A multifunction murine Col4a1 allele reveals potential gene therapy parameters for Gould syndrome

Basement membranes (BMs) are specialized extracellular matrix (ECM) structures that are conserved among metazoans (Yurchenco, 2011). In mammals, these sheet-like structures line endothelial and epithelial basal surfaces and surround fat, muscle, and Schwann cells. In addition to their mechanical properties, BMs can serve as a reservoir for signaling molecules and regulate cell behaviors directly or indirectly through biophysical or biochemical mechanisms (Jayadev and Sherwood, 2017; Pozzi et al., 2017). The core BM components consist of type IV collagens, laminins, nidogens, and perlecan; however, the specific composition and architecture can differ across time and among tissues and can be modulated by physiological and pathological processes (Jayadev et al., 2022; Keeley et al., 2020; Khalilgharibi and Mao, 2021; Nyström et al., 2017; Sekiguchi and Yamada, 2018). Current methods are typically restricted to static descriptions of BM composition (presence and/or relative abundance of various proteins), and the functional relevance of this spatiotemporal heterogeneity is poorly understood owing, at least in part, to the lack of appropriate resources to investigate spatiotemporal BM dynamics in vivo.

Type IV collagens are encoded by six genes (Col4a1–Col4a6) that make at least three distinct heterotrimers: α1α1α2(IV), α3α4α5(IV), and α5α5α6(IV) (Khoshnoodi et al., 2008). The collagen α3α4α5(IV) and α5α5α6(IV) networks have relatively restricted distributions, and mutations in COL4A3, COL4A4, and COL4A5 cause Alport syndrome, which primarily affects the glomerular and cochlear BMs. In contrast, the collagen α1α1α2(IV) network is ubiquitous during development and present in nearly all mature BMs. Consistent with this, mutations in COL4A1 and COL4A2 cause a rare congenital multisystem disorder referred to as Gould syndrome (Boyce et al., 2021; Mao et al., 2021). Clinical manifestations of Gould syndrome are highly heterogeneous and often include cerebrovascular, ocular, muscular, and renal defects (Gasparini et al., 2024; Gould et al., 2005; Plaisier et al., 2007; Rannikmäe et al., 2020; Whittaker et al., 2022).

The mature COL4A1 and COL4A2 proteins consist of a long triple helical domain flanked by a 7S domain at the amino terminus and a non-collagenous domain (NC1) at the carboxy terminus responsible for interchain crosslinking and trimer recognition and assembly, respectively. The triple helical domain is characterized by Gly–Xaa–Yaa repeats, and glycine missense mutations are the most common type of pathogenic variant (Jeanne and Gould, 2017; Meuwissen et al., 2015; Rannikmäe et al., 2020; Whittaker et al., 2022). The molecular consequences of COL4A1 and COL4A2 mutations are complex, nuanced, and not fully understood, and can depend on which gene is mutated, the nature of the mutation, where it occurs along the amino–carboxy axis, and whether or how efficiently the mutant proteins are secreted. However, the general paradigm is that heterozygous glycine missense mutations impair collagen α1α1α2(IV) secretion via a dominant-interfering mechanism leading to extracellular deficiency. Mouse models with glycine missense mutations (herein collectively called Col4a1^+/mut) faithfully recapitulate Gould syndrome hallmarks and serve as useful tools to study disease pathophysiology and develop treatment interventions (Chen et al., 2016; Favor et al., 2007; Gould et al., 2005, 2006; Labelle-Dumais et al., 2011, 2019; Mao et al., 2024; Ratelade et al., 2018; Thaung et al., 2002; Van Agtmael et al., 2005). Notably, we have shown that TGFβ signaling is elevated in Col4a1^+/mut mice and that reducing TGFβ signaling partially alleviated ocular, cerebrovascular, and skeletal abnormalities (Branyan et al., 2023; Labelle-Dumais et al., 2024; Mao et al., 2022, 2024; Yamasaki et al., 2023a). However, the rescuing effects were incomplete suggesting that additional mechanisms may be involved and interventions targeting more proximal insults (intracellular) may be required to simultaneously address diverse distal insults (extracellular). To this point, we and others have shown that promoting collagen α1α1α2(IV) secretion using 4-phenylbutyric acid (a small molecule with chemical chaperone properties) ameliorated defects across multiple tissues in Col4a1^+/mut mice (Branyan et al., 2023; Hayashi et al., 2018; Jeanne and Gould, 2017; Jones et al., 2019; Labelle-Dumais et al., 2019; Mao et al., 2022). However, promoting secretion of mutant collagen α1α1α2(IV) can also exacerbate certain pathology (Kuo et al., 2014; Labelle-Dumais et al., 2019; Mao et al., 2015a), suggesting that secreted mutant collagen α1α1α2(IV) can also change the biochemical or biophysical BM properties.

Base editing approaches leverage the power of CRISPR/Cas9 genome editing tools and are attractive interventional candidates to address the root cause of genetic diseases. However, there are collectively ∼900 glycine residues in COL4A1 and COL4A2, and when considering all potential codon changes at the nucleotide level, “n of 1” base editing therapies become impractical. Gene editing approaches that effectively knockout (induce frameshifts or introduce stop codons) or knockdown (antisense oligonucleotides) the mutant allele may have broader applicability and earlier adoption than base editing. However, the biological consequences of deleting one COL4A1 or COL4A2 allele are not clear. Loss-of-function variants for both COL4A1 and COL4A2 are significantly under-represented in the general population, suggesting that they are selected against (https://gnomad.broadinstitute.org). Furthermore, although rare, mutations leading to presumed loss-of-function (frameshift or nonsense, in contrast to dominant interfering glycine missense variants) have been reported in individuals with Gould syndrome, indicating that they can be pathogenic (Lemmens et al., 2013; Meuwissen et al., 2015; Singh et al., 2024; Verbeek et al., 2012; Yoneda et al., 2013; Zagaglia et al., 2018). In contrast, in murine studies, mice homozygous for a Col4a1/a2 double knockout (Col4a1^−/−;Col4a2^−/−) die during mid-gestation (Pöschl et al., 2004) whereas heterozygous Col4a1/a2 double knockout mice may have only subtle abnormalities under challenged conditions (Pöschl et al., 2004; Steffensen et al., 2021). Similarly, although both are lethal, homozygous Col4a2^−/− embryos survived longer than embryos homozygous for glycine missense mutations (Col4a2^mut/mut) (Favor et al., 2007; Reissig et al., 2019). Moreover, Col4a2^+/− mice have much milder and later-onset cerebral and renal defects compared with Col4a2^+/mut mice (McNeilly et al., 2024). Collectively, these preclinical data suggest that heterozygosity for a null allele may be preferable to the presence of a dominant interfering mutation. If so, allele-specific knockout or knockdown may be beneficial and offer therapeutic potential.

