Abstract

   Apical-basal polarity (ABP) establishment and maintenance is necessary
   for proper brain development, yet how it is controlled is unclear.
   Galectin-3 (Gal-3) has been previously implicated in ABP of epithelial
   cells, and, here, we find that it is apically expressed in human
   embryonic stem cells (hESCs) during neural induction. Gal-3 blockade
   disrupts ABP and alters the distribution of junctional proteins in
   hESC-derived neural rosettes and is rescued by addition of recombinant
   Gal-3. Transcriptomics analysis shows that blocking Gal-3 regulates
   expression of genes responsible for nervous system development and cell
   junction assembly, among others. Last, Gal-3 blockade during embryonic
   development in vivo reduces horizontal cell divisions, disturbs
   cortical layering of neural progenitors, and induces gyrification.
   These data uncover a regulatory mechanism for ABP in the brain and
   warrant caution in modulating Gal-3 during pregnancy.
     __________________________________________________________________

   Blocking Galectin-3 function during mice pregnancy led to de novo
   cortical gyrification in mouse embryos.

INTRODUCTION

   Apical-basal polarity (ABP) is an important feature in development,
   homeostasis, and disease, partitioning specific proteins into cellular
   locations to support distinct functions ([56]1–[57]4). Early in
   development, ABP of neural progenitor cells regulates morphogenesis of
   the neural tube. Then, ABP regulates horizontal versus vertical cell
   division, delamination of progeny from the apical surface of the neural
   tube, and basal-directed migration. In mammals, vertical cell divisions
   with cleavage furrows parallel to the ventricular surface generate an
   apical daughter cell that remains a stem cell and another daughter cell
   that is more differentiated. This can be a neuron, in direct
   neurogenesis or an intermediate progenitor that usually divides
   symmetrically into two neurons, in a process called indirect
   neurogenesis ([58]5, [59]6). In contrast, cell divisions in which the
   cleavage furrow has an oblique angle generate an apical progenitor cell
   and an outer radial glia (oRG) ([60]7, [61]8). This later cell
   population has been linked with gyrencephalic species and is sparse in
   mice ([62]9). The balance of horizontal versus vertical divisions thus
   controls neurogenesis, making ABP essential for appropriate nervous
   system development.

   ABP perturbation can lead to disruption in the telencephalon causing
   the displacement of neural progenitors in a model of hydrocephaly
   ([63]10). Similarly, decreases in ABP-related proteins led to neuronal
   mispositioning in fragile X syndrome ([64]11). Other disorders, such as
   Down’s syndrome, schizophrenia, and autism spectrum disorders, have
   been related to ABP impairment ([65]12). Thus, ABP is essential for
   healthy brain growth with severe consequences when disrupted ([66]3).

   Many studies have begun to unravel cellular and molecular mechanisms
   important for ABP ([67]1). Polarized exocytosis and membrane insertion
   of proteins is an essential aspect of ABP ([68]1). Cell adhesion is
   strengthened via lateral junction proteins near the apical pole and is
   necessary for ABP function ([69]1). Neuroepithelial cells in the neural
   tube, radial glia (RG), and in vitro neural stem cells (NSCs) derived
   from embryonic stem cells (ESCs) have lateral junctions and apical
   poles that express a variety of proteins including zonula occludens-1
   (ZO-1), Par-3, N-cadherin, and γ-tubulin ([70]1). The main effector of
   the Hippo pathway, the transcriptional coactivator Yes-associated
   protein (YAP), can also be recruited to the junctional complex of NSCs,
   where it is phosphorylated and inactivated, resulting in neuronal
   differentiation ([71]13).

   ABP can be modeled in human ESC (hESC)–derived neural rosettes,
   pinwheel-like cell clusters arising from apical cell constriction and
   basal expansion of NSCs ([72]1, [73]14, [74]15). Thus, in addition to
   molecular ABP, cells can exhibit morphological ABP. For example, RG
   exhibit a distinctive morphological ABP consisting of apical primary
   cilia and long basal processes. In addition, apical versus basal
   positioning of cells establishes tissue level ABP ([75]1, [76]12).
   Canonical RG cell bodies are positioned apically in the ventricular
   zone (VZ) lining the ventricles while oRG cells are basally positioned
   ([77]16). ABP also features in the postnatal ventricular-subventricular
   zone (V-SVZ). RG-like NSCs in this niche have apical primary cilia that
   contact the ventricles and basal processes that contact blood vessels
   ([78]17). Ependymal cells (EC) exhibit another form of ABP in the
   V-SVZ. ECs have apical tight junctions (TJs) and motile cilia that are
   disrupted in Galectin-3 (Gal-3) knockout (KO) mice ([79]18).

   Gal-3 binds to β-galactoside–bearing proteins and regulates many
   functions critical in cancer biology including apoptosis, metastasis,
   and immune surveillance ([80]19). Gal-3 is also important in apical
   trafficking and protein sorting ([81]20–[82]22), and its absence
   promotes aberrant protein distribution, which ultimately impairs
   epithelial morphogenesis in several models ([83]23–[84]27). Gal-3 also
   participates in ciliogenesis and spindle pole organization during cell
   division ([85]28–[86]31). In other systems, it regulates cell-cell and
   cell-matrix interactions ([87]32–[88]34) and facilitates cell growth
   and proliferation ([89]34, [90]35). During embryogenesis, Gal-3 is
   expressed in the trophectoderm ([91]36), notochord, and embryonic
   macrophages ([92]37). Thus, Gal-3 is expressed early during
   embryogenesis, but its functions in the developing brain remain
   unclear.

   Gal-3 is expressed by reactive glia in the injured brain
   ([93]38–[94]40), is generally pro-inflammatory, and regulates
   angiogenesis and chemokine expression during disease ([95]41, [96]42).
   In addition to its well-known functions in disease, we discovered that
   it is necessary for V-SVZ neuroblast migration ([97]18). Gal-3 also
   increases gliogenesis the postnatal V-SVZ, whereas reducing its
   expression reduces gliogenesis ([98]43). Increasing Gal-3 expression
   induced astrogenesis and suppressed oligodendrogenesis, an effect
   dependent on BMP signaling ([99]43). Gal-3 binds to the BMPR1α and
   increases V-SVZ BMP signaling ([100]43). Gal-3 also binds to β-catenin,
   an important component of adherens junctions ([101]44), and Gal-3
   knockdown (KD) increases Wnt signaling ([102]45).

   Gyrification is a complex aspect of fetal cerebral cortex growth
   ([103]3, [104]4, [105]46). Cerebral cortical folding includes the
   formation of ridges (gyri) and valleys (sulci), and it serves to
   increase cortical surface area, allowing this evolutionarily expanded
   part of the brain to occupy less volume and thus to fit into the skull.
   Studies point to differential proliferation of intermediate RG in the
   VZ and subventricular zone (SVZ) as regulating sulci and gyri ([106]47,
   [107]48). In humans, a massively expanded outer SVZ (oSVZ) is thought
   to give rise to the large number of neurons in our cerebral cortex and
   the corresponding gyrification ([108]49). Several other models of
   cortical folding exist, but there is no consensus as to which is the
   most realistic ([109]46, [110]50, [111]51).

   Lower mammalian species such as mice do not exhibit cortical folds but
   instead have lissencephalic cortices (smooth brain). It is unclear
   whether lissencephalic species have mechanisms that inhibit cortical
   folding versus lacking mechanisms that cause gyrification ([112]52).
   Recent work showed that YAP increases proliferation and causes mouse
   cortex gyrification ([113]53).

   Here, we found Gal-3 expression in stem cell niches of human and murine
   embryonic brain tissue, and we explored Gal-3 function during embryonic
   neurodevelopment. Gal-3 was necessary to establish ABP in human-ESC
   neural differentiation and the apical distribution of junctional
   complex proteins was lost by Gal-3 blockers in vitro. Fetal brains from
   female mice treated with a Gal-3 blocker unpacked TJs in the VZ, caused
   early delamination of NSCs and increased vertical divisions.
   Unexpectedly, both pharmacological blockade and genetic deletion caused
   sulci to appear in the developing mouse cerebral cortex.

RESULTS

Gal-3 is expressed apically by hESCs

   Neural induction medium (NIM) induces hESCs to develop into NSCs that
   spontaneously organize into neural rosettes ([114]Fig. 1A). After 6
   days in NIM, rosette cells exhibit ABP with constricted apical poles
   that contact the interior lumen and with expanded basal poles that
   contain nuclei ([115]Fig. 1B). Neural rosettes are characterized by the
   apical localization of TJ proteins, such as ZO-1 ([116]Fig. 1B). During
   rosette formation, components of the cytoskeleton, such as ZO-1 and
   F-actin, become apical as they reorganize expression toward the lumen
   ([117]Fig. 1B). Western blotting demonstrated that hESCs express Gal-3
   during neural induction with a maximum at day 2, just before the onset
   of rosette formation (fig. S1A). We next demonstrated Gal-3 expression
   in rosettes and showed that it is preferentially but not exclusively
   distributed toward the lumen after 6 days in NIM ([118]Fig. 1, C and D,
   and fig. S1B).

Fig. 1. Gal-3 is apically expressed in hESC during neural rosette formation.

   [119]Fig. 1.
   [120]Open in a new tab

   (A) Schematic of neural rosettes formation in neural induction medium
   (NIM). Cells progressively acquired ABP until neural rosettes
   containing a lumen are observed at day 6. (B) Immunostaining of ZO-1
   and F-actin proteins in hESCs over neural induction days until neural
   rosettes are observed at day 6. Confocal single optical section. Scale
   bar, 15 μm, for all panels. (C) Immunocytochemistry of Gal-3 in hESCs
   over neural induction days until full neural rosettes are formed at day
   6. Confocal single optical section. Scale bar, 50 μm, applies to all
   panels. (D) Higher magnification images showing Gal-3 and ZO-1 in human
   NSCs (hNSCs) at day 6 of neural induction. Schematic showing the
   details of Gal-3’s speckled distribution and its colocalization with
   ZO1. Scale bar, 5 μm, in both panels. (E) Confocal z-projections of
   double immunocytochemistry of Gal-3 and Rab11 in day 6 hNSCs and
   quantification of the Manders coefficient indicating the portion of
   Gal-3 particles colocalized with different vesicles, and the portion of
   vesicles colocalized with Gal-3. Note the extensive Rab11
   colocalization. Scale bars, 10 μm. (F) Super-resolution images of
   triple-positive EVs using an ONI NanoImager. As detected with
   immunocytochemistry, Gal-3 was colocalized with EVs (pan-EV and
   tetraspanin) at different time points. Higher magnification images show
   the details of one single triple-positive EV. Scale bars, 10 μm and 100
   nm.

