An mTurq2-Col4a1 mouse model allows for live visualization of mammalian basement membrane development

The basement membrane (BM) is an ancient specialized extracellular matrix present in all metazoans (Özbek et al., 2010). It is a thin, dense sheet (∼50–100 nm) that underlies all epithelial and endothelial tissues. The BM serves multiple functions, including structural support, maintenance of tissue integrity, and regulation of cell adhesion, migration, proliferation, and differentiation (Yurchenco, 2011). It is a complex structure composed of a network of proteins, glycoproteins, and proteoglycans and consists of three main layers: the lamina lucida, lamina densa (also called the basal lamina), and in some structures, the lamina fibro-reticularis (Timpl, 1996). Although it has been suggested that the lamina densa and lamina lucida subdivisions of the BM may be artifacts of conventional EM processing, it is more likely that these spatial compartments do exist but are masked by retention of soluble components in high-pressure freezing/freeze substitution EM sample preparation (Keene and Tufa, 2020). The lamina lucida is in direct contact with the basal surface of epithelial or endothelial cells and is made up primarily of laminins, nidogen, and other glycoproteins that are essential for cell adhesion and migration (Stanley et al., 1982; Schwarzbauer, 1999; Keene and Tufa, 2020; McGrath et al., 1994). The lamina densa, composed predominantly of type IV collagen, provides structural stability and acts as a barrier, separating the epithelial/endothelial compartment from the underlying mesenchyme (Pozzi et al., 2017; McGrath et al., 1994; Sauter et al., 1998).

Importantly, the BM acts as a selective filter, controlling the passage of cells and molecules between these different tissue compartments (Kelley et al., 2014). Many normal and pathological events involve cells transiting across BMs, including immune cell infiltration, cell migration during development, epithelial to mesenchymal transition, and cancer metastasis (Morrissey and Sherwood, 2015; Nakaya and Sheng, 2013; Micalizzi et al., 2010; Ratzinger et al., 2002). As well as acting as a selective barrier, BMs are essential for cell migration as they provide an active substrate to which cells can adhere. Laminins in BMs act as ligands to integrin receptors, coupling extracellular environmental cues to the actin cytoskeleton of cells (Hynes, 1992; Sekiguchi and Yamada, 2018b). In doing so, the BM also regulates cell behaviors. In terms of cell guidance, the BM is a reservoir of signaling cues and matrix molecules that guide many cell types along their migration path, including neurons during axonal pathfinding and neural crest-derived cells (Yasunaga et al., 2010; Leonard and Taneyhill, 2020; Banerjee et al., 2013). Tissue morphogenesis itself is regulated by BMs, via both molecular and biophysical processes. In the Drosophila embryo, the knockdown of collagen IV reduces adult wing size due to a reduction in BMP signaling (Ma et al., 2017), and heterogeneity of BM stiffness in Drosophila is important for egg elongation (Chlasta et al., 2017). Overall, the BM performs a wide variety of functions, including as a barrier, substrate, growth regulator, and signaling platform, from early development throughout the whole of adult life (Sherwood, 2021).

Studying BM dynamics or the crossing of cells through BMs in vivo is particularly challenging. BM crossing events are stochastic in nature and occur deep within tissues. Therefore, much of what is known about how cells cross BMs is derived from in vitro studies (Glentis et al., 2017; Bahr et al., 2022; Albini and Noonan, 2010), which, although informative, still require validation in the in vivo context. Our knowledge of the in vivo dynamics of BMs has come from elegant experiments in Drosophila and Caenorhabditis elegans, in which endogenous components of the BM have been fluorescently labeled. The Vkg::GFP (collagen IVa2) Drosophila line, created as part of a protein-trap strategy (Morin et al., 2001; Buszczak et al., 2007), has been invaluable in studying dissemination of tumor cells through the BM, as well as in BM-controlled shaping of growing organs during development (Tamori et al., 2016; Rei et al., 2021,Preprint; Harmansa et al., 2023; Page-McCaw et al., 2007; Page-McCaw, 2008). The stereotyped invasion of the C. elegans anchor cell through the uterine and vulval BMs during development, together with the successful endogenous tagging of over 60 C. elegans BM proteins has greatly increased our understanding of BM composition, dynamics, and the cellular and molecular mechanisms underpinning the invasion process (Schindler and Sherwood, 2013; Sherwood and Sternberg, 2003; Costa et al., 2023; Kelley et al., 2017, 2019; Keeley et al., 2020; Kenny-Ganzert and Sherwood, 2024).

A key bottleneck in our understanding of BM dynamics in mammals (and indeed vertebrates) is the inability to visualize the BM live, in vivo. Tagging endogenous BM components presents unique challenges because many are subject to extensive proteolytic cleavage and post-translational modifications (Frantz et al., 2010). Moreover, the addition of a fluorescent tag to a protein monomer that is subsequently highly crosslinked and integrated into a complex and dense protein network can interfere with the formation of that network via steric hindrance. Several attempts to visualize the vertebrate BM in vivo have been made, for example, the lamC1:lamC1-sfGFP zebrafish line recapitulates the reported mRNA expression pattern of endogenous lamC1, but the tag itself is not inserted into the endogenous genomic locus and the fusion protein is expressed as an additional copy (Yamaguchi et al., 2022; Sztal et al., 2011). Similarly, Futaki et al. (2023) have generated a R26-CAG-Nid1-mCherry mouse line in which the BM can be visualized in vivo; however, again this tag is not inserted in the endogenous locus; the fusion protein is expressed from a ubiquitous promoter that drives ectopic expression of Nid1-mCherry and causes reduced localization of endogenous Nid1. More recently, Morgner et al. (2023) generated an endogenous Lamb1-Dendra2 mouse model in which heterozygous animals have fluorescently labeled laminin β-1-containing BMs, which enabled tumor-cell mediated BM turnover to be measured using intravital microscopy. A limitation of this model, however, is that only BMs containing the laminin β-1 isoform can be visualized. In order to visualize all BMs and truly appreciate the structural complexity of the BM as a barrier, the ideal component to fluorescently tag is collagen IV, as it makes up ∼50% of all BMs (LeBleu et al., 2007).

Collagen IV is a heterotrimeric molecule, expressed ubiquitously in BMs, with all the collagen IV in mammals being derived from six different α-chains, encoded by the genes COL4A1, COL4A2, COL4A3, COL4A4, COL4A5, and COL4A6 (Netzer et al., 1998; Mak and Mei, 2017). All BMs are composed solely of α1 and α2 chains during development, with some specialized structures adding a secondary network during maturation, such as α3/α4/α5 in the kidney glomerulus and α5/α5/α6 in mucosal epithelia (Kruegel and Miosge, 2010; Komori et al., 2018). Collagen IV is not required to initiate BM deposition during development, but the loss of both Col4a1 and Col4a2 is embryonically lethal in mice by E10.5 due to the lack of BM structural integrity and the breakdown of Reichert’s membrane (Pöschl et al., 2004). As the α1 and α2 chains are ubiquitously expressed in BMs, we sought to label the endogenous Col4a1 gene locus in mice to enable all BMs to be visualized.

