Neurons Generated by Mouse ESCs with Hippocampal or Cortical Identity Display Distinct Projection Patterns When Co-transplanted in the Adult Brain.

The capability of generating neural precursor cells with distinct types of regional identity in vitro has recently opened new opportunities for cell replacement in animal models of neurodegenerative diseases. By manipulating Wnt and BMP signaling, we steered the differentiation of mouse embryonic stem cells (ESCs) toward isocortical or hippocampal molecular identity. These two types of cells showed different degrees of axonal outgrowth and targeted different regions when co-transplanted in healthy or lesioned isocortex or in hippocampus. In hippocampus, only precursor cells with hippocampal molecular identity were able to extend projections, contacting CA3. Conversely, isocortical-like cells were capable of extending long-range axonal projections only when transplanted in motor cortex, sending fibers toward both intra- and extra-cortical targets. Ischemic damage induced by photothrombosis greatly enhanced the capability of isocortical-like cells to extend far-reaching projections. Our results indicate that neural precursors generated by ESCs carry intrinsic signals specifying axonal extension in different environments.

Introduction

In recent years, the use of pluripotent stem cells has allowed the production of neurons with specific identities. This has been made possible by the modification of existing methods of in vitro neuralization and patterning of pluripotent cells, by fine-tuning the signaling pathways that normally orchestrate the acquisition of distinct types of neuronal identities during embryonic brain development (Hansen et al., 2011, Lupo et al., 2014). The ability to obtain virtually any particular type of neuronal identity starting from pluripotent cell cultures has generated new expectations of feasible and reliable protocols of neuronal cell transplantation for the potential treatment of many different neurodegenerative diseases. In fact, neurons suitable for transplantation must be able to integrate into the host tissue, produce the appropriate type of neurotransmitter and neurotransmitter receptors, and develop functional synapses with the host neurons. All these capabilities are normally displayed by in vitro produced neurons (Espuny-Camacho et al., 2013, Michelsen et al., 2015, Yu et al., 2014). However, a crucial requirement for successful transplants is the ability of transplanted neurons to generate specific connections with functionally relevant targets.

So far, the regional identity of the neurons produced in vitro through the neuralization of pluripotent cells has mainly been established by their molecular characterization through variable degrees of analysis of their gene expression, ranging from the simple study of their neurotransmitter phenotype (Eiraku et al., 2011, Shi et al., 2012, Shiraishi et al., 2017, Yu et al., 2014) to a deeper investigation of their molecular nature by methods of global gene expression analysis (Bertacchi et al., 2013, Bertacchi et al., 2015a, Bertacchi et al., 2015b, Edri et al., 2015, Espuny-Camacho et al., 2013, Van de Leemput et al., 2014, Yao et al., 2017). Even so, ascertaining the identity of a nerve cell produced in vitro by comparison of its global gene expression profile with that of neurons in vivo is very useful but not sufficient. Indeed, the expression of markers of different positional identities in the CNS often depends on the developmental time of the analysis, thus making a given combination of markers specific to a type of neural cell only in a narrow time window.

A crucial aim for cell replacement protocols is the ability to produce the wanted type of neural cell to be replaced. The molecular identity of a neural cell by itself might not be predictive of its ability to extend appropriate projections and contact the right targets once transplanted in vivo, and the adult brain is the favorite host structure for assaying this ability in transplantation experiments. Isocortex and hippocampus are main targets of neurodegenerative diseases, thus making them attractive candidates for cell replacement studies (Iqbal et al., 2015, Poewe et al., 2017). The hippocampus is a natural niche of adult neurogenesis and a source of newly formed neurons that are capable of forming new connections in adult brain (Eriksson et al., 1998, Van Praag et al., 2002). This makes hippocampus the region of choice to assay the ability of in vitro produced neural cells to make projections and to send them to appropriate targets. Eventually, the similarity of the isocortex and hippocampus in terms of developmental origin makes the isocortex an ideal brain structure to be compared with hippocampus in transplantation studies.

In this work, we assayed the differential capability of neural cells obtained in vitro, which showed a molecular identity of precursor cells of isocortex or hippocampus, to generate long-range projections after transplantation in adult healthy or damaged brain. By a fine-tuning of Wnt and bone morphogenic protein (BMP) signaling during the in vitro differentiation of mouse embryonic stem cells (ESCs), we obtained neural precursor cells with global gene expression profile clustering with the profile of embryonic hippocampal or isocortical cells. When transplanted in adult healthy hippocampus, only hippocampal-like cells were able to extend long-range projections from the site of transplantation, contacting target regions that were appropriate for hippocampal neurons. Instead, when transplanted into healthy or damaged isocortex, isocortical-like cells were also capable of extending both cortical and extra-cortical far-reaching processes. Our study indicates that the molecular identity acquired in vitro by neuralized ESCs dramatically affects their ability to form projections when transplanted in distinct brain regions.

