Abstract

   Brain development is orchestrated by a complex interplay of genetic and
   environmental signals, with endocytosis serving as a pivotal process in
   integrating extracellular cues. However, the specific role of
   endocytosis in neurogenesis remains unclear. We uncover a critical
   function of the interferon‐induced transmembrane protein, IFITM2,
   essential for endocytic processes in radial glial cells (RGCs). IFITM2
   is highly expressed near the ventricular surface in the developing
   brain. Loss of IFITM2 impairs endosome formation and disrupts RGC
   maintenance. Mechanistically, we confirmed that the YXXø endocytic
   motif on IFITM2 is essential for its subcellular localization, with
   mutations in this motif reducing endocytic vesicles. Additionally, the
   K82 and K87 residues of IFITM2 interact with phosphoinositides to
   promote endocytic vesicle formation. Polarized localization of
   phosphatidylinositol 3,4‐bisphosphate (PI(3,4)P2) on the ventricular
   side suggests its role in vesicle formation. IFITM2 deficiency also
   leads to reduced phosphorylation of AKT and GSK3β. These findings
   highlight the essential role of IFITM2 in regulating endocytosis in
   RGCs, which is critical for maintaining neural stem cells and proper
   brain development, offering new insights into the connection between
   cellular signaling and neurogenesis in both mouse and human models.

   Keywords: brain development, endocytosis, IFITM2, neural stem cell,
   phosphoinositide
     __________________________________________________________________

   This study reveals IFITM2 as a key regulator of endocytosis in radial
   glial cells (RGCs) during brain development. IFITM2's YXXø motif and
   phosphoinositide‐binding residues (K82/K87) are essential for endosome
   formation. IFITM2 deficiency disrupts the embryonic cortical
   neurogenesis, revealing a conserved mechanism linking endocytic
   dynamics to neural stem cell homeostasis in mammals.

   graphic file with name ADVS-12-2501593-g001.jpg

1. Introduction

   The cerebral cortex, responsible for high‐order cognitive and
   sensorimotor processing, evolves through complex mechanisms involving
   the proliferation and differentiation of neural progenitor cells.^[
   [36]^1 ^] These dynamic processes are regulated by intrinsic genetic
   programs and various environmental cues. Recent evidence has
   underscored the critical role of endocytosis, particularly the
   asymmetric inheritance of endolysosomes—organelles formed from the
   fusion of endosomes and lysosomes—in determining the fate of radial
   glial progenitor cells (RGCs).^[ [37]^2 ^] This pivotal role highlights
   the broader importance of the endocytic pathway in neocortical
   development. Further, the dynamics of endosomes are increasingly
   recognized as a conserved evolutionary strategy for cortical
   neurogenesis, evidenced by similar mechanisms in Drosophila, where
   endosome behavior affects Notch signaling during neural progenitor cell
   division.^[ [38]^3 ^] Despite these insights, the detailed molecular
   and mechanistic aspects of endocytosis in neocortical development
   remain poorly defined.

   This study seeks to delineate the role of interferon‐induced
   transmembrane protein 2 (IFITM2) in the developing mammalian neocortex.
   Known primarily for its antiviral defenses, IFITM2 inhibits viral
   infection by altering membrane curvature, impacting the stages of
   cellular entry and viral fusion.^[ [39]^4 ^] Intriguingly, elevated
   expressions of IFITM2 in RGCs suggest its potential beyond
   immunological functions, possibly impacting neurodevelopmental
   processes. This notion is supported by analogous functions observed in
   IFITM3, which has been implicated in modulating neurodegenerative^[
   [40]^5 ^] and oncogenic pathways.^[ [41]^6 ^] Our findings reveal a
   novel function of IFITM2 in regulating endocytosis in the developing
   brain and elucidate how IFITM2 interacts with polarized PI(3,4)P2
   signaling in RGCs, offering fresh insights into the molecular
   orchestration of brain development.

2. Results

2.1. Spatiotemporal Consistency of Endocytosis and IFITM2 Expression in Mouse
and Human Developing Neocortex

   In the developing mouse neocortex, we used immunostaining to label
   radial glial progenitor cells (RGCs), intermediate progenitor cells
   (IPs), and differentiated neurons with SOX2, TBR2, and TUJ1,
   respectively (Figure [42] 1A). The ventricular zone (VZ) is
   predominantly populated by RGCs, the sub‐ventricular zone (SVZ) by IPs,
   and the intermediate zone (IZ) and cortical plate (CP) by post‐mitotic
   neurons. To assess endocytosis in the neocortex, we stained early
   endosomes with EEA1 on embryonic day (E) 15.5, revealing a high density
   of early endosomes near the lateral ventricles, indicating robust
   endocytic activity in the RGCs near the ventricular surface
   (Figure [43]1B,C). We tracked early endosome numbers on the ventricular
   surface at E11.5, E13.5, and E15.5, observing a progressive decline
   from early to late neurogenesis, suggesting a reduction in the
   endocytic capacity of RGCs over time (Figure [44]S1A–C, Supporting
   Information). In addition, EEA1 signaling exhibits regional specificity
   near the ventricles. Compared to the ganglionic eminence region
   adjacent to the ventricles, the neocortex shows a higher density of
   endocytic signals (Figure [45]S1A–C, Supporting Information). These
   findings suggest that endocytosis in the developing neocortex is
   spatially regulated, with RGCs near the ventricular surface displaying
   robust endocytic activity, which gradually decreases as neurogenesis
   progresses. This regional variation in endocytic activity might reflect
   distinct functional roles of endocytosis in different progenitor
   populations during cortical development.

Figure 1.

   Figure 1
   [46]Open in a new tab

   Endocytosis and IFITM2 expression show spatiotemporal consistency in
   the developing mouse neocortex and human cerebral organoids. A)
   Representative image of E15.5 neocortex stained for SOX2 (green), TBR2
   (red), TUJ1 (magenta), and nuclei with DAPI (blue). Scale bar, 50 µm.
   VZ, ventricular zone; SVZ, subventricular zone; IZ, intermediate zone;
   CP, cortical plate. B,C) Representative images of E15.5 neocortex
   stained for EEA1 (early endosome antigen 1, green), TBR2 (red), and
   nuclei with DAPI (blue). The red rectangle indicates the ventricular
   surface region of the neocortex, which is displayed at higher
   magnification at the bottom. The white rectangle marks the region where
   EEA1 fluorescence intensity was quantified in panel (C). Scale bars in
   (B), 50 µm (above) and 25 µm (below). D) UMAP visualization of the
   scRNA‐seq data from mouse E13.5 somatosensory neocortex.^[ [47]^11a ^]
   Unsupervised Leiden clustering and classification of cells after
   integration with Harmony algorithm. E) UMAP visualization of expression
   levels (normalized) of marker genes for radial glial cells (Sox2),
   intermediate progenitors (Eomes), newborn neurons (Neurod1), and
   Ifitm2. F) Heatmap showing processes enriched in different cell types
   from scRNA‐seq data for E13.5 mouse neocortex. G) Q‐PCR analysis of
   Ifitm family genes expression from the E11.5 mouse brains (n = 3
   independent experiments). All Ct values were normalized to Gapdh
   control. Data are shown as mean ± s.e.m. One‐way ANOVA followed by
   Tukey's post‐hoc test was used. H) Q‐PCR analysis of Ifitm2 mRNA levels
   in CD133^+CD24^− RGCs at E12.5, E14.5, and E16.5 stages (n = 3
   independent experiments). All Ct values were normalized to Gapdh
   control. Data are shown as mean ± s.e.m. Two‐tailed unpaired Students’
   t‐test was used. I,J) Representative images of day 58 human cerebral
   organoid immunostaining with IFITM2 (red), SOX2 (green), and nuclei
   with DAPI (blue). Scale bar, 20 µm. The white rectangle marks the
   region where IFITM2 fluorescence intensity was quantified in (J).K,L)
   Representative images of day 58 human cerebral organoid immunostaining
   with EEA1 (red), SOX2 (green), and DAPI (blue). Scale bar, 20µm. The
   white rectangle marks the region where EEA1 fluorescence intensity was
   quantified in (L).

