Reconstitution of the embryonic kidney identifies a donor cell contribution to the renal vasculature upon transplantation

The kidney possesses a highly organised vasculature that is required for its filtration function. While recent advances in stem cell biology have enabled the in vitro generation of kidney tissues, at least partially, recapitulation of the complicated vascular architecture remains a huge challenge. Herein we develop a method to reconstitute both the kidney and its vascular architecture in vitro, using dissociated and sorted mouse embryonic kidney cells. Upon transplantation, arteriolar networks were re-established that ran through the interstitial space between branching ureteric buds and eventually entered glomeruli. Using this system, we found that donor-derived endothelial cells significantly contributed to the arterioles and glomerular capillaries formed after transplantation. Unexpectedly, the near-complete depletion of canonical endothelial cells from the donor embryonic kidney suggested the existence of unidentified donor-derived endothelial precursors that were negative for canonical endothelial markers, but still contributed significantly to the vasculature in the transplants. Thus, our protocol will serve as a useful platform for identification of renal endothelial precursors and induction of these precursors from pluripotent stem cells.

Scientific RepoRts | (2019) 9:1172 | https://doi.org/10.1038/s41598-018-37793-z iPSC-derived kidney organoids revealed the existence of human ECs in the glomeruli to some extent 14 . Therefore, it is critical to determine the donor versus host contribution to renal ECs after transplantation, especially in the kidney organoid setting, for recapitulation of the renal vasculature. In addition, kidney organoids would be beneficial for detection of renal EC precursors and their development, because candidate cell fractions can be sorted and reaggregated into organoids. While reaggregation of the embryonic kidney from a single-cell suspension was reported 15,16 , few have succeeded in reconstituting the arteriolar structure and glomerular vasculature. We recently reported the reconstitution of a higher-order kidney structure from mouse ESCs and dissociated embryonic kidneys 17 . Our method utilised three progenitors of the kidney: Itga8+ NPs that give rise to glomeruli and renal tubules, Pdgfra+ stromal progenitors (SPs) that form interstitial cells, and the ureteric bud (UB) that gives rise to collecting ducts and ureters. When NPs and UBs (isolated from embryonic kidneys or induced from mouse ESCs) and embryonic kidney-derived SPs were aggregated, three-dimensional kidney-like structures were generated in vitro, consisting of branching UB trees with nephrons (glomeruli and renal tubules) distributed at the UB tips 17 . We reasoned that incorporation of ECs into this system would facilitate the generation of vascularised organoids, and also elucidate the donor versus host origin of ECs after transplantation. In this study, we successfully generated vascularised kidney organoids equipped with branching UBs and nephrons. Upon transplantation, intrarenal arteriolar structures were formed in the space between the branching UBs, and the glomeruli were vascularised. Furthermore, we unexpectedly identified the existence of EC precursors that can dominantly contribute to the vasculature upon organoid transplantation.

Results
Extrarenal and intrarenal vasculature framework is established by E15.5. To examine when and how the renal vasculature is formed in vivo, we stained serial sections of embryonic day (E) 12.5 and E13.5 mouse embryos for CD31 (Fig. 1A,B). We then reconstructed three-dimensional images by colouring the CD31 + ECs continuous with the aorta in red (putative arteries and arterioles) and those continuous with the vena cava in blue (putative veins) ( Supplementary Fig. S1). While well-formed renal arteries were observed at E13.5, those at Whole-mount staining of arteries (Cx40) and the kidney (UB: CK8) at E12.5 (E,F) and E13.5 (G,H). Dorsal views of the right kidneys are shown. (F' ,H') Digitised images of (F,H), respectively. Arrows: renal arteries budding from the aorta (ao) and invading into the kidneys; arrowheads: collateral arteries connecting the renal arteries and the common iliac artery (ci). At E13.5, the collateral arteries had disappeared. (I) Whole-mount staining of arteries (Cx40) at E15.5. A dorsal view of the right kidney is shown. Arrow: a single stalk for the renal artery is formed. (J) Whole-mount staining of intrarenal arterioles (Cx40) connected to glomeruli (Nephrin). (J') Digitised image of (J). Three mice were examined at each stage and representative images are shown. Scale bars: 100 µm.
