FOXD1 is required for 3D patterning of the kidney interstitial matrix

The interstitial extracellular matrix (ECM) is comprised of proteins and glycosaminoglycans and provides structural and biochemical information during development. Our previous work revealed the presence of transient ECM‐based structures in the interstitial matrix of developing kidneys. Stromal cells are the main contributors to interstitial ECM synthesis, and the transcription factor Forkhead Box D1 (Foxd1) is critical for stromal cell function. To investigate the role of Foxd1 in interstitial ECM patterning, we combined 3D imaging and proteomics to explore how the matrix changes in the murine developing kidney when Foxd1 is knocked out.


| INTRODUCTION
The extracellular matrix (ECM) is a network of proteins and glycosaminoglycans that provides structural support and is a reservoir for growth factors during nephrogenesis. 1,2 ECM components in the kidney can be classified as being part of either the basement membrane or interstitial matrix. 3 The basement membrane, a dense meshwork of network-forming proteins directly surrounding cells, is well-studied as it is critical for normal kidney function. 4 The interstitial matrix includes fibrillar and network ECM in a matrix of amorphous proteins, and is rich in collagens (COL), elastin, microfibrils, proteoglycans, and non-collagenous glycoproteins. 2 Throughout kidney development, the ECM is arranged in complex 3D structures with vertical fibers in the cortex and medullary ray sheath fibers at the corticomedullary junction. 3 Additionally, distinct ECM proteins are enriched at developing timepoints relative to adult, which may have unique functions. 3 Clinical correlates suggest the interstitial matrix is critical for kidney development because patients with connective tissue disorders have increased kidney cysts and congenital anomalies of the kidney and urinary tract. [5][6][7][8] Stromal cells are responsible for synthesizing the interstitial ECM, and are identified by the expression of various transcription factors, including Forkhead Box D1 (Foxd1). 9 During kidney development, stromal cells surround the ureteric bud and metanephric mesenchyme that eventually form the collecting duct system. The ureteric bud induces the surrounding cap mesenchyme to condense into the pre-tubular aggregates and epithelize to form the kidney vesicle, comma-shaped body, and Sshaped body. 10 The S-shaped bodies develop into the glomerulus, proximal tubule, loop of Henle, and distal tubule. With the invasion of endothelial cells and mesangial cells into the S-shaped body, glomerulogenesis occurs through the subsequent formation of loops, maturation of the basement membrane, and differentiation of cells in three sequential stages: capillary loop, maturing, and mature. 10,11 Stromal cells are critical for kidney development: when Foxd1 is knocked out, there is altered nephrogenesis with small, abnormally formed kidneys (kidney hypodysplasia). 12,13 In this model, stromal cells are retained, but have abnormal gene expression. 12,14,15 The hypodysplasia is in part due to the overexpression of the ECM protein decorin (DCN) and other interstitial matrix proteins, which inhibit the availability of the growth factor bone morphogenetic protein 7 (BMP7) and prevent the maturation of the cap mesenchyme. 14 However, the mechanisms by which the loss of Foxd1 modulates overall 3D ECM patterning are unknown.
To assess how the interstitial matrix in kidney development can vary in a murine model of kidney hypodysplasia (Foxd1 GC/GC ), we combined 3D imaging and quantitative proteomics for a descriptive characterization. Stromal cell abnormalities in Foxd1 GC/GC kidneys resulted in clustered vertical fiber patterning and fused glomeruli; however, medullary ray sheath patterning was unaffected. While interstitial fiber abnormalities did not precede ureteric bud branching dysmorphogenesis in Foxd1 GC/GC kidneys, the changes in interstitial ECM corresponded to disruptions in stromal cell patterning.
