Serum Concentrations of Markers of TNF␣ and Fas-Mediated Pathways and Renal Function in Nonproteinuric Patients with Type 1 Diabetes

Monika A. Niewczas,*†‡ Linda H. Ficociello,* Amanda C. Johnson,* William Walker,* Elizabeth T. Rosolowsky,*§ Bijan Roshan,* James H. Warram,* and Andrzej S. Krolewski*† *Research Division of Joslin Diabetes Center, Boston, Massachusetts; †Department of Medicine, Harvard Medical School, Boston Massachusetts; ‡Department of Immunology, Transplant Medicine and Internal Diseases, Warsaw Medical University, Warsaw, Poland; and §Division of Endocrinology, Children’s Hospital, Boston, Massachusetts

Background and objectives: The aim of our study was to examine serum markers of the TNF and Fas pathways for association with cystatin-C based estimated glomerular filtration rate (cC-GFR) in subjects with type 1 diabetes (T1DM) and no proteinuria. Design, setting, participants, & measurements: The study group (the 2nd Joslin Kidney Study) comprised patients with Impaired renal function (cC-GFR <90 .(304 ؍ or microalbuminuria (MA) (n (363 ؍ T1DM and normoalbuminuria (NA) (n ml/min) was present in only 10% of patients with NA and 36% of those with MA. We measured markers of the ␣ (TNF␣) pathway [TNF␣, soluble TNF receptor 1 (sTNFR1), and 2 (sTNFR2)], its downstream effectors [soluble intercellular and soluble vascular adhesion molecules (sICAM-1 and sVCAM-1), 8 (IL8/CXCL8), chemoattractant -1 (MCP1), and IFN␥ inducible protein-10 (IP10/CXCL10)], the Fas pathway [soluble Fas (sFas) and (sFasL)], CRP, and IL6. Results: Of these, TNF␣, sTNFRs, sFas, sICAM-1, and sIP10 were associated with cC-GFR. However, only the TNF receptors and sFas were associated with cC-GFR in multivariate analysis. Variation in the concentration of the TNF receptors had a much stronger impact on GFR than clinical covariates such as age and albumin excretion. Conclusions: Elevated concentrations of serum markers of the TNF␣ and Fas-pathways are strongly associated with decreased renal function in nonproteinuric type 1 diabetic patients. These effects are independent of those of urinary albumin excretion. Follow-up studies are needed to characterize the role of these markers in early progressive renal function decline. Clin J Am Soc Nephrol 4: 62–70, 2009. doi: 10.2215/CJN.03010608

he traditional model of the development of end-stage Low-grade chronic inflammation is thought to be involved in renal disease in type 1 diabetes mellitus (T1DM), in the pathogenesis of diabetic nephropathy (3,4). Tumor necrosis T which microalbuminuria (MA) leads to proteinuria and factor alpha (TNF␣/TNF) is a key mediator of inflammation then proteinuria is followed by renal function loss, has been and plays a role in apoptosis. In animal models, its effects on challenged recently. Increase in urinary albumin excretion is an kidneys include reduced glomerular filtration rate (GFR) and important determinant of diabetic nephropathy progression, increased albumin permeability (3). It mediates its signal via but it does not entirely explain this phenomenon. For example, two distinct receptors, TNF receptor 1 (TNFR1/TNFRSF1A) the loss of renal function commences earlier than previously and TNF receptor 2 (TNFR2/TNFRSF1B), which are mem- recognized and precedes the onset of proteinuria (1). In our brane-bound and also present in soluble form in serum (5). longitudinal study of T1DM (The First Joslin Study of Natural TNF␣ mediates its inflammatory effects by induction of a broad History of Microalbuminuria), renal function decline began spectrum of , including (IL8/CXCL8); with the onset of MA in about one-third of patients and pro- chemotactic protein-1 (MCP-1/CCL2); IFN-␥ induc- gressed in a linear fashion from normal kidney function to renal ible protein-10 (IP-10/CXCL10); and adhesion molecules such insufficiency. Also, renal function decline occurred in a notice- as intercellular adhesion molecule-1 (ICAM-1) and vascular able proportion of patients with T1DM and normal albumin adhesion molecule-1 (VCAM-1) (6,7). excretion (1,2). The Fas pathway mediates apoptosis and may play a role in the progression of diabetic nephropathy (8–11). The binding of Received June 18, 2008. Accepted September 8, 2008. Fas ligand (FasL) to Fas, its membrane-bound receptor that is also present in serum in soluble form (sFasL, sFas), leads to an Published online ahead of print. Publication date available at www.cjasn.org. apoptotic response (12,13). Correspondence: Dr. Andrzej S. Krolewski, Section on Genetics and Epidemiol- ␣ ogy, Joslin Diabetes Center, One Joslin Place, Boston, MA, 02215. Phone: 617-732- Most studies on serum markers of TNF -mediated inflam- 2668; Fax: 617-732-2667; E-mail: [email protected] mation and apoptosis in diabetic nephropathy have explored

Copyright © 2009 by the American Society of Nephrology ISSN: 1555-9041/401–0062 Clin J Am Soc Nephrol 4: 62–70, 2009 Markers of TNF, Fas Pathways and Renal Function 63