   Given the limited understanding of the molecular basis of endocytosis
   in RGCs, we analyzed single‐cell RNA transcriptional profiles from the
   E13.5 developing somatosensory cortex.^[ [48]^7 ^] Initial
   identification of cell types in the neocortex highlighted a correlation
   between RGCs and endocytic biological processes (Figure [49]1D–F).
   Notably, our analysis identified a high expression of IFITM2 in RGCs
   (Figure [50]1E). Moreover, IFITM2 expression was significantly higher
   than that of IFITM1 and IFITM3 within the IFITM family at E11.5, with
   negligible expression of other family members (Figure [51]1G). Flow
   cytometry of isolated CD133^+CD24^− RGCs further confirmed high and
   progressively decreasing IFITM2 expression from early to late
   neurogenesis (Figure [52]1H; Figure [53]S1D, Supporting Information).
   Next, we stained human cerebral organoid slices for IFITM2. We observed
   a dense presence of IFITM2 near ventricle‐like structures
   (Figure [54]1I,J; Figure [55]S1E, Supporting Information), which was
   similar to the EEA1 staining near the ventricular surface
   (Figure [56]1K,L; Figure [57]S1F, Supporting Information). These
   observations, in conjunction with single‐cell sequencing data, suggest
   that IFITM2 may play a role in modulating the extensive endocytosis
   observed in the developing neocortex.

2.2. Loss of IFITM2 Promotes Differentiation of RGCs Near the Ventricular
Surface

   To determine the role of the IFITM2 in developing the neocortex, we
   engineered a mouse with a conditional Ifitm2 mutant allele, using the
   CRISPR‐Cas9 mediated double‐nicking strategy (Figure [58]S2A,B,
   Supporting Information). These mice were crossed with Nestin ^cre mice,
   which express Cre recombinase in neural progenitor cells. Genotyping of
   pups on the first postnatal day showed a Mendelian ratio, and the pups
   survived to adulthood. Moreover, Ifitm2 deletion did not result in high
   expression compensation in other members of the IFITM family (Figure
   [59]S2C, Supporting Information).

   During embryonic neurogenesis, neural stem cells exhibit dynamic
   nuclear migration, a phenomenon first described by Sauer et al. in
   1935.^[ [60]^8 ^] This process is characterized by the oscillatory
   movement of the RGC nucleus during the cell cycle, with mitosis
   occurring near the ventricular region. To investigate cell fate
   determination, we focused on the region adjacent to the ventricle,
   where cells have just completed division (Figure [61] 2A). We found
   that IFITM2 deficiency led to an increase in TBR2^+ IPs at E12.5, along
   with a significant reduction in SOX2 fluorescence near the ventricular
   surface (Figure [62]2A–E). Additionally, staining for the neural
   progenitor cell marker PAX6 showed decreased fluorescence intensity in
   the same region (Figure [63]S2D,E, Supporting Information). TBR2
   expression was found closer to the ventricular surface in the Ifitm2
   conditional knockout (cKO) neocortex compared to wild‐type, further
   suggesting enhanced differentiation (Figure [64]2D). These results
   indicate that the removal of IFITM2 promotes RGC differentiation near
   the ventricular surface, leading to an increased density of IPs during
   early neurogenesis.

Figure 2.

   Figure 2
   [65]Open in a new tab

   Impact of IFITM2 deletion on neural progenitor dynamics and cortical
   development. A) Representative images of E12.5 wild‐type and Ifitm2 cKO
   cerebral cortex sections stained with SOX2 (red), TBR2 (green), and
   DAPI (blue). The yellow arrow indicates the weakened signal of SOX2,
   and the white arrow indicates the appearance of the TBR2 signal near
   the ventricle area in Ifitm2 cKO. Scale bars, 50 µm (left) and 20 µm
   (right). B,C) Quantification of the number of TBR2^+ cells and SOX2^+
   cells in the wild‐type and Ifitm2 cKO cortices (n = 5 brains per
   group). Data are shown as mean ± s.e.m. Two‐tailed unpaired Students’
   t‐test was used.D) Quantification of the number of TBR2^+ cells in the
   areas 0–50, 50–100, and 100–150 µm from the ventricle (n = 3 brains per
   group). Data are shown as mean ± s.e.m, ^* p < 0.05, ^** p < 0.01.
   Two‐tailed unpaired Students’ t‐test was used. E) Quantification of
   SOX2 fluorescence intensity in cells located on the surface of the
   brain ventricle (n = 60 cells from 3 brains per group). Two‐tailed
   unpaired Students’ t‐test was used. F) Representative images of
   wild‐type and Ifitm2 cKO neocortex that were injected with EdU at E12.5
   and analyzed 24 h later. The tissues were stained with EdU (green),
   Ki67 (red), and DAPI (blue). Scale bar, 50 µm. G) Quantification of the
   percentage of EdU^+Ki67^+ cells among total EdU^+ cells (n = 4 brains
   per group). Data are shown as mean ± s.e.m. Two‐tailed unpaired
   Students’ t‐test was used. H) Representative images of E11.5, E14.5,
   and E16.5 wild‐type and Ifitm2 cKO cerebral cortical sections stained
   with SOX2 (red), TBR2 (green), and DAPI (blue). Scale bars, 50 µm. I,J)
   Quantification of the ratio of the TBR2^+ cells over SOX2^+ cells (I)
   and the number of the SOX2^+ cells (J) (n = 3 brains per group). Data
   are shown as mean ± s.e.m. Two‐tailed unpaired Students’ t‐test was
   used. K) Representative images of P0 wild‐type and Ifitm2 cKO cerebral
   cortex sections stained with TBR1 (red), a marker of layer VI neurons,
   and DAPI (blue). Scale bar, 200 µm. L) Quantification of the number of
   TBR1^+ cells in the wild‐type and Ifitm2 cKO cortices (n = 5 brains per
   group). Data are shown as mean ± s.e.m. Two‐tailed unpaired Students’
   t‐test was used. M) Representative images of P10 wild‐type and Ifitm2
   cKO brain sections stained with CTIP2 (red), CUX1 (green), and DAPI
   (blue). White rectangles indicate a dorsal region, which is shown at a
   higher magnification cortical area in the right. Scale bars, 1 cm
   (left) and 200 µm (right). N,O) Quantification of the number of CUX1^+
   neurons (N) and CTIP2^+ neurons (O) per 200 µm column (n = 5 brains per
   group). Data are shown as mean ± s.e.m. Two‐tailed unpaired Students’
   t‐test was used.

   We then used the FlashTag method that enables in utero labeling of RGCs
   during mitosis^[ [66]^9 ^] to trace the behaviors of RGCs after cell
   division. The RGC progeny were analyzed 12 h after labeling.
   FlashTag‐labeled cells in the Ifitm2 cKO neocortex revealed an
   increased proportion of TBR2^+ cells and a decreased presence of
   SOX2^+TBR2^− cells compared to the wild‐type neocortex (Figure
   [67]S2F,G, Supporting Information). This result suggests that the
   removal of IFITM2 leads to an increased differentiation of RGC progeny.
   Next, we used EdU to label proliferating cells and analyzed them 24 h
   later. The results showed that IFITM2 deficiency led to a higher
   proportion of EdU^+KI67^− cells exiting the cell cycle
   (Figure [68]2F,G). Additionally, neural progenitor cells isolated from
   E12.5 cortices revealed that neurospheres formed from Ifitm2 cKO neural
   progenitor cells were smaller than those formed from the wild‐type
   neural progenitor cells, corroborating the in vivo results with similar
   findings in vitro (Figure [69]S2H,I, Supporting Information).