Results

Timely Manipulation of Wnt and BMP Signaling during Mouse ESC Neuralization Generates Neural Precursor Cells with a Molecular Isocortical or Hippocampal Identity

Wnt and BMP signaling profoundly affects the fate of prosencephalic cells. In fact, during development, their repression is first required for acquiring a dorsal telencephalic identity. Subsequently, the dorsal midline of the telencephalic vesicle invaginates, forming the median wall of the hem and the choroid plexus (Figure S1A). Secreted Wnt factors from the hem are necessary for establishing the hippocampal identity in the adjacent presumptive cortex (Lee et al., 2000, Machon et al., 2007). Therefore, we assayed the effect of inhibiting or activating the two signaling pathways during defined time windows of the ESC neuralization protocol (Figure 1A; DIV, days of in vitro neuralization).

Figure 1

Timely Regulation of Wnt and BMP Signaling Affects the Regional Identity of ESC-Derived Neurons

(A) Scheme of the ESC differentiation protocol. DIV, days of in vitro differentiation.

(B) Schematic drawing of an E9.5 embryo with the domains of expression of A/P markers.

(C) Color map shows the mRNA fold change of the A/P markers in (B), as evaluated by RT-PCR in ESC-derived neurons. n = 3 independent experiments were pooled together and analyzed by qRT-PCR; each experiment contained n = 2 in vitro technical replicates.

(D and E) Global gene expression analysis of embryonic E15.5 regions: isocortex, hippocampus, midbrain, and ventral telencephalon. Explants from three embryos were pooled together. (D) Color map showing hierarchical clustering analysis of the distribution of Pearson correlation among transcriptome subsets of different embryonic brain regions. The subset of transcriptome analyzed consists of the 487 genes with the highest variance among samples (top 2%). Ctx, isocortex; hip, hippocampus; mes, mesencephalon; vt, ventral telencephalon. (E) Pie chart reporting the % of genes with a 3-fold differential expression between isocortex and the other embryonic regions.

Neuralizing ESCs required combined Wnt/BMP inhibition to acquire a dorsal telencephalic molecular identity (Bertacchi et al., 2015b, Yao et al., 2017). This is shown by the relative expression of anteroposterior (A/P) embryonic CNS markers (Figure 1B) in Wnt/BMP double-inhibited cells (WiBi cells) compared with control cells (Figure 1C). Notably, double Wnt/BMP inhibition was no longer required after DIV8 to induce the expression of the telencephalic markers FoxG1 and to repress the expression of the posterior markers Irx3, Otx2, En1, and Krox20. However, the reactivation of Wnt signaling from DIV8 by CHIR treatment (CHIR8 cells, Figure 1C) exerted a significant activation of Emx1 and Emx2 together with low expression of Irx3, Otx2, En1, and Krox20, and significant expression of FoxG1. Accordingly, CHIR treatments beginning before DIV8 (CHIR4–7) posteriorized the identity of cells, while CHIR treatments after DIV8 (CHIR9–10) were less effective in sustaining Emx2 expression. Emx1 and Emx2 are expressed in a similar caudal-rostral gradient in the E9.5 embryonic cortex, both are highly expressed in the developing archicortex, and Emx2 expression is crucial for hippocampal development (Fukuchi-Shimogori and Grove, 2003, Theil et al., 2002, Yoshida et al., 1997). Moreover, CHIR treatment before DIV8 (CHIR5–7) dramatically increased the expression of the markers of hem and choroid plexus Lmx1a and Ttr (Imayoshi et al., 2008, Kuwamura et al., 2005, Millonig et al., 2000), while decreasing the dorsal telencephalic marker Lhx2 (Monuki et al., 2001, Subramanian et al., 2009) (Figures S1A and S1B).

Isocortex and hippocampus are two dorsal telencephalic structures with a common developmental origin and very similar profiles of gene expression. A main feature distinguishing them is the later development of hippocampus and its persistent neurogenesis in the adult. We performed gene expression profiling and compared the global gene expression of mouse E15.5 hippocampus, isocortex, midbrain, and ventral telencephalon. Accordingly, we found higher correlation of hippocampus and isocortex compared with midbrain or ventral telencephalon (Figures 1D and 1E).

