Systemic Soluble TNF Receptors 1 and 2 are Associated with Severity of Diabetic Retinopathy in Hispanics

Abstract

Purpose

To investigate the associations of serum amyloid A protein (SAA) and soluble tumor necrosis factor receptors 1 and 2 (sTNF-R1 and sTNF-R2) with diabetic retinopathy (DR) in Hispanics.

Design

Prospective, nonrandomized, cross-sectional, family-based observational cohort study.

Participants

Four hundred seventy-three Hispanics type II diabetic subjects, in families ascertained via proband with DR.

Methods

Levels of SAA, sTNF-R1, and sTNF-R2 were measured with enzyme-linked immunosorbent assay (ELISA). DR was assessed by fundus photography and graded using modified Airlie House classification.

Main Outcome Measures

Levels of SAA, sTNF-R1, and sTNF-R2 to severity of DR with and without covariates.

Results

A direct association of sTNF-R1 (2.37±0.13, 2.15±0.09, 3.09±0.24, 3.25±0.46, 5.02±0.61 ng/ml; p<0.0001) and sTNF-R2 (6.04±0.20, 6.25±0.52, 7.96±0.70, 8.14±1.13, 14.83±1.68 ng/ml; p<0.0001) was found for no DR, mild non-proliferative DR (NPDR), moderate NPDR, severe NPDR, and proliferative DR (PDR), respectively. These associations remained significant after adjusting for age, gender, body mass index, hemoglobin A1c, diabetes duration, systolic blood pressure, and serum creatinine (p<0.0001 for sTNF-R1 and p=0.0004 for sTNF-R2). A similar pattern was observed when we adjusted for urinary albumin-to-creatinine ratio in place of serum creatinine (p=0.005 for sTNF-R1 and p=0.02 for sTNF-R2).

Conclusions

Levels of sTNF-R1 and sTNF-R2 are highly correlated with severity of DR, suggesting that inflammation and insulin resistance may play a critical role in the development of DR. These may be useful biomarkers for DR, aiding in etiologic studies, and possibly in identifying at-risk patients for active intervention.

Diabetic retinopathy (DR), a frequent microvascular complication of diabetes, is the leading cause of preventable blindness in working age adults.¹ Hispanics, the largest and fastest growing ethnic minority in the United States, have a high prevalence of type II diabetes and acquire the disease at a younger age. They also have a higher risk for developing DR than Caucasians.^2–5 This is not explained by traditional risk factors, e.g., glycemic and blood pressure control.⁵ Inflammation, a nontraditional risk factor, has been suggested as being involved in the pathogenesis of DR and other microvascular and macrovascular complications.^{6, 7} However, the role that inflammatory processes play in DR has not been elaborated extensively and current evidence remains unclear and inconsistent.^{8, 9} In this study, we measured the plasma levels of three inflammatory biomarkers, serum amyloid A protein (SAA) and soluble tumor necrosis factor receptors 1 (sTNF-R1) and 2 (sTNF-R2), in Hispanic type II diabetic subjects with and without DR. We chose to study these particular biomarkers because they have been linked to key etiologic components of inflammation and insulin resistance in diabetes and obesity. SAA, an acute-phase protein, is produced by the liver and adipose tissue during an acute inflammatory process. It has been shown that SAA directly mediates obesity-associated inflammation and is a sensitive marker for obesity-associated diseases, such as cardiovascular disease and diabetes.¹⁰ To the best of our knowledge, the role of SAA with DR in Hispanics has not been examined previously.

Another key inflammatory marker, tumor necrosis factor-α (TNF) and its receptors could also have an important role in DR. Previous studies have shown that the level of soluble TNF receptors increases in the serum and vitreous of proliferative diabetic retinopathy (PDR) patients with type I diabetes.^{11, 12} The association of soluble TNF receptors with DR in type II diabetics and in Hispanics with diabetes has not been examined previously. The purpose of this study was to evaluate the association of SAA, sTNF-R1, and sTNF-R2 with severity of DR in a large group of family-based Hispanic type II diabetics.

