Cardiovascular Research (2017) 113, 1551–1559 doi:10.1093/cvr/cvx177

Interleukin-9 mediates chronic kidney disease- dependent vein graft disease: a role for mast cells

Lisheng Zhang1*, Jiao-Hui Wu1, James C. Otto2, Susan B. Gurley3, Elizabeth R. Hauser4,5,6, Sudha K. Shenoy1,7, Karim Nagi1,7, Leigh Brian1, Virginia Wertman1, Natalie Mattocks3, Jeffrey H. Lawson2, and Neil J. Freedman1,7*

1Cardiology, Department of Medicine, 2Surgery, 3Nephrology, Department of Medicine, 4Biostatistics and Bioinformatics and 5Molecular Physiology Institute, Duke University Medical Center, Durham, NC, USA; 6Cooperative Studies Program Epidemiology Center Durham Veterans Affairs Medical Center, Durham, NC, USA; and 7Cell Biology, Duke University Medical Center, Durham, NC, USA

Received 14 March 2017; revised 14 July 2017; editorial decision 19 August 2017; accepted 29 August 2017; online publish-ahead-of-print 31 August 2017 Time for primary review: 46 days

Aims Chronic kidney disease (CKD) is a powerful independent risk factor for cardiovascular events, including vein graft failure. Because CKD impairs the clearance of small , we tested the hypothesis that CKD exacerbates vein graft disease by elevating serum levels of critical that promote vein graft neointimal hyperplasia...... Methods We modelled CKD in C57BL/6 mice with 5/6ths nephrectomy, which reduced glomerular filtration rate by 60%, and results and we modelled vein grafting with inferior-vena-cava-to-carotid interposition grafting. CKD increased vein graft neointimal hyperplasia four-fold, decreased vein graft re-endothelialization two-fold, and increased serum levels of -9 (IL-9) five-fold. By quantitative immunofluorescence and histochemical staining, vein grafts from CKD mice demonstrated a two-fold higher prevalence of mast cells, and a six-fold higher prevalence of activated mast cells. Concordantly, vein grafts from CKD mice showed higher levels of TNF and NFjB activation, as judged by phosphorylation of NFjB p65 on Ser536 and by expression of VCAM-1. Arteriovenous fistula veins from humans with CKD also showed up-regulation of mast cells and IL-9. Treating CKD mice with IL-9-neutralizing IgG reduced vein graft neointimal area four-fold, increased vein graft re-endothelialization two-fold, and reduced vein graft total and activated levels two- and four-fold, respectively. Treating CKD mice with the mast cell stabilizer cromolyn reduced neointimal hyperplasia and increased re-endothelialization in vein grafts. In vitro, IL-9 promoted endothelial cell but had no effect on smooth muscle cell proliferation...... Conclusion CKD aggravates vein graft disease through mechanisms involving IL-9 and mast cell activation...... Keywords Neointimal hyperplasia • Cytokines • Mast cells • Endothelial cells • Smooth muscle cells

᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿᭿

. 4 . die than they are to suffer progression to renal replacement therapy. 1. Introduction . . Furthermore, CKD substantially increases failure rates of coronary stent- 2 . 5 Defined as a glomerular filtration rate (GFR) of <60 mL/min per 1.73 m . ing and aortocoronary vein grafts: among patients with vein grafts that of body surface area for 3 months or more, chronic kidney disease . are patent 1–11 years after implantation, CKD increases the subsequent . 6 (CKD) is an important independent predictor of cardiovascular morbid- . vein graft failure rate by 62%. This problem of vein graft failure in CKD 1 . ity and mortality. CKD affects 20 million adults in the United States, . patients is magnified by the observation that, in CKD patients, coronary and the prevalence of CKD is estimated to be 8 to 16% worldwide.2 . artery bypass grafting appears to confer a lower long-term risk of myo- . 7 Cardiovascular event rates increase 10% for every 10-unit fall in GFR . cardial infarction and revascularization than coronary stenting. 3 . below even 81 mL/min/1.73 sq m, and CKD patients are 40 times . Vein graft failure initiates with neointimal hyperplasia, a process involv- more likely to suffer a cardiovascular event and 15 times more likely to . ing proliferation and migration of medial smooth muscle cells (SMCs) .

* Corresponding authors. Tel: 919 684 6826; fax: 919 681 0718, E-mail: [email protected] (L.Z.); Tel: 919 684 6873; fax: 919 684 8875, E-mail: [email protected] (N.J.F.) Published on behalf of the European Society of Cardiology. All rights reserved. VC The Author 2017. For permissions, please email: [email protected]. 1552 L. Zhang et al.

. into the sub-endothelial intima as well as an accumulation of extracellular . 2.6 Systemic neutralization of IL-9 8 . matrix in the neointima. Vein graft neointimal hyperplasia has been . At the time of vein graft surgery, CKD mice were injected with one of shown to be exacerbated by inflammatory cytokines,9,10 and to predis- . . two Armenian Hamster IgG2/j molecules: (a) D9302C12, which neu- 11 . . 19 pose vein grafts to accelerated atherosclerosis. Because elevated . tralizes IL-9, or (b) control IgG, which binds no known mouse . serum levels of pro-inflammatory cytokines are found in humans with . (BioLegend). Subsequently, mice were injected with the same IgG 3 12 . CKD, we hypothesized that inflammatory cytokines could mediate . times per week (300 lgi.p.),19 until 48–72 h before vein graft harvest at CKD-dependent acceleration of vein graft disease. To test whether iden- . . 4 week after implantation. tifying CKD-associated cytokines could facilitate therapies for CKD- . associated vein graft disease, we used the 5/6ths nephrectomy model of . . CKD in mice and carotid interposition vein grafting. . . 2.7 Systemic treatment with cromolyn . . To prevent mast cell degranulation in CKD mice, we treated mice with 2. Methods . the mast cell stabilizer cromolyn.17 One day prior to vein grafting, CKD . . mice were injected i.p. with either PBS or cromolyn (50 mg/kg). After 2.1 Mice . vein grafting, mice were injected with cromolyn or PBS twice per week . C57BL/6 J mice were used in all experiments. All animal experiments . for 4 weeks, until vein graft harvest. were performed according to protocols approved by the Duke . . Institutional Animal Care and Use Committee; these protocols complied . . with the Guide for the Care and Use of Laboratory Animals (National . 2.8 Endothelial and smooth muscle cell . Academies Press, 2011). . experiments . . Mouse aortic endothelial cells (ECs) and smooth muscle cells (SMCs) 2.2 Nephrectomy model of CKD . were isolated by enzymatic digestion, then cultivated as we described.20 . 5/6ths nephrectomy was performed in two sequential operations, modi- . ECs for proliferation experiments were grown in medium containing 13 . fied from published protocols. Anesthesia with pentobarbital (50 mg/ . . 20% (vol/vol) serum from 5/6ths-nephrectomized or 1/3rd-nephrec- kg i.p.) was used for all surgeries; see Supplementary material online. . tomized mice; SMCs used 10% mouse serum. At each proliferation time After the second stage of the nephrectomy surgeries, mice recovered . 20 . point, cells were quantitated with a dye-binding assay, as we described. for 2 week before undergoing either GFR measurement or carotid inter- . SMC proliferation and EC apoptosis assays were conducted ±IL-9 as position vein grafting. . . described in Supplementary material online. GFR Measurement was performed under 2% isoflurane anesthesia, as . we described.14 . . . 2.3 Carotid interposition vein grafting . 2.9 Human vein samples from . Interposition vein graft surgery was performed as we described.15 . arteriovenous fistulae (AVFs) . All procedures with human volunteers were approved by the Duke Inferior venae cavae from WT donor mice were anastomosed end-to- . side to the right common carotid artery of 5/6ths or 1/3rd-nephrectom- . Institutional Review Board. Vein specimens were obtained by a single . surgeon (J. H. L.) from five humans with end-stage renal disease when ized recipient mice. . . they underwent two-stage basilic vein transposition AVF surgery or . 21 2.4 Histology . radiocephalic AVF surgery. Samples of ‘native vein’ were obtained dur- . All measurements and calculations on histologic specimens were made by . ing the initial construction of the AVF, and ‘AVF vein’ samples were . obtained during the surgery to transpose the arterialized basilic vein. observers blinded to specimen identity. Vein graft morphometry was per- . formed with NIH ImageJ as we have described.9,15 Immunofluorescence . . microscopy and data analysis with ImageJ were performed as we . described.9 The extent of vein graft re-endothelialization was determined . 2.10 Statistical analyses . as the percentage of luminal perimeter surfaced by von Willebrand factor- . All values were plotted in figures as means ± SE and cited in the text as 16 . positive cells. . means ± SD. GraphPad PrismVR software was used to perform t tests 17 . Toluidine blue staining for mast cells was performed as described. . (for comparing two groups), or 2-way ANOVA with Sidak’s post-hoc Mast cells were identified by their metachromatic (purple) granules, . . test for multiple comparisons. A P value of less than 0.05 was considered which are distinct from the orthochromatic (blue) staining of all other . significant. cells. Mast cells were dichotomized as (a) quiescent, with densely packed . . To analyse data from serum assays, the distributions of each granules in the cytoplasm, or (b) activated, or degranulating, with dis- . cytokine were examined. We made two comparisons of ‘treatment’ vs. 17 . gorged and loosely packed granules. A minimum of 100 mast cells per . ‘control’: (1) CKD mice vs. control-GFR mice, and (2) anti-IL-9-IgG- specimen were counted. . . injected CKD mice vs. control IgG-injected CKD mice. We compared . the medians between the treatment and control groups using the non- . 2.5 Serum cytokine assays . parametric Wilcoxon Test with exact P value calculation, under the null Serum was harvested by ventricular puncture from pentobarbital- . hypothesis that there was no difference between the treatment and con- . anesthetized mice at the indicated time points. Mouse sera were . trol groups. A conservative multiple comparisons correction for 23 inde- TM . analysed with the Bio-Plex Pro Mouse Cytokine Assay and with an . pendent cytokines was used to indicate a significant test: P < 0.05/23, or ELISA for serum amyloid A18 (see Supplementary material online). . P < 0.00217. Chronic kidney disease-dependent vein graft disease 1553

