pericyte timp3 and adamts1 modulate vascular stability after

BASIC RESEARCH www.jasn.org

Pericyte TIMP3 and ADAMTS1 Modulate VascularStability after Kidney Injury

Claudia Schrimpf,*†‡ Cuiyan Xin,†‡ Gabriella Campanholle,†‡ Sean E. Gill,‡ William Stallcup,§

Shuei-Liong Lin,| George E. Davis,¶ Sina A. Gharib,‡ Benjamin D. Humphreys,*and Jeremy S. Duffield*†‡

*Renal Division, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts; †Division ofNephrology and ‡Center for Lung Biology, Institute for Stem Cell and Regenerative Medicine, University ofWashington, Seattle, Washington; §Cancer Center, Sanford-Burnham Medical Research Institute, La Jolla, California;|National Taiwan University Hospital, Taipei, Taiwan; and ¶Department of Medical Pharmacology and Physiology, DaltonCardiovascular Research Center, University of Missouri, Columbia, Missouri

ABSTRACTKidney pericytes are progenitors of scar-forming interstitial myofibroblasts that appear after injury. Thefunction of kidney pericytes as microvascular cells and how these cells detach from peritubular capillariesand migrate to the interstitial space, however, are poorly understood. Here, we used an unbiased ap-proach to identify genes in kidney pericytes relevant to detachment and differentiation in response toinjury in vivo, with a particular focus on genes regulating proteolytic activity and angiogenesis. Kidneypericytes rapidly activated expression of a disintegrin andmetalloprotease with thrombospondinmotifs-1(ADAMTS1) and downregulated its inhibitor, tissue inhibitor ofmetalloproteinase 3 (TIMP3) in response toinjury. Similarly to brain pericytes, kidney pericytes bound to and stabilized capillary tube networks inthree-dimensional gels and inhibited metalloproteolytic activity and angiogenic signaling in endothelialcells. In contrast, myofibroblasts did not have these vascular stabilizing functions despite their derivationfrom kidney pericytes. Pericyte-derived TIMP3 stabilized and ADAMTS1 destabilized the capillary tubularnetworks. Furthermore, mice deficient in Timp3 had a spontaneousmicrovascular phenotype in the kidneyresulting from overactivated pericytes and were more susceptible to injury-stimulated microvascular rar-efaction with an exuberant fibrotic response. Taken together, these data support functions for kidneypericytes inmicrovascular stability, highlight central roles for regulators of extracellular proteolytic activityin capillary homoeostasis, and identify ADAMTS1 as a marker of activation of kidney pericytes.

J Am Soc Nephrol 23: 868–883, 2012. doi: 10.1681/ASN.2011080851

Recent studies have identified pericytes of the peri-tubular capillaries of the kidney as a major precursorpopulation of scar-forming myofibroblasts,1,2 andstudies from embryonic development and cancergrowth demonstrate vital angiogenic and vascularstabilizing functions for pericytes.3,4 Prompted bythese studies, we sought to understand the role ofkidney pericytes in homeostasis and the molecularmechanisms regulating pericyte attachment tocapillaries in the kidney and detachment andmigra-tion from capillaries in response to injury.

Pericytes aremesenchyme-derivedmural cells ofcapillaries that have processes embedded withincapillary basement membrane and they may playa direct role in conjunction with endothelial cells

(ECs) to stimulate vascular basement membranematrix assembly.5,6 Each tissue bed has pericytes de-rived from organ-specified mesenchyme,7 and thusthere are likely to be organ-specific differences inpericyte functions that are poorly understood. The

Received August 24, 2011. Accepted January 9, 2012.

Published online ahead of print. Publication date available atwww.jasn.org.

Correspondence: Dr. Jeremy S. Duffield, Department of Ne-phrology and Lung Biology, Institute for Stem Cell and Re-generative Medicine, University of Washington, 815 MercerStreet, Seattle, WA 98105. Email: [email protected]

Copyright © 2012 by the American Society of Nephrology

868 ISSN : 1046-6673/2305-868 J Am Soc Nephrol 23: 868–883, 2012

http://www.jasn.org

mailto:[email protected]

direct interaction of pericyte processes with endothelium is be-lieved to be the location of cell: cell signaling.5 Pericytes of thekidney derive during embryogenesis from a layer of embryoniccapmetanephric mesenchyme lying outside of the mesenchymethat specifies epithelium. This mesenchyme activates the fork-head transcription factor Foxd1, specifically during embryonicspecification.8 In normal adult kidney, pericytes are Foxd1-derived, PDGF receptor b (PDGFRb+), CD73+, Coll1a1+,CD45-, a smooth muscle actin (aSMA-) cells.2,9

During angiogenesis in other organs, heparan sulfate–tethered PDGF-BB released by tip ECs is thought to signalvia PDGFRb to promote pericyte migration along the form-ing capillary and stable attachment to endothelial cells.10

However, signaling via pericyte PDGFRb in formed kidneyvessels, in response to microvascular injury, stimulates pericytedetachment, migration, and differentiation into myofibro-blasts.9,11 Similarly, vascular endothelial growth factor A(VEGFA) signaling from pericytes to angioblasts in organ de-velopment is important in successful angiogenesis; however,in response to injury, expression of alternate VEGFA splicevariant by kidney pericytes stimulates pericyte detachmentand differentiation into myofibroblasts.9,12 Therefore, it islikely that multiple signaling pathways and extracellular inter-actions that are poorly explored are important in determiningpericyte attachment, migration, and activation.

The kidney is particularly susceptible to ischemic and toxicinjuries, due in part to the concentrating functions of thenephron and to its unusual vasculature. The kidney vascula-ture comprises a primary capillary network in the glomer-uli specialized for filtration followed by a secondary capillaryplexus that supplies the remainder of the kidney, resorbs sol-utes, and uptakes water being returned to the circulation. Thespecialized functions of this secondary capillary plexus, peri-tubular capillaries (PTCs), are facilitated by its low oxygentension, with low flow.Moreover, the secondary plexus is at thevagariesof theglomerular plexus.Because thekidneycortex andmedulla perform complex transport functions, they havesignificant oxygen and nutrient demands. Hence, the kidneyis susceptible to ischemia if PTC flow/supply is insufficient.Subtle kidney ischemia due to loss of PTCs is widely believedto be a major cause of essential hypertension and a major driv-ing force in the progression of CKD and in the failure of AKIto resolve.13–15

Previous studies, particularly in cancer growth and woundrepair, have implicated proteinase-mediated degradation ofthe capillary basement membrane and surrounding extracel-lular matrix as critical for vascular instability and vascular re-gression (rarefaction).16,17 Moreover, proteinases and factorsthat regulate their activity are now recognized to serve as keyregulators of cell migration and spreading in cancer biologyand development, model systems that may be instructive instudying kidney injury.18–20

In this study, we investigated the role of kidney pericytes invascular stabilization and the central role that pericytes play inregulating protease activity of the vasculature.

RESULTS

Microarray Analyses of Pericytes in Early InjuryResponses in Kidney Identify Enriched Pathways andRegulated GenesTo explore mechanisms that orchestrate attachment and de-tachment of pericytes, we used an unbiased approach to identifyregulated genes and activated pathways. Most biologic processesentail coordinated interaction among functionally coherentgroups of genes known as modules.21 We postulated that peri-cyte differentiation is characterized by the temporal activation ofdistinct biologic modules and that key regulators of this processreside within these coherent pathways. We purified pericytesfrom normal Coll-GFP kidneys (day 0) by flow cytometry andat early time points after onset of injury (day 2 and day 7) in themurine unilateral ureteral obstruction (UUO) model in whichpericyte detachment iswell defined, usingwell establishedmeth-ods of cell purification1,22–24 (Supplemental Figure 1) and as-sessed their transcriptome bymicroarray analysis. TheColl-GFPreporter mouse used in this disease model specifically labelspericytes and myofibroblasts, and there are no significantfibrocytes contaminating the collection (Supplemental Fig-ure 1).1,22–24 In addition to pericytes, small numbers of peri-vascular fibroblasts are collected because they also expressColl-GFP, but these represent ,5% of all cells expressing GFPunder regulation of the collagen1a1 promoter (Coll-GFP+cells) in normal kidney.1 Principal component analysis of thetranscriptomes revealed grouping of biologic replicates thatcorresponded to the time of injury, implying global perturba-tions in temporal gene expression as pericytes detach and dif-ferentiate into myofibroblasts (Supplemental Figure 2A). Morethan 800 genes were differentially expressed when control peri-cytes (day 0) were compared with postinjury myofibroblasts(days 2 and 7) (Supplemental Figure 2B and SupplementalTable 1). To elucidate the temporal patterns of gene expres-sion and pathway activation in response to injury, we im-plemented a clustering algorithm optimized for analysis ofshort time series25 and identified two highly significant geneclusters—each mapping to very distinct functional categories(Figure 1). One cluster corresponded to upregulated genes andwas highly enriched in processes involved inmatrix production,cell proliferation,migration, innate immune responses, and che-motaxis. In addition, genes mapping to angiogenesis and vascu-lar development were members of this upregulated group. Thesecond cluster was composed of genes downregulated duringinjury, and was functionally enriched in processes involved inmitochondrial function, redox status, and metabolism. A smallnumber of genes regulating extracellular matrix production andangiogenesis were also downregulated. A disintegrin and metal-loproteinase with thrombospondin motif-1 (ADAMTS1) washighly significantly upregulated early after injury onset andamong the 15 most upregulated genes early after injury in peri-cytes (Figure 1 and Supplemental Figure 2C). Moreover, it was astrong candidate effectormolecule in regulating detachment andmigration. ADAMTS1 has been reported to cleave capillary

J Am Soc Nephrol 23: 868–883, 2012 Pericytes Stabilize Capillaries 869

www.jasn.org BASIC RESEARCH

http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2011080851/-/DCSupplemental







basementmembraneproteins and inhibit angiogenesis and there-fore may reasonably facilitate pericyte detachment.26,27 Tissueinhibitor of matrixmetalloproteinase 3 (TIMP3), an endogenousinhibitor of ADAMTS1, is highly expressed at baseline in peri-cytes28 but is significantly downregulated as disease progresses(Figure 1 and Supplemental Figure 2C). TIMP3 has been dem-onstrated or implicated in silencing the activityofmetalloprotein-ases including ADAM17 and some matrix metalloproteinases

(MMPs) such as MMP14, and has been reported to regulate vas-culogenesis and angiogenesis and to promote tissue regenerationafter injury in the lung.29–31 Thus, the downregulation of TIMP3as pericytes transition to myofibroblasts in vivomay have impor-tant consequences for the kidney microvasculature. Given themarked regulation of these two functionally related genes andcandidacy as regulators of pericyte function in response to injury,we aimed to study their function in kidney pericytes.

