Presented here is a method to sequence single nuclei isolated from the mouse dentate gyrus that excludes most neurons through fluorescence-activated nuclei (FAN)-sorting. This approach generates high-quality expression profiles and facilitates the study of most other cell types represented in the niche, including scarce populations such as neural stem cells.
Adult Hippocampal Neurogenesis (AHN), which consists of a lifelong maintenance of proliferative and quiescent neural stem cells (NSCs) within the sub-granular zone (SGZ) of the dentate gyrus (DG) and their differentiation from newly born neurons into granule cells in the granule cell layer, is well validated across numerous studies. Using genetically modified animals, particularly rodents, is a valuable tool to investigate signaling pathways regulating AHN and to study the role of each cell type that compose the hippocampal neurogenic niche. To address the latter, methods combining single nuclei isolation with next generation sequencing have had a significant impact in the field of AHN to identify gene signatures for each cell population. Further refinement of these techniques is however needed to phenotypically profile rarer cell populations within the DG. Here, we present a method that utilizes Fluorescence Activated Nuclei Sorting (FANS) to exclude most neuronal populations from a single nuclei suspension isolated from freshly dissected DG, by selecting unstained nuclei for the NeuN antigen, in order to perform single nuclei RNA sequencing (snRNA-seq). This method is a potential steppingstone to further investigate intercellular regulation of the AHN and to uncover novel cellular markers and mechanisms across species.
The continuous generation of hippocampal neurons in adulthood, also known as Adult Hippocampal Neurogenesis (AHN), is associated with cognitive functions such as learning, memory acquisition/clearance, and pattern separation and appears to be an important mechanism of resilience in aging and neurodegenerative diseases to prevent cognitive deficits1,2,3. Rodents have been the model of choice to study AHN using several methods, including immunocytochemistry and next-generation sequencing (NGS) methods. The translation of these results to other species remains controversial. Indeed, AHN has been observed in most species but the extent to which it persists throughout life, particularly in humans4,5,6,7,8, is regularly debated.
To date, various intrinsic and extrinsic signaling pathways have been confirmed to modulate AHN1. However, the impact of intercellular communication on AHN is only just emerging9. This could first be attributed to insufficient specificity of the currently known cell markers to conduct in vivo analysis with genetically modified animals. Indeed, many studies have relied on markers such as doublecortin or glial fibrillary acidic protein (GFAP) that are expressed in multiple cell types1. Second, the complexity and the high degree of cell diversity in the adult hippocampal niche10 brings technical challenges to profile every cell type. This is particularly the case for bioinformatic analysis with overlapping cellular markers used in analytic pipelines for different populations, such as NSCs or glial cells, resulting in controversial conclusions when assessing AHN7,11. Third, the vast number of neurons undermine the investigation of less abundant cell populations, such as astrocytes, oligodendrocytes or ependymal cells, even though their role in the fine-tuning regulation of AHN is becoming prominent9. Together, these limitations impact the ability to translate results from rodents to other species. This is particularly amplified by the difficulty to recapitulate in vitro a complex tissue, such as the hippocampal neurogenic niche, and by the many hurdles to access high quality tissue together with a lack of standardized protocols for tissue processing in studies involving human tissues12,13. It is therefore critical to develop new approaches to profile cell populations and identify new cellular markers within the dentate gyrus (DG) that ultimately will lead to a better understanding of the different contributions of each cell type to AHN regulation.
To achieve this, single cell (sc) and single nuclei (sn) isolation combined with RNA sequencing has become instrumental to investigate complex tissues such as the DG14. As such, strategies of cellular enrichment to isolate single cells from the mouse adult hippocampal niche have been performed mostly to examine NSCs15,16. An interesting strategy to enrich non-neuronal cells from the DG was applied by sequencing GluR1/Cd24 double-negative single cells that resulted in 1,408 cells being sequenced without distinct clusters between astrocytes and NSCs after bioinformatic analysis17. This could be due to the harsh enzymatic digestion required for single cell preparation that damages cell integrity and RNA. To bypass this technical issue, several methods using single nuclei isolation instead have been developed and are particularly suited to intricate tissues11,18. However, the predominance of neurons within the DG or more broadly within the hippocampal-entorhinal system generates a sampling bias to study the entirety of the cell populations present within these brain areas. In addition, the limited number of cells to load for the preparation of single cell libraries accentuates the presence of the major cell population in analytic pipelines of sequenced single nuclei. Indeed, large neuronal clusters are often annotated and analyzed while other cell populations are being underrepresented or missed5,11.