3. Results 3.1 CKD increases vein graft neointimal hyperplasia and inflammation To model CKD in mice, we performed 5/6ths nephrectomy as described,13 and assessed GFR 2 week later by FITC-inulin clearance. Compared with sham-operated control mouse GFR, CKD mouse GFR was 60 ± 10% less (Figure 1A)—a GFR reduction that models human chronic kidney disease stage 3.1 To determine the effect of CKD on vein graft disease, we performed carotid interposition vein grafting15 on CKD and control-GFR mice, 2 week after 5/6ths or 1/3rd nephrectomy, respectively. Vein grafts were harvested 4 week post-operatively, and cross sections were evaluated by planimetry.15 Although vein graft tunica media areas were equivalent in control-GFR and CKD mice, the neointimal area was 4 ± 0.5-fold larger in CKD mice (Figure 1B and C). Despite their larger neointimal areas, vein grafts from CKD mice demonstrated a structural composition comparable to that found in vein grafts from control-GFR mice, as judged by the prevalence of SMCs and the density of collagen I20 (see Supplementary material online, Figure S1). However, the prevalence of proliferating SMCs was two-fold greater in vein grafts from CKD mice than in vein grafts from control-GFR mice (see Supplementary material online, Figure S1). Vein grafts from CKD mice also demonstrated (Figure 1B and C)a50% reduction in the degree of vein graft re- endothelialization, which inversely correlates with vein graft neointimal hyperplasia.16,22 Because CKD is associated with increased plasma levels of inflam- matory cytokines in humans,12,23 we tested whether vein grafts from CKD mice evinced greater levels of inflammatory signalling in the neointimal and medial layers. We adduced as evidence of inflamma- tory signalling the phosphorylation on Ser536 of the NFjBsubunit p65; this phosphorylation is effected by IjBkinase-b, and augments NFjB transcriptional activity.24 Whereas vein grafts from control- GFR and CKD mice showed equivalent expression of total p65 NFjB, vein grafts from CKD mice showed 75 ± 9% higher levels of phospho-p65(Ser536) (Figure 2), and 60 ± 10% higher levels of p65 nuclear translocation (see Supplementary material online, Figure S2). Thus, vein grafts in CKD mice appeared to have more extensive acti- vation of NFjB. Concordantly, vein grafts from CKD mice expressed two-fold higher levels of the NFjB-dependent,24 pro-inflammatory product VCAM-1 (Figure 2), and showed 30% greater preva- lence of adventitial macrophages—even though CKD and control vein grafts had equivalent levels of adventitial T (see Supplementary material online, Figure S1). Furthermore, vein grafts from CKD mice also had 60 ± 10% higher levels of TNF and 60 ± 10% Figure 1 CKD exacerbates vein graft neointimal hyperplasia. (A), greater expression of TNF receptor-1 (see Supplementary material Mice were subjected to 5/6ths nephrectomy (CKD) or sham surgery online, Figure S1), which we previously found mediates most of the (control); 2 week later, they were injected with FITC-inulin and sub- NFjB activation in vein graft SMCs.9 Taken together, these data sug- jected to sequential blood collections at the indicated times after injec- gest that CKD engenders greater vein graft inflammation. tion; plasma fluorescence at each time point is plotted for individual CKD and control mice, at left. GFR was calculated from these decay curves and plotted (right panel) as means ±SE from 4 mice in each group. 3.2 CKD augments serum IL-9 levels Compared with control: *P <0.05(t test). (B), CKD and 1/3rd-nephrec- tomized (control) mice were subjected to carotid interposition vein Does increased vein graft inflammation in CKD mice result from higher grafting, and vein grafts were harvested 4 week later after perfusion fixa- levels of plasma pro-inflammatory cytokines? To address this question, tion. Serial cross sections were stained with a modified Masson we analysed individual sera from 9 control-GFR and 12 CKD mice, trichrome (top) or with IgG targeting von Willebrand factor (vWF, matched for age and sex. Sera were harvested 2 week after 5/6ths neph- green) and Hoechst 33 342 (DNA, blue). The area delineated by the rectomy, at a time when serum amyloid A levels were equivalent in both dashed boxes (top) are enlarged and presented in the middle panels, in CKD and control mice (1.13 ± 0.02 vs. 1.19 ± 0.03 mg/mL), and thus which dotted white and blue lines delineate the boundaries between when CKD mice had recovered from post-op inflammation.25 Of 23 1554 L. Zhang et al.

. cytokines tested, only five had serum levels that were significantly higher neointima (‘N’) and media (‘M’), and between media and adventitia (‘A’), . respectively. Scale bars = 50 lm. (C), Cross sectional areas for neointima, . in CKD than in control mice (Figure 3). Of these, the cytokine with by media, lumen and ‘total vein’ (excluding adventitia) were calculated by . far the greatest CKD-associated increment was interleukin-9 (IL-9) . planimetry, and plotted as means ± SE from six mice in each group. ‘Vein . (Figure 3); consequently, we focused on IL-9 in subsequent studies. graft endothelialization’ was calculated as the fraction of vein graft total . Increased serum levels of IL-9 in CKD mice correlated with 3.0 ± 0.6- . luminal perimeter covered by vWF-positive (endothelial) cells, and plot- . fold higher vein graft levels of IL-9 in CKD mice (Figure 4). In addition, tedasthemeans±SEfrom6miceineachgroup.Comparedwithcon- . vein grafts from CKD mice demonstrated 1.7 ± 0.3-fold higher levels of trol mice: *P < 0.05 (2-way ANOVA with post-hoc Sidak test). . . the IL-9 receptor-a (IL-9Ra), which binds to and mediates signalling . evoked by IL-9.26 Because IL-9 is a critical for mast cells,26 . . and because mast cells contribute to vascular inflammation in a variety of . 17 . ways, we asked whether at least some of the excess IL-9Ra-expressing . . vein graft cells in CKD mice were mast cells. . To address this question, we used three complementary . . approaches to quantitate the prevalence of mast cells in the adventi- . tia, where, like others,17 we found virtually all of the vein grafts’ mast . . cells: (a, b) immunostaining for the mast cell-specific proteases tryp- . tase (mouse mast cell proteases 6 and 7)27 and mouse mast cell . 27 . protease-5 (chymase), as well as (c) histochemically staining for . mast cells with toluidine blue17 (Figure 4). By all three approaches, . . mast cells were 2–3 times more prevalent in vein grafts from CKD . than from control-GFR mice (Figure 4). Furthermore, as assessed by . . toluidine blue staining, the prevalence of activated (degranulating) . mast cells17 was six-fold higher in vein grafts from CKD mice than . . from control-GFR mice (Figure 4). . . . . 3.3 IL-9 promotes CKD-dependent vein . graft disease through mast cell-dependent . . mechanisms . . To discern whether IL-9 plays a causative role in CKD-induced vein graft . disease, we injected CKD mice with IL-9-neutralizing or isotype control . . IgG for 4 week after vein grafting. Control IgG-injected CKD mice devel- . oped vein graft neointimal hyperplasia equivalent to that of untreated . . CKD mice (Figures 1 and 5). Remarkably, anti-IL-9 IgG treatment ...... Figure 2 CKD augments NFjB activation in vein grafts. (A), Serial . sections of vein grafts from Figure 1 were immunostained with rabbit . . IgG targeting the following proteins: NFjB p65 subunit phosphorylated . on Ser536 (phospho-p65); NFjB p65 subunit, phosphorylated or not . (p65); VCAM-1; or no specific protein (Neg Control). Alexa 546-conju- . Figure 3 CKD increases serum cytokine levels. Mice subjected to . gated anti-rabbit IgG and Hoechst 33 342 were used on all specimens. . 5/6ths nephrectomy (CKD) or sham surgery (control-GFR) were Specimens from control and CKD mice were stained concurrently. . exsanguinated 2 weeks after the final surgery. Individual serum samples . (B), The ratios of red (protein) to blue (DNA) pixels in the neointima . were analysed in triplicate for 23 cytokines. Serum cytokine levels were . plotted as means ± SE from 9 (control-GFR) or 13 (CKD) individual plus media were quantitated by Image J, and normalized to the cognate . ratios obtained for control specimens in each staining cohort, to obtain . mice in each group. Compared with control-GFR mouse sera: ‘CKD/control’, plotted individually and as the means ± SE for each . *P < 0.002 (Wilcoxon test). IL-12, interleukin-12 (p40 and p70 subu- . group. Compared with control: *P < 0.01 (2-way ANOVA with post- . nits). Cytokines not plotted (see Supplementary material online)were hoc Sidak test). . measured at equivalent levels in CKD and control mice. . Chronic kidney disease-dependent vein graft disease 1555

Figure 4 CKD augments vein graft mast cell prevalence and activation. (A), Serial sections of vein grafts from Figure 2 were stained with Hoechst 33 342 (DNA) and immunostained with rabbit IgG specific for IL-9, IL-9Ra, mouse mast cell proteases -6 and -7 (Tryptase), mouse mast cell protease-5 (mMCP5), or no particular protein (Neg Control). Scale bars = 50 lm. (B), Protein immunofluorescence throughout the vein graft wall was quantitated as in Figure 2,to yield ‘% of control,’ plotted individually and as means ± S.E. Compared with vein grafts from control-GFR mice: *P < 0.05 (2-way ANOVA with Sidak post- hoc test). (C), Serial sections were stained with toluidine blue, to identify metachromatic (purple) mast cell granules. Arrows indicate quiescent (upper panel) and activated (degranulating, lower panel) mast cells. Scale bars = 10 lm. (D), The prevalence of quiescent and activated mast cell in vein grafts were assessed by toluidine blue staining, and plotted as means ± S.E. of five vein grafts per group (>_100 adventitial mast cells were counted per specimen). Compared with vein grafts from control-GFR mice: *P < 0.05 (2-way ANOVA with Sidak post-hoc test).

. reduced this vein graft neointimal hyperplasia in CKD mice by four-fold, . isotype control IgG-injected CKD mice. Furthermore, in vein grafts from and reduced medial thickness by 1.6-fold (Figure 5). Consonant with its . anti-IL-9 IgG-injected CKD mice the prevalence of activated mast cells . reduction in vein graft neointimal hyperplasia, anti-IL-9 IgG treatment . was 75 ± 6% lower than it was in vein grafts from control IgG-injected . also improved vein graft re-endothelialization, by 46 ± 5% (Figure 5), and . CKD mice (Figure 5C). reduced vein graft levels of NFjB phospho-p65(Ser536), IL-9Ra, and IL- . Anti-IL-9 treatment of our CKD mice reduced the serum levels of five . 9by40% (see Supplementary material online, Figure S3). Renal function . cytokines, among the 23 cytokines assayed: levels of IL-9, IL-1a, IL-6, IL- was equivalent, however, in anti-IL-9-treated and control IgG-treated . 10, and IL-12 were 30-60% lower in anti-IL-9-treated than in control 13 . mice, as judged by creatinine values determined on post-mortem . IgG-treated mice (see Supplementary material online, Figure S4). Each serum samples (25 lmol/L). Thus, IL-9 appears to contribute substan- . of these cytokines is secreted by mast cells.28 Thus, anti-IL-9 IgG treat- . tially to CKD-dependent vein graft inflammation and neointimal . ment evoked changes in serum cytokine levels that are consistent hyperplasia. . with the changes it evoked in vein graft mast cell levels and mast cell . Because anti-IL-9 IgG treatment so dramatically reduced CKD- . activation. dependent vein graft inflammation and neointimal hyperplasia, we tested . To test more directly whether mast cell degranulation contributes to . whether anti-IL-9 IgG treatment also reduced the CKD-dependent . CKD-dependent vein graft disease, we treated vein-grafted CKD mice increase in the prevalence and activation of vein graft mast cells. In vein . with cromolyn, an inhibitor of mast cell degranulation.17 As compared . grafts from control IgG-injected CKD mice, the prevalence of mast cells . with vehicle-treated CKD mice, cromolyn-treated CKD mice developed was comparable to that found in untreated CKD mice (Figures 5C and 4). . 40% less vein graft neointimal hyperplasia and demonstrated 50% . However, in vein grafts from anti-IL-9 IgG-injected CKD mice, the preva- . more vein graft re-endothelialization (see Supplementary material lence of mast cells was 45 ± 8% lower than it was in vein grafts from . online, Figure S5). However, unlike anti-IL-9 IgG therapy, cromolyn 1556 L. Zhang et al.