Figure 1. Global analysis of kidney pericyte transcriptome in response to UUO injury. Temporal changes in gene expression map to twodistinct clusters characterized by progressively increasing or decreasing patterns. In each graph, the dark gray line corresponds to theaverage change in gene expression in response to injury relative to baseline expression (day 0) and the light gray lines identify the 90thand 10th percentiles of each group. The temporal expression patterns of two genes, Adamts1 and Timp3, are depicted with red lines.Gene members of each cluster underwent functional analysis based on Gene Ontology database annotations. Enriched functionalcategories among upregulated and downregulated genes are shown using color-coded bars corresponding to enrichment P valueswith the most highly significant shown in red and the least significant shown in dark blue.

870 Journal of the American Society of Nephrology J Am Soc Nephrol 23: 868–883, 2012



Kidney Pericytes Prevent Capillary Tubular NetworkCollapseTo study pericyte function, we generated primary pure culturesbymagnetic bead immunoaffinity frommouse kidneys. Primarymouse pericytes expressed typical pericyte markers (Figure 2, Aand B) and morphology, which share characteristics with bonemarrow mesenchymal stromal cells and lacked epithelial, endo-thelial, and leukocyte markers and lacked the podocyte markersWT1 and synaptopodin.32,33 Compared with primary culturesof myofibroblasts, pericytes expressed more angiogenic factorsincluding Angiopoietin-1 (Angpt1) and Angiopoietin-2 (Angpt2),fewer Collagen1a1, and intermediate filament transcripts.

One of the defining features of pericytes is their ability to bindto and stabilize capillaries.12,34–36 We tested their capacity tostabilize a capillary tubular network formedby three-dimensional(3D) EC culture. Human ECs spontaneously form a capillarytube network in collagen gel (Figure 3, A–C). The capillary tubenetwork can be destabilized and undergoes regression in thepresence of coagulation proteases, which function by a numberof mechanisms including activation of endothelial matrix metal-loproteinases.37,38 Because the lumina of the forming capillarytubes do not contain any collagen gel, tube regression can be vi-sualized by gel retraction (collapse) (Figure 3D). In 3D co-culture,mouse kidney pericytes bind to EC capillary tubes (Figure 3Eand Supplemental Movie 1). Moreover, the cells have long pro-cesses that form invaginations on the tube abluminal surfacesimilar to the peg and socket processes seen by electron micros-copy.9,39 When EC capillary tubes were exposed to the serineproteinase kallikrein (KLK), both capillary tube regression andgel collapse were stimulated in a time- and concentration-dependent manner (Figure 3F). In the absence of KLK, EC cap-illary tubes did not regress.When kidney pericyteswere added tothe gel in a 10%/20%ratio (1PC for every 10 ECs/2 PCs for every10 ECs), the collapse of the gel in the presence of KLK was re-tarded (Figure 3G). When 30% (3 PCs for every 10 ECs) kidneypericytes were co-cultured, vascular regression was completelyprevented and the gel did not collapse (Figure 3G). Similar ob-servationsweremadewhen humanvascular brain pericytes weresubstituted for kidney pericytes (Figure 3H); however, kidneymyofibroblasts, which are derived in vivo from kidney pericytes,did not stabilize the capillary network (Figure 3H). Kidney peri-cytes suppressed the amount of active EC-derived MMP9 in thecollagen gels as determined by gelatin zymography (Figure 3I),and suppressed activation (phosphorylation) at the EC-restricted vascular endothelial growth factor receptor 2(VEGFR2) (Figure 3J) triggered by KLK. These observations in-dicate that kidney pericytes and brain pericytes have similar ca-pacity to stabilize capillary tubes, and that pericytes lose vascularstabilizing functions when they differentiate into myofibroblasts.

TIMP3 and ADAMTS1 Are Regulated during KidneyPericyte Detachment In VivoThe microarray studies (Figure 1) indicated that Timp3 tran-scripts were highly expressed in kidney pericytes and down-regulated in response to kidney injury. By contrast, among a

number of metalloproteinases that were regulated in pericytesin response to injury (Table 1), Adamts1 transcripts werehighly regulated early and most markedly. We postulatedthat the dynamic regulation of these kidney pericyte genesserved as important determinants of the detachment of peri-cytes and vascular instability that occurs in response to kidneyinjury.1,9,40 To validate the microarray findings, we performedin situ hybridization and immunofluorescence studies to lo-calize gene expression (Figure 4, A and B). Timp3 transcriptswere readily seen in kidney pericytes of normal adult mousekidney. Timp3 transcripts were detectable at day 2 after kidneyinjury onset in the interstitium. At day 7 after injury onset, theinterstitium in which myofibroblasts are located was almostdevoid of signal for Timp3 (Figure 4A). Timp3 expression wasseen in some injured epithelium and also localized to podo-cytes, but notmesangium in the glomerulus (Figure 4A). Coll-GFP+ cells were purified from normal kidneys (pericytes) orkidneys day 7 after disease (myofibroblasts). Timp3 transcriptswere downregulated in Coll-GFP+–purified myofibroblastscompared with Coll-GFP+ pericytes purified from normalkidneys (Figure 4C). Consistent with these observations, pri-mary cultures of kidney pericytes generated higher levels ofTIMP3 protein compared with primary cultures of myofibro-blasts (Figure 4D). By contrast, in situ hybridization of normaladult kidney did not detect any transcripts for Adamts1 (Figure4A). However, by day 2 after kidney injury, Adamts1 was readilydetected both interstitial cells and epithelial cells, and it was seenin epithelium and to a lesser extent in interstitial cells on day 7(Figure 4A). Immunostaining of kidney tissues with antibodiesagainst ADAMTS1 revealed a similar pattern (Figure 4B) withexpression in the vast majority (83%65.6%) of Coll-GFP+ cellsat day 2 after injury onset and 96%66.1% after day 7. ADAMTS1proteinwas also clearly detected in epithelial cells, predominantlyat the apical surface in the region of the brush border. As ex-pected, when kidney pericytes or myofibroblasts were purifiedfrom normal or diseased (day 2) kidneys, respectively, Adamts1transcripts were highly upregulated (Figure 4E) in the myofi-broblasts. In keeping with this observation, primary culturesof pericytes expressed much lower levels of Adamts1 transcriptcompared with primary myofibroblasts (Figure 4F). To deter-mine the relationship between Timp3 transcripts and protein,immunoblots of whole kidney lysates confirmed the downregu-lation of TIMP3 protein levels in mouse kidney in response todisease (day 2) and culturedmyofibroblasts expressed lower lev-els of TIMP3 protein compared with pericytes (Figure 4G).

ADAMTS1 Destabilizes and TIMP3 Stabilizes CapillaryTubular NetworksTIMP3 has multiple biologic activities that may contribute tovascular stability. Foremost, it inhibits activity of a number ofmetalloproteinases includingADAMTS1andendothelialMMPs.In addition, in high concentrations, it may regulate signalingdirectly at endothelial VEGFR2.28,29,41To explore the role ofTIMP3 in vascular integrity further, we applied recombinantTIMP3 to the capillary tube regression assay. In lower




Figure 2. Characterization of primary kidney pericyte cultures. (A) Fluorescent confocal images of kidney pericyte cultures for typicalpericyte markers and markers of endothelium or podocytes (bar=25 mm). (B) Histogram plots and dot plot from FACS analysis of primarykidney pericyte cultures for typical pericyte markers and markers of potential contaminating cells (gray areas are the isotype controlsand white areas are the specific antibody). (C) Quantitative PCR for typical pericyte markers and myofibroblast markers in primarykidney pericyte compared with myofibroblast cultures (n=3–6/group; ***P,0.001).