A few markers of positional identity are known to be differentially expressed between embryonic hippocampus and isocortex (Abellán et al., 2014) (Figure 2A). The analysis of these markers (Lhx2, Lhx9, Lmo3, Lmo4, and Lef1) suggested a hippocampal rather than a posterior isocortical identity of CHIR8 cells (Figure 2B). In addition, CHIR-treated cells showed expression of the dorsal telencephalic markers TBR1, CTIP2, and SATB2 and TBR2 (Figures 2C–2E and S1C), as well as expression of the neuronal terminal differentiation markers βIII TUBULIN, NEUN, and MAP2 (Figures 2F and 2G). To further characterize the molecular identity of the cells in which Wnt/BMP signaling was continuously inhibited (WiBi cells) or cells in which Wnt/BMP signaling was reactivated from DIV8 (CHIR8 cells), we compared their global gene expression profiles with the profiles of embryonic cells by principal component analysis (PCA) and clustering analysis (Figures 3A and 3B). The second and the third components of PCA discriminated between E15.5 isocortex, striatum, hippocampus, midbrain, and cells treated with WiBi CHIR at different times. While WiBi-treated cells located closer to isocortex and striatum, CHIR-treated cells were closer to hippocampus (Figure 3A). Notably, the most significant loadings of the third component, which better separated hippocampus and isocortex, showed a significant clustering of WiBi and CHIR8 cells to E15.5 isocortex and hippocampus, respectively (Figure 3B). All in all, our results indicate that the differential activation from DIV8 of one single signaling, Wnt, induced a molecular identity of hippocampus (CHIR8 cells) instead of isocortex (WiBi cells).

Figure 2

Wnt Signaling Manipulations Activate the Expression of Isocortical or Hippocampal Markers in Neuralized ESCs

(A) Schematic drawing of an E15.5 coronal section at the hippocampal level, showing the domains of expression of markers differentially expressed between isocortex and hippocampus.

(B) RT-PCR analysis of expression of the markers of hippocampal and isocortical identity shown in (A), in ESC-derived neurons. Color map show log2 mean-centered expression. n = 3 independent experiments were pooled together and analyzed by qRT-PCR; each experiment contained n = 2 in vitro technical replicates.

(C) Percentage of control (0.1% DMSO) and CHIR8 and WiBi cells positive for the dorsal telencephalic markers SATB2, CTIP2, and TBR1. Error bars, SD.

(D–G) Immunocitodetection (ICD) of different neuronal markers in CHIR8 cells. TBR1, CTIP2, and NESTIN: DIV13; βIII TUBULIN, MAP2, and NEUN: DIV25.

Figure 3

Wnt Signaling Induction or Repression Induce, Respectively, Hippocampal or Isocortical Global Gene Expression Profile in Neuralized ESCs

(A) PCA of the first quartile of most differentially expressed genes (n = 3,145) among WiBi cells, CHIR cells, and E15 embryonic regions shown in labels (see Experimental Procedures). n = 3 independent cell cultures or embryonic explants for each treatment/embryonic region were pooled together.

(B) Color map and hierarchical clustering (Pearson correlation) of the top 5% (n = 157) loadings of the third PCA component shown in (A). Clustering scores were computed via multiscale bootstrap resampling (pvclust, see Experimental Procedures). Scale indicates log2 mean-centered gene expression.

Hippocampal Cells, but Not Isocortical Cells, Can Generate Long-Range Projections to Regions Targeted by Adult Dentate Gyrus Neurons, after Transplantation in Adult Hippocampus

We compared the behavior and projection pattern of the neural precursor cells produced via different Wnt/BMP signaling manipulation after grafting in vivo. First, we labeled WiBi and CHIR8 cells by lentiviral transduction with prenylated forms of EGFP and mCherry. Before transplantation, cells were treated with N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester (DAPT) and AraC to force their differentiation and negatively select dividing progenitors, respectively (see Experimental Procedures). The two types of cells were first co-injected into the dentate gyrus (DG) of 2-month-old mice (Figure 4A), because this is a niche of adult neurogenesis (Eriksson et al., 1998, Van Praag et al., 2002), and it is also a permissive environment for the extension of new neuronal processes. Two months after injection, both WiBi and CHIR8 cells persisted in the host tissue (Figure 4B), showing expression of the terminally differentiated neuronal marker NEUN (Figures S2B and S2C). Importantly, 1 month after grafting, the proportions of NEUN/GFP and NEUN/Cherry double-positive cells in the transplant remained unchanged, suggesting an equal survival ratio for both cell types (Figure S2A). While isocortical cells were not able to extend long processes, only CHIR8 cells displayed abundant projections outside the graft, already 1 month after transplantation (Figures 4C–4G). These projections mainly contacted the ipsilateral CA3, CA1, subiculum, and entorhinal cortex (Figures 4C–4G). Fibers appeared to form potential synaptic contacts, as shown by staining with the vesicular glutamate transporter VGLUT1 (Figure S2D).

Figure 4

Cell Transplantation in Hippocampus

(A) Method and site of cell grafting.

(B) ICD of CHIR8 (red) and WiBi cells (green) 2 months post transplantation (mo p.t.) in hippocampus. Blue, Hoechst nuclear counterstaining.

(C–F) ICD of CHIR8 and WiBi fibers at different times post transplantation and in different regions, as indicated by labels.