2 | RESULTS 2.1 | RNA encoding various ECM was enriched in different stromal cell types within the kidney Recent studies indicated that there is considerable heterogeneity of gene transcription in the stromal cell population of the kidney, including Foxd1 subtypes ( Figure 1A). 17,18 However, associated changes in the presence and distribution of ECM proteins were not assessed. ECM-specific transcripts from single-cell RNA (scRNA) sequencing data of E18.5 kidney stromal cells were analyzed, 17 which suggested that ECM were differentially expressed in various stromal compartments. For example, transcripts of the following were enriched in the respective compartments: Fbn2 in the cortical and nephrogenic interstitium, Postn in the cortical, proximal tubule interstitium, and interstitium medullary to proximal tubule (outer medulla), Col1a1 in the cortical, proximal tubule interstitium, interstitium medullary to proximal tubule (outer medulla), papillary, and ureteric interstitium, Tnc and Col5a1 in the proximal tubule interstitium and interstitium medullary to proximal tubule (outer medulla), Emilin1 in the proximal tubule and outer stripe of the inner medulla, and Col26a1 in the papillary interstitium ( Figure 1B). This suggested that, by E18.5, stromal cells were expressing ECM RNA in cell/location type-specific patterns.
2.2 | The distribution of interstitial matrix proteins was similar in the developing kidney but diverged in the adult To assess the spatiotemporal location of interstitial matrix proteins that are more highly expressed in embryonic kidneys compared to adult, 3 and had distinct RNA expression patterns at the scRNA level (Figure 1), we stained cryosections from embryonic days (E)14.5, E18.5, postnatal day (P)3 (coronal sections), and adult (transverse sections) kidneys for COL26A1, FBN2, EMILIN1, and TNC. Antibodies against the basement membrane components perlecan (HSPG2) and FRAS1-related extracellular matrix 2 (FREM2) were used to highlight the nephron (Figure 2A-D). We previously showed the distribution of COL5, COL1, and POSTN. 3 At E14.5, the distribution of FBN2 and COL26A1 was medullary and cortical, EMILIN1 was medullary, and TNC was cortical (Figure 2A). In contrast to the scRNA data ( Figure 1), at E18.5 and P3, these interstitial matrix proteins surrounded the developing nephron and localized to the medullary ray sheath and vertical fibers ( Figure 2B,C). TNC was enriched around the proximal tubule as previously observed 17 ; however, the pattern of these interstitial matrix proteins diverged in the adult F I G U R E 1 RNA encoding various ECM was enriched in different stromal cell types within the kidney. (A) Diagram highlighting the stromal cell groups (colored, numbered 1-16) modified from 16 overlayed with ECM structures (gray). 3 The stromal cell groups were determined by single cell RNA (scRNA) sequencing clustering and validated via in situ hybridization in England et al. 17 (B) Heat map of interstitial matrix 3 transcripts from E18.5 mouse kidney scRNA sequencing data from England et al. 17 The rank of the RNA identified in each cluster was plotted and grouped manually. Transcripts that were not ranked in the top 500 genes for any cluster were not included. Clusters: 1, 2, 3 = cortical interstitium; 4, 5 = nephrogenic interstitium; 6, 7, 8 = proximal tubule interstitium; 9 = interstitium medullary to proximal tubule (outer medulla); 10 = outer stripe of the inner medulla interstitium; 11, 12 = papillary interstitium; 13 = ureteric interstitium; 14 = vascular smooth muscle; 15 = pericyte; and 16 = mesangium; and 17 = indeterminate signature. RNA transcripts listed in red are discussed in the text. White box indicates the transcript was not found in that cell type ( Figure 2D). EMILIN1 was enriched in the inner medulla and inner stripe of the outer medulla, while TNC was sporadically found within the inner medulla and inner stripe of the outer medulla and mature glomeruli, whereas FBN2 disappeared ( Figure 2D). COL26A1 surrounded the Dolichos biflorus Agglutinin + (DBA + ) collecting ducts ( Figure 2E).