Figure 3. Kidney pericytes stabilize capillary tubes in a 3D gel assay. (A) Schema showing the addition of ECs (red) to gel in wells thatspontaneously form capillary networks. Addition of pericytes to this assay permits migration and binding of pericytes to capillary tubes. (B)Toluidine blue–stained gel showing capillary tube network (ECs only) within the gel (bar=100 mm). (C) Kidney pericyte (GFP+) in culture(bar=25 mm). Note cell processes extending the length of several cell bodies. (D) Low-power light images of gels containing capillary tubes(ECs only). Under the influence of the coagulation cascade serine protease, KLK, endothelial tubes are destabilized in the gel and vesselsbecome disorganized leading to progressive collapse of the gel (bar=100 mm). (E) Confocal image of 3D gel with YZ and XZ stacks showingcapillary tube (red, CD34) and kidney pericyte (green, Coll-GFP). Note the attachment of kidney pericytes to the capillary tube and nu-merous processes attached to the capillary tube. Z stacks (arrowheads) show yellow color at points of direct interaction of pericyte pro-cesses with capillary tubes, suggestive of peg and socket junctions (bar=25 mm). (F) Dose-response curves measuring collapse of collagengel (ECs only) induced by KLK. (G) Gel collapse curves induced by KLK (625 ng/ml) in the presence of increasing numbers of kidneypericytes. Note 30% kidney pericytes completely prevent gel collapse (n=12–16/timepoint). (H) Gel collapse curves in response to KLK (625ng/ml) in the presence of 30% human vascular brain pericytes (HVBPs) or 30% mouse kidney myofibroblasts (kMF). (I) Gelatin zymographyof supernatants from collagen gels in the presence of KLK showing multiple bands of gelatinase activity. Note that an approximately 72-kDband of activity, representing MMP9, is completely lost by the addition of 20% or 30% pericytes and the major approximately 60-kD bandof MMP2 is attenuated by the presence of 30% pericytes. (J) Phosphoblot for proteins from the collagen gels of KLK-activated gels de-tecting phospho-VEGFR2. Note that kidney pericytes complete suppress signaling at VEGFR2.



concentrations, TIMP3 retarded capillary tube regression; how-ever, at higher concentrations, it completely blocked capillarytube regression (Figure 5, A and B). Thus, TIMP3 can com-pletely replace pericyte function (Figure 3) in this assay. Bycontrast, application of recombinant ADAMTS1 to the vascu-lar assay in the absence of pericytes weakly accelerated capillarytube regression in the presence of KLK (Figure 5C), but did nottrigger regression (not shown). Much more strikingly, how-ever, is the finding that kidney pericytes were no longer ableto stabilize capillary tubes in the capillary tube regression assayin the presence of recombinant ADAMTS1 (Figure 5D). Thesefindings indicate that ADAMTS1 was sufficient to overcomethe vascular stabilizing functions of pericytes (Figure 5D). Totest specifically whether TIMP3 in pericytes stabilizes capillarytubes, TIMP3 protein expression was silenced using specificsmall interfering RNA (siRNA) (Figure 5E). Using sequentialsilencing methods, a .70% reduction in protein level wasachieved in pericytes. Whereas pericytes with control genesilencing completely stabilized capillary tubes, pericytes withTIMP3 silencing showed a nonsignificant, weakly impaired ca-pacity to stabilize capillary tubes during KLK-mediated regres-sion (Figure 5F), in keeping with pericyte-derived TIMP3playing a direct role in vascular stability.

TIMP3 Mutant Mice Have a Kidney MicrovascularPhenotypeOur findings suggested that TIMP3 may play a role in vivo inregulating vascular stability in the kidney and perhaps morebroadly in pericyte function. Timp3-deficient mice surviveinto adult life but have been noted to have microvascular defectsin the retina and in cancer growth, and also fail to resolve lunginflammation after injury. Furthermore, recent reports suggestthatTimp3-deficientmice have kidneyfibrosis.31,42,43We thereforeinvestigated the microvasculature of normal Timp32/2 kidneys,the response of the kidney microvasculature to acute ischemia

reperfusion injury (IRI), and the function of pericytes in thesevascular responses. Adult Timp32/2 mice were smaller thanlittermate controls but had normal renal function and no al-buminuria (Table 2). The microvascular density of kidneyPTCs of adult mice, assessed by two different methods, wasincreased with areas of abnormal capillaries (Figure 6, A andE and Supplemental Figure 3) and a greater number of PTCECs were in the cell cycle (Figure 6F). There were discrete ex-pansions of Pdgfrb+, CD73+, aSMA2 microvascular peri-cytes in PTC areas and also aSMA+ myofibroblasts in theTimp32/2 kidneys (Figure 6B) not seen in Timp3+/+ kidneys.Small increases of interstitial fibrillar collagen deposition weredetected in normal Timp32/2 adult kidneys (Figure 6, C, H, andI), confirming previous reports.42 Furthermore, unlike pericytesof wild-type normal kidneys, pericytes of Timp32/2 normalkidneys expressed ADAMTS1 (Figure 6, D, J, and L). In addi-tion to the interstitial changes, there were mild changes toglomeruli of Timp32/2 normal kidneys comprising mildmesangial hypercellularity and matrix expansion (not shown).Given that themajor site of expression of Timp3 in the kidney isthe pericyte, these findings collectively suggest that Timp32/2

pericytes in vivo are more activated.To determine whether Timp32/2 pericytes were intrinsically

overactivated as opposed to activated in response to vascularchanges, we studied primary cultures of pericytes in prolifera-tion andmigration assays (Figure 6M) and expression ofaSMA.Timp32/2 pericytes showed enhanced aSMA expression,increased proliferation, and enhanced motility, in keepingwith an overactivated phenotype. Enhanced migration byTimp32/2 pericytes was cell specific because cultured Timp32/2

macrophages migrated normally (not shown).To test the effect of injury responses in themicrovasculature,

we used a mild kidney IRI model that has a well recognizedmicrovascular injury, rarefaction, followed by incompleteangiogenesis.9,40,44 We observed that ischemic injury resultedin a greater rarefaction of the vasculature in Timp32/2 kidneyscompared with wild-type kidneys in this model (Figure 6, Aand E), and IRI stimulated more ECs into the cell cycle inTimp32/2 kidneys (Figure 6F). In concert with this observation,there was excessive expansion of pericyte/myofibroblast popu-lation (Figure 6, B, G, H, and K) after IRI. Using the aSMAmarker to define pericyte transition to myofibroblasts, therewas increased expansion of these myofibroblasts, excessive ex-pression of aSMA in response to injury, and increased inter-stitial fibrosis in Timp32/2 kidneys compared with Timp3+/+

kidneys by day 5 after IRI (Figure 6, C, I, and K). After kidneyIRI, there was a persistent microvascular leak in Timp32/2

kidneys, as detected by Evans blue dye assay (Figure 6, N andO) and this leak extended to vascular beds outside of the kidney,suggesting that generalized microvascular injury responses inthesemice results in persistent increase inmicrovascular perme-ability. The enhanced microvascular leak suggests that pericytefunctions were more compromised in Timp32/2mice comparedwith Timp3+/+ mice after vascular injury. Consistent with thiscompromised pericyte function, phosphorylated VEGFR2 was

Table 1. Quantitative PCR of metalloproteinases andinhibitors regulated in pericytes during detachment atday 0, day 2, and day 7 after UUO injury

Metalloproteinasesand Inhibitors

Day 0Day 2 afterInjuryOnset

Day7 afterInjuryOnset

Adamts-1 0.8958 9.8249 10.3441a

Adamts-2 0.4564 0.6883 6.42167b

Adamts-4 0.2562 1.3983 1.94952Adamts-5 0.2412 0.4097 0.94389Mmp2 2.85 2.58 2.87Mmp7 0.33 0.87 1.48a

Mmp9 2.14 3.03 4.33b

Mmp11 1.0536 2.1585 7.43393Mmp14 4.92 12.96 22.00Mmp23 0.7695 0.8386 8.73210Timp-3 137.4332 66.172 45.0715a

Timp-1 0.1280 8.1431 26.6397a

aP,0.01.bP,0.05.




Figure 4. Characterization of TIMP3 and ADAMTS1 expression in pericytes, in vivo and in vitro. (A) In situ hybridization of kidneysections from normal kidney (day 0) and on day 2 and day 7 after injury onset (disease) for Timp3, Adamts1, and control epithelialmarker Ngfa. Positive-stained pericytes and interstitial cells are shown (arrows) and positive-stained epithelial cells are shown(arrowheads) (bar=50 mm). (B) Split panels showing immunostaining for Adamts1 in normal kidneys (day 0) and on day 2 and day 7 afterinjury kidney sections from Coll-GFP mice. Coll-GFP+ pericytes do not express Adamts1 (arrows) but early after injury onset (day 2)pericytes and cellular processes (arrowheads) now express Adamts1. Kidney tubule expression is also increased. (C and D) QuantitativePCR for Timp3 comparing (C) Coll-GFP+ purified pericytes (normal kidney) with Coll-GFP+ purified myofibroblasts (disease kidney day2) or (D) primary pericyte cultures with primary myofibroblast cultures. (E and F) Quantitative PCR for Adamts1 comparing (E) Coll-GFP+purified pericytes (normal kidney) with Coll-GFP+ purified myofibroblasts (disease kidney day 2) or (F) primary pericyte cultures withprimary myofibroblast cultures. (G) Western blot showing expression of TIMP3 (upper blot) in normal kidney compared with diseasedkidney (day 7) or (lower blot) pericyte cultures compared with myofibroblasts. Note that the 24-kD Timp3 band is expressed at higherlevels in pericytes (n=3–5/group; *P, 0.05, ***P,0.001).



increased in the injured Timp32/2 kidneys compared with in-jured Timp3+/+ kidneys indicative of destabilized microvascu-lature surrounded by overactivated pericytes (Figure 6P).9,12

Because these studies indicated that the injury responses inTimp32/2 kidneys resulted in deleterious microvascular re-sponses, we testedwhether acute IRI resulted in greater functionaldeficiency. Twenty four hours after 21 minutes of kidney IRI,Timp32/2 mice (n=6) had plasma creatinine levels of 0.4961.2mg/dl, whereasTimp31/1mice (n=6) hadplasma creatinine levelsof 0.3160.03 mg/dl (P=0.06). Mice exposed to sham surgery hadcreatinine levels of 0.1760.03 and 0.1560.02 mg/dl, respectively.

