granule neurons at P22, although both hGD regions remained weaker than that of the NS1 genomic region (Fig. 4I, Fig. S4B). Consistent with these findings, the 3D distance between hGD1 and the nearest nuclear speckle border significantly decreased between P6 and P22, while the NS1 locus maintained a closer proximity to nuclear speckles even at the later stage (Fig. 4I, Fig. S4E). Together, our findings suggest that during granule neuron differentiation, hGD regions move in concert with their tethered chromocenters from the nuclear periphery toward the interior, where nuclear speckles reside. However, throughout development, these hGD regions maintain partial spatial separation from other nuclear bodies over a distance of a few hundred nanometers, which may reflect their distinct spatial compartmentalization.

Given our observation that individual hGD regions displayed significant nuclear changes during granule neuron development using microscopy, we next determined whether the 3D organization of the entire hGD subcompartment, including their genomic interactions across chromosomes, might be similarly altered. We therefore sought to take advantage of biochemical approaches such as Hi-C, which can map chromosomal organization with genome-wide resolution, to study granule neurons at discrete stages of their development. We first identified suitable protein markers for isolating the nuclei from subpopulations of developing granule neurons for downstream biochemical analyses. Well established nuclear markers include Pax6, which is expressed selectively in the excitatory granule neuron lineage, and Ki67, which labels actively proliferating cells (Engelkamp et al., 1999; Weisheit et al., 2006). In immunohistochemical analyses of the cerebellar cortex from P6 mice, we observed granule neuron precursors expressing high levels of Ki67 and Pax6 throughout the outer EGL (Fig. 5A). By contrast, immature post-mitotic granule neurons lacking Ki67 expression but expressing moderate levels of Pax6 were found predominantly in the inner EGL and the nascent IGL (Fig. 5A). We next used FANS to isolate the nuclei of granule neuron precursors and immature post-mitotic granule neurons by using the Ki67 and Pax6 markers in P6 mice (Fig. 5B). In RNA-seq analyses of the Ki67^high/Pax6^high and Ki67^low/Pax6^mid FANS-isolated populations, we identified 6759 differentially expressed genes, including Mki67 mRNA, which is enriched in granule neuron precursors, and Nebl mRNA, which is enriched in immature post-mitotic granule neurons (Figs 3A and 5C). The top genes enriched in the two populations can be visualized on a snRNA-seq UMAP projection of developing granule neurons, which includes both granule neuron precursors and immature post-mitotic granule neurons (Fig. 5D) (Rosenberg et al., 2018).

Besides enriching for granule neuron populations at earlier developmental stages, we also sought to isolate granule neurons at later stages, as these neurons integrate into cerebellar circuits. In immunohistochemical analyses of the cerebellar cortex of P10 mice, we visualized granule neurons using the protein markers Pax6, which is partially downregulated as neurons mature, and the splicing factor NeuN, encoded by the Rbfox3 gene, which is upregulated throughout neuron differentiation (Fig. 5E). Importantly, the relative abundance of the Pax6 and NeuN markers distinguished at least two major granule neuron populations in the IGL, as observed using FANS and immunohistochemical analyses of P10-P11 mice (Fig. 5E,F, Fig. S5A,B). The Pax6^mid and Pax6^mid-low granule neuron populations exhibited low DCX protein levels (Fig. S5C,D), suggesting that these neurons are post-migratory. RNA-seq analyses of the NeuN^mid/Pax6^mid and NeuN^high/Pax6^mid-low populations revealed distinct transcriptional signatures, which mapped onto immature and mature post-mitotic granule neuron populations, respectively (Fig. 5G,H).

