Wnt signaling plays a critical role in development across species and is dysregulated in a host of human diseases. A key step in signal transduction is the formation of Wnt receptor signalosomes, during which a large number of components translocate to the membrane, cluster together and amplify downstream signaling. However, the molecular processes that coordinate these events remain poorly defined. Here, we show that Daam2 regulates canonical Wnt signaling via the PIP2–PIP5K axis through its association with Rac1. Clustering of Daam2-mediated Wnt receptor complexes requires both Rac1 and PIP5K, and PIP5K promotes membrane localization of these complexes in a Rac1-dependent manner. Importantly, the localization of Daam2 complexes and Daam2-mediated canonical Wnt signaling is dependent upon actin polymerization. These studies – in chick spinal cord and human and monkey cell lines – highlight novel roles for Rac1 and the actin cytoskeleton in the regulation of canonical Wnt signaling and define Daam2 as a key scaffolding hub that coordinates membrane translocation and signalosome clustering.
Wnt signaling is a highly conserved signaling pathway required for cell fate determination during development and homeostasis in adult tissues (Logan and Nusse, 2004; Nusse and Clevers, 2017), and emerging evidence has implicated Wnt dysregulation in both genetic and non-genetic neurological disorders (Fancy et al., 2014). Wnt signaling effectors have relatively low affinities for each other, preventing unregulated interactions at their physiological levels and allowing fine-tuning of proper signaling activation at the appropriate time (Kishida et al., 2001; Niehrs, 2012; Schwarz-Romond et al., 2007). Therefore, a key regulatory step for Wnt signaling activation is the strengthening of modest interactions between Wnt effectors when the Wnt ligand binds the receptor. As a result, transmembrane receptors and signal transducers assemble to form the Wnt signalosome (Gammons et al., 2016; MacDonald et al., 2009). Dynamic polymerization of the Dishevelled proteins functions at the core of the Wnt signalosome by interacting with both the Frizzled Wnt receptors and low-density lipoprotein receptor-related protein 5/6 (LRP5/6), leading to recruitment of Axin proteins from the β-catenin destruction complex (MacDonald et al., 2009; Schwarz-Romond et al., 2007). However, the exact composition and mechanisms of signalosome assembly at the plasma membrane remain unclear.
The hallmark of canonical Wnt signaling is the accumulation and translocation of β-catenin into the nucleus for gene transcription, whereas non-canonical Wnt signaling is β-catenin independent and involves assembly/disassembly of the actin cytoskeleton, polarized cell shape changes and cell migration (Niehrs, 2012; Schlessinger et al., 2009). Rho family GTPase proteins play important roles in actin/microtubule organization such as cell adhesion and migration in all eukaryotic cells (Hodge and Ridley, 2016). Although Rho GTPases are known to be mainly involved in the non-canonical Wnt/planar cell polarity signaling pathways (Niehrs, 2012; Schlessinger et al., 2009), evidence indicates a potential link between Rho GTPases and canonical Wnt signaling, with Rac1 emerging as a possible major player (Schlessinger et al., 2009). Several studies in different systems have reported that Rac1 promotes nuclear import of β-catenin (Wu et al., 2008), enhances β-catenin–LEF-1 complex assembly (Jamieson et al., 2015) and interacts directly with Dishevelled proteins (Dvls) (Cajanek et al., 2013; Soh and Trejo, 2011), resulting in the activation of canonical Wnt signaling. Moreover, emerging studies reported the association between Rac1 and phosphatidylinositol 4-phosphate-5 kinase (PIP5K) proteins, which mediate phosphatidylinositol 4,5-bisphosphate (PIP2) production for signalosome formation (Pan et al., 2008; van den Bout and Divecha, 2009; Weernink et al., 2004). This finding raises an important question regarding a direct functional relationship between Rho GTPases and proximal events of canonical Wnt receptor complex signaling, which remains to be addressed.
Previously, we found that dishevelled associated activator of morphogenesis 2 (Daam2) is required for the clustering of Wnt receptor complexes and canonical Wnt signaling through its association with PIP5K and, consequently, PIP2 production (Lee and Deneen, 2012; Lee et al., 2015). The presence of a GTPase-binding domain in Daam2 and previous studies implicating Rho GTPase Rac1 in PIP5K activation suggested a possible role for Rac1 in Daam2–PIP5K-mediated canonical Wnt signaling. Here, we show that the association of Daam2, Rac1 and PIP5K is critical for Wnt receptor complex clustering, membrane translocation and subsequent activation of canonical Wnt signaling during cell fate determination. Moreover, pharmacological manipulation of actin dynamics regulates Daam2-mediated Wnt signal transduction during central nervous system development. Taken together, our study provides in vivo and biochemical evidence that Rac1 is required for canonical Wnt signaling, and that its function in actin polymerization is crucial for Daam2–PIP5K-mediated complex translocation at the membrane.
