articles
nature publishing group
Intervention and Prevention
Pigment Epithelium-Derived Factor, Insulin Sensitivity, and Adiposity in Polycystic Ovary Syndrome: Impact of Exercise Training Anju E. Joham1,2, Helena J. Teede1,2, Samantha K. Hutchison1,2, Nigel K. Stepto3, Cheryce L. Harrison1, Boyd J. Strauss4, Eldho Paul1 and Matthew J. Watt5 Pigment epithelium-derived factor (PEDF) is upregulated in obese rodents and is involved in the development of insulin resistance (IR). We aim to explore the relationships between PEDF, adiposity, insulin sensitivity, and cardiovascular risk factors in obese women with polycystic ovary syndrome (PCOS) and weight-matched controls and to examine the impact of endurance exercise training on PEDF. This prospective cohort intervention study was based at a tertiary medical center. Twenty obese PCOS women and 14 non-PCOS weight-matched women were studied at baseline. PEDF, cardiometabolic markers, detailed body composition, and euglycemic–hyperinsulinemic clamps were performed and measures were repeated in 10 PCOS and 8 non-PCOS women following 12 weeks of intensified aerobic exercise. Mean glucose infusion rate (GIR) was 31.7% lower (P = 0.02) in PCOS compared to controls (175.6 ± 96.3 and 257.2 ± 64.3 mg.m−2.min−1) at baseline, yet both PEDF and BMI were similar between groups. PEDF negatively correlated to GIR (r = −0.41, P = 0.03) and high-density lipoprotein (HDL) (r = −0.46, P = 0.01), and positively to cardiovascular risk factors, systolic (r = 0.41, P = 0.02) and diastolic blood pressure (r = 0.47, P = 0.01) and triglycerides (r = 0.49, P = 0.004). The correlation with GIR was not significant after adjusting for fat mass (P = 0.07). Exercise training maintained BMI and increased GIR in both groups; however, plasma PEDF was unchanged. In summary, PEDF is not elevated in PCOS, is not associated with IR when adjusted for fat mass, and is not reduced by endurance exercise training despite improved insulin sensitivity. PEDF was associated with cardiovascular risk factors, suggesting PEDF may be a marker of cardiovascular risk status. Obesity (2012) 20, 2390–2396. doi:10.1038/oby.2012.135
Introduction
Polycystic ovary syndrome (PCOS) affects 6–18% of women of reproductive age (1), depending on diagnostic criteria and populations studied (2). PCOS is diagnosed based on oligo or amenorrhea, clinical or biochemical hyperandrogenism, and polycystic ovaries on ultrasound (3). Insulin resistance (IR) plays a central pathophysiological role in the majority of women with PCOS, independent of body weight; however, IR in PCOS is further exacerbated by the high prevalence of associated obesity (4). IR underpins significant metabolic complications in PCOS including dyslipidemia, impaired glucose tolerance, and type 2 diabetes (5–7), yet the aetiology of IR in PCOS remains unclear. Pigment epithelium-derived factor (PEDF) is a glycoprotein that belongs to the superfamily of serine protease inhibitors. PEDF promotes neuronal differentiation and survival (8) and is a potent inhibitor of angiogenesis (9) and endothelial
cell injury in vitro, suggesting a role in atherosclerosis (10). Our group recently reported that plasma PEDF is elevated in rodent models of obesity and reduced upon weight loss and insulin sensitization (11). We further showed that PEDF administration in lean mice stimulated adipose tissue lipolysis and caused IR, whereas neutralizing PEDF’s actions in obese mice improved insulin sensitivity (11). Evidence for a causal role of PEDF in the development of human IR is limited. PEDF correlated with visceral adiposity in a Japanese cohort (12) and PEDF was upregulated in insulin-resistant individuals with metabolic syndrome (10,13), impaired glucose tolerance, and type 2 diabetes (14,15). However, a major limitation of these studies is determining whether PEDF primarily relates to obesity-induced IR or whether it is related to IR independent of obesity (12,16). A recent study in a Chinese cohort demonstrated correlation of PEDF to insulin sensitivity measured during clamp studies, independent of BMI in both lean and
1 Women’s Public Health Research, School of Public Health and Preventive Medicine, Monash University, Melbourne, Australia; 2Diabetes Unit, Southern Health, Melbourne, Australia; 3Institute of Sport, Exercise and Active Living, Victoria University, Melbourne, Australia; 4Department of Medicine, Southern Clinical School, Monash University, Melbourne, Australia; 5Department of Physiology, Biology of Lipid Metabolism Laboratory, Monash University, Melbourne, Australia. Correspondence: Helena J. Teede (
