Sex differences in the association of skin advanced glycation endproducts with knee osteoarthritis progression

Osteoarthritis (OA) is characterized by the progressive destruction of articular cartilage and is strongly and positively associated with chronological age, but the mechanism by which aging contributes to this increased susceptibility is largely unknown. Recently the hypothesis has been promulgated that accumulation of advanced glycation endproducts (AGEs) that are associated with cumulative glycemic exposure, oxidative stress and aging might explain some or most of this association [1]. Alterations in the extracellular matrix and loss of resident cells to repair this matrix damage induced by oxidative stress and AGEs have been suggested as explanations for the progressive articular damage associated with aging. AGEs are the endproducts of spontaneous reactions of reducing sugars with proteins, or non-enzymatic glycation found in tissues throughout the human body. These sugar-protein products are stabilized in humans through a group of complicated biochemical reactions (Maillard reactions) that produce a range of pigmented, fluorescent and glucose-derived protein cross-links that are collectively known as AGEs. Most AGEs have not been characterized in human systems; therefore a few markers of the non-enzymatic glycation process have been used to study their effects on human biology. AGEs form at accelerated rates in many tissues in patients with diabetes mellitus and therefore have been largely studied to better understand this disease and its complications [2‐4]. AGEs have been associated with the several other age-related chronic diseases including Alzheimer’s dementia, and cardiovascular disease [5‐8]. Pentosidine, a fluorescent AGE formed by lysine and arginine residues, is a highly sensitive marker of AGE and is often used to characterize the whole family of AGEs in human tissues. Pentosidine has been shown to accumulate in the collagen matrix of skin, lens, and articular cartilage [1].

Of particular importance in OA, low turnover tissues such as articular cartilage appear to be particularly sensitive to the accumulation of AGEs regardless of glucose metabolic status [1, 7]. Several lines of evidence have been reported that suggest that the accumulation of AGEs in articular cartilage might be an important mechanism in the association between aging and OA. First, articular chondrocytes express the receptors of advanced glycation endproducts in humans [9]. Second, in dog models of experimentally induced OA, elevated levels of AGEs lead to more severe OA based upon histologic grading, increased collagen damage and enhanced release of proteoglycans [10]. Third, human articular cartilage incubated in AGEs demonstrates a dose-dependent increase in AGEs that is associated with increased stiffness of the collagen network and decreases in the tensile strength of the collagen network suggesting that accumulation of AGE crosslinks might be an important putative mechanism in the development or progression of OA [11]. Senoult et al. [12] found increased levels of pentosidine, an advanced glycation endproduct in the serum and synovial fluid of patients with knee OA compared to controls. Both the serum and synovial fluid pentosidine levels correlated with synovial fluid cartilage oligomeric matrix protein (COMP) but not with radiographic staging of the knee OA. Vos et al. found in cross-sectional analysis that skin pentosidine was associated with increased severity and progression of knee and hip OA [13, 14].

This evidence adds additional support to the concept that AGEs may act as novel biomarkers of the incidence and progression of knee OA. However, not all studies are supportive of this hypothesis. Hunter et al. did not find correlation between urinary pentosidine and cartilage loss in the Boston Osteoarthritis of the Knee Study [15]. In addition, Vos et al. identified an inverse relationship of cartilage pentosidine levels and microscopic, histological and biochemical cartilage damage in samples obtained in total knee replacement [16].

Advanced glycation endproduct serum assays are not widely available due to a complex mixture of trace compounds in tissue proteins and the sophisticated and expensive techniques, such as gas chromatography and mass spectrometry, which are needed to reliably and precisely measure these products. However, when pentosidine is exposed to light in the near-UV and blue spectrum, it emits characteristic wavelengths of light called autofluorescence. Levels of the pentosidine found in skin biopsy specimens are highly correlated with collagen-linked fluorescence of the skin. Recently several groups have developed non-invasive means to measure skin fluorescence and thus indirectly measure skin AGE levels [17‐19]. These skin fluorescence levels have been shown to be correlated with collagen-linked fluorescence of the skin and AGE levels in skin biopsy specimens in both human and animal models [20, 21].

We were therefore interested in exploring the association of skin fluorescence with the progression of knee OA by measuring changes in joint space narrowing (JSN) and changes in joint space width (JSW) over a 4-year period.

sAGE scores ranged from 18.7 to 69.5 in men and 17.8 to 59.7 in women. We evaluated 160 men (255 knees) and 287 women (478 knees). Sex-specific ranges for each tertile of sAGEs and the cross-sectional associations between sAGEs and baseline characteristics are given in Table 1. Only older age and hypertension were associated with increasing levels of sAGEs in both sexes. Past and current smoking was associated with increasing levels of sAGEs in men only. Of note in women, the lowest tertile of sAGEs was associated with the greatest abdominal circumference and the percentage of women with obesity, contrary to the direction that was expected.

Due to multiple exclusions and measurement of sAGEs at a single site, we were concerned about potential selection bias and issues of generalizability. We therefore compared our sample to those excluded and to the overall OAI sample (Additional file 1: Table S1). The analytic sample differed from those excluded at the site where sAGEs were measured in terms being a slightly younger age compared to the excluded sample, and we noted fewer changes in JSW over the 4 years of measurement in the analytic sample when compared to the excluded sample at the site of measurement of sAGEs. The participants at the site where sAGEs were measured differed from the eligible participants in the overall OAI sample in that a higher percentage were women, there were fewer African Americans, more past or present smokers, greater abdominal circumference, lower diastolic blood pressure, less hypertension, lower K/L grades and less change in JSW.