Table 1. Characteristics of the study group according to albuminuria statusa

Characteristics Normoalbuminuria (n ϭ 363) Microalbuminuria (n ϭ 304) P

AERb (␮g/min) 15 (11 to 21) 69 (45 to 131) By design Age (yr) 39 Ϯ 12 41 Ϯ 12 Ͻ0.05 Male (%) 44% 61% Ͻ0.0001 Diabetes duration (yr) 20 Ϯ 923Ϯ 10 Ͻ0.0001 HbA1cc (%) 8.3 Ϯ 1.2 8.6 Ϯ 1.5 Ͻ0.01 cC-GFRd (ml/min/1.73 m2) 118 Ϯ 24 99 Ϯ 27 Ͻ0.0001 cC-GFR categories: Ͼ130 ml/min 30% 10% 90 to 130 61% 54% 60 to 89 9% 28% Ͻ60 1% 8%

aData are mean Ϯ SD, median (quartiles), or percent. bAER: median albumin excretion rate during the preceding 2-yr window cHbA1c: mean hemoglobin A1c during the preceding 2-yr window dcC-GFR: estimated GFR based on serum cystatin-C

their association with MA and proteinuria rather than with be more effective if implemented 5 to 10 yr earlier in the disease GFR (14). course. The goal of this large cross-sectional study was to investigate In this study, the GFR was estimated by a cystatin C-based whether serum concentrations of markers mediated by TNF␣ formula (cC-GFR), previously shown as an accurate way of (sTNFR1, sTNFR2, sICAM-1, sVCAM-1, IL8, MCP-1, IP-10) or evaluating renal function in patients with diabetes (15,16) involved in Fas-related apoptosis (sFasL and sFas) are associ- ated, independently from albuminuria, with variation in renal Materials and Methods function in patients with T1DM who do not have proteinuria or The Committee on Human Subjects of the Joslin Diabetes Center advanced renal function impairment. This knowledge should approved the protocol and informed consent procedures for this study. facilitate the development of new diagnostic tools for identify- The study group was selected from the population attending the ing patients with early renal function decline and help the Joslin Clinic, a major center for the treatment of patients of all ages with search for intervention protocols for high-risk patients that may T1DM or type 2 diabetes mellitus (T2DM). The population is about 90%

Table 2. Characteristics of the study group according to albuminuria status and group-specific median cC-GFRa

Variable Normoalbuminuria Microalbuminuria Group Contrast

Characteristic cC-GFR Ͼ 115 cC-GFR Ͻ 115 cC-GFR Ͼ 101 cC-GFR Ͻ 101 AER cC-GFR N 183 180 152 152 Pb Pc AER (␮g/min) 13 (10-18) 18 (12-23) 56 (42-100) 85 (51-161) By Design Ͻ0.0001 Age (yr) 37 Ϯ 11 40 Ϯ 13 36 Ϯ 12 45 Ϯ 11 Ͻ0.05 Ͻ0.0001d Diabetes duration (yr) 19 Ϯ 921Ϯ 10 20 Ϯ 926Ϯ 9 Ͻ0.0001 Ͻ0.0001d HbA1c (%) 8.3 Ϯ 1.2 8.3 Ϯ 1.2 8.7 Ϯ 1.6 8.4 Ϯ 1.4 Ͻ0.005 NS Body mass index (kg/m2) 25.6 Ϯ 3.6 26.7 Ϯ 4.3 27.2 Ϯ 4.8 27.7 Ϯ 5.2 Ͻ0.0005 Ͻ0.05 Systolic blood pressure 118 Ϯ 12 120 Ϯ 13 124 Ϯ 12 125 Ϯ 15 Ͻ0.0001 NS (mmHg) ACEI or ARB Rx (%)d 18% 21% 49% 55% Ͻ0.0001 NS Antihypertensive Rx (%) 7% 16% 14% 30% Ͻ0.001 Ͻ0.0001 Serum cholesterol 183 Ϯ 29 181 Ϯ 29 190 Ϯ 33 193 Ϯ 30 Ͻ0.0001 NS (mg/dl) Lipid-lowering Rx (%) 24% 34% 31% 42% Ͻ0.05 Ͻ0.005 Current smoker (%) 9% 12% 19% 18% Ͻ0.005 NS

aData are mean Ϯ SD, median (quartiles), or %. bP-value for the albuminuria main effect in an ANOVA cP-value for the cC-GFR main effect in an ANOVA. dACEI, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; Rx, treatment. 64 Clinical Journal of the American Society of Nephrology Clin J Am Soc Nephrol 4: 62–70, 2009

Table 3. Serum concentrations of markers of inflammation or apoptosis according to AER group and cC-GFR above or below mediana

Normoalbuminuria Microalbuminuria Group Contrast

Variable cC-GFR Ͼ 115 cC-GFR Ͻ 115 cC-GFR Ͼ 101 cC-GFR Ͻ 101 AER cC-GFR

(n ϭ 182) (n ϭ 181) (n ϭ 152) (n ϭ 152) Pb Pc

TNF-mediated pathway TNF␣ ( pg/ml) 3.6 (2.3, 4.8) 3.9 (2.8, 5.8) 4.0 (2.6, 5.4) 4.8 (3.3, 6.4) NS Ͻ0.005 sTNFR1 (pg/ml) 1.2 (1.0, 1.4) 1.4 (1.2, 1.7) 1.4 (1.2, 1.6) 2.0 (1.6, 2.5) Ͻ0.0001 Ͻ0.0001 sTNFR2 (ng/ml) 2.1 (1.7, 2.6) 2.6 (2.1, 3.6) 2.3 (1.9, 2.9) 3.2 (2.5, 5.4) Ͻ0.0001 Ͻ0.0001 Potential downstream effectors Chemokines IL8 (pg/ml) 4.4 (2.4, 10.4) 6.1 (3.4, 13.3) 7.6 (3.8, 18.3) 7.0 (4.0, 15.5) Ͻ0.05 NS IP-10 (pg/ml) 107 (79, 136) 122 (88, 171) 102 (75, 141) 115 (80, 158) NS Ͻ0.001 MCP-1 (pg/ml) 124 (75, 184) 120 (77, 184) 113 (78, 191) 105 (77, 174) NS NS Adhesion molecule sICAM-1 (ng/ml) 133 (109, 152) 137 (119, 169) 149 (123, 173) 152 (123, 191) Ͻ0.0005 Ͻ0.005 sVCAM-1 (ng/ml) 386 (301, 481) 389 (303, 489) 376 (295, 467) 394 (330, 495) NS NS Fas-mediated pathway sFasL (pg/ml) 0.00.12 (0.08, 0.19) 0.13 (0.07, 0.20) 0.12 (08, 0.18) 0.11 (0.06, 0.16) NS NS sFas (ng/ml) 3.8 (3.0, 4.7) 4.5 (3.7, 5.5) 4.5 (3.6, 5.6) 5.4 (3.7, 6.9) Ͻ0.0001 Ͻ0.0001 Other inflammatory markers CRP (␮g/ml) 1.2 (0.5, 3.2) 1.1 (0.6, 2.7) 1.4 (0.5, 3.9) 1.6 (0.8, 3.2) Ͻ0.05 NS IL6 (pg/ml) 0.8 (0.6, 1.4) 0.9 (0.7, 1.5) 0.8 (0.4, 1.3) 0.9 (0.6, 2.2) NS NS

