Volume 247, Issue 1 p. 194-200
Techniques
Free Access

A morphology independent approach for identifying dividing adult neural stem cells in the mouse hippocampus

Lachlan Harris

Lachlan Harris

The School of Biomedical Sciences, The University of Queensland, Brisbane, Queensland, Australia

Search for more papers by this author
Oressia Zalucki

Oressia Zalucki

The School of Biomedical Sciences, The University of Queensland, Brisbane, Queensland, Australia

Search for more papers by this author
Sabrina Oishi

Sabrina Oishi

The School of Biomedical Sciences, The University of Queensland, Brisbane, Queensland, Australia

Search for more papers by this author
Thomas H. Burne

Thomas H. Burne

Queensland Brain Institute, The University of Queensland, Brisbane, Australia

Queensland Centre for Mental Health Research, The Park Centre for Mental Health, Wacol, Queensland, Australia

Search for more papers by this author
Dhanisha J. Jhaveri

Dhanisha J. Jhaveri

Queensland Brain Institute, The University of Queensland, Brisbane, Australia

Mater Research Institute, The University of Queensland, Queensland, Australia

Search for more papers by this author
Michael Piper

Corresponding Author

Michael Piper

The School of Biomedical Sciences, The University of Queensland, Brisbane, Queensland, Australia

Queensland Brain Institute, The University of Queensland, Brisbane, Australia

Correspondence to: Michael Piper, The School of Biomedical Sciences, The University of Queensland, Brisbane, 4072, Australia. E-mail: [email protected]Search for more papers by this author
First published: 07 July 2017
Citations: 6

Abstract

Background: Type 1 adult hippocampal neural stem cells (AH-NSCs) continue to generate neurons throughout life, albeit at a very low rate. The relative quiescence of this population of cells has led to many studies investigating factors that may increase their division. Current methods of identifying dividing AH-NSCs in vivo require the identification and tracing of radial processes back to nuclei within the subgranular zone. However, caveats to this approach include the time-intensive nature of identifying AH-NSCs with such a process, as well as the fact that this approach ignores the relatively more active population of horizontally oriented AH-NSCs that also reside in the subgranular zone. Results: Here we describe, and then verify using Hes5::GFP mice, that labeling for the cell cycle marker Ki67 and selection against the intermediate progenitor cell marker TBR2 (Ki67+ve; TBR2−ve nuclei) is sufficient to identify dividing horizontally and radially oriented AH-NSCs in the adult mouse hippocampus. Conclusions: These findings provide a simple and accurate way to quantify dividing AH-NSCs in vivo using a morphology-independent approach that will facilitate studies into neurogenesis within the hippocampal stem cell niche of the adult brain. Developmental Dynamics 247:194–200, 2018. © 2017 Wiley Periodicals, Inc.

Introduction

The continued generation of dentate granule neurons from adult hippocampal neural stem cells (AH-NSCs) is important for multiple cognitive processes, including learning (Akers et al., 2014; Goncalves et al., 2016) and mood regulation in mice (Yun et al., 2016). The level of adult hippocampal neurogenesis in humans is comparable to that in mice, suggesting that ongoing neurogenesis is functionally important for both species (Spalding et al., 2013; Bergmann et al., 2015; Ernst and Frisen, 2015). As a consequence of this, therapies are being developed that aim to specifically increase the levels of hippocampal neurogenesis to circumvent disorders such as depression, where neurogenesis is reduced (Harris et al., 2016b; Yun et al., 2016).

There are three major cell types within the neurogenic lineage of the adult mouse hippocampus (reviewed in Goncalves et al., 2016). These are the AH-NSCs (Type 1 cells), most of which are quiescent; the transient but highly proliferative intermediate progenitors (IPs or Type 2 cells); and, finally, neuroblasts (Type 3 cells). These cell types are defined by their relative division capacity, lineage potential, and expression of certain markers (von Bohlen and Halbach, 2011). Despite the intense interest in neurogenesis in the adult brain, one of the major challenges in the field is to clearly define these cellular populations; particularly challenging in this respect is unequivocally identifying dividing AH-NSCs in vivo. The ability to accurately identify dividing AH-NSCs is crucial to determine the relative rates of quiescence or division among this population. A recent study using a novel and purified to homogeneity almost the entire population of neurosphere-forming precursors comprising both dividing and quiescent AH-NSCs (Jhaveri et al., 2015); however, this approach is not easily applied to the histological identification of these cells. In most studies, the histological criteria used to identify dividing AH-NSCs in vivo involve linking a radial process—stained for glial fibrillary acidic protein (GFAP), nestin, brain lipid-binding protein (BLBP), or epidermal growth factor receptor (EGFR) expression—back to a nucleus within the subgranular zone (SGZ) and determining whether the nucleus is positive for a cell cycle marker, such as Ki67 (von Bohlen and Halbach, 2011; Hussaini et al., 2013; Jhaveri et al., 2015). Despite the large number of studies to have utilized this approach, there are a number of inherent difficulties evident with this paradigm for identifying dividing AH-NSCs.