(G) Density of WiBi and CHIR8 fibers at different distances from the graft, 1 month after transplantation. Error bars, SD.

(H) Heatmap of CHIR8 fiber density in different hippocampal regions 1 and 3 months after cell grafting. n ≥ 3 transplanted animals for each time point were analyzed.

The density of the fibers reaching CA3 was significantly higher than the density of the fibers projecting to other hippocampal structures, especially when analyzed 2 months after grafting (Figures 4D and 4H). These observations were consistent with the capability of naive hippocampal cells to extend new long-range processes contacting CA3 during adult neurogenesis. Notably, 30.1% of transplanted CHIR8 cells expressed CALBINDIN1 (CALB1, Figures S3A and S3B), a marker for DG cells (Sloviter, 1989), 1 month after the transplantation. Moreover, most of the CHIR8 projections found in the Mossy Fiber pathway, which is a natural target of DG (Blackstad et al., 1970, Swanson et al., 1978, Gaarskjaer, 1978, Claiborne et al., 1986), were CALB1 positive (82.0%), while only 8.9% of the projections in CA1 expressed this marker (Figures S3C–S3G). Overall, 88.4% of the CHIR8/CALB1 double-positive projections were found in the Mossy Fibers (Figure S3H). This result suggested that at least a subpopulation of CHIR8 cells displays DG identity and retains the projection pattern of endogenous DG cells.

Isocortical Cells Can Send Far-Reaching Projections When Transplanted in Adult Motor Cortex

When transplanted into adult motor cortex (Figure 5A), both WiBi and CHIR8 cells persisted at least 2 months after injection, and their ratio remained unchanged (Figures 5B and S2E–S2G). As in the case of transplantations in hippocampus, CHIR8 cells were able to efficiently extend projections, either inside the cortex (Figures 5C–5F, 5H, 5I, 5K, and 5L) or toward extra-cortical regions (Figures 5G, 5J, and 5M). WiBi cells transplanted in adult motor cortex were also able to extend processes. However, they generated far-reaching processes less efficiently than CHIR8 cells 1 month after transplantation (Figures 5C–5J). Interestingly, 2 months after grafting, WiBi cells were even more effective than CHIR8 cells in extending processes within the motor cortex (Figures 5K and 5L) and almost as efficient as them in sending axons to extra-cortical targets (Figure 5M). Specifically, while WiBi and CHIR8 cells shared most of the cortical and subcortical targets, some regions, such as the piriform cortex (p < 0.01), the auditory cortex (p < 0.01), the visual cortex (p < 0.05), the entorhinal cortex (p < 0.05), and the hippocampal formation (p < 0.01), were mostly targeted by CHIR8 cells (Figures 5L and 5M), indicating cell-specific axonal outgrowth. We also analyzed other subcortical targets, including the thalamus, midbrain, and corticospinal tract, but we could not detect significant projections for either cell type.

Figure 5

Cell Transplantation in Normal Motor Cortex

(A) Scheme of transplantation.

(B) ICD of CHIR8 (red) and WiBi cells (green) 2 months after transplantation in healthy motor cortex. Blue, Hoechst nuclear counterstaining.

(C–G) ICD of CHIR8 and WiBi fibers at different times post transplantation and in different regions, as indicated by labels.

(H and K) Density of WiBi and CHIR8 fibers at a different distances from the graft, 1 (H) and 2 months (K) after transplantation.

(I, J, L, and M) WiBi and CHIR8 fiber density in cortical (I and L) and extra-cortical (J and M) regions, 1 (I and J) and 2 months (L and M) after transplantation; Error bars, SD. ∗p < 0.05; ∗∗p < 0.01 (two-tailed Student's t test). n ≥ 3 transplanted animals for each time point were analyzed.

Altogether, these results indicated that cortical cells are slower than hippocampal cells in extending far-reaching axons when transplanted into the isocortex.

One important issue is whether the projection patterns of ESC-derived neurons resemble those of hippocampal and cortical cells derived from mouse embryos. To this aim, we co-transplanted fetal (E15.5) cortical and hippocampal neurons into the hippocampus (Figures S4A–S4F) or motor cortex (Figures S5A–S5F) of adult mice. We found that embryonic cells displayed similar patterns of projection compared with ESC-derived neurons, 1 month after transplantation. The main similarity concerned the differential projection pattern after transplantation into hippocampus. Indeed, only embryonic hippocampal, but not cortical cells, extended abundant projections and contacted CA3 via the Mossy Fiber pathway after grafting into the DG (Figures S4D and S4E).

