ABSTRACT
Embryonic development is a complex and dynamic process that unfolds over time and involves the production and diversification of increasing numbers of cells. The impact of developmental time on the formation of the central nervous system is well documented, with evidence showing that time plays a crucial role in establishing the identity of neuronal subtypes. However, the study of how time translates into genetic instructions driving cell fate is limited by the scarcity of suitable experimental tools. We introduce BirthSeq, a new method for isolating and analyzing cells based on their birth date. This innovative technique allows for in vivo labeling of cells, isolation via fluorescence-activated cell sorting, and analysis using high-throughput techniques. We calibrated the BirthSeq method for developmental organs across three vertebrate species (mouse, chick and gecko), and utilized it for single-cell RNA sequencing and novel spatially resolved transcriptomic approaches in mouse and chick, respectively. Overall, BirthSeq provides a versatile tool for studying virtually any tissue in different vertebrate organisms, aiding developmental biology research by targeting cells and their temporal cues.
INTRODUCTION
Embryonic development has been defined as the sequential unfolding of the events of an embryo from which an individual emerges (Barresi and Gilbert, 2000). This process is not homogeneous in time, and most developmental events occur – and need to occur – at specific time points. According to Von Baer's influential laws of embryology (Abzhanov, 2013), earlier events in embryonic development tend to represent a more general feature of a group of species and usually have a greater influence on the formation of the individual. Development is thus a sequential program that combines the production of increasing numbers of cells with the diversification of different cell types (Duboule, 1994; García-Moreno and Molnár, 2020). Developmental biology is making great progress in identifying both the mechanisms and the role of time in the developmental process.
The impact of developmental time in the formation of the central nervous system is a well-documented example of how a given stem cell progenitor is capable of generating a range of different cell fates along embryonic time (Guillemot et al., 2006; Nieuwenhuys and Puelles, 2016; Klingler et al., 2021; Telley et al., 2016). The paradigmatic case is that of cortical progenitors in the mammalian telencephalon: one individual progenitor cell can generate sequentially a variety of pyramidal neurons for the above-lying neocortex, along the cortical formation period (Gao et al., 2014; Telley et al., 2019). Changing the developmental time at which this progenitor divides can modify the cell fate of their daughter cell (García-Moreno and Molnár, 2015). There is evidence of time ruling genomic and proteomic cues. The opening of chromatin in the Hox cluster must occur at a specific time in order for the next Hox gene to be expressed in sequence (Gaunt, 2015; Montavon and Duboule, 2013). When a Hox gene is activated, it triggers a sequence of genomic interactions that, after a given period of time, enables the expression of the next Hox gene in the cluster. This sequential expression allows the determination of different cell types based on the transcriptional Hox code expressed by the cell (Di Bonito et al., 2013), which is directly related to the time of its formation. As for the proteome, its stability sets a developmental tempo. Proteomic stability encodes for species-specific information about how the spinal cord should develop (Iwata, 2022; Rayon et al., 2020). More recently, time has been shown to play a crucial role in establishing a global temporal program for the identity of neuronal subtypes identified in the development of the mouse central nervous system (Sagner et al., 2021). There are probably many more cases in which time is translated into genetic instructions, which in turn drive embryonic development. However, the lack of specific tools allowing the simultaneous exploration of time and gene expression limits our range of exploration.
Here, we introduce BirthSeq, a solution for the shortage of techniques focused on isolating and analyzing cells based on their birth date. This innovative method enables in vivo labeling of cells at the time when they are generated, isolation of cells by fluorescence-activated cell sorting (FACS) at relevant developmental time points, and analysis of cell fate using cutting-edge high-throughput techniques. With a diverse range of administration options, BirthSeq is a versatile tool that is potentially suitable for many vertebrate species, allowing research on most tissues of the organism. Our tests of the various components of BirthSeq in the brain development of three vertebrate species (mouse, chick and gecko) have shown evidence of its effectiveness for single-cell RNA sequencing (scRNA-seq) of dated populations of neurons in mouse brain, as well as novel, spatially resolved transcriptomic techniques in chick embryonic brain. Here, we provide first the transcriptomic profile of early-born pallial neurons in mouse, and, second, we uncover the genetic identity and anatomical location of early-born neurons in the hypothalamus and diencephalon of the chick.
RESULTS AND DISCUSSION
FlashTag application in non-mammalian embryos
A permanent labeling system is crucial to reach the goal of studying specifically dated cell populations. 5-Bromo-2′-deoxyuridine (BrdU) and tritiated thymidine, two of the most widely used methods, have a drawback: they make the isolation of live cells impossible, as the staining needed for their visualization is cytotoxic. These two reagents are incorporated into newly synthesized DNA during the S phase of the cell cycle. However, there is another option that bypasses the DNA synthesis requirement: carboxyfluorescein succinimidyl ester (CFSE). When incorporated into the cytoplasm of dividing cells, it becomes fluorescent and acts as a birthdating marker. The application of CFSE in the brain ventricles, known as FlashTag (Govindan et al., 2018), has the potential to boost research on birthdated populations, at least for the developing brain. Therefore, we tested its capabilities in different species.
We replicated FlashTag experiments in chick embryos to test whether it was a useful method in non-mammalian species. With the objective of defining the optimal CFSE concentration to birthdate embryonic chick neuronal populations, we injected different CFSE concentrations in the ventricle of embryonic day (E) 4 embryos, when pallial neurogenesis begins in chick (Tsai et al., 1981), and analyzed brain cell populations by histology (Fig. S2) and flow cytometry (FC; Fig. S3) at E7, 3 days after injection.
Our results showed that neurons birthdated with low CFSE concentrations were almost undetectable and that their brightness was limited (Fig. S3); this was also observed in histological samples (Fig. S2). However, this scenario shifted when 10 mM CFSE in DMSO was used (Fig. S3F-I). Cell death percentage (Fig. S3F-I) was maintained below 1% in all tested conditions except for 10 mM CFSE, when it increased to almost 2%, although this increase has no impact on CFSE validity.
As 10 mM is the recommended CFSE concentration to label cells in mouse experimental model (Govindan et al., 2018; Telley et al., 2016, 2019), and our FC analysis results showed that this concentration ensures the detection of CFSE+ cells by FC, we decided to check the validity of the pulse-labeling birthdate in chick tissue. Histological analysis revealed that, 3 days after the injection, in the ventricular zone (VZ) of some telencephalic regions, such as dorsal pallium (DPall) and subpallium (SPall), there were still many CFSE+ neural progenitors (Fig. S2B). Therefore, newborn neurons may become CFSE+ even 3 days after CFSE administration at 10 mM, and cells were not labeled with lower concentrations in our hands (Fig. S2B). Owing to these unexpected results, the CFSE injections did not seem to be the optimal method for neuronal birthdating in developing chick embryos. However, its validity in other model organisms is well documented. Additionally, we observed that CFSE was toxic for chick embryos, as it caused brain hemorrhages and most of the embryos died after the injection (Figs S2A, S3I). This could be attributed to CFSE being diluted in DMSO, a reagent potentially very toxic to chick embryos, which remain immobile inside the egg.
We encountered several difficulties in the use of CFSE on chick embryos in our hands, related to embryo survival or lineage labeling. However, FlashTag has been used in the past by other authors in chick embryos, although the experimental conditions were markedly different (Baek et al., 2018). Also, none of our issues has been reported before in FlashTag injections in mouse. Importantly, because the goal of our project is the comparison between species, we needed to move towards other reagents, although we appreciate CFSE as a valuable choice for mammalian-centered projects. In particular, FlashTag would be the best method to distinguish the cell lineage derived from direct versus indirect neurogenesis, which is relevant in several brain regions (Baumann et al., 2023 preprint).
We have revitalized the idea of analyzing cells carrying a DNA mark with our novel method, BirthSeq. BrdU has been widely used in the past, but the antigen retrieval process required for detecting it within chromatin kills the cells, making live cell research impossible. To address this issue, BirthSeq is based on 5′-ethynyl-2′-deoxyuridine (EdU), which labels newly generated cells in a similar manner to BrdU, but without the need for harsh treatments on cells to reveal their DNA mark (Salic and Mitchison, 2007; Endaya et al., 2016). Ideally, EdU can be detected on disaggregated cells while preserving their viability. Our detailed protocol for BirthSeq (Fig. 1A) involves incorporating EdU into dividing cells through various administration methods, isolating live birthdated cells through FACS sorting, and analyzing them with RNA-seq tools. In the following sections, we describe tests for the suitability and optimal conditions for each of the steps of the BirthSeq protocol.
BirthSeq, a novel method to isolate birthdated cells. (A) Schematic of the BirthSeq methodology. BirthSeq consists of the birthdating of any vertebrate tissue by a systemic injection of EdU, followed by the extraction and dissociation of the birthdated tissue of interest. Next, EdU+ cells are detected with a low copper EdU detection reaction mix and isolated by FACS. Finally, the transcriptome of those EdU+ viable cells is analyzed by bulk or scRNA-seq. (B) Schematic of NeurogenesISS, a modified version of BirthSeq that is compatible with spatial transcriptomics.
