A single-cell transcriptome atlas of the maturing zebrafish telencephalon

Integrated graph clustering identifies cell types in the zebrafish telencephalon

To analyze the molecular features of the cells from the zebrafish telencephalon, we isolated and performed single-cell RNA-seq on the telencephalons of 6-day-postfertilization (dpf), 15-dpf, and adult zebrafish in nine, four, and two experimental batches, respectively (Fig. 1A; Supplemental Figs. S1B, S2). At 6 and 15 dpf, we dissected the telencephalon, including the pallium, subpallium, and olfactory bulb, as well as regions of the preoptic area that were continuous with the subpallium. In the adult, the olfactory bulb was removed during dissection to retain more single-cell coverage in the two main regions of interest: the pallium and subpallium. We collected a total of approximately 64,000 cells across these three different ages using the 10x Genomics platform. Quality filtering resulted in a total of 21,247 cells from the 6-dpf, 15,476 cells from 15-dpf, and 27,125 cells from the adult telencephalon (Fig. 1B).

To integrate the cells from the three different time points and identify shared and divergent clusters, we used Seurat's data integration pipeline, which combines canonical correlation analysis (CCA) and mutual nearest neighbor (MNN) analysis (Stuart et al. 2019). After integration, the cells were clustered and labeled based on the marker genes conserved across all three time points. Of the 38 clusters identified in the integrated analysis of the telencephalon, 26 clusters were neuronal, and 12 clusters were nonneuronal (Fig. 1C).

Among the nonneuronal cells, we found clusters expressing prototypical markers for progenitors, microglial cells, oligodendrocytes, and endothelial cells (Fig. 1C, side panels; Supplemental Table S1). The progenitors included radial glial cells (Lange et al. 2020) that express markers such as fabp7a and mdka, cycling progenitors that express markers such as pcna, and proneural precursors that express markers such as neurog1 (Supplemental Fig. S3). Of two progenitor pools, Progenitor_01 is the larger and labeled by fabp7a, a gene known to be expressed in the telencephalic ventricular zone in adult zebrafish (Adolf et al. 2006). The smaller Progenitor_02 cluster is labeled by fabp7b, and the differential expression (DE) of these two genes was previously described (Liu et al. 2004). We also found a cluster that expressed canonical mammalian astrocyte markers such as slc1a3b and slc1a2b, highlighted by a recent zebrafish study (Chen et al. 2020) and identified as dormant radial glia by past studies (Lange et al. 2020). In addition, these putative astrocytes are enriched in mfge8a, a marker reported to be specific for telencephalic astrocytes in mice (Zeisel et al. 2015). Based on RNA-FISH of cspg5b, a gene also enriched in this cluster, we spatially localized the cells to the medial region of the telencephalon lining the ventricles (Supplemental Fig. S4).

To assign an approximate regional identity to the mature neuronal clusters, we identified a host of markers for each (Supplemental Table S1). A large majority of clusters were defined by at least one gene that was a largely specific marker of that cell type (Fig. 2). We assigned clusters as belonging to pallium and subpallium based on the expression of known markers of these regions (Mueller et al. 2008; Ganz et al. 2012, 2014). Consistent with expression patterns in mammals, dlx1a, dlx2a, dlx2b, dlx5a, dlx6a, tbr1b, and neurod1 distinguish between the zebrafish pallium and subpallium. Among the neuronal clusters, we also found one cluster belonging to the olfactory bulb (Supplemental Fig. S4), two clusters belonging to the habenula (Pandey et al. 2018), and two belonging to the preoptic area. These clusters, established based on our previously published work or their spatial location in situ, were not analyzed further.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 2.

Marker genes of the neuronal cell types in the larval telencephalon. (A) Expression profiles of marker genes that are specific or enriched in the pallium subtypes. (B) Expression profiles of marker genes that are specific or enriched in the subpallium subtypes. The purple bar on the bottom represents previously defined markers that were used to subcategorize the region, and the green bar represents markers identified by the integrated analysis. For the expression of additional genes, see Supplemental Figure S5 and Supplemental Table S1.