Daam2 regulates Wnt signaling through Rac1 and PIP5K
To further dissect the direct interaction of Daam2 and PIP5K (Lee et al., 2015), we performed co-immunoprecipitation (co-IP) assays exploring the nature of their association. Our results revealed that PIP5K associates with the GTPase-binding domain (GBD) of Daam2, an essential domain for canonical Wnt signaling (Fig. 1A; Fig. S1A,B). These data, together with our previous studies (Lee et al., 2015), suggest that the Daam2 GBD functions through PIP5K to regulate canonical Wnt signaling. The GBDs of Daam family proteins have been shown to interact with Rho GTPases (Habas et al., 2001), and Rac1 binds to activate PIP5K (Weernink et al., 2004; Wei et al., 2002), suggesting that Rac1 or other Rho GTPases may contribute to Daam2 function in this context. To examine the link between Daam2 and Rho GTPases, we performed co-IP assays in human embryonic kidney (HEK)293T cells and found, through western blot analysis, that Daam2 associates with Rac1 and Cdc42 and that these associations require the GBD domain of Daam2 (Fig. 1A; Fig. S1C,D).
Next, we examined the ability of Rac1 or Cdc42 to rescue dorsal patterning and canonical Wnt signaling in the spinal cord in the absence of Daam2. We performed functional epistatic experiments utilizing a Daam2-targeted short hairpin RNA interference (shRNAi) (or mutant shRNAi control) system we previously employed (Lee and Deneen, 2012), together with expression of Rho GTPases and canonical Wnt reporter TOP–nRFP in the chick spinal cord at embryonic day 4 (E4). Effective knockdown of endogenous Daam2 was verified by in situ hybridization (Fig. 1F,J,N), and overexpression of GFP-tagged Rho GTPases was verified by GFP fluorescence (Fig. 1K,O). We found that Rac1 overexpression partially restores canonical Wnt signaling (TOP–nRFP) and dorsal progenitor marker expression (Pax7) in the absence of Daam2 during chick spinal cord development (Fig. 1J–M,R–S), while leaving ventral marker expression (Nkx2.2) unaffected. However, overexpression of another Rho-GTPase, Cdc42, did not rescue canonical Wnt activity or dorsal marker expression in the absence of Daam2 (Fig. 1N–S), indicating that this function is specific to Rac1.
Previously, we showed that Daam2 is required for canonical Wnt signaling through PIP2 production in developing spinal cord (Lee et al., 2015). To examine whether Rac1 regulates Daam2 to induce PIP2 production, we performed PIP2 enzyme-linked immunosorbent assays (ELISAs) with these chick spinal cords. Indeed, we found that Rac1 restored PIP2 production in the absence of Daam2 (Fig. S1E). Collectively, these observations indicate that Rac1 overexpression rescues Daam2 function in canonical Wnt signaling and dorsal patterning during spinal cord development.
Rac1 is required for dorsal patterning and Wnt signaling
Having established that Rac1 positively regulates Daam2 function in developing spinal cord, we next examined the role of Rac1 during early spinal cord development. We first confirmed that Rac1 is expressed in ventricular zone populations during development by performing in situ hybridization on embryonic chick spinal cord from E2 to E4 (Fig. S1F). To determine the effect of loss of Rac1 on spinal cord development, we performed knockdown experiments by Rac1 shRNAi and assessed canonical Wnt activity as well as expression of progenitor markers along the dorsal ventral axis at E4. The effective knockdown of endogenous Rac1 was verified by in situ hybridization (Fig. 2E). We found reduced canonical Wnt activity and a specific loss of dorsal progenitor marker expression (Pax7), but not ventral marker (Nkx2.2) expression (Fig. 2E–H,M,N).