BirthSeq, a novel method to isolate birthdated cells. (A) Schematic of the BirthSeq methodology. BirthSeq consists of the birthdating of any vertebrate tissue by a systemic injection of EdU, followed by the extraction and dissociation of the birthdated tissue of interest. Next, EdU+ cells are detected with a low copper EdU detection reaction mix and isolated by FACS. Finally, the transcriptome of those EdU+ viable cells is analyzed by bulk or scRNA-seq. (B) Schematic of NeurogenesISS, a modified version of BirthSeq that is compatible with spatial transcriptomics.
In ovo administration of EdU labels all dividing and birthdated cells
We conducted a test similar to that performed on FlashTag to assess the validity of EdU as a reagent for pulse-labeling in birthdating assays. We administered low doses of EdU during the embryonic development of chick, mouse and gecko embryos and evaluated its birthdating properties. Within 30 min of administration, dividing cells in the S-phase region of the ventricular zone of chick embryos acquired the EdU tag (Fig. 2A). After 2 h and 5 h, the labeled cells moved their nuclei towards the abventricular region of the ventricular zone where they eventually divided during the M phase (Fig. 2B,C). Importantly, 5 h after administration, there was no new incorporation of EdU in the S-phase region of the germinative zone, indicating that EdU at these low doses was rapidly washed from the ventricular zone and is an effective pulse-labeling factor in chick embryos. The administration of low doses of EdU did not result in an increase in embryo mortality (n>100) and did not generate the aberrant isochronic clusters that have been reported with high-dose BrdU in chick embryos (Rowell and Ragsdale, 2012).
In vivo EdU administration is an effective birthdating pulse-labeling assay. (A-C) Short-term birthdating of chick ventral pallial cells was performed at E4. (A) Thirty minutes after birthdating, EdU+ cells appeared at the most basal region of the neuroepithelium. (B) Two hours after the EdU injection, some of the EdU+ cells had started their migration to the ventricular surface, following their interkinetic nuclear movement. (C) Finally, 5 h after EdU administration, many of the EdU-labeled cells were undergoing mitosis (as determined by PH3 counterstaining) in the most apical region of the neuroepithelium. In addition, the S-phase region appeared mostly devoid of EdU labeling, as no new EdU was being incorporated at this time after the injection. (D-F) Long-term examples of EdU incorporation and labeling in brain coronal sections of E15 chick, injected at E6 (D); E27 gecko, injected at E7 (E); and P16 mouse, injected at E11 (F). In all of the researched species, there is an evident lack of EdU-labeled cells in the VZ of the telencephalon and the presence of many scattered EdU+ cells throughout the MZ. DAPI counterstain in blue. Dashed white lines in A-C demarcate the neuroepithelial region where neural stem cells duplicate their DNA (S phase, S), and in D-F demarcate anatomical boundaries. Yellow arrows in B,C mark the direction of cellular movement. Images to the right in D-F show enlargements of the boxed areas. Scale bars: 50 µm (A-C); 500 µm (main panels of D,F); 100 µm (main panel of E); 25 µm (insets of D,F); 5 µm (insets of E). Cx, cortex; DC, dorsal cortex; DVR, dorsal ventricular ridge; HC, hippocampus; HT, hypothalamus; LPall, lateral pallium; M, region for M phase in the VZ; MZ, mantle zone; Se, septum; Th, thalamus.
In vivo EdU administration is an effective birthdating pulse-labeling assay. (A-C) Short-term birthdating of chick ventral pallial cells was performed at E4. (A) Thirty minutes after birthdating, EdU+ cells appeared at the most basal region of the neuroepithelium. (B) Two hours after the EdU injection, some of the EdU+ cells had started their migration to the ventricular surface, following their interkinetic nuclear movement. (C) Finally, 5 h after EdU administration, many of the EdU-labeled cells were undergoing mitosis (as determined by PH3 counterstaining) in the most apical region of the neuroepithelium. In addition, the S-phase region appeared mostly devoid of EdU labeling, as no new EdU was being incorporated at this time after the injection. (D-F) Long-term examples of EdU incorporation and labeling in brain coronal sections of E15 chick, injected at E6 (D); E27 gecko, injected at E7 (E); and P16 mouse, injected at E11 (F). In all of the researched species, there is an evident lack of EdU-labeled cells in the VZ of the telencephalon and the presence of many scattered EdU+ cells throughout the MZ. DAPI counterstain in blue. Dashed white lines in A-C demarcate the neuroepithelial region where neural stem cells duplicate their DNA (S phase, S), and in D-F demarcate anatomical boundaries. Yellow arrows in B,C mark the direction of cellular movement. Images to the right in D-F show enlargements of the boxed areas. Scale bars: 50 µm (A-C); 500 µm (main panels of D,F); 100 µm (main panel of E); 25 µm (insets of D,F); 5 µm (insets of E). Cx, cortex; DC, dorsal cortex; DVR, dorsal ventricular ridge; HC, hippocampus; HT, hypothalamus; LPall, lateral pallium; M, region for M phase in the VZ; MZ, mantle zone; Se, septum; Th, thalamus.
We tested the long-term viability of EdU as a birthdating reagent in mouse, chick and gecko embryos, and our results showed that the birthdating was successful in all three species (Fig. 2D-F). Our examination of their brains several days after the administration of EdU revealed a differential staining pattern that was consistent with brain development. The neurons generated during the time of administration were labeled, whereas many other cells were not. This confirmed that the pulse-labeling was effective, as there was no remaining EdU mark in the progenitor cells days after the injection (Fig. 2D-F; Rueda-Alaña and García-Moreno, 2022).
To maximize the potential of BirthSeq, EdU can be administered by different methods to the embryo. In mouse experiments, it was administered through intraperitoneal injections via the pregnant dam. Chick and gecko embryos were mostly administered with intracardiac or intravenous injections. These three administration routes proved to be advantageous as they are systemic, meaning that the EdU molecule reaches all cells in the organism. It allows all organs to be researched with a single systemic dose, providing a more reliable birthdating analysis. In the case of the brain, as cells divide in several regions and neuroepithelial strata, some are not in an optimum position for labeling from the ventricle. We eliminated this limitation as systemic EdU also reaches dividing cells in subventricular positions, making the birthdating even more comprehensive. Intraventricular injections, although not systemic, were also able to reach all dividing cells in the developing brain. However, in order to disentangle the differential lineage of ventricular and subventricular progenitors, CFSE would be the preferential labeling marker, as it allows these populations to be labeled separately.
Our results demonstrate that EdU is a highly effective birthdating reagent for histological analysis, capable of being used in many amniote species and for the study of any developing organ or tissue through systemic administration.
Labeling and FACS isolation of viable EdU birthdated cells
Next, we evaluated the feasibility of using EdU labeling for FACS isolation of living cells. Owing to the scarcity of gecko embryos and the low number of neurons within their brains, we continued validation of the BirthSeq protocol in mouse and chick embryos. Our initial efforts to isolate live birthdated cells were unsuccessful, primarily because of the toxicity of the labeling reaction mixture used to reveal the EdU labeling. Because the reagents used in the labeling reaction are known, we were able to make modifications to the labeling protocol to maintain cell viability while still achieving successful fluorescent labeling. We experimented with different exposure times to the reaction mixture and reaction temperatures in chick embryos (Fig. S4), but the effective improvement was reducing the amount of Cu (II) in the reaction mixture (Fig. 3).
The reduction of Cu (II) concentration in the EdU detection cocktail increases cell viability without losing EdU detection capacity. (A) Experimental design of the FC analysis of dated chick embryonic neurons after Cu (II) volume variations in the EdU detection cocktail. (B-G) FC profiles of dissociated chick embryonic neural cells 1 day after EdU injection and EdU detection with 0 µl (B), 0.1 µl (C), 0.25 µl (D), 1 µl (E), 2.5 µl (F) and 20 µl (G) of Cu (II). (H-L) Graphical representations of FC analysis. Relative proportion of dead (H,I) and EdU+ (J,K) cells of total (H,J) and parental (I,K) populations were analyzed. EdU+ cell brightness was assessed by 513-17 FITC/Median (L). FC analysis demonstrated that copper is toxic for the cells, as shown by the high cell death rates observed in the samples revealed with high Cu (II) concentrations and its dramatic reduction after lowering the concentration. However, reducing copper concentration did not diminish the EdU detection capacity of the cocktail, making it possible to isolate EdU+ viable cells. Dead cells (P3, red boxes) were defined as PI+ in FSC versus dsRed dotplots (B-G, left panels), whereas EdU+ (P4, green boxes) and EdU− (P5, purple boxes) cells were identified as GFP+ and GFP− in FSC versus GFP dotplots (B-G, right panels), after establishing the threshold using a control sample. Data represented in the graphs was obtained from a single cytometry experiment, and analyzed samples contained three pooled brains.
The reduction of Cu (II) concentration in the EdU detection cocktail increases cell viability without losing EdU detection capacity. (A) Experimental design of the FC analysis of dated chick embryonic neurons after Cu (II) volume variations in the EdU detection cocktail. (B-G) FC profiles of dissociated chick embryonic neural cells 1 day after EdU injection and EdU detection with 0 µl (B), 0.1 µl (C), 0.25 µl (D), 1 µl (E), 2.5 µl (F) and 20 µl (G) of Cu (II). (H-L) Graphical representations of FC analysis. Relative proportion of dead (H,I) and EdU+ (J,K) cells of total (H,J) and parental (I,K) populations were analyzed. EdU+ cell brightness was assessed by 513-17 FITC/Median (L). FC analysis demonstrated that copper is toxic for the cells, as shown by the high cell death rates observed in the samples revealed with high Cu (II) concentrations and its dramatic reduction after lowering the concentration. However, reducing copper concentration did not diminish the EdU detection capacity of the cocktail, making it possible to isolate EdU+ viable cells. Dead cells (P3, red boxes) were defined as PI+ in FSC versus dsRed dotplots (B-G, left panels), whereas EdU+ (P4, green boxes) and EdU− (P5, purple boxes) cells were identified as GFP+ and GFP− in FSC versus GFP dotplots (B-G, right panels), after establishing the threshold using a control sample. Data represented in the graphs was obtained from a single cytometry experiment, and analyzed samples contained three pooled brains.
