Introduction

Skeletal muscle is a unique tissue composed of elongated multinucleated cells known as myofibers (Grefte et al. 2007). In response to injury, skeletal muscle has the capacity to repair the injured myofiber in a process called muscle regeneration (Chen and Shan 2019). Muscle regeneration is dependent on the resident muscle stem cells, known as satellite cells (Relaix and Zammit 2012). Satellite cells are located between the myofiber plasma membrane and the basement membrane (BM), the latter of which is a layer of extracellular matrix material composed of collagen, glycoproteins, and proteoglycans (Seale et al. 2000; Webster et al. 2016). Satellite cells express high levels of Pax7, which is a paired domain-containing transcription factor, and remain quiescent under normal conditions (Le Grand and Rudnicki 2007; Cheung and Rando 2013; Yin et al. 2013). Upon injury, satellite cells are activated and then proliferate and differentiate into fusion-competent muscle cells. Once the satellite cell-derived muscle cells, which will be referred to as SCMs, fill the space within the BM remnants, known as ghost fibers, they would fuse with each other and/or with injured myofibers to regenerate the muscle (Le Grand and Rudnicki 2007; Lepper et al. 2009; Cheung and Rando 2013; Yin et al. 2013; Webster et al. 2016; Collins et al. 2024). SCM fusion occurs rapidly between day 3.5 to 5 post injury (dpi) and persists till ∼dpi 10 (Collins et al. 2024). Despite the importance of SCM fusion in skeletal muscle regeneration, the molecular and cellular mechanisms underlying SCM fusion during muscle regeneration remain poorly understood. To date, only two proteins, the bi-partite myoblast fusogens myomaker (MymK) and myomixer (MymX)/myomerger/minion, have been shown to be required for SCM fusion (Millay et al. 2014; Bi et al. 2018). Identifying additional components of the SCM fusion machinery will not only facilitate our understanding of muscle regeneration but also provide more options to enhance muscle regeneration upon injury.

The Arp2/3 complex and its actin nucleation-promoting factors (NPFs), Neural Wiskott Aldrich Syndrome Protein (N-WASP) and WASP-family verprolin-homologous protein (WAVE), are known to generate branched actin filaments and have been shown to play a role in myoblast fusion during embryogenesis (Richardson et al. 2007; Gruenbaum-Cohen et al. 2012). These actin regulators are required for generating invasive protrusions to promote plasma membrane juxtaposition and fusion at the fusogenic synapse (Sens et al. 2010; Luo et al. 2022; Lu et al. 2023). However, their potential function in muscle regeneration is completely unknown. Here, using satellite cell- specific knockout (KO) mice of Arp2/3 and NPFs, we show that branched actin polymerization is indispensable for muscle regeneration. In particular, Arp2/3 and NPFs are required for the formation of invasive protrusions during SCM fusion, but not satellite cell proliferation, differentiation or migration. Thus, we have identified new components of the SCM fusion machinery and demonstrated a critical function of branched actin- propelled invasive protrusions in skeletal muscle regeneration.

Results

The redistribution of muscle cells and macrophages at the early stage of skeletal muscle regeneration

To examine the behavior of muscle cells during fusion after injury, we injured the tibialis anterior (TA) muscles by BaCl₂ injection (Fig. 1A), and labeled the differentiating SCMs using anti-NCAM, a cell adhesion molecule highly expressed in these cells (Capkovic et al. 2008), and ghost fibers using anti-Laminin, a major component of the BM (Webster et al. 2016) (Fig. 1B). Since a previous study showed that macrophages are present in the ghost fiber to clear the necrotic debris of the damaged myofibers (Collins et al. 2024), we also labeled macrophages with anti-MAC-2, a member of lectin family expressed on macrophage cell-surface (Hohsfield et al. 2022).

Figure 1.

At dpi 2.5, the differentiating SCMs and macrophages were two major cell populations residing within the ghost fibers, occupying 47.8% and 39.9% of the total volume, respectively. By dpi 3.5, SCMs occupied 98.2% of the ghost fiber volume, whereas macrophages only accounted for 1.8% (Fig. 1C), with most of the macrophages residing in the interstitial space outside of the ghost fibers (Fig. 1B). This would account for the high overall number of macrophages in the regenerating muscle tissues in this time period (Collins et al. 2024). How macrophages extravasated from the ghost fibers to the interstitial space remained an interesting question. Strikingly, our confocal (Fig. 2A and Supplementary Video 1 & 2) and transmission electron microscopy (TEM) (Fig. 2B) analyses of regenerating TA muscles at dpi 3 revealed narrow openings (∼1 μm diameter) on the BM of ghost fibers through which macrophages (MAC-2⁺) with a ∼20 μm diameter (as a round cell) were traversing, suggesting that the large macrophages squeeze through tiny openings on the BM to enter and/or escape the ghost fibers. The ability of macrophages to traverse the ghost fibers in and out makes it possible to spatiotemporally coordinate the ghost fiber occupancy by macrophages and SCMs during muscle regeneration.