Having established an approach to profile granule neurons at different developmental stages, we employed high-throughput chromosome conformation capture (Hi-C) on FANS-isolated granule neuron populations to determine how the hGD subcompartment was reorganized during development (Fig. 6A). The Hi-C approach measures spatially proximal DNA-DNA interactions within a few hundred nanometers, making it suitable for characterizing the organization of genomic subcompartments, which may include interactions across chromosomes (Dekker et al., 2013; Finn et al., 2019). We observed a significant increase in inter-chromosomal interactions between regions within the hGD subcompartment when comparing Ki67^high/Pax6^high granule neuron precursors with Ki67^low/Pax6^mid immature post-mitotic granule neurons in the cerebellum of P6 mice (Fig. 6B, left). However, we observed little or no changes in inter-chromosomal interactions between hGD regions at later developmental stages when comparing NeuN^mid/Pax6^mid immature granule neurons with NeuN^high/Pax6^mid-low mature granule neurons in P11 mice. In contrast to the early developmental changes observed in the hGD subcompartment, we found that nuclear speckle-associated genomic loci (Quinodoz et al., 2018) maintained robust inter-chromosomal interactions throughout granule neuron development (Fig. 6B, right). These observations complement our DNA FISH analyses, which revealed that radial repositioning of a hGD region occurred as granule neurons became post-mitotic, and specifically after neuronal migration (Fig. S1B-E). These findings suggest that the strengthening of hGD subcompartment interactions and the nuclear relocalization of these regions occurred selectively during early post-mitotic granule neuron development.

Our imaging analyses had also revealed that hGD regions were physically tethered to heterochromatic chromocenters and exhibited modest association with transcriptionally active nuclear speckles (Fig. 4). We therefore next investigated how inter-chromosomal interactions between the hGD subcompartment and various other subcompartments might be developmentally changed. We observed that transcriptionally active hGD regions formed somewhat stronger inter-chromosomal interactions with nuclear speckle-associated loci compared to heterochromatic B compartment loci or nucleolar-associated loci and these differences persisted throughout development (Fig. 6C,D). Nevertheless, genomic interactions between hGD and nuclear speckle-associated loci were still lower compared to interactions between regions within each of these subcompartments (Fig. 6B,D), indicating that hGD regions and nuclear speckles remained as distinct transcriptionally active subcompartments. In addition, we found that inter-chromosomal interactions between hGD regions and genomic loci organized by nuclear bodies exhibited little overall changes during granule neuron development (Fig. 6D). These findings suggest that although microscopy revealed the coordinated repositioning of hGD regions and heterochromatic chromocenters toward the nuclear interior, hGD regions maintain a distinct genomic subcompartment that remains spatially separated from other nuclear structures throughout cerebellar granule neuron development.

Lastly, we investigated the relationship between the developmental reorganization of the hGD subcompartment and gene transcription. We found that genes residing within hGD regions and nuclear speckle-associated regions showed modest or no net changes in mRNA levels across granule neuron development (Fig. 6E). Moreover, changes in mRNA levels within hGD regions showed little or no correlation with changes in their inter-chromosomal subcompartment interactions across various stages of granule neuron development (Fig. 6F). These results suggest that the reorganization of the hGD subcompartment in developing granule neurons is not associated with gene transcription.

Altogether, we define the early developmental window in granule neurons during which hGD genomic regions are radially repositioned in the nucleus and their compartmentalization is strengthened, with these structural changes occurring independently of transcriptional changes in this subcompartment (Fig. 6G).

Post-mitotic neurons undergo significant changes in the structural organization of their chromosomes in the developing brain. Conserved themes have emerged, where large genomic domains marked by heterochromatin show extensive repositioning away from the nuclear periphery during neuronal differentiation (Tan et al., 2021, 2023; Winick-Ng et al., 2021). In cerebellar granule neurons, we find that transcriptionally active gene-dense regions, located linearly adjacent to these heterochromatin domains, show coordinated radial repositioning together with heterochromatic chromocenters during development. These changes are associated with the overall strengthening of the active hGD subcompartment in developing granule neurons.

Interestingly, our results reveal that the structural reorganization of hGD regions occurred in immature granule neurons after they completed migration in the internal granule layer, as indicated by the downregulation of the microtubule-associated protein DCX. Prior to this, translocating nuclei are under frequent mechanical stress as they are pulled by cytoskeletal microtubules, thus requiring specific mechanisms to maintain nuclear structure during neuronal migration (Schreiner et al., 2015; Stephens et al., 2017; Tariq et al., 2017; Wu et al., 2018). Lamin and lamin receptor proteins are known to play roles in tethering heterochromatin to the nuclear lamina and in regulating nuclear stiffness during processes such as cell migration (Chen et al., 2019; Jung et al., 2013; Solovei et al., 2013). As post-mitotic neurons differentiate, lamin proteins may be downregulated, resulting in the detachment of heterochromatin regions from the nuclear lamina (Chang et al., 2022; Clowney et al., 2012; Harr et al., 2015; Solovei et al., 2013). Our finding that hGD regions are radially repositioned after the conclusion of granule neuron migration is consistent with the period when interactions between heterochromatic chromocenters and the nuclear lamina are weakened.

