A Novel Three–Dimensional Human Peritubular Microvascular System

BASIC RESEARCH www.jasn.org

Giovanni Ligresti,*† Ryan J. Nagao,* Jun Xue,* Yoon Jung Choi,* Jin Xu,* Shuyu Ren,† Takahide Aburatani,† Susan K. Anderson,† James W. MacDonald,‡ Theo K. Bammler,‡ Stephen M. Schwartz,§ Kimberly A. Muczynski,† Jeremy S. Dufﬁeld,†| Jonathan Himmelfarb,†| and Ying Zheng*¶

Departments of *Bioengineering, †Medicine, ‡Environmental and Occupational Health Sciences, and §Pathology, |Kidney Research Institute, and ¶Institute for Stem Cell and Regenerative Medicine, University of Washington, Seattle, Washington

ABSTRACT Human kidney peritubular capillaries are particularly susceptible to injury, resulting in dysregulated angiogenesis, capillary rarefaction and regression, and progressive loss of kidney function. However, little is known about the structure and function of human kidney microvasculature. Here, we isolated, purified, and characterized human kidney peritubular microvascular endothelial cells (HKMECs) and reconstituted a three-dimensional human kidney microvasculature in a flow-directed microphysiologic system. By combining epithelial cell depletion and cell culture in media with high concentrations of vascular endothelial growth factor, we obtained HKMECs of high purity in large quantity. Unlike other endothelial cells, isolated HKMECs depended on high vascular endothelial growth factor concentration for survival and growth and exhibited high tubulogenic but low angiogenic potential. Furthermore, HKMECs had a different transcriptional profile. Under flow, HKMECs formed a thin fenestrated endothelium with a functional permeability barrier. In conclusion, this three- dimensional HKMEC-specific microphysiologic system recapitulates human kidney microvascular structure and function and shows phenotypic characteristics different from those of other microvascular endothelial cells.

J Am Soc Nephrol 27: 2370–2381, 2016. doi: 10.1681/ASN.2015070747

The kidneys play an essential role in the body microvessels are highly susceptible to rarefaction af- to eliminate harmful substances from blood, in- ter exposure to toxins, xenobiotics, or injury.9,10 After cluding endogenous metabolic waste products, injured, they exhibit limited regenerative capacity, exogenously administered xenobiotics, and envi- which may contribute to tissue ischemia, tubular dys- ronmental toxins. As a major recipient of cardiac function, inflammation, fibrosis, and the develop- output (approximately 25%) and the primary filter ment of CKD.11 of exogenous drugs and toxins, kidneys are highly vascular, and the tubulointerstitium is particularly Received July 8, 2015. Accepted October 29, 2015. susceptible to injury, clinically resulting in AKI1 G.L. and R.J.N. contributed equally to this work. and contributing to the incidence and progression of CKD.2,3 Present addresses: Dr. Giovanni Ligresti, Department of Physiology & Biomedical Engineering, Mayo Clinic, Rochester, Minnesota; and Of the two major components in the kidney Dr. Jeremy S. Duffield, Biogen Idec, Cambridge, Massachusetts. tubulointerstitium, the kidney microvasculature has Published online ahead of print. Publication date available at received relatively less attention in human studies. www.jasn.org. These vessels play a critical role in delivering nutrients Correspondence: Dr. Jonathan Himmelfarb, 325 Ninth Avenue, to tubular epithelial cells, possess unique transport Seattle, WA 98104, or Dr. Ying Zheng, 850 Republican Street, – properties,4 7 and participate in the tubular secretion Seattle, WA 98109. Email: [email protected] or yingzy@ and reabsorption of solutes.8 Studies over the past uw.edu decade have shown that kidney peritubular Copyright © 2016 by the American Society of Nephrology