Foremost among the problems associated with tracing a radial process is that the current approach ignores a population of horizontally oriented cells that are putative NSCs (Lugert et al., 2010). These putative horizontal NSCs (not to be confused with IPs, which are also horizontal dividing precursors) have different characteristics from their radial NSC counterparts. For example, horizontal AH-NSCs divide more frequently, and their numbers change more acutely in response to age or exercise (Lugert et al., 2010). This shortcoming implies that the many studies that have not analyzed this cellular population may have underrepresented the number of dividing AH-NSCs present. Although further work, such as lineage tracing, needs to be performed to confirm these horizontal cells as bona fide NSCs with the capacity to generate neurons and/or glia, an inclusive approach that accounts for the potential morphological heterogeneity of AH-NSCs should be developed. The second problem with the existing radial identification approach is that it is potentially inaccurate and extremely time-consuming. Dividing AH-NSCs and IPs are frequently found together in clusters (Hodge et al., 2008), and it therefore becomes difficult to determine to which cell a radial AH-NSC process is linked. Finally, there is a lack of standardization as to which radial markers are appropriate to use when identifying dividing AH-NSCs, demonstrated by the variety employed in many studies (Gulbins et al., 2013; Li et al., 2013; Martynoga et al., 2013; Andersen et al., 2014; Kandasamy et al., 2014; Andreu et al., 2015; Nicola et al., 2015; Yousef et al., 2015). Together, these issues pose great challenges to the interpretation and comparison of the many studies in this field.

To address these issues, we developed a morphology-independent approach for identifying dividing AH-NSCs. Our method is based solely on the exclusion of the IP marker TBR2 (Hodge et al., 2008; Hodge et al., 2012). We found that Ki67+ve; TBR2−ve cells have large nuclei, express higher levels of SOX2, and are Hes5::GFP+ve. Most important, we verified in the notch reporter Hes5::GFP mouse strain (Jhaveri et al., 2010; Lugert et al., 2010) that a subset of these cells also have a horizontal GFP+ve morphology, consistent with the description of horizontal NSCs by Lugert and colleagues (2010). The approach outlined here is accurate and fast and accounts for the heterogeneous morphology of AH-NSCs, representing a significant advance in previous approaches used to identify and quantify this population in vivo.

Results

Selecting a Nuclear Marker for Dividing AH-NSCs

To develop a morphology-independent approach for identifying dividing AH-NSCs, we first needed to identify a suitable nuclear protein to use as a marker. As no protein has been identified that alone specifically labels this population, we sought to identify a proxy marker, which could then be used in conjunction with other proteins to identify AH-NSCs. Two criteria were assessed: First, the percentage of dividing AH-NSCs that are labeled by the nuclear protein (penetrance), and second, the other cell types that express the protein (specificity).

Two proteins satisfied these criteria and appeared to be reasonable candidates, namely Achaete-scute like 1 (ASCL1) and Ki67. The basic helix-loop-helix protein ASCL1 is a proneural factor required for AH-NSCs to enter the cell cycle (Andersen et al., 2014; Urban et al., 2016). Because ASCL1 protein is not expressed by quiescent AH-NSCs and labels only a fraction of IPs (Kim et al., 2011), it displays good specificity. However, ASCL1 is not a completely penetrant marker, as it expressed by only by one-third of dividing AH-NSCs (Andersen et al., 2014). Moreover, ASCL1 also has functions in fate specification and differentiation in the embryonic nervous system (Castro et al., 2011; Guillemot and Hassan, 2016), and when ectopically expressed in the adult hippocampus, it leads to aberrant formation of oligodendrocytes (Braun et al., 2015). Therefore, in mouse models where fate specification is altered, it may no longer serve as a reliable readout of AH-NSC division.