aData are medians (quartiles); analyses were done on concentrations transformed to their logarithms. bP-value for the albuminuria main effect in an ANOVA. cP-value for the cC-GFR main effect in an ANOVA. sTNFR1, soluble TNF receptor 1; sTNFR2, soluble TNF receptor 2; IP-10, IFN-␥ inducible protein-10; MCP-1, monocyte chemotactic protein-1; sICAM-1, soluble intracellular adhesion molecule-1; sVCAM-1, soluble vascular adhesion molecule-1; sFasL, soluble Fas ligand; sFas, soluble Fas; CRP, C-reactive protein.

Caucasian, and most reside in eastern Massachusetts. Between January study group included 667 participants: 304 with MA and 363 with 1, 2003 and December 31, 2004, patients with T1DM attending the Joslin normoalbuminuria (NA). Clinic were recruited into the Second Joslin Study on the Natural History of Microalbuminuria. Detailed descriptions of the Joslin Clinic population and the recruitment protocol for this study have been Assessment of Exposure Variables published previously (17). Eligibility criteria included residence in New Current and past use of medications (particularly angiotensin con- England, diabetes diagnosed before age 40 yr, treatment with insulin, verting enzyme inhibitors, angiotensin II receptor blockers, and other current age 18 to 64 yr, diabetes duration 3 to 39 yr, and multiple antihypertensive drugs) was recorded during the enrollment interview measurements in the preceding 2-yr interval of hemoglobin A1c and supplemented by examination of clinic records to confirm prescrip- (HbA1c) and urinary albumin-to-creatinine ratio (ACR). For each pa- tion dates. We also extracted all archived clinical laboratory measure- tient, the measurements of HbA1c were summarized by the mean, and ments of HbA1c, ACR, and serum cholesterol. Details of the assays the measurements of ACR by the median. Exclusion criteria included used were described previously (18,19). ACR values were converted to proteinuria (median ACR Ն 250 for men and Ն 355 ␮g/min for albumin excretion rate (AER) according to a formula published previ- women), end-stage renal disease, other serious illness, extreme obesity ously (19). For characterizing patients’ recent exposures, repeated mea- (body mass index Ͼ 40 kg/m2), or a median HbA1c less than 6.5% (near sures were summarized by their median (AER) or mean (HbA1c, cho- normoglycemia). lesterol, lipids).

Enrollment and Examination Sample Collection and Laboratory Measurements Trained recruiters administered a structured interview and brief Enrollment blood samples were drawn by venipuncture into sterile examination to eligible patients at a routine visit to the clinic (i.e., the collection tubes (SST Plus BD Vacutainer, BD, New Jersey); centrifuged enrollment visit). The interview solicited the history of diabetes and its at 3600 rpm for 10 min at 6°C (Centrifuge 5810 R); and then aliquoted treatment, other health problems, and use of medications. The recruiter into 1.5-ml sterile, nontoxic, nonpyrogenic tubes cryogenic tubes (Cryo- measured seated blood pressure twice (5 min apart) with an automatic Tubes CryoLine System; NUNC TM Serving Life Science) and frozen at monitor (Omron Healthcare, Inc), averaged them to reduce variability, Ϫ80°C until further analysis. Length of storage, defined as the interval and obtained samples of blood and urine. At of the end of 2004, this between the dates of sample collection and assay determination (range Clin J Am Soc Nephrol 4: 62–70, 2009 Markers of TNF, Fas Pathways and Renal Function 65

A. B. Serum Markers of TNF␣ and Fas-Mediated Pathways All markers were measured by immunoassay. Samples were thawed, 135 135 p<0.0001 ] ] 2 p<0.0001 2 vortexed, and centrifuged, and measurements were performed in the 125 125 supernatant. We measured sTNFR1, sTNFR2, and IL6 by ELISA 115 115 (DRT100, DRT200, and high-sensitive immunoassay HS600B, respec- 105 105 [ml/min/1.73m [ml/min/1.73m tively; R&D, Minneapolis, Minnesota) according to the manufacturer’s 95 95 p<0.0001 85 85 p<0.0001 protocol. We measured IL6 in only a subset of the study group (156 NA NA mean cC-GFR 75 mean cC-GFR 75 individuals). We measured the serum concentrations of the other pro- MA MA TR 1 TR 1 TR 2 TR 2 tein markers in a multiplex assay run on the Luminex platform. This is TR 3 TR 3 sTNFR1 sTNFR2 a multiplex particle-enhanced, sandwich type, liquid-phase immuno- assay with laser-based detection system based on flow cytometry. We C. D. used -panel B (HADK2-61K-B; Linco-Milipore) to measure

135 ␣ 135 TNF ; human sepsis-apoptosis panel (HSEP-63K; Linco-Milipore) to ] ]

2 p=0.043 2 p=0.0008 125 125 measure sFas, sFasL, sICAM-1, and sVCAM-1; and Beadlyte human

115 115 multi- detection (48-011; Upstate-Milipore) with protocol B to

105 105 measure IL8, IP-10, MCP-1. Protocols provided by vendors were fol- [ml/min/1.73m [ml/min/1.73m