We assigned nine neuronal clusters to the pallium, the dorsal region of the telencephalon that is thought to contain structures homologous to the mammalian cortex, hippocampus, and some sections of the amygdala (Cheng et al. 2014; Ganz et al. 2014). Previously, the adult zebrafish pallium was subdivided into subregions such as the dorsal, medial, ventral, and lateral based on the combinatorial expression of marker genes such as eomesa, emx1, emx2, and emx3 (Ganz et al. 2014). Most of these genes were broadly expressed across multiple clusters in our data set (Fig. 2A). Other broad pallium markers observed in our data set included bhlhe22, rprml, and zbtb18.

These previously described broad markers were used to ascribe a proposed regional identity for some pallium clusters (Fig. 2; Supplemental Figs. S1A, S5). Cells in clusters Pallium_02 and Pallium_03 are both eomesa⁺ and express low levels of emx1. Pallium_02, unlike Pallium_03, is also characterized by its expression of emx3. These expression patterns are consistent with a region near the posterior subdivision of the dorsal pallium (Dp) or the adjacent ventral subdivision of the lateral zone of the dorsal pallium (Dlv) (Ganz et al. 2014). Clusters Pallium_05 and Pallium_06 are heterogeneous (Supplemental Fig. S5). The majority of cells in Pallium_05 and Pallium_06 represent at least two subtypes of glutamatergic cells marked by expression of either c1ql3b.1 or efna1b. The efna1b⁺ cell type expresses eomesa but not emx1, emx2, or emx3, whereas the c1ql3b.1⁺ cell type has low expression of eomesa, emx1, emx2, and emx3. These expression patterns indicate localization of these cell types to the dorsal subdivision of the lateral zone of the dorsal pallium (Dld) that may be related to the hippocampus (Ganz et al. 2014). The remaining cells in Pallium_05 are eomesa⁻/emx3⁺, which is consistent with expression in the medial zone of the dorsal pallium (Dm) (Ganz et al. 2014), the proposed pallial amygdala. It is unclear whether the excitatory neurons in Pallium_05 and Pallium_06 occupy discrete regions of the telencephalon or whether cells in these clusters represent cell types shared between regions.

Cells in the remaining pallial clusters were identified based on markers from a recent paper describing the teleostean amygdaloid complex (Porter and Mueller 2020). The strong expression of pvalb7⁺ in Pallium_07 is consistent with this cluster contributing to the zebrafish “integrative olfactory pallium” (IOP) and the medial pallium or dorsolateral pallial territory (Dl), which is a putative hippocampal homolog (Porter and Mueller 2020). The pyya⁺ Pallium_01 cluster may also contribute to the IOP. Pallium_01 has some pvalb7 expression and the strongest emx1 expression of all clusters; in adult zebrafish, emx1 is expressed only in Dp, consistent with these cells contributing to the IOP. Cells in clusters Pallium_04, Pallium_08, and Pallium_09 express members of the LIM homeobox family, including lhx9, lhx2b, lhx5, and lhx1a. The zebrafish genes lhx5 and lhx1, as well as tbr1b, mark the thalamic eminence (Wullimann and Mueller 2004; Turner et al. 2016), suggesting that cells in these clusters may derive from this structure. A subset of cells in Pallium_04 coexpresses lhx5 and otpa, which is consistent with the recently identified posterior division of the medial amygdala (Porter and Mueller 2020). The gene emx2 is also expressed in Pallium_08 and a subset of cells in Pallium_09, which is exclusively expressed in the central division of the pallium (Dc) in the adult telencephalon (Ganz et al. 2014).

