The lncRNA Snhg11, a new candidate contributing to neurogenesis, plasticity and memory deficits in Down syndrome

Down syndrome (DS) stands as the prevalent genetic cause of intellectual disability, yet comprehensive understanding of its cellular and molecular underpinnings remains limited. In this study, we explore the cellular landscape of the hippocampus in a DS mouse model through single-nuclei transcriptional profiling. Our findings demonstrate that trisomy manifests as a highly specific modification of the transcriptome within distinct cell types. Remarkably, we observed a significant shift in the transcriptomic profile of granule cells in the dentate gyrus (DG) associated with trisomy. We identified the downregulation of a specific small nucleolar RNA host gene, Snhg11, as the primary driver behind this observed shift in the trisomic DG. Notably, reduced levels of Snhg11 in this region were also observed in a distinct DS mouse model, the Dp(16)1Yey, as well as in human postmortem tissue, indicating its relevance in Down syndrome. To elucidate the function of this long non-coding RNA (lncRNA), we knocked down Snhg11 in the DG of wild-type mice. Intriguingly, this intervention alone was sufficient to impair synaptic plasticity and adult neurogenesis, resembling the cognitive phenotypes associated with trisomy in the hippocampus. Our study uncovers the functional role of Snhg11 in the DG and underscores the significance of this lncRNA in intellectual disability. Furthermore, our findings highlight the importance of the DG in the memory deficits observed in Down syndrome.


Introduction
Down syndrome (DS) is caused by trisomy of Homo sapiens chromosome 21 (HSA21) and is the most common cause of genetic intellectual disability, affecting more than 5 million people globally. DS alters central nervous system development and function, impairing cognition, and adaptive behavior. Deficits in hippocampal-mediated learning and memory processes are hallmarks of DS [1,2], and molecular and cellular defects have been detected in post-mortem fetal DS hippocampus [3,4]. DS is a disorder of gene expression deregulation, as the triplication of HSA21 results in a global disturbance of the transcriptome that is proposed to contribute to the phenotypic manifestations of DS [5]. This global gene expression deregulation is likely caused by alterations intrinsic to the extra copy of HSA21, such as the overexpression of genes involved in epigenetic regulation. In fact, several studies have suggested chromatin dysfunction in DS [6][7][8][9][10][11]. However, other possible mechanisms associated with the regulation of chromatin function are still unexplored.
Despite HSA21 being the smallest autosome, it is highly enriched in long-non coding RNAs (lncRNAs) [12], which are transcripts with a length of more than 200 nucleotides that are not translated into functional proteins. Moreover, a high number of lncRNAs are abnormally expressed in DS [13][14][15]. Interestingly, a growing body of evidence from recent studies emphasize the role of lncRNAs in brain function, including learning and memory [16][17][18] and adult neurogenesis [19,20], but li le is known about their direct function.
Epigenetic mechanisms, including the expression of lncRNAs, are highly cell-type specific, thereby providing a layer of regulation for precise transcriptional control in each cell type. Therefore, their deregulation is expected to have a differential impact on the transcriptome of each cellular subtype. In fact, although bulk RNA-sequencing studies have provided evidence of global disturbance of the transcriptome in the trisomic brain [5,12,21], the high cell heterogeneity of the brain tissue greatly hampers the capacity of these studies to elucidate the full complexity of gene expression deregulation in the trisomic brain and to identify specific genes responsible for specific clinical phenotypes.
Here, we used single nucleus RNA sequencing to dissect transcriptional dysregulation associated with specific cell types in the DS mouse model Ts65Dn. Of the DS mouse models generated to date only Ts65Dn, Tc1, Ts66Yah and TcMAC21 are true aneuploid models with a freely segregating supernumerary chromosome, which may be important for some of the DS features not found in other DS mouse models with an intrachromosomal segmental duplication. We selected for our study the Ts65Dn, as it recapitulates many of the features found in DS. However, Ts65Dn mice also carry a triplication of 43 coding genes which are non-orthologous to HSA21, and are not triplicated in human DS. As such, we validated some of our results in a second mouse model of DS the Dp(16)1Yey (Dp16), which has a duplication of the HSA21 orthologous region on MMU16 [22] and also in human postmortem dentate gyrus of DS brains.
Our study revealed a cell-type specific alteration of the transcriptome and detected previously unknown differentially expressed genes in specific neuronal populations.
Strikingly, we identified Snhg11, a lncRNA, to be specifically downregulated in the trisomic dentate gyrus and we provide evidence for its involvement in neuronal function, adult neurogenesis and hippocampal-dependent memory.