95 95 lowed. Briefly, the method included use of 96-well filter plates (Mili- p=0.008 p=0.0013 85 85 pore), the capture antibodies specific for each analyte bound covalently NA NA mean cC-GFR 75 mean cC-GFR 75 to fluorescently labeled microspheres, biotinylated detection antibod- MA MA TR 1 TR 1 TR 2 TR 2 ies, and streptavidin-phycoerythrin. Detection incorporates two lasers TNFa TR 3 sFas TR 3 and a high-tech fluidics system (Luminex 100S, Austin, Texas). Values E. F. of median fluorescence intensity were fitted to a five-parameter logistic standard curve (20). 135 135 p=0.043 p=0.0087 Assay sensitivities were: TNF␣, 0.14 pg/ml; sTNFR1 and sTNFR2, ] ] 2 125 2 125 0.77 pg/ml; sFas, 7 pg/ml; sFasL, 6 pg/ml; sICAM-1, 30 pg/ml; sV- 115 115 CAM-1, 33 pg/ml; IL8, 0.7 pg/ml; IP-10, 1.2 pg/ml; MCP-1, 1.9 pg/ml; 105 105 [ml/min/1.73m [ml/min/1.73m and IL6, 0.04 pg/ml. If required, samples were diluted (sTNFR1, sT- 95 95 NFR2, sFAS, sFASL, sICAM-1, and sVCAM-1). The number of freeze- 85 p=0.0014 85 ns NA ␣ NA 75 thaw cycles was one for all measurements of TNF , IL8, IP-10, MCP-1, mean cC-GFR mean cC-GFR 75 MA TR 1 MA TR 1 TR 2 and for most measurements of the other analytes. The number did not TR 2 TR 3 TR 3 IP10 exceed two for any measurement. sICAM-1 Two internal serum controls were prepared in the same manner as study samples and were stored in a large number of aliquots at Ϫ80°C. Normoalbuminuria (NA) Microalbuminuria (MA) Aliquots of the two controls were included in each assay (21) for estimating the interassay coefficient of variation (CV). For most assays, Figure 1. Mean serum cystatin C GFR (cC-GFR) in the study interassay CV was between 8.5 and 15.8% (15.8% TNF␣, 13.0% sTNFR1, population of individuals with type 1 diabetes according to 12.7% sTNFR2, 8.5% sFas, 13.5% sFasL, 8.1% sVCAM-1, and 14.7% albuminuria status (NA, normoalbuminuria; MA, microalbu- IP-10). It was higher for the remaining three (25.2% sICAM-1, 33.3% minuria) and tertile (T1, T2, T3) of an inflammatory marker: (A) IL8, and 28.4% MCP-1). Immunoassay for TNF␣, sFas, and sFasL de- sTNFR1; (B) sTNFR2; (C) TNF␣; (D) sFas; (E) sICAM-1; and (F) tects the free form of the protein, whereas ELISA for sTNFR1 and IP-10. P value for trend across the tertiles in NA and MA, sTNFR2 detects the total amount of protein, free and bound with their respectively. ligand TNF␣, (information provided by manufacturer).

Statistical Analyses 2 to 5 yr), was included as a covariate in the analysis to estimate the Analyses were done in SAS (SAS Institute, Cary, North Carolina, extent of degradation of each analyte during storage. version 9.1.3). For continuous variables and frequencies, t-tests and ␹2 tests with alpha ϭ 0.05 were used, respectively. Analyses in Tables 2 and 3 and Figure 1 were ANOVA for unbalanced design. Linear re- cC-GFR gression with cC-GFR as the dependent variable was used for multi- Serum cystatin C concentration (Dade Behring Diagnostics) was as- variate analysis. AER and serum concentrations of the markers were sayed on a BN Prospec System nephelometer (Dade Behring, Inc., transformed to their logarithms for analysis. Missing data for serum Newark, Delaware). The range of detection is 0.30 to 7.50 mg/L, and markers never decreased the study sample by more than 5% in any the reported reference range for young, healthy persons is 0.53 to 0.95 model, so no remedial action was taken. mg/L. In our laboratory, the intraindividual coefficient of variation for subjects with diabetes is 3.8 and 3.0% in samples from the lowest and highest quartiles of the cystatin C distribution, respectively (1). Results The estimated GFR (cC-GFR ml/min/1.73 m2) is the reciprocal of Characteristics of the Study Population cystatin C (mg/L) multiplied by 86.7 and reduced by subtracting 4.2. This project included 667 patients with T1DM from the Sec- MacIsaac et al. recently developed this formula as a reliable estimate of ond Joslin Study on the Natural History of Microalbuminuria GFR in patients with diabetes. Our method for measuring cystatin-C who were recruited according to their urinary AER during the was similar with respect to assay, equipment, and coefficient of varia- 2-yr interval preceding enrollment: 363 with NA and 304 with tion as that reported by MacIsaac et al. (15). MA. Selected characteristics at their enrollment are summa- 66 Clinical Journal of the American Society of Nephrology Clin J Am Soc Nephrol 4: 62–70, 2009

Table 4. Spearman correlation coefficients between cC-GFR, AER, and serum markers of inflammation and apoptosis in the study group

AER TNF␣ sTNFR1 sTNFR2 sFas IP-10 sICAM

cC-GFR Ϫ0.31 Ϫ0.15 Ϫ0.57 Ϫ0.56 Ϫ0.27 Ϫ0.13a Ϫ0.17 AER 1.00 0.11 0.41 0.28 0.04c Ϫ0.12b 0.20 TNFa 1.00 0.11a 0.20 0.34 0.19 0.17 sTNFR1 1.00 0.81 0.26 0.20 0.21 sTNFR2 1.00 0.32 0.26 0.27 sFas 1.00 0.14a 0.12a IP-10 1.00 0.14a sICAM 1.00