In contrast, Ki67 appears to be a more suitable choice for a proxy marker of AH-NSC division. For example, Ki67 is expressed throughout the cell cycle and therefore should label all dividing AH-NSCs (high penetrance) (Gerdes et al. 1983; Endl and Gerdes 2000). In the SGZ, it is expressed only by dividing AH-NSCs, dividing IPs, and dividing neuroblasts, and thereby also displays reasonable specificity (Scholzen and Gerdes, 2000). Moreover, because the function of Ki67 is restricted to chromatin organization during cell division (Cuylen et al., 2016), it serves as a direct readout of division, which is unlikely to be altered in mouse models, where fate specification is altered. Consequently, we chose to investigate whether Ki67, in combination with other markers, could be used to specifically identify dividing AH-NSCs in the hippocampus.

Negative Selection Against TBR2 is Sufficient to Exclude Dividing IPCs and Neuroblasts

To distinguish between dividing nuclei (Ki67+ve) that are AH-NSCs and those that are dividing IPs or neuroblasts, we co-stained 12- to 16-week-old C57BL/6J mice with Ki67, the IP marker TBR2, and the neuroblast marker DCX (Fig. 1A,B). We hypothesized that the exclusion of TBR2+ve and DCX+ve cells would leave a population of Ki67+ve; TBR2−ve; DCX−ve cells within the SGZ that would comprise the dividing AH-NSC pool. To assess the plausibility of this experimental design, we first examined the percentage of IPs and neuroblasts that expressed Ki67. We focused our analysis to the SGZ specifically (see Experimental Procedures). Consistent with previous reports, we found that the majority of TBR2+ve cells were labeled with Ki67 (83.1%, Fig. 1C,D) (Hodge et al., 2008), while only 13.9% of DCX+ve cells were labeled with Ki67 (Fig. 1E,F). At this stage, we also made an important observation: All dividing DCX+ve neuroblasts had an immature IP-like morphology and expressed TBR2 (Fig. 1F). Because all dividing neuroblasts also expressed TBR2, this led us to posit that selection against TBR2 alone should be sufficient to identify dividing AH-NSCs.

Details are in the caption following the image

Negative selection against TBR2 is sufficient to exclude dividing IPs and neuroblasts.

A: Coronal section of an adult mouse brain at the level of the hippocampus stained for DAPI (white). B: A schematic showing the dividing (Ki67+ve) cell types in the adult hippocampus. C,E: The same coronal section as in A showing channels for (in C) TBR2 (blue), Ki67 (magenta), or (in E) TBR2, Ki67 and DCX (green). Boxed regions in C and E are presented in C′ and E′, respectively. D: The majority of TBR2+ve cells (arrows in C′) co-labeled with Ki67. Surprisingly, all dividing DCX+ve neuroblasts also expressed TBR2 (arrowheads in E′), as quantified in F. Scale bars A,C,E = 168 µm. Scale bars C′,E′ = 22.5 µm.

Ki67+ve; TBR2−ve Cells are Dividing AH-NSCs

To assess whether the dividing TBR2−ve cells were AH-NSCs, we examined three parameters. First, we measured the relative expression of SOX2. In AH-NSCs, the transcription factor SOX2 is down-regulated as these cells differentiate, so that AH-NSCs express relatively higher levels of SOX2 than IPs (Hodge et al., 2012; Shin et al., 2015). We stained adult mouse hippocampal sections for Ki67, TBR2, and SOX2. Supporting their identity as AH-NSCs, dividing TBR2−ve cells had a much higher expression of SOX2 than dividing TBR2+ve cells (Fig. 2A,B; P < 0.001). It is important to note that SOX2 expression may not be confined to AH-NSCs within the SGZ. To address this point, we co-labeled sections with SOX2 and S100ß, a marker for astrocytes, or with the oligodendrocyte marker OLIG2. Quantification of co-labeled cells located within the SGZ revealed that fewer than 10% of the SOX2+ve cells in the SGZ niche were S100ß-expressing cells of the astrocyte lineage, whereas fewer than 3% of the SOX2+ve cells were immunopositive for OLIG2. As such, the majority of cells expressing high levels of SOX2 within the SGZ are likely to be AH-NSCs. We also measured nuclear size. Consistent with previous observations that NSCs are comparatively large (Rietze et al., 2001), the average area of the nucleus of dividing TBR2−ve cells was approximately two-fold larger than Ki67+ve; TBR2+ve cells (Fig. 2A,C; P < 0.001). Finally, if dividing TBR2−ve cells are AH-NSCs, then we would expect some of these cells, except for those that were horizontally oriented or out of focal plane (Lugert et al., 2010), to have a radial GFAP+ve process. We investigated this by staining for Ki67, TBR2, and GFAP and observed that, indeed, many of the dividing TBR2−ve cells were clearly linked to a radial GFAP+ve process (Fig. 2D,D′). Together these data suggest that a significant proportion of dividing TBR2−ve cells are AH-NSCs.