Unbiased identification of neuronal subtypes in hippocampus
To address the question of how the different neuronal populations are affected in the trisomic (TS) hippocampus, we isolated the NeuN+ population by fluorescence activated nuclear sorting (FANS; see Methods). After FANS, using 10X we sequenced 27602 and 28545 nuclei with high integrity from four WT and four TS mice, respectively (Fig. 1a). Since single nuclei RNA-seq (snRNA-seq) profiles nuclear RNA, our gene expression profile data reflect nascent transcription, as well as the cellular transcriptome. Single-cell feature-barcode matrices were used to embed cells in a K-nearest neighbor graph that defines cell clusters in an unbiased manner. Nuclear transcriptomes were visualized using a uniform manifold approximation and projection (UMAP) plot (Fig. 1b). We detected 17 clusters of cells sharing similar gene expression pa erns. These clusters did not result from technical or batch effects ( Supplementary Fig. 1). To determine the identity of each cluster, we identified a total of 1191 cluster marker genes for the different hippocampal neuron subpopulations (Supplementary Table 1). Classical gene markers, such as Slc17a7 for excitatory neurons and Gad1 for interneurons showed a clear enrichment in the corresponding clusters ( Fig. 1c). Glutamatergic cells were further mapped to five hippocampal subregions, by analyzing the expression pa ern of the top subregional marker genes (Fig. 1d) Table 2). Using this same approach, we were also able to identify the major subregions of the RHP, namely presubiculum, subiculum and parasubiculum ( Supplementary Fig. 2b).

Differential expression analysis reveals neuronal subtype-specific alterations in TS hippocampi
The gene expression analyses of the trisomic and euploid major neuronal subtypes (CA1, CA2, CA3, DG, RHP and interneurons) revealed a total of 291 differentially expressed genes (DEGs; Supplementary Table 3). The highest portion of DEGs, mostly upregulated, were located to the mouse chromosome 16 (Mmu16) (Fig. 2a), a portion of which is triplicated in Ts65Dn. Among the 42 Mmu17 genes that are triplicated in Ts65Dn but are non-orthologous to HSA21 genes, 7 of them were deregulated. We found that the impact of the trisomy on the transcriptome is cell type specific (Fig. 2b,   Supplementary Fig. 3 and Supplementary Fig. 4a). We also found common DEGs in more than one neuronal subtype. The highest number of shared DEGs between two cell types was found between CA1 and CA3, reflecting their close biological identity and functionality. As expected, a decreasing number of genes was found to be commonly deregulated, being the trisomic genes more prone to be overexpressed across cell subtypes ( Supplementary Fig. 4a). The cell type specific impact of the trisomy is also illustrated by the finding that most of the DEGs identified at the single nuclei level were not detected in the pseudobulk analysis (i.e. pooling the gene expression values of individual cells, and then calculating the average expression levels of each gene across all cells) of the same samples ( Supplementary Fig. 4b).
The identification of cell subtype specific DEGs resulting from the trisomy might be of interest to elucidate the mechanisms that lead to DS hippocampal dysfunction ( Fig. 2c, d). For instance, Epha6, a key regulator of neuronal and spine morphology, was specifically downregulated in CA1 trisomic neurons. Instead, Eid1, which has been associated with an impaired synaptic plasticity of CA1 pyramidal neurons [24], is upregulated in this region. Interestingly, in the DG we found altered expression of genes related to neurogenesis (Fig. 2d) The trisomy leads to a major shift in the dentate gyrus transcriptomic profile To visualize cell-type specific changes in the transcriptomic profiles of WT and trisomic neurons, we overlapped the UMAP plot from each genotype (Fig. 3a).
Strikingly, we observed a major shift in the two dimensional embedding of trisomic granule cells from the trisomic DG compared to WT, whereas no shift was detected in other neuronal subtypes. The trisomic DG shift was quantified by testing if the euclidean distance, a measure of the difference between the gene expression profiles of cells of the two genotypes, was larger than expected by chance, confirming the observed shift in the DG and revealing a milder shift in CA1 ( Supplementary Fig. 5 a).
To validate this transcriptomic shift, we subse ed the DG nuclei and repeated the clustering and the two-dimensional reduction by an independent approach, the Diffusion Map, an approach that allows to identify which genes are contributing the most to the position of each cell in the embedding [39,40]. As in UMAP, we also observed a shift that is specific to the trisomic DG (Fig. 3b) (Fig. 3c), which indicates that this gene is the main driver of the shift in the two-dimensional embedding, as global gene relevance identifies drivers of the overall embedding, and local gene relevance identifies those of a defined sub-region. Snhg11 is a member of the non-coding small nucleolar RNA host gene (SNHGs) family. Interestingly, this long non-coding gene is differentially expressed specifically in the trisomic DG, where its expression is strongly reduced (Fig. 3d). The reduced expression of this gene was also observed in the DG of an independent mouse model, the Dp16 (Fig. 3e) and, most importantly, in the DG from DS patients ( Fig. 3f and Supplementary Fig. 5c) compared to unaffected controls. Netrin G1, Ntng1 is also dramatically downregulated in this subregion (Fig. 2c, Fig. 3d) and also shows a high global gene relevance in the two dimensional embedding shift of the trisomic DG (Fig. 3c). Ntng1 is a known marker of mature granule cells [41], which suggested that cellular and functional identity of trisomic granule cells could be compromised. In fact, we observed a significant loss of DG marker genes, as shown by the significantly lower Jaccard index (Fig. 3g) and a reduced number of neurons in the trisomic DG (Fig. 3h). This reduction of marker identity was specific to the DG, suggesting that the identity of granule cells is compromised. Furthermore, while no compositional changes are observed in the rest of hippocampal subregions, except for a significant increase in interneurons, a finding that is in line with previous studies in DS [31][32][33][34], we found the granule cell population to be reduced in the trisomic hippocampus (Fig. 3h, Supplementary Fig. 6