In an attempt to overcome these biases and to be able to profile cell types other than neurons present in the mouse DG, a method was devised in this study using the principle of Fluorescence Activated Nuclei Sorting (FANS)18 that excludes most neuronal populations by negative selection of stained single nuclei with neuronal nuclear antigen (NeuN, also known as Rbfox3). This choice of antigen was guided by the literature describing NeuN as a reliable neuronal marker19 and by the necessity to use a nuclear protein for this approach. NeuN-negative FACS-sorted cells were then prepared for RNA sequencing on a 10x Genomics platform. The results demonstrate that exclusion of NeuN-expressing cells allows a cell-type specific, high-quality transcriptomic profiling of glial and rare cell populations.
Animal care and experimental procedures were performed in accordance with the guidelines of the Francis Crick Institute, as well as the UK Home Office guidelines and laws.
Figure 1: Preparation of a single nuclei suspension from the dissected DG of adult mice for snRNA-seq of non-neuronal populations. Flow diagram describing the main steps of the protocol that include dissection of mouse DG, preparation of a suspension of single nuclei, NeuN immunostaining, and negative NeuN-FANS-sorting before proceeding with snRNA-seq. Please click here to view a larger version of this figure.
1. Dissection of the DG (Timing: 15 min)
2. Tissue dissociation, single nuclei isolation, and anti-NeuN immunostaining (Timing: 2 h)
3. Fluorescence activated nuclei sorting (FANS) to exclude neuronal populations (Timing: 45 min)
Figure 2: Isolation and transcriptomic profiling of non-neuronal cell populations from the DG. (A–C) Gating strategy to isolate NeuN-AF488 negative single nuclei and exclude cell debris. (A) FANS dot plot of a representative sample of isolated nuclei, depicting the gate setting for the selection of DAPI+ nuclei and exclusion of cell debris and aggregates. (B) Further selection of relevant single nuclei using FSC-area and SSC-area. (C) The gates for NeuN-AF488 to exclude the positive population and sort for the negative single nuclei. (D) Micrograph of a good single nuclei suspension with minimum amount of debris and higher proportion of good quality nuclei (round shape, black arrow) compared to bad quality nuclei (white arrow). Scale bars = 50 µm, 10 µm (inset). (E,F) Analysis of snRNA-seq data and profiling of the distinct cell populations isolated from the DG of 22-month-old C57BL/6J male mice. Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) plots of single nuclei profiles from the (E) non-FACS-sorted cells and (F) NeuN-negative FACS-sorted cells, colored by cell type. (G) Pie charts comparing the frequencies of identified cell types in both samples. (H) Respective metrics for the sequenced samples: number of nuclei, median number of genes, and transcripts per nucleus. (I) Violin plots showing the distribution of the number of genes and transcripts detected for each cell type in both the samples. Astr. = astrocytes, Olig. = oligodendrocytes, Vasc. = vascular cells, CRCs = Cajal-Retzius cells, Neur. = neurons, Imm. = immune cells, OPCs = oligodendrocytes precursor cells. Please click here to view a larger version of this figure.
4. Preparation of the single nuclei suspension to perform single nuclei RNA sequencing (Timing: 30 min)
5. Library preparation and sequencing
NOTE: Description of the following steps is based on the in-house sequencing platform used in this study (see Table of Materials). Therefore, some settings might differ when using a different platform. Here, only the key steps are described and each parameter should be determined following guidance and protocols from the chosen manufacturer albeit with optimization before first use. It is critical to ensure that preparation of libraries is performed as quickly as possible after concentrating sorted nuclei suspensions to avoid RNA degradation and ensure optimal quality of the sequencing.
The protocol presented here describes a method to prepare a suspension of non-neuronal single nuclei isolated from the DG to perform snRNA-seq. With or without FANS, bioinformatic clustering revealed well separated groups of nuclei corresponding to known cell types within the DG (Figure 2E,F). Within the non-FACS-sorted sample, the majority of the high-quality nuclei that were sequenced comprised three groups of neurons (84.9% of total nuclei for this sample, Figure 2E,G,H). Such results are expected, considering that the most represented cell populations in the DG are granule neurons, other excitatory neurons (labeled excitatory neurons), and inhibitory neurons10. The non-neuronal clusters identified were mostly made of glial cell types (11.1%), including astrocytes, oligodendrocytes and oligodendrocyte precursor cells (OPCs), immune cells (3.3%), and Cajal-Retzius cells (0.6%). When performing FANS to exclude NeuN positive populations (NeuN-negative FACS-sorted sample; Figure 2F,G,H), clusters of glial cells became predominant (81.3%). The isolation of a greater number of glial nuclei permits a better segmentation of different populations that would cluster together without FANS. Indeed, on re-clustering and analyzing specific genes either expressed in NSCs or in astrocytes, four sub-clusters separated out (Supplemental Figure 2A,B). Looking at more specific cellular markers and assessing gene expression levels across the cell-types, a small cluster of NSCs was detected segregating separately from the main astrocytic populations with higher expression of Hopx and Notch2 and almost no expression of Aldh1a1 or Aqp4 (Supplemental Figure 2C). However, because of the overlap in gene expression between astrocytes and NSCs, further analysis would be required to specifically profile and identify different sub-types of cells. Moreover, the NeuN-negative FANS sample had additional clusters labeled as vascular cells (2.3%) that encompass endothelial cells, pericytes, and vascular leptomeningeal cells when cross-referenced for expression of cell specific markers (data not shown).