As hGD regions move away from the nuclear periphery in post-migratory granule neurons, they appear to settle in the nuclear interior, which is enriched for nuclear speckles. hGD regions share similar genomic features with nuclear speckle-associated regions, including a dense concentration of genes and enrichment for RNA polymerase II (Quinodoz et al., 2018; Zhao et al., 2023 preprint). However, the key difference is that hGD regions are linearly adjacent to large heterochromatin domains and are physically tethered to heterochromatic chromocenters in the nucleus. This likely limits the ability of transcriptionally active hGD regions to fully associate with nuclear speckles within a shared spatial compartment. Indeed, hGD regions were observed to maintain a partially separated compartment from other nuclear bodies, including nuclear speckles. Interestingly, we previously observed that the hGD subcompartment and nuclear speckles form higher-order genomic interactions using SPRITE, which captures interactions in the nuclear interior of granule neurons at spatial distances of up to a few micrometers (Quinodoz et al., 2018; Zhao et al., 2023 preprint). Because gene proximity to nuclear speckles is correlated with increased splicing of mRNA transcripts (Bhat et al., 2024), the higher-order genomic interactions between the hGD compartment and nuclear speckles may provide hGD regions with greater access to splicing factors concentrated within nuclear speckles. While our RNA-seq analyses failed to reveal overall changes in the transcription of genes within the hGD subcompartment during granule neuron development, it will be important in future studies to assess how other nuclear processes, including the splicing and nuclear export of transcripts, are regulated by structural changes in the hGD subcompartment.

In conclusion, our study reveals the specific stages of neuronal development during which chromosomes are reorganized. An active question remains as to what molecular mechanisms regulate these genome architectural changes. Besides the downregulation of nuclear lamina proteins, other genome architectural proteins, such as topoisomerases, may play a role by helping to resolve the entanglement of genomic DNA as chromosomes undergo significant reorganization, particularly following cell cycle exit (Hildebrand et al., 2024). Indeed, Top2b is abundantly expressed in post-mitotic neurons and plays key roles in their differentiation and function (Li et al., 2014; Segev et al., 2024; Tiwari et al., 2012). In addition to genome architectural proteins, chromatin remodeling enzymes such as Chd4 and Chd7 also have prominent functions in genome organization during granule neuron development (Goodman et al., 2020; Reddy et al., 2008). Elucidating how the modification of nucleosomes by these major chromatin enzymes influences chromosomal organization will provide valuable insights into how the genome is coordinately regulated across multiple length scales to drive granule neuron development.

CD1 or C57BL/6 mice were purchased from Charles River Laboratories. Both male and female mice were used. All animal experiments were done according to protocols approved by the Animal Studies Committee of Northwestern University in accordance with the National Institutes of Health guidelines.

Antibodies to DCX (Abcam, ab18723, lot# GR3441268-1), Ki67 (Abcam, ab15580, lot# GR3285096-1), NeuN (Abcam, ab190565, lot# 1001571-36 or ab279296), Nucleolin (Abcam, ab22758, lot# 1033948-1), Pax6 (Invitrogen, MA1-109), and SF3a66 (Abcam, ab77800) for immunohistochemistry or FANS experiments were purchased.

Statistical analyses were performed using R. Normality of data was assessed using the Shapiro–Wilk test, and homogeneity of variances was assessed using Bartlett's test. For experiments in two groups with non-normal distributions, the Mann–Whitney-Wilcoxon rank sum test (unpaired data) or the two-sample Brown-Mood median test (heterogeneously distributed data) was used. Comparisons of all samples across multiple groups were performed using the Kruskal–Wallis rank sum test followed by Dunn's post hoc testing with P-values adjusted using the Benjamini-Hochberg procedure for unpaired non-normal distributions.