Figure 2.

Branched actin polymerization is required for mammalian skeletal muscle regeneration

Given that the Arp2/3 complex-mediated branched actin polymerization is required for myoblast fusion during mouse embryogenesis (Lu et al. 2023), we asked whether the Arp2/3 complex is required for skeletal muscle regeneration in adults. Toward this end, we generated satellite cell-specific, tamoxifen-inducible knockout mice for ArpC2, a subunit of the Arp2/3 complex (Goley and Welch 2006), by breeding Pax7^CreERT2 with ArpC2^fl/fl mice (Lepper et al. 2009; Wang et al. 2016). The conditional knockout (cKO) mouse line Pax7^CreERT2; ArpC2^fl/fl will be referred to as ArpC2^cKO hereafter. The littermates of the Pax7^CreERT2 male mice without the floxed ArpC2 allele were used as wild-type controls. To induce genetic deletion of ArpC2 in satellites, we performed intraperitoneal injection of tamoxifen to the control and mutant mice every two days over a period of ten days. satellite cell-specific ArpC2 KO did not affect TA muscle weight and size in uninjured muscle (Fig. S1A-S1C). However, muscle injury by BaCl₂ resulted in a significant decrease (87.7 ± 2.0%) in the cross-section area (CSA) of regenerated myofibers in ArpC2^cKO mice compared to their littermate controls at dpi 14 (Fig. 3A-3C) and dpi 28 (Fig. S3A and S3B). Consistent with this, the frequency distribution of CSA displayed a significant shift toward the small size in the mutant mice (Fig. S2). Taken together, these data demonstrate that the Arp2/3-mediated branched actin polymerization is essential for skeletal muscle regeneration.

Figure 3.

Branched actin polymerization is required for SCM fusion

To pinpoint the specific step of skeletal muscle regeneration – satellite cell proliferation, differentiation, migration, and SCM fusion – in which branched actin polymerization is required, we performed immunostaining using antibodies that mark different steps.

These experiments revealed that the percentages of muscle cells positive for the proliferation marker (Ki67) and the muscle differentiation marker (MyoG) in the injured TA muscles were similar between control and ArpC2^cKO mice (Fig. S4A-S4E). In addition, live imaging analysis showed that the cultured ArpC2 KO SCMs exhibited similar migration and cell-cell contact (Supplementary Video 3). Thus, branched actin polymerization is dispensable for the proliferation, differentiation, and migration of SCs during skeletal muscle regeneration. The reduced muscle size in the ArpC2 mutant mice is therefore likely due to defects in SCM fusion.

Since SCM fusion is mostly completed at dpi 4.5 (Fig. 1B)(Collins et al. 2024), we examined the regenerating TA muscles of the control and ArpC2^cKO mice at this time point (Fig. 3D). While most of the SCMs within the ghost fibers had fused in the control animals, the ghost fibers in the ArpC2^cKO mice contained differentiated (NCAM⁺) but mostly unfused SCMs that could be readily observed in cross sections, similar to those in the Mymx^cKO mice (Fig. 3E, 3F and Fig. S5B). Consistent with this, the frequency distribution of SCM number per ghost fiber displayed a dramatic shift toward higher numbers in the ArpC2^cKO mice (Fig. S5C). These results indicate that the actin cytoskeleton plays an essential role in SCM fusion as the fusogenic proteins.

Interestingly, expression of the fusogenic proteins, MymK and MymX, was up-regulated in the TA muscle of these mice (Fig. S4F), suggesting that fusogen overexpression is not able to rescue the SCM fusion defect resulted from defective branched actin polymerization. Furthermore, cultured ArpC2^KO satellite cells exhibited a severe myoblast fusion defect despite normal differentiation (Fig. 3G-3I), and pharmacologically inhibiting Arp2/3 with CK666 had the same effects (Fig. S5D-3G). Taken together, these results demonstrate that branched actin polymerization is specifically required for SCM fusion during skeletal muscle regeneration.