aP Ͻ 0.01, bP Ͻ 0.05, cP ϭ NS, otherwise all other P Ͻ 0.0001. rized in Table 1 according to AER group. In the NA group, the clinical characteristics to the large variation in renal function 25th, 50th, and 75th percentiles of the AER distribution (11, 15, within the study group, we divided the NA and MA groups at and 21 ␮g/min) were centered in the NA range (Ͻ30 ␮g/min), the group-specific median cC-GFR (115 and 101 ml/min, re- but in the MA group these AER percentiles (45, 69, 131 ␮g/min) spectively) into groups with higher and lower cC-GFR (Table were entirely in the lower half of the MA range (30 to 300 2). The median (25th, 75th percentiles) of the resulting distribu- ␮g/min). In comparison with the NA group, the MA group had tions of cC-GFR in the NA groups was 136 (125,148) and 102 an older age, higher proportion of men, longer duration of (92,109) ml/min, and in the MA groups was 115 (108,124) and diabetes, higher HbA1c, and significantly lower cC-GFR. The 82 (64,91) ml/min. All of the characteristics in Table 2 were difference in cC-GFR between the two study groups was clearer significantly different between NA and MA groups, but many when renal function was grouped into categories, the latter two were not significantly different between the groups with higher of the four corresponding to mild and moderate renal function and lower cC-GFR (two-way ANOVA). For example, the ex- impairment, present in 36% of the MA group but only in 10% of pected associations of higher HbA1c, systolic blood pressure, the NA group. and serum cholesterol with MA were present, as were the To distinguish the relative contributions of AER and various associations of cigarette smoking and treatment with an angio-

Table 5. Mean cC-GFR at the 25th, 50th, and 75th percentiles of each significant covariate and the corresponding estimates adjusted for the other covariates

Univariate Analysis Multivariate Analysisa Covariate Percentile cC-GFR cC-GFR 2 P 2 P (ml/min/1.73m ) (ml/min/1.73m )

Age (yr) Ͻ0.0001 Ͻ0.002 31 25th 115 114 40 50th 109 112 48 75th 104 110 AER (␮g/min) Ͻ0.0001 Ͻ0.0001 22 25th 119 115 39 50th 111 112 79 75th 102 108 sTNFR1 (pg/ml) Ͻ0.0001 Ͻ0.0001 1216 25th 121 120 1442 50th 112 112 1764 75th 101 103 sFas (pg/ml) Ͻ0.0001 Ͻ0.008 3.63 25th 112 113 4.50 50th 110 112 5.72 75th 107 111