2370 ISSN : 1046-6673/2708-2370 JAmSocNephrol27: 2370–2381, 2016 www.jasn.org BASIC RESEARCH

The mechanisms underlying the microvascular response to fetal kidneys at that stage contained established nephrons, peritubular kidney injury, however, remain unclear, in part including glomeruli, tubules, and interstitium (Figure 1A), and because of difficulties resulting from in vivo imaging as well as have begun to produce urine. Compared with the mature adult challenges in isolating human kidney microvascular cells for in kidneys (Figure 1B), fetal kidneys showed signs of ongoing vitro study.12 Although glomerular endothelial cells have been development, exemplified by having a higher cellularity successfully isolated and characterized,13,14 little progress has [(1.6660.04)3106 cells per mm3 in fetal versus (0.6160.08) been made on human kidney peritubular microvascular cells. 3106 cells per mm3 in adults] and smaller glomeruli diameter Much of our understanding of kidney capillary formation and (73.75617.5 mm in fetal versus 166.25622.5 mm in the adult). maintenance has been extrapolated from the study of other en- In the interstitium of fetal kidneys, the microvasculature has dothelial cells,9,15 which may not capture specific properties of established a network around the tubules, and endothelial cells the human kidney peritubular microvasculature. New evidence strongly express CD31 (Figure 1A.2). The adult kidneys have from genetic fate–mapping studies in mice suggests that the fully established tubular structure surrounded with a robust microvascular endothelium of the internal organs may not arise peritubular microvascular network that strongly expresses from a single–yolk sac progenitor as was originally thought but CD31 (Figure 1B.2), VE Cadherin, and CD34 (Supplemental rather, from discrete organ–specific mesenchymal cells that ap- Figure 1, A and B). This peritubular microvascular endothe- pear early in embryogenesis and subsequently, give rise to mul- lium is enveloped by a scattered layer of PDGFRb+ stroma tiple organ–specific populations, including the endothelium.16 (Supplemental Figure 1, C and D). In addition, the peritubular Odd Skipped–Related 1–positive progenitors likely give rise to microvascular endothelium was distinguished from the glo- all cell populations in the kidney, including the microvascular merular endothelium by low granular expression of vWF and endothelium. This suggests that, rather than the endothelium high expression of a plasmalemma protein (PV1) delineating being imposed on by the organ developing around it, organ- fenestral diaphragms (Figure 1, A.3 and B.3). In adult kidneys, specific characteristics might be intrinsic to the endothelium.17 glomerular endothelium did not express PV1, whereas partial Another important characteristic of the kidney microvascu- staining of PV1 was still present in the fetal kidney glomeruli, lature istheconstantsubjectionto highblood flow andtransport. suggesting the development stage of the fetal kidneys with im- Conventional planar cultures of endothelial cells fail to recreate mature glomerular capillaries. This is consistent with previous the in vivo physiology of the microvasculature with respect to the observations that PV1 is present in the glomerular capillaries three-dimensional (3D) geometry (lumen and axial branching) during early development but no longer expressed in postnatal and the interactions of the endothelium with blood flow and glomerular endothelium.19 extracellular matrix. To address these challenges, we have re- The endothelial cell population (CD31+CD452) in human cently engineered functional vascular networks on the basis of kidneys was identified with flow cytometric analysis and microfluidic design principles that permit precise control of accounted for 3.161.5% of cells in fetal kidney tissues (Figure vascular cell types, branching architecture, lumen diameter, 1C) and 2.262.1% of cells in adults (Figure 1D). The direct and flow dynamics.18 This approach allows us to now recon- sorting of this population from flow cytometry, however, struct human kidney microvessels under physiologic geometry resulted in limited cell survival, lack of attachment, and and flow conditions. impurity when cultured. Our modified enrichment protocol In this study, we present new methods to isolate, purify, and (Figure 2A) showed that depletion of epithelial cells and supple- expand human kidney peritubular microvascular endothelial cells menting with a high concentration of vascular endothelial (HKMECs) and recreate the kidney microvasculature with growth factor (VEGF) in culture media were critical for enhanc- appropriate geometry and flow. We show that HKMEC-formed ing endothelial cell growth (Figure 2B) and achieving purified microvessels have kidney-specific properties, exemplified by the HKMECs (VE Cadherin+/CD31+ CD452 PDGFRb2)inlarge presence of fenestral diaphragms on the endothelial membrane, quantities. low angiogenic potential, and increased sensitivity to flow-induced Isolated HKMECs showed expression of typical endothelial biophysical changes. These experiments indicate a functioning transcripts by RT-PCR, and expressed genes include PECAM, human kidney microvasculature can be recapitulated in vitro. VE Cadherin, VEGFR2, TIE2, vWF,andPDGF-BB as well as We discuss the potential application of our system for modeling genes restricted to microvascular ECs, including ROBO4 and nephrotoxicity as well as the onset and progression of kidney CD146 (Figure 2C, Supplemental Table 1). The purity was also disease. verified by showing that these cells lack CD45 and E Cadherin expression. Isolated HKMECs from both fetal and adult tissue formed sheets in two-dimensional (2D) culture (Figure 2, D RESULTS and E) with consistent expression of CD31 (Figure 2, D.1 and E.1) and VE Cadherin (Figure 2, D.3 and E.3) at cell-cell Purification, Expansion, and Characterization of contacts. Expression of Claudin-5 was also found near the HKMECs junctions between adjacent cells (Supplemental Figure 2); HKMECswerepurifiedfrombothadultnephrectomyspecimens however, this expression followed a sawtooth distribution typ- and fetal kidneys between 100 and 135 days postconception. The ically associated with the lack of tight junctions.20 Both fetal

J Am Soc Nephrol 27: 2370–2381, 2016 Peritubular Microvessels In Vitro 2371 BASIC RESEARCH www.jasn.org