   We subsequently explored whether RGCs are progressively decreased from
   early to late stages of neurogenesis following the removal of Ifitm2.
   By employing SOX2 and TBR2 immunostaining in the neocortex from E11.5
   to E16.5, we observed higher ratios of TBR2^+ cells over SOX2^+ cells
   from early to late neurogenesis (Figure [70]2H–J), together with fewer
   SOX2^+ cells in the E14.5 and E16.5 cerebral cortex of cKO compared
   with wild‐type mice, indicating a gradual decrease of RGCs. These
   findings suggest that the removal of IFITM2 leads to increased
   differentiation of RGCs and reduced RGC pool. Moreover, using EdU
   labeling to assess neurogenesis in postnatal day (P) 21 and adult mice,
   we found that the loss of IFITM2 did not significantly affect postnatal
   neurogenesis (Figure [71]S3A–H, Supporting Information).

   During embryonic cortical neurogenesis, the generation of neurons
   follows a highly orchestrated inside‐out pattern, with deeper layer
   neurons born first and upper layer neurons born later.^[ [72]^10 ^] To
   examine the impact of increased differentiation of RGCs and reduced RGC
   pool on the generation of cortical neurons in Ifitm2 cKO brains, we
   stained P0 brain sections for TBR1, a marker for layer‐VI pyramidal
   neurons, and observed an increase in the number of TBR1^+ deep‐layer
   neurons, suggesting an increase in early‐born neurons
   (Figure [73]2K,L). Then, P10 brain sections were stained for CTIP2, a
   marker for layer‐V and layer‐VI neurons, and CUX1, a marker for layers
   II‐IV neurons (Figure [74]2M–O).^[ [75]^11 ^] We found an increased
   number of CTIP2^+ neurons, and a significant decreased number of CUX1^+
   neurons in Ifitm2 cKO brains compared with the wild‐type, indicating an
   increase in early‐born deep‐layer neurons and a reduction in later‐born
   superficial‐layer cortical neurons.

2.3. Loss of IFITM2 Impairs RGC Endocytosis

   In human‐derived cerebral organoids, we observed the co‐localization of
   IFITM2 with both EEA1 and LAMP1, confirming its localization to early
   endosomes and endolysosomes (Figure [76] 3A,B). Additionally, by
   electroporating HA‐IFITM2 and GFP‐2 × FYVE into mouse RGCs, which
   labels phosphatidylinositol 3‐phosphate (PI3P)‐containing endocytic
   vesicles, we observed that IFITM2 co‐localized with PI3P^+ vesicles in
   RGCs, with the distribution of IFITM2 also observed near the
   ventricular surface (Figure [77]3C). To investigate the effects of
   Ifitm2 deletion on the endocytic process, we performed immunostaining
   for RAB7, a late endosome marker, and found a significant reduction in
   the number of late endosomes in the Ifitm2 cKO cortices
   (Figure [78]3D,E). To further confirm the impact on endocytosis, we
   injected 10 kD FITC‐dextran into the lateral ventricle of E14.5 brains
   and collected samples after 20 min. A marked reduction in fluorescence
   intensity of FITC‐dextran was observed in the Ifitm2 cKO neocortex
   compared to the wild‐type, indicating a reduced uptake of dextran
   (Figure [79]3F,G). Our previous work demonstrated that endolysosomes
   are the predominant type of lysosomes in embryonic RGCs.^[ [80]^2 ^]
   Here, using LAMP1 to label lysosomes, we found that IFITM2 deletion led
   to a significant reduction in the number of lysosomes
   (Figure [81]3H,I). Together, these findings underscore the crucial role
   of IFITM2 in regulating endocytosis during early cortical neurogenesis.

Figure 3.

   Figure 3
   [82]Open in a new tab

   Ifitm2 deletion disrupts endocytosis on the cortical ventricular
   surface. A) Representative images of day 55 human cerebral organoid
   stained with IFITM2 (green), LAMP1 (red), and DAPI (blue). Arrows in
   the magnified images indicate the colocalization of IFITM2 with LAMP1^+
   lysosomes. Scale bars, 10 and 5 µm. B) Representative images of day 55
   human cerebral organoid stained with IFITM2 (red), EEA1 (green), and
   with DAPI (blue). Arrows in the magnified images indicate the
   colocalization of IFITM2 with EEA1^+ endosomes. Scale bars, 10 and
   5 µm. C) Representative images of RGC introduced with HA‐IFITM2 (red),
   GFP‐2 × FYVE (green), and mCherry (blue) in mice. The white arrow in
   magnified image indicates the apical polarized localization of IFITM2
   in RGC, while the yellow arrow indicates the colocalization of IFITM2
   with PI3P^+ vesicle. Scale bars, 5 and 2 µm. D) Representative images
   of the E14.5 wild‐type and Ifitm2 cKO stained with RAB7 (gray) in the
   ventricular zone. Scale bars, 50 µm. E) Quantification of relative
   fluorescence intensity for RAB7 (n = 4 brains per group). Data are
   shown as mean ± s.e.m. Two‐tailed unpaired Students’ t‐test was used.
   F) Representative images of E14.5 wild‐type and Ifitm2 cKO brains
   20 min after injecting 10 kD FITC‐Dextran. Scale bar, 10 µm. G)
   Quantification of the FITC‐Dextran fluorescence intensity (n = 4 brains
   per group). Data are shown as mean ± s.e.m. Two‐tailed unpaired
   Students’ t‐test was used. H) Representative images of E14.5 the
   wild‐type and Ifitm2 cKO stained with LAMP1, a marker for endolysosome.
   Scale bar, 20 µm. I) Quantification of relative fluorescence intensity
   of LAMP1 in the VZ region (n = 5 brains per group). Data are shown as
   mean ± s.e.m. Two‐tailed unpaired Students’ t‐test was used. J)
   Representative TEM images of three stages of the endocytosis process on
   the cortical ventricular surface. Scale bar, 100 nm. K) Representative
   TEM images of the E14.5 wild‐type and Ifitm2 cKO cortical ventricular
   surface. The red arrows indicate U‐shaped stage one, the blue arrow
   indicates the Ω‐shaped stage two, and the yellow arrows indicate the
   endo‐vesicles of stage three. Scale bar, 250 nm. L) Quantification the
   number of structures in three stages in each cell (n = 30 cells from 3
   embryos per group). Data are shown as mean ± s.e.m. Two‐tailed unpaired
   Students’ t‐test was used. M) Representative images of wild‐type and
   Ifitm2 cKO primary neural stem cells endocytosing 10 kD FITC‐Dextran.
   The samples were collected at 10, 20, 30, and 60 min. Scale bar, 10 µm.
   N) Quantification of the relative FITC intensity (n = 12 cells per
   group). Data are shown as mean ± s.e.m. Two‐tailed unpaired Students’
   t‐test was used.