The percentage of men and women with worsening JSN are given by tertile of sAGEs in Table 2, comparing a crude model and fully adjusted models of age, BMI, K/L grade, rim distance and beam angle. In all models there was a monotonically increased relationship between worsening JSN and sAGEs in men but not in women, with 9%, 16.9% and 20.9% progression in men in the base models and 13.7%, 19.6% and 13.4% in women in the base models. Adjusting for age, BMI, and K/L grade individually in men had little effect on the estimates and with full adjustment the test for trend remained statistically significant (p = 0.03). For women the null results (p for trend = 0.93) persisted with adjustment for potential confounding (p for trend = 0.33).

Measured change in JSW at a fixed point in the medial tibia as a measure of progression of OA by tertile of sAGEs are given in Table 3, comparing crude, and fully adjusted models of age, BMI, and K/L grade. In the crude model in men there was a significant monotonically increased relationship between JSW by tertile of sAGEs (p for trend = 0.02). However in the full model (p for linear trend = 0.15) the relationship was no longer statistically significant. Similar to the JSN analysis, there was no relationship between tertiles of sAGEs and loss of JSW in women over serial yearly measurements.

We performed several sensitivity analyses, adding patients with diabetes mellitus to our study population, and only evaluating the white participants due to concerns that the skin autofluorescence would be affected by skin pigmentation. Including patients with diabetes mellitus did not materially affect our results, with a significant p value for trend for association between tertiles of sAGEs and JSN in men (p trend = 0.02) and no relationship in women (p = 0.27). Similarly, in comparing tertiles of sAGEs and changes in JSW there was a non-significant trend (p = 0.14) in men and no relationship in women (p = 0.84). Evaluating white participants only reduced the sample size, and while a similar result was found in terms of trends in men compared to women, they were no longer statistically significant (p = 0.08 and 0.39, respectively, for JSN and p = 0.26 and 0.94, respectively, for changes in JSW.

In this study of progression of knee OA, using semi-quantitative measurements (of JSN) we found that higher levels of skin AGEs were associated with progression of medial-tibial knee OA in men but not in women, and identified a trend using serial measurements of loss of JSW. These results provide preliminary evidence supporting the hypothesis that AGEs may play an important role in the progression of knee OA, particularly in men.

In the CHECK cohort, Vos et al. [13, 14] demonstrated an association between skin pentosidine and more severe knee OA at baseline, but only a modest association with progression of a combined score of hip and knee OA. Using skin biopsies and measurements of skin pentosidine levels, they did show a weak association (r = 0.167, p = 0.02) between skin pentosidine levels and K/L score for progression in the hip and knee. This relationship persisted when adjusted for age and BMI. They did not present any sex-specific findings or evaluate a sex-based interaction effect. Using measurements of urinary pentosidine, Hunter et al. did not find any association with cartilage loss in the knee [15]. However, the ability of urinary pentosidine in the setting of normal renal function to reflect the effect of AGEs on cartilage is unknown.

In our study, we identified progression of OA with increased skin AGE levels in men and not in women, suggesting sex differences in the effect of skin AGEs on the progression of OA. This finding could be spurious or could suggest that there are unique sex-specific mechanistic pathways that may lie behind such differences. Sex differences in the prevalence, incidence and severity of radiographic and clinical OA are well-described [29, 30], but specific examination of sex differences in MRI biomarkers or in biomarkers of joint metabolism are few, and results vary considerably [31, 32].

Men compared women have less prevalent and incident knee OA and less severe OA after the age of 55 years [29, 30]. MRI suggests that women have a smaller volume of lower knee cartilage compared to men [31]. Women have higher levels of bone resorption biomarkers [32] whereas men have higher levels of serum hyaluronan (HA), an indicator of synovial inflammation [33]. Karvonen-Guteirrez, Sowers and Heeringa [34] reported relationships between cardiometabolic risk factors and knee OA that were unique to each gender and to obesity status. In non-obese men, BMI and homeostasis model assessment-insulin resistance (HOMA-IR) and leptin, were the risk factors related to knee OA. In obese men, BMI was not related to knee OA, but HOMA-IR was strongly related to it. In non-obese and obese women, BMI continued to be a strong predictor of knee OA; however, HOMA-IR was inversely related to knee OA in obese women. In non-obese and obese men, leptin was found to be inversely related to the presence of radiographic OA, whereas leptin was positively associated with radiographic knee OA in obese women. These findings suggest that the effects of metabolic factors on knee OA are different in men and women, similar to our findings related to sAGEs. Indeed, insulin resistance, elevated insulin-like growth factor (IGF)-1, and AGEs are all related to body fat distribution and adipocyte biology, which might partially explain the gender difference found in our study [35‐39].

While biomechanical factors are clearly involved in the development and progression of knee osteoarthritis, systemic pathways including innate immunity, inflammation and advanced glycation endproducts are increasingly being recognized as playing a potentially important role in the natural history of osteoarthritis. This study adds credence to this view of the pathophysiology of knee osteoarthritis and suggests that future studies should be performed to examine the advanced glycation endproduct hypothesis.