Photothrombotic Damage Dramatically Enhances the Ability of Isocortical Cells to Extend Long-Range Projections

To assay the impact of a brain lesion on the projections of transplanted cells, we grafted WiBi and CHIR8 cells in a mouse model of ischemic stroke in the motor cortex. Cells were injected 3 days after inducing photothrombotic damage (Figure 6A; see Experimental Procedures) and their survival and pattern of axonal connectivity were analyzed 1 and 2 months after stroke (Figures 6B–6N). Both WiBi and CHIR8 cells integrated into the host tissue and were detectable at least 2 months after transplantation (Figures 6B, S2E, S2H, and S2I). As when transplanted in healthy motor cortex, CHIR8 cells efficiently extended axons into both cortical and extra-cortical regions, 1 month (Figures 6H–6J) and 2 months (Figures 6K–6M) after transplantation. However, WiBi cells formed intra-cortical projections more efficiently than CHIR8 cells already 1 month after grafting (Figures 6H and 6I). This was paralleled by an enhanced ability of WiBi cells to target subcortical regions 1 month after transplantation (Figure 6J). However, 2 months after grafting, a refinement of axonal targeting occurred, making CHIR8 projections into septal nuclei and hippocampus more abundant than WiBi projections (Figures 6M and 7). Finally, the different behavior of WiBi and CHIR8 cells was even more striking considering the total amount of fibers in all the regions analyzed (Figure 6N). Interestingly, CHIR8 cells appeared to reach a plateau in projection density already 1 month after transplantation, and the total amount of fibers was not influenced by the ischemic insult, while WiBi cell projections increased with time and were specifically promoted by the lesion. We concluded that the photothrombotic damage further enhanced the process extension of WiBi cells, but CHIR8 and WiBi cells retained their cell-autonomous ability to project toward specific cortical and extra-cortical targets.

Figure 6

Cell Transplantation in the Ischemic Motor Cortex

(A) Method and site of transplantation.

(B) ICD of CHIR8 (red) and WiBi (green) cells 2 months after transplantation in photothrombotic motor cortex. Blue, Hoechst nuclear counterstaining.

(C–G) ICD of CHIR8 and WiBi fibers at different times post transplantation and in different regions, as indicated by labels.

(H and K) Density of WiBi and CHIR8 fibers at different distances from the graft, 1 (H) and 2 months (K) after transplantation.

(I, J, L, and M) WiBi and CHIR8 fiber density in cortical (I and L) and extra-cortical (J and M) regions, 1 (I and J) and 2 months (L and M) after transplantation. Error bars, SD.

(N) Total fiber density of WiBi and CHIR8 cells in healthy and ischemic brains, 1 and 2 months after grafting. (I, J, and L–N) ∗p < 0.05; ∗∗p < 0.01 (two-tailed Student's t test); n ≥ 3 transplanted animals for each time point were analyzed.

(O) Percentage of foot faults made with the contralesional forelimb in the gridwalk test. Values were normalized on the 2 days post stroke value (initial deficit before transplantation). After the injection of WiBi cells (green, n = 9) at day 3, mice showed a significant improvement in the motor performance compared with mice injected with vehicle (black, n = 5) (two-way repeated-measures ANOVA followed by Holm-Sidak test: 16 days, ∗∗p = 0.004; 23 days, ∗p = 0.015). Error bars, SD.

Figure 7

Cell-Autonomous and Environmental Cues Regulate Axonal Extension after Grafting

Schematic showing the different cortical and subcortical regions inspected in the differential projection analysis: the total amount of fibers (size pie chart, log10) and the proportions of CHIR8 and WiBi projections in each region, as reported in Figures 4, 5, and 6. Isocortical and hippocampal cells co-transplanted in intact and ischemic motor cortex show different patterns of projection.

We also asked whether cell grafting had an impact on motor performance in animals with ischemic damage. We chose to implant WiBi cells since they showed better long-term integration than CHIR8 cells in our stroke model (Figures 6N and 7). Thus, mice received WiBi cell transplantation 3 days after experimental stroke, and motor function was assessed longitudinally via the gridwalk test (Lai et al., 2015, Alia et al., 2016). We found that while the motor deficit remained stable in the mice treated with vehicle solution, grafted mice showed a decline (starting from 16 days post stroke) in the number of foot faults made with the contralesional forelimb, indicating functional restoration (Figure 6O). However, we speculate that this very early response did not depend on the specificity of WiBi cell target innervation compared with CHIR8 cell innervation, since the differences between the two types of transplants emerged at later times (see Discussion).

Discussion

We compared the connectivity of two types of neural cells generated in vitro after their transplantation in intact or lesioned regions of the adult mouse brain. Our results indicate that the different molecular identity acquired by the cells after treatment with a single molecule (CHIR99021) in a very precise time window, profoundly affected their connectivity and their ability to integrate into different host environments.

We showed that the timely manipulation of Wnt signaling during mouse ESC neuralization generated neural precursor cells with an isocortical or hippocampal molecular identity. The reactivation of Wnt signaling soon after the expression onset of early dorsal telencephalic markers initiated a hippocampal program of differentiation, while prolonged Wnt/BMP inhibition induced an isocortical identity.