   Gal-3 exhibited a speckled distribution in the cytoplasm of hESC at day
   6, suggesting vesicular localization ([121]Fig. 1D). We queried whether
   the Gal-3 punctate expression was due to its location in specific
   vesicles. We found that Gal-3 was mostly colocalized with Rab11^+
   vesicles (recycling endosomes) ([122]Fig. 1E), which matches what has
   been found in the literature ([123]21). We found slightly less Gal-3
   expressed in Rab9^+ vesicles (late endosomes). The lowest amount of
   Gal-3 was found in EEA1^+ (early endosomes) and GOLGA2^+ (Golgi
   complex) ([124]Fig. 1E and fig. S2A). We next treated cells with human
   recombinant Gal-3 (hrGal-3), expecting Gal-3 to be captured by early
   endosomes or recycling endosomes. However, there was no difference in
   the portion of Gal-3 or in the portion of the vesicles after treatment
   (fig. S2C). However, after exposure to hrGal-3, Western blots of human
   NSC (hNSC) lysates showed an increase in Gal-3 expression, suggesting
   that cells are able to capture the protein from the medium (fig. S1E).

   We next used super-resolution imaging to verify whether Gal-3 is
   secreted via extracellular vesicles (EVs) ([125]Fig. 1F). EVs were
   detected using double immunocytochemistry for pan-EV and tetraspanin.
   We observed Gal-3^+ EVs in all the conditions ([126]Fig. 1F and fig.
   S2, B and D); however, the percentage was higher in the medium of hESCs
   at days 2 to 5 ([127]Fig. 1F and fig. S2D). Although not significant,
   this result suggests that at least a portion of the high levels of
   Gal-3 observed in cell lysates of hESCs at early days of neural
   induction is also being secreted via EVs and could have a nonautonomous
   role. Last, we detected almost four times more particles per milliliter
   in pluripotent cell (day 0) conditioned medium than in neural
   differentiation medium (fig. S2D). These in vitro data indicate that
   Gal-3 expression is tightly regulated during brain development and may
   have functional roles in ABP.

Gal-3 l.o.f. disrupts formation and maintenance of ABP in hESCs

   To determine Gal-3’s role in ABP, we carried out small interfering RNA
   (siRNA) loss-of-function (l.o.f.) studies in rosette formation
   ([128]Fig. 2, A to C; and fig. S1, C to F). ABP was assessed via the
   number of neural rosettes per area. ZO-1 expression in untreated
   rosettes is visible at low magnification as small circles of
   immunofluorescence ([129]Fig. 2B and fig. S1B). The effectiveness of
   Gal-3 siRNA was validated with Western blotting and immunocytochemistry
   (fig. S1, C and D). Gal-3 KD did not affect the expression of the early
   neural marker and fate determinant Pax6 (fig. S1, E and F). However, it
   significantly disrupted neural rosette formation at day 6 after NIM
   ([130]Fig. 2, B and C). Rosette formation could be rescued by adding
   hrGal-3 to the medium for 6 days ([131]Fig. 2, B and C). These data
   indicate that Gal-3 specifically regulates ABP rather than the
   acquisition of neural fate.

Fig. 2. Disruption of neural rosettes by Gal-3 blockers is dynamic and
reversible.

   [132]Fig. 2.
   [133]Open in a new tab

   (A) Schematics of the effects of Gal-3 KD (red arrow) and rescue with
   hrGal-3 (blue arrow, right). (B) Neural rosettes density (ZO-1 circles)
   at days 3 and 6. hESCs were either not transfected at day 0 (untreated
   control) or transfected with a nontargeting siRNA (siNT) or siGal-3.
   The rescue group received hrGal-3 (1 μg/ml) daily in the medium from
   days 0 to 6. Scale bars, 50 μm. (C) Quantification of (B). Graphs are
   means with SEM; n = 9 replicates (technical and biological). Two-way
   analysis of variance (ANOVA) followed by Tukey’s multiple comparisons
   test. **P < 0.01; ***P < 0.001; n.s., nonsignificant. (D) Neural
   rosettes density upon treatment with MCP or αGal-3, compared to
   untreated controls and control antibody αIgG, respectively. Scale bar,
   50 μm. (E) MCP or αGal-3 treatment until day 3 (red arrow) led to a
   recovery in density of neural rosettes at day 6. Scale bar, 50 μm. (F)
   MCP or αGal-3 treatment from day 3 (red arrow) abolished already formed
   neural rosettes. Scale bars, 50 μm. (G) Quantification of MCP effects
   in (D) to (F). (G′) (day 3) and (G″) (day 6) are quantification of (D),
   and significance was assessed using a two-way ANOVA followed by Tukey’s
   multiple comparisons test. (G‴) (day 6) and (G⁗) (day 6) are
   quantification of (E) and (F), and significance was assessed using an
   unpaired t test. Graphs are means with SEM; n = 3 biological
   replicates; *P < 0.05. (H) Quantification of αGal-3 effects in (D) to
   (F). (H′) (day 3) and (H″) (day 6) are quantification of (D), and
   significance was assessed using a two-way ANOVA followed by Tukey’s
   multiple comparisons test. (H‴) (day 6) and (H⁗) (day 6) are
   quantification of (E) and (F), and significance was assessed using an
   unpaired t test. Graphs are means with SEM; n = 3 biological
   replicates; *P < 0.05; **P < 0.01.

   We next confirmed these results with modified citrus pectin (MCP), a
   nutraceutical-derived Gal-3 inhibitor ([134]54–[135]56). The number of
   neural rosettes per square millimeter was significantly decreased at 3
   and 6 days by MCP ([136]Fig. 2, D, G′, and G″). MCP treatment led to a
   concentration-dependent decline in neural rosette numbers, where 0.25%
   MCP partially blocked and 0.5% MCP completely blocked neural rosette
   formation at both time points ([137]Fig. 2D, G′ and G″). Blocking
   antibodies (αGal-3) were previously used to show Gal-3 is necessary for
   neuroblast migration ([138]18). Here, rosette formation was
   significantly reduced, compared to the α immunoglobulin G (αIgG)
   control after 3 days of exposure to αGal-3 ([139]Fig. 2, D, H′, and
   H″).

   The loss of rosettes observed after 3 days of MCP treatment was
   significantly reversed when MCP was removed between days 3 and 6. This
   suggests that ABP is dynamic and requires continuous Gal-3 to be
   maintained ([140]Fig. 2, E and G‴). Rosettes were also found at day 6
   when αGal-3 was only given for 3 days compared with being given
   throughout ([141]Fig. 2, E and H‴).

   We next asked whether Gal-3 blockade could disrupt already generated
   rosettes. MCP (0.5%) included in the medium between days 3 and 6
   significantly disrupted rosette formation ([142]Fig. 2, F and G⁗).
   Rosettes already formed between days 0 and 3 were also disrupted by
   αGal-3 in the medium between days 3 and 6 ([143]Fig. 2, F and H⁗).
   Neither MCP nor αGal-3 treatments changed activated cleaved caspase-3
   expression, suggesting that apoptosis was not the cause of reduced
   rosette density (fig. S1, G and H).

   Thus, l.o.f. experiments using all three Gal-3 inhibitors blocked
   rosette formation as established by loss of apical ZO-1. These data
   provide strong evidence that Gal-3 has an important role in promoting
   and maintenance of ABP formation during hESC rosette formation.

Gal-3 blockade inhibits organization of apical junctions

   Gal-3 participates in the delivery of proteins to the apical pole via
   recycling endosomes in epithelial tissue ([144]20, [145]24, [146]25,
   [147]28). Thus, we queried whether Gal-3 effects on apical polarization
   were limited to ZO-1 or were more general. Both 0.5% MCP and αGal-3
   added for 6 days to the NIM of ESCs caused loss of apical polarization
   of N-cadherin, ARL13-B, Par3, Notch, and Rab11 ([148]Fig. 3, A to E).
   During normal rosette formation, the ciliary protein ARL13-B was
   localized in discreet apical puncta, which dispersed in a seemingly
   random distribution upon Gal-3 blockade ([149]Fig. 3B). Notch
   expression was lost from its apical localization and moved into the
   nucleus after MCP exposure ([150]Fig. 3D). We also found that MCP added
   to the medium from days 4 to 6 could disrupt already formed Par3^+
   rosettes ([151]Fig. 3C). Treatment with αGal-3 from days 0 to 3 only,
   resulted in Par3^+ rosettes reappearing by day 6 ([152]Fig. 3C).

Fig. 3. Gal-3 regulates distribution of junctional proteins.

   [153]Fig. 3.
   [154]Open in a new tab

   (A) Apical distribution of N-cadherin observed via immunocytochemistry
   at day 6 is disrupted upon treatment with MCP and αGal-3 since day 0.
   Arrows indicate apical region of fully formed neural rosettes.
   Arrowheads indicate redistribution of N-cadherin. Scale bar, 50 μm. (B)
   Apical distribution of Arl13-B observed via immunocytochemistry at day
   6 is disrupted upon treatment with MCP and αGal-3 since day 0. Arrows
   indicate apical region of neural rosettes. Arrowheads indicate
   redistribution of Arl13-B. Scale bar, 50 μm. (C) Immunocytochemistry
   for Par3 reveals that its apical localization is disrupted upon
   treatment with MCP from days 0 to 6 or from days 3 to 6. In contrast,
   stopping the treatment with αGal-3 at day 3 allowed recovery of apical
   localization of Par3 at day 6. Arrows indicate apical region of fully
   formed neural rosettes. Arrowheads indicate redistribution of Par3.
   Scale bar, 50 μm. (D) Immunocytochemistry for Notch shows apical
   localization in controls and loss of this pattern after MCP and αGal-3
   (compared with αIgG control). Scale bar, 50 μm. (E) Immunocytochemistry
   for Rab11 shows apical localization in controls and loss of this
   pattern after MCP and αGal-3 (compared with αIgG control). Scale bar,
   50 μm. (F) Orthogonal views of tight and adherens junction proteins
   ZO-1, Par3, and N-cadherin in untreated and 0.5% MCP-exposed hESCs at
   day 6 of neural induction. Note that their expression is quite
   specifically localized to the apical poles of rosettes. After rosette
   disruption by the Gal-3 blocker MCP, these proteins moved toward the
   basolateral domain within the cells. Schematics on the right show the
   redistribution of TJ and AJ proteins after Gal-3 block. TJ, tight
   junctions; AJ, adherens junctions. Scale bars, 10 μm.