Regardless of the specific approach, the therapeutic window for eventual gene editing interventions needs to be determined, and thus, understanding the rate of collagen deposition and turnover across different tissues, throughout lifespan, and in pathological settings is critical. Collagens are among the most common and longest-lived proteins in the body, and many characteristics including insolubility, extensive crosslinking, and heterogeneous cellular sources have contributed to a gap in our understanding of the rates of collagen deposition and turnover. Recent studies indicate that type IV collagens in developing tissues have rapid turnover rates (Keeley et al., 2020; Matsubayashi et al., 2020; Wuergezhen et al., 2025) and that COL4A1 production may drop significantly in postnatal mice (Lartey et al., 2023). In humans, fibrillar collagens have half-lives on the order of decades (15 and 117 years for skin and cartilage, respectively) (Verzijl et al., 2000), and BM components including type IV collagens can be stable for months in adult tissues (Decaris et al., 2014; Haddad and Bennett, 1988; Liu et al., 2019, 2020; Price and Spiro, 1977; Trier et al., 1990). Taken together, these results support a need for early therapeutic intervention. Consistent with this, the efficacy of promoting collagen α1α1α2(IV) secretion with chemical chaperones was greatest with an earlier intervention (Hayashi et al., 2018). However, the extent to which collagen α1α1α2(IV) turns over in BMs of the postnatal cerebral vasculature—a parameter that is critical for targeting one of the most clinically consequential aspects of Gould syndrome—remains unknown.

To address these key gaps in knowledge, we developed a mouse model in which the endogenous Col4a1 locus was edited to fuse COL4A1 with affinity (FLAG or hemagglutinin [HA]) and fluorescent tags to facilitate in vivo biochemical and biophysical studies and BM visualization, respectively. The construct enabled the generation of two independent lines of reporter mice expressing either 3xFLAG/mCherry or 3xHA/eGFP-tagged COL4A1 isoforms. We arranged LoxP sites to enable Cre-dependent isoform-switching in vivo, which allows us to distinguish collagen α1α1α2(IV) deposited from distinct cellular sources or at different times. Moreover, in the presence of Flp recombinase, the construct resolves into a knockout allele with an intracellular 3xFLAG/mCherry tag that marks Col4a1-expressing cells. We validated the key isoform-switching properties of this novel and versatile resource and demonstrated its utility for answering two fundamental questions for potential gene editing approaches for individuals with Gould syndrome. Given the ubiquity of collagen α1α1α2(IV) and the limited understanding of its biological functions, we expect this model will be a useful addition to the growing resources to study the biology of collagens and BMs.

To develop next-generation resources to study BMs in vivo, we used CRISPR/Cas9-mediated homologous recombination to genetically tag the endogenous Col4a1 locus with affinity tags for biochemical and biophysical analyses and with genetically encoded fluorescent proteins for in vivo visualization (Fig. 1). Tagging of BM proteins can affect protein biosynthesis, localization, or network formation, which can all impair BM structure or function (Futaki et al., 2023; Keeley et al., 2020; Shaw et al., 2020). Tagging the NC1, triple helical, or 7S domains risks interfering with heterotrimer initiation, triple helix formation, and network formation, respectively, and tagging the amino terminus would result in the removal of the tag when the signal peptide is cleaved. Notably, a Drosophila gene-trap screen successfully tagged Viking (the COL4A2 ortholog) by inserting GFP two amino acids after the signal peptide (Morin et al., 2001). Likewise, tagging of EMB-9 (the COL4A1 ortholog) in Caenorhabditis elegans at a similar position was also tolerated (Keeley et al., 2020). In murine COL4A1, a ∼45-kb intron separates the signal peptide (encoded by exon 1) from the rest of the coding sequences. We took advantage of this genomic arrangement to create a multifunctional allele (Col4a1^Tag1) that uses a gene-trap-like strategy with splice acceptor and donor sites to incorporate tags into the endogenous protein (Fig. 1 A). To this end, we designed a targeting construct containing 3xFLAG/mCherry sequences and 3xHA/eGFP sequences in opposite orientations and flanked by splice acceptor and donor sites, followed by a reversely oriented nuclear localization signal (NLS) and polyadenylation signal. The targeting construct also contains pairs of LoxP and Lox2272 sites and Frt and Frt14 sites that can be separately recognized by Cre and Flp recombinases, respectively. We inserted the construct into the first intron ∼260 nucleotides from exon 1 (Fig. 1 B). At baseline, the construct should generate a COL4A1 protein fused with the 3xFLAG/mCherry tag that is inserted one amino acid after the signal peptide (between K₂₈ and G₂₉) (3xFLAG/mCherry-COL4A1) (Fig. 1 C). Cre-mediated recombination is predicted to induce isoform switching and generate an otherwise identical COL4A1 protein tagged with 3xHA/GFP (3xHA/eGFP-COL4A1), enabling spatiotemporal distinction of COL4A1 proteins deposited by different cell types or at different time points (Fig. 1 C). Flp-mediated recombination, on the other hand, is predicted to generate a fusion protein consisting of the signal peptide, 3xFLAG/mCherry tags, NLS, and polyadenylation sequence (3xFLAG/mCherry-NLS) that terminates translation leading to a null allele (Fig. 1 C). We predicted that the 3xFLAG/mCherry-NLS fusion would not label BMs but instead would label cells actively expressing Col4a1. Details of the recombinase-dependent rearrangements are shown in Fig. S1, A and B.

We generated heterozygous Col4a1^Tag1 mice and crossed them to the ubiquitous Actb^Cre line (Lewandoski et al., 1997). Progeny that had germline recombination was used to generate an independent 3xHA/eGFP-COL4A1 reporter line. For simplicity, we referred to the 3xFLAG/mCherry and 3xHA/eGFP reporter lines as Col4a1^mCherry and Col4a1^eGFP, respectively. Both strains were independently backcrossed to the C57BL/6J (B6) genetic background. Heterozygous Col4a1^+/mCherry and Col4a1^+/eGFP mice are grossly normal and fertile, without evidence of reduced lifespan (Table S1). Homozygous Col4a1^{mCherry/mCherry} and Col4a1^eGFP/eGFP mice had reduced viability and represented 11.8% and 2.1% of progeny from heterozygous intercrosses (expected 25%; Table S1). Surviving Col4a1^{mCherry/mCherry} animals are fertile and can live for at least 15 mo. In this study, we focused on heterozygous Col4a1^+/mCherry and Col4a1^+/eGFP mice. To validate the properties of each tagged isoform, we performed western blot analyses on cell lysates from cultured primary mouse embryonic fibroblasts (pMEFs) isolated from the reporter lines. We detected signals at the expected molecular weight using antibodies against COL4A1, FLAG, mCherry, HA, and eGFP and confirmed the presence of COL4A2 (Fig. 2 A). We performed similar analyses on conditioned media collected from cultured Col4a1^+/mCherry and Col4a1^+/eGFP pMEFs and confirmed that the 3xFLAG/mCherry-COL4A1 and 3xHA/eGFP-COL4A1 and their binding partner COL4A2 can be secreted, suggesting that the tagging strategy does not disrupt the biosynthesis of collagen α1α1α2(IV) heterotrimers. We next examined if we could detect fluorescent signals in vivo using each reporter line (Fig. 2, B–D). To this end, we examined sections from the ocular lens at postnatal day 7 (P7). The lens capsule is a prominent BM that is synthesized by a single layer of lens epithelial cells in the anterior region and by elongated lens fibers in the posterior region (Fig. 2 B) (Danysh and Duncan, 2009). The thick and flat morphology of the lens capsule and the polarized, cuboidal lens epithelium allows easy visualization and interpretation, and we observed a clearly delineated layer of mCherry or eGFP signals surrounding the lens (Fig. 2 C). Importantly, we did not detect intracellular mCherry, suggesting that the tagged COL4A1 proteins are efficiently secreted and deposited in BMs (Fig. 2 D).