While cryosections can provide insight into general ECM distribution (cortical vs medullary), methods to view 3D networks are essential to resolve the intricate F I G U R E 2 The distribution of interstitial matrix proteins was similar in patterning in the developing kidney but diverged in the adult. (A-D) Coronal sections of E14.5, E18.5, and P3 kidneys and transverse sections of adult timepoints were stained for COL26A1, FBN2, EMILIN1 (green) and FREM2, HSPG2, TNC (red). (E) COL26A1 surrounded DBA + (red) collecting ducts in the adult kidney. (F-O) In the E14.5 kidney, EMILIN1 + medullary ray sheath fibers (green; *) surrounded the developing nephron (FREM2 + , blue; WGA + , red) and ran parallel to FREM2 À /WGA + blood vessels (open arrowheads). At E18.5 and P3, EMILIN1 + and COL26A1 + (green) medullary ray sheath fibers (*) and vertical fibers (arrows) surrounded the developing nephron and ELN + blood vessels. patterning of the ECM. Therefore, we decellularized kidneys to remove light-scattering lipids and retain the insoluble ECM. 3 The WGA was used to delineate the general architecture of the ECM. At E14.5, the EMILIN1 + interstitial matrix surrounded the developing nephron (FREM2) and ran parallel to the developing blood vessels F I G U R E 3 Legend on next page.
2.3 | Loss of Foxd1 in stromal cells resulted in abnormally clustered vertical fibers that did not cross the capsule at E18.5 To determine how stromal defects influenced ECM patterning, a functional knockout of Foxd1 (Foxd1 GC/GC ) was compared with control kidneys. In the Foxd1 GC/GC model, cells with Foxd1 promoter activity were retained, 12,15 but had an abnormal phenotype that was a phenocopy of kidneys when Foxd1 + cells are ablated. [19][20][21] There were defects in nephrogenic precursors in the Foxd1 null kidney 14 ; therefore, descriptions are relative to the nephron structures.
Overall, Foxd1 GC/GC kidneys were smaller, fused together in a horseshoe shape with 100% penetrance, and ventrally rotated with marked divots on the surface where superficial blood vessels were located ( Figure 3A-D), consistent with prior Foxd1 knockouts. 20 The POSTN + capsule was thicker ( Figure 3E-F 0 ), and COL26A1 + vertical fibers were abnormally, perpendicularly aligned relative to the branching nephron in Foxd1 GC/GC kidneys ( Figure 3G-H 0 ). No vertical fibers was observed crossing the capsule ECM of the fused Foxd1 GC/GC kidneys ( Figure 3I,J).

|
The distribution of medullary ray sheath fibers was unaltered despite the disassociation with arcuate blood vessels at E18.5 We next assessed the localization of the arcuate blood vessels relative to the medullary ray sheath fibers at the corticomedullary junction 3 ( Figure 4A-O). In control kidneys, the arcuate artery was ELN + /FREM2 À /COL4 + / WGA + . When we assessed Foxd1 GC/GC kidneys, ELN + /FREM2 À /COL4 + /WGA + luminal blood vessels were present in a subcapsular location consistent with a prior study 20 and were surrounded by a COL5 + interstitial matrix ( Figure 4A-F 0 ). At the corticomedullary junction, EMILIN1 + and POSTN + medullary ray sheath fibers are present in both the control and Foxd1 GC/GC kidneys despite the lack of associated ELN + blood vessels ( Figure 4G-J 0 ).

|
The Foxd1 GC/GC kidney matrisome had a reduced amount of basement membrane components at E18.5 E18.5 kidneys were analyzed using LC-MS/MS to further assess changes in ECM proteins, or matrisome, in Foxd1 GC/GC and control mice. We identified 170 matrisome components, consistent with other kidney development proteomic studies 3,22,23 (Table S1). There was a significant decrease in matrisomes relative to cellular proteins in Foxd1 GC/GC kidneys compared to controls ( Figure 5A), suggesting either increased cellularity, an increase in ECM extractability, or a reduction in ECM synthesis. 24 Despite the change in interstitial ECM organization, we did not identify significant changes in the relative amount of interstitial matrix ( Figure 5B). However, using LFQ values normalized to the matrisome fraction, there was a significant increase in BMP binding endothelial regulator (BMPER), a protein implicated in growth factor regulation 25 ( Figure 5C-E). GO analysis on proteins significantly elevated in, or exclusive to, Foxd1 GC/GC kidneys generated terms, including "microfibril" and "regulation of TGF-β production" ( Figure 5D).