DNA FISH assays were performed as described with modifications (Bolland et al., 2013; Yamada et al., 2019). The cerebellum of CD1 mice was fixed with 4% PFA and 4% sucrose by transcardiac perfusion. Sagittal cryosections (12 µm thickness) from 6-, 10-, 14-, and 22-day-old mice were prepared using a cryostat (Epredia) and cerebellar sections containing the vermis area were used for analyses. Sections were treated with sodium citrate solution (10 mM sodium citrate, pH 6.0) for 10 min at 70°C and cooled down to room temperature. Sections were dehydrated with sequential treatment of ice-cold 70% EtOH for 2 min, ice-cold 90% EtOH for 2 min, and ice-cold 100% EtOH for 2 min, and air-dried for 5 min. For hybridization, genomic DNA was denatured by incubating glass plates with buffer containing 2× SSC and 70% formamide for 2.5 min at 73°C, followed by incubation with 2× SSC and 50% formamide for 30 s.

All probes were made from BAC clones (BACPAC Resources Center) including RP24-97O16 (hGD1 locus), RP24-153C13 (hGD2 locus), RP24-97C4 (H1 locus), RP24-372C9 (H2 locus), and RP24-358P20 (NS1 locus) by nick translation with chemical coupling using an Alexa Fluor succinimidyl ester (Alexa 647 for hGD1 and NS1; Alexa 555 and Alexa 488 for hGD2, H1, and H2) (Thermo Fisher Scientific). The hybridization solution contained ∼70-80 ng of each labeled probe, 6 µg of mouse Cot-1 DNA (Thermo Fisher Scientific), and 10 µg of sheared salmon-sperm DNA (Thermo Fisher Scientific) in hybridization buffer (10% dextran sulfate, 50% formamide, 2xSSC). The probe mixtures were denatured at 73C for 5 min before use. Denatured sections and probes were sealed with coverslips and incubated at 37°C overnight. On the next day, coverslips were removed and washed once in 2xSSC and 50% formamide solution for 15 min at 45°C, two times in 2× SSC for 5 min at 45°C, and one time in 2xSSC for 5 min at room temperature with gentle agitation. Sections were washed once with 1xPBS and stained with the Hoechst 33342 DNA dye (Sigma, B2261). After washing with PBS, sections were mounted with Fluoromount-G mounting medium (Southern Biotech). Images were acquired on a confocal laser scanning microscope (Leica Microsystems, TCS SP8 confocal) using a 63× objective lens (NA 1.40). A stack of 0.3 µm thick optical sections was acquired for each field of view in the UV, red, and far-red channels. Two DNA FISH biological replicates were performed in all experiments.

Cerebellar sections were prepared and labeled with the relevant antibodies as previously described (Yamada et al., 2014). Secondary antibodies included anti-mouse or anti-rabbit Alexa 488 (Thermo Fisher Scientific, A11029 or A11034) or anti-mouse Alexa 647 (Thermo Fisher Scientific, A21235).

Cerebellar sections from 6-, 14-, and 22-day-old CD1 mice were prepared and treated with sodium citrate solution as described above. Sections were then labeled with the relevant primary and secondary antibodies. After extensive washing of secondary antibodies using PBS, sections were fixed again with 4% PFA in PBS for 10 min at room temperature, washed with PBS for 5 min twice, then 2xSSC twice for 5 min, and treated with ice-cold HCl solution (0.1 M HCl, 0.7% Triton-X) for 25 min at room temperature. Subsequently, DNA FISH procedures including prehybridization, hybridization, washing, and imaging were performed as described above. Two biological replicates were performed in all experiments.

For each cell, DNA FISH signals were subjected to automated identification using custom macros in ImageJ, as previously described (Xu et al., 2022; Yamada et al., 2019). The centroid of DNA FISH signals was detected using a 3D gaussian filter to reduce noise followed by 3D spot segmentation. For co-localization between pairs of DNA FISH probes, the Euclidian distance to the nearest alternate DNA FISH probe was calculated using the 3D Image Suite in ImageJ (Ollion et al., 2013).

Radial positioning of DNA FISH or immunofluorescence signals was analyzed using a single z-section in the middle third (z-axis) of a granule neuron nucleus to ensure a maximal x-y area for calculations. For DNA FISH analyses, the z-section containing the centroid of the DNA FISH probe signal was used. Loci failing to meet these criteria were excluded from analysis. For immunohistochemical analyses, the top 1% or top 10% pixels by intensity for each neuron were identified. The radial distances from the centroid of the nucleus to the DNA FISH probe signal or immunofluorescence signals were then calculated, where 0 represents the nuclear center and 1 represents the nuclear membrane.