In addition to the mature neuronal clusters, we also identified three immature neuron (IN) clusters in the pallium that were all enriched in markers such as tubb5 (Supplemental Fig. S4) but likely contribute to different mature neuron clusters. The committed pallium precursors are distinguished from these IN clusters by the coexpression of progenitor genes such as her4.2 (Fig. 2A). Pallium_IN01 was enriched in the expression of bhlhe22, a marker that is broadly expressed in Pallium_05, Pallium_06, and Pallium_07, suggesting that these INs contribute to these three pallium clusters. Pallium_IN02 is eomesa⁺, bhlhe22⁻ and likely contributes to the eomesa⁺, bhlhe22⁻ Pallium_02 and Pallium_03. Pallium_IN03 expresses tbr1b but not the pallium markers eomesa, bhlhe22, emx1, emx2, or emx3. This pattern of transcription factor expression most closely resembles the remaining transcripts in the mature clusters Pallium_08 and Pallium_09 (Fig. 2A). Pseudotime analysis supports the divergent developmental trajectories of Pallium_IN01 and Pallium_IN02, as these cells differentiate into Pallium_06 and Pallium_02 clusters, respectively (Supplemental Fig. S6). Defining gene modules that change as a function of pseudotime uncovered new marker genes dividing the telencephalic neuron types (Supplemental Fig. S7). In contrast to the multiple subtypes of INs in the pallium, a single IN cluster was present in the subpallium (Fig. 2B).

We identified eight mature neuronal clusters belonging to the subpallium, the ventral region of the telencephalon, which is thought to contain structures homologous to the mammalian basal ganglia (Mueller et al. 2008). Previously, the zebrafish subpallium was divided based on the expression of genes such as dlx2a, dlx5a, and lhx6 (Mueller et al. 2008; Ganz et al. 2012). These genes do not mark single clusters, whereas our clustering subdivides the subpallium into transcriptionally distinct neuron types based on multiple new marker genes such as pdrm12b (Fig. 2B).

The zebrafish subpallium was found to contain GABAergic interneurons, a cholinergic population, cells with similarity to medium spiny neurons, and an excitatory septal cluster. Subpallium_01 expresses nkx2.1, lhx6, sst1.1, and npy, suggesting that this cluster represents somatostatin⁺, neuropeptide Y⁺ interneurons. The Subpallium_03 cluster expresses markers of parvalbumin⁺ interneurons, including nkx2.1, sox6, etv1, nxph1, and pvalb6. Subpallium_02 and Subpallium_05 express markers of striatal medium spiny neurons, such as tac1 (substance P), sp8a, synpr, penkb (enkephalin), six3b, cxcl14, and foxp1b (Aguda et al. 2021). These clusters also express the neuropeptides pyya and pyyb. Subpallium_06 expresses nkx2.1, lhx6, and the cholinergic markers lhx8a, gbx1, and isl1. This cluster likely represents a recently identified population of neurons that synthesizes both acetylcholine and GABA and is involved in social behavior in zebrafish (Ncube et al. 2022). Unlike other subpallium clusters, Subpallium_04 expresses the glutamatergic marker slc17a6b (vglut2) and low levels of the GABAergic markers slc32a1 (vgat), slc6a1b (gat), gad1b (gad67b), and gad2 (gad65). This cluster is defined by septal markers, including isl1, zic1, zic2a, zic4, and zic5 (Supplemental Fig. S5; Woych et al. 2022), as well as several unique marker genes (Fig. 2B). The Subpallium_07 cluster is heterogeneous, with subsets of cells expressing septal markers (zic1, zic2a, zic4, and zic5) (Inoue et al. 2007; Zeisel et al. 2018), cholinergic markers (nkx2.1, lhx6, lhx8a, and gbx1) (Sandberg et al. 2016), interneuron markers (calb1, calb2b, pvalb6, and sst1.1) (Bugeon et al. 2022), and the dopamine receptor drd2b. An unidentified subset of cells coexpresses sp8a, rgs4, and c16h2orf66.

Taken together, these results define neuronal subtypes of the pallium and subpallium and provide a catalog of markers for their further study.