Figure 5 IL-9 promotes CKD-induced vein graft disease and augments mast cell prevalence in CKD mouse vein grafts. Mice were subjected to 5/6ths

nephrectomy and vein graft surgery as in Figure 1. At the time of vein graft surgery, mice were injected with hamster IgG2/j—either anti-IL-9, or isotype con- trol: 50 lgi.v.plus250lg i.p., followed by 300 lgi.p.3/week until vein graft harvest 4 week post-op. (A), Serial sections of perfusion-fixed vein grafts were stained with a modified trichrome or, immunofluorescently, for von Willebrand factor (vWF, green) and Hoechst 33 342 (blue, DNA). Scale bars = 200 lm. (B), The indicated cross sectional areas were calculated as in Figure 1 and plotted as means ± SE from seven mice in each group. ‘Vein graft endothelialization’ was calculated as in Figure 1, and plotted as means ± SE from five mice in each group. Compared with control mice: *P <0.01.(C), The prevalence of quies- cent and activated mast cells in vein grafts were assessed by toluidine blue staining, as in Figure 4. Shown are means ± S.E. from five vein grafts from each IgG- injected mouse group. Compared with control: *P < 0.05 (2-way ANOVA with Sidak post-hoc test).

. therapy did not reduce the prevalence of total or quiescent mast cells, . able to collect native vein during the first stage of the surgery and arteri- . 21 assessed by anti-tryptase immunostain or toluidine blue stain, respec- . alized vein during the second stage of the surgery. By using paired sam- tively; rather, cromolyn therapy reduced the prevalence of activated . ples from each end-stage-CKD human subject, we were able to . (degranulating) mast cells, by 70 ± 5% (see Supplementary material . compare vein samples that were identical genetically but distinct by vir- online, Figure S5C). In contrast to anti-IL-9 IgG therapy, cromolyn therapy . tue of adaptation to haemodynamic conditions. With arterialization of . in CKD mice did not affect levels of 23 serum cytokines (see . the vein in these end-stage CKD subjects, there was a 45–65% increase Supplementary material online, Figure S5D and data not shown). Thus, . in the expression of IL-9, IL-9Ra, and mast cell tryptase (Figure 7 and see . mast cell degranulation in CKD mice appears to augment vein graft neo- . Supplementary material online, Table S1). These data are consistent intimal hyperplasia and to inhibit vein graft re-endothelialization. . with the possibility that IL-9 and mast cells may play a role in human . CKD mice demonstrated impaired vein graft re-endothelialization, . CKD-dependent vein pathology in AVFs, in which vein neointimal hyper- and this defect was substantially mitigated by treatment with anti-IL-9 . plasia is a principal cause of AVF failure.21 . IgG or the mast cell inhibitor cromolyn (Figures 1 and 5;see . Supplementary material online, Figure S5). To address whether the . . serum of CKD mice could be directly toxic to ECs, we cultured mouse . 4. Discussion ECs in the presence or absence of serum derived from mice with either . . normal GFR or CKD, and quantitated EC number after 4 days. Serum . This study provides the first direct evidence that CKD promotes vein from control-GFR mice promoted EC growth, but serum from CKD . graft disease, through mechanisms requiring IL-9 and mast cell activation. . mice did not (Figure 6). The addition of IL-9-neutralizing IgG to control- . We found that either anti-IL-9 or mast cell-inhibiting interventions miti- GFR mouse serum had no effect on EC proliferation. However, the addi- . . gated CKD-dependent vein graft disease and promoted vein graft re- tion of IL-9-neutralizing IgG to CKD mouse serum augmented EC prolif- . endothelialization in mice. Furthermore, the possibility that these findings eration (Figure 6). Thus, it appears that CKD mouse serum may be . . pertain to human disease is suggested by our observation that IL-9 and directly toxic to ECs, and that part of this toxicity may be mediated by . mast cell prevalence both up-regulate in the arterializing vein of AVFs in . IL-9 itself. To test this possibility more directly, we cultured ECs in the . humans with end-stage renal disease. absence or presence of the [IL-9] observed in CKD mouse serum . That CKD promotes vascular disease generally is not only sug- . 3–6 (600 pg/mL). Under these conditions, IL-9 engendered a three-fold . gested by abundant observational data in humans but also directly increase in EC apoptosis, judged by the prevalence of ECs positive for . demonstrated in animal models. With the 5/6ths nephrectomy model 16 . -/- cleaved caspase-3 (Figure 6B and C). Despite the apparent toxicity to . in Apoe mice, CKD exacerbates atherosclerosis through mecha- ECs of CKD serum and IL-9, neither one adversely affected SMC prolif- . nisms involving augmented hypercholesterolemia29 and angiotensin . 30 eration (Figure 6D and E). . II-promoted LDL oxidation. However, the pro-atherogenic role of To determine whether IL-9-related mechanisms could be involved in . CKD-augmented cytokine levels has not been dissected from the . CKD-related vein neointimal hyperplasia in humans, we enrolled human . role of CKD-augmented hypercholesterolemia in the pathogenesis of subjects with chronic kidney disease stage V who were undergoing . accelerated atherosclerosis.13 . basilic vein transposition surgery to create arteriovenous fistulae (AVFs) . Although they differ substantially from vein grafts haemodynamically . 31 for haemodialysis access (see Supplementary material online, Table . because they carry >_3 times more blood flow, AVFs have been used in S1).21 This surgery is performed in two stages; consequently, we were . mice to demonstrate the effects of CKD on vein arterialization. CKD Chronic kidney disease-dependent vein graft disease 1557

Figure 6 CKD mouse serum inhibits endothelial cell proliferation. Mice were subjected to 5/6ths (CKD) or 1/3rd nephrectomy (Control); 2 week after the final surgery, serum was harvested and pooled from 12 mice in each GFR group. (A), Mouse ECs were cultured in medium containing 20% (v/v) serum from either CKD or control mice, along with isotype control (Non-immune) or anti-IL-9 IgG, each at 2 lg/mL. After 4 days, EC number was quantitated (see Methods). Within each experiment, the number of ECs was divided by that obtained with control serum and control IgG, and multiplied by 100 to obtain ‘% of control’. Plotted are means ±SE from three independent experiments performed with distinct mouse EC lines. Compared with control serum: *P < 0.05; compared with isotype control IgG: #, P < 0.05 (2-way ANOVA with post-hoc Sidak test). (B and C), Mouse ECs were grown on culture slides in the absence (None) or presence of IL-9 (600 pg/mL) for 48 h, and then stained for von Willebrand factor (vWF) and the apoptosis-associated cleaved cas- pase-3 (scale bar = 25 lm). The prevalence of cleaved caspase-3-positive ECs was plotted (mean ± SE) from three independent experiments with two inde- pendent lines of mouse ECs. Compared with ‘none’: *P <0.05(t test). (D), Mouse SMCs were grown in medium containing 10% serum from CKD or control mice; SMC numbers were quantitated at the indicated time points and plotted as means ± SE from three independent experiments performed with distinct SMC lines. (E), Mouse SMCs used in panel D were grown in medium containing 2.5% fetal bovine serum lacking (None) or containing IL-9 at 600 pg/mL; SMC numbers were quantitated at the indicated time points and plotted as means ± SE from three independent experiments.

. 38 promotes neointimal hyperplasia in the venous segment of jugulocarotid . contributes to allergic inflammation and perhaps to other diseases: 32–34 . AVFs through mechanisms that—unlike those we found in vein . levels of IL-9 are elevated in the plasma of humans with atherosclero- 32 . 39 grafts—may be independent of SMC proliferation. This vein neointimal . sis and in human atherosclerotic plaques ; the prevalence of (IL-9- 32 . þ hyperplasia in AVFs can be reduced by bone morphogenetic protein-7, . secreting) Th9 cells among all CD4 lymphocytes is higher among . 40 calcitriol (or genetic deficiency of intermediate-early response gene . acute coronary syndrome patients than among control subjects ; 35 36 . -/- 40 X-1), glucagon-like -1 receptor agonist, VEGF-A-targeting . and IL-9 promotes atherosclerosis in Apoe mice. Although CKD 37 31 . shRNA, heme oxygenase activity, and inhibition of endothelial cell . elevates plasma IL-9 levels in mice (Figure 3) and may elevate plasma Notch signalling.34 Our study extends these previous studies by demon- . levels of IL-9 in diabetic humans,41 it remains to be determined how . strating that CKD also potentiates neointimal hyperplasia and endothe- . CKD elevates plasma IL-9 levels: by augmenting IL-9 secretion from . lial toxicity in vein grafts, and that it does so largely by augmenting IL-9 . particular cell types, by reducing glomerular filtration of IL-9, or levels. . through a combination of these processes. . IL-9 is a 14.2 kDa polypeptide (32 kDa when fully glycosylated) . The role of mast cells in vascular diseases has been explored with a secreted by CD4þ T cells (Th9, Th2) and, to a lesser degree, granulo- . variety of models. In dinitrofluorobenzene-sensitized Apoe-/- mice, sys- 26,38 . cytes and mast cells. By binding to its heterodimeric receptor . temic or focal activation of mast cells augments atherosclerosis through comprising the IL-9Ra and the common c-chain receptor, IL-9 affects . mechanisms that can be inhibited by cromolyn and that appear to involve . mast cells, lymphocytes (Th2, Th9, Th17, Treg, B cells), airway epithe- . mast cell-dependent increases in leucocyte adhesion and plaque micro- lial cells and SMCs.26,38 IL-9 is a critical growth factor for mast cells; it . vessel permeability.17 Jugulocarotid AVFs develop markedly less vein 26 . 42 W/Wv also promotes proliferation of B lymphocytes and eosinophils. . neointimal hyperplasia in mast cell-deficient c-Kit mice than in Whereas IL-9 is important for immunity to parasites, it also . cognate wild-type mice.43 Congruent with these findings—and with our 1558 L. Zhang et al.