   Orthogonal views of higher magnification confocal images revealed the
   polarized distribution of tight and adherens junctions in neural
   rosettes at day 6. This was evidenced by the presence of bigger or
   smaller discs of junctions surrounded by 4′,6-diamidino-2-phenylindole
   (DAPI)^+ nuclei ([155]Fig. 3F). Upon MCP treatment, this neural
   epithelium lost its polarity at the same time that junctional proteins
   moved to the basolateral domain ([156]Fig. 3F). Among the proteins
   studied, N-cadherin was the most affected by MCP. These experiments
   showed that loss of Gal-3 function caused redistribution of all apical
   markers tested and was not restricted to ZO-1.

   Yap is implicated in ABP ([157]57, [158]58) and is an important
   component of the junctional complex in ECs ([159]59). Yap localization
   at TJs is also important for VZ integrity and correct delamination of
   progenitors ([160]60). Here, hESC-derived fully formed rosettes
   expressed YAP at the apical pole in TJs where it colocalized with ZO-1
   (fig. S3A). Apical polarization of YAP in rosettes was lost upon
   exposure to αGal-3 compared to αIgG controls, with most of the YAP
   found in the nucleus (fig. S3A). MCP treatment also promoted a shift of
   YAP localization from the TJ to the nucleus (fig. S3, B and C).
   Quantification of YAP and ZO-1 colocalization showed a significant loss
   upon MCP exposure, whereas YAP and DAPI colocalization increased (fig.
   S3D). Together, these data indicate that Gal-3 is important for the
   correct distribution of several apical proteins and for the recruitment
   of YAP to the TJs (fig. S3D).

RNA-seq reveals dysregulation in neurodevelopmental pathways upon MCP
treatment

   Multiple mechanisms may participate in Gal-3’s regulation of ABP.
   Therefore, we carried out bulk RNA sequencing (RNA-seq) on control and
   MCP-treated hNSCs at days 3 and 6 of neural induction. First, principal
   components analysis of gene expression values fragments per kilobase
   million (FPKM) confirmed good segregation between the groups (fig.
   S4A). We then identified genes that were up- or down-regulated in
   MCP-treated hESCs at days 3 and 6, using stringent conditions [false
   discovery rate (FDR) < 0.05; |log[2](fold change)| > 0.5] ([161]Fig. 4A
   and table S2). At day 3, there were 342 genes differentially expressed
   in MCP-treated cells, of which 126 were up- and 216 were down-regulated
   (fig. S4B). This number was higher at day 6 where 1258 genes were
   differentially expressed (655 up and 603 down) (fig. S4B). Most genes
   where uniquely enriched in day 3 or day 6 MCP groups, with an overlap
   of 99 differentially expressed genes between the two time points
   ([162]Fig. 4B). We found that several genes involved in Wnt signaling
   were overexpressed ([163]Fig. 4A). Gene Ontology (GO) analysis showed
   that at day 3, most dysregulated biological pathways (BPs) upon MCP
   treatment were related to nervous system development and epithelium
   morphogenesis ([164]Fig. 4C and table S3). At day 6, besides
   neurodevelopmental pathways, altered gene expression was also in
   pathways of cell junction assembly, cytoskeleton remodeling, and
   extracellular matrix (ECM) organization ([165]Fig. 4C), which was
   supported by the cellular components analysis (fig. S4C). For example,
   CDH2 was up-regulated and CDH10 down-regulated at day 3 (table S2).
   CDH2 and CDH9 were up-regulated, while CDH3 and CDH17 were
   down-regulated at day 6. At this time point, we also found changes in
   expression of integrin receptors, for example, up-regulation of alpha 1
   subunit of integrin receptors (ITGA1) and ITGA2 and down-regulation of
   ITGA4 (table S2). Pathways regulating vesicle-mediated transport were
   also perturbed at day 6 (table S2).

Fig. 4. RNA-seq reveals developmental genes regulated by Gal-3.

   [166]Fig. 4.
   [167]Open in a new tab

   (A) Volcano plot of genes differentially expressed (FDR < 0.05) in
   MCP-treated versus control hNSCs at days 3 and 6 of neural induction. A
   subset of genes significantly up-regulated [Log[2](fold change) > 0.5]
   is shown. (B) Venn diagram representing the number of genes
   differentially expressed in MCP versus control groups and the overlap
   between different days. (C) Gene Ontology (GO)–biological processes
   enrichment analysis of genes up-regulated in MCP versus control group
   at days 3 and 6. (D) Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathways enrichment analysis of genes up-regulated and down-regulated
   in MCP versus control group at days 3 and 6. MAPK, mitogen-activated
   protein kinase; TNF, tumor necrosis factor; PI3K, phosphatidylinositol
   3-kinase.

   We also questioned the Kyoto Encyclopedia of Genes and Genomes (KEGG)
   database for dysregulated biological functions upon MCP treatment
   ([168]Fig. 4D and table S4). With this analysis, we found dysregulation
   of Notch and Wnt signaling at day 3, as well as ECM-receptor
   interaction at day 6. We additionally found the Hippo pathway being
   dysregulated at day 6 ([169]Fig. 4D). Up-regulation of Notch was also
   observed with GO analysis, including its ligands Delta-like canonical
   Notch ligand 1 (DLL1) and Jagged1 (JAG1) at day 3 ([170]Fig. 4C and
   table S2). In contrast, at day 6, we found dysregulation of Wnt and BMP
   signaling pathways ([171]Fig. 4C).

   As expected, comparison between control at day 6 versus day 3 and MCP
   at day 6 versus day 3 showed a high overlap of differentially expressed
   genes, indicating that time of differentiation affects both groups
   similarly (fig. S4D). GO analysis showed similar enrichment in pathways
   regulating neural differentiation in both groups (fig. S4E).

Human and murine embryonic brain stem cell niches express Gal-3

   Because hESC expressed Gal-3 in vitro, we hypothesized that it would
   also be expressed in embryonic NSCs in vivo ([172]18, [173]43). Gal-3
   was expressed in the VZ, the inner SVZ (iSVZ) and oSVZ and in the
   intermediate zone (IZ) of postconception week (PCW) 17 human brain
   ([174]Fig. 5, A to C). Higher magnification revealed Gal-3^+ cells
   lining the LV as well as Gal-3^+ cells in the iSVZ and oSVZ ([175]Fig.
   5, D to H). The morphology and position of Gal-3^+ cells suggested that
   they were ventricular RG (vRG) ([176]Fig. 5, F to H), oRG, or
   neuroblasts. Some Gal-3^+ cells in the developing human forebrain
   expressed the RG marker p-Vimentin ([177]Fig. 5, I and J). We did not
   detect Gal-3 immunoreactivity in the human subpallium. Gal-3
   immunofluorescence was also found in human brain in the cortical sling
   (fig. S5D). These results indicate the presence of Gal-3 in NSC in
   human embryonic brains. They are commensurate with human single-cell
   RNA sequencing (scRNA-seq) studies showing LGALS3 transcripts (Gal-3
   gene) in multiple subtypes of embryonic RG ([178]Fig. 5, K and L)
   ([179]61).

Fig. 5. Gal-3 is expressed in human and murine embryonic NSC.

   [180]Fig. 5.
   [181]Open in a new tab

   (A to J) Gal-3 expression was analyzed via immunohistochemistry in
   coronal sections of a postconception week (PCW) 17 human embryonic
   brain. MZ, marginal zone; CP, cortical plate; SP, subplate;
   IZ, intermediate zone; oSVZ, outer subventricular zone; iSVZ, inner
   subventricular zone; VZ, ventricular zone; LV, lateral ventricle. (A)
   Robust Gal-3 expression was observed in the VZ, iSVZ, oSVZ, and IZ
   layers. Scale bar, 500 μm. (B and C) Details of Gal-3
   immunohistochemistry in the VZ, iSVZ, and oSVZ. Scale bar, 200 μm. (D)
   Gal-3 expression in the VZ and iSVZ, confocal z-stack projection. Scale
   bar, 100 μm. (E) Gal-3 expression in the oSVZ, single optical section.
   Scale bar, 20 μm. (F and G) Higher magnification showing Gal-3–positive
   VZ cells with ABP, with apical process evident, and RG-like morphology.
   Scale bar, 10 μm. (H to J) The oSVZ showing Gal-3–positive and
   p-Vimentin (p-Vim)–positive RG fibers (arrow) and migrating
   neuroblasts. Scale bar, 50 μm. (K) tSNE plot showing LGALS3 gene
   expression from scRNA-seq in radial progenitor populations (oRG, vRG,
   and RG-div2) across stages of human embryonic cortical neurogenesis.
   Generated using UCSC Cell Browser ([182]61). Cell type clusters are
   annotated. Clusters RG-div1, RG-div2, oRG, vRG and tRG are boxed.
   PC-MGE, intermediate progenitors of medial ganglionic eminence;
   IPC-nEN, intermediate progenitors of newborn excitatory neurons; nIN,
   newborn interneurons; IN, interneurons; nEN, newborn excitatory
   neurons; EN, excitatory neurons; M, microglia; Ast, astrocytes; OPC,
   oligodendrocyte progenitors; Endo, endothelial cells; RG-early, Early
   radial glial cells; RG-div1 and 2, dividing radial glial cells cluster
   1 and 2; oRG, outer radial glial cells; tRG, truncated radial glial
   cells. (L) Human LGALS3 enrichment measure in transcript per million
   (TPM) + 1. Graph bar is minimum to maximum, and line represents mean.
   Wilcoxon’s rank sum test. *P < 0.05; **P < 0.01; ****P < 0.0001. (M and
   N) PCR showing murine Lgals3 expression in different ages (M) and
   across regions (N). Fold change values were calculated relative to
   left-most columns. Graphs show mean with SEM. N = 3 for each region
   except for cerebellum P10 (N = 2) and midbrain E18 (N = 2) and midbrain
   P2. Ordinary one-way ANOVA followed by Dunnett’s multiple comparison
   test was used for statistical analysis. *P ≤ 0.05, ***P ≤ 0.001, ****P
   < 0.0001. (O and P) Gal-3 expression in E17.5 mouse lateral ganglionic
   eminence (LGE) (O) and septum (Spt) (P). LV, lateral ventricle. Scale
   bar, 50 μm. (Q and R) Gal-3 is expressed in Sox2^+ cells lining the
   lateral ventricles. Scale bar, 50 μm.