Figure 1. HKMECs were characterized in human kidney peritubular microvessels in vivo. (A and B) Histologic images of (A) human fetal and (B) adult kidney tissues containing (A.1 and B.1) established nephron structures. In the interstitial stroma, microvascular endothelial cells strongly expressed (A.2 and B.2) CD31 (red) and (A.3 and B.3) PV1 (red) but not (A.3 and B.3) vWF (green). Blue indicates nuclei. (C and D) From flow cytometric analysis, roughly 1.67% and 0.77% of the total cell population are endothelial cells (CD31+CD452)in(C) fetal and (D) adult kidneys, respectively. G, glomeruli; S, stroma; T, tubule. Scale bar, 50 mm. and adult endothelial cells showed abundant PV1 (Figure 2, vessel networks displayed a vessel density nearly four- D.2 and E.2) and low vWF expression (Figure 2, D.4 and E.4), fold higher than vessel networks formed with HUVECs respectively. This expression was consistent between the tis- (Figure 3F). sue sources, with no apparent difference in the HKMECs In a separate experiment, we tested the angiogenic potential with respect to their morphology and surface markers. The of HKMECs compared with HUVECs (Figure 3, G and H). consistent PV1 expression was indicative of a peritubular HKMECs formed a confluent monolayer that was quiescent microvascular endothelial cell phenotype. In addition, these anddevoidofanysproutinginthepresenceof40ng/ml HKMECs expressed consistent VEGFR2 throughout the cell VEGF. In contrast, HUVECs readily formed tip cells and in- (Supplemental Figure 2). vaded the collagen matrix in response to VEGF. Quantification of the invasion frequency revealed a significantly higher num- HKMECs Are Highly Tubulogenic but Not Angiogenic ber of sprouts per area in HUVECs compared with HKMECs To understand the functional characteristics of HKMECs, we (Figure 3I). evaluated their capacity to self-assemble into complex tubular branching structures in 48 hours using a tubulogenic assay as Molecular Characteristics of HKMECs previously described.21,22 When exposed to VEGF at 40 ng/ml, Togain a better understanding of the molecular characteristics of HKMECs showed a high degree of tubulogenic activity evident HKMECs, we compared gene expression by microarray analysis from extensive complex 3D structures with connected net- of HKMECs and HUVECs cultured under the same conditions. works (Figure 3, A and B). These networks formed an enclosed We identified 1748 significantly differentially expressed genes lumen with an average diameter of roughly 25 mm (Figure 3A, between HKMECs and HUVECs using criteria of 1.5-fold arrows). Each connected segment was highly 3D and com- difference and a 0.05 false discovery rate. In HKMECs, 964 posed of hundreds of cells. In contrast, at identical culture genes are upregulated, whereas 784 genes are downregulated conditions, HUVECs formed vacuoles, with isolated tube- compared with HUVECs. These data are summarized on the like structures comprised of ,10 cells. These structures basis of fold difference using a list of representative genes (Table were attenuated and poorly branched, and no complex net- 1). Upregulation of genes, including CD34, PV1, ANGPT2, works were formed (Figure 3, C and D). The average vessel CXCL12, JAM2, and VEGFR2, and downregulation of genes, diameter was approximately 10 mm, significantly smaller (less such as MMP1 and RGS5, were verified by real-time PCR (Figure than one half) than that found in the HKMEC-formed vas- 4A). In particular, the expression of structural protein PV1 cular networks (Figure 3E). Additionally, HKMEC-formed was .10-fold higher in HKMECs than in HUVECs, consistent

2372 Journal of the American Society of Nephrology J Am Soc Nephrol 27: 2370–2381, 2016 www.jasn.org BASIC RESEARCH

Figure 2. HKMECs were isolated, purified, and characterized in vitro. (A) Summarized procedure for HKMEC enrichment. (B) Flow cytometric analysis of a single-cell suspension of isolated kidney cells after 5 days of culture indicating the proportion of endothelial cells in three distinct culture conditions: without VEGF-A (top panel), with 40 ng/ml VEGF-A (middle panel), and with VEGF-A and prior depletion of epithelial cells from the single-cell suspension (bottom panel). (C) RT-PCR confirmed the endothelial cell expression of PECAM, vWF, VE Cadherin, VEGFR2, TIE2, PDGF-BB, and the microvascular markers CD146 and ROBO4 and the absence of CD45 and E Cadherin. ECs: HKMECs; M: DNA marker; NC: Negative control. (D and E) Isolated cultured HKMECs show uniform morphology with purity .98%, and strongly express (D.1 and E.1) CD31, (D.2 and E.2) PV1, and (D.3 and E.3) VE Cadherin but not (D.4 and E.4) vWF in both (D) fetal and (E) adult kidneys. Scale bar, 50 mm. with the abundance of fenestrae observed when evaluating cell to injury. The significant downregulation of RGS524 and morphology (Figure 2D). ANGPT2 is known to be associated MMP125 and upregulation of DLL426 support a lack of angio- with vessel instability,23 and its upregulation in HKMECs may be genic potential for HKMECs, which is consistent with the lack of correlated to the susceptibility of microvessels within the kidney sprouting seen in response to VEGF (Figure 3, G–I).

J Am Soc Nephrol 27: 2370–2381, 2016 Peritubular Microvessels In Vitro 2373 BASIC RESEARCH www.jasn.org

significantly upregulated in HKMECs compared with HUVECs (with z score .2) (Figure 4C). These upregulated pathways in- cluded Tec Kinase signaling, Integrin signaling, Phospholipase C signaling, thrombin signaling, the role of NFAT, Gaq signaling, NGF signaling, NF-kB signaling, and the Wnt/Ca+ pathway. These data support the hypothesis that isolated HKMECs maintain intrinsic properties that are significantly distinct from HUVECs, despite identical culture conditions for .5days. These molecular signatures suggest that HKMECs intrinsically lack regenerative growth capabilities and angiogenic potential, which may be implicated in their susceptibility to and induction in kidney-specificinjury.