Scientific RepoRts | (2019) 9:1172 | https://doi.org/10.1038/s41598-018-37793-z E12.5 appeared to arise from not only the abdominal aorta, but also the common iliac arteries (Fig. 1A). Veins were formed as plexuses around the arteries, as confirmed by whole-mount staining of a vein marker, Nrp2 18,19 ( Supplementary Fig. S1). We further performed whole-mount immunofluorescence staining for Connexin 40 (Cx40), an arteriole marker 20,21 , at the initial stages of kidney development. At E11.5, the kidney was located on the ventral side of the common iliac arteries. While small vessels surrounded the kidney, no clear vascular connection to the kidney was observed (Fig. 1C,D, D'). At E12.5, short stalks budded out from the aorta and started to branch into the kidney (Fig. 1E,F, F'). One of the branches elongated caudally and connected to the common iliac artery, consistent with the findings in Fig. 1A. At E13.5, this connection disappeared, but branched arteries invaded deep into the kidney through the space between the UBs (Fig. 1G,H, H'). At E15.5, the stems of the arteries were unified into one renal artery, and many tips branched out from the arterioles and connected to the glomeruli (Fig. 1,J,J'). These results suggested that the initial vascular connection in the kidney occurred at E11.5-E12.5 and the overall framework of the extrarenal and intrarenal vasculature was established by E15.5.
Endothelial cells invade the kidney at E11.5. Flow cytometry analysis of the dissociated E11.5 kidneys showed the presence of CD31+ and/or Flk1+ populations ( Fig. 2A). Because the majority, if not all, of the CD31+ cells were Flk1+, we utilised Flk1-GFP mice 22 to analyse the detailed spatial distribution of ECs around the kidney at E11.5. Although no clear arterial connection was observed as described above, numerous Flk1+ ECs surrounded the developing kidney (Fig. 2B). On the dorsal side of the kidney, ECs appeared to be derived from the common iliac arteries and reached, but did not penetrate, the NP region in the kidney (Fig. 2B,C). On the ventral side, ECs were distributed in the space along the Wolffian duct (Fig. 2B), and some of them ran toward the kidney along the UB (Fig. 2D). We also examined Tie2Cre;tdTomato mice 23,24 at E11.5 and found that the majority of CD31+ cells were tdTomato+ by flow cytometry (Fig. 2E). The spatial distribution of tdTomato+ ECs was similar to that in Flk1-GFP mice ( Fig. 2F-H). Thus, ECs invade into the dorsal and ventral sides of the kidney at E11.5. These findings were at least partially consistent with recent reports on kidney vasculature development 25-27 . Vascular development is limited in the reconstituted kidney organoids in vitro. We recently reported the reconstitution of a higher-order kidney structure from dissociated mouse embryonic kidneys 17 . To examine the vascular development in the reconstituted kidney organoids, we dissociated E11.5 kidneys and combined them with ECs. CD31+ and/or Flk1+ cells were sorted as ECs, and the cells in the CD31−/Flk1− fraction Expression of Cx40 was almost undetectable and much weaker than that in the freshly isolated intact kidney, even when ECs were combined for kidney organoid formation (Fig. 3F,K, Supplementary Fig. S2). Section staining showed the presence of ECs in the interstitial space of the kidney ( Supplementary Fig. S2), but the cells failed to integrate into the glomeruli (Fig. 3G,L). These results showed that our reaggregation method could generate vascularised kidney organoids in vitro, although arteriolar development and glomerular vascularisation were still limited.
Transplanted kidney organoids form an arteriolar network and glomerular capillaries. Because of the limited vascular development in the kidney organoids in vitro, we transplanted reconstituted kidney organoids (day 3 of culture) under the renal capsules of immunodeficient mice (Fig. 4A) using our previously reported method 8,29 . When harvested at day 7 after transplantation, the tissues were significantly larger than their size at transplantation (Fig. 4B) and the GFP+ UBs had branched extensively (Fig. 4C). A Cx40+ branching network was observed to run through the space between the UBs (Fig. 4C' ,C") and reach the glomeruli (Fig. 4D,D'), indicating arteriole formation in the kidney organoids. Furthermore, CD31+ ECs were not only detected in the interstitial regions of the tissues, but also integrated into all the NEPHRIN+ glomeruli examined (n = 19; Fig. 4E), in sharp contrast to the situation in vitro. Thus, reconstitution of embryonic kidneys and subsequent transplantation successfully restored the arteriolar network development and glomerular vascularisation, although the running patterns of the arterioles did not completely recapitulate those in vivo.  To examine the necessity for ECs in the transplantation assay, we removed CD31+and/or Flk1+ECs from the dissociated metanephric mesenchyme and performed the same reconstitution and transplantation procedures (Fig. 4A). Unexpectedly, the tissue growth (Fig. 4B), UB branching, and arteriolar network formation were comparable to those with ECs (Fig. 4F,F' ,F") and the glomeruli were vascularised with ECs (n = 25; Fig. 4G,G' ,H). These results raised two possibilities: the host ECs compensate for the absence of donor ECs, or some EC precursors still exist in the donor tissues.