We next tested the ability to induce isoform switching to visually distinguish BMs produced at different times. To this end, we first used a ubiquitously expressed tamoxifen-inducible Cre line (R26^CreERT) (Badea et al., 2003) and assessed the presence of different isoforms in cultured pMEFs and in the lens from Col4a1^Tag1 mice with and without Cre (R26^CreERT;Col4a1^+/Tag1 and R26⁺;Col4a1^+/Tag1, respectively) (Fig. 3). As expected, in the absence of Cre or tamoxifen, cells synthesized and secreted only 3xFLAG/mCherry-COL4A1. In contrast, in tamoxifen-treated R26^CreERT;Col4a1^+/Tag1 cells, 3xFLAG/mCherry-COL4A1 was completely replaced by 3xHA/eGFP-COL4A1 (Fig. 3 A). In the developing lens, the newly synthesized BMs are deposited immediately adjacent to lens cells in the inner layer of the capsule as the lens grows (Danysh and Duncan, 2009). Following tamoxifen injection at P1, P2, and P3, we observed an outer layer of red fluorescence signal surrounding an inner layer of green fluorescence signal in the anterior lens capsule from P7 R26^CreERT;Col4a1^+/Tag1 mice, representing COL4A1 deposited before and after tamoxifen administration, respectively (Fig. 3 B). The green layer was relatively thicker than the red layer, consistent with a fast turnover or growth of the lens capsule during this stage of development (Danysh and Duncan, 2009). Neither experiment revealed residual secretion of 3xFLAG/mCherry-COL4A1, indicating highly efficient recombination.

Finally, we tested if Flp-mediated recombination would generate a null allele and mCherry labeling of Col4a1-expressing cells (Fig. 3, C and D). To this end, we used a R26^Flpe (Farley et al., 2000) line with ubiquitous, constitutive (preimplantation) Flp expression to generate double heterozygous mice (R26^Flpe;Col4a1^+/Tag1) and littermate controls (R26⁺;Col4a1^+/Tag1) and evaluated the tagged isoforms in pMEFs and P7 lens. In the absence of Flp, the 3xFLAG/mCherry-COL4A1 and COL4A2 were detected both in the cell lysate and conditioned media, suggesting normal collagen α1α1α2(IV) biosynthesis. In contrast, in the presence of Flp, a smaller protein containing both FLAG and mCherry was detected only in the cell lysate (Fig. 3 C). Furthermore, we did not observe any fluorescent signal in the lens capsule from R26^Flpe;Col4a1^+/Tag1 mice as expected, and we detected intracellular mCherry signal in lens epithelial cells (Fig. 3 D). In contrast to our expectations, the mCherry signal was not nuclear but it nonetheless enables the identification of Col4a1-expressing cells.

Having validated the key features, we next used the Col4a1^Tag1 allele to evaluate parameters that are relevant when considering gene therapy for individuals with COL4A1 mutations. We focused primarily on the cerebral vasculature because it is the most common and clinically consequential aspect of Gould syndrome. We chose to use antibodies against mCherry and eGFP to maximize detection sensitivity compared with the endogenous fluorescent signals (Fig. S1, C–E). Antibodies against mCherry and eGFP labeled the cerebral vasculature in Col4a1^+/mCherry and Col4a1^+/eGFP mice, respectively, and colocalized with signals detected by an anti-COLIV antibody with broad reactivity, confirming that tagged COL4A1 isoforms correctly localized to the cerebrovascular BMs.

The majority of pathogenic COL4A1 and COL4A2 mutations are de novo, and individuals with Gould syndrome are often diagnosed postnatally. Therefore, the therapeutic window for potential gene editing will depend in part on the rates of postnatal collagen α1α1α2(IV) deposition. However, almost nothing is known about the kinetics of collagen α1α1α2(IV) turnover and how it might change with age. Therefore, we sought to establish whether there is postnatal turnover of collagen α1α1α2(IV) in the murine cerebral vasculature (Fig. 4). To this end, we used R26^CreERT;Col4a1^+/Tag1 mice to differentiate between collagen α1α1α2(IV) that was deposited before (mCherry) and after (eGFP) tamoxifen injection in newborn and young adult mice (Fig. 4 A). To evaluate collagen α1α1α2(IV) turnover in newborn pups, we injected tamoxifen at P1, P2, and P3 and harvested tissues at P7 (Fig. 4, B and C). We found that 6 days after the first tamoxifen injection (+TAM, 6d), cerebral blood vessels from R26^CreERT;Col4a1^+/Tag1 mice had much less mCherry signal compared with littermate controls that did not inherit Cre (Fig. 4, B and C). Consistent with this observation, cerebral vessels in R26^CreERT;Col4a1^+/Tag1 mice were broadly labeled with eGFP, with a signal intensity that was comparable with samples from age-matched Col4a1^+/eGFP mice (Fig. 4, B and C), suggesting a rapid turnover of the collagen α1α1α2(IV) network in the cerebral vasculature during the first postnatal week. To test whether the turnover rate is uniform or changes with age, we performed a similar experiment in young adult mice by injecting tamoxifen starting at P35 and examined fluorescence signal changes after 7 days (+TAM, 7d) (Fig. 4, D and E). In contrast to the early postnatal period, a strong mCherry signal remained in most blood vessels in R26^CreERT;Col4a1^+/Tag1 mice, with eGFP labeling being only weak and sporadic (Fig. 4, D and E). To test whether turnover is present but slower in young adult mice, we generated another cohort of mice that were harvested 14 days after tamoxifen injection (+TAM, 14d). In this case, we observed more blood vessel segments labeled with eGFP, yet the intensity and coverage were not comparable to that observed in age-matched Col4a1^+/eGFP mice (Fig. 4, D and E). Taken together, our results demonstrate that there is continued collagen α1α1α2(IV) deposition in young adult cerebrovascular BMs; however, the turnover rate is slower than in the early postnatal period. To our knowledge, this is the first description of turnover dynamics for a cerebrovascular BM component, which may help inform the selection of a postnatal therapeutic window for gene therapy interventions.