   We next performed transmission electron microscopy (TEM) to visualize
   the membrane structure on the cortical ventricular surface (Figure
   [83]S4, Supporting Information). Endocytosis exhibits three stages:
   stage one involves the formation of a U‐shaped invagination in the cell
   membrane, stage two entails the contraction of the cell membrane into
   an Ω‐shaped half‐vesicle, and stage three represents the completion of
   membrane contraction during endocytosis (Figure [84]3J).^[ [85]^12 ^]
   Comparing the wild‐type and the Ifitm2 cKO neocortices, we found that
   the knockout of IFITM2 resulted in an increase in stage one, a slight
   but insignificant reduction in stage two, and a significant reduction
   in stage three (Figure [86]3K,L). This result suggests that deletion of
   IFITM2 delays the transition from stage one to stage two, that is, the
   closure of endocytosis vesicles. Furthermore, in vitro experiments
   conducted on primary NPCs revealed a decrease in endocytosis of the
   dextran in Ifitm2 cKO cells compared with wild‐type cells at 10 and
   20 min, while there was no difference after 30 min, suggesting that
   IFITM2 is a rate‐limiting factor for endocytosis (Figure [87]3M,N).
   Together, these results suggest that IFITM2 deletion hinders the
   endocytosis near the brain ventricle surface, specifically delaying
   endocytic vesicle closure.

2.4. Structural Insights and Functional Regulation of IFITM2 in Endocytosis

   The functionality of a protein is intrinsically linked to its
   structural configuration. Through a detailed analysis of the IFITM2
   amino acid sequence, we identified several key structural features,
   including a single transmembrane domain, an amphipathic helix, a YXXø
   endocytic motif, and a phosphoinositide (PtdIns) binding motif
   (Figure [88] 4A). Structural modeling of IFITM2 further suggests that
   it is a single‐pass transmembrane protein, with its N‐terminal region
   predominantly localized intracellularly, while the C‐terminal
   extracellular domain is composed of only a few amino acids, and thus
   this topology indicates that IFITM2 is unlikely to interact with
   extracellular ligands (Figure [89]4B).

Figure 4.

   Figure 4
   [90]Open in a new tab

   Structural features and functional mechanisms of IFITM2 in endocytosis
   regulation. A) Amino acids sequence and potential modification sites of
   mouse IFITM2 protein. B) Predicted binding model of IFITM2 to the
   plasma membrane. C) Helical wheel projection plots of the amphipathic
   helix of IFITM2 protein. Hydrophobic residues are displayed as yellow,
   and polar uncharged residues are displayed as blue. Arrow represents
   the magnitude and orientation of the mean hydrophobic moment. D,E)
   Representative images of Neuro‐2a cells transfected with HA‐IFITM2 (D),
   HA‐Y19A mutant (D), HA‐K82A mutant (E), or HA‐K87A mutant (E), and
   stained with HA antibody (red), Wheat Germ Agglutinin (green), and DAPI
   (blue). Scale bars, 10 µm. F) Representative images of primary neural
   progenitor cells transfected with HA‐IFITM2, HA‐Y19A, HA‐K82A, or
   HA‐K87A lentivirus for 2 days, followed by incubation with 10 kD
   RB‐Dextran for 20 min. Scale bar, 10 µm. G) Quantification of the
   fluorescence intensity of endocytic RB‐Dextran signals per cell (n = 8
   cells per group). Data are shown as mean ± s.e.m. Two‐tailed unpaired
   Students' t‐test was used. H) Representative images of RGCs
   electroporated with the GFP reporter (green) and EV, HA‐IFITM2, or
   HA‐Y19A. Sections were stained with EEA1 (red). Scale bar, 10 µm. I)
   Quantification of the number of EEA1 in RGC (n = 20 cells from 3
   embryos per group). Data are shown as mean ± s.e.m. Two‐tailed unpaired
   Students' t‐test was used. J) Co‐immunoprecipitation assays were
   performed to detect the interaction between AP2M1 and IFITM2 or Y19A
   using HA‐tagged beads in HEK293T cells.

   The amphipathic helix is characterized by the juxtaposition of
   hydrophilic and hydrophobic residues on opposite faces of the helix,
   enabling it to interact with both aqueous environments and lipid
   membranes (Figure [91]4C).^[ [92]^4d ^] This amphipathic nature
   facilitates the formation of a “wedge” effect, wherein the helix
   inserts into one leaflet of the bilayer, thereby inducing changes in
   membrane curvature—a crucial mechanism in vesicle formation.

   In addition, the YXXø endocytic motif and the PtdIns binding motif
   appear to be pivotal regulatory elements governing the function of
   IFITM2. We initially focused on investigating the YXXø endocytic motif.
   Through mutation screening of potential phosphorylation sites, we found
   that the Y19A mutant predominantly localized at the plasma membrane,
   and thus pinpointed the Y19 residue within the YXXø motif as a critical
   phosphorylation site (Figure [93]4D; Figure [94]S5A,B, Supporting
   Information). Electroporation of the Y19A mutant into cells, followed
   by dextran uptake assays, revealed that overexpression of Y19A
   significantly impaired endocytosis, indicating that this mutation
   disrupts IFITM2's functional capacity (Figure [95]4F,G). In addition,
   in vivo analysis demonstrated a substantial reduction in the number of
   endocytic vesicles in RGCs expressing the Y19A mutant
   (Figure [96]4H,I). Importantly, the phosphorylation of the Y19 residue
   is consistent with the conserved YXXø motif, which is a hallmark of the
   IFITM family.^[ [97]^13 ^] This motif serves as a sorting signal in the
   cytosolic tails of transmembrane proteins, orchestrating the
   recruitment of adaptor protein complexes essential for clathrin‐coated
   vesicle formation at the plasma membrane. The Y19A mutation, however,
   abolishes the interaction with adaptor protein 2 (AP2), thereby
   inhibiting vesicle formation (Figure [98]4J). Together, these findings
   suggest that the YXXø motif of IFITM2 is a critical phosphorylation
   site that regulates endocytosis in neural progenitor cells.

   Additionally, previous studies have demonstrated that IFITM3 acts as a
   scaffold for phosphoinositides, amplifying their signaling.^[ [99]^14
   ^] Given the high sequence conservation between IFITM2 and IFITM3, we
   hypothesized that IFITM2 might also play a similar role. We
   investigated the PtdIns binding motif by introducing mutations at K82
   and K87, which led to the loss of IFITM2 localization at the plasma
   membrane (Figure [100]4E). We introduced these mutants alongside GFP‐2
   × FYVE into the developing neocortex and observed that the mutations
   significantly reduced co‐localization with GFP‐2 × FYVE^+ vesicles and
   diminished their size (Figure [101]S5C–E, Supporting Information).
   These results underline the critical roles of residues K82 and K87 in
   binding PtdIns and regulating endocytosis. Together, these findings
   suggest that the PtdIns binding motif of IFITM2 is also critical for
   its regulation of endocytosis. Encouraged by these results, it is
   necessary to further understand the roles of phosphoinositides in the
   developing neocortex.

2.5. Regulation of Endocytosis by Phosphoinositides and IFITM2 in Developing
Neocortex

   Phosphoinositide signaling plays a pivotal role in the regulation of
   endocytosis by modulating vesicle formation and trafficking
   (Figure [102] 5A). PI3P primarily localizes to intracellular vacuoles,
   while PI(3,4)P2 and phosphatidylinositol trisphosphate (PI(3,4,5)P3)
   predominantly localize to the plasma membrane.^[ [103]^15 ^] In
   particular, PI(3,4)P2 is crucial for membrane constriction and the
   formation of endocytic vesicles.^[ [104]^16 ^]

Figure 5.