Following the guidance for the chosen protocol to generate libraries for sequencing, high-quality expression profiles were obtained with or without FANS. For samples sequenced at 50,000 reads/nucleus, 2,510 genes were detected on average per nuclei for the non-FACS-sorted sample (5,578 transcripts, Figure 2H) and 1,665.5 genes (3,508 transcripts) for NeuN-negative FANS sample, after filtering out low-quality nuclei (Figure 2H,I). These metrics confirm that this protocol generates high-quality transcriptomic profiling of single nuclei comparable to studies using different approaches22,23 and that the process of FACS sorting does not damage nuclei for subsequent snRNA-seq. Notably, the difference in the number of genes and transcripts per nuclei between the two samples is not due to lower data quality but to the high proportion of neurons in the non-FACS-sorted sample (84.9% compared to 1.7% in NeuN-negative FANS sample), which have a higher transcriptional activity (2,660 genes/nucleus and 6,170 transcripts/nucleus in non-FACS-sorted sample) than the average transcriptional activity of all the non-neuronal cell types (1,090 genes/nucleus and 1,785 transcripts/nucleus, Figure 2I).
Together, these representative results show that the selection of NeuN negative nuclei using FANS is a powerful tool to isolate low abundance cell types from freshly dissected brain tissue and perform high-quality single nuclei transcriptomic profiling of these distinct cell populations via snRNA-seq methods.
Supplemental Figure 1: Validation of the immunostaining for FANS. Nuclei suspension was incubated (A) without the anti-NeuN-AF 488 antibody as a negative control or (B) with the antibody and run through the FACS sorter to validate immunostaining conditions. Please click here to download this File.
Supplemental Figure 2: Gene expression analysis and re-clustering of the astrocyte cluster. (A) Uniform Manifold Approximation and Projection for Dimension Reduction (UMAP) plot showing the clustering of 4968 nuclei based on genome-wide expression profiles from Figure 2F. Cell-type calls were done based on cell-type markers. (B) Astrocyte cluster comprised of 2579 nuclei chosen from (A) for further sub-setting to investigate potential cellular sub-types. Four sub-types were detected by Seurat (0-3) clustering, shown by different colors. (C) Gene expression levels of specific cellular markers across the four cell-types. All plots were obtained using the Seurat R package24. Briefly,RNA-seq counts were normalized for each cell by the total expression and multiplied by the scale factor (10,000). This result was then log transformed. The transformed values were scaled (variance scaled to one) and centered (mean set to zero) within each cell before UMAP was applied to calculate the embeddings, which were used as values on x and y axes. Graphs represent the output of a dimensional reduction technique on a 2D scatter plot where each point represents a cell with respective x and y co-ordinates based on the cell embeddings determined by the reduction technique. Cells with similar gene signatures are positioned close to one another by the embeddings. Please click here to download this File.
Supplemental Figure 3: Gene expression analysis of NeuN in the neurogenic lineage. (A) UMAP plot showing the clustering of neurogenic lineage from publicly available dataset15. UMAPs were generated as in Supplemental Figure 2. (B) Gene expression levels of specific cellular markers across the neurogenic lineage showing Astrocyte (Aquaporin 4 = Aqp4), NSCs (Homeodomain-only protein = Hopx), NeuN/Rbfox3 (NSCs and intermediate progenitor cells [IPCs]),and cycling cells (Cyclin-dependent kinase 6 =Cdk6). Please click here to download this File.
Supplemental Table 1: Compositions of media and buffers used in the study. Please click here to download this File.