Spatial localization of neuronal cell types in the zebrafish telencephalon

To determine the spatial location of the markers for a subset of these single-cell clusters and support their subpallial or pallial identity, we collected fluorescent in situ hybridization (FISH) data for some marker genes (Fig. 3; Supplemental Table S2) and registered these image stacks into a shared coordinate space (Supplemental Video S1). Similar to previously published work (Pandey et al. 2018), we found that the neuronal cell types in the pallium and subpallium were largely regionalized. By observing the larger domains within the pallium using broad markers such as bhlhe22, rprml, and zbtb18, we conclude that there are two distinct regions of the pallium along the anterior–posterior axis that encompass neuronal cell types identified by our analysis (Fig. 3A). The posterior half, defined by the expression of rprml, consists of neuronal cell types such as pdyn⁺ Pallium_04 and lhx9⁺ Pallium_08 that are located dorsally (Fig. 3A). Whereas pdyn occupies the medial region of the dorsal posterior pallium, lhx9 is located in the lateral half of the dorsal posterior pallium. Other posterior neuronal cell types include pyya⁺ Pallium_01 and prkcda⁺ Pallium_03, both of which are localized laterally along different Z-planes of the ventral pallium. In the bhlhe22⁺ anterior pallium, in correspondence with the cluster abundances in the single-cell data, Pallium_06 occupied the largest expression domain (Fig. 3A). The INs also occupy regionalized niches within the pallium (Fig. 3A). Published in situ data for some marker genes in the adult telencephalon (Supplemental Fig. S8; Diotel et al. 2015) are consistent with the larval data, adult subdivisions suggested by previous work (Ganz et al. 2014), and pseudotime clustering. Markers of Pallium_02 and Pallium_03 are found in Dp/Dlv, whereas markers of Pallium_05 and Pallium_06 are found in Dld, Dc, and Dm.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 3.

Spatial distribution of neuronal cell types in the larval telencephalon. (A) In situ expression patterns of cluster-specific marker genes for selected subclusters in the pallium at 10 dpf (dorsal view). RNA-FISH (green) was performed with a total-Erk (pale gray) costain for anatomy. Approximate demarcations of the pallium, based on the Z-Brain atlas masks (Randlett et al. 2015), are indicated by white dotted lines. (B) In situ expression patterns of cluster-specific marker genes for selected subclusters in the subpallium at 10 dpf (dorsal view). Approximate demarcations of the subpallium only (Subpallium_04, Subpallium_05, Subpallium_06) or subpallium and preoptic area (all other panels), based on the Z-Brain atlas masks (Randlett et al. 2015), are indicated by white dotted lines. The scale bar represents 50 µm. For the three-dimensional stack and depth information, see Supplemental Video S1.

Subpallium neuronal cell types were also regionalized along the anterior–posterior and dorso–ventral axis in the ventral part of the telencephalon. Neuronal clusters Subpallium_05 (penkb⁺) and Subpallium_02 (synpr⁺) were located in the anterior half of the subpallium along the dorso–ventral axis (Fig. 3B). In contrast, sst1⁺ Subpallium_01 is localized to the posterior half of the subpallial domain. Other neuronal clusters, including the prdm12b⁺ Subpallium_07, are localized to a band of neurons in the medial part of the subpallium along the anterior–posterior axis (Fig. 3B). We also localized the otpa⁺ and pdyn⁺ cluster of preoptic area neurons, PoA_01 and PoA_02, respectively, which were readily identified by their spatial position (Fig. 3B).

Together, these results provide validation and spatial localization for the majority of neuronal clusters found in the telencephalon through our single-cell analysis.

Comparison of neuronal cell types in the zebrafish and mouse forebrain

To further identify the zebrafish telencephalic neuron types, we compared zebrafish markers (Supplemental Fig. S5) to the mammalian literature (Supplemental Table S3) and performed computational integration with an adolescent mouse data set (Fig. 4A,B; Zeisel et al. 2018).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 4.

Integrated clustering of zebrafish and mouse forebrain cells. (A) UMAP representation of mouse and zebrafish mature neurons analyzed without integration. (B) UMAP representation of mouse and zebrafish mature neurons integrated with Harmony. (C) UMAP representation of the integrated analysis colored according to original identities in zebrafish samples. Gray cells (NA) represent mouse cells. (D) UMAP representation of the integrated analysis according to the original locations in mouse samples. Gray cells (NA) represent zebrafish cells.