   Figure 5
   [105]Open in a new tab

   IFITM2 modulates PI(3,4)P2 signaling to regulate endocytosis in RGCs.
   A) Schematic diagram of the relation between the endocytosis process
   and different types of phosphoinositide. B) Representative images of
   Neuro‐2a cells transfected with GFP‐2 × FYVE (green) and stained with
   EEA1 (white) and LAMP1 (red). White arrows indicate the colocalization
   of GFP‐2 × FYVE and EEA1^+ early endosome, and red arrows indicates the
   co‐localization of GFP‐2 × FYVE and LAMP1^+ lysosome. Scale bar, 5 µm.
   C) Representative IUE image of neocortex ventricular zone
   electroporated with the GFP‐2 × PH‐TAPP1 (green) at E13.5, analyzed at
   E15.5. Yellow arrows indicate an enrichment of PI(3,4)P2 signaling at
   the ventricular surface. Scale bar, 10 µm. D) Representative IUE images
   of the E14.5 wild‐type and Ifitm2 cKO neocortex electroporated with
   GFP‐2 × PH‐TAPP1 (green) and HA‐IFITM2 (red) plasmids. Scale bar, 5 µm.
   E) Quantification of the ratio of fluorescence intensity of TAPP1 at
   the ventricular surface to the average fluorescence intensity across
   the entire cell (n = 10 cells per group). Data are shown as mean ±
   s.e.m. Two‐tailed unpaired Students’ t‐test was used. F) Representative
   IUE images of the neocortex electroporated with GFP‐2 × PH‐TAPP1
   (green) and HA‐IFITM2 (red) plasmids. A 10 µm concentration of Src
   inhibitor 1, an ATP‐competitive and selective dual‐site Src tyrosine
   kinase inhibitor, was administered into the lateral ventricle 20 min
   prior to tissue collection. Arrow heads indicate the co‐localization of
   PI(3,4)P2 signaling with HA‐IFITM2 at the apical membrane of RGCs.
   Scale bars, 5 µm. G) Quantification of the ratio of fluorescence
   intensity of TAPP1 at the ventricular surface to the average
   fluorescence intensity across the entire cell after Src inhibitor 1
   treatment (n = 11 cells per group). Data are shown as mean ± s.e.m.
   Two‐tailed unpaired Students’ t‐test was used.

   We first confirmed that PI3P co‐localized with EEA1^+ early endosomes
   and LAMP1^+ endolysosomes in Neuro‐2a cells (Figure [106]5B). Building
   on these observations, we then explored the role of PI(3,4)P2 in the
   developing neocortex. By introducing GFP‐2 × PH‐TAPP1, a specific probe
   for PI(3,4)P2, into RGCs via in utero electroporation, we observed a
   polarized distribution of PI(3,4)P2 signaling at the ventricular
   surface (Figure [107]5C). Next, we investigated the potential role of
   IFITM2 as a scaffold for phosphoinositides and its impact on this
   polarized distribution of PI(3,4)P2 signaling. In Ifitm2 cKO brains, we
   found that the absence of IFITM2 resulted in a significant reduction in
   PI(3,4)P2 signaling on the apical side of RGCs (Figure [108]5D,E). This
   finding suggested that IFITM2 functions as a PI(3,4)P2 scaffold, which
   is crucial for membrane constriction and the formation of endocytic
   vesicles.

   We next explored the relationship between IFITM2 phosphorylation and
   PI(3,4)P2 signaling. Src kinase, a known activator of IFITM2,^[
   [109]^17 ^] was inhibited in the developing cerebral cortex using Src
   inhibitor 1. This treatment resulted in the loss of IFITM2's polarized
   localization and a reduction in PI(3,4)P2 signaling at the ventricular
   surface (Figure [110]5F,G). These findings collectively suggest that
   Src kinase promotes the polarization of IFITM2, which in turn
   stabilizes PI(3,4)P2 signaling, highlighting the complex interaction
   between IFITM2 and phosphoinositide signaling in regulating endocytic
   processes.

2.6. IFITM2 Modulates FAK‐PI3K‐AKT Signaling in Neural Development

   Finally, we explored which signaling pathways are affected at the
   transcriptional level in IFITM2‐mediated endocytosis. To investigate
   this, we performed RNA sequencing (RNA‐seq) to compare differentially
   expressed transcripts between the E13.5 wild‐type and Ifitm2 cKO
   neocortex (Figure [111] 6A). The knockout efficiency of Ifitm2 was
   validated (Figure [112]6B). We found a smaller number of upregulated
   genes, and neither KEGG nor GO analysis revealed specific pathways for
   these genes. In contrast, GO analysis of the downregulated genes showed
   a strong association between IFITM2 and brain development, cell fate
   transitions, and cell membrane polarity (Figure [113]6C). Furthermore,
   KEGG pathway enrichment analysis identified significant changes in the
   focal adhesion pathway (Figure [114]6D). Representative genes related
   to focal adhesion were shown in the volcano plots (Figure [115]6E). The
   focal adhesion pathway is crucial for regulating cell proliferation,
   adhesion, spreading, and survival.^[ [116]^18 ^] In this context,
   phosphorylation of focal adhesion kinase (phospho‐FAK) triggers the
   phosphoinositide 3‐kinase (PI3K)‐AKT signaling cascade, which is
   essential for cell growth and proliferation.^[ [117]^19 ^]
   Additionally, introducing a knockdown vector for FAK, Ptk2‐shRNA, into
   mouse cortices via in utero electroporation promoted neural
   differentiation (Figure [118]S6A–C, Supporting Information).

Figure 6.

   Figure 6
   [119]Open in a new tab

   IFITM2 regulates focal adhesion and PI3K‐AKT signaling in the
   developing neocortex. A) Heat map of the differentially expressed
   transcripts in wild‐type and Ifitm2 cKO E13.5 neocortex. The genes are
   clustered according to the degree of expression similarity on the left.
   The red and blue colors indicate upregulation and downregulation,
   respectively. B) Ifitm2 transcript expression of wild‐type and Ifitm2
   cKO by RNA sequencing. C) Gene ontology (GO) enrichment analysis of the
   down‐regulated genes for biological process (BP). D) Kyoto Encyclopedia
   of Genes and Genomes (KEGG) pathway enrichment analyses of
   down‐regulated genes. E) Volcano plots illustrating the downregulated
   (blue) and up‐regulated (red) genes in the Ifitm2 cKO brain neocortex
   compared with the wild‐type cortex. Representative differentially
   expressed genes associated with the FAK pathway are labeled. F) The
   co‐immunoprecipitation experiments to detect the interaction between
   IFITM2 and PTK2 with Flag‐tag beads in HEK293T cells. G) Representative
   images of the neural progenitor cells electroporated with
   mCherry‐IFITM2 (red) and GFP (green) and stained with p‐FAK (Tyr379)
   (gray). Yellow arrows indicate the colocalization of mCh‐IFITM2 and
   p‐FAK (Tyr379). Scale bar, 5 µm. H) Intensities of p‐FAK (Tyr379)
   (gray) and mCh‐IFITM2 (red) on the arrow. I) Western blot analysis of
   AKT, p‐AKT, GSK3β, and p‐GSK3β levels in E13.5 cortices from wild‐type
   and Ifitm2 cKO mice. J,K) Qualification of the p‐AKT and p‐GSK3β level
   (n = 4 independent experiments). Data are shown as mean ± s.e.m.
   Two‐tailed unpaired Students’ t‐test.