Donor-derived ECs contribute to the majority of the renal vasculature upon transplantation. To examine the possibilities described above, we utilised Tie2Cre;tdTomato mice 23 , which express tdTomato in ECs. Initially, we transplanted intact E11.5 kidneys from Tie2Cre;tdTomato mice, and found that the donor cells contributed the majority of the vasculature in the grafted tissues ( Supplementary Fig. S3), consistent with previous results [11][12][13] . We then sorted Itga8+ NPs and Pdgfra+ SPs from the CD31−/Flk1−/tdTomato− fraction and CD31+ and/or Flk1+ /tdTomato+ ECs from Tie2Cre;tdTomato mice, and aggregated them with Hoxb7-GFP+ UBs to reconstitute kidney organoids (Fig. 5A, Supplementary Fig. S3). When cultured for 7 days in vitro, tdTomato+ vascular structures were formed in the organoids (Fig. 5B), consistent with the results shown in Fig. 3. Upon transplantation (day 7), tdTomato signals were detected in the entire organoids (Fig. 5C). When exogenous ECs were not combined for organoid generation, formation of a tdTomato+ vasculature in vitro was modestly reduced (p < 0.05; Fig. 5D, Supplementary Fig. S2). However, upon transplantation, tdTomato signals were unexpectedly detected in the entire organoids (Fig. 5E). Section staining showed that tdTomato+ cells contributed to the Cx40+ arterioles and glomerular capillaries (Fig. 5F,G), irrespective of the presence of exogenous ECs in the transplanted organoids (Fig. 5H,I). Flow cytometry analysis further revealed that tdTomato+ cells constituted 75-85% of the CD31+ ECs in both cases (Fig. 5J,K), although the presence of exogenous ECs produced slightly higher percentages (83.6 ± 3.8% with ECs; 74.4 ± 4.2% without ECs). These data suggest that the CD31−/Flk1−/Tie2− fraction in the E11.5 kidney may contain EC precursors that can contribute more efficiently to the vasculature of the transplanted organoids than the host ECs. Finally, we combined CD31+ and/or Flk1+ /tdTomato+ ECs from E11.5 Tie2Cre;tdTomato mice with NPs and SPs from the CD31−/Flk1− fraction, as well as UBs, from wild-type mice. Upon transplantation, only 4.57 ± 0.88% of CD31 + ECs in the organoids were positive for tdTomato (n = 7 from 3 independent transplantation experiments). Because the majority of the vasculature in the transplanted organoids was derived from the donor, as shown in Fig. 5J,K, endothelial recruitment from the CD31−/Flk1− fraction is likely to occur even in the presence of exogenous ECs.

Discussion
Herein, we have established in vitro and in vivo protocols that can enable vascular development in kidney organoids. Because we identified that the vasculature started to invade the kidney at E11.5, we isolated four populations, NPs, SPs, UB, and ECs, from this stage of kidney development and reconstituted the kidney structures. The kidney organoids in vitro exhibited vascular development along with extensive UB branching and glomerulus formation. Transplantation of the organoids further conferred arteriolar network formation, as well as glomerular vascularisation. By using these assays, we revealed the presence of unidentified EC precursors in the donor embryonic kidney that can efficiently contribute to the vasculature upon transplantation.
Our data are consistent with previous reports showing the dominance of donor-derived ECs after transplantation of genetically labelled intact embryonic kidneys [11][12][13] . However, one of the main differences is that ECs, as well as NPs and SPs, were completely dissociated in our study. Even in this situation, reaggregated embryonic donor cells could reorganise the arteriolar architecture and dominate over the infiltrating host adult ECs. These findings suggest a potential advantage of embryonic ECs over adult ECs for renal vasculature reorganisation upon transplantation.
Unexpectedly, we found that the CD31−/Flk1−/Tie2− fraction in the embryonic kidney contains EC precursors that contributed to the majority of the vasculature upon transplantation. Although the cell types comprising these EC precursors with such robust potential remain unknown, our data are at least partially consistent with a previous report showing the existence of CD146+/CD31− EC precursors that can contribute to the vasculature during in vitro culture 30 . Thus, it would be worthwhile to test the CD146+ fraction in our transplantation assays. Once the fraction containing the EC precursors is identified, it could be an ideal target for induction from ESCs/ iPSCs for the generation of vascularised kidney organoids. While CD31+ EC-like cells were detected in human iPSC-derived kidney organoids 2 , their inefficient contribution to the vasculature upon transplantation 8 may suggest an incomplete quality of the induced CD31+ ECs or an absence of the CD31− EC precursors. The ECs in the organoids may be derived from the intermediate mesoderm, because they are induced simultaneously with the kidney 2 . However, the precise origin and developmental pathway of physiological renal ECs remain largely unknown. Elucidation of these issues, combined with the present assay, will facilitate the induction of fully competent renal ECs from ESCs/iPSCs.