   Others have shown low, but detectable, levels of Lgals3 in developing
   mouse RG with scRNA-seq ([183]62, [184]63). We detected Lgals3 with
   reverse transcription quantitative polymerase chain reaction (RT-qPCR)
   in the mouse brain at embryonic day 15.5 (E15.5), E18.5, postnatal day
   2 (P2), P10, and adulthood ([185]Fig. 5M). All areas examined (fig.
   S5A) expressed detectable Lgals3 transcripts with the cerebral cortex
   expressing the lowest amount ([186]Fig. 5N). Using
   immunohistochemistry, we found robust Gal-3 expression in the VZ of the
   lateral ganglionic eminence (LGE) ([187]Fig. 5O) and septum ([188]Fig.
   5P) of E17.5 mice. Gal-3 immunofluorescence in cell bodies lining the
   ventricles suggested that they were RG. The highest Gal-3 expression
   was apical, in cell bodies, whereas basal processes contained little to
   no detectable Gal-3 ([189]Fig. 5, Q and R). Gal-3 expression also
   colocalized with Sox-2 (which is expressed by RG stem cells) in LGE and
   septal cells lining the lateral ventricle ([190]Fig. 5, Q and R). Gal-3
   also colocalized with ZO-1 (fig. S5B). Gal-3 immunofluorescence was not
   detected in the mouse E17.5 pallium although we detected it there with
   RT-PCR. As anticipated, Gal-3 expression was detected in some Iba1^+
   microglial cells in the embryonic mouse brain especially in the
   cortical sling, which includes the developing corpus callosum and
   septum (fig. S5, C and E).

Gal-3 inhibition leads to sulcus formation

   Our findings suggested that Gal-3 l.o.f. may also disturb
   neurodevelopment in vivo during embryogenesis. MCP is known to affect
   the central nervous system upon systemic injection ([191]55, [192]64),
   and we administered it to pregnant mice in their drinking water and
   analyzed brains of embryos at E17.5. There was no difference in water
   consumption or weight gain between control and MCP-treated females
   (fig. S6A). Control embryos had smooth cerebral cortices as expected
   since mice are lissencephalic ([193]Fig. 6, A and B; and fig. S6, B and
   C). MCP-treated brains, in contrast, had sulci in the cerebral cortex
   of 57.1% of the embryos ([194]Fig. 6, C to F; and fig. S6, B and C).
   This was seen across all anterior-posterior and mediolateral regions of
   the cerebral cortex ([195]Fig. 6, D to F, and fig. S6C). Most embryos
   administered MCP had one sulcus, but two embryos had more than one
   ([196]Fig. 6E). The sulci were measured using sulcus depth (SD) and
   local gyrification index (LGI), adapted from ([197]65) (fig. S6, D and
   E). A higher SD indicates deeper folding, while a higher LGI signifies
   narrower and deeper folding. Foldings were classified as sulci if the
   SD value exceeded 100 μm and the LGI value exceeded 0.5 (fig. S6, D and
   E). The cortical layer markers Satb2 and Ctip2 exhibited proper
   orientation, indicating that these folds reflected cortical folds and
   not tissue damage ([198]Fig. 6, B and F, and fig. S6C). We then
   measured the gyrification index (GI), which is the ratio between the
   detailed outline of the pial surface and outer contour of the cortex
   (fig. S6F). Higher GI values indicate more brain folding. There was a
   9.2% increase in the GI value in the MCP group when compared to the
   control (fig. S6F).

Fig. 6. Gal-3 inhibition leads to sulci formation in E17.5 mice embryos.

   [199]Fig. 6.
   [200]Open in a new tab

   (A) DAPI staining of control coronal hemi-sections at an anterior
   level. Note the smoothness of the cortical surface. lv, lateral
   ventricle. Scale bar, 500 μm. (B) Control cerebral cortex at an
   anterior level with immunohistochemistry for Satb2 and Ctip2. Scale
   bar, 100 μm. (C) Quantification of the percentage of embryos displaying
   sulci in all litters studied (two litters per group). Quantification
   showed that controls had no sulci. (D) Anterior level coronal
   hemi-section from an embryo exposed to MCP. DAPI staining of
   MCP-treated embryos at similar anterior posterior level as (A) showing
   representative sulcus (red box). Box shown at higher magnification in
   (D′). Scale bar, 500 μm. (D′) 100 μm. (E) Posterior level coronal
   hemisection showing two adjacent sulci in an MCP-treated embryo. Green
   box shown at higher magnification in (E′). Scale bar, 500 μm. (E′) 100
   μm. (F) Satb2 and Ctip2 immunohistochemistry with a typical sulcus (red
   arrow). Scale bar, 100 μm. (G) A Gal-3^−/− embryo with a typical
   sulcus. Scale bar, 100 μm. (H) Gal-3^−/− embryo with a deep sulcus (red
   arrows). Scale bar, 100 μm. (I) A Gal-3^−/− embryo with an intermediate
   depth sulcus (red arrow). Scale bar, 100 μm. (J) Satb2 and Ctip2
   immunohistochemistry of the sulcus shown in (J). Scale bar, 400 μm. (K)
   Gal-3^−/− embryo sulcus (red arrow) stained for Satb2 and Ctip2. Note
   that it almost reached the VZ. Scale bar, 400 μm.

   To confirm these results, we examined two litters of Gal-3^−/− mice
   (Gal-3 KO) at E17.5 of gestation. Gross anatomical examination of brain
   sections revealed the unexpected result that 37.5% mice had sulci in
   the cerebral cortex similar to the ones that we observed after MCP
   ([201]Fig. 6, C and G to K). The depth of the sulci varied from
   somewhat shallow ([202]Fig. 6G) to quite deep ([203]Fig. 6, H and I),
   which can be verified with the SD and LGI values (fig. S6, D and E).
   The cortical layer markers Satb2 and Ctip2 immunofluorescence appeared
   to be normal and were expressed in and around the sulci ([204]Fig. 6, J
   and K). There was a 19.5% increase in the GI value in Gal-3 KO compared
   to control brains (fig. S6F). In addition to the cortical sulci, we
   observed several examples of indentations along the walls of the
   lateral ventricles (fig. S6G). Quantification of these VZ notches shows
   that there were significantly more in the KO compared to control mice
   (fig. S6H).

Gal-3 blockade disrupts ABP and cell division in forebrain embryogenesis in
vivo

   Abnormal VZ, SVZ, and IZ development can induce or limit gyrification
   ([205]66). We therefore examined these zones in the MCP-treated
   animals. Our rosettes showed MCP and other Gal-3 blockers disrupting
   the ABP of ZO-1. We assessed this in vivo with immunohistochemistry for
   ZO-1 in the pallial VZ. Control embryos had a very distinct expression
   of apical VZ ZO-1 expression ([206]Fig. 7A and fig. S7B). In contrast,
   MCP-exposed embryos had broad ZO-1 expression, suggesting TJ disruption
   and expansion of the apical VZ ([207]Fig. 7, A and B). This effect was
   most pronounced anteriorly and gradually diminished toward the middle
   and posterior portions of the pallium ([208]Fig. 7, A and B; and fig.
   S7, A to E).

Fig. 7. Gal-3 inhibition in the developing cortex: disrupted ABP, increased
vertical cell division, and delamination from apical surface.

   [209]Fig. 7.
   [210]Open in a new tab

   (A) Representative image showing ZO1 immunohistochemistry in the
   pallial VZ lining the lateral ventricle (LV). Anterior cerebral cortex,
   coronally sectioned. Arrows indicate the thickness of the apical VZ
   region, which is much expanded after MCP treatment. Note also that
   blood vessels (BV) are ZO1^+. Scale bars, 50 μm. (B) Quantification of
   thickness of apical region of control and MCP-treated E17.5 mice
   embryos in anterior sections. Graphs are represented as means with SEM;
   n = 6. Mann-Whitney test; **P < 0.001. (C) Representative image of DAPI
   staining of mouse E17.5 brain sections. Square on the left image shows
   enlarged area on the right images. The red lines show the angle between
   the cortex-lateral ventricle border and the cleavage plane of neural
   progenitors in anaphase. Scale bar, 350 μm. (D) Distributions of
   cleavage plane angles in separate control and MCP-treated E17.5
   embryos. Graphs are represented as means with SEM; n = 3. Mann-Whitney
   test; ***P < 0.001. (E) Quantification of percentage of vertical cell
   divisions in control and MCP-treated embryos. Vertical cell divisions
   were classified as anaphases with a cleavage plane angle of less than
   60°. Graphs are represented as means with SEM; n = 3. Mann-Whitney
   test; ***P < 0.001. (F) Representative images of Pax6^+Ki67^+
   immunohistochemistry and Tbr2^+Ki67^+ immunohistochemistry. Images are
   from anterior cortex in control and MCP-treated E17.5 embryos. Arrows
   point toward example of PAX6^+Ki67^+ and Tbr2^+Ki67^+ double-positive
   cells. Overlapping expression shown in high magnification panels on the
   right. Scale bars, 50 μm. (G) Quantification of number of PAX6^+,
   PAX6^+Ki67^+, Tbr2^+, and Tbr2^+Ki67^+ cells per square millimeter in
   anterior coronal sections. Graphs are represented as means with SEM;
   n = 4. Unpaired t test; *P < 0.05.

   ZO-1 regulates the cytoskeleton and is critical for the organization of
   the spindle pole during cell divisions ([211]67). Horizontal cell
   divisions have cleavage planes with ~90^o orientation with respect to
   the lateral ventricle, and this angle becomes more parallel to the LV
   in vertical divisions ([212]Fig. 7C). To determine whether the
   preponderance of horizontal versus vertical divisions changes in
   MCP-exposed embryos, we measured the cleavage plane angle of dividing
   cells in the VZ of the neocortex. Angles below 60% were categorized as
   vertical divisions. Notably, the MCP-exposed embryos had greater than
   four times more vertical divisions compared to controls ([213]Fig. 7, D
   and E).