   Then, we performed co‐immunoprecipitation assays to explore the
   interaction between IFITM2 and FAK in 293T cells. Our results confirmed
   that Flag‐tagged FAK robustly pulled down HA‐tagged IFITM2
   (Figure [120]6F). Immunofluorescence staining further demonstrated that
   phospho‐FAK (Tyr379) co‐localized with IFITM2 in primary neural
   progenitor cells (Figure [121]6G,H). To elucidate the role of IFITM2 in
   the FAK‐PI3K‐AKT signaling pathway, we performed immunoblotting
   analysis. The results revealed decreased phosphorylation levels of AKT
   and GSK3β in Ifitm2 cKO neocortex compared to the wild‐type
   (Figure [122]6I–K). It is well‐established that endocytosis faciliates
   the activation of AKT on endosomes, which in turn inhibits GSK3β
   activity.^[ [123]^20 ^] Thus, the reduced endocytic activity leads to
   decreased AKT activation and increased activation of GSK3β in Ifitm2
   cKO neocortex, affecting cellular processes such as proliferation and
   growth. In addition, despite these signaling changes, we observed no
   significant alterations in the actin cytoskeleton within the
   ventricular zone or at the ventricular surface, suggesting that the
   loss of IFITM2 predominantly affects endocytosis and associated
   signaling without markedly influencing actin dynamics (Figure
   [124]S6D–H, Supporting Information).

3. Discussion

   During early embryonic neural development, endocytosis plays a crucial
   role in the internalization of key signaling molecules and receptors
   for neural stem cells. This study provides substantial evidence
   supporting the idea that endocytosis is vital for maintaining stemness
   in neural stem cells. Moreover, we have demonstrated a non‐immune role
   of IFITM2 in regulating endocytosis in the mouse and human developing
   neocortex.

   Endocytosis in the developing cortex regulates receptor
   internalization, degradation, and responses to extracellular
   environments.^[ [125]^21 ^] For example, overexpression of Numb
   promotes trafficking and degradation of Notch 1.^[ [126]^22 ^]
   Furthermore, Numb mutants with defective binding to endocytic proteins
   like α‐adaptin fail to promote Notch 1 degradation. However, we still
   have a limited understanding of the regulatory mechanisms that initiate
   endocytosis in RGCs. Here we offered a mechanistic insight, showing
   that IFITM2 enhances endocytosis through its interaction with PI(3,4)P2
   signaling activated by FAK. Consequently, IFITM2 deficiency inhibited
   endocytosis in RGCs and promoted their differentiation.

   Recent studies have extensively investigated the IFITM family, focusing
   on their roles in immunology and oncology. In terms of immunity, IFITMs
   are known for their antiviral function, which is reported to inhibit
   viral entry by inducing negative membrane curvature.^[ [127]^23 ^] In
   cancer, IFITMs are identified as predictors of poor prognosis in
   several types, including melanoma, glioma, leukemia, lymphoma, and
   pancreatic ductal adenocarcinoma.^[ [128]^24 ^] IFITMs have also
   recently emerged as markers of therapeutically resistant and aggressive
   cancers.^[ [129]^14 , [130]^25 ^] Interestingly, interferons can induce
   the expression of IFITMs during combating viral infections and can also
   activate anti‐cancer immune responses in immunotherapy.^[ [131]^26 ^]
   Considering the role of IFITM2 in regulating endocytosis in the
   developing neocortex, it is possible that IFITMs also affect
   endocytosis in cancer cells, potentially leading to abnormal cell
   proliferation and the loss of cell surface antigens. Thus, in clinical
   settings, combining interferon therapy with inhibition of IFITM
   expression may yield better outcomes.

   Mechanistically, the topology of IFITM proteins suggests that they are
   unlikely to have natural ligands. Three different affinity proteomic
   studies found different sets of IFITM interacting proteins with minimal
   overlap.^[ [132]^27 ^] Consequently, we hypothesized that IFITM2 might
   primarily interact with lipid signaling pathways. This is supported by
   a study in B cell leukemia and lymphoma, which found that IFITM3 could
   act as a PIP3 scaffold.^[ [133]^14 ^] In our research, we discovered
   that IFITM2 binds to PI(3,4)P2 on the mouse embryonic brain ventricular
   surface. This interaction involves the K82 and K87 basic residues of
   IFITM2, which are proximal to two potential palmitoylation modification
   sites and an amphipathic helix. We speculated that the
   phosphatidylinositol signals may regulate the reversible
   post‐translational modifications of IFITM2, subsequently promoting the
   insertion of the amphipathic helix into the lipid bilayer, ultimately
   promoting membrane curvature (Figure [134]S7A, Supporting Information).
   This mechanism might explain that IFITM2 removal led to a TEM‐observed
   increase in type I endocytic vesicles and a decrease in type III
   endocytic vesicles on the ventricular surface.

   Despite the insights from this study, several questions remain
   unanswered. For instance, how the specific extracellular matrix
   proteins and integrins interact to regulate endocytosis on the brain
   ventricle surface is still unclear. Through digging RNA in situ
   hybridization database, we found high expression of ECM proteins, such
   as Vtn, at the ventricular surface, and integrins like Itgav and Itga6
   at the ventricular zone (Figure [135]S7B–D, Supporting Information).^[
   [136]^28 ^] The regulatory mechanisms of these region‐specific ECM and
   integrin proteins warrant further systematic investigation.
   Furthermore, the role of endocytosis in neurogenesis is a vast area of
   study, with previous research on molecules like Numb and Notch also
   implicating endocytosis. More recent studies suggest that the
   subcommissural organ regulates brain development via secreted peptides,
   with gene knockout of these peptides causing significant cortical
   developmental defects.^[ [137]^29 ^] The interactions between RGCs, the
   cerebrospinal fluid, and the extracellular matrix are key areas
   requiring more in‐depth exploration. Additionally, while our study
   focused on endocytosis at the ventricular surface, the potential role
   of IFITM2 in intracellular vesicle trafficking remains unexplored.
   Given that the anti‐viral strategies of IFITMs are involved in
   inhibiting endosome‐lysosome fusion,^[ [138]^13a ^] the potential role
   of IFITM2 in vesicular transport cannot be ruled out. Lastly, the
   molecular mechanisms underlying IFITM2's functions remain unclear;
   however, new techniques such as DNA PAINT,^[ [139]^30 ^] cryo‐electron
   tomography (ET),^[ [140]^31 ^] and cryo‐correlative electron and light
   microscopy (CLEM)^[ [141]^32 ^] may provide high‐resolution in situ
   structural insights, further advancing our understanding of the
   structural biology of IFITMs in membrane dynamics.

4. Experimental Section

Mouse

   The Ifitm2 conditional knockout mice were generated by using the
   CRISPR‐CAS9 mediated double nicking strategy. Guide RNA sequence: gRNA1
   (5′‐ATGAAACTGCCTGGGGATTCTGG‐3′); gRNA2 (5′‐
   AGGCTTCTGGTCTCCAGATATGG‐3′). Nestin^cre mice were used to recognize and
   cut the flox site. All animal experiments were approved by the Animal
   Care and Use Committees of the Institute of Zoology, Chinese Academy of
   Sciences. The ethical approval number for animal studies is
   IOZ20190080.