Next, to determine whether Col4a1 is haplosufficient or haploinsufficient, we aged R26^Flpe;Col4a1^+/Tag1 (hereafter referred to as Col4a1^+/−) mice for 6 mo and tested for the presence of cerebrovascular, ocular, muscular, and renal pathologies associated with Gould syndrome. We first sought to validate Flp-mediated knockout using lung tissue. We found that Col4a1 expression was reduced to ∼50% of control R26^Flpe;Col4a1^+/+ mice (Fig. S2A) and that COL4A1 protein levels in lung lysates from Col4a1^+/− mice were reduced to ∼40% of controls (Fig. S2B). In contrast, the Col4a2 mRNA and COL4A2 protein levels were comparable between the two groups (Fig. S2, A and B). Next, we sought to evaluate Flp-mediated recombination in the cerebral vasculature (Fig. S2, C and D). We showed that in Col4a1^+/Tag1 mice without Flp, the mCherry signal localized to blood vessel BMs labeled by the anti-COLIV antibody. In contrast, the mCherry signal was almost completely eliminated from blood vessels in the brains from Col4a1^+/− mice, suggesting that the recombination efficiency is nearly complete (Fig. S2 C). As expected, we also observed intracellular mCherry signals from Col4a1-expressing cells (Fig. S2 D). Taken together, these results strongly support the successful deletion of one copy of Col4a1.

To maximize the possibility of detecting phenotypes, we first challenged the mice with treadmill exercise that exacerbates cerebrovascular and muscular defects (Hayashi et al., 2018; Jeanne et al., 2015; Labelle-Dumais et al., 2011, 2019). In Col4a1^+/mut mouse models of Gould syndrome, cSVD manifests as intracerebral hemorrhages (ICHs) that can be detected using Prussian blue staining. Prussian blue staining typically reveals numerous prominent subcortical lesions in the basal ganglia and thalamus, and diffuse staining in periventricular white matter (Hayashi et al., 2018; Jeanne et al., 2015). In contrast to the highly penetrant and severe ICH phenotype observed in Col4a1^+/mut mice (Jeanne et al., 2015; Ratelade et al., 2018), we did not observe pathology in 9 out of 12 exercise-challenged Col4a1^+/− mice at 6 mo of age. In the remaining three mice, one had enlarged ventricles and two had faint and small periventricular Prussian blue staining but did not have any staining in the basal ganglia or thalamus (Fig. 5 A). No pathology was detected in control R26^Flpe;Col4a1^+/+ mice. Col4a1^+/mut mice can also have compromised blood–brain barrier (BBB) in the absence of ICH (Ratelade et al., 2018) that is detectable by extravasation of 3 kDa fluorescein-conjugated Dextran. Unlike mice with Col4a1 glycine missense mutations that frequently have large, multifocal fluorescent signals (Ratelade et al., 2018), we detected only two small foci of positive signal in one out of three brains from Col4a1^+/− mice injected with fluorescent Dextran (Fig. 5 B).

Ocular defects are also commonly reported in individuals with Gould syndrome, and Col4a1 mutant mice typically have severe ocular anterior segment dysgenesis—developmental defects affecting tissues in front of the vitreous (Gould et al., 2007; Kuo et al., 2014; Mao et al., 2015b, 2022, 2024). These defects are detectable by slit-lamp examination and are characterized by corneal opacities, enlarged pupils, pigment dispersion, enlarged and tortuous iris vasculature, cataracts, and increased anterior chamber depth, and some of these features worsen with age (Gould et al., 2007; Kuo et al., 2014; Mao et al., 2015b, 2022, 2024). Unlike Col4a1^+/mut mice (Kuo et al., 2014), most Col4a1^+/− eyes appeared normal even at 6 mo of age, with 8 out of 50 eyes examined having only subtle iris vessel abnormalities (Fig. S3 A). Anterior segment dysgenesis is a risk factor for glaucoma affecting the retinal ganglion cells and their axons that project to the brain via the optic nerve. Histologically, Col4a1^+/mut mice have early onset iridocorneal adhesions that lead to high intraocular pressure, retinal nerve fiber layer thinning and cell loss, and optic nerve head excavation, which are hallmarks of glaucoma pathology (Mao et al., 2015b, 2024). In contrast, Col4a1^+/− eyes have normal ocular drainage structures, and normal retinal and optic nerve head morphology (Fig. S3 B).

Myopathy and renal defects are less frequent phenotypes reported in individuals with Gould syndrome (Labelle-Dumais et al., 2011; Plaisier et al., 2007). Col4a1^+/mut mice have an increased number of muscle fibers with non-peripheral nuclei (NPN), which is a hallmark of myopathy (Kuo et al., 2014; Labelle-Dumais et al., 2011, 2019). We quantified muscle fibers with NPN in quadriceps and found that both Col4a1^+/− and control R26^Flpe;Col4a1^+/+ mice have a similarly low percentage of NPN even with the treadmill exercise challenge that typically exacerbates this phenotype (Fig. S4). Histological renal defects in adult Col4a1^+/mut mice typically involve cystic glomeruli characterized by enlarged Bowman’s spaces and retracted capillary tufts, proliferative and hypertrophic parietal epithelial cells underlying the Bowman’s capsule, and tissue atrophy in the medulla and papilla (Chen et al., 2016; Jones et al., 2016; Van Agtmael et al., 2005). In contrast, both Col4a1^+/− and R26^Flpe;Col4a1^+/+ kidneys appeared normal (Fig. S5). Collectively, our results indicate that Col4a1 loss of function can be pathogenic, but that penetrance is low, and phenotypes are mild compared with the pathology caused by glycine missense mutations. These data suggest that allele-specific knockout or knockdown could be beneficial for individuals with Gould syndrome caused by glycine missense variants—the most common class of pathogenic mutations.

Here, we developed and validated a multifunctional mouse model with Cre-dependent isoform switching and Flp-dependent knockout potential to facilitate in vivo studies of BM biology. By using a ubiquitous constitutive Cre line, we generated two independent strains of COL4A1 reporter mice—3xFLAG/mCherry-COL4A1 and 3xHA/eGFP-COL4A1. We demonstrated that most Cre- and Flp-dependent functions worked as expected in vitro and in vivo. One aspect of the design—mCherry nuclear localization in Col4a1 expressing cells—did not behave as planned; however, the intracellular mCherry retention still enables the identification of Col4a1-producing cells. Using this novel model, we demonstrated postnatal collagen α1α1α2(IV) turnover in cerebrovascular BMs and determined that the turnover rate is slower in young adults compared with newborn mice. Furthermore, we show that pathology in Col4a1^+/− mice is both infrequent and mild, suggesting that Col4a1 deletion is minimally pathogenic in B6 mice and may be preferable to the presence of a glycine missense mutation that represents the majority of individuals with Gould syndrome.