Details are in the caption following the image

Ki67+ve; TBR2−ve nuclei have AH-NSC characteristics.

A–A″: Coronal section of an adult mouse brain at the level of the hippocampus, showing DAPI (white, A), SOX2 (green, A′) TBR2 (blue, A′′), and Ki67 (magenta, A′′) staining. The dashed lines demarcate the SGZ of the dentate gyrus. Ki67+ve; TBR2−ve cells (e.g., arrowhead in A) have higher SOX2 fluorescence (A′, B), and relatively larger nuclei (A–A″, C) than Ki67+ve; TBR2+ve cells (arrows in A–A″). D,D′: Coronal section of an adult mouse hippocampus showing TBR2 (blue, D), GFAP (red, D and D′), and Ki67 (magenta, D′) staining. Many Ki67+ve; TBR2−ve cells (arrowhead) were linked to a radial GFAP+ve process (arrows). ***P < 0.001, t-test. Scale bar A = 9 µm.

Ki67+ve; TBR2−ve Cells Include Horizontal AH-NSCs

Are all dividing TBR2−ve cells in the SGZ AH-NSCs? Or does some fraction of these cells have another identity? For example, could some dividing, TBR2−ve cells be IPs that do not express TBR2? To address this, we stained Hes5::GFP mice with Ki67 and TBR2, predicting that if dividing TBR2−ve cells were AH-NSCs, they should report high notch activity, which is indicated by GFP fluorescence in this line (Basak and Taylor, 2007; Lugert et al., 2010). Significantly, we found that all of the dividing TBR2−ve cells analyzed expressed GFP (24/24 cells) (Fig. 3A–C). This finding was confirmed via the analysis of a second marker of proliferation, PCNA, in the Hes5::GFP line, in which all of the PCNA+ve; TBR2−ve cells expressed GFP (31/31 cells; Fig. 3E–H). The co-localization of GFP with all of the dividing TBR2−ve cells suggests that this population consists of AH-NSCs.

Details are in the caption following the image

Ki67+ve; TBR2−ve nuclei include horizontally oriented Hes5::GFP AH-NSCs.

A–A″′, B–B″′: Coronal section of a mouse hippocampus, showing the dentate gyrus, with Hes5::GFP in (green), Ki67 (magenta), and TBR2 (blue). A–A″′: Some Ki67+ve; TBR2−ve nuclei (arrowhead) had a radial morphology (arrows), while others (B–B″′) had a horizontal morphology (arrowheads). C: All Ki67+ve; TBR2−ve nuclei expressed GFP. D: Radial and horizontal Ki67+ve; TBR2−ve nuclei were present in approximately equal numbers. E–E″′, F–F″′: Coronal section of a mouse hippocampus, showing the dentate gyrus, with Hes5::GFP in (green), PCNA (magenta), and TBR2 (blue). E–E″′: As with the Ki67 data, some PCNA+ve; TBR2−ve nuclei (arrowhead) had a radial morphology (arrows), while others (F–F″′) had a horizontal morphology (arrowheads). G: All PCNA+ve; TBR2−ve nuclei expressed GFP. H: Radial and horizontal PCNA+ve; TBR2−ve nuclei were present in approximately equal numbers. Scale bar A = 17 µm.

The advantage of the nuclear-only stain described here is that it should enable the identification of dividing AH-NSCs regardless of their orientation. The findings of Lugert and colleagues (2010) revealed that up to two-thirds of dividing AH-NSCs, as defined by Hes5::GFP expression, are in a horizontal orientation, whereas only the remaining one-third were the classic radially oriented AH-NSCs. To test whether the dividing TBR2−ve population included both horizontal and radial cells, we stained for Ki67 and TBR2 in Hes5::GFP mice and used the cytoplasmic GFP label to examine cellular morphology. We found that approximately half the dividing TBR2−ve cells exhibited a classical radial process (54%, 13/24 cells) (Fig. 3A,D), while the remaining GFP+ve; TBR2−ve cells exhibited only a short horizontal process, fitting the description of horizontal AH-NSCs by Lugert and colleagues (2010) (46%, 11/24 cells) (Fig. 3B,D). Together these data demonstrate that dividing TBR2−ve cells represent a population of AH-NSCs that includes cells with both radial and horizontal morphology.