METHODS

From October 2007 to December 2010, a cross-sectional study of 473 Hispanic type II diabetic subjects were recruited in a joint study of Cedars-Sinai Medical Center (CSMC) and the Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center (HUMC), following informed consent and approval by the Institutional Review Board (IRB)/Ethics Committee. This study assessed siblings of a proband (with type II diabetes and either known DR or a diabetes duration of ≥ 10 years) and when possible, one or both of their parents. Information on subjects’ demographics, medical history, family history and current medications were collected along with measured anthropometric indices (height, weight, waist circumference, bioimpedance for estimation of body fat), vital signs (blood pressure and pulse), and blood and urine samples (fasting serum glucose, glycosylated hemoglobin (HbA1c), serum lipids, serum creatinine, and urinary albumin-to-creatinine ratio). Additionally, siblings and parents not known to have diabetes at the time of the study were given an oral glucose tolerance test and the diagnosis of diabetes was based on standard criteria of the American Diabetes Association.¹³

Biochemical assays

Standard clinical and urinary assays were performed in the general clinical laboratory of HUMC. Investigative biochemical assays were performed at the biochemistry laboratory at CSMC. Specifically, we determined the plasma levels of SAA (ELISA, Abazyme, LLC) and sTNF-R1 and sTNF-R2 (EASIA, Biosource) using commercially available enzyme-linked immunosorbent assays (ELISA).

Ocular phenotypes

All subjects underwent dilated ophthalmic eye examination with 30° digital stereoscopic color fundus photography of the Early Treatment Diabetic Retinopathy Study (ETDRS) 7 standard fields at HUMC and retinal images were assessed by masked photo graders at the Ocular Epidemiology Reading Center (OERC) in the University of Wisconsin, Madison, using modified Airlie House classification scheme and the ETDRS severity scale.¹⁴

Statistical analysis

For quantitative analysis, retinopathy level from the ETDRS diabetic severity scale was utilized. For qualitative analysis, retinopathy level was grouped into five levels of increasing severity: none (levels 10–13), mild non-proliferative diabetic retinopathy (NPDR) (levels 14–20), moderate NPDR (levels 31–43), severe NPDR (levels 47–53), and PDR (levels 60–85). The eye with the more severe retinopathy was used for assigning retinopathy level for each participant. To account for the dependency among family members, the generalized estimating equations methods (GEE1) as implemented in the GENMOD procedure of SAS 9.1 (SAS Institute Inc., Cary, North Carolina, USA) was utilized in the association analysis by using family as the cluster factor, i.e., members from the same family are assumed to be correlated and those from different families are assumed to be independent. A normal distribution was assumed for quantitative traits, while logistic regression with multilevel link function was used for qualitative trait analysis. All statistical analyses were performed with JMP and GENMOD in SAS.

RESULTS

A total of 473 type II diabetic subjects with DR (cases, n=266) and without DR (controls, n=207) were included in the analysis (demographics of the cohort summarized in Table 1). The mean ± standard error of the mean (SEM) age was 53.7 ± 0.8 years in the controls and 53.6 ± 0.7 years in the cases. There were 66 (31.9%) and 114 (42.9%) male subjects in the controls and cases, respectively. The mean body mass index (BMI) ± SEM were 33.4 ± 0.5 and 32.1 ± 0.4 in the controls and cases, respectively, with the controls having slightly higher BMIs.

Table 1.

Demographics and laboratory results of cohort.

		CONTROLS	CASES	p-value*
		Diabetic + No Retinopathy (n=207)	Diabetic + Retinopathy (n=266)
Demographics	Age – yrs	53.7 ± 0.8	53.6 ± 0.7	NS
	Gender (% male)	31.9	42.9	0.01
	BMI (kg/m2)	33.4 ± 0.5	32.1 ± 0.4	0.01
	HbA1c (%)	7.9 ± 0.1	8.9 ± 0.1	p<0.0001
	DM duration (yrs)	7.2 ± 0.5	13.9 ± 0.5	p<0.0001
Blood Pressure	SBP (mmHg)	127 ± 1.3	133 ± 1.1	P=0.0007
	DBP (mmHg)	70 ± 0.7	71 ± 0.6	NS
Lipid Profiles	Cholesterol (mg/dL)	183 ± 3.2	180 ± 2.8	NS
	TG (mg/dL)	179 ± 9.3	183 ± 8.2	NS
	HDL (mg/dL)	40.7 ± 0.8	40.9 ± 0.7	NS
	LDL (mg/dL)	106 ± 2.5	104 ± 2.2	NS
Renal function	Serum creatinine (mg/dL)	0.7 ± 0.06	1.1 ± 0.05	p<0.0001
	Urinary albumin/creatinine ratio	35 ± 28.7	235 ± 26.1	p<0.0001
Inflammatory Markers	sTNF-R1 (ng/mL)	2.43 ± 0.2	3.28 ± 0.2	p<0.0001
	sTNF-R2 (ng/mL)	6.17 ± 0.5	9.14 ± 0.4	p<0.0001
	sTNFR2/sTNFR1 ratio	2.65 ± 0.05	2.59 ± 0.04	NS
	SAA (ug/mL)	7.1 ± 0.6	6.1 ± 0.5	NS