Plasmids

   Mouse IFITM2 cDNA (accession number [142]NM_030694) was obtained by
   polymerase chain reaction (PCR) and cloned into pCAG‐HA to generate
   pCAG‐HA‐IFITM2. Mouse LAMP2 cDNA (accession number [143]NM_001017959)
   and RAB7A cDNA (accession number [144]AB232599) were obtained by
   polymerase chain reaction (PCR) and cloned into pCAG‐N1‐mCherry/GFP to
   generate pCAG‐LAMP2‐mCherry. Y19A plasmid was obtained by homologous
   recombinant HA‐IFITM2. Mouse GRP1 cDNA (accession number
   [145]NM_001163548) was obtained by polymerase chain reaction (PCR) and
   cloned into pCAG‐GFP to generate pCAG‐GFP‐2 × FYVE and pCAG‐GFP‐2 × PH‐
   TAPP1 were cloned from pEGFP‐2xFYVE (MIAOLING Plasmid, #[146]P31053)
   and GFP‐2 × PH‐TAPP1 (MIAOLING Plasmid, #[147]P36199). The shRNA
   fragments were synthesized and constructed into pSicoR. The sequences
   of shRNAs targeting PTK2 are as follow: Ptk2‐sh1,
   5′‐GCCTTAACAATGCGTCAGTTT‐3′, and Ptk2‐sh2, 5′‐CCTGGCATCTTTGATATTATA‐3′.

Immunostaining

   Brains were fixed with 4% PFA (4% paraformaldehyde dissolved in PBS)
   overnight and then transferred into 30% sucrose to dehydrate. Brains
   were sectioned at 15 µm for immunostaining. For immunostaining, use 4%
   PFA for 20 min at room temperature, and wash 3 times with PBST (10%
   Triton X‐100) for 10 min. Decant 5% BSA in PBST for 2 h at room
   temperature, then change with the primary antibodies overnight. Wash
   the slices 3 times for 10 min and incubate them with secondary
   antibodies for 2 h at room temperature in the dark. In the end, wash
   the slices in PBS and seal them with 50% glycerin. For some
   immunostaining, nuclei were labeled by DAPI (Sigma, Cat# D9542) for
   2 min before the final seal. For EdU staining, the BeyoClick EdU Cell
   Proliferation Kit with Alexa Fluor 488 (Beyotime Cat#C0071S) was used
   following the manufacturer's instructions.

Human Cerebral Organoid Generation

   The STEM diff cerebral organoid kits (STEM, Cat#08570, Cat#08571) were
   used to generate cerebral organoids. All steps were followed by the
   protocol of the products. There are four steps: embryoid body
   formation, induction, Expansion, and organoid maturation. The H9 ES
   cell line was used to form EBs. In the first three steps, the EBs were
   cultured in the 96‐well or 24‐well low attachment plate. On day 12, the
   tissues were transferred into a 3D spinning bioreactor (Pfeiffer
   spinner, Cat# 183001).

FlashTag Labeling

   FlashTag labeling was performed as previously described.^[ [148]^2 ^]
   Briefly, timed pregnant mice were anaesthetized with
   2,2,2‐tribromoethanol and 2‐Methyl‐2‐butanol, and the uterine horns
   were exposed for the experiment. A 2 mm solution of carboxyfluorescein
   succinimidylester (CellTrace, Thermo Fisher, Cat# [149]C34570) was
   injected into the lateral ventricle of the embryonic brain to label the
   M‐phase apical juxtaventricular radial glial cells. The CellTrace
   compounds function by covalently binding the intracellular proteins
   with the fluorescent dye carboxyfluorescein. The term FT refers to both
   these compounds and the pulse‐labeling procedure.

Antibodies

   Rabbit anti‐SOX2 (Cell Signaling Technology Cat# 3728S; RRID:
   AB_2194037); Rabbit anti‐IFITM2 (Proteintech Cat# 12769‐1‐AP; RRID:
   AB_2122089); Mouse anti‐IFITM2 (Santa Cruz Biotechnology Cat#
   sc‐373676; RRID: AB_10915739); Rabbit anti‐TBR1 (Abcam Cat# ab31940;
   RRID: AB_2200219); Rat anti‐CTIP2 (Abcam Cat# ab18465; RRID:
   AB_2064130); Rabbit anti‐CUX1 (Proteintech Cat# 11733‐1‐AP; RRID: AB_
   2086995); Mouse anti‐TUJ1 (Millipore Cat# MAB1637; RRID: AB_ 2210524);
   Rabbit anti‐PAX6 (Millipore Cat# AB2237; RRID: AB_ 1587367); Rat
   anti‐TBR2/EOMES (eBioscience Cat# 14‐4875‐82; RRID: AB_11042577);
   Rabbit anti‐Ki67 (Abcam Cat# ab15580; RRID: AB_443209); Rabbit
   anti‐Rab7 (Cell Signaling Technology Cat# 9367; RRID: AB_1904103); Rat
   anti‐LAMP1 (BD Biosciences Cat# 553792; RRID: AB_2134499); Rabbit
   anti‐EEA1 (Cell Signaling Technology Cat# 3288T; RRID: AB_2096811);
   Rabbit anti‐HA (Cell Signaling Technology Cat# 3724S; RRID:
   AB_1549585); Rabbit anti‐FAK (ABclonal Cat# A11195; RRID: 2758455);
   Rabbit anti‐p‐FAK(Y397) (ABclonal Cat# AP0302; RRID: 2771470); Rabbit
   anti‐p‐GSK‐3β (S9) (Cell Signaling Technology Cat# 5558T; RRID:
   AB_10013750); Rabbit anti‐AKT (Cell Signaling Technology Cat# 4865S;
   RRID: 2225340); Rabbit anti‐p‐AKT(Ser473) (Cell Signaling Technology
   Cat# 3787S; RRID: 331170); Rabbit anti‐β‐actin (Proteintech
   Cat#20536‐1‐AP; RRID: AB_10700003); PE/Cyanine7 anti‐mouse CD133
   Antibody (BioLegend Cat# 141209; RRID: AB_2564068); FITC anti‐mouse
   CD24 Antibody (BioLegend Cat# 101806; RRID: AB_312839); 594‐Phalloidin
   (Proteintech Cat#PF00003; RRID: AB_3675292); Wheat Germ Agglutinin
   (Thermo fisher scientific Cat# [150]W11261).

Transmission Electron Microscopy (TEM)

   TEM was performed as previously described.^[ [151]^33 ^] The E15.5
   wild‐type and Ifitm2 cKO cortices were cut into sections and fixed with
   2% PFA and 2.5% glutaraldehyde overnight at 4 °C. After washing with
   PBS three times, the sections were post‐fixed in 1% osmium for 1 h.
   Then washed and dehydrated by an ethanol series. Use 100% acetone to
   wash sections, followed by soaking in acetone and epoxy resin. The
   sections were embedded with epoxy resin for 48 h. Serial sections of
   60 nm were cut by Leica UC7 ultramicrotome. Images of the brain
   ventricular surface were acquired with a transmission electron
   microscope (Tecnai G2 F20 TWIN TMP).

Cells

   The Neuro‐2a cells and HEK 293T cells were cultured in high glucose
   DMEM (Gibco) with 1% penicillin/ streptomycin (Invitrogen) and 10%
   fetal bovine serum (Gibco). The H9 cells (hECs) were cultured in the
   Essential 8 medium kit (Thermo). The primary neural progenitor cells
   were isolated from the E12.5 cortices.

Flow Cytometry

   Neural stem cells were isolated from embryonic mouse cortices in a
   sterile bench. The tissue was first dissociated using a papain
   digestion solution at 37 °C for 5 min. After digestion, the suspension
   was filtered through a 70 µm cell strainer to remove any remaining
   undigested tissue. The cell suspension was centrifuged at 300 g for
   5 min, and the cell pellet was resuspended in 1 mL of DPBS to generate
   a single‐cell suspension. For cell surface marker staining, the cells
   were incubated with PE/Cyanine7 anti‐mouse CD133 antibody (Biolegend,
   Catalog #141209) and FITC anti‐mouse CD24 antibody (Biolegend, Catalog
   #101806) on ice for 30 min, according to the recommended
   concentrations. After incubation, the cells were washed twice with DPBS
   and resuspended in 1 mL of DPBS. The cell population was then sorted
   using a BD FACSAria Fusion Flow Cytometer, and the data were analyzed
   using FlowJo software.