In this work, we present the novel mTurquoise2-Col4a1 (mTurq2-Col4a1) mouse line, in which endogenous collagen 4a1 (COL4A1), and therefore all BMs, are labeled. Despite being homozygous lethal, which is a developmental phenotype that requires further analysis, heterozygous animals are healthy and fertile, with normal BM architecture and composition in the dermal–epidermal junction BM of the skin, suggesting incorporation of the mTurquoise2 protein into one copy of endogenous Col4a1 can support normal functionality of murine BMs. Furthermore, the stability and brightness of mTurquoise2 mean that BMs are clearly visible in heterozygous animals, allowing for long-term live imaging to be undertaken. Using an innovative mounting and imaging methodology, we show that the dermal–epidermal junction BM (hereafter referred to as the epidermal BM for simplicity) in embryonic mouse skin can be imaged long-term using explants from mTurq2-Col4a1 mice. This has revealed new insights into epidermal BM morphology and behavior during hair follicle development, suggesting a previously underappreciated inherent pliability. Moreover, this flexibility of the epidermal BM allows dividing basal progenitor cells to maintain adhesion to the BM even during division, challenging the notion that all dividing cells reduce or lose their adhesion to the substrate when they round up in metaphase. We describe novel observations of COL4A1 in the dermis, including a potential role in stabilizing the developing dermal condensate. Finally, we show that despite the significant growth that occurs during E13.5 and E15.5, COL4A1 is strikingly stable in the BM of the developing epidermis, suggesting that the inherent pliability observed may be what allows for growth and morphogenetic changes rather than extensive proteolysis and degradation of the core components. We envisage that the mTurq2-Col4a1 mouse model will be a useful addition to the field of BM research as it allows direct visualization of BM assembly, deformation and remodeling during development, and should be applicable to pathological conditions. Moreover, this model will allow many of the important observations made in vitro and in invertebrates to be visualized directly in the clinically relevant mouse model.

The collagen IV α chains are characterized by a central triple helical collagenous region of Gly-XY repeats, flanked by a 7S domain at the amino terminal and a globular NC1 carboxy domain (Fig. 1 A). The central collagenous region confers both structural stability and flexibility to the molecule, the NC1 domain is required to assemble the collagen α1α1α2(IV) heterotrimer, and the 7S domain is involved in the assembly of the formed heterotrimers into a more complex lattice network to form the BM (Timpl, 1989; Paulsson, 1992). It therefore follows that inserting a fluorescent protein in any position that interferes with or disrupts these domains may interfere with protein function and/or proper assembly of the BM. To inform our strategy, we studied the position of the Drosophila Vkg::GFP protein trap insertion, which is viable and labels all BMs (Morin et al., 2001; Buszczak et al., 2007). In this line, GFP is inserted between the N′ terminal signal sequence, which is cleaved within the cell before secretion, and the highly conserved 7S region. We mapped the Vkg::GFP insertion locus to the analogous position in the mouse Col4a1 gene, which corresponds to a position toward the 5′ end of exon 2 (Fig. 1 B), specifically a target insertion site on the COL4A1 protein between amino acids K28 and G29.

Many fluorescent proteins, despite being termed monomeric, have a tendency to oligomerize under physiological conditions (Costantini et al., 2012). We therefore chose mTurquoise2 as our fluorescent protein tag as it does not homodimerize under any of the currently established gold-standard tests (Goedhart et al., 2012). mTurquoise2 also has an “in-built” linker region consisting of a short N′ terminal β-sheet juxtaposed away from the main β-barrel, which precludes the need for artificial linker sequences, which may be required for other fluorescent protein fusions such as mCherry (Goedhart et al., 2012). Moreover, mTurquoise2 has a low acid sensitivity (pKa 3.1), meaning it is stable in the lower pH of the lumenal and extracellular space (Denda et al., 2000; Chan and Mauro, 2011). It is also bright and photostable (t½ = 90 s compared with ∼50 s for EGFP), and its excitation/emission spectrum means it can be used in combination with YFP, mCherry/tdTomato, and other far-red fluorescent proteins (Lambert, 2019).

To insert the mTurquoise2 ORF into the target locus of the Col4a1 mouse gene, we used highly efficient two-cell homologous recombination CRISPR/Cas9-mediated mutagenesis as previously described (Gu et al., 2018, 2020). A circular plasmid repair template containing the mTurquoise2 ORF flanked by ∼1,000 bp homology arms, together with a sgRNA and Cas9 mRNA, was injected into the cytoplasm of both blastomeres of two-cell mouse embryos. Using this approach, we successfully generated an mTurquoise2-collagen4a1 mouse line (mTurq2-Col4a1), in which the mTurquoise2 ORF is inserted downstream of the cleaved signal sequence and upstream of the conserved 7S domain (Fig. 1 B). Histological analysis of adult skin from mTurq2-Col4a1/+ mice confirmed robust and clear localization of mTurq2-COL4A1 signal in the epidermal BM (Fig. 1, C–E and Fig. S1 A). In the kidney, mTurq2-COL4A1 signal was observed in Bowman’s capsule, the mesangial matrix, and tubular BMs. (Fig. 1, F–I and Fig. S1 C). The mTurq2-COL4A1 signal in each tissue recapitulated the expression pattern of collagen IV antibody staining (Fig. 1, C”, D”, and G’). We also labeled both these tissues for perlecan, the most abundant heparan sulfate proteoglycan in the BM (Fig. 1, E, H, and I; and Fig. S1, B, E, and F). Aggregated perlecan connects the laminin polymers of the BM into the collagen IV network, in a “spot-welding” manner, thus connecting these two distinct and critical BM networks (Behrens et al., 2012). The pattern of mTurq2-COL4A1 expression coincided with perlecan expression in all BMs observed.

The BM is established during embryonic development, with laminins expressed around gastrulation followed by collagen IV α1 and α2 chains shortly thereafter (Yurchenco et al., 2004). Studies have shown that although collagen IV is not required to initiate BM assembly, it is required for functional stabilization of BMs from as early as E10.5 (Guo et al., 1991; Guo and Kramer, 1989; Pöschl et al., 2004). Moreover, unlike laminins, which express additional subunit isoforms in post-natal life, collagen α1 and α2 chains are laid down mainly during embryogenesis, with only a small number of BMs replacing some of their α1 and α2 chains with α3/α4/α5 chains (Gunwar et al., 1998). We therefore examined the epidermal BM in embryonic mouse skin. Both WT and mTurq2-Col4a1/+ embryos were labeled with antibodies against collagen IV (ColIV), laminin β-1 (Lamb1), and perlecan (Fig. 2). The epidermal BM was clearly labeled in mTurq2-Col4a1/+ embryonic skin (Fig. 2, B''' and E'''), and its localization overlapped with that of the ColIV antibody (Fig. 2 B").