DISCUSSION

These studies describe novel vascular stabilizing functions forkidney pericytes, and identify genes involved in regulation ofpericyte detachment and activation. Specifically, they identifythe protease regulator TIMP3 and one of its targets, ADAMTS1,

as highly regulated kidney pericyte genes that play roles inmicrovascular stability of the peritubular capillaries. Moreover,ADAMTS1 expression is a marker of activated pericytes.

Tissue ischemia is increasinglybelieved tobeacentralproblemin failure to regenerate the kidney normally after injury, a drivingforce for progressive organ dysfunction after injury, and a causeforhypertension andpossibly cardiovasculardisease.45Central tokidney tissue ischemia is loss of the microvasculature, which isalso known as capillary rarefaction. The recent identification of arelatively large population of pericytes in the cortex andmedullaof normal kidney and the observation that they detach fromcapillaries in response to injury, migrate into the interstitialspace, and transition to scar-forming myofibroblasts have raisedthe possibility that chronic diseases in the kidney are a stateof functional pericyte deficiency. Studies have suggested that in-jury of the kidney is a state of vascular instability that can easilylead to rarefaction despite endothelial proliferation and that fac-tors that prevent pericyte detachment can prevent this instabil-ity.9,11 The fact that the findings presented here reveal kidney

Figure 5. Characterization of TIMP3 and ADAMTS1 functions on capillary tube stability in a 3D tube assay. (A and B) Graph of KLK-inducedcapillary tube regression (ECs only) showing (A) retardation of collapse in the presence of low concentrations of recombinant TIMP3 and (B)complete prevention of collapse (ECs only) in the presence of high concentrations of recombinant TIMP3 (doses stated in ng/ml). (C) Graph ofKLK-induced capillary tube regression (ECs only) showing modest acceleration by recombinant ADAMTS1 (dose stated in ng/ml). (D) Graphof KLK-induced capillary tube regression in the presence of kidney pericytes and recombinant ADAMTS1. Thirty percent pericyte addition isno longer able to stabilize capillaries when ADAMTS1 is present (doses stated in ng/ml). (E) Western blots showing Timp3 levels in pericytesat time points after introduction of siRNA by transfection. (F) Graph of KLK-induced capillary tube regression in the presence of 30% kidneypericytes after Timp3 silencing or mock silencing (n=12–16/timepoint; *P,0.05, **P,0.01, ***P,0.001).



pericytes to have a similar capacity to pericytes from otherorgans to prevent microvascular instability ex vivo stronglysuggests that interventions that will promote pericyte attach-ment in the kidney will lead to preservation of capillaries. Inkeeping with the hypothesis that chronic diseases have func-tional pericyte deficiency due to pericyte transition to my-ofibroblasts, kidney myofibroblasts were unable to stabilizevessels in the capillary tube assay.

The microarray analysis indicated that there is a markedphenotypic change in pericytes when they become myofibro-blasts. The analysis we performed was based on collecting cellsthat express Coll-GFP. We previously characterized these cellsin the kidney and showed that they are pericytes. We identifiedvery few circulating fibrocytes (leukocytes that produce collagenmatrix) in this kidney injurymodel and our purificationmethodexcludes all leukocytes.1,2,9,39 We previously identified perivas-cular fibroblasts around arterioles of the kidney that also expressColl-GFPand expand to becomemyofibroblasts. These cells werealso purified along with pericytes because there is currently nomarker described to distinguish them. Although the microarrayanalysis requires validation due to potential contamination fromother kidney cells that may have been inadvertently collectedduring the flow cytometric sorting procedure, it is striking thatactivated pericytes are involved in innate immune responses, che-motaxis, and represent amajor source of cytokines. This is highlyrelevant because recent studies that target the pericyte directly byblocking PDGFRb or PDGFRa signaling in the kidney not onlyattenuate fibrosis and capillary rarefaction but also markedly at-tenuate the inflammatory response.11 Therefore, it is likely thatactivated pericytes are a major contributing cell to the inflamma-tory response. Further studies are required to understandwhether this response can be specifically targeted in the kidney.

Ourmicroarray studies identifiedTimp3 as a pericyte factorthat is lost during pericyte transition to myofibroblasts, and ourin vitro and in vivo studies showed TIMP3 to function as a mo-lecular factor promoting microvascular stability. AlthoughTIMP3 is an inhibitor of metalloproteinases, it is also a pleio-tropic, extracellular matrix–anchored molecule that is ableto inhibit a number of enzymes that breakdown capillary base-ment membrane, including ADAMTS1, and other matrixcomponents. In addition, TIMP3 is able to specifically inhibitsignaling at the endothelial receptor VEGFR2, initiated by allVEGFA isoforms. Interestingly, TIMP2 is also capable of in-hibiting VEGFR2 signaling in a manner independent of itsproteinase inhibitor activity, and previous work demonstrateda role for both pericyte-derived TIMP3 and EC-derived TIMP2in capillary tube stabilization.16 In recent studies, we identi-fied excessive signaling in response to injury at VEGFR2,

particularly in response to dysangiogenic VEGFA isoforms,released by activated pericytes/myofibroblasts in the kidney,as a major factor in microvascular instability.9,11 Therefore,“dialing down” the degree of signaling at this receptormay bedesirable. Our studies in vitro and in vivo indicate that kidneypericytes and TIMP3 function to limit endothelial VEGFR2 sig-naling potentially by both direct and indirect mechanisms.Timp3 mutation in humans is associated with retinal micro-vascular disease and Timp3-deficient mice exhibit increaseddysangiogenesis (increased angiogenesis with leaky, poorlyfunctioning vessels) in growth of some tumors and increasedangiogenesis in growth of others.29,46–49 Consistent with thesefindings, we identified dysangiogenesis in the kidney after injurywhen Timp3 was mutated, particularly in response to microvas-cular injury, and showed that pericytes are more activated andexhibit excessive migration and proliferation. Although furtherstudies are required, measurement of plasma creatinine in acuteinjury of Timp3-deficient kidneys suggests that TIMP3 mayserve to protect kidneys from acute injury, as we previouslyshowed for the lung.31 Further studies are required to under-stand whether TIMP3 regulates pericyte properties through reg-ulating MMPs or through regulating cleavage of other factorssuch as semaphorins.29 The finding that TIMP3 canmimic peri-cytes in vascular stabilization in vitro but that TIMP3 silencinghad aweak effect on pericyte function in vascular stabilization invitro was surprising but may be instructive. Timp3-deficientmice are born healthy and have a subtle microvascular pheno-type only. Therefore, it may not be surprising that only a modestdefect was seen in the short vascular stabilization assay. Thesefindings also indicate that pericytes rely on factors other thanTIMP3 to stabilize capillary tubes. These other factors are cur-rently unknown. In the capillary tube stabilization assay, peri-cytes inhibit MMP production primarily from ECs, but thesignal that mediates this inhibition is unknown. AlthoughTIMP3 may be a factor in this regulation, other factors mustbe contributing. TIMP3-deficient pericytes have a number offunctional abnormalities, including enhanced migration prolif-eration and activated phenotype as judged by aSMA expressionandADAMTS1 expression. It is unclear howTIMP3 regulates allof these processes, but these studies suggest that TIMP3must beable to affect transcription by either direct or indirect means.

Although a number of metalloproteinases includingADAMTS2 and MMP11 are activated in the microvasculaturein response to injury, Adamts1was most remarkably regulatedin pericytes of the kidney. ADAMTS1, like TIMP3, is highly teth-ered to the extracellular matrix and acts locally. We identifiedADAMTS1 regulation in epithelial cells, as well as pericytes, inthe kidney. ADAMTS1 expression in kidney development waspreviously reported.44,50 Because ADAMTS1 acts locally, the sig-nificance of epithelial expression of ADAMTS1, particularly, atthe brush border proximal tubule, is unclear. However, it is likelythat pericyte expression of ADAMTS1 early after injury exertsactivity at the pericyte-endothelial interface. ADAMTS1 has beenreported to cleave capillary basement membrane components,including aggrecan and versican; to play an anti-angiogenic role

Table 2. Characterization of Timp3-deficient mice

Timp3+/+ Male 12 wks Timp32/2 Male 12 wks

Weight (g) 29.960.9a 24.760.8Albuminuria ND ND

ND, not detected.aP,0.01.