EdU ‘click’ technology relies on the creation of a covalent bond. This chemical reaction is catalyzed by Cu (I), which is created in situ from Cu (II) obtained from CuSO4 (Rostovtsev et al., 2002; Salic and Mitchison, 2007; Tornøe et al., 2002). It is well known that Cu (II) is toxic for cells, as it reduces mRNA transcription and promotes RNA degradation (Halliwell and Gutteridget, 1984; Halliwell et al., 1992). However, there is evidence that, in fixed cells, an 85% reduction in Cu (II) concentration was able to replicate the standard EdU detection capacity (Ng et al., 2017). Hence, we decided to analyze by FC chick brain birthdated samples after modifying the Cu (II) concentration in the EdU cocktail (Fig. 3), following previous experiments suggesting the efficiency of such a decrease for maintaining RNA integrity (Endaya et al., 2016).
Experiments in chick embryos showed that cell death percentages decreased proportionally with Cu (II) reduction, until almost undetectable levels were reached after decreasing Cu (II) to 1.25% and 0.5% of the original concentration (Fig. 3B-G left, H,I). At the same time, we observed that EdU+ cells were still distinguishable, although their brightness was slightly reduced (Fig. 3B-G left, L), and that this low Cu (II) cocktail was capable of detecting EdU+ cells also on histological samples (Fig. S5). Further analysis showed that samples revealed with low Cu (II) levels were enriched with EdU+ cells compared with the rest of the samples (Fig. 3B-G left, J), which is an expected consequence of the reduced cytotoxicity. In addition, we observed that the EdU parental population (EdU+ viable cells with the size and shape of interest) was smaller in those samples revealed with low copper (Fig. 3B-G left, K), which, together with brightness loss, could indicate that the Cu (II) concentration decrease reduces the capacity to detect EdU.
Considering all these results, we decided to work with 0.5% of the original Cu (II) concentration (as in Fig. 3C). Although our EdU detection capacity was reduced with this Cu (II) concentration, its cytotoxicity was largely negligible. We would still be able to isolate birthdated cells that had divided only once after the injection (so the EdU of their DNA would not have been diluted and would be enough to be detected), while maintaining a high survival rate. For our particular experimental setting, this is actually an advantage, allowing more accurate birthdating.
As mentioned above, Cu (II) can interfere with RNA expression (Halliwell and Gutteridget, 1984; Halliwell et al., 1992; Ng et al., 2017). Consequently, we decided to examine whether the new EdU cocktail [containing low Cu (II) concentration] caused altered cell gene expression. For this, we purified by FACS a non-revealed control population and a treated population, and compared their gene marker expression by RT-qPCR (Fig. S6). First, FACS results showed that the treatment was not modifying population size and that it was safe for cell viability (Fig. S6B-E). However, RT-qPCR analysis showed that some mRNA expression was altered in the treated sample (Fig. S4F), and that the new EdU cocktail could diminish the number of mRNA reads in transcriptomic experiments. Surprisingly, these experiments showed the opposite results to our preliminary experiments and also to previously published literature (Endaya et al., 2016). Following this RT-qPCR analysis, we decided to continue our optimization of the protocol, to assess whether this caveat of BirthSeq could be a limitation for scRNA-seq and other omics technologies.
Once the EdU detection protocol was optimized, the next step was to verify whether it was sensitive enough to selectively isolate birthdated cells. In order to do that, we compared the mRNA expression of two neural cell types that, at the time of the isolation, were at two different stages of maturation in the E7 chick brain: neural progenitors (birthdated 30 min before isolation) and immature neurons (birthdated 3 days before isolation) (Fig. 4A-E). We found statistically significant differences in the expression of all the analyzed markers: neural progenitors had a higher mRNA expression of progenitor markers than immature neurons (Fig. 4F). This difference of expression tendency shifted its orientation when neural markers were analyzed (Fig. 4G). Importantly, the diminished mRNA expression caused by exposure to the EdU detection cocktail did not interfere with the identification of the two different populations. Taken together, these results suggested that use of this protocol to selectively isolate EdU+ cells was valid.
EdU-mediated cell isolation is specific for birthdated cells. (A) Experimental design in chick embryos used to analyze gene expression of birthdated neuronal progenitors and immature neurons. (B,D) Coronal sections and high-power views of E7 chick telencephala injected with EdU 30 min (B) or 3 days (D) before the analysis. Images clearly show that E7 chick embryos birthdated 30 min before the analysis only have EdU+ progenitors located in the VZ, and there is a lack of EdU-labeled cells in the MZ. However, 3 days after being birthdated, EdU+ cells present in E7 chick telencephalon are detected in the MZ and correspond to immature neurons. Images to the right show enlargements of the boxed areas. (C,E) FC profiles of birthdated chick neuronal progenitors (C) and neurons (E). (F,G) Expression of progenitor (F) and neuronal (G) markers in neuronal progenitors (E7+30 min) versus immature neurons (E4-E7) cells by RT-qPCR. Progenitor cells expressed higher levels of progenitor markers, whereas neuronal markers were expressed at higher levels in immature neurons than in neural progenitors. RPL7 was selected as a reference gene. DAPI counterstain in blue. Dashed white lines demarcate anatomical boundaries. Scale bars: 250 µm (B,D, main panels); 25 µm (B,D, insets). EdU+ (P4, green boxes) and EdU− (P5, purple boxes) cells were identified as GFP+ and GFP− in FSC versus GFP dotplots (C,E, right panels), whereas dead cells (P3, red boxes) were defined as PI+ in FSC versus dsRed dotplots (C,E, left panels). *P<0.05; **P<0.01; ***P<0.001 (unpaired Student's t-test) (F,G). Bars represent mean±s.e.m. Dots show individual samples. Each sample contained seven pooled brains. a.u., arbitrary units; DPall, dorsal pallium; IHC, immunohistochemistry; MPall, medial pallium; MZ, mantle zone.
EdU-mediated cell isolation is specific for birthdated cells. (A) Experimental design in chick embryos used to analyze gene expression of birthdated neuronal progenitors and immature neurons. (B,D) Coronal sections and high-power views of E7 chick telencephala injected with EdU 30 min (B) or 3 days (D) before the analysis. Images clearly show that E7 chick embryos birthdated 30 min before the analysis only have EdU+ progenitors located in the VZ, and there is a lack of EdU-labeled cells in the MZ. However, 3 days after being birthdated, EdU+ cells present in E7 chick telencephalon are detected in the MZ and correspond to immature neurons. Images to the right show enlargements of the boxed areas. (C,E) FC profiles of birthdated chick neuronal progenitors (C) and neurons (E). (F,G) Expression of progenitor (F) and neuronal (G) markers in neuronal progenitors (E7+30 min) versus immature neurons (E4-E7) cells by RT-qPCR. Progenitor cells expressed higher levels of progenitor markers, whereas neuronal markers were expressed at higher levels in immature neurons than in neural progenitors. RPL7 was selected as a reference gene. DAPI counterstain in blue. Dashed white lines demarcate anatomical boundaries. Scale bars: 250 µm (B,D, main panels); 25 µm (B,D, insets). EdU+ (P4, green boxes) and EdU− (P5, purple boxes) cells were identified as GFP+ and GFP− in FSC versus GFP dotplots (C,E, right panels), whereas dead cells (P3, red boxes) were defined as PI+ in FSC versus dsRed dotplots (C,E, left panels). *P<0.05; **P<0.01; ***P<0.001 (unpaired Student's t-test) (F,G). Bars represent mean±s.e.m. Dots show individual samples. Each sample contained seven pooled brains. a.u., arbitrary units; DPall, dorsal pallium; IHC, immunohistochemistry; MPall, medial pallium; MZ, mantle zone.
Finally, we wanted to ratify the versatility of the method and verify that the new EdU cocktail was valid to detect various other types of birthdated cells. To do this, EdU was intravenously injected into chick embryos to reach all cells of the embryo systemically (Fig. S7). Three days after the injection, embryonic hearts and limbs were dissociated and EdU was detected on those two tissue samples (Fig. S7A). FC analysis of parental populations showed that cell death percentage was low in both cardiac and limb cells (Fig. S7B,C left, D), suggesting that the EdU cocktail is safe for cells obtained from different tissues. We were also able to visualize (and, therefore, isolate if needed) bright EdU+ populations in both samples (Fig. S7B,C right, E,F), which, once again, validated use of the new EdU cocktail to detect birthdated cells.