[email protected])
Received 20 October 2011; accepted 4 May 2012; advance online publication 21 June 2012. doi:10.1038/oby.2012.135
2390
VOLUME 20 NUMBER 12 | december 2012 | www.obesityjournal.org
articles Intervention and Prevention overweight PCOS and non-PCOS women; however, interpretation of this data was limited as no clamps were performed on the overweight non-PCOS women (17). The etiology of IR in PCOS is poorly understood and existing research into this area is limited to humans, with no ideal animal models. Current etiological theories suggest that intrinsic IR (purportedly genetic, inherent, and unique to PCOS) (18), is exacerbated by extrinsic IR (lifestyle/obesity related) (4,19,20). As PEDF appears to have a causal role in obesity-related IR in animal studies (11), we hypothesized that PEDF may be involved in the etiology of extrinsic lifestyle-related IR in PCOS and as such may be improved with exercise intervention. IR is central to the metabolic and reproductive disturbances in PCOS and thus lifestyle modification including exercise is a first-line therapy in PCOS management. There are limited studies examining the effects of exercise in PCOS. Previous studies assessing the impact of exercise therapy in PCOS report improved IR following exercise using indirect measures of IR (21). A recent systematic review on exercise in PCOS demonstrates clear gaps in knowledge in this area (21). Recent findings by our group demonstrate that intensified exercise enhances insulin sensitivity in women with and without PCOS independent of weight loss (22); however, it is unclear whether PEDF is affected by improved IR following exercise in the context of PCOS. Therefore, the primary aim of this study was to assess the relationship between PEDF, adiposity, and IR in two groups of obese women: a non-PCOS control group and an insulin-resistant PCOS group. In addition, we aimed to assess the relationship between PEDF and known cardiometabolic risk factors. We also assessed the impact of exercise on IR and PEDF. Methods and Procedures Subjects Women were recruited from community advertisements for a larger study (23). Thirty-four women were eligible (n = 20 PCOS and n = 14 non-PCOS controls) and had completed 3 months run-in and baseline data collection as previously described (23). In this PEDF substudy, we analyzed baseline data from 32 women (n = 18 with PCOS and n = 14 without PCOS) and post-exercise data from 18 women (n = 10 with PCOS and n = 8 without PCOS) based on sample availability for PEDF analysis. As previously described, PCOS was diagnosed (1990 National Institutes of Health (NIH) criteria) by a qualified endocrinologist (S.K.H.) based on perimenarchal onset of irregular cycles (>35 days), combined with clinical (hirsutism, acne) or biochemical (elevation of at least one circulating ovarian androgen) hyperandrogenism (3,24). Hyperprolactinemia, thyroid dysfunction, and specific adrenal disorders were excluded clinically and where indicated, biochemically. All women without PCOS had regular menses and no evidence of clinical or biochemical hyperandrogenism. Exclusion criteria were smoking, diabetes, participation in regular physical activity, recent weight change, and pregnancy. The Southern Health Research Advisory and Ethics Committee approved the study and all participants gave written informed consent. The clinical trial registration number is ISRCTN84763265. Study design At screening (3 months before baseline), standard diet and lifestyle advice were delivered (Heart Foundation of Australia recommendations (http://www.heartfoundation.org.au)) as previously described obesity | VOLUME 20 NUMBER 12 | december 2012
