Self-organization of axial polarity, inside-out layer pattern, and species-specific progenitor dynamics in human ES cell–derived neocortex

Significance

Using 3D culture of human ES cells, we show new self-organizing aspects of human corticogenesis: spontaneous development of intracortical polarity, curving morphology, and complex zone separations. Moreover, this culture generates species-specific progenitors, outer radial glia, which are abundantly present in the human, but not mouse, neocortex. Our study suggests an unexpectedly wide range of self-organizing events that are driven by internal programs in human neocortex development.

Abstract

Here, using further optimized 3D culture that allows highly selective induction and long-term growth of human ES cell (hESC)-derived cortical neuroepithelium, we demonstrate unique aspects of self-organization in human neocorticogenesis. Self-organized cortical tissue spontaneously forms a polarity along the dorsocaudal-ventrorostral axis and undergoes region-specific rolling morphogenesis that generates a semispherical structure. The neuroepithelium self-forms a multilayered structure including three neuronal zones (subplate, cortical plate, and Cajal-Retzius cell zones) and three progenitor zones (ventricular, subventricular, and intermediate zones) in the same apical-basal order as seen in the human fetal cortex in the early second trimester. In the cortical plate, late-born neurons tend to localize more basally to early-born neurons, consistent with the inside-out pattern seen in vivo. Furthermore, the outer subventricular zone contains basal progenitors that share characteristics with outer radial glia abundantly found in the human, but not mouse, fetal brain. Thus, human neocorticogenesis involves intrinsic programs that enable the emergence of complex neocortical features.

The mammalian neocortex has a multilayered structure (layers I–VI) (1). The neocortex arises from the neuroepithelium (NE) of the dorsal telencephalon, which evaginates to form a semispherical brain vesicle (Fig. S1A) (2). The dorsocaudal side of the neocortex is flanked by the cortical hem, whereas its ventrorostral side is neighbored by the lateral ganglionic eminence (LGE; striatum anlage) and septum. Layer I [its fetal primordium is called the marginal zone (MZ); Fig. S1B] is qualitatively different from other layers, as this superficial-most layer is mainly composed of Reelin⁺ Cajal-Retzius (CR) cells, which are largely derived from neighboring tissues such as the cortical hem and septum (3) (in the case of human cortex, some Reelin⁺ cells also arise directly from neocortical NE) (4). The rest of the cortical layers are generated with the “inside-out” pattern: the deeper the layer, the earlier the neurons are born from cortical progenitors (Fig. S1B) (5, 6).

A detailed understanding of early human corticogenesis remains elusive because of the limited access to fetal brain tissues. We previously established a 3D culture of mouse and human ES cell (hESC) aggregates that recapitulates early steps of corticogenesis [or serum-free floating culture of embryoid body-like aggregates with quick reaggregation (SFEBq)] (7–9). This method has been also applied to human induced pluripotent stem (iPS) cell culture (10). In this self-organization culture, large domains of cortical NE self-form within a floating hESC aggregate and spontaneously develop ventricular zone (VZ), cortical plate (CP) (mostly deep-layer neurons), and MZ by culture day 40–45. This cortical NE was still immature, mimicking human corticogenesis during the first trimester (Fig. S1C) (7).

Here, using an optimized culture, we revealed unique self-organizing aspects of human corticogenesis. Moreover, the optimized culture generates species-specific progenitors in the outer subventricular zone (oSVZ), called outer radial glia (oRG), which are abundantly present in the human neocortex (11, 12) but rare in the mouse cortex (13, 14). Thus, an unexpectedly wide range of self-organizing events is internally programmed within the cortical NE.

Results

Intracortical Polarity in Self-Organized Cortical NE.

For the improved SFEBq culture (Fig. S2 A and A′), we formed aggregates by plating 9 × 10³ dissociated hESCs into each well of V-bottomed 96-well plates (15) and culture them in medium supplemented with the Rho kinase inhibitor (16) (Fig. S2A). Then, the aggregates were transferred to petri dishes and cultured under 40% O₂ conditions. The addition of chemically defined lipid concentrate, FBS, heparin, and a low concentration of Matrigel improved long-term maintenance of VZ progenitors, whereas the addition of a TGFβ inhibitor and a Wnt inhibitor for the first 18 d moderately promoted telencephalic differentiation.