The spatial relationship between the two major structural components of the BM, the lamina lucida and the lamina densa, is critical to BM function. The supramolecular laminin-containing network of the lamina lucida is linked to the collagen IV lamina densa network by the HSPG perlecan, with a potential role also played by agrin (Behrens et al., 2012; Hohenester and Yurchenco, 2013). In keeping with this, mutations in human COL4A2 result in a failure of COL4A2 and laminin colocalization in the epidermal BM (Murray et al., 2014). Similarly, a 40% reduction in perlecan staining in all major BMs was observed in mice carrying Col4a1 mutations (Jones et al., 2016). We, therefore, asked whether the relationships between these critical BM proteins are affected by the incorporation of mTurq2-COL4A1. To do so, we measured the linear correlation between Lamb1 and ColIV antibody labeling in WT and mTurq2-Col4a1/+ E18.5 epidermal BMs at pixel-level scale in high-resolution images (see Materials and methods). As expected, we saw that the Pearson’s Correlation Coefficient (PCC) r value was variable, as both supramolecular networks will not occupy the same space, but crucially, no difference was detected between WT and mTurq2-Col4a1/+ backskins, suggesting that the mTurq2-Col4a1/+ allele does not interfere with the network interactions of laminin β-1 and collagen IV (Fig. 2 C). We repeated this analysis measuring the correlation between Lamb1 and perlecan antibody labeling and observed that the perlecan localization profile in mTurq2-Col4a1/+ E18.5 epidermal BMs resembled that of WT littermates (Fig. 2, D’’, E’’, and F). Taken together, these data demonstrate that the structure of the epidermal BM in mTurq-Col4a1/+ embryos resembles that of WT littermates.

Despite numerous (>10) heterozygous crosses at several stages during development of the mTurq2-Col4a1 line, we never observed the birth of any homozygous pups. This suggests that the mTurq2-Col4a1 line is embryonically homozygous lethal, although we saw no apparent defects in heterozygous mice, which resemble WT littermates in terms of longevity, fertility, and overall health. The fact that heterozygous animals were viable, healthy, and fertile suggested that incorporation of the mTurq2-COL4A1 protein was not overtly deleterious to BM function when present as a single genomic copy. Furthermore, our correlation coefficient data in skin showed no difference in the spatial relationships between core BM components, as would be expected if incorporation of the mTurq2-COL4A1 protomer was detrimental to BM formation. However, despite attaining high-resolution images to complete this analysis, our data were inherently limited by the resolution limits of light microscopy. To study the BM of mTurq2-Col4a1/+ epidermis at the ultrastructural level, therefore, we next turned to electron microscopy.

Using transmission electron microscopy (TEM), we performed a detailed analysis of skin architecture and the BM at the dermal–epidermal junction of E18.5 mTurq2-Col4a1/+ embryos (Fig. 3). The overall skin architecture in mTurq2-Col4a1/+ embryos was similar to WT littermates, consisting of five distinct layers (dermis, basal, spinous, granular, and stratum corneum) that displayed characteristic morphologies associated with each layer (i.e., dense keratin bundles in the spinous layer and keratohyalin granules in the granular layer; Fig. 3, A and B; and Fig. S2). High magnification views of the junction between the basal epithelial layer and the underlying dermis showed numerous and regularly spaced BM thickenings at sites of hemidesmosome attachments (Fig. 3, C–F and Fig. S2). These electron-dense sites of dermal–epidermal adhesion consisted of inner and outer hemidesmosome plaques associated with intracellular keratin intermediate filaments (Fig. 3, G and H; and Fig. S2). Beneath the plasma membrane of basal epidermal cells, the lamina lucida and lamina densa layers of the BM were each clearly distinguishable. Furthermore, the thickness of the lamina densa, where type IV collagen accumulates, was comparable between the skins of mTurq2-Col4a1/+ and their WT littermates (Fig. 3, G and H; and Fig. S2). Thus, at the ultrastructural level, the skin of mTurq2-Col4a1/+ embryos displayed all the hallmarks of normal dermal–epidermal junction assembly.

Having shown that our mTurq2-Col4a1/+ mice display robust mTurq2-COL4A1 signal in both adult and late embryonic stage skin, we now sought to characterize the distribution of mTurq2-COL4A1 during BM assembly and maturation in earlier embryonic stages. To do this, we harvested embryos at E15.5 and labeled the backskins of both WT and mTurq2-Col4a1/+ embryos with antibodies against laminin β-1 (Lamb1), collagen IV (ColIV), and perlecan (Fig. 4). XZ optical reconstructions of Z stack images of E15.5 skins clearly show robust mTurq2-COL4A1 signal in the epidermal BM of both the IFE and developing hair follicles that recapitulates the ColIV antibody labeling (Fig. 4 B). As described above, we then calculated the PCC of the spatial relationship between Lamb1 and ColIV antibody labeling in both WT and mTurq2-Col4a1/+ E15.5 epidermal BMs (Fig. 4, A and B) and detected no difference between the genotypes (Fig. 4 C), further validating our previous findings that incorporation of mTurq2-COL4A1 into the epidermal BM does not alter the interaction between these two supramolecular layers, at least in the skin.

To take this analysis further, we next measured how much of the laminin β-1 localization domain within the epidermal BM overlaps with that of the collagen IV localization domain. We analyzed high-resolution images of the epidermal BM from both WT and mTurq2-Col4a1/+ backskins and applied a custom FIJI macro to calculate the fraction of each protein that overlaps with the other (overlap fraction; see Materials and methods). Our results demonstrate that a higher fraction of collagen IV (∼70%) is found to overlap with laminin β-1 localization at E15.5 (Fig. 4 E). Laminin β-1 is more frequently observed not colocalizing with collagen IV (<50% overlap; Fig. 4 D). Crucially, we observed no difference between the overlap fractions of laminin β-1 and collagen IV in WT and mTurq2-Col4a1/+ backskins (Fig. 4, D and E). We repeated these analyses with the laminin β-1 and perlecan localization domains (Fig. 4, H–J). Here, we observed a similar contribution to the area of colocalization by both laminin β-1 and perlecan (Fig. 4, I and J), and again we saw no difference between WT and mTurq2-Col4a1/+ in either analysis. As these results reflected what we observed in E18.5 embryos, we expanded our analysis to E12.5 embryos, a developmental time prior to both hair follicle development and stratification, when the basal progenitor layer is one-cell thick (Fig. 4, K–Q). We hypothesized that due to the significant growth of the embryo and expansion of the skin between these stages, the epidermal BM morphology and overlap composition would differ in E12.5 embryos. Again, we observed well-defined labeling of the epidermal BM with mTurq2-COL4A1 that recapitulated ColIV antibody labeling (Fig. 4, K–L) and detected no differences in the spatial relationships between laminin β-1 and collagen IV/perlecan in E12.5 WT and mTurq2-Col4a1/+ BMs, respectively (Fig. 4, K–O and P–T). Surprisingly, we observed very similar overlap fractions of laminin β-1 and collagen IV/perlecan in E12.5 compared with E15.5. This suggests that the morphology and composition of the epidermal BM, with respect to these core components, does not change significantly over time and is established during the early stages of embryonic skin development.