(22,23). Medications affecting end-points including insulin sensitizers, antiandrogens, and hormonal contraceptives were ceased for 3 months. End-point data was collected at 3 months (baseline) and again following 12 weeks of exercise intervention. Data was collected in the follicular phase of the menstrual cycle wherever feasible. A subset of PCOS (n = 10) and control (n = 8) women completed a 12-week intensified aerobic exercise program as described previously (22,23). Anthropometric measurements Participants were weighed lightly clothed without shoes (TBF310; Tanita, Tokyo, Japan). BMI was calculated, (weight (kilogram)/height squared (metre2)), (Stadiometer; Holtain, Wales, UK). Waist circumference was measured at the umbilicus by an experienced operator. Euglycemic–hyperinsulinemic clamp Insulin sensitivity was measured using the euglycemic–hyperinsulinemic clamp technique as previously described (22,23,25). Fasting venous blood samples were collected, centrifuged, and stored for assessment of glucose, insulin, PEDF, testosterone, sex hormone-binding globulin, total cholesterol, high-density lipoprotein (HDL), low-density lipoprotein, and triglycerides as previously described (22,23). Testosterone was measured on Beckman Coulter Unicel DXI 800 analyzer (Beckman Coulter Diagnostics Australia, Gladesville, Australia) using an automated competitive binding immunoenzymatic assay and the free androgen index was calculated from free androgen index = (testosterone/sex hormone-binding globulin) × 100 as previously reported (23). The glucose infusion rates (GIRs) were calculated during the last 30 min of the euglycemic–hyperinsulinemic clamp and expressed as glucose (ml) per hour per body surface area (m2) per minute. Body composition and adipose tissue distribution Fat mass, abdominal fat mass, and fat-free mass were measured by dual-energy X-ray absorptiometry as described elsewhere (23). Singleslice computed tomography images were acquired at the level of L4–L5 intervertebral disc space and abdominal visceral fat (AVF) and abdominal subcutaneous fat cross-sectional areas (centimeters squared) were calculated as previously described (23). Biopsies were performed on all participants to obtain thigh adipose tissue samples for analysis. Exercise intervention Participants completed a 12-week intensified aerobic exercise program on a motorized treadmill (Biodex 500/Life Fitness 95T; Life Fitness, Lawrence, NY) as described (22). Participants attended three, 1 h sessions each week which sequentially alternated between moderate intensity (walking or jogging at 70% of VO2 max or 75–85% HRmax) and high-intensity interval training (6 × 5-min intervals with 2-min recovery period at ~95–100% of VO2 max or ~95–100% HRmax). Participants progressed to eight repetitions in the high-intensity training sessions by week 4, and reduced recovery time to 1 min by week 8 of training. Target exercise intensity (percentage VO2 max) and heart rates for each participant were achieved by altering speed (kph) and workload (gradient; %) on the treadmill with individual increases in fitness. Plasma PEDF Plasma PEDF was determined using a commercial ELISA according to the manufacturers’ procedure (BioVender Human PEDF ELISA; Evropska, Modrice, Czech Republic) using a single assay. The within assay coefficient of variation was 8.1 ± 1.5% (n = 8 samples in duplicate, values are mean ± s.e.m.). Quantitative reverse transcription-PCR RNA was extracted from thigh adipose tissue using the RNeasy Lipid Tissue Mini Kit (Qiagen, Valencia, CA) and 1 µg mRNA was reverse transcribed (iScript cDNA Synthesis Kit; Bio-Rad Laboratories, Hercules, CA). Quantitative real-time PCR was performed on a realplex 2391
articles Intervention and Prevention Mastercycler (Eppendorf, North Ryde, Australia) using the TaqMan Universal PCR Master Mix and TaqMan Gene Expression Assays (Applied Biosystems, Foster City, CA). The relative quantification was calculated using the ΔΔCt method (18S as the housekeeping gene) and results are normalized to values of the control group. Statistics All data were analyzed using STATA software version 11.0 (Stata, College Station, TX). PEDF levels were assessed for normality and found to be well-approximated by a normal distribution. A comparison between obese PCOS and non-PCOS subjects was conducted using Student’s t-test for normally distributed variables and Mann–Whitney U-test for variables that are not normally distributed. Linear regression was used for assessment of factors associated with plasma PEDF levels. Age adjusted estimates were also derived due to a statistically significant difference in age between the two groups. Results from the regression analysis are reported as parameter estimates (s.e.), with an R2 statistic to indicate the amount of variation explained. Correlation of PEDF with lipids and cardiovascular risk factors was assessed using Pearson or Spearman rank correlation, wherever appropriate. The effect of exercise within groups was assessed using paired t-tests. Continuous data are presented as mean ± s.d. or median (interquartile range) as appropriate. Statistical significance was set at α-level of 0.05. Results