aAdjusted r2 for the multivariate model was 0.45, whereas it was 0.41 after adjustment for sTNFR1 and sFas only. Adjustments for gender, HbA1c, body mass index, antihypertensive and lipid-lowering treatments, and duration of storage samples did not significantly modify the associations. Clin J Am Soc Nephrol 4: 62–70, 2009 Markers of TNF, Fas Pathways and Renal Function 67 tensin converting enzyme inhibitor or angiotensin II receptor is the most pronounced for sTNFR1, and it is hardly changed blocker. However, none of these characteristics were associated by multivariate adjustment. Adjustment for the other poten- with lower cC-GFR. In contrast, older age and longer diabetes tially relevant clinical covariates—such as gender, HbA1c, body duration were significantly associated with both MA and lower mass index, renoprotective and other antihypertensive treat- cC-GFR, as was evidenced by medical attention represented by ment and lipid-lowering treatment, and duration of storage of treatment with antihypertensive or lipid lowering agents. serum specimens—did not modify the association of sTNFR1 and Fas with cC-GFR. When the analysis was repeated using Markers of Inflammation or Apoptosis and Impaired Renal sTNFR2 instead of sTNFR1, the result was similar, indicating Function—Univariate Analyses that measurement of either receptor yields roughly the same Serum concentrations of markers of inflammation or apopto- information. sis were examined in the same manner as the characteristics shown in Table 2. Four markers (sTNFR1, sTNRF2, sFas, and Discussion sICAM-1) were significantly associated both with AER and In this large cross-sectional study we examined twelve serum with cC-GFR (Table 3). TNF␣ and IP-10 were significantly markers of inflammation and apoptosis. We sought to discover associated only with the cC-GFR group and two (IL8 and a profile of markers that is associated with renal function C-reactive protein) were significantly associated only with the impairment in patients with T1DM and either NA or MA. To AER group. our knowledge, the novelty of this study is its primary focus on For the six markers significantly associated with cC-GFR in cC-GFR (not albuminuria) as an outcome in early diabetic Table 3, the patterns of association are illustrated in Figures 1A nephropathy and its attempt to differentiate the observed effect through 1F. Separately for the NA and MA groups, patients of markers on GFR from their potential associations with AER. were grouped according to the tertiles of the distribution of We attempted this approach in both uni- and multivariate each marker, and the mean cC-GFR for each subgroup was analyses. In univariate analyses, six markers were unrelated to depicted as a vertical bar. In both AER groups, the decrease in renal function (C-reactive protein, IL6, IL8, MCP-1, sVCAM-1, cC-GFR with increasing marker concentration was steepest for and sFasL) and six were significantly associated with variation sTNFR1 and sTNFR2. The pattern was similar for TNF␣, but in cC-GFR (TNF␣, sTNFR1, sTNFR2, sFas, sICAM-1, and IP-10). the differences among subgroups were smaller. For all three Among the six, the associations of TNF receptors with de- markers, the decrease appeared steeper in the MA group than creased cC-GFR were the strongest. in the NA group. For the remaining three markers (sICAM-1, Of the six markers, only the concentrations of sTNFR1, sT- IP-10 and sFas), a pattern of differences among subgroups was NFR2, and sFas contributed independently to cC-GFR. The less obvious effect of TNF receptors on cC-GFR was much more pronounced We studied these markers further by examining their corre- than the effects of clinical covariates such as age and AER lations with each other, and with the two nephropathy mea- (Table 5). Furthermore, serum concentrations of sTNFR1 and sures, cC-GFR and AER (Table 4). The negative correlations sTNFR2 are highly correlated (Spearman r ϭ 0.81) and show between the six markers and cC-GFR recapitulate the negative roughly the same associations with cC-GFR. Therefore, further associations shown in Table 3 and Figure 1. All pairs of markers studies are required to determine whether measurement of are significantly correlated, but the coefficients are generally both rather than just one is worthwhile. Our study provides modest. Only the correlation of the two receptors (sTNFR1 and evidence for the first time that markers of TNF␣- and Fas- sTNFR2) with cC-GFR and with each other exceeded 0.50. Note mediated pathways are strongly associated with variation in the poor (although significant) correlations between TNF␣ and cC-GFR in patients with T1DM and early diabetic nephropathy. its receptors (r ϭ 0.11 for TNF␣/sTNFR1 and r ϭ 0.20 for This association is independent of the association of these TNF␣/sTNFR2). markers with AER. Our findings support the hypothesis that The independence of the associations of these six markers of inflammation and apoptosis are involved in early renal func- inflammation or apoptosis with cC-GFR was examined in mul- tion decline in T1DM. tiple regression models. Only sTNFR1, sTNFR2, and sFAS re- Other cross-sectional studies in T1DM reported that serum mained significant when all were included in the model. Al- concentrations of TNF␣-related markers were elevated in com- though sTNFR2 was statistically significant in this model, its parison with healthy subjects and that the higher concentra- contribution was small because of its high collinearity with tions of these markers were associated with elevated urinary sTNFR1, so it was not retained in subsequent modeling. Most albumin excretion (14,22). Cross-sectional association between notable about this model was that the serum markers alone serum concentrations of sTNFR and variation in GFR has been (sTNFR1 and Fas) explained 41% of the variation in cC-GFR shown in type 2 diabetes mellitus (23), as well as in nondiabetic (adjusted r2), and addition of age and AER to the model only individuals (24,25). In the prospective CARE study, high serum increased the adjusted r2 to 45% (Table 5). Addition of the other concentrations of sTNFR2 were found to be associated with clinical covariates from Table 2 did not improve the adjusted r2 faster progression of renal function loss (26); however, all sub- (data not shown). The relative influence of these covariates on jects in that study had chronic kidney disease (GFR Ͻ 60ml/ cC-GFR is summarized in Table 5 by the cC-GFR estimated at min/1.73m2) at baseline. the 25th, 50th, and 75th percentiles of each covariate, with and Whatever mechanisms could underlie a causal relationship without adjustment for other covariates. The effect on cC-GFR between renal function decline and elevated serum concentra- 68 Clinical Journal of the American Society of Nephrology Clin J Am Soc Nephrol 4: 62–70, 2009 tions of sTNFR1 or TNFR2 (and whether those mechanisms weak, and disappeared in multivariate analyses, as one would include an activated TNF␣ pathway) remain to be discovered. expect if their effect were not independent of the TNF receptors Soluble receptors bind TNF␣ and may serve as a slow-release or sFas. reservoir of TNF␣ in a diabetic (and possibly low-grade inflam- Analysis of the Fas-mediated pathway revealed an indepen- matory) state (27). There is also some experimental evidence for dent effect of the serum concentration of sFas on variation in activation of TNF␣-pathway in diabetes (3,28). Possible factors cC-GFR and a lack of an effect of the serum concentration of that could influence serum concentrations of TNF receptors sFasL. A similar pattern of disparate effects of sFas and sFasL include their upstream regulators in serum, such as TNF␣ or was previously demonstrated in individuals with advanced IL1, and intramembrane activity of ADAM17 (TNF receptor kidney disease (11). Also, in a few individuals with T1DM and sheddase) (29). Concentrations of interleukin 1 below our de- without proteinuria, sFas was reported to correlate with both tection limits prevented us from measuring it reliably (data not ACR and GFR (35). shown), and measurement of ADAM17 was not possible with The mechanism of action of soluble Fas receptor has not been the methods we used in our study. well known but may be similar to that of TNF receptors in that How elevated concentrations of soluble TNF receptors may it leads to an enhanced Fas-mediated response in the kidney. lead to renal injury is not known. If they represent an activated The Fas-related system is involved mainly in regulation of TNF␣-pathway, several mechanisms may be involved. The apoptosis (10), whereas the TNF-system regulates apoptotic TNF␣-pathway has a broad range of inflammatory and apo- and inflammatory responses. Consistent with this is the tubu- ptotic properties. Dysregulation of these processes may contrib- lointerstitial apoptosis seen in strepotozocin-induced diabetic ute to injury of the diabetic kidney. In addition, the TNF␣- rats (8) and in human diabetic kidneys (9). Some evidence also pathway directly increases glomerular vasoconstriction and suggests that TNF␣ may induce Fas-mediated apoptosis albumin permeability. Exposure of the kidney to TNF␣ in- (36,37). In our study serum concentrations of TNF␣ and sFas creases mRNA expression of TNF receptors in renal tubuloint- were markedly correlated. erstitium and triggers cell death. Also, an apoptotic response The main limitation of our study is its cross-sectional study follows exposure of human kidney cells to sTNFR as well. This design; therefore inferences about the causality of the associa- effect is more pronounced after exposure to sTNFR1 than to tions remain tentative. Furthermore, high interassay CV of sTNFR2 (3,30,31). Therefore, our findings may support a hy- some of the measured markers would have weakened or ob- pothesis that elevated serum concentrations of soluble TNF scured true associations with cC-GFR. Furthermore, the assay receptors, by themselves or as markers of activation of TNF␣ for TNF␣ only measures the free form of TNF␣. In low-grade pathway, contribute to early renal function decline. chronic inflammation (which we expect to be the case in this One may argue that the association of TNF␣ receptors and condition), most circulating TNF␣ is bound to its receptors and cC-GFR simply reflects impaired renal handling of these pro- undetected by the assay used. This fact may account for the teins. Indeed, these receptors are cleared mainly by the kidneys noticeably poor correlations between TNF␣ and its receptors as as shown by tracer studies of radiolabeled sTNFR2 in animals well as its association with cC-GFR being weaker than that of its (32). Also, serum concentrations of soluble TNF receptors in- receptors. crease in advanced renal failure, as demonstrated in binephrec- In conclusion, this study provides the first clinical evidence tomized mice (32) and in human studies (33). However, most that markers of the TNF- and Fas-mediated pathways are patients in our study had normal renal function, and even the strongly associated with GFR in patients with T1DM and NA or renal function loss resulting from uninephrectomy does not MA. sTNFR1, sTNFR2, and sFas are the markers most strongly raise serum sTNF receptor concentrations in animals (32). representing these associations. These findings support the hy- Moreover, serum concentrations of sFasL, which has a molec- pothesis that inflammation and apoptosis are involved in renal ular mass similar to soluble TNF receptors, is not associated function decline in patients with T1DM and early diabetic with cC-GFR, whereas the receptors are strongly associated nephropathy. with variation in cC-GFR. On the basis of those data, potentially decreased clearance of those molecules has to be mentioned Acknowledgments here, but it does not stand for the most likely explanation of our This study was supported by grant DK-41526 from the National findings. Institutes of Health. Adhesion molecules and chemokines are potential down- M.A. Niewczas was supported by American Diabetes Association stream effectors of the TNF–sTNFRs inflammatory pathway (6). mentor-based fellowship # 7-03-MN-28. The authors thank Harry Expression of IL8, MCP-1, and IP-10 mRNA is induced in Spaulding for his assistance in the preparation of this manuscript. TNF␣-activated peripheral blood mononuclear cells taken from individuals with diabetes, but not from healthy ones (7). Ex- Disclosures pression and serum concentrations of chemokines and adhe- None. sion molecules, VCAM-1 and ICAM-1, increase as diabetic nephropathy develops (7,34). In our univariate analysis, serum concentrations of IP-10 and sICAM-1 were associated with References variation in cC-GFR and they correlated with their potential 1. Perkins BA, Ficociello LH, Ostrander BE, Silva KH, Wein- upstream regulators. Nevertheless, the observed effects were berg J, Warram JH, Krolewski AS: Microalbuminuria and Clin J Am Soc Nephrol 4: 62–70, 2009 Markers of TNF, Fas Pathways and Renal Function 69