HKMECs Form Stable Microvessels and Adapt to Flow To recapitulate the physiologic conditions found in vivo, we engineered 3D microvascular networks in collagen gel using lithographic processes that we described Figure 3. HKMECs had different tubulogenic and angiogenic characteristics compared previously.18 HKMECs were seeded in with HUVECs. (A and C) Confocal images of vascular tube formation (tubulogenesis assay) the lumen of this microphysiologic system for (A) HKMECs and (C) HUVECs: xy plane (left panel) and (A.1, A.2, C.1, and C.2) two yz (MPS) and cultured under gravity-driven cross–sectional planes at the dashed lines. Arrows indicate the enclosed lumen. Scale fl – fl bar, 50 mm. (B and D) 3D reconstruction of vessel tubes in A and C with a 200-mm depth ow for 3 14 days to achieve a con uent in the z direction, showing the (B) connected microvascular network formed by HKMECs layer of endothelium on luminal walls – and (D) disconnected tubes formed by HUVECs. (E) Quantification of vessel diameters (Figure 5, A C). This endothelium ex- indicated that HKMECs formed connective networks with an average vessel diameter pressed CD31 and VE Cadherin at regions of approximately 25 mm, significantly larger than in HUVEC networks (approximately of cell-cell contact (Figure 5, D and E). 10 mm). ***P,0.001. (F) Quantification of vessel density within HKMEC networks was vWF granules became more abundant in around fourfold higher than that in HUVEC networks. (G and H) Confocal images of (G) HKMECs comprising an endothelium along HKMEC and (H) HUVEC monolayers remaining on the surface of a 2 mg/ml collagen microvessels compared with those cultured gel in the xy plane (left panel) and yz cross-sections at the dashed lines (right panel) under static 2D conditions (Figures 2, D and after 72 hours of culture. HUVECs readily sprouted into matrix, whereas HKMECs did E and 5F). A similar increase of vWF was also not. (I) Quantification of the number of sprouts per area for HKMECs and HUVECs. observed in HUVECs cultured in a 3D mi- croenvironment, which was also greater than To better understand the regulation of genes involved in HKMECs when cultured in the same manner (Figure 5F). In angiogenesis and tubulogenesis functions, we further analyzed addition, HKMECs were more sensitive to shear stress than differential gene expression using Ingenuity Pathways Analysis. HUVECs, determined by enhanced alignment with the direction There were 62 significantly upregulated and 116 significantly of flow in microvessels (Figure 5G). The same shear-sensitive downregulated genes that pertain to angiogenesis in HKMECs effect was also consistently found in mouse KMECs (Supple- compared with HUVECs (Figure 4B). Within tubulogenesis mental Figure 4). The recapitulated kidney microvessels also function, 69 genes were significantly upregulated, and 62 formed a barrier to the transfer of solutes from the lumen into genes were significantly downregulated (Supplemental Figure the matrix. A basement membrane consisting predominantly of 3). The upstream analysis revealed that the transcription fac- collagen 4 was found to be deposited by HKMECs along the tors TP53 and SMAD3 were activated in HKMECs, indicating vessel walls (Figure 5H). the inhibition of angiogenesis and an upregulation of TGFb We further measured the vessel permeability or molecular signaling. Furthermore, transcription factors, such as MYC sieving effect of a large molecule, 40-kD FITC-dextran, for and CCND1, were inhibited in HKMECs, indicating their in- both HKMECs and HUVECs microvessels. Perfusion of the hibited cell growth and proliferation potential compared with molecules was driven by the hydrostatic pressure drop through HUVECs. Through canonical pathway analysis, we also ob- the vessel. The oncotic pressure drop across the vessel wall was served 12 of a total of 407 canonical pathways that were approximately zero given the same culture media, and

2374 Journal of the American Society of Nephrology J Am Soc Nephrol 27: 2370–2381, 2016 www.jasn.org BASIC RESEARCH