   Previous studies have shown that oblique cell divisions in the VZ lead
   to generation of basal progenitors (oRGs), a population of progenitors
   enriched in gyrencephalic species. We predicted that altered cleavage
   planes would cause abnormal numbers of basal progenitors (or oRGs) in
   mouse. These cells can be detected as Pax6^+ proliferative progenitors
   in the IZ ([214]9, [215]48). We found significantly more Pax6^+
   progenitor cells in the IZ of embryos exposed to MCP, at all anterior
   posterior levels ([216]Fig. 7, F and G; and fig. S7, F to I). We also
   found more Pax6^+Ki67^+ cells in the IZ at all anterior posterior
   levels in MCP-exposed embryos ([217]Fig. 7, F and G; and fig. S7, F to
   I). We also show that the number of Tbr2^+ and Tbr2^+Ki67^+ cells was
   not significantly different between controls and MCP exposed embryos
   ([218]Fig. 7, F and G), indicating that the pool of intermediate
   progenitors was not affected. Together, these data suggest that Gal-3
   blockade causes an increase in the pool of basal progenitors, as
   indicated by proliferative Pax6^+ cells in the IZ, a phenotype also
   seen in Gal-3 binding protein (GAL3BP) mutants ([219]66).

DISCUSSION

   We have shown that Gal-3 l.o.f. caused ABP disruption in vitro and in
   vivo, and a summary of our data is shown in [220]Fig. 8. Additionally,
   Gal-3 KO and the Gal-3 blocker MCP during embryonic development caused
   sulci to appear in the murine cerebral cortex. These findings suggest
   that Gal-3 has fundamental roles in early mammalian brain development.
   MCP is readily available over the counter as a nutraceutical without
   health warnings. However, blocking Gal-3 function may inhibit ABP and
   thus be harmful for the developing fetus in pregnant women. Thus, our
   work cautions against the use of MCP as medication during pregnancy.

Fig. 8. The effects of Gal-3 inhibition on ABP in neural embryonic
development.

   [221]Fig. 8.
   [222]Open in a new tab

   Blocking Gal-3 in vitro resulted in disruption of the polarization of
   the neural epithelium, as evidenced by a decrease in the number of
   neural rosettes. In vivo, disruption of ZO1 could be observed around
   the VZ of E17.5 mouse embryos upon Gal-3 inhibition. The number of
   vertical cell divisions increased, and more Pax6^+Ki67^+ cells were
   found in the intermediate zone, indicating delamination of oRG. Last,
   sulci were observed in MCP and Gal-3 KO mice. TJ, tight junctions; CP,
   cortical plate; SP, subplate; IZ, intermediate zone; V-SVZ,
   ventricular-subventricular zone; aRGs, apical radial glia; oRG, outer
   radial glial.

Gal-3 regulates ABP in hESC rosette formation

   Using neural induction of hESCs in vitro, we observed a peak of Gal-3
   expression at day 2, preceding important early neural commitment
   events, such as ABP and Pax6 expression. Gal-3 KD did not alter
   expression of Pax6, suggesting that Gal-3 works in a Pax6-independent
   fashion. Gal-3 l.o.f. experiments did show, however, a decrease in the
   number of neural rosettes in culture, suggesting a role in cellular
   events leading to ABP.

   Gal-3 was previously shown in other models to affect epithelialization
   and apical-basal compartmentalization, targeting protein cargos to the
   apical side ([223]25, [224]26). We observed Gal-3^+ intracellular
   puncta in hESC-derived NSC. We confirmed that the Gal-3 puncta were
   colocalized with vesicles and enriched in the Rab11^+ compartment,
   which labels recycling endosomes. Secreted Gal-3 can form lattices in
   the cell membrane upon its interaction with transmembrane receptors,
   which can result in incorporation into carrier vesicles
   ([225]68–[226]70). We found that Gal-3 secreted via exosomes, with a
   higher percentage in the early days of neural differentiation. It is
   also possible that Gal-3 participates in cell polarization via
   interaction with the extracellular domain of junctional proteins, as
   described in previous work ([227]71–[228]73). Gal-3 has also been
   implicated in ciliogenesis and is associated with centrosomal
   abnormalities ([229]28). Here, we observed that junctional,
   cytoskeletal, and ciliary proteins, such as ZO-1, Par3, N-cadherin,
   F-actin, and ARL13-B, lack typical ABP organization upon Gal-3
   blockade, although no direct interaction between those components was
   analyzed.

   These findings suggested that Gal-3 could affect ABP in a number of
   divergent cellular and extracellular domains. This was confirmed by the
   variety of tools we used and the location of Gal-3 expression that they
   target. Gal-3 siRNA would initially have reduced cytoplasmic and
   nuclear Gal-3. The function blocking Gal-3 antibody and MCP would have
   limited Gal-3 incorporation, binding Gal-3 extracellularly and blocking
   its function. Recombinant Gal-3, on the other hand, would have, at
   first, increased the extracellular concentration of Gal-3 and then
   affected intracellular compartments. Thus, over time, it is likely that
   location of Gal-3 gain of function or l.o.f. would have spread into
   other domains.

RNA-seq reveals potential mechanisms influencing Gal-3 function in ABP

   The function of Gal-3 in healthy mice and in disease models in the
   postnatal and adult SVZ has been extensively studied ([230]18,
   [231]39–[232]43, [233]45, [234]74). Early studies revealed that Gal-3
   helps maintain V-SVZ neuroblast migration in adult mice ([235]18,
   [236]40), and, later, we showed that it regulates cell fate choices in
   postnatal mice via Wnt and BMP signaling ([237]43, [238]45). ECs have
   apically localized motile cilia, and these were stunted and fewer in
   number in Gal-3^−/− mice, suggesting abnormal apical pole maintenance
   ([239]18). However, not much is known about pathways regulated by Gal-3
   in the embryonic brain.

   Our transcriptomics GO analysis data revealed that first Notch at day 3
   and then Wnt and BMP pathways at day 6 are up-regulated upon treatment
   with the Gal-3 inhibitor MCP. The regulation of BMP is supported by the
   previous literature showing involvement of Gal-3 in these pathways
   ([240]43). Our data also support previous literature showing that Gal-3
   regulates Notch signaling, via down-regulation of JAG1 and DLL4,
   including via binding to JAG1 ([241]75, [242]76). KEGG analysis
   additionally showed Hippo pathway up-regulated upon MCP treatment. The
   Hippo pathway is not only important for development and neural
   differentiation but also essential for cell-cell signaling ([243]77,
   [244]78). Its main effector YAP is not only present at the junctions,
   but loss of junctional YAP led to loss of N-cadherin in ECs and of ZO-1
   and β-catenin from cellular junctions in RGs in the developing mouse
   brain ([245]59, [246]79). We also found that Yap expression becomes
   disorganized in response to Gal-3 l.o.f. and shifts from TJs to the
   nucleus. YAP nuclear localization led to the loss of apical surface
   integrity, hydrocephalus, and gyrification in mice ([247]60). Whereas
   Gal-3 expression clearly affects Yap localization, we are uncertain
   whether YAP is necessary for Gal-3 function. Last, we also saw
   dysregulation of several cadherin genes upon MCP treatment. Gal-3 has
   been shown to interact with cadherins ([248]71).

Gal-3 is expressed in embryonic NSC niches

   As predicted by the hESC findings, we found robust Gal-3 expression in
   the developing human brain. Gal-3 was located in the proliferative
   layers, VZ, SVZ, and also in the IZ. These data suggested that it was
   positioned to have effects on the human stem cell niche.

   In the mouse, RT-qPCR revealed Lgals3 across different developmental
   stages and at different regions of the brain, with the lowest
   expression in the cerebral cortex. Gal-3 expression was also examined
   via immunostaining in E17.5 mouse embryos. Unexpectedly, Gal-3
   immunoreactivity was in the subpallium and Gal-3 could not be detected
   in the pallium. We predict that more sensitive immunohistochemical
   techniques such as enzymatic amplification would detect Gal-3 in the
   pallium. We found strong Gal-3^+ expression apically in VZ cells and
   coexpression with the stem cell marker Sox2.

Loss of Gal-3 function generated sulci in the mouse cortex and disrupted
embryonic ABP

   In examining Gal-3 KO and MCP-treated brains, we noticed that
   approximately one-third to half of the embryos (respectively) had sulci
   in the cerebral cortex. A similar phenotype has been observed after
   overexpression of mutant versions of the human GAL3BP (LGALS3BP) by the
   Cappello group ([249]66). They suggested that the phenotype resulted
   from a disruption of apical junctions and increased delamination of
   NSCs from the VZ ([250]66). How Gal-3 and Gal3BP functionally interact
   in the developing human nervous system is poorly understood. Notably,
   mice do not express LGALS3BP; thus, our in vivo findings cannot be
   explained by Gal-3 interactions with the GAL3BP.

   We suspected that the de novo sulci were correlated with disrupted ABP
   in the germinal zones and showed that Gal-3 inhibition by MCP disrupted
   ZO-1^+ apical junctions. ZO-1 is expressed apically in a linear and
   tight pattern in VZ cells in controls. However, MCP caused its
   expression to spread in a honeycomb pattern, similar to the in vitro
   spread of ZO-1 from the apical lumen of neural rosettes in response to
   Gal-3 blockade. Because the large majority of cells lining the VZ
   during embryogenesis are RG stem cells, we believe that the ZO-1
   dispersion occurred in these cells. These data suggest that Gal-3 is
   essential for ABP in the developing cerebral cortex.

   Migration may directly influence cortical folding, and we showed that
   Gal-3 is necessary for postnatal SVZ neuroblast migration ([251]18).
   Gal-3 KOs exhibited slower and complex “exploratory” migration and
   resulted in overall decreases in olfactory bulb neurogenesis ([252]18).
   Similarly, it may be that, in the developing cortex, Gal-3 KO or MCP
   blockade disrupts rates of migration form the mitotic zones into
   developing cortical layers. Future live imaging of slices may shed
   light on this interesting possibility.

   We also suspected that the ABP disruption in the VZ would affect cell
   division and layering of progeny. We found that Gal-3 inhibition
   increased the percentage of VZ vertical cell divisions by fourfold.
   Gal-3 binds to NuMa, affecting mitotic spindle cohesion and
   localization in HeLa cells ([253]31). Disrupted cellular junctions may
   also contribute to the phenotype, as the LGN/NuMA complex localizes at
   cellular junctions and is crucial for division angle determination
   ([254]80). The switch to vertical divisions suggests that the VZ may
   generate more oRG progenitor cells. The number of Pax6^+ and
   Pax6^+Ki67^+ cells in the IZ was increased upon MCP exposure, which was
   not observed for intermediate progenitors detected with Tbr2. This
   could have contributed to the observed cortical folding, because oRG
   cells have been linked with gyrus formation and the oRG population is
   increased in gyrencephalic species ([255]49). Alternatively, cortical
   folding may also be due to early delamination of progenitors from the
   VZ, and TJ disruption has been seen before in other studies ([256]60,
   [257]65).