In contrast, there was a decrease in the overall fraction of basement membrane in the Foxd1 GC/GC kidneys ( Figure 5B). Additionally, COL4A4, COL4A5, (found in glomerular basement membranes [GBMs] and some tubular basement membranes [TBMs]), 26 COL4A6 (specific to Bowman's capsule), 26 TINAG (GBM, TBM), 27,28 and TINAGL1 (GBM, TBM), 29 were significantly reduced F I G U R E 3 Loss of Foxd1 in stromal cells resulted in abnormally clustered vertical fibers that did not cross the capsule at E18.5. (A-B 0 ) Control and Foxd1 GC/GC kidneys were decellularized to resolve the 3D ECM structure (a: adrenal gland; b: bladder).  Figure 5C,E). GO analysis of proteins significantly elevated or exclusive to control kidneys generated terms related to "collagen type IV trimer" and "laminin-5 complex," suggesting delayed maturation of the Foxd1 GC/GC kidney basement membrane ( Figure 5D).
2.6 | Foxd1 GC/GC kidneys contained COL4A4 + fused glomeruli at E18.5 In control kidneys, maturing glomeruli were found near the corticomedullary junction and S-shaped and commashaped bodies in the cortex. This radial change in The distribution of medullary ray sheath fibers was unaltered despite the disassociation with arcuate blood vessels at E18.5. (A-D) ELN + blood vessels (green; open triangles) were located near the medullary ray sheath fibers (*) in the control kidneys. In Foxd1 GC/GC kidneys, ELN + blood vessels were found in the cortex and were surrounded by interstitial matrix (arrows). (E-F 0 ) COL5 + fibers surrounded the superficial blood vessels (COL4, blue; WGA, red). (G-J 0 ) EMILIN1 + and POSTN + medullary ray sheath fibers (*) were observed in the glomeruli maturity was disrupted in Foxd1 GC/GC kidneys, where maturing stage glomeruli were found throughout the cortex. However, there was no marked difference in the COL4A4 staining, a marker of GBM maturation 30 in Foxd1 GC/GC kidneys ( Figure 6A-B 0 ).
Interestingly, we observed sporadic, fused glomeruli with two glomerular tufts within one Bowman's capsule ( Figure 6C-D 0 ; Videos S1 and S2) at E18.5. Similar structures were not observed at E14.5 (data not shown). These glomeruli were at different stages of maturation and F I G U R E 5 The Foxd1 GC/GC kidney matrisome had a reduced amount of basement membrane components at E18.5. (A) The percent matrisome of different cellular compartments was significantly reduced in the Foxd1 GC/GC kidneys (two-tailed t test, P = 0.0070). (B) There was a significant decrease in basement membrane proteins in the IN fraction in the Foxd1 GC/GC kidneys (two-tailed t test P = 0.0110). In contrast, there was no significant difference in the interstitial matrix. (C) Volcano plot comparing control and Foxd1 GC/GC kidneys based on log 2 scaled label free quantification (LFQ) values. Significance was based on P < 0.05 and jfold changej > 2 (gray lines). (D) GO analysis terms generated using proteins significantly elevated in, or exclusive to, the different genotypes. (E) Log 10 -scaled LFQ intensity heat map revealed minimal changes in the interstitial matrix, whereas there was a reduction in some basement membrane proteins. The analysis included proteins identified in n = 2-3 biological replicates. White boxes signify zero intensity values  Figure 6E-F 0 ; Videos S3-S5). EMI-LIN1 + vertical fibers were abnormally oriented relative to both single glomeruli and fused glomeruli in the knockout ( Figure 6G-H 0 ). The anatomy of the fused glomeruli varied. In general, there were two macula densa, one or two urinary poles, and either connected or branched afferent and efferent arterioles ( Figure 6I; Videos S1-S5).