the risk for early progressive renal function decline in type Binns AL, Roshan B, Warram JH, Krolewski AS: High- 1 diabetes. J Am Soc Nephrol 18: 1353–1361, 2007 normal serum uric acid is associated with impaired glo- 2. Caramori ML, Fioretto P, Mauer M: Low glomerular filtra- merular filtration rate in nonproteinuric patients with type tion rate in normoalbuminuric type 1 diabetic patients: An 1 diabetes. Clin J Am Soc Nephrol 2008 indicator of more advanced glomerular lesions. Diabetes 52: 18. Ficociello LH, Perkins BA, Silva KH, Finkelstein DM, Ig- 1036–1040, 2003 natowska-Switalska H, Gaciong Z, Cupples LA, Aschen- 3. Navarro JF, Mora-Fernandez C: The role of TNF-alpha in grau A, Warram JH, Krolewski AS: Determinants of pro- diabetic nephropathy: Pathogenic and therapeutic implica- gression from microalbuminuria to proteinuria in patients tions. Cytokine Rev 17: 441–450, 2006 who have type 1 diabetes and are treated with angiotensin- 4. Galkina E, Ley K: Leukocyte recruitment and vascular converting enzyme inhibitors. Clin J Am Soc Nephrol 2: injury in diabetic nephropathy. J Am Soc Nephrol 17: 368– 461–469, 2007 377, 2006 19. Krolewski AS, Laffel LM, Krolewski M, Quinn M, Warram 5. Macewan DJ: TNF receptor subtype signaling: Differences JH: Glycosylated hemoglobin and the risk of microalbu- and cellular consequences. Cell Signal 14: 477–492, 2002 minuria in patients with insulin-dependent diabetes mel- 6. Segerer S, Nelson PJ, Schlondorff D: Chemokines, chemo- litus. N Engl J Med 332: 1251–1255, 1995 kine receptors, and renal disease: From basic science to 20. Gottschalk PG, Dunn JR: The five-parameter logistic: A pathophysiologic and therapeutic studies. J Am Soc Nephrol characterization and comparison with the four-parameter 11: 152–176, 2000 logistic. Anal Biochem 343: 54–65, 2005 7. Wong CK, Ho AW, Tong PC, Yeung CY, Kong AP, Lun 21. Aziz N, Nishanian P, Fahey JL: Levels of and SW, Chan JC, Lam CW: Aberrant activation profile of immune activation markers in plasma in human immuno- cytokines and mitogen-activated protein kinases in type 2 deficiency virus infection: Quality control procedures. Clin diabetic patients with nephropathy. Clin Exp Immunol 149: Diagn Lab Immunol 5: 755–761, 1998 123–131, 2007 22. Schram MT, Chaturvedi N, Schalkwijk CG, Fuller JH, Ste- 8. Kumar D, Zimpelmann J, Robertson S, Burns KD: Tubular houwer CD: Markers of inflammation are cross-sectionally and interstitial cell apoptosis in the streptozotocin-diabetic associated with microvascular complications and cardio- rat kidney. Nephron Exp Nephrol 96: e77–e88, 2004 vascular disease in type 1 diabetes—the EURODIAB Pro- 9. Kumar D, Robertson S, Burns KD: Evidence of apoptosis in spective Complications Study. Diabetologia 48: 370–378, human diabetic kidney. Mol Cell Biochem 259: 67–70, 2004 2005 10. Schelling JR, Nkemere N, Kopp JB, Cleveland RP: Fas- 23. Lin J, Hu FB, Rimm EB, Rifai N, Curhan GC: The associa- dependent fratricidal apoptosis is a mechanism of tubular tion of serum lipids and inflammatory biomarkers with epithelial cell deletion in chronic renal failure. Lab Invest 78: renal function in men with type II diabetes mellitus. Kidney 813–824, 1998 Int 69: 336–342, 2006 11. Perianayagam MC, Murray SL, Balakrishnan VS, Guo D, 24. Keller CR, Odden MC, Fried LF, Newman AB, Angleman King AJ, Pereira BJ, Jaber BL: Serum soluble Fas (CD95) S, Green CA, Cummings SR, Harris TB, Shlipak MG: Kid- and Fas ligand profiles in chronic kidney failure. J Lab Clin ney function and markers of inflammation in elderly per- Med 136: 320–327, 2000 sons without chronic kidney disease: The health, aging, 12. Baba K, Minatoguchi S, Sano H, Kagawa T, Murata I, and body composition study. Kidney Int 71: 239–244, 2007 Takemura G, Hirano T, Ohashi H, Takemura M, Fujiwara 25. Knight EL, Rimm EB, Pai JK, Rexrode KM, Cannuscio CC, T, Fujiwara H: Involvement of apoptosis in patients with Manson JE, Stampfer MJ, Curhan GC: Kidney dysfunction, diabetic nephropathy: A study on plasma soluble Fas lev- inflammation, and coronary events: A prospective study. els and pathological findings. Nephrology (Carlton) 9: 94–99, J Am Soc Nephrol 15: 1897–1903, 2004 2004 26. Tonelli M, Sacks F, Pfeffer M, Jhangri GS, Curhan G: Bi- 13. Ortiz A, Lorz C, Egido J: The Fas ligand/Fas system in omarkers of inflammation and progression of chronic kid- renal injury. Nephrol Dial Transplant 14: 1831–1834, 1999 ney disease. Kidney Int 68: 237–245, 2005 14. Zoppini G, Faccini G, Muggeo M, Zenari L, Falezza G, 27. Aderka D, Sorkine P, Abu-Abid S, Lev D, Setton A, Cope AP, Targher G: Elevated plasma levels of soluble receptors of Wallach D, Klausner J: Shedding kinetics of soluble tumor TNF-alpha and their association with smoking and micro- necrosis factor (TNF) receptors after systemic TNF leaking vascular complications in young adults with type 1 diabe- during isolated limb perfusion. Relevance to the pathophys- tes. J Clin Endocrinol Metab 86: 3805–3808, 2001 iology of septic shock. J Clin Invest 101: 650–659, 1998 15. Macisaac RJ, Tsalamandris C, Thomas MC, Premaratne E, 28. Navarro JF, Mora C: Role of inflammation in diabetic com- Panagiotopoulos S, Smith TJ, Poon A, Jenkins MA, Rat- plications. Nephrol Dial Transplant 20: 2601–2604, 2005 naike SI, Power DA, Jerums G: The accuracy of cystatin C 29. Bell JH, Herrera AH, Li Y, Walcheck B: Role of ADAM17 in and commonly used creatinine-based methods for detect- the ectodomain shedding of TNF-alpha and its receptors by ing moderate and mild chronic kidney disease in diabetes. neutrophils and . J Leukoc Biol 82:173–176, 2007 Diabet Med 24: 443–448, 2007 30. Navarro JF, Mora C, Gomez M, Muros M, Lopez-Aguilar 16. Perkins BA, Nelson RG, Ostrander BE, Blouch KL, C, Garcia J: Influence of renal involvement on peripheral Krolewski AS, Myers BD, Warram JH: Detection of renal blood mononuclear cell expression behavior of tumor ne- function decline in patients with diabetes and normal or crosis factor-alpha and interleukin-6 in type 2 diabetic elevated GFR by serial measurements of serum cystatin C patients. Nephrol Dial Transplant 23: 919–926, 2008 concentration: Results of a 4-year follow-up study. JAm 31. Al Lamki RS, Wang J, Vandenabeele P, Bradley JA, Thiru S, Soc Nephrol 16: 1404–1412, 2005 Luo D, Min W, Pober JS, Bradley JR: TNFR1- and TNFR2- 17. Rosolowsky ET, Ficociello LH, Maselli NJ, Niewczas MA, mediated signaling pathways in human kidney are cell 70 Clinical Journal of the American Society of Nephrology Clin J Am Soc Nephrol 4: 62–70, 2009