Figure 6. TIMP3 deficiency in vivo predisposes to microvascular instability of kidney peritubular capillaries due to overactivatedpericytes. (A) Confocal images showing microvascular density and pericyte coverage of normal and day 5 post-IRI kidneys. (B) Confocalimages showing areas of expanded pericytes (Pdfgfr§+) in normal kidneys of Timp32/2 mice, some of which express aSMA (arrows) andsome of which do not (arrowheads). Five days after IRI, the pericyte/myofibroblasts (Pdgfrb+) show greater expansion and higher levelsof aSMA, particularly in Timp32/2 kidneys. (C) Low-power images of Sirius red–stained kidneys. (D) Fluorescence images of Adamts1.Note that normal Timp32/2 kidneys have pericytes that express Adamts1 (arrowheads). (E through K) Graphs showing quantificationof (E) vascular density, (F) endothelial cell proliferation, (G) pericytes, (H) myofibroblasts, and (I) fibrosis in interstitial (J) Adamts1+ cells.(K) Western blot showing aSMA and Pdgfrb expression in kidneys from Timp32/2 and Timp3+/+ mice. (L) Western blot quantifyingAdamts1 in kidney lysates after IRI. (M) Graphs showing primary pericyte cultures undergoing migration in a scratch closure assay,proliferation, and quantitative PCR expression of Acta2. (N and O) Images of (N) Evans blue–labeled mouse skin and sectioned kidneysand (O) Evans blue quantification indicating a microvascular leak 5 days after IRI. (P) pVEGFR2 expression in normal plus day 5 post-IRIkidneys (bar=50 mm; n=6–8/group; *P,0.05, **P,0.01, ***P,0.001). g, glomerulus; a, arteriole.



in both in vivo and in vitro angiogenesis assays; and to regulatemigratory capacity of a number of cells through cleavage of anumber of substrates, including syndecans and semaphorins. Itis therefore a protease thatmaymediate detachment of pericytes.Moreover, it selectively binds to and sequesters VEGFA164 butnot dysangiogenic VEGFA isomers.51 We recently showed thatwhen pericytes detach from capillaries in the kidney, they switchfrom VEGF164 to VEGF120 and VEGF188 production. Thelatter two VEGFA forms contribute to unstable angiogenesis.Therefore, it is likely that ADAMTS1 also promotes unstableangiogenesis by regulating VEGFR2-mediated signaling. Inkeeping with this, Timp3-deficient mice show overexpressionof ADAMTS1 in pericytes, have increased but dysfunctional vas-cularity under physiologic conditions, and have increased vesselrarefaction after injury compared with littermate controls. Ourstudies predict therefore that ADAMTS1 blockade would bene-ficially affect the kidney by limiting pericyte detachment afterinjury, promoting vessel stability, and reducing interstitialfibrosis.Adamts1mutantmicehavemicrovasculardefects aswell as a severekidney developmental phenotype resembling hydroureter.52,53

The severe hydroureter phenotype renders them unsuitable forstudy of pericyte function in response to injury. One possibleexplanation for the developmental kidney phenotype is im-pairedmigration of kidney stroma during nephrogenesis leadingto impaired nephron lengthening and papilla formation.53 Fur-ther studies will be required to determine whether Adamts1 is atargetable gene in kidney injury to prevent pericyte detachment.

In summary, kidney pericytes stabilize capillaries ex vivoand detach from capillaries in interstitial disease, and pericytefactors TIMP3 and ADAMTS1 are key regulators of these peri-cyte functions in vivo.

CONCISE METHODS

MaterialsAll reagents were from Sigma Aldrich, all plastic ware was from BD

Falcon, and all media were from Cellgro unless otherwise stated.

AnimalsColl1a1-GFP (Coll-GFP) transgenic mice were generated and validated

as previously described on the C57BL6 background.1,9,23 Timp32/2

gene–targeted homozygous mice and littermate controls were gen-

erated as previously described31,43 and maintained on the C57BL6

background.

Animal Models of Kidney InjuryUUO and unilateral and bilateral kidney IRI models were performed

in adult mice (aged 10–12 weeks) as previously described.1,22,23 The

ischemic time was 21 minutes. In some experiments, BrdU (50 mg/g)

or Evans blue dye (1%) was injected intraperitoneally or intravenously,

respectively, 1 hour before euthanasia. All animal experiments were

performed under protocols approved by the Animal Resources and

Comparative Medicine Committee of Harvard University and the De-

partment of Comparative Medicine at the University of Washington.

Pericyte or Myofibroblast Isolation from KidneyFor pericyte isolation from normal kidney of C57BL6 WTor Coll-GFP

mice, the kidney was decapsulated, diced, and then incubated at 37°C

for 30 minutes with Liberase (0.5 mg/ml; Roche) and DNase (100 U/ml;

Roche) in DMEM/F12. Digestion was curtailed by the addition of 10%

FCS to the single cell suspension. The suspension was filtered (40 mm)

several times, centrifuged (3003g, 5 minutes), and resuspended in

MACS (Miltenyi Biotech, Bergisch Gladbach, Germany) buffer or 1%

BSA buffer containing EDTA. In some experiments, cells were incu-

bated with anti-rabbit Pdgfrb polyclonal antibodies generated against a

segment of the full-length receptor54 (800 ng per kidney on ice) for 15

minutes. After washing, cells were incubated with goat anti-rabbit

IgG microbeads (Miltenyi Biotech) (15 minutes at 4°C) and resus-

pended and positive selection was performed by MACS magnetic

bead separation. Selected cells were resuspended in RLT buffer (Qia-

gen) or cultured in DMEM/F12 with 10% FBS and 1% penicillin/strep-

tomycin and 1% ITS, in six-well plates precoated with gelatin. Primary

cultured cells used in all experiments were between passages 2 and 5.

For other purposes, RNA was isolated using the RNeasy Mini Kit

(Qiagen). To note, this purification does not distinguish pericytes

from perivascular fibroblasts. In other experiments, single cells from

Coll-GFP mice were separated by green fluorescent protein fluores-

cence by FACS Aria cell sorting, directly into RLT buffer for RNA anal-

ysis or into medium (cell culture). In these experiments, kidney single

cell suspensions were also incubated with anti-CD45-APC, anti-CD31-

PE, anti-F4/80-APC,or anti-Tim1 antibodies (eBioscience) as de-

scribed1,22–24 to exclude these cell populations from contaminating

the Coll-GFP cells. To purify kidney myofibroblasts, single cell prepa-

ration and purification were achieved as described above but kidneys

from day 2 or day 7 after UUO were used.

Cell CultureHuman umbilical cord ECs (HUVECs) (American Type Culture

Collection) were used from P3 to P6 and grown with endothelial cell

basal medium (American Type Culture Collection) including one

Endothelial Cell Growth Kit BBE (American Type Culture Collection)

per 500mlmedium. Human brain pericytes (ScienCell) were cultured

in DMEM (Cellgro) containing 10% FCS (Invitrogen) and used from

P3 to P6. Primary mouse kidney pericytes (see below) were cultured on

0.1% gelatin-coated T75 flasks (BD Falcon) or six-well plates (BD

Falcon) and used from P2 to P5. Mouse kidney–derivedmyofibroblasts

were established as previously described1,9 and were maintained under

the same conditions as pericyte cultures and used from P3 to P7.

Three-Dimensional Vascular Tube Regression AssayECs were placed in 3D vascular tube regression assays following a

modified protocol of Saunders et al.30,37 HUVECs (106 cells/ml) were

suspended in 3.75mg/ml of rat tail collagen type I (4°C) (Invitrogen),

transferred to a 37°C incubator in which polymerization occurred,

and were allowed to form 3D networks in the presence of VEGF165

(40 ng/ml) and FGF2 (40 ng/ml) (Peprotech),38 in 384-well plates or

96-well 1/2 plates. For quantitative analysis of gel contraction, 15 ml

of ECs alone or ECs and varying amounts of primary kidney PCs

(as indicated) in collagen matrix were placed in 384-well plates (BD

Falcon) (n.10/condition). Addition of mediumwith or without hu-

man plasma KLK (Enzyme Research Laboratories) denoted the



starting point of the assay. Cultures were monitored every 4 hours for

gel contraction. Upon initiation of gel contraction, the regressed area

was quantified using an ocular grid. In every experiment, at least one

condition (n=5–10) with HUVECs in collagen matrix without the

addition of KLK was used as an internal control. In some experi-

ments, recombinant mouse Adamts1 or recombinant human

TIMP3 (R&D Systems) was applied to the media. For experiments

inwhich rmAdamts1 was applied, HUVECs were allowed to form tubes

for 24hours before additionofKLKand recombinant protein. Toluidine

blue staining of gels was performed as described previously.30 Condi-

tioned media or whole gels were collected to examine differential pro-

tein expression at various time points and were analyzed by SDS-PAGE

gels and Western blotting or zymography.

Pericyte Migration and Proliferation AssaysForty-eight-well plates (BD Falcon) were coated with 0.1% gelatin for

2 hours at 37°C. Cells were seeded with the same density of cells per

well and incubated with DMEM/F12 10% FCS until confluent. A

scratch “wound” was performed using a 200-ml pipet tip, wells were

washed twice with PBS, and DMEM/F12 with 2% FCS was applied. At

pre-fixed points along the scratches, photomicrographs were performed

at 0 hours and 12 hours after initialization of the wound. Migration of

cells and quantification of the wound area and its closure after 12 hours

were analyzed using ImageJ software. To quantify proliferation, 105

Timp32/2 and Timp3+/+ pericytes were seeded onto gelatin-coated glass

1.5-cm diameter cover slips (n=9 cover slips/group). BrdU (Sigma) (10

mM/ml) was applied to the medium and incubated for 3 hours. Cells

were fixed in acid ethanol, washed, and incubated with 2M HCL, then

0.1M sodium borate, and then 0.2% Triton in 3% BSA. After addi-

tional washing, cells were incubated with an anti-BrdU FITC anti-

body (1:200; eBioscience) in 3% BSA and mounted with Vectashield

and 4’,6-diamidino-2-phenylindole (DAPI) (Vector Laboratories), and

BrdU+ nuclei were quantified in 15 random fields per coverslip.