Table 1. Selected genes that were significantly upregulated and downregulated in HKMECs compared with HUVECs in microarray studies HKMECs/HUVECs No. of Genes Representative Genes and Fold Changes Upregulation .10 184 RPS4Y1, CD34,a COL15A1, ITM2A, COL1A1, COL6A3, KCNAB1, NTNG1, SOX11, TUBB2B, EPDR1, COL1A2, GPR116, S100A4, PDGFRB, PLVAP,a CXCL12,a ANGPT2,a NOSTRIN, PRKAR1A, IGFBP3, PIEZO2, RCAN2, RBP1, JAM2,a TNFSF10, VCAM1, COL4A2, SPRY1, PDE2A, PDGFB, FGF12, CYBA, EDNRA, VAV3, SERP2, ATP1B1, KISS1, CLDN5, ITGA11, ITGA3, PPARG, IL1B, THBS2 5–10 242 FBLN1, NOTCH3, PTP4A3, FGF13, WNT5A, DLL4,a ADAMTS7, COL3A1, SOCS2, LRP10, LAMB1, FABP5P3, PGF, CDH13, NOS2, NRARP, SPP1 3–5 429 TNFRSF25, NGF, SELE, RARA, MAPK7, KDR,a JAG1, FLT4, HAS2, MMP15, CDH5 1.5–3 119 ITGB1BP1, RPS6KA2, PRKCD, HIF1A, ICAM1, FAS, CXCR4, RHOB, NRP2, AKT1, S1PR1, MMP14, MMP7, VEGFB, DLL1 Downregulation ,1/10 121 SERPINB2, SPOCK1, MGARP, IFI27, RGS5,a EFEMP1, BST2, TSPAN8, SULT1B1, GRB14, TACSTD2, IL1RL1, NRG1, NRCAM, ADAM23, PLAT, MMP1,a HHIP, MMP10, PTGS2, NTSR1, NLGN1, S1PR3,VEGFC, RARB, SLIT2, NOTCH2NL, LRP5 1/10–1/5 180 PDGFA, GPRC5A, GPR37, CDC45, RRM2, RGS2, FOXC1, FOXC2, CD44,a BMP4, TNFSF4, CDH2, PTGER4, NREP, ADAMTS1, TFPI, CDCA2, CDC25A, ATP5G1, NOTCH2, RL18R1 1/5–1/3 404 PDGFC, CTNNB1, RGS4, RGS11, CDC27, HSPA4L, VEZT, TJP2, STK26, ACVR1B, CDC20, FLNC, NRG1, CKAP2L, SOX4, ITGB3, RAC1, GATA3, PTGFRN 1/3–1/1.5 105 HSPB8, YAP1, VEGFA, CDC42, ATP2A2, ANGPT1, IGFLR1 aThese genes are verified in real time quantitative PCR, shown in Figure 4. thus, protein content was saturated on the abluminal and (Supplemental Figure 5, B and C). The presence of abundant luminal sides of the endothelium. Focal permeation of 40-kD fenestrae in HKMECs microvessels increases the total area per- FITC-dextran occurred along the endothelium (Figure 5I), missible to transport molecules moving across the vessel wall, leading to an increased spatially averaged permeability, which which likely contributed to the increase of vessel permeability represents the increased convective flux of the molecules across compared with HUVEC microvessels. the endothelium. We used the distribution of fluorescence intensity during the transient flow of dextran to estimate the averaged permeability coefficient of the kidney endothelium DISCUSSION 2 to be K=0.1660.06 mms 1, including the focal leaky regions. These permeability coefficients were approximately five times There is currently great interest in developing microfluidic higher than that observed in HUVEC-formed microvessels. organs on chips, including the vasculature, which carry the The reconstructed kidney microvessels strongly express the potential to create predictive models of human disease and plasmalemma protein, PV1, throughout the endothelium. The interrogate the organ-specific physiologic and pathophysio- expression of PV1 was ubiquitously distributed throughout the logic responses to drugs and environmental toxins.28 Within cytoskeleton in the cell periphery (Figure 6A), whereas PV1 was the kidney is a complex interplay between the tubular and not detectable in HUVECs vessels (Supplemental Figure 4). Ul- vascular units comprising filtration and resorption. Specifi- trastructural examination of these engineered kidney microves- cally, the peritubular microvessels are highly susceptible to sels substantiated the immunohistochemistry, because HKMEC rarefaction after exposure to toxins, xenobiotics, or injury. endothelium displayed abundant closed fenestratae along the Characterizing the relevant cells and recreating a representa- microvessel walls (Figure 5, B–D). HKMECs formed a thin layer tive microvasculature are crucial for the understanding of along the luminal wall and appeared to be polarized between the drug transport, drug-induced nephrotoxicity, and capillary lumen and the matrix. Junctions formed at focal contacts of rarefaction in progressive kidney diseases. Although compar- lining HKMECs (red arrows in Figure 5, B and C), numerous atively more attention has been focused on the glomerulus, fenestrated diaphragms were clearly visible, and all contained a the peritubular capillaries represent an understudied and central knob (black arrows in Figure 5, B and D). The average size important physiologic system to investigate given their critical of fenestrae was 59611 nm, which is similar to the size of fenes- importance. In this work, we developed methods to isolate, trae in kidney microvessels in vivo (approximately 62–68 nm).27 purify, and culture HKMECs. The success of the isolation The ultrastructure of HUVEC-formed microvessels revealed a and subsequent culture were particularly dependent on continuous endothelium lacking fenestrae but rich in caveolae, VEGF concentration and epithelial cell depletion before en- which may be a result of the high VEGF culture conditions richment and flow cytometric sorting. HKMECs exhibited