N-WASP and WAVE families have partially redundant functions in regulating SCM fusion

Activation of the Arp2/3 complex requires the actin NPFs, including the WASP and WAVE family of proteins (Goley and Welch 2006). To examine their potential functions in mammalian muscle regeneration, we generated single and double cKO mice for N- WASP [the WASP family member with high expression in SCMs (Lu et al. 2023)] and CYFIP1 [a subunit of the WAVE complex (Chen et al. 2014)], respectively. Pax7^CreERT2; N-WASP^fl/fl will be referred to as N-WASP^cKO, Pax7^CreERT2; CYFIP1^fl/fl as CYFIP1^cKO, and Pax7^CreERT2; N-WASP^fl/fl; CYFIP1^fl/fl as dcKO hereafter.

At dpi 14, immunostaining revealed a moderate but significant decrease of TA myofiber CSA by 48.2 ± 0.1% in N-WASP^cKO and 67.7 ± 3.3% in CYFIP1^cKO mice, respectively, compared to their littermate controls (Fig. 3B and 3C). The myofiber CSA of dcKO mice was further decreased by 80.9 ± 1.8%, comparable to the CSA deduction observed in the ArpC2^cKO mice (in which both N-WASP and WAVE could not exert their functions) (87.7 ± 2.0%) (Fig. 3B and 3C) and in the Mymx^cKO mice (89.3 ± 1.9%). Consistent with the reduction of myofiber size, the N-WASP^cKO and CYFIP1^cKO single KO mice exhibited moderate myoblast fusion defects at dpi4, which were exacerbated in dcKO mice (Fig. 3E and 3F), despite normal satellite cell proliferation and differentiation, as well as fusogenic protein upregulation in the dcKO mice (Fig. S4A-S4F). Thus, our data have revealed partially redundant functions between N-WASP and WAVE NPFs in promoting myoblast fusion during skeletal muscle regeneration.

Branched actin polymerization promotes invasive protrusion formation during SCM fusion

To investigate the mechanism by which branched actin polymerization regulates SCM fusion during muscle regeneration, we first examined the cellular structure at the fusogenic synapse of cultured SCMs. Live cell imaging of cultured SCMs expressing LifeAct-mScarleti (mScar) and Arp2-mNeongreen (mNG) at day two in differentiation medium (DM) revealed Arp2- and F-actin-enriched finger-like protrusions projecting from the invading cells into their fusion partners (receiving cells), prior to cell membrane fusion (Fig. 4A and Supplementary Video 4). Consistent with this, TEM analysis of TA muscle at dpi 3.5 in wild-type mice revealed invasive finger-like protrusions at muscle cell contact sites (28.7 ± 3.9% of SCMs exhibited invasive protrusions, n = 83 SCMs from 20 ghost fibers examined) (Fig. 4B and 4C). The average length and width of the finger-like protrusions were 422 ± 200 nm (n = 32 invasive protrusions) and 121 ± 73 nm (n = 32 invasive protrusions), respectively (Fig. S6A and S6B). In contrast, the muscle cells in ArpC2^cKO mice seldom formed invasive protrusions (1.4 ± 0.7% of the muscle cells exhibited invasive protrusions, n = 147 SCMs from 20 ghost fibers examined) (Fig. 4B), whereas protrusions in SCMs of MymX^cKO mice appeared normal (n = 29 invasive protrusions) (Fig. 4B, 4C, Fig. S6A and S6B). Therefore, branched actin polymerization, but not the fusogenic protein MymX, is required for invasive protrusion formation to promote myoblast fusion during adult muscle regeneration.

Figure 4.

Discussion

In this study, we show that the Arp2/3 complex-mediated branched actin polymerization is indispensable for SCM fusion, but not for satellite cell proliferation, migration, or differentiation during muscle regeneration. The Arp2/3 NPFs, N-WASP and WAVE, exhibit partially redundant functions in regulating SCM fusion. Our live cell imaging and electron microcopy analysis revealed actin-propelled invasive protrusions at the fusogenic synapses of SCMs and a requirement for branched actin polymerization in generating these protrusions. Taken together, we propose that branched actin polymerization promotes mammalian muscle regeneration by generating invasive protrusions at the fusogenic synapse (Fig. 4D).