^*=Adjusted for family structure; BMI = body mass index; DBP = diastolic blood pressure; DR = diabetic retinopathy; HbA1c = hemoglobin A1c; HDL=High-density lipoprotein; LDL=Low-density lipoprotein; n = number; NS = not significant; yrs = years; SAA=serum amyloid A; SBP = systolic blood pressure; TG=triglyceride; sTNF-R1=soluble tumor necrosis factor (TNF) receptor-1; sTNF-R2=soluble TNF receptor-2. Value depicted as mean ± standard error of the mean.

Two well-characterized risk factors for DR, HbA1c and diabetes duration, were also investigated and found to be statistically different between these two groups (p<0.0001 for both traits), with worse glycemic control and longer duration of diabetes observed in the cases. We found differences in gender, BMI, HbA1c, duration of diabetes, and systolic blood pressure to be statistically significant between the two groups (Table 1).

We first examined a series of serum and plasma variables in controls versus cases. Serum lipid levels were similar between the two groups with no significant difference observed for total cholesterol, triglyceride, high-density lipoprotein (HDL), and low-density lipoprotein (LDL). In the cases, both serum creatinine level and urinary albuminto- creatinine ratio were significantly higher than the controls (p<0.0001). Of particular interest, sTNF-R1 (p<0.0001) and sTNF-R2 (p<0.0001) were elevated in those with DR compared to those without DR, whereas SAA was not different (Table 1).

We then subdivided our data into various levels of DR severity for further characterization of the relationship of the TNF receptor levels to DR in qualitative and quantitative analyses. Diabetes without DR was found in 207 (43.8%) subjects. DR was present in 266 (56.2%) subjects, of whom 74 subjects had mild NPDR, 105 had moderate NPDR, 22 had severe NPDR, and 62 had PDR. Three subjects had un-gradable fundus photos and were not included in the analysis. The mean values of sTNF-R1 ± SEM were: no DR 2.37 ± 0.13, mild NPDR 2.15 ± 0.09, moderate NPDR 3.09 ± 0.24, severe NPDR 3.25 ± 0.46, PDR 5.02 ± 0.61 ng/ml; and of sTNF-R2 ± SEM were: no DR 6.04 ± 0.20, mild NPDR 6.25 ± 0.52, moderate NPDR 7.96 ± 0.70, severe NPDR 8.14 ± 1.13, and PDR 14.83 ± 1.68 ng/ml, respectively (Table 2). Stratification by severity of DR indicates that sTNF-R1 and sTNF-R2 exhibit an increasing trend with severity of DR estimated both qualitatively (Figure 1) and quantitatively (Figure 2).

Table 2.

Analysis by degree of diabetic retinopathy with different models of covariate adjustments.

		No DR	Mild NPDR	Moderate NPDR	Severe NPDR	PDR	Models & Significance
							1	2	3	4	5
Inflammatory Markers	sTNF-R1 (ng/mL)	2.37±0.13	2.15±0.09	3.09±0.24	3.25±0.46	5.02±0.61	1.84×10−13	1.85×10−13	8.05×10−11	3.77×10−6	0.005
	sTNF-R2 (ng/mL)	6.04±0.20	6.25±0.52	7.96±0.70	8.14±1.13	14.83±1.68	5.83×10−14	6.22×10−15	3.07×10−11	0.0004	0.02

Mean Values and Significance of Inflammatory Markers by Severity of Diabetic Retinopathy Adjusted for Family Structure. Value depicted as mean ± standard error of the mean.