BirthSeq cells can be used for scRNA-seq
One of the final goals of BirthSeq is to conduct high-throughput sequencing experiments on cell populations for which birthdate is known. We tested the applicability of BirthSeq to scRNA-seq experiments. With our comparative goal in mind, we shifted our experimental focus to mouse embryos for two primary reasons. First, unlike chick pallium, there is a wealth of available information regarding the cortical populations generated at specific time points in mouse development. Second, leveraging scRNA-seq data from other publications provided us with a valuable opportunity to compare our BirthSeq extracted cells. We utilized the protocol for extracting E12.5-generated cortical cells from postnatal day (P) 3 mouse pups (Fig. 5A,B). At this stage of postnatal development, the cortex already harbors populations of neurons and glial cells (Fig. 5B). This is relevant because it is known that neurons are more sensitive to the pressure and aggressive conditions of FACS sorting than are glial cells (Pan and Wan, 2020; Ryan et al., 2021). When we applied BirthSeq to the brains of these mice, we obtained the expected outcome in the scRNA-seq dataset: several populations of glial cells were present in the dataset as they survived better under the BirthSeq conditions, as well as the expected and birthdated populations of neurons (Fig. 5C,D, Fig. S8). Specifically, the injection of EdU at E12.5 labeled Cajal–Retzius cells, immature deep layer cortical neurons, and other pallial populations known to be generated at this time point, including GABAergic interneurons (Fig. 5E,F). Additionally, we found earliest-generated upper layer cortical neurons, which may correspond to the latest divisions of EdU+ progenitors (Fig. 5E,F). All these cells exhibited the typical markers of the populations to which they belonged (Fig. 5F), thereby confirming that BirthSeq was capable of generating an scRNA-seq dataset enriched in the populations of cells generated shortly after the EdU injection. Cajal–Retzius cells are the best example because they constitute a minor population of cortical neurons and are known to disappear from the tissue at postnatal stages (Meyer, 1999). Any cortical dataset from P3 mouse cortex would reveal very few of these cells (Di Bella et al., 2021), whereas, in our birthdated dataset, it is a highly enriched cluster (Fig. 4G). Indeed, this enrichment of early-born cells was evident when comparing our BirthSeq dataset with two other available full cortical datasets of early postnatal mice (Fig. S8). These data confirm BirthSeq as a suitable method to isolate and analyze birthdated populations using scRNA-seq methodologies. EdU has been suggested to induce DNA damage to some extent (Kohlmeier et al., 2013), although our results confirmed that, if any, this damage might not be relevant for the validity of the tool because (1) scRNA-seq populations seem transcriptomically similar to equivalent cells in available datasets in which EdU was not employed (Fig. S8), and (2) our injections, performed shortly after organogenesis in mouse and chick, did not induce morphological changes, which would be expected after DNA damage at such a sensitive developmental stage.
BirthSeq allows scRNA-seq of birthdated populations in the mouse brain. (A) Experimental design. Pregnant dams were injected at E12, when neurogenesis is starting in the neocortex of mouse embryos. After surgery, embryos were allowed to develop to P3, when we applied the BirthSeq protocol followed by scRNA-seq. (B) EdU staining (yellow) of P3 mouse telencephalon showing the location of birthdated cells within the tissue in a littermate; DAPI counterstain is shown in blue. (C,D) UMAP distribution of E12 birthdated cells after BirthSeq. All main neural populations were identified. Representative gene plots for marker genes of each main cell type are shown in D. (E) UMAP distribution of the neuronal populations only (encircled by dashed line in C), including glutamatergic and GABAergic neurons. (F) Heatmap of gene markers for each of the nine neuronal clusters identified and generated at E12, organized into four GABAergic and five glutamatergic clusters. (G) Feature plots of different Cajal–Retzius cell markers to show the enrichment of this early-born population in the sequenced dataset.
BirthSeq allows scRNA-seq of birthdated populations in the mouse brain. (A) Experimental design. Pregnant dams were injected at E12, when neurogenesis is starting in the neocortex of mouse embryos. After surgery, embryos were allowed to develop to P3, when we applied the BirthSeq protocol followed by scRNA-seq. (B) EdU staining (yellow) of P3 mouse telencephalon showing the location of birthdated cells within the tissue in a littermate; DAPI counterstain is shown in blue. (C,D) UMAP distribution of E12 birthdated cells after BirthSeq. All main neural populations were identified. Representative gene plots for marker genes of each main cell type are shown in D. (E) UMAP distribution of the neuronal populations only (encircled by dashed line in C), including glutamatergic and GABAergic neurons. (F) Heatmap of gene markers for each of the nine neuronal clusters identified and generated at E12, organized into four GABAergic and five glutamatergic clusters. (G) Feature plots of different Cajal–Retzius cell markers to show the enrichment of this early-born population in the sequenced dataset.
This is one of the first examples of the application of scRNA-seq to cells for which birthdate is known a priori. In addition, we have performed the equivalent experiment in chick embryos (Lee et al., 2023 preprint). As such, BirthSeq may be the first birthdating method coupled to omics applications that is not limited to a specific cell type, organ or species.
NeurogenesISS: BirthSeq along spatially resolved transcriptomics
EdU labeling can be detected by fluorescent microscopy, which allowed us to combine BirthSeq with spatial-omics techniques. To test the compatibility of the EdU label with in situ sequencing (ISS) (Ke et al., 2013; Lee et al., 2023 preprint), we worked on an E15 chick brain sample treated with BirthSeq at E4 and used ISS to detect the expression of 84 genes at single-cell resolution (Fig. 6). At the end of the ISS detection cycles, we revealed the EdU label using click-chemistry. Given our focus on the generation of diencephalic neurons, we named this combined tool ‘NeurogenesISS’. Our use of both techniques proved successful, enabling us to determine the transcriptomic profile and brain location of the diencephalic and hypothalamic neurons generated at E4 (Fig. 6A).
NeurogenesISS resolves spatial transcriptomics of birthdated populations in the chick brain. (A) Anatomical location of the brain region under research, the diencephalon (highlighted in pink on the left), in a coronal section. The embryos were injected at E4, and the EdU staining was revealed at E15, depicted by the red dots in the right-hand panel (in the isolated diencephalon). (B) Merge image of 16 genes from the panel obtained through ISS reveals broad regional differences between the thalamus, prethalamus and hypothalamus. (C) Example of the gene expression pattern of 16 relevant genes detected after ISS. (D) A high-magnification detail image illustrates the ISS mapped reads of two types of cells: one EdU+ cell from cluster 17 (left), and one EdU− cell from cluster 9 (right). Pie charts show the percentage of mRNA reads detected for each gene of the panel, and these percentages were used to assign a cell identity. (E) Distribution of 18 cell clusters in the diencephalon detected after clustering of the cells, informed by the quantitative expression of 84 analyzed genes. It shows the identification of diencephalic and hypothalamic nuclei, guided by ISS gene expression and the distribution of cell clusters. (F) UMAP distribution of cells belonging to the 18 clusters, grouped by major cell type. (G) Equivalent UMAP distribution, highlighting in red the cells that were EdU+ (generated at E4). (H) Distribution of three cell classes in the diencephalic section, organized by major cell type. Green dots represent cells of a given cluster, and red dots represent cells of the cluster that were EdU+ (generated at E4). (I) Dot plot representing the expression of ISS gene markers on each cell cluster. The depth of the color represents increased expression with respect to the average, and the size of the dot represents the percentage of cells in the cluster expressing the gene marker (as shown in the key on the right). The axes of the graph are cluster identity (y-axis; color-coded by major cell type as in J) and gene marker (x-axis). (J) Proportion (percentage) of cells within each cluster that were generated at E4, following the administration of EdU. A13, A13 group of dopaminergic neurons in the zona incerta; al, ansa lenticularis; DIA, dorsal intermediate anterior nucleus of the thalamus; DLA, dorsolateral anterior nucleus of the thalamus; DLG, dorsal lateral geniculate nucleus of the thalamus; DMA, dorsomedial anterior nucleus of the thalamus; DSS, dorsal somatosensory nucleus of the thalamus; ICL, intercalated nucleus of the prethalamus; lfb, lateral forebrain bundle; LHA, lateral hypothalamic area; ot, optic tract; PG, pregeniculate nucleus; PVN, paraventricular hypothalamic nucleus; Rot, rotundus nucleus; SCh, suprachiasmatic nucleus; SubT, subthalamic nucleus; SPa, subparaventricular nucleus; SubG, subgeniculate nucleus; TRN, thalamic reticular nucleus; ZI, zona incerta.