To assess the specificity of the Rac1 knockdown phenotype, we generated a mutant version of the Rac1 shRNAi containing five nucleotide substitutions and it showed no effect on expression of endogenous Rac1 or dorsal marker expression (Fig. 2A–D,M,N). We next confirmed that the effects of the Rac1 shRNAi are the result of the loss of Rac1 by performing rescue experiments through co-expression with tagged human versions of Rac1. We found that expression of human Rac1 is sufficient to restore the expression of dorsal markers and TOP–nRFP reporter in the absence of endogenous chick Rac1 (Fig. 2I–N). Together, these experiments reveal that knockdown of Rac1 is responsible for the loss of dorsal progenitors and Wnt activity, indicating a causal effect of Rac1 ablation that mirrors Daam2 knockdown.
Rac1 and PIP5K regulate assembly and membrane localization of Daam2–Dvl3 complexes
Because Daam2, Rac1 and PIP5K interact to mediate Wnt signaling, we next investigated whether their interactions regulate the assembly and clustering of Wnt receptor complexes. Previously, we found that Daam2 promotes the assembly of Dvl3 complexes and that this clustering activity is tightly correlated with its role in Wnt signaling and dorsal patterning (Lee and Deneen, 2012). To define whether Rac1 is required for Daam2-mediated Wnt complex clustering, we performed Daam2/Dvl3 overexpression in conjunction with pharmacological inhibition or small interfering RNA (siRNA) knockdown of Rac1 in Cos7 cells. We then measured intracellular protein assemblies, which have previously been correlated with stabilized Daam2–Dvl3–Axin2 complexes (Lee et al., 2015). Intriguingly, we found that Daam2/Dvl3 form protein aggregates; however, inhibition of Rac1 resulted in the formation of discrete punctate structures instead of aggregates, suggesting that the loss of Rac1 blocks clustering of Daam2–Dvl3 complexes (Fig. 3A–I; Fig. S2A–G).
Our previous studies revealed that PIP5K is necessary for Daam2-mediated Wnt receptor clustering (Lee et al., 2015). Next, we tested whether PIP5K can stabilize Daam2/Dvl3 binding in Cos7 cells in a Rac1-dependent manner. When co-expressing PIP5K with Daam2 and Dvl3 in Cos7 cells, we observed significantly increased membrane localization of Daam2–Dvl3–PIP5K aggregates compared to Daam2/Dvl3 alone (Fig. 3J–N; Fig. S2I–K). Moreover, these phenotypes were abolished by RNA interference (RNAi)-mediated loss of Rac1 (Fig. 3O-R,AA; Fig. S2H,L–N). These observations reveal that PIP5K is required for Daam2-mediated Wnt receptor complex membrane translocation to the membrane through a process requiring Rac1. To examine whether Daam2–Dvl3–PIP5K complex membrane localization influences Wnt signalosome formation, we then assessed LRP6 localization and phosphorylation (Bilić et al., 2007; Pan et al., 2008). We observed an increase in the amount of gross phosphorylated LRP6 (pLRP6) by immunoblotting (Fig. 3CC, lane 6) and in pLRP6 signal through immunostaining (Fig. 3N; Fig. S2U–X).
We then examined whether Rac1 is required for signalosome assembly in the context of canonical Wnt stimulation. Treatment with the Wnt ligand Wnt3a led to modest increases in membrane colocalization of signalosome components and levels of pLRP6 (Fig. 3S–V,AA; Fig. S2O–Q), which may be due to already high levels of Wnt ligands in the co-expression system. On the other hand, pharmacological inhibition of Rac1 in conjunction with Daam2/Dvl3/PIP5K overexpression continued to abrogate membrane translocation of signalosome components and LRP6 phosphorylation even in the presence of the added Wnt3a (Fig. 3W–BB; Fig. S2H,R–T,GG–JJ), suggesting that Rac1 is necessary for Wnt-mediated assembly of signalosome receptor complexes. We independently validated these results in HeLa cells (Fig. S3). Taken together, these data indicate that the association between Daam2/PIP5K/Rac1 is essential for the Wnt signalosome by regulating the clustering and membrane localization of key components of the Wnt receptor complex.