Under these conditions, all hESC aggregates were positive for foxg1::Venus (marking telencephalic tissue) (2) on days 26 (Fig. 1A and Fig. S2B), and >75% of total cells (day 34) expressed foxg1::Venus, in contrast to the previous culture (Fig. 1B and Fig. S2C). The foxg1::Venus⁺ NE contained epithelial domes with a ventricle-like cavity inside (Fig. 1C; day 42). These thick epithelia had a cell-dense VZ positive for Pax6 and Sox2 on the luminal side (Fig. 1 D and E), whereas pH3⁺ progenitors under mitosis were found exclusively in its innermost part (Fig. 1F), as seen in vivo. TuJ1⁺ neurons occupied the zone outside of the VZ (CP; Fig. 1G), and expressed markers of early CP neurons such as Ctip2 and Tbr1 (1, 2) (Fig. 1H and Fig. S2D). The neuronal zone also contained Reelin⁺ cells (Fig. 1I), and a Laminin-rich zone near the surface (Fig. 1J) (4). Thus, self-organizing lamination occurs in this hESC-derived cortical NE.

Fig. 1.

Axial polarity in cortical NE self-organizes from hESCs. (A) hESC aggregates containing cortical NE visualized with foxg1::Venus on day 26. (B) Representative FACS analysis for foxg1::Venus⁺ populations. Gray, control (day 1 culture); green, day 34 culture under the new conditions. (C–J) Immunostaining of semispherical cortical structures self-formed from foxg1::venus hESCs. VZ, ventricular zone; CP, cortical plate. (K–N) Self-formation of axial polarity seen in hESC-derived cortical NE. Cortical hem-like tissues (Otx2⁺; M) were located in the flanking region of cortical NE on the side strong for the dorsocaudal markers Coup-TF1 (K) and Lhx2. A higher level of pErk signals (bracket) was observed on the side opposite to Coup-TF1 expression (N). Gradient and polarity of expression are indicated by triangles. Arrowhead, VZ (note that the gradients of marker expression are seen in the VZ). (O and P) Fgf8 treatment suppressed CoupTF1 and expanded the expression of the ventrorostral marker Sp8. (Scale bars, 1 mm in A; 200 μm in C–P.) Nuclear counter staining (blue), DAPI.

Interestingly, the cortical NE frequently had an axial polarity. Expression of Coup-TF1 in the VZ (Fig. 1K, red), which forms a dorsocaudal-to-rostroventral gradient in the fetal brain (Fig. S2 E and F), was stronger on one side of the hESC-derived cortical NE, whereas the ventrorostral marker Sp8 was expressed in the reverse pattern (Fig. 1K, white). Consistent with this idea, Lhx2 expression (forming a dorsal-to-ventral gradient of signal intensity in vivo) was also strong on the same side with Coup-TF1 (Fig. 1L and Fig. S2G). The reverse gradient pattern of Coup-TF1 and Sp8 was already observed on day 35. In the mouse embryo, the dorsocaudal cortical area is flanked by the cortical hem (Fig. S2 H–J), which later gives rise to the fimbria of the hippocampus. Consistent with this, the cortical hem markers Otx2 and Zic1 were expressed in the tissue flanking the cortical NE on the side with strong Coup-TF1 expression (Fig. 1M and Fig. S2K).

These findings indicate that hESC-derived NE spontaneously acquires an intracortical dorsocaudal-ventrorostral polarity. In the embryo, FGF8 promotes rostral specification of neocortical NE (17). Interestingly, a higher level of pErk signals (working downstream of FGF signaling) was observed in the hESC-derived cortical NE on the side opposite to Coup-TF1 expression (Fig. 1N). Conversely, treatment with exogenous FGF8 caused broad expression of Sp8 at the expense of Coup-TF1 expression (Fig. 1 O and P and Fig. S2L), suggesting an active role of FGF-MAPK signaling in the rostral specification.

Polarized Curving Morphogenesis of Self-Organized Cortical NE.