NeurogenesISS resolves spatial transcriptomics of birthdated populations in the chick brain. (A) Anatomical location of the brain region under research, the diencephalon (highlighted in pink on the left), in a coronal section. The embryos were injected at E4, and the EdU staining was revealed at E15, depicted by the red dots in the right-hand panel (in the isolated diencephalon). (B) Merge image of 16 genes from the panel obtained through ISS reveals broad regional differences between the thalamus, prethalamus and hypothalamus. (C) Example of the gene expression pattern of 16 relevant genes detected after ISS. (D) A high-magnification detail image illustrates the ISS mapped reads of two types of cells: one EdU+ cell from cluster 17 (left), and one EdU− cell from cluster 9 (right). Pie charts show the percentage of mRNA reads detected for each gene of the panel, and these percentages were used to assign a cell identity. (E) Distribution of 18 cell clusters in the diencephalon detected after clustering of the cells, informed by the quantitative expression of 84 analyzed genes. It shows the identification of diencephalic and hypothalamic nuclei, guided by ISS gene expression and the distribution of cell clusters. (F) UMAP distribution of cells belonging to the 18 clusters, grouped by major cell type. (G) Equivalent UMAP distribution, highlighting in red the cells that were EdU+ (generated at E4). (H) Distribution of three cell classes in the diencephalic section, organized by major cell type. Green dots represent cells of a given cluster, and red dots represent cells of the cluster that were EdU+ (generated at E4). (I) Dot plot representing the expression of ISS gene markers on each cell cluster. The depth of the color represents increased expression with respect to the average, and the size of the dot represents the percentage of cells in the cluster expressing the gene marker (as shown in the key on the right). The axes of the graph are cluster identity (y-axis; color-coded by major cell type as in J) and gene marker (x-axis). (J) Proportion (percentage) of cells within each cluster that were generated at E4, following the administration of EdU. A13, A13 group of dopaminergic neurons in the zona incerta; al, ansa lenticularis; DIA, dorsal intermediate anterior nucleus of the thalamus; DLA, dorsolateral anterior nucleus of the thalamus; DLG, dorsal lateral geniculate nucleus of the thalamus; DMA, dorsomedial anterior nucleus of the thalamus; DSS, dorsal somatosensory nucleus of the thalamus; ICL, intercalated nucleus of the prethalamus; lfb, lateral forebrain bundle; LHA, lateral hypothalamic area; ot, optic tract; PG, pregeniculate nucleus; PVN, paraventricular hypothalamic nucleus; Rot, rotundus nucleus; SCh, suprachiasmatic nucleus; SubT, subthalamic nucleus; SPa, subparaventricular nucleus; SubG, subgeniculate nucleus; TRN, thalamic reticular nucleus; ZI, zona incerta.
We initiated our analysis by determining the expression pattern of the 84 genes tested within the tissue sample (Fig. 6B,C). We then segmented the tissue into individual cells based on the nuclear signal, and assigned to each cell a set of neighboring ISS signals situated within its segmentation mask (Fig. 6D,E). At this stage, the data consist of a cell by gene expression matrix, which can be mainly treated in the same way as a single-cell RNA dataset. We proceeded to perform clustering of the diencephalon cells based on their gene expression similarity (Fig. 6D-G). Our analysis produced a comprehensive spatial map of cell types, detailing the location of each cell type within the diencephalon (Fig. 6F). In total, we identified 18 clusters corresponding to 34,717 total mapped cells (Fig. 6D,F,G and Fig. S9). Thalamic clusters, with their glutamatergic nature, were easily identifiable, as were GABAergic neurons of the prethalamus and hypothalamus. These cells belong to major classes of cells such as: five classes of GABAergic neurons mainly distributed across the prethalamus and regions of the hypothalamus (ISLR2, NPY, DLX2, GAD1, SST1, PENK), five classes of glutamatergic neurons mainly located within the limits of the thalamus and regions of the hypothalamus (NEFM, CCK, SYT4, SV2A, CHL1, SLC17A6), two types of astrocyte cells with a different distribution pattern (PTN, RGMA, AQP4), and another two types of oligodendrocytes of varied maturation degree (OLIG2, PLP1, PDGFRA). Additionally, one cluster belonging to blood cells (HBA1) and three clusters of unknown identity were detected.
Finally, by means of the EdU administered at E4 we determined the diencephalic and hypothalamic neurons generated at that specific time point. We extracted the EdU intensity for each cell, and defined as EdU-positive all the cells that had a median intensity superior to a given conservative threshold. We then extracted the identity and location of all the EdU-positive cells. This enabled us to visualize the location and genetic profile of chick neurons generated at E4 (Fig. 6H,I). Our findings revealed that GABAergic cells of the prethalamus and hypothalamus were the most abundant populations of cells generated at the E4 stage (Fig. 6J). A good example is the distribution of cluster 1, comprising GABAergic cells located in the prethalamus and hypothalamus. In this cluster, identified by enriched expression of ISLR2, up to 15% cells were generated at E4 as labeled by EdU (Fig. 6J). These populations were specifically located in the reticular thalamic nucleus, the pregeniculate nucleus, the intercalate nucleus of the prethalamus, and the paraventricular and subparaventricular nuclei of the hypothalamus. Also, nearly 20% of all NPY+ GABAergic cells, located in the surroundings of the rotundus nucleus (cluster 17 in Fig. S6), were generated at E4. By contrast, the percentage of glutamatergic neurons generated at E4 was consistently lower than that of GABAergic cells, ranging from 2.5% in the case of CCK+ neurons of the dorsal anterior nucleus of the thalamus to 7.5% of other glutamatergic thalamic neurons labeled with SV2A (clusters 9 and 5, respectively, in Fig. S6). As expected from an early labeling time point, the percentage of glial cells generated was the lowest among the neural cells detected, with a maximum of 3% of either astrocytes or oligodendrocytes of the four classes detected being generated as early as E4 (Fig. 6G,J). This dataset represents one of the first instances of the coupling of spatially resolved transcriptomics and birthdating analysis.
The potential of BirthSeq and its spatial-omics variant NeurogenesISS has yielded highly promising results, establishing their position as versatile tools to explore the crucial role of developmental timing in generating diverse cell populations. With their flexibility, these techniques can be effectively applied across various vertebrate species and cell types from many organs, paving the way for further research in the dynamic fields of proteomics and epigenomics. BirthSeq tools are versatile and can be used in conjunction with various scRNA-seq platforms. They could also be combined with either scATACseq (Baek and Lee, 2020) or multiomics approaches to uncover changes in the epigenetic profiles of cells during differentiation. When combined with CITE-seq (Stoeckius et al., 2017), BirthSeq tools could enable the identification of cell populations and their ages based on their surface protein expression. Additionally, when used alongside Perturb-seq (Dixit et al., 2016), BirthSeq tools could help pinpoint specific genes that play a crucial role in the formation of these dated cell populations. BirthSeq development represents a relevant breakthrough in research into the intricate mechanisms underlying cellular differentiation and developmental processes.
MATERIALS AND METHODS
Experimental animals
All animal experiments were approved by the University of the Basque Country (UPV/EHU) Ethics Committee (Leioa, Spain) and the Diputación Foral de Bizkaia, and conducted in accordance with personal and project licenses in compliance with the current normative standards of the European Union (Directive 2010/63/EU) and the Spanish Government (Royal Decrees 1201/2005 and 53/2013, Law 32/107). Fertilized chick eggs (Gallus gallus) were purchased from Granja Santa Isabel (Córdoba, Spain). They were incubated at 37.5°C in humidified atmosphere until required developmental stage. The day when eggs were incubated was considered E0. Adult C57BL/6 mice (Mus musculus) were obtained from a mouse breeding colony at Achucarro Basque Center for Neuroscience (Spain). Animals were housed in a 12:12 h light/dark cycle (08.00 h, lights on), at constant temperature (19-22°C) and humidity (40-50%); and provided with ad libitum food and water. The day when the vaginal plug was detected was considered E0. Fertilized ground Madagascar gecko eggs (Paroedura picta) were obtained from a breeding colony at Achucarro Basque Center for Neuroscience (Spain). Adult geckos were housed in a 12:12 h light/dark cycle (08.00 h, lights on, 27°C; 20.00 h lights off, 22°C) and provided with ad libitum food (live crickets) and water. Gecko eggs were incubated at 28°C in a low humidified atmosphere until required developmental stage. The day when eggs were harvested from the terrarium was considered E0.
In ovo procedures and birthdating experiments
Manipulation of chick and gecko embryos was performed as previously described (Rueda-Alaña and García-Moreno, 2022; Rueda-Alaña et al., 2018). Briefly, eggs were incubated in a vertical position at either 37.5 or 28°C. Birthdating molecules were administered via intraventricular, intracardiac or intravenous injections for chick embryos or via intraventricular injection for gecko embryos, by mouth-pipetting using a fine pulled-glass needle. Drops of sterilized Hank's balanced salt solution (HBSS; HyClone Cytiva) were added to prevent embryonic dehydration, eggs were sealed and embryos were incubated until needed.
EdU birthdating
EdU (EdU Cell Proliferation Kit, baseclick) was diluted in sterile PBS (0.1 M, pH 7.6) with 0.1% Fast Green FCF (Sigma-Aldrich), and the dose was adjusted for each species and developmental stages. Chick embryos were injected with a single dose of 1 µl EdU: E4 embryos were injected with 2.5 µg/µl EdU; E6 and E7 embryos were injected with 5 µg/µl EdU. Gecko E7 embryos were administered a single 0.5 µl dose of 25 ng/µl EdU. Pregnant mice dams were intraperitoneally injected with 150 mg/kg EdU at E11 and E12.
CFSE birthdating
CFSE (CellTrace™ CFSE, Life Technologies) was diluted in DMSO (Sigma-Aldrich) with 0.1% Fast Green FCF. For the FlashTag FC experiments and histology, E4 chick embryos were intraventricularly injected with 1 µl of 1-10 mM CFSE. For tissue dissociation optimization experiments, E4 chick embryos were injected intraventricularly with 1 µl of 10 mM CFSE.