Results are presented for 32 subjects with samples available for further analysis including PEDF at baseline. Eighteen women with PCOS and 14 women without PCOS who were eligible
at screening, completed the 3-month run-in phase with a steady diet, undertook no additional regular physical activity, were compliant with withdrawal of relevant medications and completed baseline data collection. Ten PCOS and eight nonPCOS women completed the 12-week exercise intervention and completed baseline data collection. PCOS vs. non-PCOS women: clinical, metabolic, and hormonal status and PEDF levels
All baseline metabolic and clinical characteristics are presented in Table 1. The women with PCOS were younger than controls (29.8 ± 5.7 vs. 35.0 ± 4.3 years, P = 0.01). The experimental groups had similar adiposity as determined by BMI, waist hip ratio, fat mass assessment by dual-energy X-ray absorptiometry, and visceral abdominal fat content determined by computed tomography. As expected, the women with PCOS had higher androgen levels. HDL levels were lower in PCOS; other cardiovascular risk factors were not different between groups. PCOS subjects were more insulin resistant with higher fasting insulin and homeostasis model assessment and lower GIR during euglycemic–hyperinsulinemic clamp studies (Table 1). Fasting plasma and adipose tissue PEDF mRNA content were not different between groups (P = 0.22 and P = 0.24 respectively) (Table 1).
Table 1 Baseline characteristics: demographics, biochemistry, and measures of adiposity Variable
Control group (n = 14)
PCOS group (n = 18)
P value
Age (years)
35.0 ± 4.3
29.8 ± 5.7
0.01
BMI (kg/m²)
35.7 ± 4.8
37.1 ± 7.0
0.54
Waist hip ratio
0.85 ± 0.1
0.86 ± 0.06
0.72
Systolic blood pressure (mm Hg)
113.4 ± 14.7
109.9 ± 14.4
0.50
Diastolic blood pressure (mm Hg)
76.1 ± 10.6
70.4 ± 8.4
0.10
Total cholesterol (mmol/l)
4.7 ± 0.8
5.1 ± 1.3
0.41
Triglycerides (mmol/l)
1.2 ± 0.60
1.50 ± 0.9
0.30
LDL (mmol/l)
3.0 ± 0.7
3.4 ± 1.1
0.23
HDL (mmol/l)
1.2 ± 0.3
1.0 ± 0.3
0.04
Blood pressure
Lipid profile
Measures of adiposity Abdominal visceral fat (cm2) Abdominal subcutaneous fat (cm2) Fat mass (kg)
131.2 ± 58.2
0.59
512.9 (432.7–629.7)
121.5 ± 35.2
563.5 (453.9–634.6)
0.74
45.3 (39.2–56.0)
45.2 (40.5–55.5)
0.96
2.5 (2.2–3.7)
5.02 (3.9–6.9)
0.01
Measures of insulin sensitivity HOMA Fasting insulin (pmol/l)
72.8 (62.6–103.1)
141.4 (104.7–176.5)
0.01
GIR (mg.m−2.min−1)
257.2 ± 64.3
175.6 ± 96.3
0.02
11.9 ± 4.0
13.7 ± 3.9
0.22
0.96 (0.45–1.46)
0.73 (0.45–0.96)
0.24
PEDF Plasma PEDF (ng/ml) PEDF mRNA
Values are reported either as mean ± s.d. (for data with normal distribution) or median (interquartile range) (for data with skewed distribution). GIR, glucose infusion rate; HDL, high-density lipoprotein; HOMA, homeostasis model assessment; LDL, low-density lipoprotein; mRNA, messenger RNA; PEDF, pigment epithelium-derived factor; PCOS, polycystic ovary syndrome.
2392
VOLUME 20 NUMBER 12 | december 2012 | www.obesityjournal.org
articles Intervention and Prevention PEDF relationships with clinical, metabolic, and hormonal status
Effects of aerobic exercise training on PEDF, clinical, metabolic, and hormonal status
Relationship between plasma PEDF levels and cardiometabolic factors are presented in Table 2. Plasma PEDF levels were negatively related to insulin sensitivity as assessed by GIR during euglycemic–hyperinsulinemic clamps. Other predictors of PEDF on univariate analyses were systolic and diastolic blood pressure, HDL, plasma triglycerides, and AVF. Adjusting for age did not impact on these results (Table 2). We also adjusted the regression analysis for fat mass to assess the effect on IR, independent on adiposity (Table 2). The relationship of PEDF to IR just failed to reach statistical significance after adjustment for fat mass (P = 0.07). As adipose tissue is postulated to be the major source of plasma PEDF and adipocyte PEDF is increased in rodent obesity (11), a univariate analysis of PEDF mRNA in adipose tissue was performed; however this did not find any statistically significant relationships.