Foxg1 was first detected in hESC-derived NE around days 18–20. The apical side (aPKC⁺) of the NE was located on the surface of the aggregate (Fig. 2A, Lower). On day 21, the NE structure started to break into several large domains (Fig. 2A), and subsequently became apically concave (Fig. 2 B–D and Fig. S3A, Upper).

Fig. 2.

Asymmetric rounding morphogenesis in self-organized cortical NE. (A–I) Asymmetric progression of rounding morphogenesis of hESC-derived cortical NE. Arrows, boundary of a cortical NE domain in A and rolling epithelium in B–D. Red arrowheads, rolling epithelium in E. Red arrows, rounding movements of the NE (F–I). (J–L) Effect of the ROCK inhibitor Y-27632 on the rolling of cortical NE. (L) Attenuation of rolling morphogenesis with ROCK inhibitor. ***P < 0.001 in contingency table analysis (2 × 2) with Fisher’s exact test. Treatment group, n = 187 NE domains; control group, n = 130. (M and N) The rolling shape was preferentially observed on the side with strong expression of Otx2 and Coup-TF1 (dorsal and caudal markers). (O–Q) Adjacent formation of NE structures of cortex (Pax6⁺) and LGE (Gsh2⁺; with GAD65⁺ GABAergic neurons underneath) on day 35. The cortical side contacting the LGE domain was opposite to the side with strong rolling (arrow). (R) Interkinetic nuclear migration in the hESC-derived cortical NE on day 24 (two-photon imaging). Visualized with partial mixing of pax6::venus reporter hESCs with nonlabeled hESCs. Two daughter cells with both apical and basal processes were generated from an apically dividing progenitor (red dots). (Scale bars, 200 μm in A; 100 μm in B–H and J–N; 200 μm in O–Q.) Nuclear counter staining (blue), DAPI.

Each cortical NE domain had an asymmetrically curved structure. One end of the NE was characterized by an epithelium with a rolling shape (Fig. 2 B–D, arrows), whereas the other side had a blunt end. Active myosin (indicated by phosho-myosin light chain 2) was highly enriched throughout the apical surface of the cortical domain (Fig. 2C). In live imaging, the rolling side of the cortical domain approached the other end and eventually contacted it (Fig. 2 E and F and Movie S1). The main body of NE moved around in the same direction with the rolling end (Fig. 2 E–H), and generated a semispherical structure with a lumen inside (Fig. 2I and Fig. S3A, Lower).

The rolling morphogenesis of the cortical domain was attenuated by ROCK inhibitor treatment (days 26–30) (Fig. 2 J–L), which inhibits the Rho-ROCK-myosin pathway necessary for causing apical constriction. The rolling side of NE expressed markers for the dorsocaudal side (Otx2 and CoupTF1; Fig. 2 M and N), indicating that the rolling end corresponded to the dorsocaudal side.

When the NE was partially ventralized (18, 19) by a moderate level of Hedgehog agonist (30 nM smoothened agonist or SAG for days 15–21), a substantial portion of foxg1::Venus⁺ NE expressed Gsh2, a marker for LGE (20) (Fig. 2O, arrowhead, and Fig. S3B). A mass of GAD65⁺ GABAergic neurons was generated underneath this LGE NE, as seen in vivo (19) (Fig. 2P, red), whereas the rest of the telencephalic NE was largely positive for the cortical NE marker Pax6 (Fig. 2Q). Higher concentrations of SAG induced the medial ganglionic eminence (MGE) marker Nkx2.1 at the cost of Pax6 and Gsh2 expression (Fig. S3 B and C). Importantly, the NE treated with moderate Hedgehog signals frequently exhibited continuous formation of cortical (Pax6⁺)-LGE (Gsh2⁺) domains, as seen in vivo, suggesting that our improved culture allows self-formation of pallial-subpallial structures en bloc. In this continuously extending NE, the rolling side of the cortical NE (Fig. 2 O–Q, arrows) was opposite to the cortex-LGE junction, consistent with the idea that the rolling and nonrolling sides represent the dorsal and ventral side of the cortical NE, respectively.

In the embryo, the developing cortex evaginates by strong rounding morphogenesis of the pallial NE, whereas the margin of the embryonic pallium is fixed to the neighboring tissues. The curvature of the embryonic NE region from the medial pallium to the dorsal part of the neocortex is particularly strong (Fig. S1A). It is therefore reasonable to infer that the rounding NE movement of the cortical domain in our hESC culture reflects the strong rounding of the embryonic dorsal cortex (Fig. S3D).