Models & Significance: 1=Unadjusted; 2=Adjusted for age, gender and body mass index (BMI); 3=Adjusted for age, gender, BMI, Hemoglobin A1c (HbA1c), and diabetes duration; 4=Adjusted for age, gender, BMI, HbA1c, diabetes duration, systolic blood pressure, and serum creatinine; 5=Adjusted for age, gender, BMI, HbA1c, diabetes duration, systolic blood pressure, and urinary albumin-to-creatinine ratio.

DR = diabetic retinopathy; NPDR=non-proliferative diabetic retinopathy; SEM=standard error of the mean; PDR=proliferative diabetic retinopathy; TNF-R1=soluble TNF receptor-1; sTNF-R2=soluble TNF receptor-2.

Figure 1.

Graph depicting changes in the levels of inflammatory markers by severity of diabetic retinopathy in a qualitative analysis. Box plots depicting soluble TNF receptor concentrations by severity of diabetic retinopathy as a function of qualitatively defined degree of retinopathy (distribution of mean values in asterisks).

Figure 2.

Graph depicting changes in the levels of inflammatory markers by severity of diabetic retinopathy in a quantitative analysis. Graph indicating the values of soluble TNF receptor levels by severity of diabetic retinopathy using Early Treatment Diabetic Retinopathy Study (ETDRS) scale. p-value for trend, sTNF-R1 p<0.0001, sTNF-R2 p<0.0001, adjusted for family structure.

We then investigated whether the association was independent by analyzing the data with various adjustments for potential covariates. The significance of this trend remained after adjusting covariates in all the different models. As serum creatinine and urinary albumin-to-creatinine ratio are significantly correlated (r²=0.12, p<0.0001), we analyzed our data adjusting for each in separate models. An association of sTNF-R1 and sTNF-R2 with severity of DR remained after adjustments in the following 5 models: 1) unadjusted (p<0.0001 for sTNF-R1 and sTNF-R2), 2) age, gender, and BMI (p<0.0001 for sTNF-R1 and sTNF-R2), 3) age, gender, BMI, HbA1c, and diabetes duration (p<0.0001 for sTNF-R1 and sTNF-R2), 4) age, gender, BMI, HbA1c, diabetes duration, systolic blood pressure, and serum creatinine (p<0.0001 for sTNF-R1 and p=0.0004 for sTNF-R2) and 5) age, gender, BMI, HbA1c, diabetes duration, systolic blood pressure, and urinary albumin-to-creatinine ratio (p=0.005 for sTNF-R1 and p=0.02 for sTNF-R2) (Table 2).

DISCUSSION

After controlling for confounding risk factors, including age, gender, BMI, HbA1c, diabetes duration, systolic blood pressure, and serum creatinine or urinary albumin-to-creatinine ratio, both sTNF-R1 and sTNF-R2 were cross-sectionally associated with increasing severity of DR in Hispanic type II diabetic subjects. These findings reveal novel systemic biomarkers, namely soluble TNF receptors, which may potentially serve as a biochemical link between two important pathophysiological processes involved — inflammation and insulin resistance. Previous studies have examined a number of systemic biomarkers of inflammation,^{8, 9, 11, 12, 15} endothelial dysfunction,⁸ hemostatic disturbances,⁹ angiogenesis,¹⁵ and homocysteinemia^{8, 9} with DR in both type I and type II diabetes, but reports for these results have been inconclusive. It has been suggested that the lack of association between DR and systemic biomarkers of inflammation, endothelial dysfunction, and hemostatic disturbances may be due to the fact that the eye is an immune privileged area, and therefore inflammatory makers do not cross the blood-retinal barrier.^{7, 9}