ASO-mediated knockdown of Snhg11 in vivo leads to trisomic-like transcriptomic alterations
We then addressed the potential involvement of Snhg11 in the DG, with its expression and function in the brain still largely unexplored. As we observed the specific downregulation of Snhg11 in trisomic mice within the DG, we directed our a ention towards this particular subregion within the WT hippocampus. Our analysis involved examining the expression of Snhg11 and Ntng1 expression, using in situ hybridization. As previously reported [41], Ntng1 signal appeared distributed along the granule cell layer as punctuated and with a cytoplasmic localization (Fig. 3f).  Fig. 7e, f).
Once the efficacy of the ASO-mediated knockdown of Snhg11 was confirmed in vitro, we investigated its impact in vivo by a bilateral injection of 50 µg of either Snhg11-ASO or Control-ASO specifically in each of the hemispheres of the dorsal DG of 4 months-old WT mice (Fig. 4a). The specific knockdown of Snhg11 in the dorsal hippocampus was confirmed by RT-qPCR (Fig. 4b). To discard possible off-target effects, the expression of Snhg11 was also measured in the ventral hippocampus ( Supplementary Fig. 8a). We also observed a reduced expression of the two small nucleolar RNAs (snoRNAs) hosted by the gene (Supplementary Fig. 8b and c).
In order to investigate the role of these snoRNAs (Gm25187 and Gm25129) in the  Fig. 8c; R2). In addition, we also found a reduction in the pU modification levels of 18S:1137 (~5% loss in mismatch frequency in both replicates ( Supplementary Fig. 8e)), which we speculate to be a target of ACA60 (Gm25129; Supplementary Fig. 8f). This analysis suggests that while minor differences in the rRNA modification levels are observed in the rRNA upon Snhg11 knockdown, it is unlikely that these extremely modest differences are responsible for the phenotypes observed.
We also tested whether the reduction in Snhg11 expression was accompanied by a loss of granule cell identity and found a reduction in Ntng1 expression ( Supplementary Fig. 8g), similarly to what we observed in the TS DG.
In order to investigate the role of Snhg11 in the DG and its contribution to the Dpp6 [53]. In line with this observation, the Gene Ontology analysis of the downregulated genes revealed that the pathways affected upon Snhg11 knockdown are mainly related to neurogenic and synaptic function categories (Fig. 4d), a similar enrichment that was observed among genes downregulated in the trisomic DG (Fig.   4d). Altogether, our results support the relevance of Snhg11 for DG-dependent deficits in DS.