Gene Silencing in Primary Kidney Pericytes with siRNAsiRNA for Timp3 and control siRNA (Santa Cruz Biotechnology)

were resuspended in universal buffer and stored (280°C, 10 mM).

For efficient transfectionof primary kidney pericytes, cellswerewashed,

trypsinized, and harvested. A single cell suspension was generated

and cells were distributed into a six-well plate with 200,000 cells/well.

siRNA (final concentration of 100 nM/well) was resuspended in serum-

free medium, vortexed, and incubated for 10 minutes at room tem-

perature with HiPerfect Transfection Reagent (Qiagen). Medium

containing siRNA was added dropwise to the wells and plates were

swirled gently. Cells were incubated under normal conditions for 3

hours, followed by substitution of additional culture medium. After

48hours, cellswere analyzed for gene silencing. In experiments inwhich

TIMP3 was silenced, cells were transfected again with siRNA as de-

scribed above 48 hours after initial transfection to achieve higher levels

of gene silencing from 72 to 96 hours.

Microarray ExperimentsKidney pericytes or kidneymyofibroblasts were purified from 12-week-

old adult kidneys ofColl-GFPmice that had no surgery (pericytes, day 0,

n=2) or had undergone UUO surgery 2 days (n=2) or 7 days (n=1)

previously (myofibroblasts) by generating single cell preparation and

FACS sorting into RLT buffer as described in the section above. RNA

was purified by the RNeasy method (Qiagen). The quality of RNAwas

tested by a Agilent Bioanalyzer and high-quality samples underwent

RNA amplification by in vitro transcription methods at Harvard Uni-

versity Core Microarray Biotechnology Center. cRNA samples were

hybridized to CodeLink mouse whole-genome microarrays using the

manufacturer’s protocols (Applied Microarrays) and scanned using

AxonGenePix scanner (Molecular Devices). Image processing, back-

ground subtraction, and median-based normalization of probe set

intensities were performed using CodeLink Expression Analysis Soft-

ware (Applied Microarrays).

Microarray Data AnalysesPrincipal component analysis using the entire transcriptome of

pericytes and myofibroblasts was performed based on the covariance

matrix of normalized gene expression data55 (Supplemental Figure

2A). Temporal profiles of gene expression (day 0, day 2, and day 7) were

produced using Short Time-series Expression Miner clustering—

an algorithm developed for analyzing short time series microarray

data.25 Replicate gene expression values were averaged at day 0 and

day 2. Highly significant clusters (P values ranging from 5.431028 to

5.93102199) were merged if they displayed similar patterns, resulting

in two distinct temporal profiles (Figure 1). Functional analysis of the

genemembers of these two clusters was performed using the Database

for Annotation, Visualization, and Integrated Discovery, a web-based

program based on the Gene Ontology database.56 To demonstrate dif-

ferential gene expression between pericytes (day 0, no surgery) versus

myofibroblasts (day 2 and day 7 after UUO surgery), we applied a Bayes-

ian implementation of the parametric t test57 and corrected for multiple

comparisons using the Q value method.58 AQ value,0.01 was chosen

for identifying significant differential expression, and genes meeting

this cutoff underwent two-dimensional hierarchical clustering using

Pearson’s correlation (Supplemental Figure 2B and Supplemental

Table 1).55

Quantitative AnalysesWe used Western blot analysis to quantify protein concentrations in

tissues. Total cellular protein extracted using radioimmunoprecipitation

assay buffer containing Complete Mini Enzyme Inhibitor Cocktail

(RocheDiagnostics) fromwhole tissue, cells, or supernatants or collagen

gel fromwells of the vascular stabilization assay. Equal protein amounts

were subjected to SDS-PAGE and Western blotting using previously

described methods.59 The following primary antibodies were used to

detect the specific protein: anti-TIMP3-C-terminus (rabbit 1:100; Mil-

lipore), anti-ADAMTS-1 (rabbit 1:100; Santa Cruz Biotechnology),

anti-b-actin (goat 1:1000; Santa Cruz Biotechnology), and anti-

phospho-VEGFR2 and anti-PDGFRb (rabbit 1:1000). To measure

MMP activity after regression in 3D cultures treated with KLK, we

used a 10% gelatin zymogram ready gel (BioRad) and loaded super-

natant of regressed gels for electrophoresis under nonreducing con-

ditions. Protein amounts were normalized by the Bradford method.

After electrophoresis, gels were washed twice in a 2.5% Triton X-100

including 0.02% NaN3 buffer for 20 minutes, switched to an incuba-

tion buffer (50 mM Tris HCl pH 8.0, % mM CaCl2, 0.02% NaN3) for

10minutes and incubated in fresh incubation buffer at 37°C for 24 to 48

hours. After incubation, gels were fixed and stained in fresh Coomassie






blue solution and destained in MeOH:AcOH:H2O (5:1:4) followed by

10% AcOH. Gels were scanned and MMP activity was measured by

densitometry of destained areas.

Total RNA was extracted using RNeasy Mini Kit. Purity was

determined by the A260/A280 ratio. cDNAwas synthesized using Iscript

(BioRad). Conventional and quantitative PCR was performed using

previously described methods.1 The specific primer pairs used for PCR

are listed in Supplemental Table 2.

In Situ HybridizationRiboprobes were designed by linearization pCytoMegaloVirus sport6

vectors containing the specific cDNA for Timp3, Adamts1, or Ngf

alpha (control) (Invitrogen) via restriction enzyme digestion. RNA

probes were then generated using a DIG labeling kit (Roche) con-

taining Sp6 or T7 polymerases to generate probes from the 39 end.

Probe sizes were confirmed by RNA agarose gel electrophoresis and

digoxigenin incorporationwas determined by membrane spotting. Hy-

bridizations were performed following methods adapted from in situ

hybridization in kidney embryos.60 In brief, paraformaldehyde (PFA)

fixed kidney cryosections (8mm) were used within 48 hours, postfixed,

proteinase-K digested, postfixed again, acetylated, and then incubated

with 750 ng/ml riboprobes at 68°C for 16 hours in hybridization buffer

(ENZ-33808; Enzo Life Sciences). Sections were digested by RNase, and

then washed again with solutions of increasing stringency. After incu-

bation with anti-digoxigenin antibody (Roche) for 2 hours at room

temperature, sections were washed and developed with nitro blue tet-

razolium chloride (Roche) for up to 24 hours. Color reactions were

stopped with fixation and sections were mounted with ProLong Gold

(Invitrogen).

Tissue Preparation, Histology, and AnalysesMouse kidneys were flushed with ice-cold PBS to remove blood, and

were harvested and fixed with paraformaldehyde, L-lysine, periodate

solution for 2 hours before cryopreservation for immunofluorescence

studies or with 10% formalin solution for 12 to 16 hours for paraffin

sections.61 Mouse kidneys for in situ hybridization were fixed overnight

with 4% PFA before cryopreservation.60 Human kidney biopsies were

either prepared for electronmicroscopy (EM) by standard methods9 or

fixed with paraformaldehyde, L-lysine, periodate solution for 1 hour,

followed by cryopreservation.61 Human or mouse cryosections (7 mm)

for immune fluorescence studies were blocked in 10% normal serum

and then incubated with primary antibody (2 hours at room tempera-

ture or 16 hours at 4°C) in blocking serum using the following primary

antibodies for mouse kidneys: anti-ADAMTS1 (rabbit 1:100; Santa

Cruz Biotechnology), anti-PDGFRb (rabbit 1:400 collaborator),

anti-CD31 (rat 1:250; R&D Systems), and anti-aSMA-cy3 (mouse

1:250; Sigma Aldrich). Anti-mouse PDGFRb antibodies were gener-

ated and characterized as previously described.54 The following anti-

bodies were used for human kidneys: anti-CD34 (1:250 Ab30375;

Abcam) (1:300; BD Biosciences), anti-PDGFRb (1:250; Cell Signal-

ing), and anti-aSMA-Cy3 (mouse 1:250; Sigma Aldrich). As necessary,

sections were incubated with appropriate affinity purified secondary

Cy3-conjugated antibodies (1:500; Jackson Immunoresearch) for 1 hour

at room temperature, postfixed in 1% PFA for 2 minutes, and mounted

with Vectashield plus DAPI or ProLong Gold plus DAPI. Quantification

of vascular density was performed as previously described,40 and mor-

phometry of Sirius red stain or other fluorescent labels was performed

as previously reported.1,22–24 In some studies, extravasated Evans blue

dye (1%)was detected inwhole kidneymeasuring absorbance.62Digital

images were captured by confocal microscopy. Sirius red staining or

periodic acid–Schiff staining were performed on 3-mm paraffin sec-

tions as previously described.63 Positive cells or area of positive stain/

fluorescence in images were quantified as previously described.1,38

Statistical AnalysesComparisons of two groups were performed using the t test, whereas

several groups were analyzed by one-way ANOVA followed by multiple

comparison testing, using GraphPad Prism software. Groups were con-

sidered significant if P values were,0.05. Error bars indicate SEM.