Clinical concern

   Proteins associated with apical polarization are essential for
   development and ABP disruption can lead to abnormal brain development.
   Junctional instability was shown to lead to disruption of radial glial
   cell morphology, disruption of the ventricular surface, defects in
   neuronal migration, and development of hydrocephalus ([258]10,
   [259]81). Blocking Gal-3 function to treat systemic inflammation may
   inhibit ABP and thus be harmful for the developing fetus in pregnant
   women ([260]82–[261]84). MCP is readily available over the counter as
   an anti-inflammatory nutraceutical yet there are no warnings that
   taking it during pregnancy is contra-indicated. Further studies on
   human mutations of the Gal-3 gene in human cohorts could help to
   establish causal relations.

Limitations of the study

   Our in vitro studies were conducted in hESC line H9 only, and other
   cell lines may vary in their response to Gal-3 l.o.f. Similarly, we
   used commercial medium for neural induction in which the components are
   not disclosed. Thus, it is unclear to what extent our in vitro results
   are dependent on the exact culture medium composition.

   With immunohistochemistry, we did not detect Gal-3 in the pallial VZ in
   E17.5 embryos, possibly caused by lack of sensitivity in detection of
   expression and/or peak expression earlier during development.

   MCP effectively blocks Gal-3 function by binding to the noncanonical
   site of its carbohydrate recognition domain ([262]82, [263]83,
   [264]85). Several studies have administered MCP to animals in drinking
   water ([265]54, [266]56, [267]86–[268]89). However, it is not clear how
   specific MCP is to Gal-3 ([269]85).

   We did not investigate whether Gal-3 regulates other aspects of ABP in
   the brain including in blood vessels. A recent study showed that Gal-3
   participates in BBB damage in a stroke model ([270]64). We showed that
   Gal-3 is necessary for vascular endothelial growth factor–dependent
   angiogenesis poststroke in mice ([271]41).

   Loss-of-function experiments were conducted in vivo in murine models
   and human cell lines. The murine model should be investigated at later
   time points to explore whether the alterations in cortical cell numbers
   persist to later stages. We did not explore whether these alterations
   in cortical folding and cell numbers have effects on behavior.

   Last, a notable paper shows that RG can assume distinct but parallel
   transcriptomics trajectories, all leading to a common newborn neuron in
   human and ferret, but not in mouse ([272]90). We do not know whether
   the trajectories described are regulated by Gal-3 but one of them had a
   high level of expression of family member, LGals1 (Gal-1).

MATERIALS AND METHODS

Neural induction of hESC line H9

   The H9 hESC line (WA09) is an XX genotype cell line originally obtained
   from WiCell. Cells were kept in an incubator supplied with 5% CO[2] and
   cultured at 37°C. Cells were maintained under feeder-free conditions on
   Geltrex (Thermo Fisher Scientific) or Matrigel hESC-Qualified Matrix
   (Corning) in mTeSR Plus medium (STEMCELL Technology).

   Pluripotent hESCs were induced to commit to the neural lineage using
   STEMdiff SMADi NIM (STEMCELL Technology). In brief, 80% confluent cells
   were washed with Dulbecco's Phosphate Buffered Saline (D-PBS)
   (Sigma-Aldrich) and incubated with 0.5 mM EDTA (Thermo Fisher
   Scientific) for 10 min at 37°C. Cells were pelleted down at 200g for 5
   min and resuspended to single-cell solution in NIM supplemented with 10
   μM Rho-associated kinase inhibitor (ROCKi) Y-27632 (Stratech
   Scientific). Cells were seeded in Matrigel hESC-Qualified Matrix
   (Corning) coated six-well plates containing glass cover slips (VWR) or
   μ-Slide 18 ibiTreat (Thistle Scientific) at densities between 2000 and
   4000 cell/mm^2. Medium was changed daily, and cells were collected at
   several time points.

hESC transfection and treatments

   To promote transient KD of Gal-3 in H9 cells, a library of siRNAs
   targeting Gal-3 was acquired from Dharmacon (siGal-3). For [273]Fig. 2
   (B and C) and fig. S1C, a nontargeting pool was acquired from Dharmacon
   and used as control [nontargeting siRNA (siNT)]. For fig. S1D, a
   siLuciferase acquired from Eurofins was used as control (siNT). Reverse
   lipid transfection was performed. Briefly, siRNAs and Lipofectamine
   (Thermo Fisher Scientific) were separately diluted in Opti-MEM (Gibco).
   Solutions were then mixed together and incubated for 15 min before
   being distributed in μ-Slide 18 ibiTreat or 24 well Matrigel-coated
   plates. Pluripotent H9 aggregates were detached into single-cell
   suspension in NIM and 10 μM ROCKi ([274]Fig. 2, B and C, and fig. S1C)
   or mTeSR and 10 μM ROCKi (fig. S1D). In both cases, the medium was
   changed 48 hours later to NIM, which was renewed daily. Cells were
   collected at different time points.

   Carrier-free hrGal-3 (R&D Systems) was diluted in D-PBS. Cells were
   treated with NIM containing hrGal-3 (1 μg/ml), which was changed every
   other day. For Gal-3 activity inhibition experiments, cells were
   treated every day with rat Gal-3 antibody clone M3/38 (1 μg/ml; Santa
   Cruz Biotechnology) or the same quantity of species-specific IgG (Santa
   Cruz Biotechnology) diluted in NIM. For cell treatment with MCP
   (ecoNugenics), MCP was diluted in NIM to a concentration of 0.25 or
   0.5%, and the medium was filtered using 0.22-μm Millex-GP Filter Units
   (Sigma-Aldrich). Medium was changed daily.

Bulk RNA-seq of MCP-treated hESCs

   Cells were treated with MCP for 3 or 6 days as described. RNA was
   extracted using the RNeasy Mini Kit (QIAGEN) according to the
   manufacturer’s guidelines. Messenger RNA was purified from total RNA
   using poly-T oligo-attached magnetic beads. After fragmentation, the
   first-strand cDNA was synthesized using random hexamer primers,
   followed by the second-strand cDNA synthesis using either dTTP for
   non–strand-specific library or 2'-deoxyuridine, 5'-triphosphate (dUTP)
   for strand-specific library. For the non–strand-specific library, it
   was ready after end repair, A-tailing, adapter ligation, size
   selection, amplification, and purification. For the strand-specific
   library, it was ready after end repair, A-tailing, adapter ligation,
   size selection, amplification, and purification. The library was
   checked with Qubit and real-time PCR for quantification and bioanalyzer
   for size distribution detection. After library quality control,
   different libraries were pooled on the basis of the effective
   concentration and targeted data amount and then subjected to Illumina
   sequencing. Raw data (raw reads) of fastq format were first processed
   through fastp software. In this step, clean data (clean reads) were
   obtained by removing reads containing adapter, reads containing ploy-N,
   and low quality reads from raw data. At the same time, Q20, Q30, and
   guanine-cytosine content (GC) of the clean data were calculated.

   All the downstream analyses were based on the clean data with high
   quality. Reference genome and gene model annotation files were
   downloaded from genome website. We used HISAT2 (2.2.1) to build the
   index of the reference genome and use HISAT2 to align paired-end clean
   reads to the reference genome. The mapped reads of each sample were
   assembled by StringTie (v2.2.3) in a reference-based approach.
   featureCounts (2.0.6) was used to count the reads numbers mapped to
   each gene. Then, FPKM of each gene was calculated on the basis of the
   length of the gene and reads count mapped to this gene. Differential
   expression analysis for two conditions/groups was performed using the
   DESeq2 R package (1.42.0). The resulting P value is adjusted using the
   Benjamini and Hochberg’s methods to control the error discovery rate.
   The threshold of significant differential expression: FDR ≤ 0.05 and
   |log[2](fold change)| ≥ 0.5. GO enrichment analysis of differentially
   expressed genes was implemented. GO terms with FDR less than 0.05 were
   considered significantly enriched by differential expressed genes. KEGG
   analysis was performed using the website
   [275]https://maayanlab.cloud/Enrichr/.

Immunocytochemistry

   Cells were fixed with 4% paraformaldehyde (PFA) in phosphate-buffered
   saline (PBS; Thermo Fisher Scientific) for 15 min and permeabilized
   with 0.5% Triton X-100 (Thermo Fisher Scientific) for 10 min. The cells
   were blocked for 1.5 hours in blocking solution [5% bovine serum
   albumin (BSA) (Thermo Fisher Scientific) and 0.2% Triton X-100 in 0.1 M
   PBS]. After that, the cells were incubated with the primary antibodies
   that were diluted in antibody solution (1% BSA and 0.2% Triton X-100 in
   PBS) overnight at 4°C. Next day, a three times 0.1 M PBS wash of 10 min
   each was performed before and after incubation with secondary antibody.
   For this, cells were incubated at room temperature for 1.5 hours with
   secondary antibodies (1:500) and DAPI (Thermo Fisher Scientific;
   1:1000) diluted in antibody solution. Secondary antibodies were species
   specific and conjugate with Alexa Fluor 488, 568, or 647 fluorophores
   (Invitrogen). For a complete list of antibodies used, see table S1.

EV isolation and characterization

   Conditioned medium (72 hours) was obtained from cells at pluripotency
   (day 0) and during neural induction (days 2 to 5 and days 5 to 7). EVs
   were purified using qEV Gen 2 size exclusion columns (IZON). The
   particle concentration of the isolated EVs was then measured using a
   flow NanoAnalyzer (NanoFCM). Super-resolution imaging was performed
   with an ONI NanoImager using the EV Profiler v2 kit (ONI), following
   the manufacturer’s instructions. Triple antibody staining was carried
   out using the tetraspanin trio detection reagent (568 nm) and the
   pan-EV detection reagent (488 nm) provided in the kit, along with an
   Alexa Fluor 647–conjugated Gal-3 antibody (BD Biosciences). The EV
   concentration for staining was adjusted to 1 × 10^10 particles/ml for
   each imaging channel.