Studies in multiple organisms, including Drosophila, zebrafish, and mouse, have demonstrated that myoblast fusion during embryogenesis is mediated by actin-propelled invasive membrane protrusions (Sens et al. 2010; Chen 2011; Kim et al. 2015a; Kim and Chen 2019; Lee and Chen 2019; Luo et al. 2022; Lu et al. 2023). These protrusions serve several functions, including enhancing the plasma membrane contact areas between the fusion partners, increasing the mechanical tension of the fusogenic synapse, and bringing the two opposing cell membranes into close proximity to engage the fusogenic proteins (Sens et al. 2010; Shilagardi et al. 2013; Kim et al. 2015b). The current study is the first revealing invasive protrusions as an integral part of the fusion machinery in adult skeletal muscle regeneration. In addition, we have identified molecular determinants of the invasive protrusions – the Arp2/3 complex and its NPFs, N-WASP and WAVE. It is striking that depleting the branched actin polymerization machinery results in a similarly severe SCM fusion defect as does depleting the fusogenic protein MymX, highlighting the crucial role for actin cytoskeletal rearrangements in SCM fusion. Indeed, our previous work using a reconstituted cell- fusion culture system in Drosophila revealed that fusogens and branched actin regulators are two minimal components of the cell-cell fusion machinery, and that actin- propelled invasive protrusions are required for fusogen engagement across the two apposing cell membranes (Shilagardi et al. 2013). It will be interesting to determine whether invasive protrusions promote the trans-interactions of fusogens at the mammalian fusogenic synapse.

We show in this study, for the first time, that macrophages can squeeze through tiny openings on the basement membrane to infiltrate and/or evacuate the ghost fibers during muscle regeneration. It is unclear whether these openings are small tears due to muscle injury or generated by macrophages through their interactions with the basement membrane. Carman et al. reported that lymphocytes use invasive podosomal protrusions to palpate the surface of, and ultimately generate transcellular pores through, the endothelium in a process called diapedesis (Carman et al. 2007). Given that macrophages can also generate invasive protrusions (Faust et al. 2019), it is conceivable that macrophages may use these invasive structures to push apart the protein meshwork on the basement membrane and thus opening small holes to get in and out of the ghost fibers. This feature would allow macrophages to infiltrate relatively intact basement membranes during micro muscle injuries to clear the debris and allow satellite cell proliferation and SCM fusion. Future studies are required to identify the mechanisms by which macrophages traverse the basement membrane during muscle regeneration.

Materials and Methods

Mouse Genetics

All animals were maintained under the ethical guidelines of the UT Southwestern Medical Center Animal Care and Use Committee according to NIH guidelines.

C57BL/6J (stock: 000664) and Pax7^CreERT2 (stock: 017763) mice were obtained from The Jackson Laboratory. The ArpC2^fl/fl (Wang et al. 2016), N-WASP^fl/fl (Cotta-de-Almeida et al. 2007), and CYFIP1^fl/fl (Habela et al. 2020) mice were previously described.

The MymX^fl/fl line (Bi et al. 2018) was generously provided by Dr. Eric Olson. The control and mutant male littermates were used in each cohort of experiments.

Tamoxifen and BaCl₂-Induced Muscle Injury

Tamoxifen (Sigma; T5648) was dissolved at 20 mg/ml in corn oil. 100 μl tamoxifen/corn oil solution was administered by intraperitoneal injection to 2-month-old male mice as schematized in Figs. 3A, 3D, S1A, S3A, S4A and S5A. To induce muscle injury, BaCl₂ (Sigma; 342920) was dissolved in sterile saline to a final concentration of 1.2%, aliquoted, and stored at −20°C. Mice were anesthetized by isoflurane inhalation, the legs were shaved and cleaned with alcohol, and TA muscles were injected with 50 μL of BaCl₂ with a 28 gauge needle.