A number of studies focused on the signaling required in vitro to steer the positional identity of neural precursor cells. Both mouse and human ESCs, or induced pluripotent stem cells (iPSCs), expressed markers of dorsal telencephalic identity upon inhibition of Wnt signaling, or combined Wnt/BMP inhibition, during in vitro neuralization (Bertacchi et al., 2015b, Chambers et al., 2009, Espuny-Camacho et al., 2013, Mariani et al., 2012, Shi et al., 2012). However, many of the markers used to ascertain the positional identity of in vitro differentiated cells (e.g., Emx2, Foxg1, Tbr1, Ctip2, Satb2) are co-expressed by cortical and hippocampal precursors or neurons (Bulfone et al., 1995, Fukuchi-Shimogori and Grove, 2003, Gyorgy et al., 2008, Hébert and McConnell, 2000, Nielsen et al., 2010), and the exact identity (isocortical versus hippocampal) of the neurons generated in vitro in these studies could not precisely be established. Gage and colleagues (Yu et al., 2014) showed that the expression of DG markers could be induced in neural precursor cells generated by human iPSCs that were initially deprived of Wnt/BMP signaling and then exposed to exogenous Wnt/BMP. Moreover, they showed that the neurons generated in this way were able to functionally integrate into mouse DG after grafting, suggesting that Wnt/BMP signaling was required to induce a hippocampal identity in precursor cells that were previously committed to a general telencephalic identity. However, a direct comparison of these cells with different cells, focusing on the capacity to integrate and extend neurites in distinct environments, was not performed.

We aimed to precisely identify the molecular identity of our in vitro generated neural progenitor and precursor cells. The analysis of specific markers of hippocampal versus isocortical differentiation, and the comparison of the global gene expression profiles of in vitro differentiated cells and of a number of embryonic regions, allowed us to identify a mechanism that is sufficient to refine the dorsal telencephalic positional differentiation. Our results indicate that in a specific time window, which corresponds to DIV8 of our protocol of neuralization, cells can acquire either an isocortical or a hippocampal molecular identity, depending on the degree of Wnt signaling at that time. Markers that show that graded expression levels between the embryonic hippocampus (Lhx9, Lhx2, Lef1, Lmo4) and isocortex (Lmo3) were differentially expressed by WiBi and CHIR8 cells. Moreover, the global gene expression profiles of WiBi and CHIR8 cells clustered with embryonic isocortex and hippocampus, respectively. Wnt signaling de-inhibition before DIV8, or later than DIV9, failed to induce a hippocampal gene expression profile, indicating the existence of a time window of cellular competence for the induction of hippocampal identity.

It is known that hippocampal cells, together with olfactory bulb cells, are the only CNS cells able to undergo efficient adult neurogenesis in mammals (Eriksson et al., 1998, Van Praag et al., 2002). Conversely, adult cortical cells are not able, in physiological conditions, to generate new neurons, although they share common origin and similar molecular identity with hippocampal cells (Magavi et al., 2000). In this study, we compared WiBi and CHIR8 cells for their ability to generate new processes in different in vivo environments. To this aim, we compared their pattern of connectivity by differentially labeling them with fluorochrome-carrying lentiviral vectors and co-transplanting them into different adult brain regions.

Both WiBi and CHIR8 precursors survived and fully differentiated in mature neurons after grafting in adult hippocampus and isocortex. However, they showed different capacity for axonal projection. In intact hippocampus, only grafted hippocampal cells were able to extend long-range projections, while isocortical cells failed in sending far-reaching processes. At least a subpopulation (about 30%) of CHIR8 cells displayed markers of DG granule cells, such as CALB1, and contacted proper targets of adult DG neurons, such as CA3. This axonal projection became more robust over time, as shown by the dramatic enhancement of fiber density 2 months versus 1 month after transplant (Figure 4H). Finally, the fact that similar results were obtained when transplanting primary cultures of fetal hippocampal or cortical neurons suggests that the different behavior of CHIR8 and WiBi cells might indeed be due to the specific respective hippocampal and cortical identity acquired in vitro.

Likewise fetal cells of the embryonic motor cortex, ESC-derived isocortical cells transplanted into either intact or photothrombotic motor cortex failed in sending projections to thalamic nuclei and midbrain, according to previous observations (Michelsen et al., 2015). In addition, a significant number of processes were found in the internal capsule of photothrombotic brains 1 month after stroke. Based on these observations, we might speculate that our cells acquired a motor identity. However, we cannot exclude that they might simply have maintained a general isocortical identity and that a further specification step would have been required to generate also thalamic and midbrain projections. Their layer identity remains to be assessed, but we speculate that a majority of them differentiated as deep-layer neurons, because at the time of their transplantation they were mostly expressing TBR1 and CTIP2 markers (Figure 2C).