2.7 | Alterations in branching preceded vertical fiber abnormalities in Foxd1 GC/GC embryonic kidneys E14.5 Foxd1 GC/GC kidneys were compared with controls to investigate the formation of the abnormal vertical fibers in relation to ureteric bud development. FREM2 + ureteric bud branches were reduced compared to the control and arrayed along the transverse axis in Foxd1 GC/GC kidneys instead of the bifurcated branching classically observed 31 ( Figure 7A-B 0 ). Between the ureteric bud in Foxd1 GC/GC kidneys, EMILIN1 + vertical fibers formed clusters instead of the regular, reticular patterning observed in controls ( Figure 7C-D 0 ). Glomeruli maturation was delayed in the Foxd1 GC/GC kidneys with a reduced number of apparent capillary loops compared to controls ( Figure 7E,F). Notably, the network of Foxd1 GFP + cells surrounding the ureteric bud tips was lost in Foxd1 GC/GC kidneys and corresponded to defects in COL5 + patterning at the surface ( Figure 7G-H⁗), suggesting a correlation between stromal cells and ECM patterning.
To determine if abnormal ECM patterning preceded branching defects, E12.5 kidneys were assessed. 32 At E12.5, the stalk of Foxd1 GC/GC kidneys was broader than the controls, with fewer branching tips ( Figure 7I-J 0 , K). The EMILIN1 + fibers did not appear dramatically different between the genotypes ( Figure 7I 00 -J 00 ).
2.8 | Foxd1 GC/+ ; Col5a1 fl/fl mice did not have any overt kidney defects Because we observed that the COL5 + patterning was disrupted in the Foxd1 GC/GC kidney ( Figure 7G-H⁗), we next assessed if COL5 is important for nephrogenesis. COL5 is critical for COL1 fiber nucleation, fiber organization, and tissue strength, [33][34][35] and we hypothesized the organization of the interstitial matrix would be affected when knocked out in stromal cells. Col5a1 was knocked out in Foxd1-expressing connective tissue cells (Foxd1 GC/ + ; Col5a1 fl/fl ) since global knockout of Col5a1 À/À is embryonic lethal at E10, prior to nephrogenesis. 33 The 3D distribution of ECM in E18.5 decellularized Foxd1 GC/ + ; Col5a1 fl/fl kidneys appeared normal but was more easily disrupted than in controls during the decellularization process, suggesting ECM stability was altered in the knockout ( Figure 8A-P). Body and kidney weight were not affected by the knockout of COL5 ( Figure 8O,R). Foxd1 GC/+ ; Col5a1 fl/fl mice were born at Mendelian ratios but did not survive to P21 at expected ratios ( Figure 8S), likely due to extrarenal causes, as proteinuria and hematuria were not observed via urine dipstick (data not shown).

| DISCUSSION
While the ECM undergoes dynamic changes in composition and structure during normal kidney development, how the ECM changes when stromal cells are disrupted were unknown. To determine whether developmental ECM structures were altered in a model of kidney hypodysplasia, we combined 3D imaging and proteomics and resolved distinct changes in cortical vertical fibers ( Figure 3G-H 0 ). The changes in ECM structure corresponded to alterations in Foxd1 + stromal cell patterning; however, these changes were not observed prior to defects in ureteric bud branching (Figure 8).
In addition, we investigated how the distribution of an ECM component that is transiently upregulated during development, COL5, affected patterning in the kidney. The kidneys in Foxd1 GC/+ ; Col5a1 fl/fl mice did not have any clear defects (Figure 8). Under normal conditions, the COL5 trimer can be composed of α1α1α2, α1α2α3, or α1α1α1 chains. 36,37 The lack of a phenotype could be due to the contribution of COL5A1 by other stromal cells, such as Tbx18-expressing cells. 9,38 Alternatively, redundancy or compensatory mechanisms, such as upregulation of other COL5 chains (eg, COL5A2) 39 or other proteins associated with COL1 fiber regulation (eg, COL3A1), 40 could contribute to the lack of a phenotype. Foxd1 GC/+ ; COL5A1 fl/fl knockout mice did not survive, likely due to extrarenal effects since we did not observe proteinuria or hematuria or gross changes in kidney morphology. Foxd1 is expressed in other tissues, including the brain (optic chiasma 41 and hypothalamus 42 ) and lung. 43 This suggests that other tissues lack COL5A1 in the knockout, which may decrease viability; for example, haploinsufficiency of COL5A1 decreases lung function. 44 Alternatively, the lack of phenotype could be due to inefficient activity of the Foxd1 GC Cre in the stromal cells, resulting in an incomplete deletion of COL5A1.