   Imaging data acquired were analyzed using CODI (Collaborative Discovery
   platform from ONI,
   [276]https://oni.bio/nanoimager/software/codi-software/). The “EV
   Cluster and Counting” analysis module was used, and all parameter
   settings were adjusted following the CODI user manual and were applied
   consistently across imaging datasets.

Western blotting

   Cells were detached, and the pellet was washed with 0.1 M PBS before
   lysis. Samples were then sonicated for ~10 s at an amplitude of 10%
   maximum in a XL-2000 Ultrasonic Liquid Processor (Misonix). Lysates
   were centrifuged at maximum speed (17,000g) for 15 min, and the
   supernatant containing proteins was collected. Protein content was
   measured using the Pierce Bradford Plus Protein Assay Reagent (Thermo
   Fisher Scientific) in a microplate reader (BMG Labtech) using standard
   curves. Proteins were treated with NuPAGE LDS Sample Buffer (Thermo
   Fisher Scientific), with 100 mM 1,4 dithiothreitol (DTT; Roche), and
   denatured by boiling at 97°C for 5 min. Samples were loaded in NuPAGE
   with a bis-tris gradient of 4 to 12%, on a 1.5-mm Mini Protein Gel
   (Thermo Fisher Scientific) and separated via electrophoresis. Proteins
   were then transferred to an Immobilon-FL polyvinylidene difluoride
   membrane of pore size of 0.45 μm (Merck Millipore). Membranes were
   blocked in 5% milk in PBS-Tween (PBS-T: 0.1 M PBS and 0.1% of Tween 20;
   Sigma-Aldrich) for 1 hour at room temperature before overnight
   incubation with primary antibodies in 5% milk at 4°C. Horseradish
   peroxidase (HRP)–labeled secondary antibodies were diluted to 1:2500 in
   5% milk, and membranes were incubated for 1 hour at room temperature.
   Protein bands were obtained in x-ray film (Fujifilm) and developed in
   ECOMAX X-ray Processor (Raytech Diagnostics) or in Chemidoc XRS Gel
   Imaging System using Image Lab Software (Bio-Rad) after 1 min
   incubation with Pierce ECL Western Blotting Substrate (Thermo Fisher
   Scientific).

   Western blotting densitometry quantification of bands was obtained in
   ImageJ software and is given in arbitrary units. Gal-3 levels in
   [277]Fig. 1C and fig. S1C were normalized to the loading control
   glyceraldehyde-3-phosphate dehydrogenase. For [278]Fig. 1C, values are
   relative to basal levels during pluripotency, which were the average of
   hESC expression. For fig. S1C, values are relative to control. Three
   different cell passages of pluripotent hESCs and four different neural
   inductions collected between days 1 and 6 were used for comparison.
   Ordinary one-way analysis of variance (ANOVA) followed by Dunnett’s
   multiple comparisons test was used for statistical analysis. Gal-3 and
   Pax6 levels in fig. S1D were normalized to the loading control H3, and
   aberrant replicates were removed. Four different neural inductions
   collected between days 1 and 7 were used for comparison. Mixed-effects
   model followed by Šídák’s multiple comparisons test was used for
   statistical analysis. For a complete list of antibodies used, see table
   S1.

Mice

   All animal experiments were approved by a local ethical review
   committee and conducted under the UK Animals (Scientific Procedures)
   Act, 1986 (ASPA), under valid personal and project licenses. Animals
   were bred and housed in individually ventilated cages on a 12-hour
   light/dark cycle in the Biomedical Sciences Building, Oxford. Water and
   food were given ad libitum. Cages were environmentally enriched with an
   activity wheel and nesting material. Wild-type (WT) C57BL/6J mice were
   acquired from Charles Rivers or from the University of Oxford
   Biomedical Services Specific Pathogens Free Breeding Unit. Gal-3 KO
   mice (Lgals3^tm1Ftl) were donated by A. Mead. At different embryonic
   stages, postnatal and adult animals were used. For embryonic studies
   where timed mating was required, breeding pairs were separated when
   females were plugged, which was considered E0.5.

MCP treatment in vivo

   Timed matings were set up by pairing C57BL/6J WT mice. Breeding pairs
   were separated on plug day E0.5. On the next day, single caged females
   were treated with 1% MCP diluted in water. Water containing the
   treatment was changed every other day for 17 days and also used to
   prepare mash. Females were tested for dehydration via pinch test, and
   no differences from control animals were found. For one female of each
   group, weight gain was calculated, and water consumption was measured
   via water bottle weight, with no statistically significant differences
   between them. The litters from two control and two MCP-treated females
   were collected at E17.5. Between both litters, there were a total of 15
   control embryos and 14 MCP-treated embryos.

RT-PCR of developmental brain sections

   Tissue from nine brain regions of C57BL/6J mice at different
   developmental stages were dissected, and RNA was extracted using the
   RNeasy Mini Kit (QIAGEN). Embryonic and postnatal tissue was pooled
   from pups within the same litter, while the adult age was from one
   mouse. RNA integrity was tested using the Agilent RNA Nano Kit with RNA
   integrity number (RIN) requirement of 8.0. One sample (one replicate of
   midbrain at P10) had RIN of 5.5 but had consistent results with other
   replicates; therefore, it was included in the final analysis. Reverse
   transcription was performed using the QuantiTect Reverse Transcription
   Kit (QIAGEN). RT-PCR was performed on an Applied Biosystems 7900HT Fast
   Real Time PCR System (Thermo Fisher Scientific) using QuantiTect Primer
   Assay (QIAGEN) and QuantiFast SYBR Green PCR Master Mix (Thermo Fisher
   Scientific), according to the manufacturer’s instructions.

   Data were initially analyzed using SDS Software V2.4 (Thermo Fisher
   Scientific), and aberrant replicates were removed. Tfrc was used as
   housekeeping gene, and it had efficiency between 90 and 110% and
   coefficient of determination (R^2) ≥ 0.998. Lgals3 had a minimum
   efficiency of 82.60% and R^2 ≥ 0.993. Ordinary one-way ANOVA followed
   by Dunnett’s multiple comparisons test was used for statistical
   analysis.

Immunohistochemistry of mouse brain tissue

   Embryos were collected at E17.5, and the brains were dissected and
   immersed in 4% PFA for 24 hours at 4°C. All brains were then
   cryoprotected in 30% sucrose until they sank (24 to 48 hours) and
   frozen down in dry ice. A mold was used for Tissue-Tek O.C.T. (Sakura)
   embedding before freezing. Coronal sections of 30- to 50-μm thickness
   were obtained using a Leica SM2000 R sliding microtome and kept in
   cryoprotectant solution at −20°C.

   Free-floating brain sections were washed three times in 0.1 M PBS
   before and after a 15-min incubation with 50 mM glycine in 0.1 M PBS to
   bind free aldehydes and reduce background staining. Porous well inserts
   were used to facilitate the transfer of the sections into different
   solutions. Sections were blocked in PBS^+ [10% donkey serum
   (Sigma-Aldrich) and 0.5% Triton X-100 in 0.1 M PBS] for 1 hour at room
   temperature, followed by incubation in primary antibodies diluted in
   PBS^+ at 4°C overnight. The next day, sections were washed with 0.1 M
   PBS before and after incubation with secondary antibodies (1:500) and
   DAPI (1:1000) in PBS^+. All washes and incubations were performed on a
   rocker at 10 to 20 oscillations/min. Sections were mounted in FluorSave
   Reagent (Merck Millipore). For a complete list of antibodies used, see
   table S1.

Human embryonic brain tissue immunofluorescence

   One 17-PCW human fetal tissue sample from a terminated pregnancy was
   provided by the joint MRC/Wellcome Trust-funded Human Developmental
   Biology Resource, University of Newcastle. The tissue was collected
   with appropriate maternal consent and approval from the Newcastle and
   North Tyneside National Health Service (NHS) Health Authority Joint
   Ethics Committee Specimens. The sample was shipped under a Material
   Transfer Agreement (MTA H-20068526) between the University of Newcastle
   and the University of Oxford as part of other studies conducted in the
   laboratory of Professor Molnár. The donated tissue was screened for
   gross brain abnormalities or common genetic defects and has been found
   to be suitable for this study by the personnel of the tissue bank
   before inclusion in the study. Sex and ethnicity of the donated
   specimen have not been disclosed as not considered relevant for the
   study. After collection, the brain sample was immediately fixed in 4%
   PFA (in 0.1 M PBS) for at least 24 hours with minimal postmortem delay,
   and it was then stored in 0.1 M PBS with 0.05% sodium azide at 4°C upon
   arrival to avoid bacterial contamination.

   The human brain sample was manually cut into coronal slabs (~0.5 to 1
   cm in thickness) before embedding in 3% low–gelling temperature Agarose
   (Sigma-Aldrich) and cut into 100-μm-thick coronal sections using a
   Vibratome VT1000S (Leica). Selected free-floating sections from a
   medial slab containing the prospective parietal cortex were subjected
   to antigen retrieval by incubation in 10 mM sodium citrate in 0.1 M PBS
   at 65°C for 20 min. Sections were then blocked in 20% normal donkey or
   goat serum (Sigma-Aldrich) 0.3% Triton X-100 (blocking solution) for at
   least 6 hours at room temperature. The tissue was then incubated with
   the primary antibodies diluted in blocking solution 48 hours at 4°C.
   Sections were washed for 1 hour with 0.1 M PBS before incubation with
   the appropriate secondary antibodies diluted in blocking solution
   (1:500). Sections were then washed for 1 hour in 0.1 M PBS,
   counterstained with DAPI in 0.1 M PBS (1:1000) for 30 min at room
   temperature, and lastly mounted on glass coverslips with FluorSave
   reagent (Merk Millipore) and sealed with nail polish to avoid drying.

   Due to the thickness and irregularity of the human tissue, z-stacks
   were acquired depending on the thickness of the slice, with ~1.0- to
   3.0-μm thickness at 63× objective in the LEICA TCS SP5 Broadband
   Confocal Laser Scanning Microscope. For a complete list of antibodies
   used, see table S1.

Human embryonic brain tissue immunohistochemistry

   One 14-PCW human fetus from a medically induced or spontaneous abortion
   was fixed in Bouin’s fixative, embedded in paraffin, and cut into a
   series of 10-μm-thick sections in the coronal plane. The material was
   part of previous studies ([279]91) and was donated to the laboratory of
   Zoltán Molnár by G. Meyer. The brain was acquired with maternal consent
   and approval by the Ethical Committee of the University Hospital La
   Laguna.