J Am Soc Nephrol 27: 2370–2381, 2016 Peritubular Microvessels In Vitro 2375 BASIC RESEARCH www.jasn.org

Figure 4. Molecular signatures were intrinsically different between HKMEC and HUVECs. (A) Real–time quantitative PCR on selective genes (bold) verified similar trends in downregulated genes MMP2 and RGS5 and upregulated genes CD34, ANGPT2, CXCL12, DLL4, JAM2, KDR, PDGFB,andPV1. (B) Ingenuity pathway analysis for angiogenesis function with significantly changed gene expression comparing HKMECs versus HUVECs. Red indicates upregulation, and blue indicates downregulation. (C) Heat map of canonical pathway analysis of HKMECs compared with HUVECs showing the most significantly differentiated genes that belong to specific ontologies. standard endothelial surface markers, adherens junctions, differences from HUVECs in phenotype, gene expression pat- and displayed an endothelial cell gene expression pattern. tern, and functional assays. Microarray analysis revealed that In addition, they maintained kidney microvascular-specific the intrinsic transcriptional signatures of HKMECs were dif- structures, including closed fenestrae, characterized by the ferent from those in HUVECs, with more tubulogenic but less plasmalemma protein, PV1. HKMECs exhibited marked angiogenic potential.

2376 Journal of the American Society of Nephrology J Am Soc Nephrol 27: 2370–2381, 2016 www.jasn.org BASIC RESEARCH

Figure 5. Human kidney peritubular microvessels were reconstructed in vitro. (A and B) Schematic diagram of 3D MPS set up with (A) 3D views and (B) cross-sections. (C) Example of kidney microvessel networks generated in a 3D MPS. Red indicates CD31, and blue indicates nuclei. (D) z-Stack projection of confocal image of engineered human kidney microvessel in xy plane (left panel) and cross-sectional yz plane at the dashed line (right panel). Red indicates CD31, green indicates vWF, and blue indicates nuclei. Scale bar, 50 mm. (E) z-Stack projection of confocal image of human kidney microvessels at a junction of the network. Red indicates VE Cadherin, and blue indicates nuclei. Scale bar, 25 mm. (F) Quantification of the amount of vWF per area for both HKMECs and HUVECs and in both 2D static culture and 3D flow-based microvessel cultures. (G) Quantification of SI (4pA/P2)forHKMECsin3D microvessels and 2D static cultures and HUVECs in 3D microvessels. (H) Immunofluorescence image of a cryosectioned HKMEC vessel network (thickness of 7 mm). Red indicates CD31, green indicates collagen IV, and blue indicates nuclei. Scale bar, 100 mm. (I) Fluorescence image of 40-kD FITC-dextran perfused through the HKMECs after 1 minute of perfusion. Arrows show good barrier, and stars show focal leakage. Scale bar, 100 mm.