The fluorescence intensity level surrounding a genomic locus was analyzed from the z-section containing the centroid of the DNA FISH probe signal using custom ImageJ macros. The average pixel intensity of the immunofluorescence signal in a 350 nm radius circle in the x-y plane surrounding the centroid was normalized to the bottom 20th and top 90th percentile of signal in the nucleus.

The borders of nuclear speckles were extracted as 3D objects using the top 2% pixels by intensity for each z-slice. The distance between the centroid of the DNA FISH probe signal and nearest nuclear speckle border was then quantified using the 3D Image Suite in ImageJ.

The cerebellum from P6 or P11 C57BL/6 mouse pups was fixed with crosslinking solution (1% PFA in PBS) by homogenizing with a glass dounce homogenizer and incubated for 10 min at room temperature. The crosslinking reaction was quenched by 125 mM glycine solution and incubated for 5 min. The cell pellet was washed twice with BSA/PBS solution (0.3% BSA, 0.1% Triton-X in PBS) and stored at −80°C. To isolate nuclei, the frozen pellet was thawed on ice, homogenized with lysis buffer (10 mM Tris-HCl pH 8, 10 mM NaCl, 0.2% Triton-X) and incubated on ice for 10 min. The lysate was pelleted by centrifugation at 700 g for 5 min at 4°C and washed with lysis buffer once. The pellet was resuspended in BSA/PBS solution and filtered using a 50 µm filter (CellTrics). Filtered nuclei were incubated with various combinations of antibodies including anti-Pax6 (1/100), anti-Ki67 (1/100), or anti-NeuN (1/500) for 1 h at 4°C with rotation. Nuclei were washed with BSA/PBS solution twice and stained with secondary antibodies including anti-mouse Alexa 488 (1/250, Thermo Fisher Scientific, A11029) and anti-rabbit Alexa 647 (1/250, Abcam, ab150075) for 30 min at 4°C with rotation, followed by washing with BSA/PBS solution twice. Stained nuclei were resuspended in BSA/PBS solution and filtered using a 50 µm filter prior to FANS. The target populations were enriched with a SH800S cell sorter (Sony) using a 100 µm nozzle sorting chip and 488/638 nm lasers. Nuclei were first sorted using the ultra-yield (P6) or normal (P11) mode and then re-sorted using the normal mode.

For RNA-seq following FANS, 1600-3500 cells were sorted and stored at −80°C. The sorted nuclei were thawed on ice and reverse-crosslinked with 100 µl of RIP buffer [100 mM Tris-HCl pH 8.0, 10 mM EDTA, 1% SDS, 1 µl of RNAse inhibitor (Promega)] containing 1 µl of Proteinase K (New England Biolab) for 1 h at 65°C. Then, nuclear RNA was purified using a RNeasy micro kit (Qiagen) according to the manufacturer's instructions. Half the amount of purified RNA from FANS-isolated granule neurons was treated with a NEBNext rRNA Depletion Kit (New England Biolabs). RNA-seq was performed using libraries prepared with a NEBNext Ultra™ Directional RNA Library Prep Kit for Illumina (New England Biolabs). RNA-seq libraries were sequenced on the Illumina NextSeq 550 platform to obtain 37 bp paired-end reads. Two biological RNA-seq replicates were performed in all experiments.

Hi-C assays were performed as previously described (Rao et al., 2014; Yamada et al., 2019). For Hi-C following FANS, 3500-4800 cells for P6 or 23,000-50,000 cells for P11 were sorted into PBS and lysis buffer (10 mM Tris-HCl pH 8, 10 mM NaCl, 0.2% Triton-X) at four times the volume was added immediately. Nuclei were pelleted and resuspended with 0.5% SDS for permeabilization as described with modifications (Rao et al., 2014; Yamada et al., 2019). Hi-C libraries were sequenced on the Illumina NextSeq 550 platform to obtain 37 bp paired-end reads. Two biological Hi-C replicates were performed in all experiments.