While IHC staining showed the distribution of ECM proteins such as COL5 were found throughout the kidney, scRNA data indicated that stromal cells express  Figures 1  and 2). 3 This suggested that some of the ECM components visualized via IHC were synthesized earlier in development and persisted in the matrix, indicating the need to combine RNA level analyses with protein composition and distribution. Further work in quantifying ECM protein turnover can clarify the discrepancy in RNA and protein localization (Figure 1).
The distribution of interstitial ECM components that were upregulated in developing kidneys (EMILIN1, FBN2, TNC, COL26A1, and COL5) was similar at perinatal timepoints but varied in the adult (Figure 2A-D). Notably, COL26A1 is localized around the adult DBA + collecting duct ( Figure 2E). In developing tissues, COL26A1 was found in the vertical fibers and was abnormally clustered in the Foxd1 GC/GC kidneys. Notably, the abnormal vertical fibers did not cross the thickened capsule in the Foxd1 GC/GC kidney ( Figure 3E-J), likely due to superficial vs parenchymal fusion resulting from a cellular capsule defect. 12 Some proteins found in the clustered vertical fibers can be grouped as part of the elastinmicrofibril axis, which was elevated during murine and human nephrogenesis. 3,23 Proteins in this axis contribute to growth factor regulation through transfoming groth factor β (TGF-β) and BMP signaling. [45][46][47][48][49][50] For example, one protein in this axis, EMILIN1 connects microfibrils and ELN and inhibits TGF-β signaling in the vasculature. 51 In the Foxd1 GC/GC kidney, the disorganized EMI-LIN1 + vertical fibers ( Figure 6G,H) could alter local growth factor availability and contribute to the ureteric bud branching dysmorphogenesis. This is supported by prior studies that found stromal cells modulate nephron patterning through altered DCN and BMP7 signaling pathways. 14 When assessing the ECM composition of the Foxd1 GC/GC kidney at E18.5, we did not observe a significant increase in DCN, suggesting the distribution had normalized by this time ( Figure 5C). The structural changes in the Foxd1 GC/GC kidney could correspond to an increase in BMP-associated proteins, such as BMPER ( Figure 5C).
The abnormalities in the vertical fibers indicate Foxd1 is important for ECM orientation and could contribute to the loss of the nephrogenic zone as previously described in the Foxd1 knockout. 14 Distinct changes in interstitial ECM were observed relative to nephron structures and correlated with changes in stromal cell location in the setting of abnormal nephron patterning in the Foxd1 knockout ( Figure 7H). [12][13][14] To clarify if the ECM changes were secondary to nephron changes or specific to stromal changes, models of renal hypodysplasia due to tubular vs stromal mutations can be compared in future studies.
In addition to the disorganized pattern of glomeruli maturity in Foxd1 GC/GC kidneys, 12 there were fused glomeruli with diverse configurations ( Figure 6A-B 0 , E-F 0 , I and Videos S1-S5). While the glomeruli expressed COL4A4, an indicator of GBM maturation, the altered anatomy will prevent accurate tubuloglomerular feedback; the combined inputs of two glomerular tufts will eliminate the direct one-to-one correlation of the afferent arterial, Bowman's capsule, and macula densa. 52 The common urinary poles suggested fusion began as early as the S-shaped body. The cause of the fused glomeruli could be a result of abnormally clustered vertical fibers, although the effect would be stochastic since not all glomeruli were affected. Alternatively, dysregulation of Foxd1 in vascular epithelial, mesangial, and parietal epithelial cells could have contributed to glomerular fusion.