Recently, complementary tools with varying degrees of similarity have been developed for type IV collagens and other BM components in invertebrates and vertebrates (Buszczak et al., 2007; Futaki et al., 2023; Jayadev et al., 2022; Keeley et al., 2020; Lartey et al., 2023; Matsubayashi et al., 2020; Morgner et al., 2023; Morin et al., 2001; Wuergezhen et al., 2025). In addition to the tagged COL4A1 and COL4A2 orthologs in Drosophila and C. elegans, murine laminin beta 1 (LAMB1), nidogen 1 (NID1), COL4A1, and COL4A2 have been tagged with fluorescent proteins. Laminins are more heterogeneous than type IV collagens (there are 16 laminin heterotrimers made of one each of five alpha, three beta, and three gamma chains), and LAMB1 tagged with Dendra2 enables labeling of many BMs (Morgner et al., 2023). Tagged NID1 labels BMs in multiple organs; however, this model also showed ectopic labeling and signal loss due to partial protein degradation (Futaki et al., 2023). There are also two recent related models for each of Col4a1 and Col4a2. The Col4a1–P2A–eGFP model was designed with a self-cleaving peptide followed by eGFP fused to the COL4A1 carboxy terminus. This results in intracellular cleavage and retention of eGFP to label Col4a1-expressing cells with normal secretion of the untagged COL4A1 protein (Lartey et al., 2023). In contrast, in the mTurq2-tagged COL4A1 model, the fluorophore labels the mature protein near the signal peptide/7S interface, similar to our Col4a1^mCherry and Col4a1^eGFP strains (Jones et al., 2024). Similarly, the Col4a2^eGFP and Col4a2^mKikGR mouse models also tag the mature COL4A2 protein near the signal peptide/7S interface (Wuergezhen et al., 2025). These latter three models have been used to advance insights into BM remodeling during hair follicle morphogenesis.

There are notable differences in homozygous viabilities of the various tagged strains that suggest chain stoichiometry, and/or choice of fluorescent protein may influence collagen IV biosynthesis or network formation. The three Col4a1-tagged models have reduced homozygous viability (Table S1 and (Jones et al., 2024)), whereas homozygous Col4a2^eGFP mice were reported to be normal (Wuergezhen et al., 2025). This may reflect differences in stoichiometry whereby collagen α1α1α2(IV) in homozygous Col4a2-tagged models only ever incorporates a single tagged alpha chain in contrast to the incorporation of two tagged alpha chains for collagen α1α1α2(IV) in homozygous Col4a1-tagged models. This is consistent with the normal viability of heterozygous Col4a1-tagged mice. It should also be noted that the eGFP tag in the Col4a2^eGFP mice was flanked by a short linker while the Col4a1 reporter models were not (Jones et al., 2024; Wuergezhen et al., 2025), which may also contribute to a more favorable protein/network structure. Furthermore, we found that homozygous Col4a1^mCherry and Col4a1^eGFP mice have different viabilities (Table S1)—an observation that strongly suggests that the choice of fluorescent tag influences protein or network structure. Despite reduced homozygous viability, heterozygous Col4a1^+/mCherry and Col4a1^+/eGFP mice are viable and fertile and are a valuable new resource for the study of BM biology.

The Col4a1^+/mCherry and Col4a1^+/eGFP lines that we have developed complement and extend these recently described models by providing additional choices of fluorescent tags and affinity tags, which give investigators flexibility for understanding collagen α1α1α2(IV) dynamics and function. More importantly, the recombinase-dependent isoform-switching features of the Col4a1^Tag1 mouse model provide additional advantages to unlock new aspects of BM biology. Instead of studying BMs as a monolith, the Col4a1^Tag1 mouse can allow molecular dissection of collagen α1α1α2(IV) with distinct cellular or temporal profiles when used in combination with spatiotemporally controlled Cre lines. BMs can be synthesized from multiple adjacent cellular sources, or by migrating or distant sources (Graham et al., 1997; Halfter et al., 2008; Matsubayashi et al., 2017; Pastor-Pareja and Xu, 2011; Tsutsui et al., 2021) that may change over time or in disease states (Morgner et al., 2023; Naba et al., 2014). Understanding if, when, and at what rate BM turns over in different tissues and the underlying cellular sources of BM dynamic changes holds biological interest and has implications for disease conditions and potential gene therapy approaches. Furthermore, it is possible that the biochemical and biophysical properties of the collagens from different cellular sources may vary (Cabral et al., 2014; Hudson et al., 2015; Ishikawa et al., 2021) and even confer distinct biological functions. We anticipate that cell type–specific tagged COL4A1 will facilitate studies investigating the role of BM dynamics in developmental, homeostatic, and pathological processes. These types of studies will be enhanced even further by the Flp-dependent functionality, which, in addition to generating a knockout allele, enables the identification of Col4a1-expressing cells visually or using FACS/scRNAseq approaches that can reveal novel or unexpected cell types. An important caveat is that mCherry-labeled cells come at the cost of Col4a1 deletion that may influence physiological processes, and thus the results should be independently validated.

Here, we used Col4a1^Tag1 mice to address two questions critical to potential gene editing strategies for individuals with Gould syndrome. First, we sought to determine the extent of collagen α1α1α2(IV) turnover in newborn and young adult mice. Unlike the rapid half-lives of many cytosolic proteins, collagens are generally thought to be relatively long-lived. Previous studies have shown that the turnover rate of type IV collagens from adult tissues can vary from weeks to months (Decaris et al., 2014; Liu et al., 2019, 2020). Moreover, a recent study using the Col4a1-P2A-eGFP reporter demonstrated Col4a1-expressing endothelial cells from mid to late gestation, but the signal was no longer detected after E18.5 (Lartey et al., 2023), raising the possibility that collagen α1α1α2(IV) synthesis in mice ceased at birth and that the embryonically synthesized collagen α1α1α2(IV) might be retained throughout life. By using a ubiquitously expressed inducible Cre (R26^CreERT) line, we activated Cre starting at P1 and found nearly complete turnover from the COL4A1^mCherry to COL4A1^eGFP isoforms by P7. These data demonstrate that there is continued Col4a1 expression at a relatively high rate in early postnatal life in the mouse. We think this discrepancy is likely due to different detection thresholds between the two models.