Discussion

This study provides a morphology-independent approach to identify dividing AH-NSCs. The significance of this study, apart from its accuracy and simplicity, lies in the fact that the majority of quantification methods for AH-NSCs do not account for the putative population of horizontal AH-NSCs, as they do not possess a radially oriented fiber. Although the true lineage potential and self-renewing capacity of horizontal AH-NSCs is yet to be established, those cells potentially account for at least one-half (this study) or two-thirds (Lugert et al., 2010; Jhaveri et al., 2015) of the dividing precursor pool in the dentate gyrus that express stem cell–specific markers and characteristics (e.g., SOX2high, Hes5::GFP+ve, TBR2−ve). Assuming these horizontal cells are bona fide NSCs, current methods would underestimate stem cell numbers or misrepresent these horizontal cells as IPs. This is particularly important, as scenarios exist in which the enhancement or inhibition of certain signaling pathways may have no effect on the relative activation of radial AH-NSCs but could affect the activation of horizontal cells. Current quantification methods would miss these effects. Indeed, there is evidence that subpopulations of AH-NSCs respond differently to signaling cues. For example, reporter strains such as the Hes5::GFP and nestin::GFP lines have previously been used in conjunction with the expression of EGFR to isolate homogeneous populations of neurosphere-forming precursors in vitro, which contain distinct subpopulations that are responsive to norepinephrine exposure or KCl-depolarization (Jhaveri et al., 2015).

There are other advantages to the method described here. For example, because the protocol uses only nuclear-localized antigens, it can be used to quickly quantify dividing AH-NSCs either manually or in an automated manner using image analysis software. This is in stark contrast to the current method that requires users to trace a radial process back to the correct nucleus in the SGZ, which is both time-consuming and difficult to automate. Our approach is also likely to be more accurate. Dividing neural progenitors are often found in cellular clusters (Hodge et al., 2008), and it becomes difficult to trace a radial process back to the correct nucleus. Moreover, because our method relies only on the co-localization of two nuclear antigens, a relatively low-resolution confocal image is sufficient for the correct identification of these cells. Finally, our method requires only a dual antibody stain (Ki67 and TBR2). Combining this stain with other markers can provide additional information if required. For example, a triple stain with GFAP would provide an estimate of the proportion of dividing AH-NSCs that are radial (GFAP+ve radial process) or horizontally oriented. Alternately, a triple stain with SOX2 could be used to also estimate the number of quiescent AH-NSCs (Ki67−ve; SOX2+ve; TBR2−ve). It is important to note there are caveats to this method. For example, although we confine our analysis to the SGZ of the dentate gyrus, we cannot rule out that Ki67+ve; TBR2−ve are not dividing oligodendrocyte precursors or microglia with only these two stains. That said, our analysis of SOX2 expression revealed that fewer than 3% of SOX2+ve cells within the SGZ co-expressed the oligodendrocyte marker OLIG2, suggesting that the vast majority of these cells are AH-NSCs. Furthermore, triple labeling with cell type–specific markers can be used to quickly determine the percentage of Ki67+ve; TBR2−ve that are not AH-NSCs, and whether the proportion of these cells changes during different experimental conditions or during aging.

Despite such caveats, the simple morphology-independent approach to identify dividing AH-NSCs described in this study is an improvement over existing methods because it allows for the rapid and accurate quantification of dividing AH-NSCs. Critically, it is inclusive of horizontally oriented cells that traditional quantifications methods ignore. Although further work is needed to establish the true identity and fate potential of horizontally oriented precursors in the dentate gyrus, our method will be pivotal to provide a comprehensive interpretation of changes to precursor cell numbers in studies that investigate the molecular control of neurogenesis.

Experimental Procedures

Animal Ethics

The work performed in this study conformed to The University of Queensland's Animal Welfare Unit Guidelines for Animal Use in Research (AEC approval numbers QBI/353/13/NHMRC and QBI/355/13/NHMRC/BREED). All experiments were performed in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes and were carried out in accordance with The University of Queensland Institutional Biosafety Committee.