Having shown that the mTurq2-Col4a1 mouse model faithfully labels the epidermal BM, we next explored its utility for live imaging developmental BM dynamics. As the epidermal BM has a thickness of around 100 nm and developing mouse skin is inherently undulating, to properly visualize BM dynamics, a sagittal view of developing skin explants is necessary. Established live imaging approaches obtain planar XY images and the sagittal view is reconstructed from the Z-stack (Fig. 5 A). However, Abbe’s formula (lateral (XY) resolution = λ/2NA and axial (Z) resolution = 2λ/NA²) states that lateral resolution will always be greater than axial resolution in both widefield and confocal microscopy (Huszka and Gijs, 2019), so obtaining a high-resolution sagittal image is a limitation of this method. To image the epidermal BM in all three dimensions with a similar resolution and without the need for software reconstruction, we developed a novel live imaging technique, PS Multiview imaging. This simple technique involves folding the skin explant over an agarose ridge allowing the focal plane to be set at either the standard planar imaging plane (XY) or the sagittal imaging plane (“XZ”; Fig. 5 B and Video 1). This sagittal view of the developing skin allows for high-resolution imaging of cells and tissues in “Z,” without the need for reconstruction. By combining PS Multiview imaging with mTurq2-Col4a1/+ backskins in which all cell membranes are labeled with tdTomato (mTmG/mTmG), we were able to long-term live image the epidermal BM in developing skin explants (Fig. 5 C and Video 1).

As we were able to long-term live image the developing epidermal BM, we next wanted to visualize BM dynamics during the morphogenesis of hair follicles, which involves an initially thickened placode of basal epidermal cells that buds into the underlying dermis. The BM has been shown to play both an active and passive role during tissue development (Walma and Yamada, 2020) and is thought to be extensively remodeled during epithelial morphogenesis. For example, during branching morphogenesis, the BM surrounding actively growing branch tips becomes thin and/or discontinuous with microperforations, which is thought to allow for tip expansion and branch elongation (Harunaga et al., 2014; Spurlin et al., 2019; Sekiguchi and Yamada, 2018a). Given the wide-ranging roles of the BM during morphogenesis, we sought to investigate BM dynamics during hair follicle budding by live imaging E14.5–E15.5 mTurq2-Col4a1/+;mTmG/mTmG backskins over several hours using PS Multiview imaging (Fig. 5 C and Video 1).

We used the fluorescence intensity of the mTurq2-COL4A1 signal as a read-out for changes to the epidermal BM under an invaginating hair follicle relative to the neighboring IFE that is not undergoing significant morphological changes (Fig. 5 D). Surprisingly, the mTurq2-COL4A1 signal intensity was significantly greater at the actively growing base of developing hair follicles than at the IFE (Fig. 5 E). Although we had expected the intensity to become fainter, more diffuse, or discontinuous as downward budding occurred, the signal became brighter and qualitatively more uniform in appearance. This suggested that the epidermal BM may be compressed by the epithelium during hair follicle invagination or that new mTurq2-COL4A1 incorporation exceeds proteolytic degradation during budding, unlike in other examples of budding/branching morphogenesis (Wang et al., 2017; Sekiguchi and Yamada, 2018a).

Another cellular force that acts upon the epidermal BM is cell division. Several studies have shown that when cells in culture approach mitosis, they detach from the ECM and disassemble their adhesions to retract from the matrix and round up in metaphase (Champion et al., 2017; Suzuki and Takahashi, 2003; Yamakita et al., 1999). U2OS cells have been shown to use reticular adhesions to mediate substrate attachment during mitosis due to the complete disassembly of focal adhesions (Lock et al., 2018). Similarly, epithelial cells are thought to reduce or lose their adhesions to the BM when undergoing division, and this breakdown of focal adhesions during mitosis has been shown to be important for macrophage infiltration in Drosophila (Matus et al., 2014; Jones et al., 2019; Akhmanova et al., 2022). Since basal cells within the epidermis divide frequently during development (Damen et al., 2021; Box et al., 2019), we wanted to examine the interface between dividing basal cells and the epidermal BM. We analyzed several movies in which numerous basal cells were seen to go from interphase through to complete cytokinesis, and, to our surprise, we observed that when dividing cells rounded up in metaphase, they did not detach from the epidermal BM (Fig. 5 F; and Fig. S3, A and B; Video 2). Instead, rounded cells remained closely adhered to the BM, which subsequently deformed around the base of the dividing cell. To quantify this BM deformation, we measured the angle of the BM at the junction between the dividing cell and its neighbors before, during, and after mitosis (interphase, metaphase, and cytokinesis, respectively). We found that the epidermal BM pinches inward by an average of 25° underneath a rounded dividing cell and then reverts to the near-straight interphase angle after division, implicating maintained adhesion to the BM throughout mitosis (Fig. 5 G and Fig. S3 C). Antibody labeling for integrin β4, the laminin receptor essential for cell-BM hemidesmosomes, showed that dividing cells maintain integrin β4–mediated adhesions during mitosis (Fig. S3, D–G). Moreover, TEM analysis of dividing cells revealed hemidesmosome attachments are not only maintained during rounding but appear unchanged from those in non-dividing cells (Fig. S3, H–J and Fig. 3). Taken together, these data show that dividing basal progenitor cells do not disassemble hemidesmosome adhesions to the BM during division. In keeping with our previous observations of epidermal BM deformation underneath invaginating hair follicles, this suggests an inherent pliability of the epidermal BM during embryogenesis. BM stiffness has been shown to direct cellular behaviors and sculpt organs during development, and stiffer BMs are associated with cell migration and epithelial-to-mesenchymal transition (Crest et al., 2017; Ramos-Lewis and Page-McCaw, 2019; Wei et al., 2015). It is perhaps therefore intuitive that a BM underlying a rapidly proliferating progenitor cell layer during stratification needs to have sufficient pliability to allow cells to remain adhered throughout mitosis and to deform during morphogenesis.