Snhg11 knockdown recapitulates DS DG-dependent neurogenesis, synaptic plasticity and memory deficits
To further investigate the relationship between Snhg11 and adult neurogenesis, we quantified the generation of new neurons in the DG of animals injected either with Snhg11-ASO or with Control-ASO (Fig. 5a). To this aim, we sacrificed the animals 3 days upon injection and processed their brains for histology. In our experiments, the density of BrdU + cells in the DG region was reduced by~30% in Snhg11-ASO injected mice compared to Control-ASO injected mice (Fig. 5b, c). In line with the observation of an altered transcriptome affecting neurogenic pathways upon Snhg11 knockdown, these results indicate that Snhg11 is required for optimal adult neurogenesis in the DG.
In order to determine the role of Snhg11 in DG synaptic plasticity we induced long-term potentiation (LTP) in mouse brain slices. We in synaptic plasticity processes (Fig. 5d, e). The affectation of some genes related with normal synaptic function in the transcriptome of Snhg11 knockdown supports the idea that Snhg11 has a crucial role in synapse plasticity.
Our results indicated that Snhg11 plays an important function in adult neurogenesis and synaptic plasticity in the DG, which are fundamental processes for hippocampal-dependent memory. To assess whether the observed alterations upon Snhg11 knockdown lead to memory deficits that recapitulate the ones observed in TS animals, we performed a behavioral ba ery in male and female WT animals injected either with Snhg11-ASO or Control-ASO injection (Fig. 6a) and also in non-injected WT and TS animals. We started the behavioural ba ery 3 days after the injection of the ASOs.
Strikingly, the reduction of Snhg11 had a dramatic impact on all the cortico-hippocampal memory paradigms tested. In the Novel Object Recognition (NOR) task, which measures recognition memory, non injected and WT mice injected with the Control-ASO properly acquired memory and recalled it both after 1 hour ( Fig.   6b) or after 24 hours (Fig. 6c) as shown by the pronounced preference for the novel object measured by the Discrimination Index (DI). Instead, WT animals that were administered with Snhg11-ASO failed to display object recognition both at short or long term to a similar extent as trisomic mice (Fig. 6b, c). In order to address more specifically the impact of Snhg11 knockdown on DG-dependent tasks, we assessed whether Snhg11-ASO-injected animals presented an altered spatial pa ern separation memory, which strictly depends on the DG [54,55]. For this, we used a paradigm known as Object Pa ern Separation [56]. In this case, the loss of Snhg11 led to an even more dramatic reduction of memory performance than the one observed in TS, whereas both WT and Control-ASO-injected mice showed a high DI (Fig. 6d).
In order to confirm the DG-specific impact of the Snhg11 knockdown, we lastly subjected a separate group of animals to the Delay Fear Conditioning paradigm, which is a memory task known to be independent of hippocampal function (Fig. 6e).
As expected, in this case we did not observe any performance differences between Control-ASO and Snhg11-ASO-injected animals. Similarly, no changes were observed on motor activity, explorative behavior or anxiety ( Supplementary Fig. 11).