Actin polymerization is required for Daam2-mediated canonical Wnt signaling and patterning
Protein transport to the cell membrane requires actin cytoskeleton assembly, which has not previously been implicated in the construction or regulation of signalosomes associated with canonical Wnt signaling. Given that Rac1 regulates cytoskeletal remodeling (Guo et al., 2006; MacHacek et al., 2009) and is required for membrane localization of Daam2–Dvl3–PIP5K complexes (Fig. 3), we examined whether manipulation of the cytoskeleton influences the membrane localization of these complexes in Cos7 cells through pharmacological means. Treatment with the actin polymerization inhibitors Cytochalasin D (CD) and Latrunculin A (LatA) blocked PIP5K-induced Dvl3–Daam2 complex membrane localization and clustering of these complexes, whereas the microtubule inhibitor Nocodazole (Noco) showed no effect (Fig. 4A–D). These data specifically implicate actin polymerization in the clustering and localization of Daam2–Dvl3–PIP5K complexes. Subsequent analysis of actin dynamics in cells revealed that membrane localization of Daam2/Dvl3 by PIP5K correlates with the presence of F-actin in the membrane (Fig. S4), further supporting a model in which membrane targeting of Daam2–Dvl3–PIP5K complexes is regulated through actin remodeling.
The foregoing data suggest that actin dynamics regulate signalosome formation and canonical Wnt signaling. To examine these results in vivo, we used organotypic chick spinal cord explants, which demonstrated loss of dorsal patterning and Wnt activity upon Rac1 inhibition (Fig. S5). We treated chick explants (Fig. 4E) with actin polymerization inhibitors (CD and LatA) and found a specific loss of dorsal markers compared to dimethyl sulfoxide (DMSO) control treatments and the microtubule inhibitor Noco (Fig. 4F–J). To measure canonical Wnt activity, chick spinal cords were co-electroporated with TOP–nRFP and PIP2–GFP before explant culture and inhibitor treatment. We found a similar loss of Wnt activity as indicated by reduced TOP–nRFP signal (Fig. 4K–N,S), which correlated with diminished PIP2–GFP levels (Fig. 4O–S), implicating actin polymerization in the regulation of PIP2-mediated Wnt signaling.
Actin stabilization rescues PIP2-mediated canonical Wnt activity in the absence of Daam2
We next examined whether promoting actin polymerization can compensate for the loss of Daam2 during canonical Wnt signaling in chick explants. Chick spinal cord was co-electroporated with Daam2 shRNAi, TOP–nRFP and PIP2–GFP, and cultured as an explant with treatment of Jasplakinolide (JPK), a compound that promotes actin polymerization (Bubb et al., 2000; Holzinger, 2009). JPK-treated explants were able to restore Wnt activity and PIP2 production despite the loss of Daam2 (Fig. 5A–C,E–H,J), whereas the administration of Taxol, a microtubule-stabilizing compound, failed to rescue either of these in the absence of Daam2 (Fig. 5D,E,I,J). This indicates that Daam2 function in canonical Wnt signaling is mediated, in part, through stabilization of actin filaments and provides evidence that actin polymerization is necessary for Wnt signaling by facilitating signalosome formation at the membrane. Taken together, our data suggest that Rac1 plays a crucial role in Wnt signaling and dorsal patterning in the early spinal cord via its actin nucleation activity and interaction with Daam2 and PIP5K (Fig. 6).
A key unresolved question in developmental biology is a mechanistic understanding of membrane localization of Wnt receptor complex components and subsequent signalosome formation. Previous work has demonstrated that subcellular trafficking of cell polarity proteins, receptors and downstream activators regulates morphogen response (Hudish et al., 2013, 2016; Kong et al., 2015); however, many of the mechanisms remain to be defined for Wnt signaling. In this study, we found that Daam2 functions as a scaffolding hub for canonical Wnt signaling, coupling the mechanisms that control membrane translocation with those regulating the clustering of Wnt receptor signalosomes (Fig. 6). Our data suggest that this effect is mediated through its GBD, which interacts with both Rac1 and PIP5K. However, this does not rule out the possibility of Daam2 acting through parallel pathways. Daam2 is a Formin family protein, and as yet no studies have demonstrated a direct action of its formin domains in actin cytoskeletal remodeling. However, it is important to note that Daam2 alone was insufficient to cause translocation of Daam2–Dvl3 complexes to the membrane (Fig. 3A–C,J–M). This suggests that its function in signalosome assembly is governed by its association with both PIP5K and Rac1, where tripartite interactions between these molecules are required to coordinate both translocation and signalosome activities. Wnt signalosome formation by this complex is mediated through the PIP2–LRP6 axis, illustrating a novel role for Daam2 in the regulation of phosphatidylinositol lipids and providing insight into the regulation of PIP2 in canonical Wnt signaling.