. . own study of mast cells in human AVFs (Figure 7)—c-Kitþ cells (i.e. pre- . 42 . dominantly mast cells )are50% more numerous in arterialized veins . of human brachiobasilic AVFs than in corresponding native veins har- . 43 . vested from the same subjects. In carotid interposition vein grafts, cro- . molyn has been used to reduce total wall thickness of vein grafts placed . -/- 44 . in Apoe mice. However, this study’s vein graft model is of question- . able relevance to vein grafts placed in humans: the vein grafts use end-to- . . end vein-to-artery attachments achieved by slipping the vein (telescopi- . cally) over the carotid artery’s cut ends, which have been everted over . . polyethylene tubing segments, and then tying the veins in place with cir- . cumferential ligatures (rather than sutured anastomoses); consequently, . . neointimal hyperplasia in this model is exceptionally aggressive (dis- . cussed in Ref. 15). In contrast, our vein grafts employ sutured, end-to- . . side vein-to-carotid anastomoses that model surgery performed on . human subjects.15 . . CKD reduced vein graft re-endothelialization in our vein graft model, . and re-endothelialization correlated inversely with the degree of vein graft . . neointimal hyperplasia. Similarly, CKD delays re-endothelialization of jugu- . 33 . lar veins used in AVFs of mice undergoing 5/6ths nephrectomy, and . 34 . reduces endothelial barrier function. In our CKD mice, anti-IL-9 IgG and . cromolyn therapies both augmented vein graft re-endothelialization and . . diminished neointimal hyperplasia. These findings accord with our pre- . 9,16 . vious work: by genetically manipulating TNF receptors or by adminis- . tering endothelial cells intravenously,22 we found that accelerating vein . . graft re-endothelialization diminishes neointimal hyperplasia. . CKD-associated endothelial deficits may help to explain why our . . CKD mouse vein grafts demonstrate IL-9 immunostaining throughout . the neointima, media, and adventitia (Figure 4,seeSupplementary mate . . rial online, Figure S3), even though the primary IL-9-responsive mast cells . and primary IL-9-producing T lymphocytes reside almost exclusively in . . the adventitia (see Supplementary material online, Figure S1). Plasma IL- . 9—at levels five-fold higher in CKD than in normal mice—permeates . . the damaged vein graft endothelium and diffuses throughout all layers of . the vein graft. The resultant high level of vein graft IL-9 does not appear . . to promote SMC proliferation (Figure 6), but does promote adventitial . mast cell proliferation and degranulation (Figure 5), a process that itself . . could subsequently promote SMC proliferation by increasing vein graft . 9 . levels of TNF (see Supplementary material online, Figure S1), among . other cytokines that can promote neointimal hyperplasia.27 In addition, . . mast cell protease-5 (‘chymase’) activates transforming growth factor b1 . 45 . and angiotensin II, both of which contribute to neointimal hyperplasia . 46–48 . in a variety of models. Indeed, inhibiting mast cell chymase reduces . 49 . canine arteriovenous fistula neointimal hyperplasia. Whether targeting . IL-9 therapeutically will prove practical in reducing vein graft disease or . Figure 7 Human veins demonstrate increased IL-9, IL-9Ra,andmast . arteriovenous fistula vein disease in humans with CKD remains to be cell prevalence during maturation of arteriovenous fistulas (AVFs) in . . determined. patients with end-stage renal disease. Vein specimens were obtained . from five humans with end-stage renal disease undergoing two- . . stage basilic vein transposition or radiocephalic AVF surgery (see . Supplementary material online, Table S1). (A), Serial frozen sections were . Supplementary material stained with trichrome (bottom) or, pairwise, with primary IgGs targeting . . IL-9, IL-9Ra, mast cell tryptase, or no particular protein (Control); sections . Supplementary material is available at Cardiovascular Research online. were counterstained with Hoechst 33 342 (DNA, blue). ‘L’ designates . . Conflict of interest: none declared. the luminal surface. Scale bars = 100 lm for immunofluorescence panels . and 500 lm for the trichrome panels. Shown are images from Subject 3 . (see Supplementary material online, Table S1); the boxed areas indicate . . Funding the regions imaged for immunofluorescence. (B), From five independent . This work was supported by the National Institutes of Health pairs of native and AVF vein specimens, protein immunofluorescence . . [P30DK096493 to SBG, LZ, ERH, and NJF; DK103082 to JHL, and from three random 200 fields was quantitated as in Figure 2,normalized . . HL121689 to NJF], the American Heart Association [15GRNT25710194 to values obtained for the native vein, and plotted as means ± S.E. . Compared with values from the native vein: *P <0.05. . to LZ], and the Edna and Fred L. Mandel, Jr. Foundation. Chronic kidney disease-dependent vein graft disease 1559

. References . 25. Fattori E, Cappelletti M, Costa P, Sellitto C, Cantoni L, Carelli M, Faggioni R, Fantuzzi . G, Ghezzi P, Poli V. Defective inflammatory response in -deficient mice. 1. Levey AS, de Jong PE, Coresh J, El Nahas M, Astor BC, Matsushita K, Gansevoort RT, . Kasiske BL, Eckardt KU. The definition, classification, and prognosis of chronic kidney . J Exp Med 1994;180:1243–1250. disease: a KDIGO Controversies Conference report. Kidney Int 2011;80:17–28. . 26. Steenwinckel V, Louahed J, Orabona C, Huaux F, Warnier G, McKenzie A, Lison D, . Levitt R, Renauld JC. IL-13 mediates in vivo IL-9 activities on lung epithelial cells but 2. Jha V, Garcia-Garcia G, Iseki K, Li Z, Naicker S, Plattner B, Saran R, Wang AY, Yang . not on hematopoietic cells. J Immunol 2007;178:3244–3251. CW. Chronic kidney disease: global dimension and perspectives. Lancet 2013;382: . 27. Wernersson S, Pejler G. Mast cell secretory granules: armed for battle. Nat Rev 260–272. . . Immunol 2014;14:478–494. 3. Anavekar NS, McMurray JJ, Velazquez EJ, Solomon SD, Kober L, Rouleau JL, White . 28. Galli SJ, Nakae S, Tsai M. Mast cells in the development of adaptive immune HD, Nordlander R, Maggioni A, Dickstein K, Zelenkofske S, Leimberger JD, Califf . responses. Nat Immunol 2005;6:135–142. RM, Pfeffer MA. Relation between renal dysfunction and cardiovascular outcomes . 29. Bro S, Bentzon JF, Falk E, Andersen CB, Olgaard K, Nielsen LB. Chronic renal failure after myocardial infarction. N Engl J Med 2004;351:1285–1295. . . accelerates atherogenesis in apolipoprotein E-deficient mice. J Am Soc Nephrol 2003; 4.GoAS,ChertowGM,FanD,McCullochCE,HsuCY.Chronickidneydiseaseandtherisks . 14:2466–2474. of death, cardiovascular events, and hospitalization. NEnglJMed2004;351:1296–1305. . 30. Bro S, Binder CJ, Witztum JL, Olgaard K, Nielsen LB. Inhibition of the renin- 5. Appleby CE, Ivanov J, Lavi S, Mackie K, Horlick EM, Ing D, Overgaard CB, Seidelin . angiotensin system abolishes the proatherogenic effect of uremia in apolipoprotein PH, von Harsdorf R, Dzavik V. The adverse long-term impact of renal impairment in . . E-deficient mice. Arterioscler Thromb Vasc Biol 2007;27:1080–1086. patients undergoing percutaneous coronary intervention in the drug-eluting stent . 31. Kang L, Grande JP, Hillestad ML, Croatt AJ, Barry MA, Katusic ZS, Nath KA. A new era. Circ Cardiovasc Interv 2009;2:309–316. . model of an arteriovenous fistula in chronic kidney disease in the mouse: beneficial 6. Wellenius GA, Mukamal KJ, Winkelmayer WC, Mittleman MA. Renal dysfunction . effects of upregulated heme oxygenase-1. Am J Physiol Renal Physiol 2016;310: increases the risk of saphenous vein graft occlusion: results from the Post-CABG . . F466–F476. trial. Atherosclerosis 2007;193:414–420. . 32. Kokubo T, Ishikawa N, Uchida H, Chasnoff SE, Xie X, Mathew S, Hruska KA, Choi 7. Bangalore S, Guo Y, Samadashvili Z, Blecker S, Xu J, Hannan EL. Revascularization in . ET. CKD accelerates development of neointimal hyperplasia in arteriovenous fistulas. patients with multivessel coronary artery disease and chronic kidney disease: . . J Am Soc Nephrol 2009;20:1236–1245. everolimus-eluting stents versus coronary artery bypass graft surgery. J Am Coll . 33. Liang A, Wang Y, Han G, Truong L, Cheng J. Chronic kidney disease accelerates Cardiol 2015;66:1209–1220. . endothelial barrier dysfunction in a mouse model of an arteriovenous fistula. Am J 8. Cai X, Freedman NJ. New therapeutic possibilities for vein graft disease in the post- . Physiol Renal Physiol 2013;304:F1413–F1420. edifoligide era. Future Cardiol 2006;2:493–501. . . 34. Wang Y, Liang A, Luo J, Liang M, Han G, Mitch WE, Cheng J. Blocking Notch in 9. Zhang L, Peppel K, Brian L, Chien L, Freedman NJ. Vein graft neointimal hyperplasia . endothelial cells prevents arteriovenous fistula failure despite CKD. J Am Soc Nephrol is exacerbated by receptor-1 signaling in graft-intrinsic cells. . 2014;25:773–783. Arterioscler Thromb Vasc Biol 2004;24:2277–2283. . 35. Brahmbhatt A, NievesTorres E, Yang B, Edwards WD, Roy Chaudhury P, Lee MK, 10. Zhang L, Brian L, Freedman NJ. Vein graft neointimal hyperplasia is exacerbated by . . Kong H, Mukhopadhyay D, Kumar R, Misra S, Sen U. The role of Iex-1 in the patho- CXCR4 signaling in vein graft-extrinsic cells. J Vasc Surg 2012;56:1390–1397. . genesis of venous neointimal hyperplasia associated with hemodialysis arteriovenous 11. Mann MJ, Gibbons GH, Kernoff RS, Diet FP, Tsao PS, Cooke JP, Kaneda Y, Dzau VJ. . fistula. PLoS One 2014;9:e102542. Genetic engineering of vein grafts resistant to atherosclerosis. Proc Natl Acad Sci USA . . 36. Chien CT, Fan SC, Lin SC, Kuo CC, Yang CH, Yu TY, Lee SP, Cheng DY, Li PC. 1995;92:4502–4506. . Glucagon-like peptide-1 receptor agonist activation ameliorates venous thrombosis- 12. Gupta J, Mitra N, Kanetsky PA, Devaney J, Wing MR, Reilly M, Shah VO, Balakrishnan . induced arteriovenous fistula failure in chronic kidney disease. Thromb Haemost 2014; VS, Guzman NJ, Girndt M, Periera BG, Feldman HI, Kusek JW, Joffe MM, Raj DS. . 112:1051–1064. Association between albuminuria, kidney function, and inflammatory biomarker pro- . . 37. Yang B, Janardhanan R, Vohra P, Greene EL, Bhattacharya S, Withers S, Roy B, file in CKD in CRIC. Clin J Am Soc Nephrol 2012;7:1938–1946. . Nieves Torres EC, Mandrekar J, Leof EB, Mukhopadhyay D, Misra S. Adventitial 13. Pedersen TX, Madsen M, Junker N, Christoffersen C, Vikesa J, Bro S, Hultgardh- . transduction of lentivirus-shRNA-VEGF-A in arteriovenous fistula reduces venous Nilsson A, Nielsen LB. deficiency dampens the pro-atherogenic effect . stenosis formation. Kidney Int 2014;85:289–306. of uraemia. Cardiovasc Res 2013;98:352–359. . . 38. Goswami R, Kaplan MH. A brief history of IL-9. J Immunol 2011;186:3283–3288. 14. Gurley SB, Mach CL, Stegbauer J, Yang J, Snow KP, Hu A, Meyer TW, Coffman TM. . 39. Gregersen I, Skjelland M, Holm S, Holven KB, Krogh-Sørensen K, Russell D, Influence of genetic background on albuminuria and kidney injury in Ins2(þ/C96Y) . Askevold ET, Dahl CP, Ørn S, Gullestad L, Mollnes TE, Ueland T, Aukrust P, (Akita) mice. Am J Physiol Renal Physiol 2010;298:F788–F795. . Halvorsen B, Khoury JE. Increased systemic and local interleukin 9 levels in patients 15. Zhang L, Hagen PO, Kisslo J, Peppel K, Freedman NJ. Neointimal hyperplasia rap- . . with carotid and coronary atherosclerosis. PLoS One 2013;8:e72769. idly reaches steady state in a novel murine vein graft model. J Vasc Surg 2002;36: . 40. Zhang W, Tang T, Nie D, Wen S, Jia C, Zhu Z, Xia N, Nie S, Zhou S, Jiao J, Dong 824–832. . W, Lv B, Xu T, Sun B, Lu Y, Li Y, Cheng L, Liao Y, Cheng X. IL-9 aggravates the 16. Zhang L, Sivashanmugam P, Wu JH, Brian L, Exum ST, Freedman NJ, Peppel K. . -/- . development of atherosclerosis in Apoe mice. Cardiovasc Res 2015;106:453–464. Tumor necrosis factor receptor-2 signaling attenuates vein graft neointima formation . 41. Vasanthakumar R, Mohan V, Anand G, Deepa M, Babu S, Aravindhan V. Serum IL-9, by promoting endothelial recovery. Arterioscler Thromb Vasc Biol 2007;28:284–289. . IL-17, and TGF-beta levels in subjects with diabetic kidney disease (CURES-134). 17. Bot I, de Jager SC, Zernecke A, Lindstedt KA, van Berkel TJ, Weber C, Biessen EA. . Cytokine 2015;72:109–112. Perivascular mast cells promote atherogenesis and induce plaque destabilization in . . 42. Dahlin JS, Hallgren J. Mast cell progenitors: origin, development and migration to tis- apolipoprotein E-deficient mice. Circulation 2007;115:2516–2525. . sues. Mol Immunol 2015;63:9–17. 18. Zhang L, Peppel K, Sivashanmugam P, Orman ES, Brian L, Exum ST, Freedman NJ. . 43. Skartsis N, Martinez L, Duque JC, Tabbara M, Velazquez OC, Asif A, Andreopoulos Expression of tumor necrosis factor receptor-1 in arterial wall cells promotes athe- . . F, Salman LH, Vazquez-Padron RI. c-Kit signaling determines neointimal hyperplasia in rosclerosis. Arterioscler Thromb Vasc Biol 2007;27:1087–1094. . arteriovenous fistulae. Am J Physiol Renal Physiol 2014;307:F1095–F1104. 19. Li H, Nourbakhsh B, Ciric B, Zhang GX, Rostami A. Neutralization of IL-9 amelio- . 44. de Vries MR, Wezel A, Schepers A, van Santbrink PJ, Woodruff TM, Niessen HW, rates experimental autoimmune encephalomyelitis by decreasing the effector . . Hamming JF, Kuiper J, Bot I, Quax PH. Complement factor C5a as mast cell activator population. J Immunol 2010;185:4095–4100. . mediates vascular remodelling in vein graft disease. Cardiovasc Res 2013;97:311–320. 20. Wu JH, Zhang L, Fanaroff AC, Cai X, Sharma KC, Brian L, Exum ST, Shenoy SK, . 45. Doggrell SA, Wanstall JC. Vascular chymase: pathophysiological role and therapeutic . Peppel K, Freedman NJ. G protein-coupled receptor kinase-5 attenuates atheroscle- . potential of inhibition. Cardiovasc Res 2004;61:653–662. rosis by regulating receptor tyrosine kinases and 7-transmembrane receptors. . 46. Wolff RA, Ryomoto M, Stark VE, Malinowski R, Tomas JJ, Stinauer MA, Hullett DA, Arterioscler Thromb Vasc Biol 2012;32:308–316. . Hoch JR. Antisense to transforming growth factor-beta1 messenger RNA reduces . vein graft intimal hyperplasia and monocyte chemotactic protein 1. J Vasc Surg 2005; 21. Conte MS, Nugent HM, Gaccione P, Roy-Chaudhury P, Lawson JH. Influence of dia- . betes and perivascular allogeneic endothelial cell implants on arteriovenous fistula . 41:498–508. remodeling. J Vasc Surg 2011;54:1383–1389. . 47. Otsuka G, Agah R, Frutkin AD, Wight TN, Dichek DA. Transforming growth factor 22. Brown MA, Zhang L, Levering VW, Wu JH, Satterwhite LL, Brian L, Freedman NJ, . beta 1 induces neointima formation through plasminogen activator inhibitor-1- . dependent pathways. Arterioscler Thromb Vasc Biol 2006;26:737–743. Truskey GA. Human umbilical cord blood-derived endothelial cells reendothelialize . vein grafts and prevent thrombosis. Arterioscler Thromb Vasc Biol 2010;30:2150–2155. . 48. Yamada T, Kondo T, Numaguchi Y, Tsuzuki M, Matsubara T, Manabe I, Sata M, Nagai 23. Pruijm M, Ponte B, Vollenweider P, Mooser V, Paccaud F, Waeber G, Marques-Vidal . R, Murohara T. Angiotensin II receptor blocker inhibits neointimal hyperplasia . through regulation of smooth muscle-like progenitor cells. Arterioscler Thromb Vasc P, Burnier M, Bochud M. Not all inflammatory markers are linked to kidney function: . results from a population-based study. Am J Nephrol 2012;35:288–294. . Biol 2007;27:2363–2369. 24. Jean-Charles PY, Zhang L, Wu JH, Han SO, Brian L, Freedman NJ, Shenoy SK. . 49. Jin D, Ueda H, Takai S, Okamoto Y, Muramatsu M, Sakaguchi M, Shibahara N, Ubiquitin-specific protease 20 regulates the reciprocal functions of b-arrestin2 in . Katsuoka Y, Miyazaki M. Effect of chymase inhibition on the arteriovenous fistula Toll-like receptor 4-promoted NFjB Activation. J Biol Chem 2016;291:7450–7464. . stenosis in dogs. J Am Soc Nephrol 2005;16:1024–1034. . . . SUPPLEMENTARY MATERIAL