RNA-seq reads were aligned to the mm10 reference genome with HISAT2 using the public server at https://usegalaxy.org/ and normalized by library size. Gene annotations were derived using GENCODE version M11. Differential gene expression analyses were performed using pair-wise negative binomial tests with edgeR (Robinson et al., 2010) and the false discovery rate (FDR) was calculated for all genes.

snRNA-seq datasets from the developing cerebellum were obtained and analyzed using Seurat (Hao et al., 2021; Rosenberg et al., 2018). Briefly, granule neurons in the P11 developing cerebellum were extracted for downstream analysis. Read counts were normalized and variance stabilized using sctransform and subjected to dimensionality reduction using PCA preprocessing followed by UMAP embedding (Hao et al., 2021). Gene modules were obtained using differentially expressed genes from FANS-RNA-seq with log₂FC>2 and FDR<0.05 for each granule neuron population. Genes that exhibited continuous upregulation or downregulation across all developmental stages, with an FDR<0.05 for each stage, were excluded from the immature granule neuron gene modules from P6 and P11 mice, which represent an intermediate developmental stage. To further distinguish the immature granule neuron modules from P6 and P11 mice, genes exhibiting higher expression in either P6 or P11 immature granule neurons were considered to represent the corresponding condition. Module scores were calculated from the average expression of genes within the module, subtracted by the levels of similarly expressed control genes (Hao et al., 2021).

ChIP-seq datasets from P6 and P22 mouse cerebellum were obtained (Zhao et al., 2023 preprint). ChIP-seq reads were aligned to the mm10 reference genome with Bowtie2 using the public server at usegalaxy.org and normalized by library size. Normalized H3K9me3 counts were derived genome-wide at 500 Kb resolution and the differential ChIP-Seq signals (diffBind) between the P6 and P22 cerebellum were calculated using pair-wise negative binomial tests with EdgeR and the FDR. The frequency distribution was calculated using the fraction of bins with differential H3K9me3 signals per chromosome, normalized by the overall fraction of bins with differential signals across all chromosomes.

Hi-C reads were aligned to the mm10 reference genome with Bowtie2 and HiC-Pro (Servant et al., 2015). Uniquely mapped paired-end reads were assigned to MboI restriction fragments and valid pairs with a minimum genomic distance of 1 Kb were filtered for PCR duplicates. Interaction matrices were normalized using Knight-Ruiz (KR) matrix balancing and observed divided by expected (o/e) transformation (Durand et al., 2016). Genomic regions associated with the hGD subcompartment, nuclear speckles, and nucleoli were obtained (Quinodoz et al., 2018; Zhao et al., 2023 preprint). A/B compartment strength was derived from the eigenvector of the Hi-C correlation matrix (Rao et al., 2014) using 500 Kb bins. Genomic loci enriched for the active histone modification H3K27ac and de-enriched for the repressive histone modification H3K9me3 were assigned a positive eigenvector and denoted as the active A compartment, while genomic loci with a negative eigenvector were denoted as the repressive B compartment.

Dip-C datasets from the developing and adult mouse cerebellum with cell type and structural stage annotations were obtained (Tan et al., 2023) and granule neurons were extracted for downstream analysis. Dimensionality reduction and visualization of Dip-C cells was performed using single-cell A/B compartment analysis as described (Tan et al., 2023), with modifications. Briefly, long-distance intra-chromosomal contacts separated by more than 3 Mb or inter-chromosomal contacts were kept for further analysis. Aligned reads were grouped into 100 Kb bins, and genomic bins with fewer than 10 long-distance contacts were excluded. For each bin, the average CpG density of contacting regions, weighted by the number of contacts, was calculated. A genomic bin x cell matrix was generated with each entry of the matrix containing the weighted CpG density of contacting regions. For bins with missing values, the weighted CpG density was interpolated from surrounding bins on the same chromosome, starting 1 bin away and extending to 2 bins if needed. Cells containing at least 5000 bins, as well as bins present in all cells, were considered for further analysis. The bin x cell matrix was then subjected to PCA preprocessing followed by UMAP embedding, as described for snRNA-seq analyses. Radial position analyses were performed from single cell 3D structures as previously described (Zhao et al., 2023 preprint), with modifications. Briefly, for 3D structures derived from F1 hybrid mice (Tan et al., 2023), the positions of genomic regions were projected onto the X-Y, Y-Z, or X-Z plane. Each plane is modeled as an ellipse based on the centroid and semi-axes calculated from datapoints in the plane. Planes with at least 40% of the maximum area among all planes in a stack were selected for downstream analysis. The 2D radial position was then calculated for each genomic region.

We thank members of the Yang Lab for helpful discussions and the Biological Imaging Facility at Northwestern University (RRID:SCR_017767) for a confocal laser scanning microsope (TCS SP8 confocal, Leica Microsystems) for DNA FISH and immunohistochemical analyses.