Animals

Two mouse lines were used in this study, wild-type C57BL/6J mice and Hes5::GFP mice (CD1 background) (Basak and Taylor, 2007; Lugert et al., 2010). These mice contain a 3-kilobase (kb) portion of Hes5 gene regulatory elements (1.6 kb upstream of the transcriptional start site, 1.4 kb downstream) controlling the expression of GFP. All mice in this study were analyzed between 12 and 16 weeks of age. Both male and female mice were used.

Primary Antibodies

The rabbit primary antibodies used in this study were anti-GFAP (Z033429-2, 1/1000, Dako), anti-TBR2 (ab23345, 1/800, Abcam), or monoclonal anti-TBR2 (EPR19012, 1/200, Abcam); the mouse primary antibodies were anti-Ki67 (550609, 1/200, BD Pharmingen), mouse anti-PCNA (ab29, 1/100, Abcam), and mouse anti-GFAP (MAB 360, 1/500, Millipore, MA); the rat primary antibodies were anti-KI67 FITC clone SolA15 (11-5698-80 1/400, eBioscience, CA) and anti-SOX2 Alex Fluor 488 clone Btjce (53-9811-82 1/400, eBioscience); the goat primary was anti-DCX (sc-8066, 1/200, Santa Cruz, CA), and chicken primary was anti-GFP (ab13970, 1/500, Abcam).

Tissue Processing and Immunofluorescence

Animals were perfused transcardially with phosphate buffered saline (PBS) followed by 4% (w/v) paraformaldehyde in PBS. The dorsal skulls of animals were removed and brains were post-fixed for 1–2 weeks. Coronal brain sections were cut with a Vibratome (Leica, Deerfield, IL) at a thickness of 50 µm.

Prior to immunostaining, sections were mounted on Superfrost slides. Heat-mediated antigen retrieval was then performed in 10 mM sodium-citrate solution in PBS (pH 6.0) at 95 °C for 15 min. Exceptions to this protocol were Hes5::GFP sections, in which heat-mediated antigen retrieval was performed for only 2 min. This shorter retrieval was required so as to not irreversibly damage the GFP epitope (Nakamura et al., 2008). A standard immunofluorescence protocol was then performed as described previously (Harris et al., 2016a).

Microscopy and Image Processing

Confocal images were acquired as 1-µm optical sections spanning a 10-µm-thick z-stack on a Zeiss inverted Axio Observer fitted with a W1 Yokogawa spinning-disk module and Hamamatsu Flash4.0 sCMOS camera and SlideBook software (3i). Image channels were pseudo-colored to allow for overlay, cropped, and minimum and maximum gray values adjusted in ImageJ (freeware). Individual images were resized in Photoshop (Adobe), and image tiles were constructed as display items in Illustrator (Adobe). The quantification of immunopositive cells from these images was restricted the SGZ, which we defined as 2 cell widths above and below the innermost granule cell of the dentate gyrus.

Fluorescence Intensity Quantification

A single in-focus plane was used to quantify SOX2 fluorescence. An outline was drawn around each cell nuclei using the DAPI channel and the area, mean gray value, and integrated density were measured (ImageJ). The total corrected cellular fluorescence (TCCF) = integrated density – (area of selected cell * mean fluorescence of background readings) was then calculated (McCloy et al., 2014).

Statistics

Student's t-tests were performed in GraphPad Prism (7.0) to compare the relative SOX2 (TCCF) and nuclei area of Ki67+ve TBR2−ve and Ki67+ve TBR2+ve cells. Data were obtained from at least five cells from each individual mouse (n = 3). Because we were interested in variability between cells within the SGZ niche, each individual cell was considered a biological replicate. This approach allows for a better measure of the biological variability per cell, rather than when averaging the values from multiple cells per mouse.

Acknowledgments

We thank Prof. Perry Bartlett for the provision of the Hes5::GFP mice. Microscopy was performed in the Queensland Brain Institute's Advanced Microscopy Facility. This work was supported by an Australian Research Council grant (DP160100368 to M.P.). M.P. was supported by a fellowship (Australian Research Council Future Fellowship; FT120100170), and D.J.J. was supported by Mater Foundation Fellowship. L.H. and S.O. were supported by Australian Postgraduate Awards.