To successfully execute this protocol, the dissection of the DG is the first critical step, which requires some practice to keep it undamaged and to limit contamination from the surrounding tissues. From experience, separation of the DG from the hippocampus could be acquired very quickly by a skilled researcher who could then work on refining their technique to increase the rapidity of dissection and therefore improve the freshness of the tissue to generate high quality data. In a similar vein, preparation and resuspension of single nuclei demands consistency across the different conditions used in a single experiment, but also avoidance of excessive pipetting that could disrupt nuclear membrane releasing ambient RNAs that will bias the sequencing results. In addition to recommendations mentioned previously to prepare high quality nuclei, the concentration of the single nuclei suspension is also to be considered before proceeding with sequencing. Indeed, according to the manufacturer's guidelines, a preparation with a concentration higher than 1,200 nuc/µL should be diluted, as this level of nuclei concentration will have higher risk of forming multiplets impacting downstream bioinformatic analyses. Of note, sequencing samples with nuclei concentrations under 500 nuc/µL might not be worthwhile due to the cost involved. It is also recommended to follow the advice of an advanced FACS user to set up all the gating and to remain consistent with the settings across samples and biological replicates. Likewise, preparation of libraries for RNA sequencing entails some training to achieve high quality results and most vendors have excellent support to achieve this efficiently. This method was only tested with fresh tissue in this study; however, FANS has also been performed with frozen tissue25. It is therefore reasonable to assume that this protocol could be performed with frozen tissue albeit with minor optimization.
This protocol has been developed with a particular downstream application in mind, which is to investigate cell populations other than neurons within the hippocampal neurogenic niche. Indeed, increasing lines of evidence indicate that impairment of AHN in ageing could be attributed to the surrounding cells within the niche1,2,3,9. In particular, astrocytes and oligodendrocytes emerge as key regulators of AHN; however, their isolation from the DG coupled with RNA-sequencing has generated mixed results, making this hypothesis challenging to assess with this technique1,17. This approach of FACS sorting NeuN-negative nuclei allowed the isolation of more astrocytes and oligodendrocytes compared to samples that were not FACS-sorted, which enables better bioinformatic analysis. This protocol is applicable at all ages across lifespan and the representative data presented here with tissues from old animals provides a proof of concept that this method is robust to investigate the ageing hippocampal neurogenic niche. To expand the use of this method and to adapt it for different biological questions, it is important to consider that other neuronal nuclear membrane antigens could be tested together with a thorough titration of the best validated antibodies for these markers. For instance, when studying the process of neuronal differentiation from NSCs in the DG, some cell types such as type 2 cells or neuroblasts start expressing NeuN (Supplemental Figure 3). Therefore, another antigen would be needed to specifically investigate these cell types. Conversely, some neurons were still identified in this study after NeuN-negative FACS-sorting possibly due to low or no expression of NeuN in these populations (e.g., Cortical Cajal-Retzius neurons19). Additionally, NeuN has been reported to be expressed in sub-populations of oligodendrocytes26, which could give biased results if these sub-populations were of interest. Thus, the choice of antigen when starting to use FANS should be carefully considered to avoid inclusion or exclusion of cell populations that would preclude an accurate answer to a specific biological question. In agreement with this, it is also recommended that each sequencing result is further validated by orthogonal assays (e.g., immunohistochemistry or RNA-scope) before validating or refuting the tested hypothesis with this protocol. Finally, the step involving FANS could be further developed to include more than one antibody with a more elaborated sorting strategy to exclude and/or include desired cell populations.
Ultimately, technologies described in this protocol could have some limitations when used with other species. For instance, the niche is very well defined in rodents with the presence of proliferative and quiescent NSCs or newly-born neurons restricted within specific sub-regions of the DG, but it is still not clear how the hippocampal neurogenic niche should be delineated in other species. Indeed, proliferative cells are not aligned within a continuous zone of the DG in non-human primates and humans but are rather scattered around it and might also be present in the amygdala7. Therefore, dissecting and isolating broader areas than the DG in other species would potentially impact the use of this protocol. Particularly, the dissociation and trituration steps for the preparation of tissue will need to be optimized while working with larger pieces of tissue27,28. Regarding bioinformatic analysis, while inbreed housed rodents have a very homogeneous and very well annotated genome, the genetic variability of the human genome combined with insufficient numbers of cellular markers to clearly distinguish different cell populations (e.g., NSCs and astrocytes) requires a lot of normalization for analysis that could lead to different conclusions when a small cluster of cells is identified7,11. In such situations, cell enrichment might still be a preferred option or should be used alongside other strategies to increase analytical power.