   Coronal 10-μm on-slide paraffin-embedded slices were used. Paraffin was
   melted for 30 min at 60°C and deparaffinized in xylene. After a wash in
   100% ethanol, endogenous peroxidase was blocked in 3% H[2]O[2] in
   methanol for 15 min at −20°C. Sections were rehydrated by an ethanol
   serial concentration wash (100, 80, and 50%) for 5 min each at room
   temperature. The sections were washed in 1× tris-buffered saline [TBS;
   10× stock solution: 0.2 M Trizma base and 1.5 M NaCl (pH7.4)] at room
   temperature for 20 min. For antigen retrieval, 10 mM citrate buffer
   solution was used to incubate the slices for 20 min at 60°C. Samples
   were cooled down and washed with TBS once before drying to allow
   circling the section with a hydrophobic pen. Unspecific antibody
   binding was blocked in 3% goat serum and TBS-T (TBS and 0.1% Triton
   X-100) for 1 hour at room temperature. Next, sections were blocked in
   avidin for 15 min and in biotin for 15 min to enhance the signal using
   the ready-to-use solution Avidin/Biotin Blocking Kit (Vector
   Laboratories). Sections were incubated with the primary antibody
   overnight at 4°C in antibody solution (1% goat serum and TBS-T). The
   following day, sections were washed three times in TBS-T. Sections were
   incubated for 3 hours at room temperature with the biotinylated
   secondary antibody diluted in 1:100 in antibody solution. Three 5-min
   washes in TBS-T and one in TBS preceded a 1-hour incubation with the
   VECTASTAIN Elite ABC-HRP Kit (Vector Laboratories), according to the
   manufacturer, to enhance the signal. A 3,3′-diaminobenzide (DAB)
   reaction was developed using a DAB 15 Substrate Kit, Peroxidase (HRP)
   (Vector Laboratories), according to the manufacturer. The reaction was
   blocked after 5 to 8 min by soaking slides in H2O and washing for 5
   min. The sections were counterstained in hematoxylin (Sigma-Aldrich)
   for a few seconds and washed in running water until clear. Next,
   sections were dehydrated in serial dilutions of ethanol (50, 80, and
   100%) for 5 min each and washed in xylene. Sections were air-dried and
   coverslipped using DePeX mounting medium (Serva). Slides were left to
   dry overnight in the flow hood and sealed with nail polish the next
   day.

scRNA-seq analysis

   Data analyzed here were previously generated ([280]61). The
   t-distributed stochastic neighbor embedding (tSNE) on Weighted Gene
   Co-expression Network Analysis (WGCNA) plot was generated using the
   UCSC Cell Browser ([281]92) and reannotated manually.
   Transcript-per-million (TPM) values for LGASL3 expression per WGCNA
   cluster were exported from USCS Cell Browser. Exported TPM values were
   converted in TPM + 1. Six RG clusters were plotted separately
   (RG-early, vRG, tRG, oRG, RG-div2, and RG-div1). All other cell types
   were plotted together, excluding the cluster Glyc, the four unknown
   clusters (U1, U2, U3, and U4) and other unidentified numbers. Outliers
   were removed using the Robust regression and Outlier removal (ROUT)
   method (Q = 1%). P values were obtained from the Wilcoxon’s rank sum
   test.

Image acquisition

   Images were taken either at ZEISS LSM 710 or 780 Laser Scanning
   Confocal Microscope using ZEN software; Leica TCS SP5 Broadband
   Confocal Laser Scanning Microscope using Leica Application Suite
   Advanced Fluorescence; or Nikon ECLIPSE Ti inverted microscope system
   using Volocity version 7.0.0.

   Mouse brain tile scans were produced using the Nikon ECLIPSE Ti
   inverted microscope with a 10× objective with a 10% overlap and
   brightness but no shading. For human brain immunohistochemistry, a
   Leica DMR Microscope Upright Brightfield Fluorescent and Neurolucida
   software (MBF Bioscience) was used. Stainings were compared to
   no-primary, secondary-only controls to assess nonspecific binding.
   Images were analyzed using ImageJ.

Image analysis

   Neural rosettes 20× images were obtained on a Nikon ECLIPSE Ti inverted
   microscope. Three images per well were obtained. The number of neural
   rosettes per image was counted manually using ImageJ and then converted
   to number of rosettes per square millimeter. For [282]Fig. 2C, an
   average of nine wells was used for comparison. Ordinary one-way ANOVA
   followed by Tukey’s multiple comparisons test was used for statistical
   analysis.

   For analysis of Gal-3 colocalization in different vesicles, 63× images
   with 4× or 6× zoom were obtained in the Zeiss LSM 710 confocal. Maximum
   projection images of five stacks at 0.5 μm of distance were obtained. A
   macro was created on Fiji for processing the images with Gaussian Blur
   for posterior removal of background and thresholding. Images of Gal-3
   and vesicle markers were then converted into a binary mask, and the
   resulting image of the operation “imageCalculator AND” on Fiji was
   considered the overlap between both channels. Particles or area of the
   overlap channel were calculated, and the percentage over Gal-3 channel
   was considered “Portion of Gal-3,” while the percentage of the vesicle
   channel was considered “Portion of vesicle.” Average of percentage was
   calculated for three biological replicates, each being the average of
   two technical replicates.

   For [283]Fig. 2 (G′, G″, H′, and H″), an average of three wells was
   used for comparison. Ordinary one-way ANOVA followed by Tukey’s
   multiple comparisons test was used for statistical analysis. For
   [284]Fig. 2 (G‴, G⁗, H‴, and H⁗), an average of three wells was used
   for comparison. Unpaired t test was used for statistical analysis.

   For in vitro activated–caspase-3 analysis, images were obtained in
   Nikon ECLIPSE Ti inverted microscope at 20×. The number of cleaved
   caspase-3 (aCasp3)–positive cells and DAPI^+ nuclei was manually
   counted using ImageJ, and a percentage was obtained. Three images per
   well were obtained. An average of three wells was used for comparison.
   Ordinary one-way ANOVA was used for statistical analysis.

   For in vitro YAP/DAPI and YAP/ZO1 colocalization analysis, confocal
   images were taken at 63× with z-stack interval of 0.5 μm. JACoP BIOP
   ImageJ plugin was used to calculate the Pearson’s correlation
   coefficient in region of interest (nuclei: DAPI signal with Gaussian
   blur = 2 σ thresholded using Huang; junctions: ZO1 signal with Gaussian
   blur = 2 σ thresholded using Li). Automatic signal thresholding was
   used: Huang for DAPI, Otsu for YAP staining, and Moments for ZO1
   staining. Nine images per group per repeat were analyzed. Unpaired t
   test was used for statistical analysis.

   For ZO1 analysis in vivo, confocal images of the ventricular and
   subventricular zones of anterior, middle, and posterior sections were
   captured at ×40 magnification with a z-stack interval of 1 μm, 20
   optical sections per z stack. The average thickness of the apical
   region was measured using ImageJ. Brains of six controls and six
   MCP-treated mice from two litters treated E17.5 mice embryos were used
   for this analysis. Mann-Whitney test was used for statistical analysis.

   For in vivo analysis for cell division angle, RGs in anaphase located
   at the ventricular surface of the cortex were imaged with confocal
   microscope z-stack intervals of 0.4 μm and between 40 and 80 optical
   sections per z stack at 63× oil lens. For each embryo, 80 to 120 NSCs
   in anaphase were imaged. The cleavage plane angle relative to the
   ventricular surface was measured using ImageJ. Cleavage plane angles
   below 60° were classified as vertical divisions, while those at or
   above 60° were considered horizontal divisions. Brains of three control
   and three MCP-treated E17.5 mice embryos from two litters were used for
   this analysis. Mann-Whitney test was used for statistical analysis.

   For Pax6, Tbr2, and Ki67 analysis in embryonic sections, tile scans of
   cortex were captured at ×63 magnification with a confocal microscope,
   with a z-stack interval of 4 μm and three optical sections per z stack.
   For each animal, anterior, middle, and posterior cortical sections
   measuring 250 μm in width were analyzed. The cortex was categorized
   into VZ and SVZ, IZ, subplate, and cortical plate on the basis of their
   morphological characteristics visualized through DAPI staining. Manual
   counting of the number of positive and double-positive cells in
   different cortical regions was performed using the Cell Counter tool in
   ImageJ. Brains of four control and four MCP-treated E17.5 embryos from
   two litters were used for this analysis. Unpaired t test was used for
   statistical analysis.

   For quantification of cortical folding, all observed gyri and sulci
   were described using SD and LGI values ([285]65). To calculate the SD
   value, a line was drawn connecting the highest point of the gyrus with
   the lowest point of the sulcus. The LGI was determined by the contour
   of the gyrus divided by the length of the surrounding gyrus. Foldings
   were classified as gyri/sulci if the SD value exceeded 100 μm and the
   GI value exceeded 0.5. Then, the percentage of animals with one or more
   cortical foldings in each mouse litter and overall was calculated.

   To determine the GI value, the total length of the contour surrounding
   the brain hemisphere, including sulci, was divided by the length of
   brain hemisphere not including the sulci. Two brain sections per animal
   were used for this calculation. All animals containing sulci were
   quantified, and the average value was used for statistical analysis
   with Kruskal-Wallis test followed by Dunn’s multiple comparisons test.

   For the notches analysis, a notch was defined as an indentation along
   the SVZ, excluding natural inflection points associated with normal
   ventricular curvature. All coronal brain sections in which the SVZ was
   visibly present were analyzed. Images were acquired at ×10
   magnification, and their scale was calibrated in ImageJ. Notches
   measuring less than 15 μm in depth were excluded to control for
   artifacts introduced by histological staining. Kruskal-Wallis test
   followed by Dunn’s multiple comparisons test was used for statistical
   analysis.

Statistical analysis

   All experiments were repeated at least three times unless specified.
   Biological replicates were composed of different animals or different
   neural inductions started at different cell passages. Technical
   replicates were composed of different brain sections of the same animal
   or separated wells of same neural induction.

   Analysis was performed using the statistical package GraphPad Prism 10,
   unless otherwise indicated. Experiments were tested for normality
   distribution with Shapiro-Wilk test. *P < 0.05; **P < 0.01;
   ***P < 0.001; ****P < 0.0001

Acknowledgments