These findings demonstrate that the hESC-derived cortical NE self-develops a cortical curvature by asymmetrical rounding morphogenesis along the self-formed dorsocaudal-ventrorostral axis. Following this topological change, the apical surface of the NE becomes located inside of the cortical semispheres. In live imaging, progenitors divided at the luminal surface while they underwent interkinetic nuclear migration, as seen in the embryonic VZ (Fig. 2R, Movies S2 and S3, and Fig. S3E; cell divisions were mostly symmetrical at these stages).

Morphological Separation of Three Cortical Neuronal Zones.

The optimized culture conditions allowed cortical NE to grow even beyond day 42. On day 70, the thickness of cortical NE was 200 µm or larger (Fig. 3 A and A′). The NE was morphologically stratified into the VZ, SVZ, intermediate zone (IZ), CP, and MZ (Fig. 3 B–G and Fig. S4 A and B). The superficial-most portion of the MZ accumulated Laminin and contained Reelin⁺ CR cells (Fig. 3 C and C′). Beneath the MZ was mainly the CP and contained deep-layer cortical neurons expressing Tbr1 and Ctip2 (Fig. 3 D and D′). The population of Satb2⁺ superficial-layer (21) was still relatively small (Fig. 3E). On the apical side, the day 70 VZ was ∼100 µm thick and cell dense with Pax6⁺ Sox2⁺ progenitors (Fig. 3 F and F′) or radial glia (22). Basally adjacent to the VZ, a SVZ formed and contained cells positive for Tbr2 (Fig. 3G).

Fig. 3.

Self-formation of multiple zones in hESC-derived cortical NE. (A) Sections of day 70 hESC-derived cortical NE. Clear separation of VZ (Pax6⁺), SVZ, intermediate zone, and CP (Ctip2⁺) was seen even at this low-magnification view. (B–H) Immunostaining of day 70 cortical NE with zone-specific markers. (I) Total thickness of cortical NE and thickness of ventricular and cortical plate zones on days 70 and 91. **P < 0.01; ***P < 0.001, Student t tests between day 70 and day 91 NE samples (n = 6, each). (J–O) Immunostaining of day 91 cortical NE with zone-specific markers. (P) Schematic of the laminar structure seen in long-term culture of hESC-derived cortical NE. (Scale bars, 400 μm in A; 50 μm in B–H″; 100 μm in J–O.) Bars in graph, SEM. Nuclear counter staining (blue), DAPI.

By this stage, a distinct cell-sparse zone developed between the CP and SVZ, reminiscent of the IZ in the fetal brain. Immediately beneath the CP was a layer of Calretinin⁺ cells with massive MAP2⁺ neurites extending into the IZ (Fig. 3 H and H′ and Fig. S4 C and D). These characteristics resemble those of neurons in the subplate (23–25). Chondroitin sulfate proteoglycans (CSPGs) are enriched in the embryonic subplate and its underlying IZ (Fig. S4F, Lower Right, bracket) (26). Similarly, strong CSPG accumulation was observed in the corresponding zones in hESC-derived cortical NE (Fig. 3H″ and Fig. S4E). These findings demonstrate that hESC-derived cortical NE can self-organize not only the CP and MZ but also the subplate and IZ in a correct apico-basal order. At this stage, no substantial accumulation of GAD65⁺ interneurons in the CP or TAG1⁺ corticofugal axons was observed (Fig. S4G).

By day 91, the cortical NE reached the thickness of 300–350 µm but still contained well-developed VZ (Fig. 3 I–K and Fig. S4 H and I). The CP also became much thicker (∼150 µm; Fig. 3I), and contained a number of superficial-layer neurons (Satb2⁺ and Brn2⁺) in addition to Tbr1⁺ and Ctip2⁺ deep-layer neurons (Fig. 3 L–N and Fig. S4J). The zone of subplate neurons (Calretinin⁺) mainly localized beneath the CP (Fig. 3O).