The results obtained after transplantation in photothrombotic motor cortex indicate that the normal cortical environment retains some cues inhibiting the extension of new processes from WiBi cells and that such cues are somehow removed after the photothrombotic damage. However, it is interesting to note that, after a first early phase when WiBi processes were much denser than CHIR8 processes both in cortical and extra-cortical regions 1 month after grafting, 2 months after, the density of CHIR8 fibers in hippocampus and septal nuclei was significantly higher than the density of WiBi processes. We speculate that, in addition to a differential capacity of axonal extension, WiBi and CHIR8 cells hold some ability in selecting specific targets of innervation that is refined over time. This is also supported by the observation that WiBi processes were less dense than CHIR8 processes in healthy motor cortex and callosum 1 month after graft, but they increase dramatically at 2 months. These data suggest that isocortical and hippocampal cells display different degrees of axonal pruning/plasticity in specific brain regions.

Altogether, our observations indicate that the distinct molecular identities acquired by CHIR8 and WiBi cells paralleled their different behavior after transplantation and suggest that WiBi and CHIR8 cells might be very similar to naive isocortical and hippocampal precursor cells, respectively. Furthermore, this interpretation is supported by the similar projection patterns established by fetal hippocampal and cortical neurons co-transplanted under the same experimental conditions as WiBi/CHIR8 cells (see Figures S4 and S5).

Fetal hippocampal cells grafted into the adult mouse hippocampus survived and successfully contacted the correct hippocampal layers and typical hippocampal target regions as the septum and the entorhinal cortex (Shetty and Turner, 1996, Zaman and Shetty, 2001). In addition, mouse and human neurons generated in vitro by pluripotent cells were also successfully grafted into mouse hippocampus (Yu et al., 2014) or hippocampal slice (Hiragi et al., 2017). Consistently, we observed a very similar pattern of connectivity after transplantation of both fetal or ESC-derived neurons in hippocampus.

A number of experiments of grafting into mouse isocortex of either fetal cells or neural precursors originated in vitro by pluripotent cells have been performed. Overall, these studies show that embryonic or fetal neural precursors transplanted in the intact adult hippocampus integrate efficiently in the host circuitry and contact the correct target in the host tissue displaying specific patterns of long-range projections, whereas the intact adult cortical parenchyma appears to be poorly permissive with the transplanted cells (Avaliani et al., 2014, Fricker-Gates et al., 2002, Gage et al., 1995, Guitet et al., 1994, Rosario et al., 1997, Sheen et al., 1999, Shetty and Turner, 1996, Shin et al., 2000). Indeed, this behavior suits well with the natural capability of the hippocampal niche, unlike the cortical environment, to support adult neurogenesis. In addition, several studies reported that experimentally induced cortical damage (e.g., aspiration, experimental ischemia, or chemically induced neurodegeneration) resulted in functional integration and higher connectivity of the transplanted cells, thus suggesting the existence of inhibitory cues in the healthy isocortex removed by the lesion (Espuny-Camacho et al., 2013, Falkner et al., 2016, Gaillard et al., 2007, Michelsen et al., 2015, Tornero et al., 2013). Our results, which are consistent with these observations, indicate that the inhibitory effect of the intact cortical environment affects isocortical but not hippocampal cells, both fetal or originating from ESCs, although the signals mediating such inhibition are currently unknown.

The possibility to successfully transplant neural precursors originated in vitro from ESCs or iPSCs into damaged isocortex has opened new opportunities for therapeutic approaches for cortical stroke. Here, we showed that transplanted cells were able to establish potential synaptic connections with the host, as indicated by axonal varicosities stained with VGLUT1 (Figures S2D and S2J). Moreover, grafting WiBi cells promoted some functional restoration of forelimb function after ischemic damage to the motor cortex. Improvements in motor output were already apparent as early as 16 days after stroke, suggesting that, in addition to network rewiring due to axon extension by the transplanted cells, bystander effects (e.g., release of trophic factors, modulation of inflammation; George and Steinberg, 2015, Lee et al., 2008, Chen et al., 2003, Borlongan et al., 1998, Modo et al., 2002) likely play a key role in the observed recovery. Finally, we speculate that the specificity of innervation that we found for WiBi and CHIR8 cells might affect later processes of motor recovery. These and other functional aspects of cell transplantation in ischemic motor cortex are currently under investigation.