Discussion
DS is a disorder of gene expression deregulation, as the triplication of HSA21 results in a global disturbance of the transcriptome that is proposed to contribute to its phenotypic manifestations. However, the mechanisms regulating gene expression are highly cell-type specific, and thus, the full complexity of gene deregulation in DS may be missed in bulk studies. This is particularly relevant in those tissues with high cellular heterogeneity such as the hippocampus, where the extent to which each cellular subtype is affected by the trisomy remains elusive.
To overcome this limitation, we have generated the first single-nuclei atlas of a trisomic hippocampus by characterizing the transcriptome of tens of thousands of individual hippocampal neurons in parallel in the Ts65Dn mouse model of DS.
Notwithstanding the fact that Ts65Dn mice carry a triplication of 43 coding genes which are non-orthologous to HSA21, and are not triplicated in human DS, we chose to perform our snRNAseq on this DS model as Ts65Dn is a true aneuploid model with an extra freely segregating marker chromosome, which may be important for some of the DS features. A new model, TcMAC21 mouse model, has been recently generated, that overcomes the drawbacks of previous models in that it is not mosaic and contains a near complete (93%) HSA21, but it was not available when we started the project.
Furthermore, there is uncertainty regarding the interactions between proteins encoded by human genes and mouse orthologs [57,58], and the presence of a significant number of non-coding human genes (>400) with uncertain effects on the mouse transcriptome.
In the present study we show that, although it has been previously suggested that illustrated by the important differences in DE genes found between pyramidal neurons located in different subregions of the CA. As discussed above, one of the limitations of our study is the use of Ts65Dn, which besides overexpressing two thirds of the triplicated HSA21 ortholog genes, it is also also trisomic for non-HSA21 orthologs.
Importantly, among the 42 Mmu17 genes that are triplicated in Ts65Dn but are non-orthologous to HSA21 genes, only 7 of them were deregulated. neurons. The expression of dual-specificity tyrosine-(Y)-phosphorylation regulated kinase 1a (Dyrk1a), a well-studied triplicated gene in DS individuals and mouse models, was not detected as modified in our study. This result is explained by the low level of expression of the gene in our nuclear RNA dataset, not reaching the minimum threshold to be included in the analyses. However, previous expression profiling studies involving the brain of different trisomic mouse models did also not find Dyrk1a differentially expressed [5]. All downregulated genes in our study were non-trisomic.
Among them, we found Grin2a, encoding the Glutamate Ionotropic Receptor NMDA Type Subunit 2A, to be dramatically reduced in the whole trisomic hippocampus regardless of cellular subtypes. Importantly, the DG was particularly affected. In this region, which is the only neurogenic niche in the adult mouse hippocampus, we found a transcriptional deregulation of genes related to neurogenesis. Although an alteration of neurogenesis has been proposed to be a neurobiological correlate of intellectual disability in DS [68], few studies have focused on this hippocampal subregion. Those have shown that cell proliferation from early postnatal stages is reduced in the subgranular zone of fetuses with DS and of Ts65Dn mice [3,69]. The number of differentiated neurons is also reduced in individuals with DS [70]. This defective neurogenesis derived from the trisomy continues into adulthood [71,72]. Here, we found a granule cell hypocellularity in TS mice, which is possibly contributed by the Strikingly, we also detected a marked reduction of the expression of Snhg11, a lncRNA involved in cell proliferation in different types of cancer [46-50] which was the main contributor to the transcriptomic shift. In recent years, there has been a growing interest on lncRNAs, which have been shown to act as essential epigenetic regulators [76] and have been functionally and mechanistically linked with neurobiological processes related with neuronal proliferation and differentiation [19,20] and with learning and memory [16][17][18], as well as with neurological disorders [77,78]. However, a conclusive link between lncRNAs and DS pathophysiology has not been reported so far. In our dataset, besides Snhg11, we found other 7 lncRNAs differentially expressed,  [79].
In addition, the small nucleolar RNAs (snoRNAs) hosted by these genes can also have an impact on cell function. Snhg11 harbours two snoRNAs that are highly conserved between human and mice: Gm25129 (ACA60 in humans) and Gm15187 (SNORA71 in humans). Both of them are H/ACA box snoRNAs, which are a class of snoRNAs involved in pseudouridylation of rRNA, a key step for ribosome biogenesis [80], which is essential for cell proliferation. In fact, the downregulation of both snoRNAs upon Snhg11 knockdown results in a small reduction of pseudouridylation levels at two ribosomal 18S positions, although these modest differences are unlikely to be responsible for explaining the phenotypes observed in this work. SnoRNAs are also increasingly regarded as key regulators of gene expression [81][82][83]. Furthermore, dynamic changes in snoRNA expression have been reported in different brain areas upon fear conditioning, indicating a possible role in learning and memory [84][85][86].
Therefore, the reduction of Snhg11 expression in the DG could have far-reaching implications for neuronal function and identity.
As expected from these observations, the in vivo knockdown of Snhg11 had an important functional impact in the WT DG. Previous reports indicated that Snhg11 is upregulated in various types of cancer [46-50], where it promotes cellular proliferation, mainly through the regulation of the Wnt/β-catenin signaling pathway, which is fundamental for adult neurogenesis [87] and is affected in DS [88]. In line with these results, we observed a significant decrease of the adult neurogenic capacity upon Snhg11 reduction. However, we did not find any significant deregulation of the In line with these results, the reduction of Snhg11 expression in the DG led to dramatic hippocampal-dependent memory deficits that recapitulate those observed in TS animals. Upon Snhg11 knockdown, we detected a strong impairment of both long-term and short-term recognition memory in Snhg11-ASO injected mice, that is similar to the one observed in TS animals, indicating that the hippocampal function is compromised in these animals. Importantly, we also observed a striking impairment of the pa ern separation function, which is unique to the DG [54,55] , showing that the DG is specifically affected. Although the specific mechanisms remain to be explored, the long and short-term memory deficits could be explained by the defective synaptic plasticity resulting from the reduction of Snhg11 expression. It has been shown that newly generated neurons are also important for long-term memories and particularly for pa ern separation [44, 93,94]. However, the link between the described phenotypes and the observed deficits in adult neurogenesis is uncertain, as the behavioral tests were conducted between 1 and 2 weeks after the Snhg11-ASO injection, while newborn granule cells take around 2 weeks to form stable functional synapses and be integrated in the hippocampal circuit [95]. Nevertheless, given the transcriptomic alteration of genes involved in neuronal differentiation and survival, it is also possible that the knockdown of Snhg11 interferes with the integration of neurons born previously to the injection of the ASO, thereby contributing to the DG-dependent memory deficits.
In conclusion, our results provide evidence of the zcell-type specificity of the  In the first three wild type and three trisomic hippocampi and in the second one hippocampus per genotype were processed. For each experiment, all libraries were pooled and sequenced on

Animals
NovaSeq 6000 S1 to an average depth of approximately 20.000 reads per cell.