Brain tissue processing: perfusion, fixation and sectioning
Immunohistochemistry
All the brains were fixed in 4% paraformaldehyde (PFA; Sigma-Aldrich) diluted in PBS. Gecko brains and the chick embryonic brains up to E7 were collected in ice-cold HBSS and fixed by immersion in PFA for 24 h, transferred to PBS and kept at 4°C. P3 mouse pups and E15 chick embryos were anesthetized by hypothermia after immersion of either pup or the chick egg in ice. P16 mouse pups were deeply anesthetized with 2.5% 2,2,2-tribromoethanol (Avertin; Sigma-Aldrich), then they were transcardiacally perfused with PBS followed by 4% PFA. Brains were removed and postfixed with the same fixative for 3 h at room temperature (RT), then transferred to PBS and kept at 4°C.
Brains were serially sectioned in the coronal plane using a Leica VT1200S vibrating blade microtome (Leica Microsystems) and stored at 4°C until use. All mouse and gecko brains were cut at 50 µm thickness. Embryonic chick brains were cut at 60 µm thickness, to prevent tissue breakdown.
ISS
E15 chick embryos were anesthetized by hypothermia. Brains were extracted in ice-cold HBSS, cryopreserved in 30% sucrose, fresh-frozen in optimal cutting temperature media (OCT, Tissue-Tek®) and stored at −80°C until sectioning. Tissue was sectioned in the coronal plane, under RNase-free conditions, at 12-16 µm thickness with a Leica CM1950 cryostat (Leica Microsystems) and collected on SuperFrost Plus microscope slides (VWR). Slides were stored at −80°C until processing.
Immunohistochemistry and birthdating staining
Single immunohistochemical reactions were performed as described previously (Rueda-Alaña and García-Moreno, 2022; Rueda-Alaña et al., 2018). Briefly, sections were incubated with blocking and permeabilization solution [PBS containing 0.5% Triton X-100 (Sigma-Aldrich) and 3% bovine serum albumin (BSA; Sigma-Aldrich)] for 3 h at RT and then incubated overnight with the primary antibodies (diluted in the same solution) at 4°C with agitation. After the incubation, the primary antibodies were removed. Sections were washed five times with 0.5% PBT (0.5% Triton X-100 in PBS) for 10 min, and once with the blocking and permeabilization solution for 10 min. Next, sections were incubated with fluorochrome-conjugated secondary antibodies diluted in the same solution for 2 h at RT in the dark. Finally, sections were washed with 0.5% PBT twice for 10 min.
In birthdated animals, at the end of the immunohistochemical process, EdU molecule was detected in the sections (EdU Cell Proliferation Kit, baseclick) according to the manufacturer's protocol using ‘Click-iT’ chemistry. Finally, sections were washed with 0.5% PBT and mounted on SuperFlost slides (Epredia™) with Mowiol mounting medium (Mowiol® 4-88, Sigma-Aldrich).
All sections were counterstained with DAPI (1:1000; Sigma-Aldrich). The following antibodies and EdU azides were used: rabbit anti-PH3 (1:100; 06-570, Sigma-Aldrich), Alexa Fluor 488 goat anti-rabbit (1:1000; A11008 Invitrogen), 6-FAM azide (1:500; Lumiprobe) and sulfo-cyanine 5 azide (1:500; Lumiprobe).
Image capture and analysis
Images were collected using a Leica SP8 laser scanning confocal microscope (Leica Microsystems), a Carl Zeiss ApoTome2 microscope (Carl Zeiss Microimaging) or a 3DHistech Panoramic Midi II digital slidescanner (3DHistech).
Images obtained with the confocal microscope were acquired with LAS X software. The signal from fluorophores was collected sequentially; green light and far-red light were collected together. The same image parameters (laser power, gain and wavelengths) were maintained for images from each slide and adjusted for new animals. For large brain sections, tile-scan images were composed. All images shown are projections from z-stacks ranging from 10 to 35 µm thickness, typically 20 µm. Images obtained with the Zeiss microscope were acquired with a 20× air objective. The same image parameters (laser power, gain and wavelengths) were maintained for images from the same brain and adjusted for new slides. For large brain sections, tile-scan images were composed. Images obtained with the slidescanner microscope were acquired in single layer mode, using a 20× objective and collecting the signal from fluorophores sequentially.
BirthSeq optimization procedures
Tissue dissociation and sample preparation
All chick samples for this set of experiments were dissociated using an adaptation of previously described protocols (Beccari et al., 2018; Huettner and Baughman, 1986; Moussaud and Draheim, 2010). Briefly, brains were extracted, collected in ice-cold HBSS and pooled. Tissue was cut into small pieces and placed into full chick enzymatic solution (see details in Table S1). Samples were incubated at 37°C for 15-20 min, manually homogenized by carefully pipetting and filtered through a 40 µm nylon strainer (Fisher Scientific) to a 15 ml Falcon tube quenched by 4 ml of 25% fetal bovine serum (Gibco) in HBSS. Afterwards, cell suspensions were centrifuged at 200 g for 10 min at RT.
FC and FACS
FC was performed on a FACS Jazz (BD Biosciences). Cells were scanned for propidium iodide (PI; BD Pharmingen; excitation, YG561 laser; detection, BP585/29 filter) and 488 signal (excitation, B488 laser; detection, BP513/17 filter). The same voltage parameters were maintained in all the samples of an experiment, and adjusted for each experiment. Cell contour plots (created with BD Sortware software) were used for data representation. FACS was performed using a FACS Jazz, with a 100 µm nozzle. Cells were scanned for PI (excitation, YG561 laser; detection, BP585/29 filter) and EdU-488 signal (excitation, B488 laser; detection, BP513/17 filter) to collect PI−/EdU+ and PI−/EdU− cell populations. Sorting pressure was always 27 PSI. Cell contour plots (created with BD Sortware software) were used for data representation.
Definition of cell populations, hierarchies and gating strategies
The following definitions of the several total cell populations (P) that were gated in the samples (Fig. S1) were used: the P2 population comprised all cells of an adequate size [defined by high forward scatter (FSC)] and complexity [defined by low side scatter (SSC)] in the sample (Fig. S1A); the P6 population comprised all singlets (individualized cells) present in the sample (Fig. S1B); the P3 population comprised all dead cells (positive for PI at 585/29-DsRed/PE) present in the sample (Fig. S1C); the NOTP3 population was formed by all the live cells (negative for PI at 585/29-DsRed/PE) present in the sample (this population was defined by a Boolean gating of P3; Fig. S1C); the P4 population comprised all CFSE+ or EdU+ green cells (positive at 513/17- FITC/GFP) present in the sample (Fig. S1D); the P5 population comprised all CFSE− or EdU− cells (negative at 513/17- FITC/GFP) present in the sample (Fig. S1D).
In order to analyze and isolate specific cell types, parental populations were defined by establishing a strict cell population hierarchy during FC and FACS procedures: (1) the P2 population was selected, excluding cell debris and nuclei; (2) the P6 parental population was defined by selecting the singlets among the P2 population; (3) the P3 parental population was defined by identifying the dead cells (PI+) among the P6 parental population; (4) the NOTP3 parental population was defined by excluding the P3 parental population from the analysis; (5) the P4 parental population was defined by selecting the green fluorescent cells among the NOTP3 parental population; (6) the P5 parental population was defined by selecting the non-green fluorescent cells among the NOTP3 parental population. All the hierarchy and parental populations are detailed in Fig. S1E,F.
The threshold for defining P4 and P5 populations was determined through a negative control. Additionally, in the absence of an appropriate isotype control to exclude potential autofluorescence from CFSE and EdU, and with the aim of selecting those events that had only undergone one division, exclusively the brightest 10% of the green fluorescence events were defined as the P4 population.
CFSE-mediated isolation of birthdated chick embryonic neurons
For the FlashTag method experiments (Govindan et al., 2018; Telley et al., 2016, 2019), E4 chick embryos were intraventricularly injected with 1-10 mM CFSE. Three days later, brains were extracted and dissociated. The obtained cell pellet was resuspended in 500 µl of sorting buffer (see details in Table S1), filtered through a 50 µm cell strainer and analyzed by FC. Analyzed samples are detailed in Table S2.
Optimization of time, temperature and reagent concentration for EdU detection
E7 chick embryos were intravenously injected with EdU and one day later (at E8) brains were extracted and dissociated. After dissociation, different protocols were used for EdU detection (see details in Table S3). After EdU detection, cells were washed by centrifugation at 200 g for 10 min and resuspended in 500 µl of HBSS. The washing step was repeated twice. Finally, cell pellets were resuspended in 500 µl of sorting buffer, filtered through a 50 µm strainer and analyzed by FC.
Optimization of the copper concentration in the EdU detection reaction cocktail
The optimal CuSO4‧5H2O concentration for EdU detection was identified in an experiment performed on E7 and E8 chick embryonic brains that had been intravenously injected with EdU 24 h before. Birthdated brains were extracted and dissociated. After dissociation, pelleted cells were resuspended in EdU detection reaction cocktail, which contained variable copper volumes (see details in Table S4). After EdU detection, samples were washed twice with 500 µl of HBSS and resuspended in 500 µl of sorting buffer. Finally, samples were analyzed by FC after being filtered through a 50 µm strainer.
Validation of the EdU detection reaction cocktail
With the goal of checking whether the EdU detection protocol was interfering with gene expression, E4 chick embryos were intravenously injected with EdU. Two days later, brains were extracted, dissociated and incubated in the new EdU detection cocktail at RT for 30 min. Next, cells were washed thrice by centrifugation at 200 g for 10 min and resuspended with HBSS. Finally, samples were passed on a 50 µm cell strainer, and all the viable PI− singlets with the size and shape of interest, regardless of being EdU+ or EdU− (P1 parental population), were isolated by FACS. The P1 population was collected in Buffer NR (NZYTech) containing 1% β-mercaptoethanol (Sigma-Aldrich) and stored at −80°C until processing. Analyzed samples are detailed in Table S5.