Interleukin-9 Mediates Chronic Kidney Disease-dependent Vein Graft Disease: a Role for Mast Cells

Lisheng Zhang,1 Jiao-Hui Wu,1 James C. Otto,2 Susan B. Gurley,3 Elizabeth R. Hauser,4 Sudha K. Shenoy,1,5 Karim Nagi,1,5 Leigh Brian,1 Virginia Wertman,1 Natalie Mattocks,3 Jeffrey H. Lawson,2 Neil J. Freedman1,5

Departments of Medicine (1Cardiology, 3Nephrology), 2Surgery, 4Biostatistics and Bioinformatics, 5Cell Biology and the 4Molecular Physiology Institute Duke University Medical Center, Durham, North Carolina, USA 4Cooperative Studies Program Epidemiology Center Durham Veterans Affairs Medical Center, Durham, North Carolina, USA

SUPPLEMENTARY MATERIALS AND METHODS

Mice All mice were C57BL/6J, housed in a barrier facility on a 12-hour light-dark cycle; they received ad libitum water and standard chow LabDiet 5053 (irradiated). All animal experiments were performed according to protocols approved by the Duke Institutional Animal Care and Use Committee; these protocols complied with the Guide for the Care and Use of Laboratory Animals (National Academies Press, 2011).

Nephrectomy model of CKD The 5/6ths nephrectomy procedure was performed in two sequential operations, modified from published protocols.1-3 For analgesia, mice were treated with acetaminophen (300 mg/kg daily for 48hrs pre-op and post-op via drinking water), and injected with buprenorhphine SR (0.5 mg/kg at the time of surgery). Anesthesia with pentobarbital (50 mg/kg i.p.) was used for all surgeries. The left partial nephrectomy was performed first: we incised left flank through the dorsoventral muscles, and then performed lateral left renal capsulotomy (the capsule subsequently contracts toward the hilum). Next, the upper and lower poles of the left kidney were removed, leaving behind the segment around the hilum. Hemostasis on the cut renal surface was achieved by cauterization and/or Gelfoam®. The musculofascial and skin incisions were closed with 5-0 vicryl suture. Ten days later, we used a similar approach for right total nephrectomy: the renal pedicle was ligated with 5-0 vicryl, the vessels and ureter were severed, and the entire kidney was removed. For mice undergoing sham surgery, we performed only bilateral renal capsulotomy. For mice undergoing 1/3rd nephrectomy, we performed the left partial nephrectomy described above and a right renal capsulotomy. After the second stage of the nephrectomy surgeries, mice were allowed to recover for 2 wk before undergoing either GFR measurement or carotid interposition vein grafting. All mice in 5/6ths nephrectomy and corresponding control cohorts (sham nephrectomy or 1/3rd nephrectomy) were matched for age and sex; they did not differ with regard to weight pre- or post-operatively (22-23 gm prior to vein graft surgery). Blood pressure in C57BL/6J mice is not altered by 5/6ths nephrectomy.4,5 All control-GFR mice for vein graft surgeries were subjected to 1/3rd nephrectomy, which does not alter GFR (because total unilateral nephrectomy does not alter GFR).2

GFR measurement Under 2% isoflurane anesthesia, CKD and control mice matched for age and gender were injected retro-orbitally with sterile FITC-inulin at time 0; saphenous vein blood (~10 μl) was collected into heparinized capillary tubes 3, 7, 10, 15, 35, 55, and 75 min afterwards. Plasma fluorescence was determined and GFR calculations were performed as described.6

Carotid interposition vein grafting Interposition vein graft surgery was performed as we described previously,7-11 on mice that previously underwent 5/6ths or 1/3rd nephrectomy. All mice were wild-type (WT), C57BL/6J mice. Inferior venae cavae from donor mice were harvested after pentobarbital-induced anesthesia (50 mg/kg, i.p.). The inferior vena cava was stored in heparinized PBS while the recipient of the vein graft was anesthetized (pentobarbital, 50 mg/kg, i.p.), shaved and prepared for surgery. The donor inferior vena cava was anastomosed end-to-side to the right common carotid artery of control-GFR or CKD recipient mice. After both vein graft anastomoses were secured, the intervening common carotid artery was ligated and cut, thereby stretching the IVC graft. All vein graft donors and recipients were matched for gender and age (10-15 wk old). Grafts were harvested 4 weeks postoperatively from pentobarbital-anesthetized mice (50 mg/kg, i.p.), after perfusion-fixation with 3.7% formaldehyde (20 min, 80 mm Hg pressure), and prepared for histochemical and immunofluorescence analysis as we described previously.7-9