10X data pre-processing
The resulting reads were aligned to the reference genome and converted to mRNA molecule counts using the Cellranger pipeline (CellRanger v3.0.1 [98]) provided by the manufacturer. For every nucleus, we quantified the number of genes for which at least one read was mapped, and then excluded all nuclei with fewer than 200 or more than 2500 detected genes, to discard low quality nuclei and duplets, respectively. Genes that were detected in fewer than six nuclei were excluded. Expression values Ei,j for gene i in cell j were calculated by dividing UMI counts for gene i by the sum of the UMI counts in nucleus j, to normalize for differences in coverage, and then multiplying by 10,000 to create TP10K (transcript per 10,000) values, and finally computing log2(TP10K + 1) using the NormalizeData function from the Seurat package v.2.3.4 [99].

Batch Correction and scaling data matrix
Since samples were processed in two different experiments, batch correction was done using Harmony [100] on the normalized dataset. The batch-corrected data was scaled using the ScaleData function from Seurat [99] with default parameters (v. Clustering was performed using the Seurat functions FindNeighbors and FindClusters (resolution = 0.6). Clusters were then visualized with UMAP. Reference anchors were identified between genotypes before integration with the IntegrateData function, and integrated data were then processed by the same methods.

Identification of marker genes of individual cell clusters
Cluster-specific marker genes were identified using the FindAllMarkers function from Seurat [99], utilizing a negative binomial distribution (DESeq2). A marker gene was defined as being >0.25 log-fold higher than the mean expression value in the other clusters, and with a detectable expression in > 20% of all cells from the corresponding cluster both in WT and TS groups. In this way, we were able to select markers that were highly expressed within each cluster, while still being restricted to genes unique to each individual cluster. We evaluated the enrichment of region-specific markers in each cluster by calculating the Jaccard overlap coefficient between cluster markers and genes enriched in each hippocampal subregion according to the Allen Brain Atlas [23, 101]. To find region-enriched genes, we used the "Differential Search" function in the Allen Brain Atlas web, se ing the region of interest as "Target Structure" and the whole Hippocampal Formation as the "Contrast Structure". We set an expression threshold of 2.5.

Identification of DEGs between WT and TS
All clusters identified as belonging to the same cell type were merged for differential gene expression analyses. Within each cell type, WT and TS samples were compared for differential gene expression using Seurat's FindMarkers function. To be included in the analysis, the gene had to be expressed in at least 10% of the cells from one of the two groups for that cell type and there had to be at least a 0.1 fold change in gene expression between genotypes. After correcting for multiple testing, only genes with a p adjusted value < 0.001 were considered for downstream analyses.

Gene set enrichment
The differential expression signatures from each cellular subtype were tested for enriched Gene Ontology processes, using a hypergeometric test (shinyGO [102]), and corrected for multiple hypotheses by false discovery rate (FDR). Processes with p adjusted value < 0.05 were reported as significantly enriched. The complete list of genes present in the dataset was used as the universe for the hypergeometric test.

Diffusion map
For the automatic identification of relevant genes from low-dimensional embeddings of

Cellular proportion analysis
To gain insight into the cell type alterations in the trisomic hippocampus, the relative proportion of the number of nuclei in each cell type was normalized to the total number of nuclei captured from each library. To determine if any changes in cell-type proportion were statistically significant, we implemented single cell differential composition analysis (scDC) [103] to bootstrap proportion estimates for our samples. We employed a linear mixed model (random effect of subject) to determine if any changes in cell-type proportion were present.

Quantitative assessment of global transcriptome shifts
The quantification was done as explained in [104]. Briefly, we generated two representative cells, one for the WT group and the other for the TS group. This was done by calculating the average gene expression of each gene for each genotype group within that cell type. We then calculate the Euclidean distance in gene expression between these representative cells as a metric to quantify the effect of the trisomy on each cell type. To determine if the observed Euclidean distance between WT and TS cells within each cell type was significantly larger than that of random cells, we estimated a null distribution by calculating the Euclidean distance between randomly sampled cells of the given cell type. This permutation approach was repeated for a total of 1000 times to generate the null distribution, which is compared to the Euclidean distance generated from the representative WT and TS cells to determine an empirical p value. To correct for multiple testing across all the cell types tested, we applied a Bonferroni correction to retrieve adjusted p values.