Although we focused primarily on the epidermal BM for this study, PS Multiview imaging enabled visualization of a sharp delineation between the epidermis and dermis, with the dermis having a diffuse mTurq2-COL4A1 signal throughout, which was completely absent in the epidermis, consistent with dermal fibroblasts being the major source of collagen IV in the skin (Zorina et al., 2023; Video 1 and Fig. 5 H, bottom panels). Moreover, several other structures in the dermis showed robust mTurq2-COL4A1 localization, including vascular BMs and as yet unidentified migrating cells containing bright puncta of mTurq2-COL4A1 (Video 1 and Video 3). We also noticed a consistent diffuse mTurq2-COL4A1 signal in the dermis that coincided with the dermal condensate, underlying each hair follicle. Upon closer inspection, it appeared that a “basket” of collagen IV strands surrounds the cells of the dermal condensate as it develops (Fig. 5 H). This raises compelling questions regarding the role of collagen IV in maintaining the structural integrity of mesenchymal condensates during development, and it would be interesting to examine other mesenchymal condensates, such as those of the lung and kidney, to see if this phenomenon is conserved (Mammoto et al., 2011, 2015).

Having established that our mTurq2-Col4a1 mouse model is amenable to live imaging approaches, we next assayed the stability of mTurq2-COL4A1 in the epidermal BM of mTurq2-Col4a1/+ backskins at E13.5 and E15.5, stages when the skin undergoes significant growth and morphological changes (stratification and hair follicle development). Several studies in invertebrates have shown that COL4A1 within the BM is highly immobile (Morrissey et al., 2016; Matsubayashi et al., 2017; Keeley et al., 2020). The half-life of BM proteins in the Drosophila embryo has been measured to be on the order of ∼7–10 h (Matsubayashi et al., 2020). Additionally, a recent study in mice found that active secretion of vascular BM COL4A1 could only be detected early in development, suggesting that secretion thereafter, throughout adult life, may occur at much lower levels (Lartey et al., 2023). We therefore hypothesized that mTurq2-COL4A1 would be highly stable in the epidermal BM but may show increased mobility at E13.5 compared with E15.5 due to the rapid growth taking place during this earlier stage. We conducted FRAP on the interfollicular epidermal BM at both E13.5 and E15.5 and observed minimal mobility, with only a small mTurq2-COL4A1 mobile fraction of <15% at both stages (Fig. 6 C). We also performed FRAP on the BM around the rim of the hair follicle (E15.5), where the mTurq2-COL4A1 signal appears brightest (Fig. 6, A and B), and again observed minimal mTurq2-COL4A1 mobility. These data suggest that epidermal BM mTurq2-COL4A1 is highly stable within the epidermal BM, even as early as E13.5, consistent with other published datasets.

Our understanding of the complexity of the BM has evolved significantly over the past few decades, and it is no longer considered a simple, static barrier between tissue compartments. Numerous studies have highlighted the critical roles that the BM plays in many physiological processes, including development, cell signaling, and disease progression (Töpfer, 2023; Pozzi et al., 2017; Randles et al., 2017; Banerjee et al., 2022). Elegant studies in Drosophila and C. elegans with fluorescently tagged BM components have significantly increased our understanding of the BM in terms of its interaction with neighboring cells and the mobility of individual components within it (Isabella and Horne-Badovinac, 2015; Keeley et al., 2020; Kelley et al., 2019; Clay and Sherwood, 2015; Van De Bor et al., 2021). However, until recently, the key bottleneck in our understanding of mammalian, and indeed vertebrate BMs, has been the lack of models in which endogenous BM components are labeled. Several attempts to generate such models have been unsuccessful (Shaw et al., 2020), and in vivo studies have relied upon knock-in reporters, which do not necessarily reflect endogenous protein expression and can be expressed ectopically and/or interfere with native protein function (Yamaguchi et al., 2022; Sztal et al., 2011; Futaki et al., 2023). The recent success by Morgner et al. (2023) in generating a Lamb1-Dendra2 mouse model provided a proof-of-principle that fluorescently tagging BM components in mice is feasible, and in this study, we have advanced one step further by generating a mouse line in which the BM ubiquitous COL4A1 is labeled. Collagen IV is the primary component of all BMs, and we, therefore, envisage that our mTurq2-Col4a1 mouse will open up new possibilities to study many different aspects of BM biology, from developmental morphogenesis and wound healing to cancer metastasis. Moreover, several human diseases are linked specifically to mutations in collagen IV genes. Mutations in COL4A1 and COL4A2 have been linked to a wide range of human multisystem phenotypes, including ocular defects, porencephaly, cerebrovascular disease, and myopathy, collectively termed Gould’s syndrome (Jeanne et al., 2012, 2015; Gould et al., 2005, 2006; Favor et al., 2007).

We focused on the epidermal BM for this study due to its accessibility and amenability to live imaging and because little is known about epidermal BM dynamics during embryonic development. Using the novel mTurq2-Col4a1 mouse model combined with an innovative PS Multiview imaging approach, we have been able to long-term live-image the developing epidermal BM in embryonic mouse skin for the first time. Our findings reveal changes in epidermal BM morphology during hair follicle morphogenesis and suggest that the BM may be compressed by the forces generated by hair follicle invagination. We show that significant BM deformation also occurs during cell division, suggesting an inherent pliability of the epidermal BM during embryogenesis. Moreover, our observations that dividing basal progenitor cells remain adhered to the BM during mitosis challenge the existing assumption that all dividing cells significantly reduce their adhesions to the BM (Matus et al., 2014; Jones et al., 2019) and are in keeping with the integrin-rich contacts observed by Dix et al. (2018) during cell division. Our FRAP experiments provide a proof-of-principle that the mTurq2-Col4a1 mouse line can be used to study and quantify COL4A1 mobility during assembly and our data agree with other studies demonstrating the highly stable nature of collagen IV within BMs, particularly the recent observations made by Lartey et al. (2023). The Col4a1-P2A-eGFP mouse line generated in their work labels cells that are actively secreting COL4A1 with eGFP. Because collagen IV is so stable, they wanted to determine when endothelial cells (EC) actively synthesize COL4A1 during development, as the presence of COL4A1 in vascular BMs does not necessarily mean it is being actively secreted. They showed that secretion of COL4A1 by EC occurs in a temporally restricted manner during development, with eGFP expression (and thus COL4A1 secretion) being detected during the period E8.5–E16.5 but not by E18.5, or in any of the adult organs they observed (including skin). Although it is possible that this observation reflects an inherent limitation of the mouse model, where subsequent COL4A1 secretion by EC is beyond the limits of detection, it is consistent with the idea that COL4A1, once established in a BM, is inherently very stable.

Importantly, although we focused on the epidermal BM in this study, the mTurq2-COL4A1 signal was observed in every physiological BM observed, in both embryos and adult tissues. Therefore, we foresee that there is a great deal to learn about mammalian BM morphology, remodeling, and mechanics that can be achieved using our mTurq2-Col4a1 mouse.