One of the advantages of our protocol is that arteriolar network formation can be observed along with extensive UB branching, the latter of which is lacking in most other published methods. Furthermore, nephrons are formed efficiently both in vitro and after transplantation, the latter of which achieved glomerular vascularisation. The failure of arteriolar character acquisition and glomerular vascularisation in vitro may arise through suboptimal culture conditions, including culture medium components, oxygen concentrations, and flow-mediated physical force. While new techniques such as organ-on-a-chip may solve these issues in the future 31 , transplantation appears to satisfy most of the necessary conditions. In particular, blood flow, which is required for EC development and maintenance [32][33][34] , is likely to promote development and re-organisation of the newly synthesised vasculature. We detected anuclear red blood cells around the UBs in the transplanted organoids ( Supplementary  Fig. S3), suggesting that the vasculature of the organoids is likely to be connected to the host circulation. However, dye injection into the host circulation should be performed for formal proof.  The transplanted organoids still differed from kidneys in vivo. Our data showed that the trunk of the extrarenal artery was formed around E13.5 in vivo, but was not observed in the transplanted kidney organoids. The branching patterns of intrarenal arterioles also differed from those in vivo. In addition, ureters were shortened within 3 days upon aggregation (prior to transplantation), and re-elongation of ureters was not observed upon transplantation (see Hoxb7-GFP UBs in Fig. 3-5). These may partly arise because we simply reaggregated NPs and SPs as bulk populations that could have lost their topological information. In particular, SPs consist of multiple populations located in distinct regions 35 , and preservation of this positional information in organoids would be necessary for recapitulation of the complex renal vascular development, as well as formation of a single elongated ureter. Ultimately, the existence of flow in the vasculature and urine production should be examined to prove the functions of the reconstituted kidneys.
In conclusion, our protocol, which can reconstruct the overall kidney and vascular structures from dissociated and sorted cells, will serve as a useful basis for the identification of renal EC precursors and induction of such precursors from ESC/iPSCs, which will eventually lead to the generation of vascularised and functional kidney organoids.
Transplantation of kidney organoids. Reaggregated kidney organoids at day 3 of in vitro culture were transplanted as previously described 8 . Because the mouse strains used as donor animals were maintained on a hybrid background, immunodeficient NOD/SCID mice were used as the host animals, and anaesthetised by peritoneal administration of normal saline containing 0.75 mg/kg medetomidine, 4.0 mg/kg midazolam, and 5.0 mg/ kg butorphanol 36 . After surgery, atipamezole was administered as an anaesthetic antagonist. The grafted organoids were harvested at 7 days after transplantation (day 10).

Immunostaining of sections.
Paraffin sections were subjected to antigen retrieval in citrate buffer. Next, paraffin or frozen sections were washed three times with PBS and blocked by incubation with 1% BSA in PBS for 1 h at room temperature. The sections were incubated overnight with primary antibodies at 4 °C, followed by incubation with secondary antibodies conjugated with Alexa Fluor 488, 568, 594, 633, or 647 for 90 min at room temperature. Nuclei were counterstained with 4,6-diamidino-2-phenylindole (Roche). The following primary antibodies were used: chick anti-GFP (Abcam; ab13970); rabbit anti-Cx40 (Alpha Diagnostic; CX40-A); mouse anti-E-cadherin (BD Biosciences; 610181); rabbit anti-CD31 (Abcam; ab28364); rat anti-CD31 (BD Biosciences; 557355); rabbit anti-Six2 (Proteintech; 11562-1-AP); rabbit anti-red fluorescent protein (RFP) (Rockland; 600-401-379); rat anti-TER-119 (MBL; D062-3); and guinea pig anti-nephrin (Progen Biotechnik; GP-N2). Fluorescence images were captured by a confocal microscope (TSC SP8; Leica). For colouring of arteries and veins, paraffin blocks of E12.5 and E13.5 embryos were serially sectioned and stained with an anti-CD31 antibody (Abcam; ab28364) using a BlueMap Kit and a Discovery System (Roche). Each section was scanned with a slide scanner (SCN400; Leica), and images of the same regions were selected for alignment with an ImageJ plugin (Stackreg). CD31+ structures that were continuous with the aorta in each section were coloured in red, while those continuous with the large veins were coloured in blue using Photoshop software (Adobe Systems Inc.). The resulting images were reconstituted into three-dimensional images using Imaris software (Bitplane).
Quantification and statistical analysis. Student's t-test was applied for statistical analysis of differences between two groups. Differences with values of p < 0.01 were considered statistically significant.