For many individuals with Gould syndrome, a genetic diagnosis comes in early childhood—although the age at diagnosis is gradually increasing as continually milder manifestations are being attributed to COL4A1 and COL4A2 mutations. To address whether appreciable levels of collagen α1α1α2(IV) turnover persist into early adulthood—which may reflect a therapeutic window for gene editing—we activated Cre at P35 and evaluated the relative levels of COL4A1^mCherry to COL4A1^eGFP isoforms at P42. Despite a similar number of elapsed days as the experiment in newborn mice, newly deposited collagen α1α1α2(IV) (COL4A1^eGFP) was barely detectable. When extended to 14 days, we detected more newly deposited collagen α1α1α2(IV); however, the distribution and intensity of “old” collagen α1α1α2(IV) were predominant. These data clearly demonstrate both ongoing postnatal collagen α1α1α2(IV) deposition in the cerebral vasculature and also that the turnover rate slows with age. Our results have implications for possible gene therapy for Gould syndrome. Although postnatal intervention cannot completely reverse prenatal damage, there is an opportunity for gene editing approaches to arrest or prevent further damage as the collagen α1α1α2(IV) network turns over. Consistent with our earlier studies to define an efficacious window for interventions that promote collagen α1α1α2(IV) secretion (Hayashi et al., 2018), these data suggest that the therapeutic window narrows with age and calls for early interventions.

Finally, we used Col4a1^+/− mice to test whether Col4a1 is haploinsufficient. We found that Col4a1^+/− mice have infrequent and mild pathology suggesting that a Col4a1 null allele is largely tolerated in this experimental setting. We found evidence for cerebrovascular pathology in only 3 out of 12 Col4a1^+/− mice at 6 mo that had been challenged with exercise to exacerbate this phenotype, subtle changes in iris vasculature in 8 out of 50 eyes, and no evidence for muscular or renal pathology. Importantly, the frequency and severity of pathology are markedly less compared with mice from an allelic series of glycine missense mutations that were maintained on the same genetic background (Jeanne et al., 2015; Kuo et al., 2014). We performed molecular and histological validation studies that demonstrated highly efficient recombination, suggesting that the infrequent and mild pathology is unlikely to be explained by failure to delete the Col4a1 allele. On one hand, it is possible that heterozygosity for a null Col4a1 allele is completely benign and that the concomitant intracellular retention of mCherry has toxic effects that contribute to this mild pathology. Excluding this possibility would require generating a “clean” knockout allele, and we believe it is more likely that the phenotypes we detected are indeed due to reduced collagen α1α1α2(IV). On the other hand, it is also possible that broader and deeper phenotyping analyses might reveal other pathologies in Col4a1^+/− mice. For example, we have recently demonstrated that Col4a1^+/mut mice have age-related cerebrovascular dysfunction (Thakore et al., 2023; Yamasaki et al., 2023a, 2023b), and it will be important to test for these outcomes in Col4a1^+/− mice. Moreover, outcomes for individuals with Gould syndrome are extremely variable and a genotype/phenotype position effect may exist whereby mutations nearer the carboxy terminus tend to be more severe (Jeanne et al., 2015; Kuo et al., 2014). Indeed, based on allele frequencies in gnomAD, glycine missense variants near the amino terminus appear to be more tolerated. Therefore, it may not be universally true that allele-specific knockout/knockdown would be preferable for all glycine missense mutations. A recent study showed similar results suggesting that Col4a2^+/− mice also have very mild pathology (McNeilly et al., 2024). Notably, the strongest data for tolerance of Col4a1 and Col4a2 loss of function comes from murine models (McNeilly et al., 2024; Pöschl et al., 2004; Steffensen et al., 2021), and conflicting results from mice and human genetic studies that identified pathogenic, loss of function variants, leave open the possibility that tolerance is species dependent.

Taken together, we have developed and validated Col4a1^Tag1 mice that we predict will be a valuable tool for the scientific community to understand spatiotemporal dynamics and functions of collagen α1α1α2(IV). Based on our current studies, we propose that further work is justified to explore whether allele-specific Col4a1 knockout or knockdown could help individuals with Gould syndrome.

All experiments were approved by the Institutional Animal Care and Use Committee at the University of California, San Francisco, CA, USA (Protocols AN182181 and AN196112) and the University of Iowa. All animals were maintained in full-barrier facilities free of pathogens on a 12-h light/dark cycle with ad libitum access to food and water. All strains were iteratively crossed to B6 mice for at least five generations, and both sexes were used for each genotype.

The Col4a1^Tag1 mouse model was generated at the University of Iowa Genome Editing Core Facility. A gene trap-like strategy was utilized to generate a targeting construct that was inserted into the first intron of the mouse Col4a1 locus by CRISPR/Cas9 targeting. The overall strategy is shown in Fig. 1. Briefly, the knock-in cassette mainly comprised 3xFLAG/mCherry and 3xHA/eGFP sequences, both flanked by splicing donor and acceptor sites, followed by a NLS and a polyadenylation site that only associates with a splicing acceptor. Splicing of the knock-in cassette results in the insertion of the tags one amino acid past the signal peptide (between K₂₈ and G₂₉). The cassette also contains pairs of LoxP, Lox2272, Frt, and Frt14 sites that enable isoform switch upon recombination. Homologous arms (415 bp and 517 bp) flanking the insertion site were cloned from B6 mice (The Jackson Laboratory) and were integrated with the main knock-in cassette into a pUC vector and linearized to generate the double-strand repair construct. Three guides (5′-GAT​GAG​GGT​TCA​AAT​CTC​AC-3′, 5′-AAG​CTT​AGC​GTG​TAC​CAC​GG-3′, 5′-CTG​AAG​CTT​AGC​GTG​TAC​CA-3′) were used to direct SpCas9 to generate double-strand breaks ∼260 bp from exon 1. A mixture of chemically modified CRISPR RNAs (crRNAs) and transactivating CRISPR RNAs (tracrRNAs) (Integrated DNA Technologies, 20 ng/μl each pair), Cas9 (Integrated DNA Technologies, 60 ng/μl), and double-strand repair DNA (10 ng/μl) were delivered to B6 zygotes using pronuclear microinjection (Miura et al., 2018). Zygotes were immediately implanted in pseudopregnant Hsd:ICR (Envigo) females, N0 founders were screened for the knock-in cassette by PCR, and the insertion was sequence verified. DNA fragments from founders were crossed with B6 mice to acquire germline transmission of the knock-in allele and the line is maintained on the B6 background. We crossed Col4a1^+/Tag1 mice to Actb^Cre mice (Lewandoski et al., 1997) and then to B6 to generate the Col4a1^+/eGFP line. We used the Rosa26^CreERT line (Badea et al., 2003) for ubiquitous inducible Cre expression and COL4A1 protein turnover analyses. To generate heterozygous Col4a1 null mice, we used the Rosa26^Flpe line which ubiquitously expresses Flp before implantation (Farley et al., 2000).