Glomerular fusion was also observed in a horseshoe kidney in a trisomy 18 human fetus. 53 Horseshoe kidneys in humans may be caused by physical entrapment or ectopic mesenchymal tissue. 54 By contrast, the Foxd1 knockout horseshoe kidney is due to a thickened capsule and adhesion to the body wall. 12 Nevertheless, fused glomeruli in both a trisomy 18 fetus and the Foxd1 GC/GC kidneys suggested a common etiology of the horseshoe kidney and glomerular fusion, for example, altered signaling pathways. 55 Even though the fused glomeruli expressed COL4A4 at E18.5, indicating maturation, proteomic analysis revealed a reduction in the percentage of basement membrane proteins, particularly those enriched in the adult ( Figure 5C-E). This reduction in mature basement membrane proteins could be associated with the delay in glomerular maturation observed at E14.5 ( Figure 7E,F). 12 Alternatively, the density of glomeruli or other nephron structures where these ECM proteins localize could be decreased. 26 While defects in the nephrogenic precursors of the Foxd1 null mice have been described, 14 the GBM in Foxd1 null kidneys has not been analyzed via electron microscopy, and further studies could reveal changes in GBM structure.
The superficial relocation of the ELN + vasculature could be associated with the role of Foxd1 in the development of blood vessel mural cells. 20 Nevertheless, interstitial matrix components still surrounded the superficial blood vessels in the Foxd1 GC/GC kidney ( Figure 4A-F 0 ). The medullary ray sheath fibers at the corticomedullary junction were not affected by the absence of the ELN + arcuate artery, indicating these structures may support aspects of development unrelated to the vasculature. 3,56 The minimal alteration in interstitial ECM composition is likely due to the persistence of stromal cells that have an active Foxd1 promoter in the Foxd1 knockout model ( Figure 7G-H 000 ). 12,15 However, kidneys in which Foxd1 + cells were ablated [19][20][21] have similar defects as Foxd1 À/À kidneys 12,13,20 with abnormalities in nephron patterning, reduced stromal cells, and mispatterned TNC expression. Alternatively, other stromal cells, such as those regulated by the transcription factor Tbx18, 9,38 could have an additional influence on interstitial ECM patterning. While the interstitial matrix present at the perinatal timepoints is stable relative to the perturbation in Foxd1 (Figures 4-6), 3 it is unclear if the interstitial matrix is unique to the kidney. The basement membrane of the E14.5 kidney was different than limb, whole embryo, and brain 3 ; however, there was little interstitial matrix at this timepoint. Future work can clarify if the development of the scaffold composition was unique to the kidney or reflects a general trend in internal organ development (such as in the liver, lung, heart, or pancreas).
In summary, we provide evidence that the establishment of the 3D ECM architecture in the developing kidney depends on normal stromal cells (Figure 9). While branching dysmorphogenesis preceded the disruption of the interstitial matrix in Foxd1 GC/GC kidneys, the disorganized ECM could contribute to late-stage defects in the nephrogenic environment and kidney hypodysplasia. Future studies that disrupt additional matrisome components that make up the vertical and medullary ray sheath fibers will help further determine the role of the interstitial matrix in nephrogenesis.