The morphological zone separation seen in these late cultures (summarized in Fig. 3P) mimics the histology of the human fetal neocortex during early second-trimester stages (25, 27). Moreover, within the hESC-derived CP, superficial-layer neurons (Satb2⁺ and Brn2⁺ cells) preferentially localized more superficially to deep-layer neurons (Tbr1⁺ and Ctip2⁺ cells) (Fig. 4 A–H). Furthermore, when 1-d pulse labeling was done with 5-ethynyl-2′-deoxyuridine (EdU) on day 50 and then with BrdU on day 70, EdU- and BrdU-labeled cells were preferentially located on the deep and superficial sides of the day 91 CP, respectively (Fig. 4 I–L). These findings indicate a biased tendency in the localization of neurons reminiscent of the inside-out pattern during fetal corticogenesis (5, 6), in which late-born CP neurons are located outside and early-born CP neurons are inside. Consistent with this idea, on day 112, the mature cortical neuron marker CaMKIIα was preferentially seen in the apical two-thirds portion of the hESC-derived CP, which predominantly expressed Tbr1 and less Satb2 (Fig. 4 M–O and Fig. S4K). Indeed, at the cellular level, the majority of these CaMKII neurons coexpressed Tbr1 but not Satb2 (Fig. S4 L and M; Fig. 4P for summary).

Fig. 4.

Basally biased localization of Satb2⁺ and Brn2⁺ cortical neurons in CP. (A–H) Cortical neurons positive for Satb2 and Brn2 (superficial-layer markers) were preferentially localized to the basal (superficial) portion of the hESC-derived CP in day 91 culture. Most of the basally located Satb2⁺ cells were negative for the deep-layer marker Tbr1. (H) Distribution of marker-positive neurons within the CP. For relative positions, the apical and basal boundaries of the CP were defined as 0 and 100, respectively. ***P < 0.001. Mann-Whitney test. Red line, median. Counted neurons: Tbr1⁺ (n = 105), Satb2⁺ (n = 58), Ctip2⁺ (n = 87), and Brn2⁺ (n = 86). (I–L) Double-pulse labeling study using EdU (day 50; red; n = 36) and BrdU (day 70; white; n = 53). Analyzed by immunostaining on day 91. Statistical analysis was done as in H. (M–O) The mature cortical neuron marker CaMKII was preferentially expressed in Tbr1⁺ neurons located in the deep portion of the CP on day 112. The cortical NE was cultured on a Transwell filter during days 78–112 to support robust survival of mature neurons. (O) Plotting was done as in H. ***P < 0.001. Kruskal-Wallis test with a post hoc multiple comparison test. Numbers of neurons counted: Tbr1⁺ (n = 293), Satb2⁺ (n = 177), and CaMKII⁺ (n = 132). (P) Schematic of neuronal distributions within the CP of hESC-derived cortical NE on days 91 and 112. (Scale bars, 100 μm in A–C, E–G, and I–K; 50 μm in D; 200 μm in M and N.) Nuclear counter staining (blue), DAPI.

Appearance of Human-Specific Basal Progenitors in the oSVZ.

Previous in vivo studies have reported preferential nonvertical division of apical cortical progenitors at an advanced stage, when many of them produce basal progenitors through asymmetrical divisions (28, 29). In our culture, proliferating apical progenitors on day 70 preferentially divided with a “vertical” cleavage plane (60–90°; Fig. 5 A–C), causing segregation of daughter cells parallel to the apical surface. In contrast, on day 91, proliferating progenitors (phospho-Vimentin⁺) showed a higher frequency of nonvertical divisions (0–60°) (Fig. 5 D–F).

Fig. 5.