Intracerebroventricular Injection and In‐Utero Electroporation (IUE)

   The target plasmids or chemical inhibitors were mixed in proportion to
   the concentration and inhaled into beveled glass micropipettes for use.
   Sterilize bench and instrument. Timed pregnant ICR mice were
   anesthetized with 2,2,2‐tribromoethanol (Sigma, Cat# [152]T48402). Cut
   the medioventral line, expose the uterine, find the fetal cerebral
   cortex, and inject the mixture into the lateral ventricle. For
   electroporation, the positive electrode was attached to the side of the
   target cortex. Five 50 ms pulses of 35–40 V with a 950 ms inter‐pulse
   interval were delivered with the tweezer electrodes connected to the
   electroporator (BTX ECM830).

Single‐Cell RNA Sequencing Analysis

   Single‐cell sequencing data were analyzed from the E13 developmental
   stage using the publicly available database (GEO SuperSeries
   [153]GSE153164) and processed with the Seurat R package.^[ [154]^7 ^]
   To begin with, DoubletFinder was employed to remove doublet cells.
   Quality control parameters were set to exclude cells with less than 10%
   mitochondrial reads and more than 500 detected genes. This step ensured
   the retention of high‐quality cells for further analysis. The
   “FindVariableFeatures” function was used to select variable genes,
   followed by Principal Component Analysis (PCA) for data reduction.
   Based on the PCA results, cell clustering was achieved using the
   “FindNeighbors” and “FindClusters” functions. For Gene Pathway Scoring,
   major cell types were identified by their characteristic gene
   expression profiles. To calculate pathway scores for specific gene
   sets, Seurat's `AddModuleScore` function was used for each cell and was
   computed based on the gene list for selected terms, and the average
   value was taken as the final score for each cell type.

Co‐Immunoprecipitation (Co‐IP)

   Mouse IFITM2 and PTK2 cDNA were obtained by polymerase chain reaction
   (PCR). Ifitm2 cloned into pCAG‐3HA to generate pCAG‐HA‐ Ifitm2 and Ptk2
   cDNA cloned into pCAG‐3FLAG to generate pCAG‐FLAG‐ Ptk2. The target
   plasmids were co‐transfected into 293T cells, and the plasmid with the
   HA label was transferred as a negative control. After two days of cell
   culture, the protein was harvested. Resuspend the cells in 200 µL of IP
   lysis buffer. Sonicate cells 4–5 times for 5 s and place them directly
   on ice. Take out 20 µL cell lysate as an Input group and boil them.
   FLAG beads were added to the cell lysate and incubated at 4 °C
   overnight. The next day, the supernatant was discarded, and keep the
   beads for immunoprecipitation. Wash the beads with PBS and resuspend
   the beads in 200 µL RIPA buffer. In the last, boil them for 10 min in
   boiled water. The samples were stored in −80 °C refrigerator.

Western Blot

   Tissues and cells were lysed in RIPA lysis buffer with 1% Cocktail and
   1% PMSF. Sonicate the lysis 4–5 times for 5 s, centrifuge at 8000 rpm
   for 10 min, and aspirate the supernatant. The lysate was added loading
   buffer and boiled for 10 min. The samples were loaded on 10% or 12%
   SDS‐PAGE. After transferring onto PVDF membranes and blocking with 5%
   milk (dissolved in PBST‐tween20), the membrane blots were incubated
   with primary antibodies at 4 °C overnight. The next day, the membranes
   were washed 3 times by PBST‐tween20 and incubated with the secondary
   antibodies at room temperature for 1 h. Finally, the membranes were
   washed 3 times with PBST‐tween 20. Develop and analyze by the imaging
   system (Odyssey, LI‐COR).

RNA Sequencing

   The samples were isolated from E12.5 wild‐type and Ifitm2 cKO. Use the
   TRIzol reagent to extract total RNAs. The quality of RNA‐seq libraries
   was evaluated by Agilent 2100 Bioanalyzer and sequenced on an Illumina
   platform by Annoroad Gene Tech. (Beijing, China) Co. Ltd. Data
   deposited at GEO accession: [155]GSE247448. GO analysis was performed
   using Metascape ([156]http://metascape.org).

Quantitative Real‐Time PCR

   RNA was extracted using a TRIzol reagent. The purified RNA was
   reverse‐transcribed into cDNA using the First‐Strand cDNA Synthesis Kit
   (Tiangen, Cat# KR106‐02). Quantitative RT‐PCR was performed with the
   SuperReal PreMix Plus (SYBR Green) Kit (Tiangen, Cat# FP205‐03) on an
   RT‐PCR Detection System (LI‐COR Biosciences). GAPDH served as the
   internal control for normalizing mRNA levels. The primers are shown in
   Table [157]S1 (Supporting Information).

In Vitro Neurosphere Culture

   Neural progenitor cells were isolated from the cortex of wild‐type and
   Ifitm2 cKO of the E12.5 period. Cells were cultured in the medium
   containing DMEM/F12 (Gibco, Cat# 11330‐032), GlutaMax supplement
   (Gibco, Cat# 35050061), B‐27 supplement (1x, Thermo Fisher, Cat#
   12587010), bFGF (5ng mL^−1, Thermo Fisher, Cat# 13256029), EGF
   (5ng mL^−1, Thermo Fisher, Cat# PHG0311) for 7 days, and changed the
   medium every 2 days.

Quantification and Statistical Analysis

   The fluorescence intensity of confocal microscopy images was assessed
   using either a Zeiss LSM 880 with Airyscan or a Leica Stellaris
   microsystem. To ensure comparability across groups within experiments,
   consistent settings were maintained, including laser power, pinhole
   size, and gain. Images were processed and analyzed using ImageJ
   software (NIH, USA). Regions of interest (ROIs) were delineated on each
   image to measure fluorescence intensity specifically. The “Measure”
   function in ImageJ quantified the mean fluorescence intensity within
   these ROIs, with background fluorescence subtracted from areas lacking
   specific staining to yield corrected fluorescence values. Fluorescence
   data were normalized by setting the mean intensity of the wild‐type
   group as 100%, with KO group intensities expressed relative to this
   benchmark. Statistical significance between groups was determined using
   Student's t‐test. For the quantification of fluorescence intensity
   distribution along the Y‐axis, rectangles of 200 µm height by 50 µm
   width were marked on selected images. The “Plot Profile” function from
   the “Analyze” menu generated plots of fluorescence intensity along the
   Y‐axis, which were used to evaluate distribution characteristics. These
   data were subsequently exported for further analysis. ImageJ 2.0.0
   software facilitated the analysis of immunofluorescence
   co‐localization, area statistics, and intensity measurements, while
   Prism 9.4.1 software was employed for statistical analysis. Parameters
   and repetition counts were detailed in the figure legends. A p‐value of
   <0.05 was deemed statistically significant in all analyses.

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   Y.L., W.Z., and L.L. These authors contributed equally to this work
   Y.L. and W.Z. conceived the project, J.J. and J.P. supervised the
   project and acquired the funding support, Y.L., W.Z., L.L., and J.L.
   performed the experiments, W.Z. and S.Z. analyzed the high‐throughput
   sequencing data. Y.L. and W.Z. wrote this manuscript with input from
   all authors.

Supporting information

   Supporting Information
   [158]ADVS-12-2501593-s001.pdf^ (7.5MB, pdf)

Acknowledgements