PIP2 is the most abundant of the seven known phosphoinositides and functions in a number of crucial processes in the cell (Czech, 2000; Lemmon, 2008; McLaughlin et al., 2002), participating in multiple pathways as second messenger precursors, protein regulators and lipid anchors (Gamper and Shapiro, 2007). Importantly, PIP2 is known to be a major regulator of the actin cytoskeleton, recruiting actin regulators to the membrane, including the ARP2/3 complex, gelsolin and CapZ proteins (Rohatgi et al., 2000; Wu et al., 2014; Yin and Janmey, 2003). The recruitment of PIP2-generating enzymes such as PIP5K to the membrane leads to local sites of PIP2 enrichment, regulating these functions (Choi et al., 2015; El Sayegh et al., 2007; Qin et al., 2009), and previous evidence showed that Wnt signaling can induce the production of PIP2 through the Daam2–PIP5K pathway (Lee et al., 2015). Our study further sheds light on a potential mechanism by which PIP2 membrane signaling is maintained: Wnt signalosome formation can induce local increases in PIP2 production, which in turn promotes actin polymerization. The polymerization of actin may then further amplify Wnt signaling by facilitating signalosome formation.
Moreover, these studies identified a new role for Rac1 in Wnt signalosome formation and canonical Wnt signaling during cell fate determination in developing chick spinal cord. Previously, Rac1 has been linked to cell motility and non-canonical Wnt signaling by regulating actin dynamics (Rosso et al., 2005; Schlessinger et al., 2009). On the other hand, Rac1 has been suggested to play a role in canonical Wnt signaling by modulating nuclear localization of β-catenin (Phelps et al., 2009; Wu et al., 2008) or stimulating downstream enhancer complex assembly in the nucleus (Jamieson et al., 2015). However, whether Rac1 participates in proximal events involving Wnt signalosome formation has been less clear. PIP5K membrane recruitment and subsequent synthesis of PIP2 are known to be crucial steps in signalosome formation and canonical Wnt signaling (Lee et al., 2015; Pan et al., 2008). In this study, we identified novel and important roles for Rac1 and actin polymerization as necessary steps in Wnt signalosome formation during early neural development in vivo. We showed that Rac1 is critical for PIP5K localization and signalosome formation at the membrane, and that Daam2 mediates this function by serving as a convergence point for Rac1 and PIP5K.
Here, we found, for the first time, that translocation of Wnt receptor complex components and subsequent canonical Wnt signaling are actin dependent. Taken together, these findings suggest distinct roles for actin in discrete Wnt pathways, where it regulates cell motility and polarity in non-canonical Wnt signaling while playing a role in membrane targeting in canonical Wnt pathways. Deciphering these modes of actin regulation adds another layer of regulatory complexity: protein translocation relies on motor proteins, whereas cell motility entails protrusive forces at the membrane. Therefore, distinguishing how these modes of actin regulation are linked to Wnt pathway components may provide new insight into the complex interplay between canonical and non-canonical Wnt signaling events.
MATERIALS AND METHODS
The following complementary DNAs (cDNAs) were used in chick electroporation experiments: full-length mouse Daam2 (a gift from Dr Terry P. Yamaguchi, National Cancer Institute, Bethesda, MD, USA), TOP–nRFP reporter (a gift from Dr Andy Groves, Baylor College of Medicine, Houston, TX, USA) and the PIP2 reporter PLCδ–PH–GFP (Balla and Várnai, 2002). To generate a specific riboprobe for chick Rac1 mRNA, we cloned cDNA fragments of chick Rac1 using RT-PCR from E4 chick spinal cord mRNA. We used the RCAS system, a replication-competent retroviral vector system for avian expression, for shRNAi experiments. We generated the RCAS shRNAi against chick Rac1 using the methods described in Lee and Deneen (2012). The target chick Daam2 sequence used to generate the shRNAi construct is located within the coding sequence from 56 bp to 76 bp, 5′-GTAGTGACCTCCCTGAAATCA-3′, as outlined in our previous work (Lee and Deneen, 2012). The target chick Rac1 sequence used to generate the shRNAi construct is located within the coding sequence from 5′-AGAGATAGGTGCAGTGAAA-3′. We also generated mutant Daam2 shRNAi (referred to as Mut-Daam2i; 5′-GTCGCGACCTCGCTCAAATCA-3′; bold characters indicate mutated bases) and Rac1 shRNAi (5′-AGCGACAGCTGTAGAGAAA-3′; bold characters indicate mutated bases). We generated the siRNAs against Daam2 and Rac1 using the On-TARGETplus siRNA design program (Dharmacon).