ACKNOWLEDGMENTS

WethankDr. Li LiHsiao (HarvardMedical School,MA),Dr. Roderick

Jensen (Virginia Tech, VA), and Michael J. Lombardi (Microarray

Biotechnology Center, Harvard Medical School) for assistance with

microarrays; Dr. William Parks and Dr. Yujiro Kida (University

of Washington, WA) and Dr. Anastasia Sacharidou (University of

Missouri, MI) for advice and assistance with the vascular assay, re-

spectively; Dr. Rama Khokha (University of Toronto, Toronto, ON,

Canada) for providing the Timp32/2 mice; and the Lynne and Mike

Garvey Microscopy Suite at South Lake Union for microscopy sup-

port. We also thank Dr. Herman Haller for ongoing support and

advice. The Duffield laboratory is funded by the National Institutes of

Health (Grants DK73299, DK84077, and DK87389), University of

Washington, Genzyme (Research in Progress Grant), and the Nephcure

Foundation, as well as by a research agreement with Regulus Thera-

peutics. C.S. was partly funded by Medizinische Hochschule,

Hannover, Germany.

DISCLOSURESJ.S.D. is on the scientific advisory board for Promedior Inc and Regulus

Therapeutics, has stock options with Promedior Inc, and has consulted for

Amira Pharamceuticals, Gilead, Abbott Pharmaceuticals, Takeda Pharma-

ceuticals, and Boehringer Ingelheim.

REFERENCES

1. Lin SL, Kisseleva T, Brenner DA, Duffield JS: Pericytes and perivascularfibroblasts are the primary source of collagen-producing cells in ob-structive fibrosis of the kidney. Am J Pathol 173: 1617–1627, 2008

2. Humphreys BD, Lin SL, Kobayashi A, Hudson TE, Nowlin BT, BonventreJV, Valerius MT, McMahon AP, Duffield JS: Fate tracing reveals thepericyte and not epithelial origin of myofibroblasts in kidney fibrosis.Am J Pathol 176: 85–97, 2010

3. Benjamin LE, Hemo I, Keshet E: A plasticity window for blood vesselremodelling is defined by pericyte coverage of the preformed endo-thelial network and is regulated by PDGF-B and VEGF. Development125: 1591–1598, 1998

4. Benjamin LE, Golijanin D, Itin A, PodeD, Keshet E: Selective ablation ofimmature blood vessels in established human tumors follows vascularendothelial growth factor withdrawal. J Clin Invest 103: 159–165, 1999




5. Armulik A, Abramsson A, Betsholtz C: Endothelial/pericyte inter-actions. Circ Res 97: 512–523, 2005

6. Armulik A, Genové G, Mäe M, Nisancioglu MH, Wallgard E, Niaudet C,He L, Norlin J, Lindblom P, Strittmatter K, Johansson BR, Betsholtz C:Pericytes regulate the blood-brain barrier. Nature 468: 557–561, 2010

7. Majesky MW, Dong XR, Hoglund V, Mahoney WM Jr, Daum G: Theadventitia: A dynamic interface containing resident progenitor cells.Arterioscler Thromb Vasc Biol 31: 1530–1539, 2011

8. Levinson RS, Batourina E, Choi C, Vorontchikhina M, Kitajewski J,Mendelsohn CL: Foxd1-dependent signals control cellularity in therenal capsule, a structure required for normal renal development. De-velopment 132: 529–539, 2005

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: Targetingendothelium-pericyte cross talk by inhibiting VEGF receptor signalingattenuates kidney microvascular rarefaction and fibrosis. Am J Pathol

178: 911–923, 201110. Stratman AN, Schwindt AE,Malotte KM, Davis GE: Endothelial-derived

PDGF-BB and HB-EGF coordinately regulate pericyte recruitmentduring vasculogenic tube assembly and stabilization.Blood 116: 4720–4730, 2010

11. Chen YT, Chang FC, Wu CF, Chou YH, Hsu HL, Chiang WC, Shen J,Chen YM, Wu KD, Tsai TJ, Duffield JS, Lin SL: Platelet-derived growthfactor receptor signaling activates pericyte-myofibroblast transition inobstructive and post-ischemic kidney fibrosis. Kidney Int 80: 1170–1181, 2011

12. Greenberg JI, ShieldsDJ, Barillas SG, Acevedo LM,Murphy E, Huang J,Scheppke L, Stockmann C, Johnson RS, Angle N, Cheresh DA: A rolefor VEGF as a negative regulator of pericyte function and vessel mat-uration. Nature 456: 809–813, 2008

13. Wühl E, Schaefer F: Managing kidney disease with blood-pressurecontrol. Nat Rev Nephrol 7: 434–444, 2011

14. Jung YJ, Kim DH, Lee AS, Lee S, Kang KP, Lee SY, Jang KY, Sung MJ,Park SK, Kim W: Peritubular capillary preservation with COMP-angiopoietin-1 decreases ischemia-reperfusion-induced acute kid-ney injury. Am J Physiol Renal Physiol 297: F952–F960, 2009

15. Rabelink TJ,WijewickramaDC, de Koning EJ: Peritubular endothelium:The Achilles heel of the kidney? Kidney Int 72: 926–930, 2007

16. Davis GE, Saunders WB: Molecular balance of capillary tube formationversus regression in wound repair: Role of matrix metalloproteinasesand their inhibitors. J Investig Dermatol Symp Proc 11: 44–56, 2006

17. Bausch D, Pausch T, Krauss T, Hopt UT, Fernandez-del-Castillo C,Warshaw AL, Thayer SP, Keck T: Neutrophil granulocyte derivedMMP-9 is a VEGF independent functional component of the angio-genic switch in pancreatic ductal adenocarcinoma. Angiogenesis 14:235–243, 2011

18. Darmanin S, Chen J, Zhao S, Cui H, Shirkoohi R, Kubo N, Kuge Y,Tamaki N, Nakagawa K, Hamada J, Moriuchi T, Kobayashi M: All-transretinoic acid enhances murine dendritic cell migration to draininglymph nodes via the balance of matrix metalloproteinases and theirinhibitors. J Immunol 179: 4616–4625, 2007

19. Ries C, Egea V, KarowM, KolbH, JochumM,Neth P:MMP-2,MT1-MMP,and TIMP-2 are essential for the invasive capacity of humanmesenchymalstem cells: Differential regulation by inflammatory cytokines. Blood 109:4055–4063, 2007

20. Vigetti D, Moretto P, Viola M, Genasetti A, Rizzi M, Karousou E, PallottiF, De Luca G, Passi A: Matrix metalloproteinase 2 and tissue inhibitorsof metalloproteinases regulate human aortic smooth muscle cell mi-gration during in vitro aging. FASEB J 20: 1118–1130, 2006

21. Hartwell LH, Hopfield JJ, Leibler S, Murray AW: From molecular tomodular cell biology. Nature 402[Suppl]: C47–C52, 1999

22. Lin SL, Li B, Rao S, Yeo EJ, Hudson TE, Nowlin BT, Pei H, Chen L, ZhengJJ, Carroll TJ, Pollard JW, McMahon AP, Lang RA, Duffield JS: Mac-rophage Wnt7b is critical for kidney repair and regeneration. Proc Natl

Acad Sci USA 107: 4194–4199, 2010

23. Castaño AP, Lin SL, Surowy T, Nowlin BT, Turlapati SA, Patel T, Singh A,Li S, LupherML Jr,Duffield JS: Serumamyloid P inhibits fibrosis throughFc gamma R-dependent monocyte-macrophage regulation in vivo. SciTransl Med 1: 5ra13, 2009

24. Lin SL, Castaño AP, Nowlin BT, Lupher ML Jr, Duffield JS: Bone marrowLy6Chigh monocytes are selectively recruited to injured kidney and dif-ferentiate into functionally distinct populations. J Immunol 183: 6733–6743, 2009

25. Ernst J, Bar-Joseph Z: STEM: A tool for the analysis of short time seriesgene expression data. BMC Bioinformatics 7: 191, 2006

26. Jönsson-Rylander AC, Nilsson T, Fritsche-Danielson R, Hammarström A,Behrendt M, Andersson JO, Lindgren K, Andersson AK, Wallbrandt P,Rosengren B, Brodin P, Thelin A, Westin A, Hurt-Camejo E, Lee-SøgaardCH: Role of ADAMTS-1 in atherosclerosis: Remodeling of carotid artery,immunohistochemistry, and proteolysis of versican. Arterioscler Thromb

Vasc Biol 25: 180–185, 200527. Vázquez F, Hastings G, Ortega MA, Lane TF, Oikemus S, Lombardo M,

Iruela-Arispe ML: METH-1, a human ortholog of ADAMTS-1, andMETH-2 are members of a new family of proteins with angio-inhibitoryactivity. J Biol Chem 274: 23349–23357, 1999

28. Rodríguez-Manzaneque JC, Westling J, Thai SN, Luque A, Knauper V,Murphy G, Sandy JD, Iruela-Arispe ML: ADAMTS1 cleaves aggrecan atmultiple sites and is differentially inhibited by metalloproteinase in-hibitors. Biochem Biophys Res Commun 293: 501–508, 2002

29. Qi JH, Ebrahem Q, Moore N, Murphy G, Claesson-Welsh L, Bond M,Baker A, Anand-Apte B: A novel function for tissue inhibitor ofmetalloproteinases-3 (TIMP3): Inhibition of angiogenesis by blockageof VEGF binding to VEGF receptor-2. Nat Med 9: 407–415, 2003

30. SaundersWB, BohnsackBL, Faske JB, AnthisNJ, Bayless KJ, Hirschi KK,Davis GE: Coregulation of vascular tube stabilization by endothelial cellTIMP-2 and pericyte TIMP-3. J Cell Biol 175: 179–191, 2006

31. Gill SE, Huizar I, Bench EM, Sussman SW, Wang Y, Khokha R, ParksWC: Tissue inhibitor of metalloproteinases 3 regulates resolution of in-flammation following acute lung injury. Am J Pathol 176: 64–73, 2010