J Am Soc Nephrol 27: 2370–2381, 2016 Peritubular Microvessels In Vitro 2377 BASIC RESEARCH www.jasn.org

Figure 6. Reconstructed human kidney peritubular microvessels were fenestrated in vitro.(A)z-Stack projection of confocal image of engineered human kidney microvessel at a junction of vessel network (left panel) and in a zoomed view (right panel). Red indicates F-actin, green indicates PV1, and blue indicates nuclei. (B–D) Transmission electron microscopy reveals the ultrastructure of HKMECs microvessels containing proper junctions (red arrows) formed at cell-cell contacts between (B and C) two adjacent cells C1 and C2 and (B and D) numerous fenestrae (black arrows) throughout the peripheral regions.

We then reconstructed a 3D human kidney microvascular Because the loss of kidney peritubular capillaries is common to network to assess the structure and function of human kidney kidney disease progression from almost all inciting mecha- peritubular microvessels under flow. We showed that the nisms of injury, this approach provides a high content system HKMECs in 3D MPS altered their alignment along flow- for identifying new therapeutic approaches to preventing directed pathways and increased granular vWF compared with human kidney failure. In addition, this approach opens new the same cells in static 2D culture conditions. HKMECs formed possibilities for mechanistic studies of heterogeneity in in- comprehensive barriers to large dextran molecules and de- terindividual susceptibility to kidney injury. posited physiologically observed basement membrane pro- teins on the vessel walls. Our data indicate that HKMECs in a 3D flow-directed MPS exhibit features highly representative of CONCISE METHODS the kidney microvasculature in vivo, while simultaneously re- capitulating standard microvascular properties. Future studies Kidney Tissue will be required to further define the unique transport prop- Adult kidney tissues were obtained from normal renal cortex of erties of these fenestrated endothelia for different size mole- nephrectomies performed for renal masses or transitional cell cules and under varying flow and pressure conditions. carcinoma. Fetal human kidneys were obtained after voluntary Our data show that the use of organ-specific microvascular pregnancy interruptions performed at the University of Washington endothelial cells in a 3D MPS can provide unique insights into Medical Center in compliance with Institutional Review Board the properties of the peritubular microvasculature, which protocol (IRB447773EA). Informed consents for the use of fetal appear to closely approximate human anatomyand physiology. tissues were obtained from patients.

2378 Journal of the American Society of Nephrology J Am Soc Nephrol 27: 2370–2381, 2016 www.jasn.org BASIC RESEARCH

Endothelial Cell Isolation, Enrichment, and Culture density of 53106 cells per ml were delivered to the inlet of an aspi- Kidney tissues were processed mechanically followed by enzymatic rated channel. Flow was allowed by replenishing media in the inlet dissociation and sieving through a cell strainer to remove glomeruli reservoir twice per day. and large vessels to obtain a single-cell suspension. The removal of fi glomeruli was con rmed by microscopic examination, showing that Imaging and Quantifications almost all glomeruli are present in the cell strainer, and no glomeruli The immunofluorescence images of the cells and tissue sections on fi were found in the single-cell suspension. Toobtainasuf cientnumber slides, 3D gels, or intact microvessels in situ were taken using a Nikon of cells with high purity, epithelial cells were depleted from the single- A1R Confocal Microscope (Nikon, Tokyo, Japan). Image stacks were fi cell suspension using a magnetic immunoaf nity column, and the rest accumulated with a z step between successive optical slices of ap- – of cells was cultured on a gelatin matrix for 72 96 hours in customized proximately 2 mm. Cross-sections, projections, and 3D reconstruc- endothelial cell proliferation media with high concentration of VEGF. tions were generated from z stacks of images using ImageJ software This expanded cell population was then resuspended and subjected to with orthogonal projection, z projection, and 3D viewer. The ultra- fl 2 2 ow-cytometric sorting for a VE Cadherin+/CD31+ CD45 PDGFRb structured images were acquired on ultrathin after–processed vessel fi subpopulation. After sorting, the puri ed cells were immediately cultured sections (70 nm) using a JEOL JEM-1400 Transmission Electron fi in conditions favoring endothelial cell growth for up to ve passages. Microscope (JEOL Ltd.) with a Gatan Ultrascan 1000XP Camera The cells were further characterized by morphology, immunostaining, (Gatan, Inc., Pleasanton, CA). and PCR to verify the purity and for confirmation of the absence of glomeruli endothelial cell contamination. Details are described in Quantification of Vessel Diameter, Density, and Sprout Supplemental Material. Frequencies For tubulogenesis assay, image stacks of vascular network in 3D gels Tubulogenesis and Angiogenesis Assays were loaded in ImageJ, and the channels were split to obtain the 3D For tubulogenesis function,22 endothelial cells were mixed with col- stack of vessel walls. An intensity threshold was selected to filter the lagen type 1 solution to reach gel density of 2 mg/ml and cell density single-channel image stack, so that the vessel walls were preserved and of 23106 cells per ml. pipetted into customized 5-mm-diameter wells well defined. The enclosed vessel lumen was then filled, and a binary of precast polydimethylsiloxane (PDMS) dishes, and allowed to gel at stack output was generated after the filtering and hole-filling 37 °C and 5% CO2 for 30 minutes. Cultures were then kept in endo- functions. The vessel density was calculated by summing the nonzero thelial growth media with 40 ng/ml VEGF and allowed to assemble areas in the stacks and normalizing with the total stack volume. The over time. The collagen invasion (angiogenesis) assay was performed vessel diameter of each stack image was measured for three well as described.22,29 Briefly, 5-mm-diameter wells of PDMS dishes were defined connected regions along the direction perpendicular to the filled with rat tail collagen to reach a flat surface and allowed to axial direction of the tubes. Two image stacks were analyzed for each polymerize for 20 minute at 37 °C. After polymerization, endothelial replicate. For angiogenesis assays, the number of sprouts per imaging cells were plated on top of each gel at density of 400 cells per mm2 field was counted and converted to that per mm2. Three images were and cultured with complete EBM-2 endothelial medium. The next counted to reach an averaged value for each replicate. For both assays, day, the cultures were rinsed twice with EBM-2 serum free and in- three to five replicates were processed for analysis. cubated with 40 ng/ml VEGF in endothelial complete medium. After – fi 48 72 hours, cultures were xed in 4% formaldehyde solution, Quantification of vWF Expression stained, and imaged under microscope. In 2D and 3D experiments, vWF expression was taken by setting a binary threshold on the vWF immunofluorescence channel. The area Microvessel Fabrication and Culture of vWF corresponded to the number of pixels counted in the image A neutralized liquid collagen solution was prepared at 7.5 mg/ml usingImageJ. This number wasdivided by the totalnumberof pixelsin from 15 mg/ml collagen stock solution for microvessel fabrication the regionof interest and then, multiplied by100 to give the percentage as described previously.18 Briefly, collagen was injected into a PEI/ of vWF present in a given area of cells. Three images were analyzed for glutaraldehyde–treated Plexiglas housing top half that formed a cav- three separate devices. Data for each device were then averaged and ity along with an oxygen plasma–sterilized PDMS stamp, thereby used to calculate SEM with n=3. forming a negative impression of a microvessel network. Inlet and outlet ports were formed by inserting stainless steel dowel pins before Quantification of Shape Index of Cells in Microvessels injecting the collagen. The Plexiglas bottom half consisted of a flat Shape index (SI) was calculated according to the equation SI=4pA/P2. layer of collagen compressed by a flat PDMS surface that was gelled on Each cell was traced and measured using ImageJ to output its perim- top of a standard coverslip. The collagen was allowed to gel for 30 eter and area. Nine cells were randomly chosen and calculated for minutes at 37 °C. After gelation, the PDMS stamp on the top piece each of three separate devices. The SI for each device was averaged was removed, revealing an open microfluidic network. The PDMS and then, used to calculate SEM with n=3. on the flat layer of collagen on the Plexiglas bottom half was used to seal the microvessel network. Culture medium was then added to the Quantification of Vessel Permeability reservoirs and incubated for 2 hours before cell seeding. To seed the The permeability of kidney microvessels was measured by delivering vessels, 10-mlinjectionsofHKMECsorHUVECsfromP1toP5ata 40-kD FITC-dextran through the endothelialized microchannel at a