Page 2

Histology Paraffin-embedded vein grafts were sectioned at 5 μm, from the distal or middle third of the vein graft specimens; all vein graft groups were matched for section location. For morphometry, sections were stained with a modified Masson’s trichrome and Verhoeff’s elastic tissue stain that facilitates the simultaneous identification of collagen (green), elastin (black), cytoplasm (red), and nuclei (black).7 Morphometry was performed on vein grafts by observers blinded to specimen identity, as we have described.7,9,12 The neointimal/medial boundary was defined as the transition from the cytoplasm-rich, disorganized neointima to the collagen-rich media. The medial/adventitial boundary was defined as the transition from the more densely organized medial collagen to the less densely organized, vasa vasora- containing collagenous network of the adventitia. Neointimal area was measured as the cross- sectional area subtended by the luminal perimeter and the neointimal/medial boundary. Medial area was measured as the cross-sectional area subtended by the neointimal/medial and medial/adventitial boundaries. For each vein graft, lumen area was normalized to the cross-sectional area encompassed by the boundary between the media and the adventitia; this ratio was plotted as “lumen / total vein graft area”. Vein graft immunofluorescence was performed as described,9,12 using rabbit IgGs specific for the following proteins: vascular cell adhesion molecule-1 (VCAM-1, Santa Cruz Biotechnology, sc-8304); proliferating cell nuclear antigen (PCNA, Santa Cruz Biotechnology, sc-7907); NFκB p65 (Santa Cruz Biotechnology, sc-109); NFκB p65 phosphorylated on the IκB kinase target Ser536 (, monoclonal IgG #3033S); interleukin-9 (IL-9, Gene Tex, #GTX51537); IL-9 receptor-α (IL-9Rα, Santa Cruz Biotechnology, sc-1030); von Willebrand factor (vWF, DAKO, A0082); tryptase (mouse mast cell proteases [mMCPs] 6 and 7, Santa Cruz Biotechnology, sc-32889); mouse mast cell protease-5 (mMCP5, or “chymase”, MyBioSource, Inc., MBS2002348). Goat IgG was used to detect TNF and CD3 (Santa Cruz Biotechnology, sc-1351 and sc-1127, respectively). Mouse monoclonal IgGs were used to immunostain TNF receptor-1 (TNFR1, Santa Cruz Biotechnology, sc-8436) and collagen I (Sigma-Aldrich, c2456). Alexa Fluor® 488-conjugated rat IgG2a anti-mouse CD68 (Biolegend, #137012) was used to stain macrophages; non-immune Alexa Fluor® 488-conjugated rat IgG2a was used for negative control sections. Negative control serial sections were incubated with equivalent concentrations of non-immune, isotype control IgG in lieu of primary antibody. Cyanine 3- conjugated 1A4 IgG (Sigma, C6198) was used to detect smooth muscle α-actin, as described.9 The DNA-binding dye Hoechst 33342 (10 µg/mL, Invitrogen) was added to secondary antibody incubations. Single microscopic fields were imaged for multiple fluorophores using narrow band-pass filter cubes on a Leica DM6 B light/fluorescence microscope, as described.9,10,12 We compared vein graft protein expression between groups: control-GFR vs. CKD, control IgG-injected vs. anti-IL-9 IgG-injected. To do so, we stained and imaged vein grafts specimens from each group simultaneously, batch-wise. Identical exposure times and incident light intensities were used to visualize each specimen; images of each vein graft were captured with a Q charge-coupled device camera (Q Imaging, Inc.) at 200×. At this magnification only ~50% of each vein graft fit into a single photograph; consequently, immunofluorescence quantitation for each vein graft required 2 photographs (the results from which were averaged for each specimen). Immunofluorescence was analyzed using NIH ImageJ software by observers blinded to specimen identity, as we described.9,10 The “total” immunofluorescence signal for a particular vein graft (vein graft wall, neointima, etc.) was measured on samples incubated with a primary IgG targeting a specific protein. The “nonspecific” immunofluorescence was measured on a serial section incubated with an isotype negative control IgG. “Specific” immunofluorescence for each specimen was calculated by subtracting “nonspecific” from “total” immunofluorescence. Specific immunofluorescence was then normalized to (Hoechst 33342-stained) DNA fluorescence intensity (Hoechst 33342-stained) obtained in cognate portions of the same vein graft section—thereby normalizing protein immunofluorescence to cellularity.9,10 These protein/DNA ratios were averaged among all vein grafts within each group; the ratios were then normalized to those obtained for the control samples within each immunostaining/imaging cohort.

Page 3

The extent of vein graft re-endothelialization was determined as the percentage of luminal perimeter surfaced by von Willebrand factor-positive cells, as judged from vein graft cross sections measured by observers blinded to specimen identity, as we described.10 Toluidine blue stain for mast cells was performed as described previously.13,14 Briefly, de- paraffinized 5-μm vein graft sections were stained with 0.1% (w/v) toluidine blue in 154 mmol/L NaCl/7% (v/v) ethanol, pH 2.5 for 2 min (25 °C), washed with water and then dehydrated with 95-100% ethanol, followed by xylene. Mast cells were identified by their metachromatic (purple) granules, which are distinct from the orthochromatic (blue) staining of all other cells. Mast cells were dichotomized as (a) quiescent, with densely packed in the cytoplasm, or (b) activated, or degranulating, with disgorged and loosely packed granules.13,14 These mast cells were quantitated by observers blinded to specimen identity; a minimum of 100 mast cells per specimen were counted. Confocal microscopy was performed as we described,15 with a Zeiss LSM 510 META confocal microscope, to identify nuclear localization of the NFκB p65 subunit in the vein graft neointima and media. Sections were stained with rabbit IgG targeting the NFκB p65 subunit or non-immune rabbit IgG, followed by Alexa® 568-conjugated anti-rabbit IgG (Thermo Scientific) along with DRAQ5™ (to stain nuclear DNA [Alexis Biochemicals]). Confocal images were acquired such that each channel was scanned successively using filter sets for multitrack sequential excitation (568, and 633 nm) and emission ( 585-615 nm, Alexa® 568; 650 nm, DRAQ5). To facilitate reliable co-localization studies, a 100× oil objective was used, and a pin-hole for each channel was matched for airy units to set optical slice thickness at 1 μm.16 Final processing of images was performed with Adobe Photoshop CS2 software; the brightness and contrast were adjusted for entire images and to the same extent for p65 and control IgG images.

Serum cytokine assays To determine how CKD affects plasma cytokine levels and to determine how CKD serum affects endothelial cell proliferation, we harvested serum 2 wk after the final nephrectomy surgery from mice subjected to 5/6ths nephrectomy, sham nephrectomy, or 1/3rd nephrectomy. To determine the effects of anti-IL-9 IgG or cromolyn therapies on plasma cytokine levels, we harvested serum from CKD mice 4 weeks after vein graft implantation (and 2-3 days after the final injection of either anti-IL-9 IgG or cromolyn). Mice were anesthetized with pentobarbital and exsanguinated by ventricular puncture. Blood was allowed to coagulate in pyrogen-free tubes for 30 minutes at room temperature, and then centrifuged at 1,000 ×g (4 °C, 10 min). Supernatant serum was transferred to new tubes and spun again (1,000 ×g at 4 °C, 10 min); supernatant serum was aliquotted and stored at minus 80 °C until it was analyzed. Serum amyloid A concentrations were measured by ELISA with a kit from Biosource International, Inc., as we described.12 Mouse sera were analyzed in triplicate using the Bio-Plex Pro™ Mouse Cytokine Assay on a BioPlex 1036 Array Reader (BioRad, Inc). Individual sera from CKD (n = 13) versus control-GFR (n = 9) mice were compared in a single run (a single plate) of the Bio-Plex Pro™ Mouse Cytokine Assay, according to the manufacturer’s instructions. Similarly, individual sera from anti-IL-9 IgG-injected (n = 7) versus control IgG-injected (n = 7) mice and cromolyn-injected versus saline-injected mice were compared in single runs of the Bio-Plex Pro™ Mouse Cytokine Assay. The Bio-Plex Pro™ Mouse Cytokine Assay detects 23 cytokines (or 22, if IL-12 p70 and p40 subunits are counted as a single cytokine): interleukin-1α (IL-1α), IL-1β, IL-2, IL-3, IL-4, IL-5, IL-6, IL-9, IL-10, IL-12(p40), IL-12(p70), IL- 13, IL-17, KC (CXCL1), G-CSF (granulocyte colony-stimulating factor), IFNγ (-γ), MIP-1α (CCL3), MIP-1β (CCL4), RANTES (CCL5), and TNF (tumor necrosis factor).

Systemic neutralization of IL-9 At the time of vein graft surgery, CKD mice were injected with one of 2 Armenian Hamster IgG2/κ molecules: (a) D9302C12, which neutralizes IL-9,17 or (b) control IgG, which binds no known mouse protein (BioLegend): 50 μg intravenously plus 250 μg i.p. Subsequently, mice were injected with the same IgG three times per week (300 μg i.p.)17 until vein graft harvest 4 wk after implantation. (Hamster IgG is not immunogenic in mice.18) Mice were sacrificed for vein graft harvest 48-72 hours after the final injection of IgG.

Page 4

Serum creatinine assay Mouse sera were assayed for creatinine concentration using an enzymatic assay with a colorimetric endpoint (Creatinine Assay Kit, Sigma-Aldrich # MAK080),19,20 according to the manufacturer’s specifications. All mouse sera were run in triplicate, in a single assay.

Systemic treatment with cromolyn To prevent mast cell degranulation in CKD mice, we treated mice with the mast cell stabilizer cromolyn as reported by others.13,21 Cromolyn sodium (10 mg/ml, Teva Pharmaceuticals) was diluted 1:1 prior to intraperitoneal injection with 2× PBS (made from 10× PBS [VWR] and USP sterile water, Mediatech, Inc.). One day prior to vein grafting, CKD mice were injected i.p. with either PBS or cromolyn (50 mg/kg). After vein grafting, mice were injected with cromolyn or PBS twice per week for 4 weeks, after which vein grafts were harvested as described above.