The previously examined inflammatory biomarkers include C-reactive protein (CRP),^{8, 9} TNF-α,^{8, 12} cytokine interleukin (IL)-6,^{8, 9} T-cell expressed and secreted RANTES/CCL5,¹⁵ stromal cell-derived factor (SDF)-1α/CXCL12,¹⁵ epithelial neutrophil activator (ENA)-78/CSCL5,¹⁵ interferon-induced protein (IP)-10/CXCL10,¹⁵ monocyte chemoattractant protein (MCP)-1/CCL2,¹⁵ macrophage inflammatory protein (MIP)- 1α/CCL3,¹⁵ and IL-8/CXCL8.¹⁵ Levels of RANTES and SDF-1α were reported to be elevated in patients with PDR.¹⁵ The latter study combined both type I and type II persons with diabetes, and combined both blacks and whites in the analysis, which may be problematic if the baseline levels of these markers varied between type I and type II diabetes and/or between different ethnicities. In the Wisconsin Epidemiologic Study of Diabetic Retinopathy (WESDR), the authors found that the inflammatory marker TNF-α was associated with the severity of DR in type I diabetic participants with kidney disease.⁸

Endothelial dysfunction, characterized by increased level of soluble vascular cell adhesion molecule (sVCAM-1) and soluble intercellular adhesion molecule (sICAM-1), has been reported to increase vascular permeability, alter blood flow, increase oxidative stress, and affect angiogenesis.^{8, 16} Only sVCAM-1 has been associated with the prevalence and severity of DR in type I diabetics with kidney disease,⁸ although in an earlier study, a positive association was not seen with either adhesion molecules.¹⁵ The Multi-Ethnic Study of Atherosclerosis (MESA) study examined markers for inflammation, homocysteine and hemostatic disturbances, namely fibrinogen, plasmin-α₂-antiplasmin complex (PAP), D-dimer, and factor VIII in type II diabetics from four different race/ethnicities (whites, blacks, Hispanics, and Chinese) and found that only PAP was associated with any DR after adjusting for renal dysfunction.⁹

Our findings support two earlier studies, one examining the level of SAA with severity of DR in Japanese with type II diabetes¹⁷ and another examining the serum levels of TNF-receptors in type I diabetes in Caucasian subjects with active PDR, quiescent PDR, no DR, and healthy controls.¹² The authors from the first study demonstrated no association of SAA with severity of DR, but instead, suggested an association with diabetic nephropathy. This observation indicates the mechanism by which SAA acts may be due to impaired metabolism in the liver and kidney and thus a different mechanism from that which may affect DR in the eyes. In the second study, the authors found that type I diabetic patients complicated by PDR exhibited significantly higher serum levels of sTNF-R1 and sTNF-R2 than those without retinopathy or healthy individuals. It is possible that marked elevation of sTNF-R1 and sTNF-R2 may constitute important clinical and pathophysiological significance in that soluble TNF receptors may play a role in modulating the biological activity of TNF-α^{12, 18} and interfere with the immunological detection of TNF-α.^{12, 19} Soluble TNF receptors may be a more sensitive marker than TNF-α in determining inflammatory and insulin resistance status in patients with DR.

In recent years, the molecular mechanisms of TNF function have been intensively investigated. One of the earliest studies linking inflammation to insulin resistance in type II diabetes involved animal models of obesity and diabetes.²⁰ Increased expression of TNF-α mRNA and protein was observed and these elevated levels were found both locally within the adipose tissue and systemically in the plasma of these animals. Neutralization of TNF-α using a recombinant TNF-α receptor-immunoglobulin G chimeric protein showed improvement in insulin sensitivity, implicating a direct role of TNF-α in the development of insulin resistance. Subsequent studies have suggested that in addition to TNF-α, increased circulating levels of TNF receptors are also associated with diabetes, obesity, and insulin resistance.^{21, 22} Our study demonstrates that soluble TNF receptors are also correlated with DR in type II diabetic patients and supports the idea that insulin resistance/impaired insulin action could have a critical role in the development of DR.

The strengths of this study include a large cohort of family-based Hispanic type II diabetic subjects and the assessment of DR by standardized gradable digital fundus photographs. However, several limitations should also be noted. First, levels of these biomarkers may vary among different race/ethnicities. We are uncertain if the observed effect is specific to Hispanics or if this is a general effect. Second, as this is a crosssectional study, our ability to evaluate a temporal sequence is limited.

In summary, data from this study of Hispanic type II diabetics indicate that soluble TNF receptors are cross-sectionally associated with DR after adjusting for potential confounders. It is possible that the biological activity of soluble TNF receptors may correlate with clinical disease severity and that this biomarker may provide a link between inflammation and insulin resistance in DR. This finding needs to be replicated in prospective studies.