Isolation of birthdated neural cells
Both immature neurons (birthdated 3 days prior to isolation) and neuronal progenitors (birthdated 30 min before the extraction) were isolated. Briefly, E7 chick brains were extracted and dissociated (see details in the ‘Tissue dissociation and sample preparation’ section). Next, EdU was detected by incubating the samples in the new EdU detection cocktail for 30 min at RT. Samples were then washed thrice by centrifugation at 200 g for 10 min and resuspended in HBSS. Finally, clogs were removed by filtering the sample through a 50 µm strainer and EdU+ viable cells were sorted by FACS. Isolated cells were collected in Buffer NR containing 1% β-mercaptoethanol and stored at −80°C until processing. Analyzed samples are detailed in Table S5.
Isolation of birthdated cardiac and limb cells
Birthdated cardiac and limb cells were obtained from E7 chick embryos that were intravenously injected with EdU at E4. Tissue was collected on ice-cold HBSS, dissociated and samples were incubated in the new EdU detection cocktail for 30 min at RT. Next, samples were washed three times by centrifugation at 200 g for 10 min, resuspended in HBSS and filtered. Finally, samples were analyzed by FC. Analyzed samples are detailed in Table S5.
RNA isolation, retrotranscription and qPCR
FACS cell RNA isolation and retrotranscription
RNA from FACS-derived EdU+ cells (neural cells >650,000) was isolated using the NZY Total RNA Isolation kit (NZYTech) according to the manufacturer's instructions. RNA was quantified in QubitTM 4 Fluorometer (Invitrogen) with the Qubit® RNA HS Assay Kit (Invitrogen).
RNA was retrotranscribed using the NZY First-Strand cDNA Synthesis Kit (NZYTech) following the manufacturer's instructions in a Veriti Thermal Cycler (Applied Biosystems).
RT-qPCR
RT-qPCR was performed following MIQE guidelines (Bustin et al., 2009) on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad). Three replicates of 1.5 µl of each cDNA were amplified using NZYSpeedy qPCR Green Master Mix (2×; NZYTech). The amplification protocol was 3 min at 95°C for denaturing; and 45 cycles of 10 s at 95°C and 30 s at 60°C for annealing. The expression level ratios of target genes to housekeeping genes were calculated, and these ratios were normalized to the expression levels of the non-treated control group or to the expression levels of the progenitor group.
Primers
Primers (Sigma-Aldrich) were designed to amplify exon–exon junctions using PrimerBlast (NIH) to avoid amplification of contaminating genomic DNA, and their specificity was assessed using melting curves. Amplification efficiency was calculated for each pair of primers using the software LinRegPCR (Ramakers et al., 2003; Ruijter et al., 2009). Primer sequences are listed in Table S6.
Two independent reference genes were compared (GAPDH and RPL7) and their expression remained constant independently of time and treatments, validating their use as reference genes. In all experiments, the pattern of mRNA expression was similar using the assigned couple of reference genes, and in each experiment the reference gene that rendered lower intragroup variability was used for statistical analysis.
scRNA-seq
Cell preparation
We applied BirthSeq for the isolation of birthdated cells on P3 mice injected with EdU at E12.5. Briefly, at P3 brains were extracted in ice-cold HBSS, the pallial region was microdissected, and meninges were removed under a stereomicroscope, collected in ice-cold HBSS, pooled and cut into small pieces. Next, tissue was incubated in full mouse enzymatic solution (see Table S1) for 15 min at 37°C. Following enzymatic digestion, tissue was manually homogenized by carefully pipetting. Then, tissue clogs were removed by filtering the cell suspension through a 40 µm nylon strainer to a 15 ml Falcon tube containing 4 ml of 25% fetal bovine serum in HBSS. Dissociated cells were then centrifuged at 200 g for 10 min at RT, and pellets were resuspended in 500 µl of the new EdU detection reaction cocktail (see Table S1) and incubated at RT for 30 min in the dark. Following reaction, cells were washed thrice by centrifugation at 200 g for 10 min and suspended in 1 ml of HBSS then passed on a 50 µm nylon strainer. PI−/EdU+ cells, gated to include only the top 10% brightest cells, were finally analyzed by FC and sorted by FACS.
FACS-isolated cells were centrifugated at 200 g for 10 min and resuspended in HBSS, at a concentration of 1200 cells per µl. Cell concentration and viability was verified using a TC20 Automated Cell Counter (Bio-Rad). Next, cells were processed for single-cell GEM (gel beads in emulsion; reaction vesicle) formation following the standard Chromium Single Cell 3′ v3.1. Briefly, cells were placed at the Chromium chip with the beads and reagents for the RNA capture and cDNA amplification. After obtaining the emulsion, cDNAs were amplified, purified and quality control checked (visualized and quantified by Bioanalyzer 2100). Post-amplified cDNAs, labeled with individual cell barcodes, were fragmented, ligated to sequencing adapters and amplified with dual indexes. After that, the PCR product was purified, quantified using Qubit 2.0 and the profiles visualized in a Bioanalyzer 2100.
The 10x libraries were sequenced in a HiSeq 4000 and a NovaSeq 6000 for an approximated 50,000 reads per cell. When aiming to identify subtle difference among closely related cell types, it is important to perform deep sequencing of the libraries.
Single-cell pre-processing
10x CellRanger v6.0.2 was employed for alignment and demultiplexing of FASTQ files to obtain feature-barcode matrices. The genome used as reference was mm10-2020-A. Afterwards, data matrices were imported to R (v4.1.0), where Seurat (v4.1.0) (Hao et al., 2021) was employed to further analysis, as described in their vignettes (https://satijalab.org/seurat/).
The cell-cycle phase was determined by CellCycleScore() function, using ‘Rrm2’, ‘Pcna’, ‘Slbp’, ‘Wdr76’, ‘Mcm5’ as S-phase genes and ‘Cenpf’, ‘Tpx2’, ‘Hmgb2’, ‘Ube2C’, ‘Bub1B’, ‘Top2A’, ‘Cenpe’, ‘Tacc3’, ‘Bub1’, ‘Aurka’, ‘Cdc20’ as G2M genes, and the mitochondrial percentage of each cell was calculated with PercentageFeatureSet(pattern= ‘^MT-‘). A manual identification of poor-quality clusters was performed and those cells were excluded for next steps. Samples of equal temporal stages were merged into one Seurat object were subjected to normalization and scaling, and regress-out cell-cycle variation.
Cluster identification
To group cells by transcriptome similarity, expression data were linearly reduced into principal components (‘RunPCA’, default parameters). A shared nearest neighbor (SNN) graph was calculated from the first principal component, which explained almost 90% of the variability, or the percentage variability explained by the next PC, which was less than 5%, with ‘FindNeighbors’, default parameters. Based on this graph, the Louvain algorithm was used to detect communities or clusters with multi-level tuning of the resolution parameter (‘FindClusters’, resolution 0.05, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.8, 1.0, 1.2, 1.8, 2.4). This resolution varied from lower values, to identify general cell types (e.g. ‘glutamatergic neurons’), to higher values, to identify cell subtypes (e.g. ‘MGE-derived GABAergic interneurons’). Finally, cells were represented into a two-dimensional space by the non-linear dimensional reduction technique uniform manifold approximation and projection (‘RunUMAP’) and also t-distributed stochastic neighbor embedding (tSNE).
To identify the neurobiological cell identity of each cluster, differential expression analysis was carried out among clusters (FindAllMarkers, min.pct=0.25, logfc.threshold=0.25). Scientific literature, in situ hybridization databases (Allen Brain Institute, Mouse Genome Informatics) and single-cell equivalent experiments from the literature (Bandler et al., 2022; Di Bella et al., 2021; Li et al., 2020; Loo et al., 2019; Telley et al., 2019) were used to assign cell type identities. As previously stated, there was a first identification step at low resolution, whereby neurons and neural progenitors were subsetted as cells of interest, followed by a high resolution step, whereby only cells of interest were classified into more specific subtypes. The genes used for the first and second assignments are displayed in several heatmaps and ‘FeaturePlots’ (Fig. 5).
ISS
ISS library preparation
Slides with tissue sections were thawed and brought to RT for 5 min, then washed twice at RT with PBS and progressively dehydrated with a 70% ethanol bath for 2 min, followed by a 100% ethanol bath for 2 min, and finally air-dried. Secure-Seal hybridization chambers (Grace Biolabs, 621502) were attached to the microscope slides to cover the samples, and filled with PBS-Tween 0.5%, followed by a PBS wash. The sections were fixed in fresh 3% PFA in PBS for 5 min, followed by three PBS washes. The tissue was permeabilized for 5 min in 0.1 N HCl and washed twice with PBS.