Nonetheless, the current approach can enable investigation of the role of understudied albeit potentially important cell populations in the regulation of AHN. This could particularly be the case for populations of astrocytes, which play a central role in the onset and progression of neurodegenerative diseases29,30. This study demonstrated that astrocytes and other rare cell populations can be identified and profiled simply by excluding the vast majority of neurons present within the DG. Other studies using different approaches have not been able to achieve similar recovery of nuclei from the same range of cell populations5,11,17. Moreover, the results from this study demonstrate that it is possible using this approach to isolate a NSC cluster without specific enrichment of this cell population15.
In conclusion, following and improving this method would be a step forward to address outstanding questions related to the contextual role of the hippocampal neurogenic niche for the modulation of AHN. In particular, it could bring new insights into gene expression levels in aged and diseased brains in cell populations associated with the regulation of AHN9, support the identification of a potential heterogeneity of the NSCs1 or address the role of the vasculature in AHN. Ultimately, this method could be adapted for other adult stem cell niches with similar questions and issues.
The authors have nothing to disclose.
The authors would like to thank Lachlan Harris and Piero Rigo for technical support and Jason M. Uslaner and Ditte Lovatt for providing feedback on the manuscript. This work was supported by grant support from the MRC and a pre-competitive research collaboration with MSD, the Francis Crick Institute, which receives its funding from Cancer Research UK (FC0010089), the UK Medical Research Council (FC0010089), the Wellcome Trust (FC0010089) and by a Wellcome Trust Investigator Award to FG (106187/Z/14/Z). We apologize to the many authors whose work we could not discuss and cite because of lack of space.
0.5ml microtube | Eppendorf | 30124537 | |
10.00µm Flouresbrite YG Carboxylate Microspheres | Polysciences | 15700-10 | |
15 mL polypropylene centrifuge tubes | Corning | 430052 | |
2 pairs of sterile Dumont #5 forceps | Fine Science Tools | 11252-30 | |
4′,6-diamidino-2-phenylindole (DAPI) | Sigma Aldrich | D9564-10MG | |
4150 TapeStation System | Agilent | N/A | |
5 mL round bottom high clarity polypropylene test tube with snap cap | Falcon | 352063 | |
5 mL round bottom polystyrene test tube with cell strainer snap cap | Falcon | 352235 | |
50 mL polypropylene centrifuge tubes | Corning | 430829 | |
70 µm cell strainer | Falcon | 352350 | |
8 peak SPHERO Rainbow Calibration Particles | BD Biosciences | RCP-30-5A | |
Accudrop Beads | BD Biosciences | N/A | |
Allegra X-30R Centrifuge | Beckman Coulter | N/A | |
Anti-NeuN antibody, clone A60, Alexa Fluor 488 conjugated | Millipore | MAB377X | |
BD FACSAria Fusion Flow Cytometer | BD Biosciences | N/A | |
Beckman Coulter MoFlo XDP | Beckman Coulter | N/A | |
Chromium Controller | 10x Genomics | N/A | |
Chromium Next GEM Single Cell 3' Reagent Kits v3.1 | 10x Genomics | PN-1000121; PN-1000120; PN-1000213 | |
BSA 7.5% | Gibco | 15260037 | |
Dithiothreitol (DTT) | Thermo Scientific | R0861 | |
Dounce tissue grinder set: mortar, loose pestle (A) and tight pestle (B) | KIMBLE | D8938-1SET | |
Eppendorf Tubes Protein LoBind 1.5ml | Eppendorf | 30108116 | |
Halt, 100x Protease inhibitor | ThermoFisher | 78429 | |
HiSeq 4000 Sequencing System | Illumina | N/A | Sequencing configuration: 28-8-0-91 |
KCl | Any chemical supplier | Laboratory made | |
LUNA-FX7 Automated Cell counter | Logos Biosystems | N/A | |
MgCl2 | Any chemical supplier | Laboratory made | |
N°10 guarded sterile disposable scalpels | Swann-Morton | 6601 | |
Nuclease-free water | Sigma Aldrich | W4502-1L | |
Pair of sterile student surgical scissors | Fine Science Tools | 91401-12 | |
PBS | Any chemical supplier | Laboratory made | |
RNase Inhibitor 40 U µl-1 | Ambion | AM2684 | |
RNasin 40 U µl-1 | Promega | N211A | |
Sterile Petri dish | Corning | 430167 | |
Sucrose | Sigma Aldrich | 59378-500G | |
Tris buffer, pH 8.0 | Any chemical supplier | Laboratory made | |
Triton X-100 10% (v/v) | Sigma Aldrich | T8787-250ML | |
Trypan blue | Invitrogen | T10282 |