32. Cai X, Lin Y, Friedrich CC, Neville C, Pomerantseva I, Sundback CA,Zhang Z, Vacanti JP, Hauschka PV, Grottkau BE: Bone marrow derivedpluripotent cells are pericytes which contribute to vascularization. StemCell Rev 5: 437–445, 2009

33. Crisan M, Yap S, Casteilla L, Chen CW, Corselli M, Park TS, Andriolo G,Sun B, Zheng B, Zhang L, Norotte C, Teng PN, Traas J, Schugar R,Deasy BM, Badylak S, Buhring HJ, Giacobino JP, Lazzari L, Huard J,Péault B: A perivascular origin for mesenchymal stem cells in multiplehuman organs. Cell Stem Cell 3: 301–313, 2008

34. Hayden MR, Karuparthi PR, Habibi J, Lastra G, Patel K, Wasekar C,Manrique CM, Ozerdem U, Stas S, Sowers JR: Ultrastructure of isletmicrocirculation, pericytes and the islet exocrine interface in the HIP ratmodel of diabetes. Exp Biol Med (Maywood) 233: 1109–1123, 2008

35. Caruso RA, Fedele F, Finocchiaro G, Pizzi G, Nunnari M, Gitto G,Fabiano V, Parisi A, Venuti A: Ultrastructural descriptions of pericyte/endothelium peg-socket interdigitations in the microvasculature ofhuman gastric carcinomas. Anticancer Res 29: 449–453, 2009

36. Suzuki K, Masawa N, Sakata N, Takatama M: Pathologic evidence ofmicrovascular rarefaction in the brain of renal hypertensive rats. J

Stroke Cerebrovasc Dis 12: 8–16, 200337. Saunders WB, Bayless KJ, Davis GE: MMP-1 activation by serine pro-

teases and MMP-10 induces human capillary tubular network collapseand regression in 3D collagen matrices. J Cell Sci 118: 2325–2340,2005

38. DavisGE, PintarAllenKA,Salazar R,Maxwell SA:Matrixmetalloproteinase-1 and -9 activation by plasmin regulates a novel endothelial cell-mediated mechanism of collagen gel contraction and capillary tuberegression in three-dimensional collagenmatrices. J Cell Sci 114: 917–930, 2001

39. Schrimpf C, Duffield JS: Mechanisms of fibrosis: The role of the peri-cyte. Curr Opin Nephrol Hypertens 20: 297–305, 2011



40. Li B, Cohen A, Hudson TE, MotlaghD, Amrani DL, Duffield JS: Mobilizedhuman hematopoietic stem/progenitor cells promote kidney repair afterischemia/reperfusion injury. Circulation 121: 2211–2220, 2010

41. Lim NH, Kashiwagi M, Visse R, Jones J, Enghild JJ, Brew K, Nagase H:Reactive-site mutants of N-TIMP-3 that selectively inhibit ADAMTS-4and ADAMTS-5: Biological and structural implications. Biochem J 431:113–122, 2010

42. Kassiri Z, Oudit GY, Kandalam V, Awad A, Wang X, Ziou X, Maeda N,Herzenberg AM, Scholey JW: Loss of TIMP3 enhances interstitialnephritis and fibrosis. J Am Soc Nephrol 20: 1223–1235, 2009

43. Janssen A, Hoellenriegel J, Fogarasi M, Schrewe H, Seeliger M, TammE,Ohlmann A,May CA,Weber BH, Stöhr H: Abnormal vessel formationin the choroid of mice lacking tissue inhibitor of metalloprotease-3.Invest Ophthalmol Vis Sci 49: 2812–2822, 2008

44. Basile DP, Fredrich K, Chelladurai B, Leonard EC, Parrish AR: Renalischemia reperfusion inhibits VEGF expression and induces ADAMTS-1,a novel VEGF inhibitor.AmJ Physiol Renal Physiol 294: F928–F936, 2008

45. Ronco C, House AA, Haapio M: Cardiorenal syndrome: Refining thedefinition of a complex symbiosis gone wrong. Intensive Care Med 34:957–962, 2008

46. Anania MC, Sensi M, Radaelli E, Miranda C, Vizioli MG, Pagliardini S,Favini E, Cleris L, Supino R, Formelli F, Borrello MG, Pierotti MA, GrecoA: TIMP3 regulates migration, invasion and in vivo tumorigenicity ofthyroid tumor cells. Oncogene 30: 3011–3023, 2011

47. Kornfeld JW, Meder S, Wohlberg M, Friedrich RE, Rau T, Riethdorf L,Löning T, Pantel K, Riethdorf S: Overexpression of TACE and TIMP3mRNA in head and neck cancer: Association with tumour developmentand progression. Br J Cancer 104: 138–145, 2011

48. Song B, Wang C, Liu J, Wang X, Lv L, Wei L, Xie L, Zheng Y, Song X:MicroRNA-21 regulates breast cancer invasion partly by targeting tis-sue inhibitor of metalloproteinase 3 expression. J Exp Clin Cancer Res29: 29, 2010

49. Qi JH, Dai G, Luthert P, Chaurasia S, Hollyfield J, Weber BH, Stöhr H,Anand-ApteB: S156Cmutation in tissue inhibitor ofmetalloproteinases-3induces increased angiogenesis. J Biol Chem 284: 19927–19936, 2009

50. Nakamura A, Sakai Y,OhataC, Komurasaki T: Expression and significanceof a disintegrin and metalloproteinase with thrombospondin motifs(ADAMTS)-1 in an animal model of renal interstitial fibrosis induced byunilateral ureteral obstruction. Exp Toxicol Pathol 59: 1–7, 2007

51. Luque A, Carpizo DR, Iruela-Arispe ML: ADAMTS1/METH1 inhibitsendothelial cell proliferation by direct binding and sequestration ofVEGF165. J Biol Chem 278: 23656–23665, 2003

52. Mittaz L, Ricardo S, Martinez G, Kola I, Kelly DJ, Little MH, Hertzog PJ,Pritchard MA: Neonatal calyceal dilation and renal fibrosis resultingfrom loss of Adamts-1 in mouse kidney is due to a developmentaldysgenesis. Nephrol Dial Transplant 20: 419–423, 2005

53. Mittaz L, Russell DL,Wilson T, BrastedM, Tkalcevic J, Salamonsen LA,Hertzog PJ, Pritchard MA: Adamts-1 is essential for the developmentand function of the urogenital system. Biol Reprod 70: 1096–1105,2004

54. OzerdemU, Grako KA, Dahlin-Huppe K,Monosov E, StallcupWB: NG2proteoglycan is expressed exclusively by mural cells during vascularmorphogenesis. Dev Dyn 222: 218–227, 2001

55. SaeedAI, SharovV,WhiteJ, Li J, LiangW,BhagabatiN,BraistedJ,KlapaM,Currier T, ThiagarajanM, Sturn A, SnuffinM, Rezantsev A, PopovD, RyltsovA, Kostukovich E, Borisovsky I, Liu Z, Vinsavich A, Trush V, Quackenbush J:TM4: A free, open-source system for microarray data management andanalysis. Biotechniques 34: 374–378, 2003

56. Huang W, Sherman BT, Lempicki RA: Systematic and integrative anal-ysis of large gene lists using DAVID bioinformatics resources. NatProtoc 4: 44–57, 2009

57. Baldi P, Long AD: A Bayesian framework for the analysis of microarrayexpression data: Regularized t -test and statistical inferences of genechanges. Bioinformatics 17: 509–519, 2001

58. Storey JD, Tibshirani R: Statistical significance for genomewide studies.Proc Natl Acad Sci USA 100: 9440–9445, 2003

59. Lin SL, Chen RH, Chen YM, Chiang WC, Lai CF, Wu KD, Tsai TJ: Pen-toxifylline attenuates tubulointerstitial fibrosis by blocking Smad3/4-activated transcription and profibrogenic effects of connective tissuegrowth factor. J Am Soc Nephrol 16: 2702–2713, 2005

60. McMahonAP,AronowBJ,DavidsonDR,Davies JA,GaidoKW,GrimmondS, Lessard JL, Little MH, Potter SS, Wilder EL, Zhang P GUDMAP project:GUDMAP: The genitourinary developmental molecular anatomy project.J Am Soc Nephrol 19: 667–671, 2008

61. Brenes F, Harris S, Paz MO, Petrovic LM, Scheuer PJ: PLP fixation forcombined routine histology and immunocytochemistry of liver bi-opsies. J Clin Pathol 39: 459–463, 1986

62. David S, Ghosh CC, Kümpers P, Shushakova N, Van Slyke P, KhankinEV, Karumanchi SA, Dumont D, Parikh SM: Effects of a synthetic PEG-ylated Tie-2 agonist peptide on endotoxemic lung injury andmortality.Am J Physiol Lung Cell Mol Physiol 300: L851–L862, 2011

63. Duffield JS, Forbes SJ, Constandinou CM, Clay S, PartolinaM, VuthooriS,Wu S, Lang R, Iredale JP: Selective depletion ofmacrophages revealsdistinct, opposing roles during liver injury and repair. J Clin Invest 115:56–65, 2005

See related editorial, “Activated Pericytes and the Inhibition of Renal VascularStability: Obstacles for Kidney Repair,” on pages 767–769.

This article contains supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2011080851/-/DCSupplemental.





pericyte timp3 and adamts1 modulate vascular stability after

Documents