For better discrimination of cell types in the brain section, the selection of genes to be assessed by ISS was led by a previous scRNA-seq experiment. We then chose significant genes, based on their specific and high expression, for the differentiation of the several cell types. A probe solution was prepared according to the following recipe: 2× SSC, 10% formamide and 10 nm of each padlock probe. The sequence for all the probes can be found in Table S7. Samples were incubated in this solution overnight at 37°C. The next day, two washes with 10% formamide in 2× SSC, followed by two washes in 2× SSC were carried out to remove unhybridized excess probes. After removing the last SSC wash, a ligation mix was prepared as follows: 50 mM Tris-HCl (pH 7.5), 1 mM DTT, 100 µM ATP, 2 mM MgCl2, 50 nM RCA primer, 1 U/µl RiboProtect (Blirt, RT35) and 0.5 U/µl T4 RNA ligase 2 (NEB, M0239L). The ligation mix was introduced to the SecureSeal chamber and samples were incubated for 2 h at 37°C. After ligation, the samples were washed twice with PBS and an amplification mix was prepared as follows: 50 mM Tris-HCl (pH 8.3), 10 mM MgCl2, 10 mM (NH4)2SO4, 5% glycerol, 0.25 mM dNTPs, 0.2 µg/ml BSA and 1 U/µl Phi29 polymerase (Blirt, EN020). The amplification reaction was carried out overnight at 30°C. The next day, the samples were washed three times with PBS and L-probes (or bridge probes) were incubated at 10 nM each for 30 min in 2× SSC and 20% formamide. Excess probes were washed out with two washes in 2× SSC, and detection oligos and DAPI were incubated for 30 min in the same conditions as L-probes. Excess detection oligos were washed out with two washes in 2× SSC. TrueBlack (Biotium, 23007) was applied to quench background fluorescence, according to the manufacturer's instructions. Samples were mounted in SlowFade Gold (Thermo Fisher Scientific, S46936), and cyclical imaging was performed. After each cycle of imaging, L-probes and detection oligos were stripped with two washes of 3 min in 100% formamide, followed by five washes in 2× SSC. Hybridization of L-probes and detection oligos for the following detection cycle was performed as above.
ISS image acquisition
Imaging was performed using a standard epifluorescence microscope (Leica DMI6000) connected to an external LED source (Lumencor® SPECTRA X light engine). Light engine was set up with filter paddles (395/25, 438/29, 470/24, 555/28, 635/22, 730/40). Images were obtained with a sCMOS camera (2048×2048, 16-bit, Leica DFC90000GTC-VSC10726), automatic multi-slide stage, and Leica Apochromat 40× (HC PL APO 40×/1.10 WATER, 11506342) objective. The microscope was equipped with filter cubes for five-dye separation (AF750, Cy5, Cy3, AF488 and DAPI) and an external filter wheel (DFT51011).
Each region of interest (ROI), corresponding to a hemisphere of each one of the sections, was marked and saved in Leica LASX software for repeated imaging. Each ROI was automatically subdivided into tiles, and for each tile a z-stack with an interval of 0.5 μm was acquired in all the channels. The tiles were defined to have a 10% overlap at the edges. The images were saved as thousands of individual tiff files with associated metadata.
EdU staining
At the end of the last ISS detection cycle, we stripped all the L-probes and detection oligos by two washes in 100% formamide, followed by five washes of 2× SSC. We then proceeded to the click-labeling of EdU with an Alexa Fluor 488 dye, using the Click-iT EdU Cell Proliferation Kit for Imaging (Thermo Fisher Scientific, C10337), following the manufacturer's instructions. Imaging was performed as described above, and the images were processed as if they were a normal ISS cycle.
Image data processing
The raw images and the associated metadata from the microscopes were fed into the pre-processing module of a custom analysis pipeline (Lee et al., 2023 preprint). The module transforms the images into a format suitable for decoding, executing the following steps. First, the images were maximum z-projected. The resulting 2D projected images (‘tiles’) were simultaneously stitched and registered across imaging cycles, using ASHLAR (Muhlich et al., 2022) software. ASHLAR captures the metadata and places the tiles correctly in the xy space before starting the alignment step. During the process, the 10% overlap was also removed to produce stitched images. Finally, the aligned stitched images were sliced again into smaller tiles, to allow faster and computationally efficient decoding. The resliced aligned tiles were denoised using Content-Aware Image Restoration (CARE) (Weigert et al., 2018), applying a custom denoising model, trained in the lab specifically on ISS data. The resulting images were then converted into the SpaceTX format (Axelrod et al., 2021), and fed into the Starfish Python library for decoding of image-based spatial transcriptomics experiments (https://github.com/spacetx/starfish). Here, the images were normalized across channels and imaging cycles, a spot detection step was performed and, for each detected spot the intensity across all channels and cycles was extracted. The color sequence of each spot across cycles was matched to a decoding table that associates a color sequence with a specific gene. The output of this decoding is a csv table, in which each row represents a detection spot with different properties (x and y position, gene identity, quality metrics). The quality metrics for each spot was computed as follows. For each spot, the normalized fluorescence intensities across all channels were extracted. The prominent channel was considered the ‘true signal’, and all the others were considered ‘background’. The score is described by the formula ‘true signal’/(‘true signal’+‘background’), and it has a theoretical maximum of 1 (perfect decoding) and a theoretical minimum value of 0.25 when decoding in four colors, which corresponds to a random assignment (i.e. all the channels have the same fluorescence intensity for that spot). The quality score for each spot was computed per cycle, allowing two parameters to be calculated: (1) the average quality across all cycles, and (2) the minimum quality across cycles. We found that filtering according to a minimum quality produces more reliable data, and we normally used a minimum quality value of 0.5.
Cell segmentation, clustering and EdU analysis
To identify cells, we used DAPI staining of nuclei. We applied the ‘2D_versatile_fluo’ model from StarDist (Schmidt et al., 2018) to the DAPI images, in order to segment all the nuclei into ROIs. We manually checked the segmentation mask to exclude the possibility of frequent segmentation artifacts. We then expanded each ROI of the mask, allowing each cell to ‘grow’ isometrically for 20 pixels in each xy direction, unless the expansion clashed with a nearby expanding cell. We then assigned all the reads contained within the expanded ROIs to each cell, and saved the resulting matrix into an anndata object for downstream analysis. We then clustered the segmented cells using Scanpy (Wolf et al., 2018), performing dimensionality reduction followed by Leiden clustering at different resolutions. We then critically assessed the spatial distribution of the inferred cell types under different clustering resolution, and when satisfied we extracted the most informative markers for each cluster and assigned it an identity based on available knowledge. We finally went back to the non-expanded DAPI mask and applied it onto the EdU image, extracted the intensity values for each pixel in every ROI, and computed the median EdU intensity for each ROI. We then filtered all the ROIs for which the median was above an arbitrary threshold that labeled only the brightest EdU cells across the tissue, and labeled them in the anndata object as EdU+ cells for downstream analysis.
Acknowledgements
We thank the ISS unit at SciLifeLab for the provision of some of the ISS imaging data. We are grateful for support from the Acucarro core facilities for specific imaging and FAC sorting.
Footnotes
Author contributions
Conceptualization: E.R.-A., M.G., F.G.-M.; Methodology: E.R.-A., E.V., S.M.S., R.S.-G., L.E., A.Q., A.B., A.M.A., F.G.-M.; Software: M.G., S.M.S.; Validation: E.R.-A., M.G., F.G.-M.; Formal analysis: E.R.-A., E.V., S.M.S., R.S.-G.; Investigation: E.R.-A., M.G., F.G.-M.; Resources: M.G., A.D., J.M.E., M.N., F.G.-M.; Writing - original draft: E.R.-A., F.G.-M.; Writing - review & editing: E.R.-A., M.G., E.V., S.M.S., A.M.A., F.G.-M.; Visualization: E.R.-A., M.G., S.M.S., R.S.-G., F.G.-M.; Supervision: M.G., F.G.-M.; Project administration: F.G.-M.; Funding acquisition: N.B.-V., J.M.E., M.N., F.G.-M.
Funding
E.R.-A. holds a predoctoral fellowship from the Basque Government (Eusko Jaurlaritza). During the duration of this research, F.G.-M. holds and held an Ikerbasque, Basque Foundation for Science Research Fellowship, grants from the Spanish Ministry MICNN (Ministerio de Ciencia, Innovación y Universidades) (PGC2018-096173-A-I00 and PID2021-125156NB-I00), grants from the Basque Government (PIBA 2020_1_0057 and PIBA_2022_1_0027) and an EASI-GENOMICS 3rd TNA grant (PID14596). The work in M.N.’s group is supported by funds from the Chan Zuckerberg Initiative, an advised fund of the Silicon Valley Community Foundation; the Erling-Persson Family Foundation (Familjen Erling-Perssons Stiftelse; a human developmental cell atlas); the Knut and Alice Wallenberg Foundation (Knut och Alice Wallenbergs Stiftelse; KAW 2018.0172); the Swedish Research Council (Svenska Forskningsrådet Formas; 2019-01238) and the Swedish Cancer Society (Cancerfonden; CAN 2021/1726). J.M.E.'s group is funded by MINECO (Ministerio de Economía y Competitividad)/MICINN (SAF-2015-70866-R; with FEDER Funds), MICINN (PID2019-104766RB), and the Basque Government (PIBA_2021_1_0018).
Data availability
scRNA-seq data have been deposited in Gene Expression Omnibus under accession number GSE270283.
Peer review history
The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202429.reviewer-comments.pdf
References
Competing interests
M.N. is advisor for the company 10x Genomics. M.G. and S.M.S. are co-founders of Spatialists.