Cells in clusters Pallium_02 and Pallium_03 were found to share expression of genes with the mouse subiculum (Supplemental Table S3), suggesting that Pallium_02 and Pallium_03 may represent the area between the hippocampus proper and the entorhinal cortex. The computational integration supports this designation (Fig. 4C,D), as these zebrafish cells cluster near mouse cells located in the subiculum and entorhinal cortex. Mouse orthologs of the markers of Pallium_02 and Pallium_03—Rprm (Pallium_02 only), Rspo2 (a subset of Pallium_03), and Cbln1 (both)—are expressed in the murine subiculum, piriform area, postpiriform transition area, piriform–amygdalar area, lateral amygdalar nucleus, basolateral amygdala, and entorhinal cortex (Supplemental Fig. S9; Supplemental Table S3). Together, expression of marker genes suggests that cells in clusters Pallium_02 and Pallium_03 share expression patterns with the murine subiculum, entorhinal cortex, and pallial amygdala.

Clusters Pallium_05, Pallium_06, and Pallium_07 are scattered among mouse cells located in the entorhinal cortex, isocortex, CA1, CA3, and piriform cortex (Fig. 4C,D). Pallium_07 is marked by high expression of pvalb7, which is expressed in the IOP region that has been proposed as the putative homolog of the mammalian entorhinal cortex (Porter and Mueller 2020) and the medial pallium (hippocampus homolog). Mouse orthologs of genes marking the efna1b⁺ population in the Pallium_05 and Pallium_06 clusters are expressed in the murine isocortex, hippocampus, and basolateral amygdala (Supplemental Figs. S5, S9). Cells expressing c1ql3b.1, another cell type found in Pallium_05 and Pallium_06, additionally express nptx1l, which is consistent with a report that NPTX1 coimmunoprecipitates with C1QL3 and that C1ql3 and Nptx1 are coexpressed in mouse cortical neurons (Sticco et al. 2021). This cell type is found in multiple mouse brain regions.

The emx2 expression in Pallium_08 and Pallium_09, marking these clusters at the Dc region, suggests that they may be located in a functional correlate of the mammalian cortex (Cecchi 2002; Ganz et al. 2014). However, we did not identify a population of cells that coexpresses emx2 and eomesa, which is also expressed in Dc (Supplemental Fig. S5). One hypothesis for these observations is that emx2⁺/eomesa⁻ cells migrate to Dc from the thalamic eminence, whereas emx2⁻/eomesa⁺ cells derive from the dorsal proliferative matrix of the dorsal pallial division (Mueller et al. 2011). This hypothesis is strengthened by the expression of markers of Cajal–Retzius cells, which in mammals migrate from regions including the dorsal septum and thalamic eminence (Jiménez and Moreno 2022). The signature of the thalamic eminence, coexpression of Lhx1 and Lhx5, is also conserved between the mouse (Adutwum-Ofosu et al. 2016) and zebrafish. Adding further support to this classification, the computational integration clustered Pallium_08 near the murine isocortex and Cajal–Retzius cells in the UMAP plots (Fig. 4C,D), marked by expression of Emx2 (Supplemental Fig. S9). Determining whether emx2⁺ cells in Dc derive from the thalamic eminence will require further investigation.