A key limitation of this study is the homozygous lethality of the mTurq2-Col4a1 mouse. Although we conducted several experiments to confirm that the BMs in mTurq2-Col4a1/+ animals are indistinguishable from WT littermates, we have not been able to explain the lethality of the homozygous genotype. For studies of collagen IV protein biochemistry, therefore, the mTurq2-Col4a1 mouse may not be the best suited model. One possible explanation for the absence of homozygous animals is that heterotrimers containing two mTurq2-COL4A1 molecules (collagen α1/α1/α2 [IV]) may cause problems with secretion; mutant COL4A1 or COL4A2 has been shown to accumulate within cells, reducing secretion and leading to extracellular deficiency (Jeanne et al., 2015). Inhibited secretion was also observed with C-terminal tagged human laminin β-1 in adenocarcinoma cells (Shaw et al., 2020). Alternatively, given that the 7S domain is important for crosslinking collagen IV into a lattice network, it may be that two mTurq2 molecules per collagen IV trimer interfere with 7S-mediated crosslinking, but one mTurq2 per trimer can be tolerated. Nevertheless, as the mTurq2-COL4A1 signal is bright and photostable in heterozygous mTurq2-Col4a1/+ mice, this limitation does not impede the use of our mouse line to visualize BM dynamics live, in vivo. Furthermore, the Lamb1-Dendra2 mouse, currently the only other mouse line in which a BM component is tagged at the endogenous level, is also largely homozygous lethal (Morgner et al., 2023). Finally, mTurquoise2 has an excitation/emission spectrum of 434/474 nm (Goedhart et al., 2012), necessitating the use of a 445-nm laser for optimal imaging results, which is often not included as part of the traditional laser confocal set-up. However, as the mTurquoise2 fluorescent protein is not excited by the 488 nm excitation wavelength for GFP, it can theoretically be used in combination with all three standard fluorescent protein combinations (i.e., GFP, mCherry/RFP, and far-red).

Using a novel mTurq2-Col4a1 mouse model combined with an innovative PS Multiview imaging technique, we have visualized previously unobserved dynamics of developing skin. Our findings reveal unique insights into the epidermal BM during mouse embryonic development and highlight its inherent pliability in the context of morphological change. Although the large degree of growth, morphogenetic movement and cell division occurring during the period studied in this work suggested a need for the collagen IV supramolecular network to be extensively degraded, our observations show that once the collagen IV network is established early in development, it is not visibly diminished to accomodate morphogenesis, which may be due to either an equilibrium of degradation and secretion or an inherent pliability requiring little degradation to accommodate morphogenetic movements. Even during periods of rapid epithelial proliferation and morphogenetic change, mTurq2-COL4A1 within the epidermal BM is remarkably stable, showing minimal mobility. These data suggest that the observed pliability of the structure is a physical attribute of a relatively stable but flexible meshwork. Given the close association between increases in ECM stiffness, aging, cancer metastasis, and disease progression, our data raise questions about the behavioral changes cells might exhibit on a more rigid BM that can no longer yield to cellular forces. With the mTurq2-Col4a1 mouse model in hand, these questions are now feasible to address.

The mTurq2-Col4a1 mouse line was generated by knocking-in the coding sequence of mTurquoise2 after the START codon and signal sequence of the endogenous Col4a1 genomic locus to create an N-terminal fusion reporter using the previously described 2C-HR CRISPR method (Gu et al., 2020, 2018). Briefly, an HR repair plasmid was constructed by amplifying homology arms 1,000 bp upstream and downstream of the mTurquoise2 insertion site by PCR from genomic DNA obtained from murine CD1 keratinocytes and inserted using InFusion cloning (Takara) into the mTurquoise2-N1 plasmid (Addgene plasmid #54843; http://n2t.net/addgene:54843; RRID:Addgene_54843, a gift from Michael Davidson [Florida State University, Tallahassee, FL] and Dorus Gadella [University of Amsterdam, Netherlands]). Insertion was designed to remove the STOP codon at the 3′ end of the mTurquoise2 ORF. Site-directed mutagenesis was then performed to silently mutate the PAM sites using NEB Q5 Site-Directed Mutagenesis kit (E0554) according to the manufacturer’s instructions. sgRNAs were designed using CRISPOR (Concordet and Haeussler, 2018) and two were selected based on high predicted cutting efficiencies (sgRNA1 and sgRNA2). Synthetic sgRNAs were then ordered from Synthego with chemical modifications at the 5′ and 3′ termini: 2′-O-methyl modified bases and 3′ phosphorothioate linkages; for sequences see Table S1; sgRNAs were provided in donor vectors by Synthego. CD1-IGS mice (Charles River strain 022) were used as embryo donors. Briefly, female CD1-IGS were super-ovulated at 5–7 wk of age using 7.5 IU PMSG (Biovendor) administered by IP injection followed by 7.5 IU HCG (Sigma-Aldrich) by IP injection 47 h post PMSG. Superovulated females were mated to CD1-IGS stud males and checked for copulatory plugs the following morning. Cytoplasmic microinjection of two-cell embryos was performed as previously described (Gu et al., 2020, 2018). Briefly, embryos were harvested at the two-cell stage at E1.5 by flushing the oviducts with M2 Media (Cytospring), and each cell was microinjected with 100 ng/μl Cas9 mRNA (made by in vitro transcription [mMESSAGE mMACHINE SP6 transcription kit; Thermo Fisher Scientific] using Addgene plasmid 122948), 30 ng/μl donor plasmid, and 50 ng/μl sgRNA using a Leica Dmi8 inverted epifluorescent microscope, an Eppendorf Femtojet, and a Micro-ePore (WPI). Embryos were immediately transferred into the oviducts of pseudo-pregnant female CD1 mice. CD1 mice were used to generate the line as they tend to produce larger litter sizes than the C57BL/6J, thus increasing the probability of identifying a founder. Positive founders (N0) were identified by PCR genotyping as follows. gDNA from pup ear punches was prepared using Purelink Genomic DNA Mini Kit (Invitrogen) before being used as the template in two PCR reactions. Genotyping PCRs were designed to (i) only amplify a fragment if insertion had occurred at the correct locus and (ii) differentiate between hetero- and homozygotes by spanning the entire insertion area. See Table S1 for primer sequences. N0 founders were then crossed to C57BL/6J mice, and N1 mice from these crosses were genotyped as above using a proofreading polymerase (Q5 DNA Polymerase; NEB). Purified amplicons from these PCR reactions were then Sanger sequenced to ensure the accuracy of the insertion site and fidelity of the genomic sequence. N1 mTurq2-Col4a1 heterozygotes were then outcrossed to C57BL/6J mice for a minimum of five generations. Heterozygote incrosses from all generations did not yield any homozygous offspring, and litters were slightly smaller as a result but were otherwise Mendelian in ratio. All experiments were conducted in biological triplicate as a minimum, with technical replicates where appropriate. Embryos were obtained from pregnant dams at E11.5–E15.5 and all embryos of the correct genotype were used, without consideration of sex. Mice were housed in an AAALAC-accredited facility following the Guide for the Care and Use of Laboratory Animals. Animal maintenance and husbandry followed the Laboratory Animal Welfare Act. Princeton University’s Institutional Animal Care and Use Committee (IACUC) approved all animal procedures, and the three “R”s were considered in all experimental design.