RNAscope
In situ hybridization was performed using RNAscope® Multiplex Fluorescent assay (V1) (Advanced Cell Diagnostics) probes and reagents. Naive animals were sacrificed and transcardially perfused with 50mL of chilled PBS followed by fixation with 50-100mL of 4% depolymerised paraformaldehyde (PFA). Following hippocampus dissection, tissues were placed in 4% PFA and post-fixed overnight at 4ºC. Tissues were then immersed in increasing concentrations of sucrose (10%, 20% and 30%) over three days, increasing the concentration every 24 hours. Once embedded in sucrose, tissues were dried and frozen at -80ºC prior to cryosectioning. 14 µm sections were obtained and mounted on Superfrost charged slides and stored at -80ºC until being processed. In situ hybridization was performed following the manufacturer's guidelines. Custom RNAscope® target specific oligonucleotide (ZZ) probes were designed by Advanced Cell Diagnostics targeting nucleotides 2-1604 for Snhg11 (NR_164123.1) and 1130 -2109 for Ntng1 (NM_030699.2). control and a no-RT negative control was included in each run. The thermal cycling was initiated at 95°C for 10 minutes followed by 45 cycles of 10 s at 95°C and 15 s at 55ºC, the optimal annealing temperature for our target genes. Melting curve analyses were carried out at the end of each run of qPCR to assess the production of single, specific products. ggaga g gccatcaacga, Gapdh reverse primer: tgaagacaccagtagactccacgac) were used.

ASO-mediated knockdown of Snhg11
Antisense oligonucleotides which included the same nucleotides in a scrambled order, did not generate any full match to identified gene sequences in the database.

In vitro knockdown of Snhg11
Cell culture and ASO transfection Mouse N2a neuroblastoma cells were maintained in high glucose Dulbecco's modified Eagle's medium (Thermo Fisher #11965084) supplemented with 10% FBS, 100 U/mL penicillin and 100ug/mL streptomycin in a humid atmosphere containing 95% air and 5% CO 2 at 37ºC.

Stereotactic injection
For in vivo injection of ASOs, the same wild type mice li ermates of Ts65Dn were used. Briefly 4 month old male and female mice (n =16 per experimental condition) were anesthetized by intraperitoneal injection of ketamine (75 mg/kg) and medetomidine (10 mg/kg) and placed on a stereotaxic frame. For each injection, a small incision was made, the skull was exposed, and a small burr hole was drilled at the proper coordinates (-2.2 mm anteroposterior, +/-1.3 mm mediolateral and −2 mm dorsoventral, relative to the bregma). Three minutes after the needle (Hamilton #65458-01) was placed into the proper coordinates, a total of 660nL of saline-diluted Snhg11-ASO or Control-ASO at a concentration 75 µg/µl was delivered into each hemisphere of the DG at an infusion rate of 50nL/minute using an injection pump. After the infusion was completed, the needle was kept in the same position to ensure no contamination of other brain areas with residual volume. The incision site was sutured using surgical glue (Cemave #1050052) and the mouse was allowed to recover in a temperature-controlled environment.  1KT). Single-stranded cDNA was synthesized from 1000 ng of total RNA using SuperScript™ III Reverse Transcriptase and oligo(dT) primers (Thermo Scientific #18080093) following the manufacturer's instructions. Quantitative Real-time PCR was performed as described above.

RNA-sequencing
The hippocampal RNA was also used for RNA-seq. Library preparation for mRNA sequencing was performed according to Illumina standard protocols (TruSeq, Illumina) from ribo-depleted total RNA. Libraries were quality-controlled and quantified using Nanodrop 2000 (Thermo Scientific), Agilent 2100 Bioanalyzer (Agilent Technologies) and Qubit (Life Technologies).
Samples were sequenced (paired end, 50 base pairs in length) using an Illumina HiSeq2500 apparatus with a depth of at least 50 million reads.

Gene set enrichment
The significantly downregulated genes in Snhg11-ASO injected animals were tested for enriched Gene Ontology processes, using a hypergeometric test (shinyGO [102]), and corrected for multiple hypotheses by FDR. Processes with p adjusted value < 0.05 were reported as significantly enriched. The complete list of genes present in the dataset was used as the universe for the hypergeometric test.