Appearance of oRG-like progenitors. (A–F) Percentages of apical progenitors with vertical (cleavage angle at 60–90°) and nonvertical (0–30° and 30–60°) cleavages (A and B) in the VZ of day 70 (C) and day 91 (D–F) hESC-derived cortical NE. p-Vimentin, M-phase marker. Arrowhead, pericentrin. Cells analyzed: n = 42 (day 70) and n = 33 (day 91). (G–I) Basal progenitors (Pax6⁺, Sox2⁺) and intermediate progenitors (Tbr2⁺) in the SVZ of day 91 culture. (H) Percentages of Sox2⁺/Tbr2⁻ and Sox2⁻/Tbr2⁺ progenitors within all progenitors (Sox2⁺ and/or Tbr2⁺) in the CP. The percentage of Sox2⁺/Tbr2⁻ progenitors increased from day 70 to day 91, whereas Sox2⁻/Tbr2⁺ progenitors decreased in proportion. ***P < 0.001, Student t tests between day 70 and day 91 samples. Non-VZ progenitors from four cortical NE domains from each day were counted. (I) On day 91, Sox2⁺/Tbr2⁻ progenitors tended to localize farther from the ventricular surface than Sox2⁻/Tbr2⁺ progenitors (Right). ***P < 0.001, Mann-Whitney test. Red line, median. (J–M) Pax6⁺ p-Vimentin⁺ progenitors had a long basal process extending toward the pia but not an apical process (J and J′), whereas these progenitors were negative for Tbr2 (K and K′). A majority (>70%) of these SVZ progenitors possessing a basal process showed a horizontal type of cleavage angle (60–90°; L and M) (n = 37). (Scale bars, 100 μm in D; 25 μm in E; 50 μm in G, J, and K; 10 μm in L.) Bars in graph, SEM. Nuclear counter staining (blue), DAPI.

Both on days 70 and 91, the SVZ contained a number of Tbr2⁺ Sox2⁻ Pax6⁻ intermediate progenitors (Fig. 3 G and M). Interestingly, on day 91, the outer portion of SVZ accumulated another population of phospho-Vimentin⁺ progenitors that were Tbr2⁻ Sox2⁺ Pax6⁺ (Fig. 5 G–G″ and Fig. S5 A–C). This progenitor population was relatively small in percentage on day 70 and became prominent by day 91 (Fig. 5H). On day 91, this Sox2⁺ Tbr2⁻ cell population was biased to localize more basally, in contrast to the apically deviated location of Sox2⁻ Tbr2⁺ intermediate progenitors (Fig. 5I, Right). Interestingly, these two populations responded differently to Notch signal inhibition, which strongly decreases apical progenitors by inducing precocious neuronal differentiation. The Notch inhibitor treatment (days 70–77) increased Sox2⁻ Tbr2⁺ intermediate progenitors, whereas Sox2⁺ Tbr2⁻ cells rarely remained after the treatment (Fig. S5 D–F).

Recent studies have shown that the oSVZ in the human cortex accumulates a Tbr2⁻ Sox2⁺ Pax6⁺ progenitor population distinct from Tbr2⁺ intermediate progenitors (Fig. S5G) (11, 12). These progenitors, termed oRG (or OSVZ progenitors) (11, 12), have a basal process extending to the pial surface and lack an apical process unlike apical progenitors. Similarly, the Tbr2⁻ Sox2⁺ Pax6⁺ progenitors in the day 91 hESC-derived cortical NE also had a basal process but not an apical process (Fig. 5 J–K′ and Fig. S5 H, H′, and I). These cells had a pericentrin⁺ basal body in the soma located in the SVZ (Fig. S5J), unlike apical progenitors, in which basal bodies are located near the apical surface. Like in vivo oRG, the cleavage plane of the hESC-derived oRG-like cells tended to be horizontal (Fig. 5 L and M). No basal processes were found in Tbr2⁺ progenitors (Fig. S5 K–K″).

Discussion

Our optimized culture allowed robust growth of hESC-derived cortical NE in long-term suspension culture, even beyond 13 wk; eventually, the cortical NE became almost 350 µm thick and contained multiple laminar zones as seen in the fetal cortex at the second trimester (starting from embryonic week 11) (30). This robust growth makes a clear contrast to the limitation of our previous 3D culture, which could support the cortical NE development up to the tissue maturation equivalent to the first trimester cortex. The optimized culture also recapitulated another aspect of second-trimester neocorticogenesis, i.e., the appearance of oRG-like progenitors on day 91 (13 wk) of culture. Thus, the developmental speed in our culture is roughly comparable to that in the fetal brain.

The self-organization shown in this study raises many important questions for future investigation. The self-forming mechanism for the intracortical polarity is an intriguing topic, and it would be also interesting to examine whether the frontal lobe-specific characters may appear in the Fgf8-treated aggregate. How the dorsocaudal domain generates a stronger curvature in the rolling morphogenesis is another important question to be addressed using our 3D culture.