Satellite Cell Isolation and Culture

Satellite cells were isolated from limb skeletal muscles of 2-month-old male mice. Briefly, muscles were minced and digested in 800U/ml of type II collagenase (Worthington; LS004196) in F-10 Ham’s medium (ThermoFisher Scientific; 11550043) containing 10% horse serum for 90 minutes at 37 °C with rocking to dissociate muscle fibers and dissolve connective tissues. The dissociated myofiber fragments were collected by centrifugation and digested in 0.5U/ml dispase (Gibco; 17105041) in F-10 Ham’s medium for 30 minutes at 37 °C with rocking. Digestion was stopped with F-10 Ham’s medium containing 20% FBS. Cells were then filtered from debris, centrifuged and resuspended in growth medium (GM: F-10 Ham’s medium supplemented with 20% FBS, 4 ng/ ml FGF2, 1% penicillin–streptomycin and 10mM HEPEs). The cell suspension from each animal was pre-plated twice on the regular 100 mm tissue culture-treated dishes for 30 minutes at 37 °C to eliminate fibroblasts. The supernatant containing mostly myoblasts was then transferred into collagen-coated dishes for culture in GM. To induce myogenic differentiation, satellite cells were cultured in DM (DMEM supplemented with 2% horse serum, 1% penicillin-streptomycin and 10mM HEPEs).

Pharmacological Treatments of Satellite Cells

To pharmacologically inhibit branched actin polymerization in SCMs, the Arp2/3 inhibitor CK666 (50 μM) were added into the DM at day 0 of differentiation of wild-type SCMs.

After 48 hours, the cells were fixed in 4% paraformaldehyde (PFA) and stained with anti-MHC and DAPI to assess their differentiation and fusion index.

To delete ArpC2 in satellite cells in vitro, satellite cells isolated from ArpC2^cKO mice were cultured in GM supplemented with or without 2 μM 4-hydroxytamoxifen (Sigma; H6278) for 10 days. Subsequently, the cells were trypsinized and plated at 70% confluency in DM. After 48 hours, the cells were fixed in 4% PFA and stained with anti-MHC and DAPI to assess their differentiation and fusion index.

Retroviral Vector Preparations and Expression

The LifeAct-mScarleti and Arp2-mNeongreen constructs were described in the previous study (Lu et al. 2023), and assembled into the retroviral vector pMXs-Puro (Cell Biolabs; RTV-012) using the NEBuilder HiFi DNA Assembly Cloning Kit (NEB; E2621L). To package the retrovirus, two micrograms of retroviral plasmid DNA was transfected into platinum-E cells (Cell Biolabs; RV-101) using the FuGENE HD transfection reagent.

Two days after transfection, the virus-containing medium was filtered and concentrated with Retro-X Concentrator (Clontech, PT5063-2) following the manufacturer’s protocol. The concentrated retroviruses were diluted in GM (with a 1:1000 dilution), mixed with polybrene (7 μg/ml), and used to infect cells. One day after infection, cells were washed with PBS and cultured in fresh GM.

Immunohistochemistry

To co-stain NCAM, MAC-2, and Laminin, the 4% PFA fixed TA muscles were dehydrated in 30% sucrose at 4°C overnight. The specimens were embedded in Tissue- Plus O.C.T. Compound (Fisher Scientific; 23-730-571) and 12-μm cryosections were collected onto Superfrost Plus Microscope Slides (Fisher Scientific;12-550-15). Then, the cryosections were incubated with blocking buffer for 20 minutes at room temperature (RT), followed by overnight incubation with rabbit anti-NCAM (1:200; Millipore; AB5032), rat anti-MAC-2 (1:200; Biolegend; 125401), and rat anti-Laminin-2 (1:500; Sigma; L0663) at 4°C. To stain for dystrophin, the freshly dissected TA muscles were snap frozen in Tissue-Plus O.C.T. Compound and 12-μm cryosections were collected onto Superfrost Plus Microscope Slides. Next, the sections were fixed in 4% PFA for 12 minutes at RT, washed three time with PBS and incubated with blocking buffer for 20 minutes at RT, followed by overnight incubation with rabbit anti-dystrophin (1:200; Abcam; ab15277) at 4°C. To co-stain Pax7, MyoG, Laminin, and Ki67, the freshly dissected TA muscles were snap frozen in Tissue-Plus O.C.T. Compound and 12-μm cryosections were collected onto Superfrost Plus Microscope Slides. Then, the sections were fixed in 2% PFA for 5 minutes at RT, washed three time with PBS and incubated with blocking buffer supplemented with M.O.M blocking reagent (1:25; Vector; MKB-2213-1) for 60 minutes at RT, followed by overnight incubation with mouse anti-Pax7 (1:2; DSHB; Pax7), mouse anti-MyoG (1:2; DSHB; F5D), rat anti-Laminin-2 (1:500; Sigma; L0663), and rat anti-Ki67 (1:500; ThermoFisher Scientific; 14-5698-82) at 4°C. After the incubation with primary antibodies, the sections were extensively washed with PBS and then incubated with alexa fluor-conjugated secondary antibodies for one hour at RT. Subsequently, the sections were washed with PBS and subjected to imaging using a Leica TCS SP8 inverted microscope. Reconstruction of z-stacks to 3- dementional images was performed using Imaris 10.1.