pMEFs were isolated from E14.5 embryos and cultured in DMEM (Dulbecco modified essential medium)/high glucose/pyruvate (11995065; Gibco) supplemented with 10% (vol/vol) fetal bovine serum (S11150; R&D Systems), Pen Strep glutamine 100× (10378016; Gibco), and 5 mM HEPES. These pMEFs were used within three passages for biochemical experiments. To activate Cre, pMEFs were treated with 25 µM 4-Hydroxytamoxifen Ready Made Solution (4-OHT, SML1666-1ML; Sigma-Aldrich) for 1 day before harvesting.

pMEFs were plated and grown to 80–90% confluency and the medium was replaced with fresh medium supplemented with 100 μg/ml ascorbic acid phosphate (013-12061; FUJIFILM Wako Chemicals) for 24 h to stimulate procollagen biosynthesis. The medium was then replaced with fresh DMEM with ascorbic acid, and the cells were cultured for an additional 24 h before media and cell lysates were collected. The medium was centrifuged to remove dead cells and the supernatant containing secreted proteins was collected. pMEFs were harvested and collected by centrifugation, and proteins from cell pellets were isolated using M-PER (Thermo Fisher Scientific) containing Halt Protease Inhibitor Cocktail, EDTA-Free (Thermo Fisher Scientific) at 4°C according to manufacturer’s instructions. Media and cell lysates were denatured in Bolt LDS sample buffer under reducing conditions for 5 min at 80°C and proteins were separated on Novex 6% Tris/Glycine SDS-PAGE gels (Thermo Fisher Scientific) and then transferred to PVDF membranes in 5 mM sodium tetraborate buffer with 0.05% SDS and incubated with primary and secondary antibodies. To detect β-actin and FLAG/mCherry-tagged isoforms in samples from R26^Flpe;Col4a1^+/Tag mice and corresponding controls, proteins were separated using Bolt Bis-Tris Plus 4–12% gels (Thermo Fisher Scientific) with MES running buffer. Blots were developed with HRP-enhanced Super-Signal West Pico Chemiluminescent Substrate (PI34580; Thermo Fisher Scientific) and detected by ChemiDoc MP imaging system (Bio-Rad) using the software Image Lab version 4.0.1 (Bio-Rad). Antibodies used in these analyses are listed in Table S2.

To activate Cre, mice were injected with tamoxifen (T5648; Sigma-Aldrich) solubilized in ethanol and diluted in corn oil (C8267; Sigma-Aldrich). For early postnatal activation, pups received one intragastric injection of tamoxifen (50 μg) daily for three consecutive days starting from P1. For activation after weaning age, mice received one intraperitoneal injection of tamoxifen (0.1 mg/g) daily for five consecutive days.

Mice were anesthetized and transcardially perfused with phosphate-buffered saline [pH 7.4] (PBS) with 100 mg/liter heparin (Thermo Fisher Scientific). Tissues were fixed by immersion in 4% paraformaldehyde (PFA) in PBS overnight at 4°C, cryoprotected in 15% and 30% sucrose/PBS, and embedded in optimal cutting temperature compound (OCT; Sakura Fineteck). Cryosections from eyes (12 μm) and brains (20 μm) were prepared using a Leica CM1950 cryostat. Sections were either counterstained with DAPI (1 μg/ml) before imaging or processed for indirect immunofluorescence analyses. After blocking in buffers containing 0.25% Triton X-100 in PBS (PBST), 10% normal donkey serum, and 1% BSA for 2 h at room temperature, sections were incubated with primary antibodies at 4°C overnight followed by species-specific Alexa Fluor 488- or 594- conjugated secondary antibodies (1:500; Thermo Fisher Scientific) for 2 h at room temperature, counterstained with DAPI (1 μg/ml), mounted in Prolong Gold Antifade Mountant (Thermo Fisher Scientific), and imaged using a Zeiss LSM900 confocal microscope equipped with Plan-Apochromat 20×/0.8 air, 40×/1.40 oil or 63×/1.40 oil immersion objectives and ZEN Blue software (v3.5; Carl Zeiss Microscopy). Lens overview images were captured using an Axiocam 712 mono camera and a Plan-Apochromat 10×/0.45 air objective. For anti-COLIV immunolabeling, sections were boiled in 10 mM citrate buffer pH 6.0 for 20 min for antigen retrieval before blocking. Antibodies: rabbit anti-GLUT1 (#ab15309, 1:500; Abcam), rabbit anti-mCherry (#RL600-401-379, 1:500; Rockland Immunochemical), goat anti-GFP (#ab5450, 1:250; Abcam), goat anti-COLIV (#1340-01, 1:500; SouthernBiotech), and rabbit anti-COLIV (1:250) (Pokidysheva et al., 2014). Fluorescence intensities were quantified using ImageJ (version 1.54). Six or eight images (897 × 896 µm²) were obtained at different regions from P7 or P42 brains, respectively. The integrated intensity of the mCherry or eGFP signal for each image was recorded and the average values from all six (or eight) images were used for each sample. Data were shown as relative values normalized to the average of the control group in each graph.

Lungs were dissected following transcardiac perfusion and immediately frozen on dry ice. Lungs were ground using mortar and pestle prechilled in liquid nitrogen. Total RNA was extracted using the RNeasy Plus Mini kit (Qiagen) and reverse-transcribed into cDNA using the iScript cDNA Synthesis kit (Bio-Rad) according to the manufacturer’s instructions. qPCR was performed using the SsoFast EvaGreen Supermix (Bio-Rad) on a Bio-Rad CFX96 Real-Time Detection System (Bio-Rad) as described previously (Mao et al., 2022). β-Actin (Actb) was used as a housekeeping gene. Primer sequences are as follows:

• Col4a1 5′-GCA​CCC​ATC​TCT​GGG​GAC​AA-3′ and 5′-TGG​TCT​GAC​TGT​GTA​CCG​CC-3′;

• Col4a2 5′-GAC​TCC​TGG​CTC​AGA​GCG​TCT​T-3′ and 5′-TAG​CAC​TGG​CAA​CCC​CCA​CT-3′;

• Actb 5′-CAC​TGT​CGA​GTC​GCG​TCC-3′ and 5′-GTC​ATC​CAT​GGC​GAA​CTG​GT-3′.

Lungs were ground in liquid nitrogen as mentioned above and lysed using T-PER (Thermo Fisher Scientific) supplemented with Halt Protease Inhibitor Cocktail, EDTA-free (Thermo Fisher Scientific) at a protein concentration of 0.2 mg/ml at 4°C. The soluble lung protein lysate was collected by centrifugation and denatured in Bolt LDS sample buffer under reducing conditions for 5 min at 80°C. Proteins were then separated on Tris/Acetate 3–8%, 6% Tris/Glycine, and 4–12% Bis/Tris SDS-PAGE gels (Thermo Fisher Scientific) for COL4A1, COL4A2, and β-actin, respectively, and analyzed by western blotting as described above in the “Protein analyses in cell lysate and conditioned culture medium” section.