Integration of the single-cell data sets between the two species suggested that the subpallium clusters were largely consistent between zebrafish and mouse. Cells in Subpallium_01 cluster near hippocamposeptal and long-range projection interneurons (Fig. 4C,D), which are Sst⁺ (Supplemental Fig. S9). In mice, the medial ganglionic eminence (MGE) gives rise to interneurons that express somatostatin or parvalbumin downstream from the transcription factors Nkx2.1 and Lhx6, also expressed in Subpallium_01 (Fig. 2B). Cells in Subpallium_03 cluster near MGE-derived neurogliaform and axo-axonic interneurons (Fig. 4C,D), which express Nxph1 (Supplemental Fig. S9). Interneurons derived from the caudal ganglionic eminence (CGE) in mammals, marked by coexpression of orthologs of Nr2f2 (Coup-TFII), Prox1, Sp8, Reln (reelin), Calb2 (calretinin), or Vip (vasoactive intestinal polypeptide), were not observed in the zebrafish telencephalon (Supplemental Fig. S5). Subpallium_06 clusters near cholinergic neurons that express Lhx8 (Fig. 4C,D), a transcription factor required for forebrain cholinergic neuron specification in mice (Zhao et al. 2003). As expected from their expression of striatum-like markers (Gokce et al. 2016) such as penkb⁺ and synpr⁺, Subpallium_02 and Subpallium_05 cluster near lateral ganglionic eminence (LGE)–derived striatal medium spiny neurons. Cells in Subpallium_02 cluster near D1 medium spiny neurons, which are marked by the expression of Tac1 (Supplemental Fig. S9). Subpallium_05 clusters near D2 medium spiny neurons and Cck interneurons and is marked by the expression of Penk. Orthologs of the dopamine receptors are minimally expressed in our data, with drd2b expressed substantially only in Subpallium_08. In addition to their presence in the zebrafish striatum equivalent, cells in Subpallium_02, Subpallium_05, and Subpallium_08 express markers indicating a potential contribution of these clusters to the LGE-derived nuclei of the extended amygdala (Porter and Mueller 2020). Cells in Subpallium_07 and Subpallium_08 are spread among multiple mouse cell types, including pallidal inhibitory neurons, basket and bistratified cells, cholinergic neurons, and interneuron-selective interneurons. Together, these analyses suggest that the cell types in the zebrafish subpallium generally resemble those in the mouse, whereas the cell types in the pallium are more divergent but can be mapped to plausible brain regions.

Comparison of the larval telencephalon to adult telencephalon in the integrated data

To assess the conservation and emergence of neuronal cell types from the larval to adult telencephalon, we investigated the changes in cellular proportions for the clusters identified by the integrated analysis across all three stages. Of the nine neuronal cell types we identified in the pallium, two expand in proportion from each representing ∼10% of the total pallium cells in larvae to >30% of the pallium in the adult telencephalon (Fig. 5A). These include the c1ql3b⁺ Pallium_05 and efna1b⁺ Pallium_06. Similarly, Pallium_07, which comprises only ∼3% of the total pallium cells in the 6-dpf animal, expands to ∼10% of the cells in the adult pallium. Pallium_07 represents a cluster of slc17a6a and slc17a6b⁺ excitatory neurons that are also marked by the expression of pvalb7 (Fig. 5B; Supplemental Fig. S5; Mueller et al. 2011). We validated the specific expression of pvalb7 in larva and adult telencephalon and found strong expression of the gene in the lateral part of the adult telencephalon but weak and spotty expression of the gene in the medial part of the larval telencephalon (Fig. 5C). Together, our ISH and single-cell data indicate that this subtype domain starts to emerge in larva and continues to grow and expand into adulthood.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 5.

Conservation and emergence of neuronal diversity between 6-dpf, 15-dpf, and adult telencephalon. (A) The contribution of select pallium clusters to the total pallium cells in 6-dpf, 15-dpf, and adult pallium. (B) Gene expression profiles of pvalb7 faceted by age. (C) In situ expression domain of pvalb7 in larval (green, RNA-FISH; pale gray, total-Erk; costain for anatomy) and adult telencephalon (magenta, RNA-FISH; yellow, Sytox nuclear stain). Closer views of the left and right expression domains in the adult are shown on the corresponding sides of the complete brain section. The scale bars represent 50 µm. (D) The contribution of select subpallium clusters to the total subpallium cells in 6-dpf, 15-dpf, and adult pallium. (E) Gene coexpression profiles for tubb5 with bhlhe22 and six3b. Statistical test for plots in A and D: unpaired two-tailed t-test with Benjamini–Hochberg correction (Benjamini and Hochberg 1995) for multiple comparisons: (*) P < 0.05, (**) P < 0.01, (***) P < 0.001, (****) P < 1 × 10⁻⁴).