Western Blot

For western blots, proteins were isolated from TA muscle using ice-cold RIPA buffer (150mM NaCl, 1% NP40, 0.1% SDS and 50mM Tris, PH7.4) containing protease and phosphatase inhibitors (Cell Signaling Technologies; 5872) for 20 minutes. The supernatants were collected by centrifugation at 140,000 x g for 15 minutes. Protein concentrations were determined using the Bradford Protein Assay Kit (Bio-Rad; 5000201). 10-30 μg total protein was loaded for each sample and separated by 10% SDS-PAGE gel and transferred to PVDF membranes (Millipore; GVHP29325). Then, the membranes were blocked for one hour at RT in PBS containing 5% nonfat dry milk and 0.1% Tween-20 (PBSBT) and subsequently were incubated with primary antibodies diluted at 1:1000 in PBSBT overnight at 4°C. The membranes were then washed with PBST and incubated with appropriate HRP-conjugated secondary antibodies diluted in PBSBT for one hour at RT. After extensive washes with PBST, the membranes were developed with the ECL western blotting substrate (ThermoFisher Scientific; 32209).

The following primary antibodies were used: sheep anti-ESGP/MymX (1:1000; R&D Systems; AF4580), mouse anti-MymK (Zhang et al. 2020) (1:1000), and rabbit anti-β- Tubulin (1:1000; Cell Signaling Technologies; 2146).

Time-lapse Imaging and Analysis

Time-lapse imaging of cells incubated in 5% CO₂ at 37 °C was performed on a Nikon A1R confocal microscope with a Nikon Biostation CT. The satellite cells were seeded on fibronectin-coated cover glass (MATTEK; P35G-0-14-C) and imaged using a 40× (0.4 NA) objective at indicated time points after switching from GM to DM. The cells were imaged at two- or five-minute interval. After time-lapse imaging, ImageJ (NIH, 64-bit Java 1.8.0_172) was used to project the z-stacks in 2D, using maximum intensity projection and the resulting 2D images were assembled into a time-lapse video.

Electron Microscopy

To observe the invasive protrusions at the contact sites of SCMs during muscle regeneration in vivo, TA muscle at dpi 3.5 were fixed in a solution containing 3% paraformaldehyde, 2% glutaraldehyde, 1% sucrose, 3mM CaCl₂ in 0.1M sodium cacodylate buffer (pH 7.4) overnight at 4°C. Samples were subsequently washed with 0.1M cacodylate buffer containing 3% sucrose and 3mM CaCl₂, and post fixed with 1% osmium tetroxide in 0.1M sodium cacodylate buffer for 1.5 hours on ice. The muscle samples were stained with 2% uranyl acetate, dehydrated and embedded in EPON resin as previously described (Zhang and Chen 2008). The embedded samples were then cut into 70-nm thick sections using LEICA ultramicrotome (UC6) and collected on copper slot grids. These sections were post-stained with 2% uranyl acetate and sato’s lead solution and examined using a JEOL 1400 transmission electron microscope.

Statistics and Reproducibility

Statistical significance was assessed using two-tailed student’s t-test. The P values were obtained using GraphPad Prism 8. The numbers of biological replicates for each experiment are indicated in the figure legends.

Acknowledgements

We thank Dr. Eric Olson for generously providing the MymX^fl/fl mice, UT Southwestern Animal Resource Center for assistance with mouse colony maintenance, and UT Southwestern Quantitative Light Microscopy Core Facility for assistance with 3D reconstruction of confocal images. This work was supported by an NIH grant (R35GM136316) to E.H.C.

Author contributions

Y.L. and E.H.C. designed the project. Y.L., T.W., P.P., and C.Z. performed experiments. Y.L., and E.H.C. analyzed the data, made the figures, and wrote the manuscript. All authors commented on the manuscript.

Data availability

The data supporting the findings of this study are available within the article and its supplementary files.

Declaration of interests

The authors declare no competing financial interests. Correspondence and requests for materials should be addressed to E.H.C. (Elizabeth.Chen@UTSouthwestern.edu) and Y.L. (Yue.Lu@UTSouthwestern.edu).