As previously described (Lee and Deneen, 2012; Lee et al., 2015), overexpression and shRNAi constructs were cloned into avian retrovirus RCAS vectors. Electroporation was performed unilaterally into the neural tube of stage 11–13 chick embryos (∼E2) through the use of a capillary needle and micromanipulator under a dissecting microscope. Using a BTX Electro Square Porator, unilateral electroporation was performed on the neural tube. Empty plasmid was utilized in co-electroporation experiments to ensure equal DNA concentrations. All expression constructs and siRNAs can be found in Table S1.
Chick explant culture
As previously described (Lee et al., 2015), the electroporated chick embryos were harvested and dissected using a micro tungsten needle in ice-cold PBS. The dissected spinal cords were then moved into Transwell plates and maintained for 48 h in Dulbecco's modified eagle medium (DMEM)/F12 containing 0.1% fetal bovine serum (FBS) and N2 supplement.
Immunostaining and in situ hybridization
Spinal cords were fixed in 4% paraformaldehyde and put in 20% sucrose for cryoprotection. After cryosectioning, we then performed in situ hybridization using mRNA probes for chick Daam2 and chick Rac1. Immunostaining and in situ hybridization were performed on chick spinal cord sections using standard conditions as we have previously described (Ding et al., 2020). Briefly, for in situ hybridization, RNA probes were generated in house, and all probes were tested for specificity with sense probes used as controls. For immunostaining slides, tissues were washed in PBS three times, for 5 min each, permeabilized with 0.3% Triton-X in PBS (PBST) for 5 min, washed with PBS, and blocked with 10% goat serum in PBST for 1 h at room temperature. Slides were then incubated with primary antibodies overnight at 4°C. Secondary antibody staining was performed by washing with PBS three times, incubating with secondary antibodies for 1 h at room temperature, washing with PBS and staining with 4′,6-diamidino-2-phenylindole (DAPI). Sections were then mounted with Vectashield mounting medium. All antibodies used for immunostaining are found in Table S1.
Cell line cultures
Cos7, Hela and HEK293T cells were cultured in DMEM (GenDEPOT) supplemented with 10% FBS (GenDEPOT) and 1% Pen-Strep (GenDEPOT). Cells were passaged with 0.25% Trypsin EDTA 1× (GenDEPOT). For co-IP assays, cell lysates were incubated with Protein A/G agarose beads (Thermo Fisher) and incubated overnight with target protein-specific antibodies at 4°C. Lysates containing beads were washed three times and boiled with 2× SDS sample buffer at 95°C for 10 min before western blot analysis. Antibodies used for biochemical assays are listed in Table S1.
Imaging and quantification
Images were taken with Zeiss Imager.M2m equipped with Apotome.2, Axiocam 506 mono and AxioCam MRc and captured using Zen 2 Software. Images were analyzed semi-automatically using Particle Analysis and Cell Counter plugins in ImageJ. Average cell intensity traces were calculated and plotted with MATLAB (MathWorks). Statistical analysis and quantitative graphs were plotted using GraphPad Prism7. All statistical details including n numbers and statistical tests performed can be found in each figure legend. Prism7 was used for graphs and statistics. One-way ANOVA or two-way ANOVA tests were performed, and P<0.05 was considered significant. Image processing and analysis were performed using ImageJ software.
We are grateful to Dr Lesley Chaboub for technical support during this study and Dr Benjamin Deneen for conceptual discussions and input on the manuscript.
Conceptualization: C.D.C., H.K.L.; Methodology: C.D.C., Q.Y., J.J., X.D., D.C., H.K.L.; Validation: J.J., C.-Y.W.; Formal analysis: C.D.C., C.-Y.W.; Resources: Z.C.; Data curation: Q.Y., X.D.; Writing - original draft: C.D.C., H.K.L.; Visualization: C.D.C.; Supervision: H.K.L.; Project administration: H.K.L.; Funding acquisition: H.K.L.
This work was supported the National Multiple Sclerosis Society (RG-1907-34551 to H.K.L.), the National Institute of Neurological Disorders and Stroke (R01NS110859-01 to H.K.L.), the Welch Foundation (AU-1731-20190330 to Z.C.), the Dan L. Duncan Cancer Center, Baylor College of Medicine and the Jan and Dan Duncan Neurological Research Institute. The project was supported in part by IDDRC grant number 1U54 HD083092 from the National Institute of Child Health and Human Development. Deposited in PMC for release after 12 months.
Peer review history
The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.251140.reviewer-comments.pdf
The authors declare no competing or financial interests.