Our results highlight the importance of the type of molecular identity acquired by the cells during their in vitro neuralization to establish proper connections. When transplanted into photothrombotic motor cortex, WiBi cells, more efficiently than CHIR8 cells, sent far-reaching processes toward the surrounding cortical regions (somatosensory, cingulate, prefrontal) and some subcortical regions (internal capsule, caudate putamen, claustrum) typically targeted by resident cortical projection neurons. Although our observations clearly indicate a functional difference between WiBi and CHIR8 cells, we cannot anticipate how their differential connectivity after grafting could relate to their different molecular identity. The analysis of genes differentially expressed between hippocampal and isocortical cells originating from ESCs in vitro showed clusters of biologically related pathways with high enrichment score comprising extracellular matrix (ECM)-cell interaction (the most enriched), cell-cell interaction, PI3K-Akt signaling pathway, and WNT signaling. Notably, ECM-cell interaction is also the most enriched pathway in the analysis of genes differentially expressed between embryonic isocortex and hippocampus (Table S1). We speculate that changed expression of some of the genes belonging to this category could account for the specificity of connectivity of WiBi and CHIR8 cells.

In conclusion, our in vitro system proved useful to finely dissect the molecular pathways generating cell diversity in CNS development. In vivo grafting in distinct adult brain regions revealed an intrinsic functional diversity of neural precursor cells generated in vitro, which might have a crucial impact for cell replacement therapies. Our findings support the importance of in vitro systems in addressing developmental biology issues and pose the bases for more focused assays of isocortical or hippocampal cell replacement.

Experimental Procedures

Cell Culture

Murine ESC lines E14Tg2A (passages 25–38) were cultured as described (Bertacchi et al., 2013, Lupo et al., 2014). For their differentiation, ESCs were seeded on gelatin-coated culture dishes (65,000 cells per cm2; DIV0) and cultured in chemically defined minimal medium (CDMM: DMEM/F12 supplemented with N2/B27; see Supplemental Information for the exact composition) plus 2.5 μM Wnt inhibitor (53AH) and 0.25 μM BMP inhibitor (LDN193189) for 3 days. At DIV3, cells were dissociated and seeded (65,000 cells per cm2) on poly-ornithine and natural mouse Laminin. Cells were cultured for 4 additional days in CDMM plus Wnt/BMP inhibitors. At DIV7, cells were dissociated and seeded (125,000 cells per cm2) on poly-ornithine and Laminin-coated wells. Subsequently, isocortical cultures were kept in CDMM plus Wnt/BMP inhibitors (WiBi 11), whereas hippocampal cultures were grown in CDMM supplemented with CHIR 3 μM (CHIR8) for 4 additional days. At DIV11, DMEM/F12 was replaced with Neurobasal to avoid glutamate-induced excitotoxicity.

Gene Expression

RT-PCR and immunocytodetection (ICD) analyses were carried out as previously described (Bertacchi et al., 2013). An Agilent SurePrint G3 Mouse Gene Expression Array (8 × 60 K, grid ID 028005) was hybridized and analyzed as described (Bertacchi et al., 2015a).

In Vivo Grafting

Animal protocols were reviewed and approved by the Italian Ministry of Health, authorization 739 /2017-PR. The photothrombotic lesion was induced as previously described (Lai et al., 2015, Alia et al., 2016). WiBi- and CHIR-treated neurons were transduced at DIV9, respectively, with a lentivirus carrying a membrane-bound form of GFP or mCherry. Lentiviral vectors were constructed swapping the original GFP in the pWPXLd vector (Addgene number 12258) with, respectively, the prenylated GFP in pME-EGFPCAAX and mCherry in pME-mCherryCAAX (Tol2kit, Kwan et al., 2007). To eliminate undifferentiated ESCs and cycling cells, gamma-secretase inhibitor 10 μM (DAPT) was added to the culture medium on days 14 and 15 and Aracytin 5 μM (AraC) on day 15. At DIV16, ESCs were dissociated and resuspended at a concentration of 150,000 cells/μL in DMEM/F12 supplemented with 10% fetal calf serum. In co-transplantation experiments, WiBi- and CHIR-treated cells were pooled and injected together in equal proportions. A volume of 1 μL of cell suspension was injected in 2-month-old mice. The gridwalk test was employed for the behavioral assessment of motor deficits after ischemia (Lai et al., 2015).

Imaging

Z-stack images (2.5-μm z-step size) of three different sections were acquired with the ZENpro software (Zeiss) and then processed with ImageJ 1.48p software (https://imagej.nih.gov/ij/index.html). Projection density was calculated as the total number of pixels above threshold normalized on the average area (expressed in millions of pixels) of the graft from which the fibers originated.

Detailed experimental procedures are available in the Supplemental Information.

Author Contributions

M.T., M.C., and F.C. designed experiments; M.T. performed cell culture, molecular biology, imaging, and gene expression data computation; I.B. and C.A. planned and carried out in vivo experimental activity and imaging; M.P. set up the cell transplantation protocol; M.D. designed and performed the microarray analysis; I.A. performed the microarray computational analysis; M.C. and F.C. wrote manuscript.