The increase in size of these three pallial clusters was accompanied by a decrease in the proportion of committed pallium precursors and INs in the adult pallium. Among the three IN clusters we identified in the pallium, only Pallium_IN02 and Pallium_IN03 disappear completely in the adult telencephalon. In contrast, Pallium_IN01, which represents the bhlhe22⁺ class of INs, continues to persist into adulthood despite decreasing in proportion (Fig. 5E, top panel). The three clusters that expand in number belong to this bhlhe22⁺ cluster of cells, suggesting that these may represent the neuronal cell types that not only expand well into adulthood but also might be capable of regeneration (Kizil et al. 2012; Lange et al. 2020). Indeed, a recent study revealed that vim is up-regulated following telencephalic injury (Kutsia et al. 2022), and vim is a marker of Pallium_IN01 in our data (Fig. 2A, 3A). In addition, the proportion of rprma⁺ Pallium_02 and pdyn⁺ Pallium_04 declines steadily from constituting ∼10% of the pallium in larval animals to 2%–3% of the pallium in adults (Fig. 5A).

In contrast, the cellular proportions of the subtypes in the subpallium remain more consistent compared with the pallium. Although the proportions of the committed subpallium precursors and subpallium INs decrease from larva to adult, the only subpallial subtype that expands significantly in size is the six3b⁺ Subpallium_05 (Fig. 5D). In line with this observation, of the INs that persist in adulthood, a proportion of these is also positive for six3b, suggesting that these INs might be resulting in the expansion of Subpallium_05 later in development (Fig. 5E, bottom panel). These changes in cell proportion of pallium and subpallium neuronal cell types were not a result of disappearing progenitors and INs, as was observed in a separate analysis of cell proportions of just mature neuronal populations (Supplemental Fig. S10A). Finally, although some Subpallium_02 cells persist in adult animals (Fig. 5D), the proportion of the population and corresponding marker genes is substantially reduced (Supplemental Fig. S10B). In addition to the neuronal subtypes, we also found an adult-specific group of endothelial cells (Supplemental Fig. S10C).

Although the key markers of the pallium and subpallium clusters identified by the integrated analysis were largely similar across the 6-dpf, 15-dpf, and adult fish, we performed a systematic comparison of the gene expression profile of each mature neuronal cluster across the three stages analyzed in our study. We generated pseudobulk expression profiles of each pallium and subpallium cluster by separately aggregating cells of different ages into different bulk expression profiles. We then performed cluster-specific DE comparison among all time points: 15 dpf versus 6 dpf (Supplemental Fig. S11A; Supplemental Table S4), adult versus 15 dpf (Supplemental Fig. S11B; Supplemental Table S5), and adult versus 6 dpf (Supplemental Fig. S11C; Supplemental Table S6). Both pallium and subpallium clusters show transcriptional changes despite the identifying markers remaining largely similar across time points. The largest number of DE genes between 15 dpf and 6 dpf as well as between adult and 6 dpf was observed in efna1b⁺ Pallium_06 (Supplemental Fig. S11A,C), whereas the largest number of DE genes in the 15 dpf versus adult comparison was in pyyb⁺ Subpallium_02 (Supplemental Fig. S11B).

Integrated analysis is a powerful approach for comparing cell types across related data sets. However, combining the data from the three ages could also mask more subtle differences between the ages. Therefore, we chose to extend our developmental comparison by analyzing the data from each age separately.

Comparison of the larval telencephalon to adult telencephalon with separate clustering

To determine additional similarities and differences between clusters of forebrain cells across developmental stages, we split the data by age and reclustered (Fig. 6A–C). We then mapped each cell to its original location in the integrated space (Fig. 6D–F). The new clusters at each age were connected to the originally defined integrated clusters (Supplemental Table S10) by combining the following information: (1) the location of the cells in relation to the originally defined conserved clusters in the UMAP plot (Fig. 1A), (2) the marker genes for the new clusters at each age (Supplemental Tables S7–S9), and (3) the expression patterns of known marker genes and those identified in the integrated analysis (Supplemental Figs. S12–S17). Clusters that mapped readily to the integrated data are colored identically in Figure 6, whereas new or split clusters are a unique color (Supplemental Table S10).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 6.