Endothelial and smooth muscle cell proliferation Aortas were harvested from mice anesthetized with pentobarbital (50 mg/kg, i.p.). Endothelial cells (ECs) and smooth muscle cells (SMCs) were isolated by enzymatic digestion and passaged as we described.22 The purity of EC preparations was assessed by the prevalence of von Willebrand factor (vWF) immunofluorescence, which was ≥90%; three independently derived mouse EC lines were used in experiments. The purity of SMC preparations was assessed by the prevalence of smooth muscle α- actin staining, which was ≥90%; three independently derived mouse EC lines were used in experiments. For EC growth experiments, we used 96-well plates coated with collagen I. ECs for the growth experiments were plated at 2.5×103 cells/well in “medium A”: Dulbecco’s modified Eagle medium (DMEM) containing 20% FBS (which was previously heat-inactivated), 1× L-glutamine, 1× non- essential amino acids, 1× sodium pyruvate (Invitrogen), 25 mmol/L Hepes (pH 7.4), 100 μg/ml heparin, 100 μg/ml EC growth supplement (Sigma-Aldrich), and 1% penicillin/streptomycin (Invitrogen). In parallel, the same pool of ECs was plated in medium A on the “standard curve” plate at serial 2-fold dilutions ranging from 2.5×103 to 10×103 cells/well. After 16 h, ECs were washed twice with “medium B”, defined as medium A that lacks FBS. Next, ECs on the growth assay plate (sextuplicate wells) were re-fed with medium B lacking (control) or containing 20% (vol/vol) serum from three cohorts of mice (6 mice/cohort): (1) 5/6ths nephrectomized, (2) 1/3rd-nephrectomized, or (3) sham- nephrectomized. Both control and mouse serum-containing medium B also contained the same Armenian Hamster IgG2/κ molecules used in the in vivo IL-9 neutralization experiment: either D9302C12, which neutralizes IL-9,17 or control IgG, at 2 μg/ml ([final]). Parallel ECs on the standard curve plate were washed with PBS, stained with crystal violet (see below), and stored dry until the conclusion of the growth assay. ECs were incubated in this medium for 4 days (37 °C in a CO2 incubator). Subsequently, ECs were washed with PBS and then stained with 0.5% (w/v) crystal violet in 5% (v/v) ethanol for 60 min. After extensive washing, ECs were dried completely, and EC-adsorbed crystal violet was dissolved in 40% acetic acid. The absorbance of each well was then read at 562 nm in sextuplicate, and corrected for the absorbance obtained in wells in which no ECs had been plated. EC protein was quantitated with a standard curve generated from crystal violet readings obtained from the standard control plate. The relationship between OD562 and EC number was linear from 0 to 10×103 ECs/well (data not shown). To determine the effect of mouse serum on EC proliferation, we subtracted from each well the number of ECs present in control wells containing just medium B. EC growth data were equivalent with mouse serum from sham-nephrectomized and 1/3rd-nephrectomized mice (data not shown). SMCs were cultivated in medium D: DMEM with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (Gibco). Prior to proliferation assays, confluent SMCs incubated for 24 hr in serum-free medium (DMEM with 20 mmol/L Hepes pH 7.4, 0.1% (w/vol) fatty acid-free bovine serum albumin (Sigma-Aldrich), 1% penicillin/streptomycin); they were then trypsinized and plated in 96-well plates at 2×103 cells/well in medium D. Four hours later, SMCs were washed with PBS and re-fed with one of the following media: (a) DMEM with 2.5% FBS and 1% penicillin/streptomycin, lacking (control)

Page 5 or containing IL-9 (R&D Systems) at 600 pg/mL; (b) DMEM with 10% (vol/vol) CKD mouse serum and 1% penicillin/streptomycin; (c) DMEM with 10% (vol/vol) control-GFR mouse serum and 1% penicillin/streptomycin. SMCs were grown in these media for 1-8 days; SMC number was assessed by a crystal violet-binding assay exactly parallel to that described above for ECs. Standard curves were linear from 0 to 20×103 SMCs/well (data not shown).

EC apoptosis experiments Mouse ECs were plated at 7,500 per well onto 8-well culture slides (Nunc/Thermo), in medium A lacking FBS and supplemented with vehicle or recombinant mouse IL-9 (R&D Systems) at a final concentration of 600 pg/mL. ECs were incubated at 37 °C in a CO2 incubator for 2 days and then processed for immunofluorescence microscopy. ECs were fixed for 2 min in ethanol:formalin:water (75:15:10, vol/vol), rinsed thrice in phosphate- buffered saline (PBS, 3 min/wash), and then blocked for 30 min (25 °C) in “blocking buffer”: 4% (vol/vol) FBS/0.2% (w/vol) gelatin/PBS. Next, ECs were incubated in blocking buffer containing rabbit monoclonal anti-cleaved caspase-3 diluted 1:25 (Cell Signaling, #9664) or non-immune rabbit IgG for 1 hr (25 °C), and then rinsed in PBS as above. ECs were then incubated for 1 hr (25 °C) in blocking buffer containing (a) FITC-conjugated sheep IgG targeting von Willebrand factor diluted 1:25 (Abcam, #ab8822) and (b) Alexa® 546-conjugated anti-rabbit IgG diluted 1:50 (Molecular Probes). After 3 rinses performed as above, slides were covered with Fluoromount-G (Southern Biotech) and coverslipped for microscopy. Non-immune rabbit IgG used as a primary IgG yielded no red immunofluorescence.

Human vein samples from arteriovenous fistulae (AVFs) All procedures with human volunteers were approved by the Duke Institutional Review Board, under Protocol Number Pro00040569, and conformed to the principles outlined in the Declaration of Helsinki. All human subjects gave their informed consent prior to their inclusion in the study and collection of vein specimens. Vein specimens were obtained by a single surgeon (J. H. L.) from 5 humans with end- stage renal disease when they underwent two-stage basilic vein transposition AVF surgery or radiocephalic AVF surgery.23 For two-stage basilic vein transposition, the basilic vein was anastomosed to the brachial artery and allowed to mature in place. Because the basilic vein resides deep within the arm (and therefore cannot be easily cannulated for hemodialysis), after maturation of the AVF, the AVF’s basilic vein must be transposed to a subcutaneous location to facilitate access for hemodialysis. Thus, between 3 and 5 months after initial AVF creation, a second surgery detached the (arterialized) basilic vein from its original anastomosis, tunneled it under the skin, and reconnected it to the brachial artery at a new anastomosis. Samples of “native vein,” 3-5 mm in length, were obtained intraoperatively during the initial construction of the AVF. “AVF vein” samples were also obtained intraoperatively, when the arterialized basilic vein transposition. In one case, we procured an “AVF vein” sample 11 months following radiocephalic AVF construction, during surgery required for AVF revision. Native and AVF vein samples were embedded in OCT and frozen in the operating room, immediately after removal from each human volunteer. Serial frozen sections were stained with trichrome or, in pairs of “native” and “AVF” vein from single subjects, with primary IgGs targeting IL-9, IL-9Rα, mast cell tryptase, or no particular protein; sections were counterstained with Hoechst 33342 (DNA, blue). Immunofluorescence data were acquired and processed as described above for mouse vein samples.

Statistical analyses All values were plotted in figures as means ± SE and cited in the text as means ± SD. GraphPad Prism® software was used to perform t tests (for comparing two groups), 1-way ANOVA with Tukey’s post-hoc test for multiple comparisons (≥3 groups), or 2-way ANOVA with Sidak’s post-hoc test for multiple comparisons. A p value of less than 0.05 was considered significant. To analyze data from serum cytokine assays, the distributions of each cytokine were examined. We made 2 comparisons of “treatment” versus “control”: (1) CKD mice versus control-GFR mice, and (2) anti-IL-9-IgG-injected CKD mice versus control IgG-injected CKD mice. We compared the medians between the treatment and control groups using the non-parametric Wilcoxon Test with

Page 6 exact p value calculation, under the null hypothesis that there was no difference between the treatment and control groups. A conservative multiple comparisons correction for 23 independent cytokines was used to indicate a significant test: p < 0.05/23, or p < 0.00217.

SUPPLEMENTARY REFERENCES

1. Bro S, Moeller F, Andersen CB, Olgaard K, Nielsen LB. Increased expression of adhesion molecules in uremic atherosclerosis in apolipoprotein-E-deficient mice. J Am Soc Nephrol 2004;15:1495-1503. 2. Qi Z, Whitt I, Mehta A, Jin J, Zhao M, Harris RC, Fogo AB, Breyer MD. Serial determination of glomerular filtration rate in conscious mice using FITC-inulin clearance. Am J Physiol Renal Physiol 2004;286:F590-596. 3. Pedersen TX, Madsen M, Junker N, Christoffersen C, Vikesa J, Bro S, Hultgardh-Nilsson A, Nielsen LB. Osteopontin deficiency dampens the pro-atherogenic effect of uraemia. Cardiovasc Res 2013;98:352-359. 4. Bernardi S, Candido R, Toffoli B, Carretta R, Fabris B. Prevention of accelerated atherosclerosis by AT1 receptor blockade in experimental renal failure. Nephrol Dial Transplant 2011;26:832-838. 5. Ma LJ, Fogo AB. Model of robust induction of glomerulosclerosis in mice: importance of genetic background. Kidney Int 2003;64:350-355. 6. Gurley SB, Mach CL, Stegbauer J, Yang J, Snow KP, Hu A, Meyer TW, Coffman TM. Influence of genetic background on albuminuria and kidney injury in Ins2(+/C96Y) (Akita) mice. Am J Physiol Renal Physiol 2010;298:F788-795. 7. Zhang L, Hagen PO, Kisslo J, Peppel K, Freedman NJ. Neointimal hyperplasia rapidly reaches steady state in a novel murine vein graft model. J Vasc Surg 2002;36:824-832. 8. Zhang L, Freedman NJ, Brian L, Peppel K. Graft-extrinsic cells predominate in vein graft arterialization. Arterioscler Thromb Vasc Biol 2004;24:470-476. 9. Zhang L, Peppel K, Brian L, Chien L, Freedman NJ. Vein graft neointimal hyperplasia is exacerbated by tumor necrosis factor receptor-1 signaling in graft-intrinsic cells. Arterioscler Thromb Vasc Biol 2004;24:2277-2283. 10. Zhang L, Sivashanmugam P, Wu JH, Brian L, Exum ST, Freedman NJ, Peppel K. Tumor necrosis factor receptor-2 signaling attenuates vein graft neointima formation by promoting endothelial recovery. Arterioscler Thromb Vasc Biol 2008;28:284-289. 11. Zhang L, Brian L, Freedman NJ. Vein graft neointimal hyperplasia is exacerbated by CXCR4 signaling in vein graft-extrinsic cells. J Vasc Surg 2012;56:1390-1397. 12. Zhang L, Peppel K, Sivashanmugam P, Orman ES, Brian L, Exum ST, Freedman NJ. Expression of tumor necrosis factor receptor-1 in arterial wall cells promotes atherosclerosis. Arterioscler Thromb Vasc Biol 2007;27:1087-1094. 13. Bot I, de Jager SC, Zernecke A, Lindstedt KA, van Berkel TJ, Weber C, Biessen EA. Perivascular mast cells promote atherogenesis and induce plaque destabilization in apolipoprotein E-deficient mice. Circulation 2007;115:2516-2525. 14. Combs JW, Lagunoff D, Benditt EP. Differentiation and proliferation of embryonic mast cells of the rat. J Cell Biol 1965;25:577-592.