DRS data pre-processing
Raw fast5 files were processed with the Master of Pores [110] (version 2 of the pipeline, which is publicly available in GitHub (h ps://github.com/biocorecrg/MOP2). The mop_preprocess module was used to process the samples, using DeePlexiCon [111]

Memory tests
We and WT injected with Control-ASO (N=7 males and 5 females) were used. In the case of injected animals, they were let to recover for 5-7 days before being subjected to any task. Three days before testing mice were handled by the experimenter. All the behavioral experiments were conducted in 4 month old male and female mice during the light cycle (7:30 am to 13:00 pm).
All mice were individually handled and habituated to the investigator for five minutes, in three separate days. Handling took place in a different room where the behavioral apparatus was located. Before each handling session, mice were transported to the handling room by a wheeled trolley to habituate them to the journey.

Novel Object Recognition
The novel object recognition test (NOR), is a relatively fast and efficient means for testing different phases of learning and memory in mice. The test was conducted as explained in previously [114]. Briefly, the NOR protocol consists of four phases, namely habituation, familiarization, short-term test and long-term sessions. During the habituation session, mice were allowed to freely explore the empty open field for 10 minutes. 24 hours later, each mouse was returned to the arena containing two identical objects placed at symmetrical positions 5 cm from the arena wall and allowed to explore them freely for 15 minutes. During the familiarization session most mice reached a minimum exploration for each object of 30 seconds.
Mice not reaching this threshold, were excluded from the analysis. After a retention interval of 1 hour, the mouse was returned to the arena in which one of the objects was replaced by a novel object and let to explore both the familiar and the novel object for 5 minutes. Similarly, 24 hours later the animal was returned to the arena in which the novel object of the short-term session was replaced by a third object validated for achieving similar levels of exploration. The animal was then let to explore both objects for 5 minutes. All sessions were video-taped.

Object Pa ern Separation (OPS)
To test pa ern separation memory, the protocol described by van Hagen et al. (2015) [56][44], was followed with few modifications. The same square chamber used for NOR was used for OPS. Two days after completing the NOR, mice were habituated again to the open field without any objects for 10 minutes. 24 hours later, mice were presented with two identical objects placed in two symmetrical spots 5 cm from the arena wall. The objects were made with building blocks stuck to the arena floor so that mice were not able to move them. Animals were let 6 minutes to explore. The test session was performed 1 hour later, where one of the objects was placed in a novel location inside the arena, 8 cm away from the initial location. Mice were let 4 minutes to explore. All sessions were video-taped.

NOR and OPS behavioural experimental variables
Locomotor activity was quantified as the total distance traveled in the apparatus during the experimental sessions. Thigmotaxis refers to the disposition to remain close to the walls of the apparatus. It is measured as distance traveled or percentage of time spent in the periphery of the apparatus. It decreases gradually during the first minutes of exploration, and can be used as an index of anxiety [115]. Exploration time is defined as the action of pointing the nose toward an object, at a maximum distance of 2 cm or touching it. Going around the objects or si ing on the object is not evaluated as exploration time. As such, the exploration time is only computed when the snout of the animal is directed toward the object, sniffing or touching it.   were promediated to obtain a response every 30 seconds in order to reduce noise.

Statistical analyses of behavioral, histological and qPCR experiments
In the cases where more than two groups were compared, either a one or two-way ANOVA was conducted depending on the number of factors (Group and Sex) to investigate statistical differences in experimental variables. For pairwise comparisons, the Saphiro-Wilk and Bartle tests were used to assess normality of values and homogeneity of variances between groups, respectively. When the two conditions were met, pairwise differences were tested by Tukey's post hoc test.
When only two conditions were compared, statistical differences were tested by Student's t-test after confirming normality and homogeneity of variances between groups by Shapiro-Wilk test and Fisher's F test, respectively.
In the cases where distribution of the data could not be estimated due to the small sample size, statistical differences were assessed by a permutation test.
All statistical analyses were two-tailed. P-values were considered to be significant when α < 0.05.

Data availability
The snRNA-seq and bulk RNA-seq data that support the findings of this study have been deposited in the Gene Expression Omnibus repository, with the series record GSE212351 and GSE212258. Basecalled FAST5 of nanopore direct RNA sequencing runs have been deposited to ENA, under accession code PRJEB58921. All the raw data and supplementary materials are available upon request.

Code availability
The software used in this study are available at the following online repositories. R

Competing interests:
The authors declare no competing interests.