Our culture system is also applicable to the study of the dorsal-ventral specification of the whole telencephalic region. Notably, under the partially ventralized conditions (Fig. 2 O–Q), the hESC-derived NE recapitulated the continuous self-formation of the cortex and LGE (striatum anlage) in adjacent positions as seen in vivo. The origin of human cortical interneurons may be an intriguing topic to be studied using this system, because they are suggested to arise not only from the subplallium but also from the cortical VZ/SVZ (31, 32) in the human fetus.

The optimized culture allowed the emergence of complex separation of cortical zones. The subplate zone is a particularly predominant structure in the fetal primate cortex (also called layer VII), and consists of early-born neurons within the neocortex (e.g., pioneer neurons) (24, 25). Although this zone is only transiently present in the fetal cortex, some of its derivatives exist in the adult brain as interstitial neurons in the white matter (33). Because the subplate disappears postnatally, its investigation is not easy, especially in humans, and thus, our culture system should be useful in studying this little understood neuronal layer. In addition, our system may be applicable to studies of the inside-out pattern formation in the human fetal cortex, including the pathogenesis of lissencephaly.

Thus far, little has been known about the mechanism of how the cortical NE grows in thickness. One possible mechanism is that the distance between the apical and basal surfaces may be gradually widen by the accumulation of neurons and precursors in the CP and SVZ. However, this idea does not seem to go along with the IZ formation in this self-organizing culture, because this zone is low in cell density and also lacks rigid structural components other than radial glial fibers. Therefore, this zone is difficult to transfer mechanical compression. Our observations suggest that the NE thickness is actively controlled by the growth of the radial glia fiber length.

Finally, our culture should also be very advantageous in studying the role of oRG progenitors in human corticogenesis. It is presumably advantageous for the gyrencephalic human neocortex to involve this type of progenitors that keep on dividing multiple times to generate a number of superficial neurons. To date, there are no specific molecular markers reported for demarcating oRG, and the distinction between oRG and apical progenitors (both are Sox2⁺, Pax6⁺, and Tbr2⁻) mainly depends on their cellular morphology, behavior, and location. Therefore, the extent of oRG study has been fairly limited in the case of dissociation culture that lacks the topological context. In contrast, our system provides a great advantage in this respect, because the 3D context of the developing human cortex can be recapitulated. Very recently, after our submission of this report, an independent study from another group also reported a similar observation of the oRG appearance in the stratified cortical tissue generated from human pluripotent stem cells (34), using a nonselective differentiation method based on only stochastic specification of brain regions (unlike our reproducibly cortex-selective differentiation culture). Of note, their study successfully demonstrated the usefulness of 3D self-organizing culture for studying congenital disorders of brain development such as microcephaly.

A future challenge for the self-organization approach in human corticogenesis study is to recapitulate the morphological separation of all neuronal layers (II/III–VI) within the CP zone, which occurs during the third trimester of human gestation (35).

Materials and Methods

Self-Organized Cortical Generation from hESCs.

hESCs were maintained as described previously (16). For cortical NE generation, hESCs were dissociated to single cells and quickly reaggregated using low-cell-adhesion 96-well plates (15) in cortex differentiation medium (9,000 cells per well) containing Glasgow-MEM, 20% Knockout Serum Replacement, and 20 µM Y-27632. IWR1e (Wnt inhibitor) and SB431542 (TGFβ inhibitor) were added to culture to reach 3 and 5 µM, respectively, from day 0 to day 18.

Long-Term Cortical NE Culture.

On day 18, the floating aggregates were transferred to a 9-cm Petri dish (non–cell adhesive) and further cultured in suspension using DMEM/F12 medium supplemented with N2 and Chemically Defined Lipid Concentrate under the 40% O₂/5% CO₂ conditions. From day 35, FBS (10% vol/vol), heparin (5 µg/mL), and Matrigel [1% (vol/vol); growth factor–reduced] were also added to the medium. The tissues were cultured using a lumox dish (SARSTEDT; high O₂ penetration) after day 56. From day 70, the concentration of Matrigel was increased [2% (vol/vol)], and B27 supplement (Invitrogen) was also added to the medium.