type-specific and differentially contribute to renal injury. 35. Protopsaltis J, Kokkoris S, Nikolopoulos G, Spyropoulou P, FASEB J 19: 1637–1645, 2005 Katsaros T, Salvanos L, Brestas P, Korantzopoulos P, Me- 32. Bemelmans MH, Gouma DJ, Buurman WA: Tissue distri- lidonis A: Correlation between increased serum sFas levels bution and clearance of soluble murine TNF receptors in and microalbuminuria in type 1 diabetic patients. Med mice. Cytokine 6: 608–615, 1994 Princ Pract 16: 222–225, 2007 33. Brockhaus M, Bar-Khayim Y, Gurwicz S, Frensdorff A, 36. Elzey BD, Griffith TS, Herndon JM, Barreiro R, Tschopp J, Haran N: Plasma tumor necrosis factor soluble receptors in Ferguson TA: Regulation of Fas ligand-induced apoptosis chronic renal failure. Kidney Int 42: 663–667, 1992 by TNF. J Immunol 167: 3049–3056, 2001 34. Nelson CL, Karschimkus CS, Dragicevic G, Packham DK, 37. Boldin MP, Mett IL, Varfolomeev EE, Chumakov I, Wilson AM, O’Neal D, Becker GJ, Best JD, Jenkins AJ: Shemer-Avni Y, Camonis JH, Wallach D: Self-association Systemic and vascular inflammation is elevated in early of the “death domains” of the p55 tumor necrosis factor IgA and type 1 diabetic nephropathies and relates to vas- (TNF) receptor and Fas/APO1 prompts signaling for cular disease risk factors and renal function. Nephrol Dial TNF and Fas/APO1 effects. J Biol Chem 270: 387–391, Transplant 20: 2420–2426, 2005 1995