| Kidney sample collection
Wild-type C57BL/6 mice (WT; Jackson Laboratory, # 000664), Foxd1 tm1(GFP/cre)Amc (Foxd1 GC/+ ; Jackson F I G U R E 9 Summary of ECM changes in Foxd1 GC/GC kidneys. E12.5: branching dysmorphogenesis was observed prior to defects in ECM patterning. E14.5: interstitial ECM was abnormally clustered around the dysmorphic ureteric bud branches in the E14.5 Foxd1 GC/GC kidneys. E18.5: vertical fibers were disorganized, fused glomeruli were observed, but medullary ray sheath fibers were not disrupted in  Laboratory, # 012463), and Col5al fl/fl34 (gift from Dr. David Birk) were used in the study. Experimental protocols were in compliance with either the Purdue University or the University of Colorado-Boulder Institutional Animal Care and Use Committee (PACUC or IACUC; protocol #1209000723 or #2705). PACUC and IACUC assessed that university researchers follow all procedures and facilities are compliant with regulations of the US Department of Agriculture, US Public Health Service, Animal Welfare Act, and Purdue's Animal Welfare Assurance. Mice were time-mated, and embryonic day (E)0.5 referred to noon of the day when the copulation plug was noted. Adult mice and pregnant dams were euthanized by CO 2 inhalation and confirmed via cervical dislocation. E18.5 and P3 pups were euthanized via decapitation. Embryos and pups were dissected in chilled PBS to isolate the kidneys. The sex of the embryos and pups was not determined. The Foxd1 GC/+ allele and Col5a1 fl/fl were genotyped using primers in Table 1. Foxd1 GC/+ , Foxd1 +/+ , and Col5a1 fl/fl genotypes were used as controls.

| Immunohistochemistry
Dissected kidneys were embedded in an optimal cutting temperature compound (Electron Microscopy Science) and stored at À80 C until cryosections were collected. Ten micrometer thickness cryosections were acquired using a Shandon Cryotome FE (ThermoFisher), adhered to charged slides, and stored at À20 C until use.
IHC was performed following the antibody dilutions described in Table 2. Sections were imaged on a Leica DMI6000 at 20Â magnification. Negative controls consisted of the same process and settings without the addition of the primary antibody. Images shown are representative of ≥2 biological replicates from two independent litters.

| Wholemount staining
E14.5 kidneys were isolated and fixed for 2 h at RT and/or overnight at 4 C. IHC was performed following 3 using dilutions found in Table 2. Sequential secondary antibody incubations were used to avoid interactions between secondary antibodies. Images are representative of n ≥ 3 biological replicates.

| 3D imaging of kidneys
Kidneys were processed for 3D imaging of the ECM following. 3 Dissected kidneys were directly incubated in an SDS-based solution or embedded in agarose for the times and at concentrations in Table 3. After decellularization, fixation, and blocking, samples were stained with primary and secondary antibodies at concentrations in  Table 4 with a line average of 2. Negative controls were obtained using the same process as samples stained with both primary and secondary antibodies, with the exclusion of the primary antibodies, and were imaged at the maximum setting used for the antibody channels. Images are representative of n ≥ 3 biological replicates. adjusted per sample for improved structural resolution. Wholemount image kidneys are visualized using maximum z-projection in FIJI. Images of cryosections were processed using FIJI. Adobe Photoshop and Adobe Illustrator were used to combine the cryosections, decellularized, and wholemount images.

| Proteomics
Proteomic analyses were performed following. 3 The peptides were analyzed at the University of Colorado Boulder Proteomics and Mass Spectrometry facility using an Ulti-Mate 3000 RSLC Nano System coupled to a QE HF-X. Peaks from the raw file were analyzed by Max-Quant 62,63 (version 1.6.7.0) with the settings shown in Table S1 against a FASTA database for Mus musculus (downloaded 12/06/2019) with canonical variants and contaminants. Fixed modifications of cysteine carbamidomethylation and variable modifications of oxidation of methionine, hydroxylysine, hydroxyproline, deamidation of asparagine, and conversion of glutamine to pyroglutamic acid were used. Match-between-runs was used for biological replicates of the same fraction. A decoy database derived from the Mus musculus database was used to control the false discovery rate to 1%. LFQ was enabled for biological replicates.

| Statistical analysis
Kidneys for proteomic analyses were collected with n = 3 biological replicates and analyzed in Prism (GraphPad, V9.0.1), and differences in protein abundance were compared using a t test. Branching tip number was compared using a t test. χ 2 analysis was used to determine if the distribution and weight of Col5a1 fl/fl; Foxd1 Gc/+ knockout mice were different from control mice.