J Am Soc Nephrol 27: 2370–2381, 2016 Peritubular Microvessels In Vitro 2379 BASIC RESEARCH www.jasn.org

2 continuous flow rate of approximately 5 mlmin 1.Fluorescence 6. Ojteg G, Nygren K, Wolgast M: Permeability of renal capillaries. II. images were acquired sequentially, and image analysis was carried Transport of neutral and charged protein molecular probes. Acta – out in MATLAB with detailed mathematics described previously18 to Physiol Scand 129: 287 294, 1987 7. Källskog O, Wolgast M: Driving forces over the peritubular capillary obtain the permeability of the vessel wall. Six regions of interest were membrane in the rat kidney during antidiuresis and saline expansion. randomly chosen for each image to obtain an averaged value for Acta Physiol Scand 89: 116–125, 1973 permeability coefficient. The fluorescence intensity at the center of 8. Jen K-Y, Haragsim L, Laszik ZG: Kidney microvasculature in health and interchannel space between endothelialized channels was measured disease. Contrib Nephrol 169: 51–72, 2011 with time to extract the permeability coefficient of the vessel wall. 9. Lin SL, Chang FC, Schrimpf C, Chen YT, Wu CF, Wu VC, Chiang WC, Kuhnert F, Kuo CJ, Chen YM, Wu KD, Tsai TJ, Duffield JS: Targeting endothelium-pericyte cross talk by inhibiting VEGF receptor signaling Statistical Analyses attenuates kidney microvascular rarefaction and fibrosis. Am J Pathol For all quantitative measurements, the entire population was used to 178: 911–923, 2011 calculate statistical significance, whereas mean values with n=3 were 10. Basile DP: The endothelial cell in ischemic acute kidney injury: Impli- used to calculate SEM and graphical confidence intervals. An F test cations for acute and chronic function. Kidney Int 72: 151–156, 2007 11. Basile DP: Rarefaction of peritubular capillaries following ischemic was used to determine variance equality. A two–tailed t test was then acute renal failure: A potential factor predisposing to progressive ne- performed on each group of interest. On each graph, error bars re- phropathy. Curr Opin Nephrol Hypertens 13: 1–7, 2004 present 62 SEM (a 95% confidence interval). For P values ,0.05, a 12. Muczynski KA, Ekle DM, Coder DM, Anderson SK: Normal human single asterisk over a solid line was used; for P values ,0.01, two kidney HLA-DR-expressing renal microvascular endothelial cells: asterisks were used, and for P values ,0.001, three asterisks were Characterization, isolation, and regulation of MHC class II expression. J – used. Am Soc Nephrol 14: 1336 1348, 2003 13. Ballermann BJ: Regulation of bovine glomerular endothelial cell Detailed methods of cell isolation, immunostaining, and molec- growth in vitro. Am J Physiol 256: C182–C189, 1989 ular characterization are described in Supplemental Material. 14. Marsden PA, Brock TA, Ballermann BJ: Glomerular endothelial cells respond to calcium-mobilizing agonists with release of EDRF. Am J Physiol 258: F1295–F1303, 1990 15. Kida Y, Ieronimakis N, Schrimpf C, Reyes M, Duffield JS: EphrinB2 re- ACKNOWLEDGMENTS verse signaling protects against capillary rarefaction and fibrosis after kidney injury. J Am Soc Nephrol 24: 559–572, 2013 We acknowledge the Microfabrication Facility and the Lynn and Mike 16. Taguchi A, Kaku Y, Ohmori T, Sharmin S, Ogawa M, Sasaki H, Nishinakamura R: Redefining the in vivo origin of metanephric nephron Garvey Imaging Core at the Universityof Washingtonand the Electron progenitors enables generation of complex kidney structures from Microscope Facility and Genomics Department at the Fred Hutch- pluripotent stem cells. Cell Stem Cell 14: 53–67, 2014 inson Cancer Research Institute. We thank Drs. Ed Kelly, Kenneth 17. Abrahamson DR, Robert B, Hyink DP, St John PL, Daniel TO: Origins Thummel, and Danny Shen for helpful discussions. and formation of microvasculature in the developing kidney. Kidney Int – This project was supported by National Institutes of Health Grants Suppl 67: S7 S11, 1998 18. Zheng Y, Chen J, Craven M, Choi NW, Totorica S, Diaz-Santana A, UH2/UH3 TR000504 (to J.H.), P30-ES07033 (to J.W.M. and T.K.B.), Kermani P, Hempstead B, Fischbach-Teschl C, López JA, Stroock AD: In DK93493 (to J.S.D.), and DP2DK102258 (to Y.Z.) and American Heart vitro microvessels for the study of angiogenesis and thrombosis. Proc Association Grants 12040023 (to J.S.D.) and 12SDG9230006 (to Y.Z.). Natl Acad Sci U S A 109: 9342–9347, 2012 19. Ichimura K, Stan RV, Kurihara H, Sakai T: Glomerular endothelial cells form diaphragms during development and pathologic conditions. J Am Soc Nephrol 19: 1463–1471, 2008 DISCLOSURES 20. Kluger MS, Clark PR, Tellides G, Gerke V, Pober JS: Claudin-5 controls None. intercellular barriers of human dermal microvascular but not human umbilical vein endothelial cells. Arterioscler Thromb Vasc Biol 33: 489– 500, 2013 21. Koh W, Stratman AN, Sacharidou A, Davis GE: In vitro three di- REFERENCES mensional collagen matrix models of endothelial lumen formation during vasculogenesis and angiogenesis . Methods Enzymol 443: 83– 1. Himmelfarb J, Ikizler TA: Acute kidney injury: Changing lexicography, 101, 2008 definitions, and epidemiology. Kidney Int 71: 971–976, 2007 22. Cross VL, Zheng Y, Won Choi N, Verbridge SS, Sutermaster BA, 2. Coresh J, Selvin E, Stevens LA, Manzi J, Kusek JW, Eggers P, Van Lente Bonassar LJ, Fischbach C, Stroock AD: Dense type I collagen matrices F, Levey AS: Prevalence of chronic kidney disease in the United States. that support cellular remodeling and microfabrication for studies of JAMA 298: 2038–2047, 2007 tumor angiogenesis and vasculogenesis in vitro. Biomaterials 31: 3. Chawla LS, Eggers PW, Star RA, Kimmel PL: Acute kidney injury and 8596–8607, 2010 chronic kidney disease as interconnected syndromes. NEnglJMed 23. Maisonpierre PC, Suri C, Jones PF, Bartunkova S, Wiegand SJ, 371: 58–66, 2014 Radziejewski C, Compton D, McClain J, Aldrich TH, Papadopoulos N, 4. Brenner BM, Falchuk KH, Keimowitz RI, Berliner RW: The relationship Daly TJ, Davis S, Sato TN, Yancopoulos GD: Angiopoietin-2, a natural between peritubular capillary protein concentration and fluid re- antagonist for Tie2 that disrupts in vivo angiogenesis. Science 277: 55– absorption by the renal proximal tubule. J Clin Invest 48: 1519–1531, 60, 1997 1969 24. Jin Y, An X, Ye Z, Cully B, Wu J, Li J: RGS5, a hypoxia-inducible apo- 5. LeBrie SJ: Renal peritubular capillary permeability to macromolecules. ptotic stimulator in endothelial cells. J Biol Chem 284: 23436–23443, Am J Physiol 213: 1225–1232, 1967 2009

2380 Journal of the American Society of Nephrology J Am Soc Nephrol 27: 2370–2381, 2016 www.jasn.org BASIC RESEARCH

25. Moses MA: The regulation of neovascularization of matrix metal- 29. Ligresti G, Aplin AC, Dunn BE, Morishita A, Nicosia RF: The acute phase loproteinases and their inhibitors. Stem Cells 15: 180–189, 1997 reactant orosomucoid-1 is a bimodal regulator of angiogenesis with 26. Leslie JD, Ariza-McNaughton L, Bermange AL, McAdow R, Johnson SL, time- and context-dependent inhibitory and stimulatory properties. Lewis J: Endothelial signalling by the Notch ligand Delta-like 4 restricts PLoS One 7: e41387, 2012 angiogenesis. Development 134: 839–844, 2007 27. Stan RV: Endothelial stomatal and fenestral diaphragms in normal vessels and angiogenesis. JCellMolMed11: 621–643, 2007 28. Bhatia SN, Ingber DE: Microﬂuidic organs-on-chips. Nat Biotechnol 32: This article contains supplemental material online at http://jasn.asnjournals. 760–772, 2014 org/lookup/suppl/doi:10.1681/ASN.2015070747/-/DCSupplemental.

J Am Soc Nephrol 27: 2370–2381, 2016 Peritubular Microvessels In Vitro 2381