Separate clustering of each age and the relationship of clusters to the integrated space. (A) UMAP representation of the 6-dpf data clustered separately. (B) UMAP representation of the 15-dpf data clustered separately. (C) UMAP representation of the adult data clustered separately. (D) Visualization of the 6-dpf cells mapped onto the integrated space. (E) Visualization of the 15-dpf cells mapped onto the integrated space. (F) Visualization of the adult cells mapped onto the integrated space.

This separated analysis mirrored some conclusions from the integrated analysis. Similar findings included the loss of the IN types Pallium_IN02 (19 at 6 dpf, 20 at 15 dpf), Pallium_IN03 (18 at 6 dpf, 23 at 15 dpf), and Subpallium_IN (nine at 6 dpf, two at 15 dpf). Pallium_IN01 is present at all three ages but is substantially reduced by the adult stage (clusters 22 and seven at 6 dpf, six and 15 at 15 dpf, 24 at adult). Findings such as the expansion of the Subpallium_05 cluster (21 at 15 dpf, one at adult) and the lack of Pallium_07 cells (not clustered at 6 and 15 dpf, five at adult) until the adult stage are visually apparent when mapping the new clusters in the integrated space (Fig. 6D–F). Other age-dependent changes, such as the reduction in the Pallium_02 and Pallium_04 subtypes (Fig. 5A), are less visually clear but are indicated by the lack of cluster formation (Supplemental Table S10). This difference and others, such as the lack of a cluster forming for Subpallium_01 at any age (Fig. 6D–F; Supplemental Table S10), may be the result of fewer cell numbers per age, leading to a decrease in the power to identify cell types in the separated analyses compared with the integrated.

Although many findings were shared between the two analysis approaches, the separated clustering also identified additional differences between the adult and larval stages. Sample age impacted both neuronal and nonneuronal cell populations. As was recently reported (Wu et al. 2020), there are two types of microglia in the adult zebrafish brain. The adult samples included both the ccl34b.1-expressing phagocytotic microglia and white-matter enriched regulatory microglia that lack ccl34b.1 and express markers such as id2a and batf3, whereas the 6 and 15 dpf only contained the phagocytotic microglia (Supplemental Fig. S18). Second, an immune cell population found in the integrated analysis was only present in the adult ages (Supplemental Fig. S19). We have designated these cells as natural killer (NK)–like, based on a previous study of immune populations in zebrafish that defined top marker genes such as ccl38.5 and il2rb (Carmona et al. 2017). NK cells were recently found in the dentate gyrus of the aging brains of mice and humans (Cuapio and Ljunggren 2021; Jin et al. 2021), supporting this designation. Among the neuronal populations, there were several new clusters formed only in the adult analysis (Fig. 6C), but most had high similarity to identified clusters (Supplemental Table S10). However, one new pallium cluster stood out, characterized by unique marker genes that did not form a cluster in the integrated analysis and were only weakly expressed at larval ages (Supplemental Fig. S20). This cluster, number 26 in the adult data, expresses the marker genes timp2b, dgkzb, rnd3b, and ngef. In mice, TIMP2 is involved in hippocampal-dependent cognition and function and is expressed in the hippocampus (Castellano et al. 2017). It is antimitogenic and promotes neuronal differentiation (Pérez-Martínez and Jaworski 2005). In mammals, DGKZ is expressed in the hippocampus and promotes neurite outgrowth (Kim et al. 2010), as does Rnd3 (Peris et al. 2012; Jie et al. 2015). It is conceivable that this cell population, which is far more prominent in the adult forebrain, is involved in the acquisition of adult-specific learning and memory behavior.

Together, the integrated and separated analysis of these single-cell data sets provides a resource for future investigation into the diversification and transcriptional programming of forebrain cell types as zebrafish develop from larval to adult stages.