The following mouse tissues were embedded fresh in OCT on dry ice and stored at −80°C: E18.5 embryos, dissected in PBS and heads removed; adult kidneys, whole; adult back skin, shaved using clippers to remove the hair and dissected into strips along the midline. Frozen OCT-embedded tissues were cryosectioned on a Leica CM3050 cryostat. Sagittal sections were cut at 10 µm thickness, collected on Superfrost Plus slides, and allowed to dry before being stored at −20°C. For immunostaining, sections were moved to a slide holder container and fixed in 4% paraformaldehyde (PFA) in PBS for 10 min at room temperature (RT), washed in PBS for 5 min, permeabilized in PBT2 (PBS with 0.2% Triton X-100) for 10 min at RT, and blocked in 2% normal goat serum, 2% normal donkey serum, and 2% bovine serum albumin (BSA) in PBT2 overnight at 4°C with gentle rocking. Slides were moved to a humidified chamber on a flat surface and incubated with a primary antibody in blocking medium for 1 h at RT; see Table S2 for primary antibodies and concentrations used. Slides were washed in slide holder containers three times for 10 min with PBT2 and incubated in a humidified chamber on a flat surface with secondary antibodies in PBT2 for 1 h at RT. Alexa Fluor-555 and -647 secondary antibodies were used at 1:2,000 (Invitrogen). Slides were washed in slide holder containers in PBT2 with Hoechst (1 µg/ml, H1399; Invitrogen) for 10 min, followed by two 10 min washes in PBT2. A final wash in PBS for 5 min was carried out before mounting in ProLong Gold (P36930; Invitrogen). Slides were stored at −20°C until imaged.

For immunofluorescence of E16.5, E15.5, and E12.5 backskins, embryos were dissected in PBS (+Mg, +Ca) and fixed in 4% PFA for 1 h (1.5 h for E16.5) at RT with gentle rocking. Backskins were then incubated in blocking medium overnight at 4°C with gentle rocking (1% BSA, 1% fish gelatin, 2% normal goat serum, and 2% normal donkey serum in PBT2 with 0.01% sodium azide). The following day, samples were incubated in primary antibody (in blocking medium) overnight at 4°C with gentle rocking. See Table S2 for primary antibody details and concentrations used. Samples were then washed a minimum of five times (30 min per wash) in PBT2 before incubating overnight in secondary antibodies (Alexa Fluor-555 and -647, 1:1,000; Invitrogen) in PBT2 at 4°C with gentle rocking in the dark. Following secondary incubation, samples were washed a minimum of five times in PBT2 (30 min per wash), with the first wash containing Hoechst (H1399, 1 µg/ml; Invitrogen), before being mounted in ProLong Gold (P36930; Invitrogen).

Images of E15.5 and E12.5 backskins were acquired on an inverted Nikon A1R confocal microscope controlled by NIS Elements software. Objectives used were Plan Apo 10/0.45NA air, Plan Apo 20/0.75NA air, and Plan Apo 60/1.4NA oil immersion (Nikon). For XZ optical reconstructions, resonance imaging was used with 8× averaging and images denoised in NIS Elements. NIS Elements and FIJI were used for all image processing. Images of E16.5 backskins were acquired on a Nikon inverted CSU-W1 SoRa Spinning Disk microscope with a Plan Apo 60× Oil/1.4NA immersion objective, controlled by NIS Elements software. Images were processed using NIS Elements and FIJI.

Images were acquired as described above using a Plan Apo 60/1.4NA oil immersion objective with 6× zoom to obtain a pixel size of 0.09 µm and an optical resolution of 0.26 µm (to attain Nyquist). Z stacks were obtained with 0.3 µm step size over a range of 3 µm with the center of the stack being set at the epidermal BM. 10 individual Z-stacks were obtained per embryo (n = 3). To calculate Pearson’s Correlation Coefficient, the Z slice at which the epidermal BM was most in focus was selected, background-subtracted, and the FIJI plugin Coloc 2 (Costes threshold regression; Costes randomizations = 100) was used on a region of interest (ROI) around the area of laminin β-1 signal (avoiding areas devoid of signal to reduce noise). Both biological (n = 3) and technical (n = 30) replicates were plotted and statistical analysis was undertaken using GraphPad Prism.

Overlap fraction analysis was undertaken using a custom macro in FIJI. Single Z planes in both channels as described above were thresholded using the Otsu method to generate a mask. The overlap map was then thresholded (Otsu) and an ROI was generated around the entire overlap region. The area of the overlap region was then divided by the area of the mask for each channel to calculate the overlap fraction. Both biological (n = 3) and technical (n = 30) replicates were plotted and statistical analysis was undertaken using GraphPad Prism.

E18.5 embryos were harvested from the euthanized dam and backskins were immediately dissected into fresh TEM fixative (2% glutaraldehyde, 4% formaldehyde, 2 mM CaCl₂ in 0.1 M sodium cacodylate buffer, pH 7.2). Samples were gently rocked at RT for 1 h before storage at 4°C until processed for imaging. Samples were postfixed in 1% osmium tetroxide and processed for Epon embedding. Ultrathin sections (60–65 nm) were counterstained with uranyl acetate and lead citrate. Images were taken with a transmission electron microscope (Tecnai G2-12; FEI) equipped with a digital camera (AMT BioSprint29).

A stencil to create a ridge in the agarose gel was created using Conair bobby pins and a standard glass microscope slide (Fisherbrand). Three bobby pins were cut at the apex, and then the flat arms from each pin were super glued side-on to a microscope slide. This created a flat surface with a large bar in the center measuring 3 mm wide, 25 mm long, and 1 mm tall. F-medium agarose plates were then made as previously described (Cetera et al., 2018) with the following additional steps. After briefly cooling to 60°C, the liquid agarose medium was poured into a 35-mm dish plate at 4°C. After 2 min had elapsed, the bobby pin glass stencil was dropped onto the solidifying gel, allowing it to float on top. After 15 min at RT, the bobby pin glass stencil was carefully removed from the gel using forceps. The ridged gel was then stored in a 37°C, 5% CO₂ cell-culture incubator while dissections took place. E14.5 and E15.5 flank skin explants were dissected in PBS (+Mg, +Ca), as previously described (Cetera et al., 2018), and transferred to the agarose-F-medium gel with the ridge. The explants were mounted dermal side down with ∼50% of the lateral flank region being placed over the edge of the ridge and the midline region was mounted firmly on the flat portion of the gel to prevent slipping into the ridge. Explant gels were then mounted into Lumox membrane dishes (Sarstedt) as previously described. Z-stacks with 3-µm step sizes were then acquired over several XY positions at 7-min intervals for 5–15 h using a Nikon inverted CSU-W1 SoRa Spinning Disk microscope with a Plan Apo Lambda S 40XC Sil/1.25NA immersion objective, controlled by NIS Elements software. Sample temperature was maintained during imaging using a Tokai Hit humidified stage top incubator and an environmental imaging chamber, both set to 37°C with 5% CO₂. Throughout all imaging experiments, areas that were close to the edge of the explants were not imaged to avoid any wound-healing response. NIS Elements and FIJI were used for movie processing.