Following exercise, a subgroup analysis was performed on 10 women with PCOS and 8 women without PCOS who completed end-point data collection following exercise intervention. There was a 44% drop out following exercise intervention. Adherence to the exercise intervention was >90% in both groups with no difference between groups as published elsewhere (22). Exercise training reduced whole group BMI (P = 0.02), but with no difference between or within groups. Insulin sensitivity was improved by exercise training in the group as a whole, as reflected by an increased GIR (P = 0.01); however, exercise training did not affect plasma PEDF in either group, nor overall (Figure 2).
Correlations
Significant correlations were found at baseline between plasma PEDF and BMI (r = 0.39, P = 0.03), systolic blood pressure (r = 0.41, P = 0.02), diastolic blood pressure (r = 0.47, P = 0.01), plasma triglycerides (r = 0.49, P = 0.004), and HDL (r = −0.46, P = 0.01) (Figure 1). PEDF correlated with AVF (r = 0.46, P = 0.01), but not with abdominal subcutaneous fat (r = 0.21, P = 0.27). GIR correlated negatively with PEDF (r = −0.41, P = 0.03). There was no significant correlation with PEDF and age (r = 0.06, P = 0.73).
Discussion
We report here that despite similar fat mass and greater IR in women with PCOS, plasma PEDF did not differ between groups. Although PEDF correlated with insulin sensitivity, this relationship, despite a trend, failed to reach significance after adjusting for fat mass. Furthermore, plasma PEDF was not reduced after prolonged exercise training, despite reduced adiposity and enhanced insulin sensitivity. A further novel finding from this study is that plasma PEDF levels correlate with both systolic and diastolic blood pressure and we confirmed the correlation of PEDF with dyslipidemia. IR, independent of obesity, is a feature present in the majority of women with PCOS (26). Lean women with PCOS are more insulin resistant than lean controls without PCOS (18). There
Table 2 Relationship of plasma PEDF levels to metabolic risk factors Variable (n)
Mean ± s.d.
PE (s.e.)
P value
R2
Adjusted for fat mass PE (s.e.)
P valuea
BMI (kg/m²)
36.5 ± 6.1
0.25 (0.11)
0.03
15.12
0.33 (0.23)
0.17
Systolic blood pressure (mm Hg)
111.4 ± 14.4
0.13 (0.04)
0.01
23.46
0.12 (0.05)
0.02
Diastolic blood pressure (mm Hg)
72.9 ± 9.7
0.19 (0.07)
0.01
21.75
0.17 (0.07)
0.01
HDL (mmol/l)
1.1 ± 0.3
−5.98 (2.13)
0.01
20.75
−5.30 (2.18)
0.02
Triglycerides (mmol/l)
1.4 ± 0.8
2.01 (0.87)
0.03
15.29
2.24 (0.82)
0.01
LDL (mmol/l)
3.2 ± 1.0
0.51 (0.74)
0.50
1.56
0.34 (0.73)
0.65
Total cholesterol (mmol/l)
4.9 ± 1.1
0.42 (0.66)
0.53
1.32
0.39 (0.64)
0.55
126.8 ± 48.7
0.04 (0.01)
0.01
20.71
0.03 (0.02)
0.04
567.1 ± 163.3
0.01 (0.004)
0.27
4.21
−0.003 (0.01)
0.73
0.03
16.57
−0.02 (0.01)
0.07
Blood pressure
Lipid profile
Measures of adiposity Abdominal visceral fat (cm2) Abdominal subcutaneous fat (cm ) 2
Measures of insulin resistance GIR (mg.m−2.min−1) Fasting insulin (pmol/l) HOMA
209.4 ± 92.7
−0.02 (0.01)
22.8 ± 16.2
0.01 (0.007)
0.16
6.94
0.01 (0.007)
0.17
5.0 ± 3.6
0.28 (0.20)
0.17
6.65
0.24 (0.19)
0.23