Page 7

15. Shenoy SK, Lefkowitz RJ. Trafficking patterns of beta-arrestin and G protein-coupled receptors determined by the kinetics of β-arrestin deubiquitination. J Biol Chem 2003;278:14498-14506. 16. North AJ. Seeing is believing? A beginners' guide to practical pitfalls in image acquisition. J Cell Biol 2006;172:9-18. 17. Li H, Nourbakhsh B, Ciric B, Zhang GX, Rostami A. Neutralization of IL-9 ameliorates experimental autoimmune encephalomyelitis by decreasing the effector T cell population. J Immunol 2010;185:4095-4100. 18. Sheehan KC, Pinckard JK, Arthur CD, Dehner LP, Goeddel DV, Schreiber RD. Monoclonal antibodies specific for murine p55 and p75 tumor necrosis factor receptors: identification of a novel in vivo role for p75. J Exp Med 1995;181:607-617. 19. Yasuhara M, Fujita S, Arisue K, Kohda K, Hayashi C. A new enzymatic method to determine creatine. Clin Chim Acta 1982;122:181-188. 20. Burris D, Webster R, Sheriff S, Faroqui R, Levi M, Hawse JR, Amlal H. Estrogen directly and specifically downregulates NaPi-IIa through the activation of both estrogen receptor isoforms (ERα and ERβ) in rat kidney proximal tubule. Am J Physiol Renal Physiol 2015;308:F522-534. 21. Wang J, Sjoberg S, Tia V, Secco B, Chen H, Yang M, Sukhova GK, Shi GP. Pharmaceutical stabilization of mast cells attenuates experimental atherogenesis in low-density lipoprotein receptor-deficient mice. Atherosclerosis 2013;229:304-309. 22. Wu JH, Zhang L, Fanaroff AC, Cai X, Sharma KC, Brian L, Exum ST, Shenoy SK, Peppel K, Freedman NJ. G Protein-coupled receptor kinase-5 attenuates atherosclerosis by regulating receptor tyrosine kinases and 7-transmembrane receptors. Arterioscler Thromb Vasc Biol 2012;32:308-316. 23. Conte MS, Nugent HM, Gaccione P, Roy-Chaudhury P, Lawson JH. Influence of diabetes and perivascular allogeneic endothelial cell implants on arteriovenous fistula remodeling. J Vasc Surg 2011;54:1383-1389.

Page 8

Table: Clinical Characteristics of Human Vein Donors

AVF Native Age Subject AVF Vein Maturation Ethnicity Sex Co-morbidities Vein (y) (months)

1 Basilic Mature 5 70 EA M HTN, DM2, CAD, CHF, HL

2 Basilic Mature 3 48 AA M HTN, DM1, PAD Mature and stenotic segments 3 Cephalic 11 32 AA M HTN, HIV (from radiocephalic AVF revision) 4 Basilic Mature 4 47 EA F HTN, DM, PAD, CVA, CHF, HL

5 Basilic Mature 4 31 LA M HTN, DM1

“AVF Maturation” denotes the time between the initial surgery (at which native vein was harvested) and the subsequent surgery (stage 2 of the brachiobasilic AVF or revision of the radiocephalic AVF). Abbreviations: EA, European American; AA, African American; LA, Latino American; HTN, hypertension; DM2, type II diabetes mellitus; CAD, coronary artery disease; CHF, congestive heart failure; HL, hyperlipidemia; DM1, type I diabetes mellitus; PAD, peripheral artery disease; HIV, human immunodeficiency virus; CVA, cerebrovascular accident.

Table S1. Characteristics of human subjects from whom AVF veins were harvested during surgery. Vein specimens were obtained from 5 humans with end-stage renal disease undergoing 2-stage basilic vein transposition or radiocephalic AVF surgery: (a) at the initial construction of the AVF (native vein), and (b) 3-5 months following AVF construction, at the time of brachiobasilic AVF transposition, or 11 months following radiocephalic AVF construction, at the time of AVF revision—when vein arterialization had already taken place (AVF vein). A Mouse GFR Control CKD Control CKD Control CKD DNA ; DNA DNA -actin α TNF; TNFR1; SM DNA DNA DNA PCNA; Ctrl IgG; Collagen-1; DNA DNA DNA

Ctrl IgG; Ctrl IgG; CD68; DNA DNA

CD3; Ctrl IgG; DNA

B * 200 * * Ctrl IgG; 150 *

100 Protein / DNA

(CKD/Control, %) 50

0 TNF CD3 α-actin PCNA CD68 TNFR1 SM Collagen-1 Vein Graft Protein

Figure S1. CKD augments vein graft neointimal hyperplasia and inflammation. (A) Serial sections of vein grafts from Figures 1 and 2 were immunostained for the indicated protein as in Figure 2, with protein-specific or isotype control IgGs from rabbit (PCNA), mouse (TNFR1, collagen I), rat (CD68), or goat (TNF, CD3). Vein grafts from control-GFR and CKD mice were stained batch-wise. Scale bars = 50 μm. (B) The ratios of red or green (protein) to blue (DNA) pixels in the vein graft wall were quantitated by Image J, and normalized to the cognate ratios obtained for control specimens in each staining cohort, to obtain “CKD/Control”, plotted as means ± SE for 3-4 vein grafts in each GFR group. Compared with control: *, p < 0.05 (2-way ANOVA with Sidak post-hoc test). Mouse GFR A Control CKD p65 DNA DNA

+ p65 DNA

+ Neg Ctl IgG B 30 *

20

10 Nuclear p65 (% of total cells) 0 Control CKD Mouse GFR

Figure S2. CKD augments NFκB activation in vein grafts. (A), Serial sections of vein grafts from Figure 2 and Supplementary Figure 1 were immunostained with rabbit IgG targeting the NFκB p65 subunit (p65) or no specific protein (Neg Ctl IgG). Alexa 546-conjugated anti-rabbit IgG and DRAQ5™ (for nuclear DNA) were used on all specimens. Specimens from control and CKD mice were stained concurrently, and imaged by confocal microscopy with identical settings (see Methods). (B), Neointimal and medial nuclei were counted manually by observers blinded to specimen identity (at least 200 per vein graft), and the percent of nuclei evincing red p65 staining was quantitated from 4 vein grafts of each GFR group, as indicated, with means ± SE. Compared with control: *, p < 0.01 (t test). A phospho-p65, DNA IL-9Rα, DNA IL-9, DNA Neg Ctrl IgG, DNA Control Anti-IL-9

CKD Mouse Injections: IgG

IgG Treatment B Control Anti-IL-9 125

100

75

50 * *

(% of control) *

25

Protein Immunofluorescence 0 phospho- IL-9Ra IL-9 p65 Vein Graft Protein

Figure S3. IL-9 exacerbates vein graft inflammation. (A) Serial sections of vein grafts from Figure 5 were stained with Hoechst 33342 (blue, DNA) and immunostained (red) with rabbit IgG specific for NFκB p65 phosphorylated on Ser536 (phospho-p65), IL-9 receptor-α (IL-9Rα), IL-9, or no particular protein (Neg Control). Scale bars = 50 μm. (B) Protein immunofluorescence throughout the vein graft wall was normalized to cognate DNA fluorescence intensity; resulting ratios were normalized within each staining group to the mean of samples from control IgG-injected CKD mice, to yield “% of control,” plotted as means±S.E. of 4 vein grafts per group. Compared with control: *, p < 0.05 (2-way ANOVA with Sidak post-hoc test). 1200 IgG Treatment 800 Control anti-IL-9 400 * * 100

50 * [Cytokine], pg/ml * * 0 IL-1a IL-6 IL-9 IL-10 IL-12 Serum Cytokine

Figure S4. Inhibiting IL-9 reduces the serum levels of multiple cytokines. CKD mice were injected 3 times per week with anti-IL-9 IgG or isotype control IgG for 4 wk. Two to three days after the final IgG injections, mice were sacrificed for the harvest of vein grafts (Figure 5) and serum. Serum from each mouse was assayed for cytokines in triplicate by a single run of the Bio-Plex Pro™ Mouse Cytokine assay (as in Figure 3). Plotted are the means ± SE for the indicated serum cytokine levels in mice treated with control or anti-IL-9 IgG (n = 6-7/group). Compared with control IgG-injected mice: *, p < 0.002 (Wilcoxon Test). V ) ein Graft Endothelialization (%) CKD Mouse Treatment 4 Mouse Treatment A Vehicle Cromolyn B Vehicle

x 10 40 2 Cromolyn 60 60 Area (%)

40 40 Area (µm ein Graft * 20 richrome T otal V * T 20 20

0 Lumen / 0 0 Cross Sectional Intima Media Lumen Endothelium

DNA Vein Graft Layer ; Immunofluorescence / DNA Mouse Treatment

C ) Vehicle 100

Cromolyn (% of control) SM α-actin 10

50 5 DNA % of adventitial cells ; ( Mast Cell Prevalence

vWF 0 * 0 Mast Cells Quiescent Activated Total Vein Graft Stain Toluidine Blue Anti-Tryptase

D Treatment 800 Vehicle l Cromolyn m /

g 400 p , ] e n i

k 100 o t y C [ 50

0 IL-1a IL-6 IL-9 IL-10 IL-12 Serum Cytokine

Figure S5. Inhibition of mast cell degranulation attenuates CKD-dependent vein graft disease. Mice were subjected to 5/6ths nephrectomy and vein graft surgery as in Figure 1. On the day prior to vein graft surgery and thrice weekly thereafter, mice were injected i.p. with PBS lacking (vehicle) or containing cromolyn sodium (50 mg/kg). Vein grafts were harvested 4 wk post-operatively, as in Figures 1 and 5. (A) Serial sections of perfusion-fixed vein grafts were stained with a modified Masson trichrome or immunofluorescently for smooth muscle (SM) α-actin or von Willebrand factor (vWF, green), along with Hoechst 33342 counterstain (blue, DNA). Scale bars = 100 μm. (B) The indicated cross sectional areas were calculated by planimetry and plotted as means ± SE from 8 mice in each group. For each vein graft, luminal area was normalized to cognate total vein graft cross-sectional area prior to plotting. “Vein graft endothelialization” was calculated as in Figure 1, and plotted as means ± SE from 5 mice in each group. Compared with control mice: *, p < 0.05. (C) Serial sections of vein grafts from panel A were stained with toluidine blue or anti-tryptase IgG to identify mast cells, as in Figure 4. The prevalence of quiescent and activated mast cells in vein grafts was assessed by toluidine blue staining as in Figure 4, and the prevalence of total mast cells was assessed by anti-tryptase staining and normalized to that found in vein grafts from vehicle-treated mice, as presented in Figure 4. Plotted are the means ± S.E. of 4 vein grafts per group (at least 100 adventitial mast cells were counted per specimen). Compared with vein grafts from vehicle-treated mice: *, p < 0.05 (2-way ANOVA with Sidak post-hoc test). (D) Serum from each mouse depicted in panel A was assayed for cytokines in triplicate by a single run of the Bio-Plex Pro™ Mouse Cytokine assay (as in Figure 3 and Supplementary Figure 4). Average values for each mouse are plotted, along with means ± SE for the indicated serum cytokine levels in mice treated with vehicle or cromolyn (n = 8/group).