The mTurq2-COL4A1 signal intensity was determined during PS Multiview imaging using FIJI software. An ROI was drawn both around the mTurq2-COL4A1 signal at the base of the hair follicle and around the interfollicular epidermis (IFE) region in view. If the hair follicle was in the center of the image, two IFE ROIs were drawn on either side of the hair follicle, and then the mean values were obtained. The average intensity of the mTurq2-COL4A1 signal in both hair follicle and IFE ROIs was then measured, and the ratio between the two was then calculated for each time point across the different movies (n = 8 movies). The two stages of hair follicle development (placode and hair germ) were determined by measuring the depth of the hair follicle relative to the IFE at every time point across all movies and then binning the different depths into two equal groups, representing the two developmental stages. The background IFE/IFE ratio was determined by taking each individual IFE mTurq2-COL4A1 signal intensity measurement and dividing it by the average IFE mTurq2-COL4A1 signal for the corresponding movie. Graphing and statistical analysis were undertaken using GraphPad Prism.

To measure BM deformation occurring during cell division, we identified dividing basal progenitor cells and measured the angle between the dividing cell’s most basal contact with that of both of its closest neighbors. The center of the basal surface of the dividing cell and the two neighboring cells, together with the basal most intersection of the two neighboring cells, were used as the points from which to measure the angles. The angles were measured at three timepoints: 18 min before the onset of division when the cell is in interphase (columnar), during metaphase when the cell rounds up, and 18 min after the daughter cells separate. The mean of both angles at each time point was calculated and plotted, n = 23 dividing cells that were followed through division. Graphing and statistical analysis were undertaken using GraphPad Prism.

BMs in live skin explants from E13.5 or E15.5 mTurq2-Col4a1/+ embryos were imaged using a Plan Apo 20/0.45NA air objective (Nikon) with appropriate optical zoom on an inverted Nikon A1R laser scanning confocal microscope equipped with a stage-top Tokai Hit incubation chamber to maintain 37°C and 5% CO₂. 1.5 µm diameter circular ROIs were bleached using a 445-nm laser, and recovery was monitored for a period of 10 min with 10-s intervals from either IFE or along the rim of developing hair follicles. Three reference prebleach images were also acquired. Magnification, laser power (both for bleach and acquisition), pixel dwell time, and acquisition rate were kept uniform across all measurements. The acquired images and measured ROIs in the time series were checked for Z-drift and corrected for the presence of any XY drift. A reference ROI was made in a non-bleached region to correct for overall bleaching during image acquisition. Background autofluorescence was measured from non-fluorescent WT skin using the same image acquisition conditions. The ROI values were extracted from drift-corrected images in NIS Elements software and subsequently processed in Microsoft Excel and GraphPad Prism. Each image time series was background/autofluorescence subtracted and bleach corrected (to be referred to as corrected intensity henceforth) and thereafter the corrected intensity profile was normalized as (F_t − F_bleach)/ (F_ini − F_bleach), where F_t is the corrected intensity of the ROI at a given time point, F_bleach is the corrected intensity at the time point immediately after bleaching, and F_ini is the mean ROI intensity of the three pre-bleach frames. Each mean recovery curve was fitted to an exponential one-phase association equation in GraphPad Prism with an r-squared value >0.9 to determine the fitted plateau and Y₀ values, which were then used to determine the immobile fraction = 1- [(Plateau-Y₀)/(1- Y₀)]. Data represented are pooled from three biological replicates.

Fig. S1, related to Fig. 1, shows an additional experimental replicate of mTurq2-COL4A1 labeling in adult mouse skin and kidney. Fig. S2, related to Fig. 3, shows TEM images from additional experimental replicates of E18.5 WT and mTurq2-Col4a1/+ embryonic skin. Fig. S3, related to Fig. 5, shows additional examples of snapshots taken from time-lapse imaging of BM curvature around dividing basal progenitor cells in mTurq2-Col4a1/+;mTmG/mTmG backskins, and quantification of the average angle of deformation and maintained hemidesmosome adhesions in dividing cells by integrin β4 antibody labeling and TEM. Video 1 shows three representative time-lapse live-imaging movies of hair follicle invagination in E14.5-E15.5 mTurq2-Col4a1/+;mTmG/mTmG backskins. Video 2 shows three representative time-lapse live-imaging movies of basal progenitor cell divisions in E14.5–E15.5 mTurq2-Col4a1/+;mTmG/mTmG backskins. Video 3 shows a Z-stack movie in planar view through a fixed E15.5 mTurq2-Col4a1/+ backskin from the basal layer (basement membrane) through to the dermis. Table S1 details sgRNA and oligonucleotide sequences used in this study; Table S2 details the primary antibodies used.

The authors thank the staff of the Princeton Laboratory Animal Resources (LAR) and the Princeton Core Imaging Facility (CIF), particularly Dr. Gary S. Laevsky and Dr. Sha Wang. Many useful discussions were had with Professor Doug Gould (UCSF), Dr. Jessica Morgner (Netherlands Cancer Institute), and Professor Dorus Gadella (University of Amsterdam), and the authors extend our heartfelt thanks.

This work was supported by National Institutes of Health (NIH)/Eunice Kennedy Shriver National Institute of Child Health and Human Development grant R01HD105009 (D. Devenport), NIH/National Institute of Arthritis and Musculoskeletal and Skin Diseases grant R01AR068320 (D. Devenport), NIH National Institute of General Medical Sciences T32GM007388 (B. Trejo), the New Jersey Alliance for Clinical and Translational Science (D. Devenport, R.A. Jones, B. Trejo), and the Ludwig Institute for Cancer Research (D. Devenport and R.A. Jones).

Author contributions: R.A. Jones and D. Devenport conceptualized the study; R.A. Jones designed and performed most experiments. B. Trejo designed the PS Multiview imaging methodology and performed live-imaging experiments with R.A. Jones. Samples prepared for EM by R.A. Jones were further prepared and EM-imaged by H.A. Pasolli. The mouse model was generated with the help of B. Joyce and E. Posfai. P. Sil performed FRAP experiments and analysis with R.A. Jones; K.A. Little also performed experiments. R.A. Jones and D. Devenport wrote the manuscript with all authors having read it and provided comments.