Data are means ± s.d. GIR, glucose infusion rate; HDL, high-density lipoprotein; HOMA, homeostasis model assessment; LDL, low-density lipoprotein; PE, parameter estimate; PEDF, pigment epithelium derived-factor; R2, coefficient of determination. a P value adjusted for fat mass. obesity | VOLUME 20 NUMBER 12 | december 2012
2393
articles
a
25
b
25
Plasma PEDF (ng/ml)
20
Plasma PEDF (ng/ml)
Intervention and Prevention
20
15 10 5 0
15 10 5 0
20
25
30
35
40
45
50
55
60
0
BMI (kg/m
d
25
Plasma PEDF (ng/ml)
Plasma PEDF (ng/ml)
c
20 15 10 5
20
250
20 15 10 5
0
f
20 15 10 5 100 120 140 Systolic BP (mm Hg)
100
200
GIR (mg/m
25
80
Plasma PEDF (ng/ml)
200
25
80
Plasma PEDF (ng/ml)
Plasma PEDF (ng/ml)
40 60 Fat mass (kg)
0
g
150
100
0
0
e
50
Abdominal visceral fat (cm2)
2)
160
300
400
2/min)
25 20 15 10 5 0
50
60
70 80 90 100 Diastolic BP (mm Hg)
110
25 20 15 10 5 0
0
1
2 3 4 Triglycerides (mmol/l)
5
Figure 1 Relationship of PEDF to adiposity measures, insulin sensitivity, blood pressure, and triglycerides. (a) Relationship of PEDF to BMI, r = 0.39, P = 0.03. (b) Relationship of PEDF to abdominal visceral fat, r = 0.46, P = 0.01. (c) Relationship of PEDF to fat mass, r = 0.31, P = 0.09. (d) Relationship of PEDF to glucose infusion rate (GIR), r = −0.41, P = 0.03. (e) Relationship of PEDF to systolic blood pressure, r = 0.41, P = 0.02. (f) Relationship of PEDF to diastolic blood pressure, r = 0.47, P = 0.01. (g) Relationship of PEDF to triglycerides, r = 0.49, P = 0.004. BP, blood pressure; PEDF, pigment epithelium-derived factor.
is emerging evidence that the underlying mechanisms of IR differ in the intrinsic and extrinsic components of IR in PCOS. This intrinsic PCOS-related IR is independent of adiposity and appears to be related to insulin-signaling abnormalities (4). 2394
Intrinsic IR is exacerbated by obesity-related extrinsic IR (4). We report here significantly greater IR in PCOS, yet no difference in plasma PEDF levels between obese insulin-resistant PCOS women compared to overweight controls. We did show a VOLUME 20 NUMBER 12 | december 2012 | www.obesityjournal.org
articles Intervention and Prevention a
Pre-exercise Post-exercise
N
on
-P C
PC
O
O
S
S
BMI (kg/m2)
45 40 35 30 25 20 15 10 5 0
BMI pre-and post-exercise P = 0.06 P = 0.14
PEDF pre-and post-exercise
b
P = 0.79
P = 0.32
13
Pre-exercise Post-exercise
12 9 6 3 0
N
on
-P
C
PC
O
O
S
S
Plasma PEDF (ng/ml)
18
GIR pre-and post-exercise
c
P = 0.07
GIR (mg/m2/min)
350 300 250 200 150
Pre-exercise Post-exercise
P = 0.03
100 50
N
on -
PC
PC O
O
S
S
0
Figure 2 Change in BMI, insulin sensitivity, and PEDF with endurance exercise training. Error bars represent s.e.m. (a) BMI pre- and postexercise. (b) PEDF levels pre- and post-exercise. (c) Glucose infusion rates (GIRs) pre- and post-exercise. PEDF, pigment epithelium-derived factor; PCOS, polycystic ovary syndrome.
relationship between PEDF and IR measured by GIR; however, this failed to reach significance after correction for adiposity. Insulin sensitivity improved after the exercise intervention, yet there was no significant change in plasma PEDF. These observations are of interest, suggesting that PEDF may not be causal for, or a specific marker of IR in humans. A recent study from China analyzed serum PEDF levels in lean (BMI