To maximize the possibility of observing phenotypic changes (Hayashi et al., 2018; Labelle-Dumais et al., 2011), mice harvested for histological analyses were challenged with a single session of treadmill exercise 2 days prior to harvesting as described previously (Labelle-Dumais et al., 2011). The exercise included a 2-min acclimation period, followed by a 30-min exercise challenge with a 15° downhill grade on a treadmill equipped with a shock plate (Exer 3/6; Columbus Instruments). Animals were started at 7 m/min and increased by 3 m/min every 2 min until a maximum speed of 12 m/min was reached.

Mice were anesthetized and transcardially perfused with PBS prior to organ collection. Brains were fixed by immersion in 4% paraformaldehyde (PFA) overnight at 4°C and cryoprotected in 15–30% sucrose gradient series in PBS at 4°C, embedded in OCT compound (Sakura Fineteck), and flash-frozen using dry ice/ethanol bath. To test for the presence of intracerebral hemorrhage, 40 μm coronal cryosections collected at regular intervals of 160 µm along the rostrocaudal axis were stained with Perl’s Prussian Blue and counterstained with nuclear Fast red (Vector Lab) as described previously (Jeanne et al., 2015; Hayashi et al., 2018).

Mice at 1 mo of age were injected with 3 kDa lysine fixable, fluorescein-conjugated dextran (40 μg/g; Molecular Probes) through the femoral vein, perfused with PBS containing 100 mg/liter heparin 30 min after injection for dextran clearance, followed by 4% PFA in PBS for brain fixation. Brains were dissected and post-fixed in 4% PFA overnight, and 40-μm coronal serial sections were prepared as described above. Three to six sections sampled at 0, -1, and -2 mm relative to bregma were imaged for each brain. Sections were counterstained with DAPI and mounted with Mowiol or FluromountG (Invitrogen). Images were acquired with a Zeiss LSM900 equipped with an Axiocam 712 mono camera using an EC Plan-NeoFluar 5×/0.16 M27 air objective.

Ocular anterior segment examinations were performed on 6-mo-old mice using a slit-lamp biomicroscope (Topcon SL-D7; Topcon Medical Systems) attached to a digital SLR camera (Nikon D200; Nikon).

Ocular histological analyses were performed as described previously (Mao et al., 2024). Briefly, eyes were enucleated and fixed in half-strength Karnovsky fixative (2% paraformaldehyde and 2.5% glutaraldehyde) in 0.1 M phosphate buffer, pH 7.4 for 24–48 h at room temp and stored at 4°C. Eyes were dehydrated in a graded series of ethanol and embedded in Technovit 7100 methacrylate (Kulzer Technik). 2 µm sections were collected from the level of the optic nerve head and stained with Hematoxylin (Gills 3) and Eosin (H&E). Five to eight consecutive sections per eye were evaluated for ocular pathology.

Histological analyses of quadriceps were performed as described previously (Labelle-Dumais et al., 2011). Briefly, quadriceps were dissected and flash-frozen in liquid nitrogen-cooled isopentane. Cryosections (10 μm) were collected at regular intervals and stained with H&E. The number of NPN was evaluated on a total of 10–15 sections for each muscle, and the percentage of muscle fibers with NPN was quantified. A total of between 1,000 and 2,000 muscle fibers per animal were counted.

Kidneys were fixed in 10% neutral-buffered formalin for 24 h at room temperature and stored in 70% ethanol for no more than 1 mo. Kidneys were processed on an automated processor and embedded in paraffin wax. Sagittal sections (7 μm) were prepared on a Leica 2235 microtome and stained with H&E. Three to six consecutive sections from the center per kidney were evaluated for pathology.

Statistical analyses were performed using GraphPad Prism v10.0 (GraphPad). Statistical differences between two groups with equal variance were determined using a two-tailed unpaired Student’s t test. Statistical differences between the three groups were determined using one-way ANOVA with Tukey’s multiple comparison test. Prior to the indicated tests, Shapiro–Wilk test was used to test normality. The observed genotype ratios for heterozygote intercrosses were compared with expected ratios using the Chi-square calculator (https://www.graphpad.com/quickcalcs/chisquared1/).

Fig. S1, A and B, related to Fig. 1, shows details of recombinase-based rearrangements resulting in final products shown in Fig. 1 C. Fig. S1, C–E, related to Fig. 4, shows that antibodies against mCherry and eGFP label the cerebrovascular BMs and colocalize with COLIV labeling in Col4a1^+/mCherry and Col4a1^+/eGFP mice, and the endogenous mCherry and eGFP signals are detectable in the cerebral vasculature. Fig. S2, related to Fig. 5, shows high recombination efficiency using Flp in R26^Flpe;Col4a1^+/Tag1 mice. Fig. S3 shows that Col4a1^+/− mice have mild ocular phenotypes. Fig. S4 shows that the Col4a1^+/− mice have no muscle defects. Fig. S5 shows that the Col4a1^+/− mice have no renal defects. Table S1 shows the genotype frequency of P18 pups from Col4a1^+/mCherry × Col4a1^+/mCherry and Col4a1^+/eGFP × Col4a1^+/eGFP crosses. Table S2 is a list of antibodies used for Western blot analyses.

The data underlying all figures are available in the article and supplemental figures.

Tagged mice were generated at the University of Iowa Genome Editing Core Facility, directed by Dr. William Paradee and supported in part by grants from the NIH and from the Roy J. and Lucille A. Carver College of Medicine. We thank Norma Sinclair, Patricia Yarolem, Joanne Schwarting, and Rongbin Guan for their technical expertise in generating transgenic mice. We thank Drs. Corinna Cozzitorto and Dawiyat Massoudi for technical guidance in imaging and histology. We thank Lourdes Flores for their technical assistance in genotyping and colony management.

This work was supported by the National Institutes of Health (NIH) under Award Numbers 1R21NS133610 and 1RF1NS128217 (D.B. Gould) from the National Institue of Neurological Disease and Stroke and from All May See Foundation. The UCSF Department of Ophthalmology is supported by an NIH Core Grant from the National Eye Institute (P30 EY002162) and an unrestricted grant from Research to Prevent Blindness.

Author contributions: M. Mao: Conceptualization, Formal analysis, Investigation, Methodology, Project administration, Validation, Visualization, Writing - original draft, Writing - review & editing, Y. Ishikawa: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Validation, Visualization, Writing - review & editing, C. Labelle-Dumais: Conceptualization, Formal analysis, Investigation, Methodology, Validation, Writing - review & editing, X. Wang: Investigation, Y.-M. Kuo: Investigation, U.B. Gaffney: Investigation, M.E. Smith: Investigation, C.N. Abdala: Investigation, M.D. Lebedev: Investigation, W.J. Paradee: Conceptualization, Methodology, D.B. Gould: Conceptualization, Formal analysis, Funding acquisition, Methodology, Project administration, Supervision, Visualization, Writing - original draft, Writing - review & editing.