Positive correlation between tumor necrosis factor ( TNF )-and cardiorespiratory fitness after six-months of regular aerobic exercise in Peruvian Amerindian women

Background: Adipose tissue is a source of infl ammatory cytokines which may infl uence insulin action. Aim: To evaluate exercise effects on plasma (p)-levels of tumor necrosis factor (TNF)-, and interleukin (IL)-6 in Peruvian Amerindian women. Material and Methods: After fi ve years of observation during which pglucose deteriorated and weight increased, 44 Peruvian Amerindian women aged 20-59 years took part in a 6-month exercise program. Plasma TNF- and IL-6 had been measured 5 years earlier, and were reassessed before and after exercise, using ELISA kits. Cardiorespiratory fi tness was derived from the results of the twelve minutes’ walk and expressed as maximum oxygen uptake (VO2max). Results: After the training program, estimated VO2max increased from 18.55 ± 3.79 to 20.91 ± 4.61 mL kg min (p < 0.01) and p-TNF- increased from 3.2 ± 3.4 to 4.3 ± 6.6 pg/ mL (p = 0.03). There were reductions in p-glucose from 5.19 ± 1.59 to 4.32 ± 1.62 mmol/L (p < 0.01) and in waist circumference (p=0.01). Weight, p-insulin, and IL-6 did not change. Levels of p-glucose and p-TNF- observed after the training period were identical to values 5 years earlier. P-TNF- was positively associated with estimated VO2max, prior to (r = .414, p < 0.01) as well as after (r = .362, p < 0.05) exercise, independently of adiposity. P-IL-6 was associated with waist circumference but not with estimated VO2max. Conclusions: P-TNF- correlated with estimated VO2max independently of adiposity. Benefi cial effects of exercise on p-glucose were observed in spite of an increase in p-TNF-. (Rev Med Chile 2011; 139: 998-1005).


S
everal prospective studies have shown lower cardiorespiratory fi tness to be associated with higher risk of glucose intolerance and type 2 diabetes in normoglycemic individuals.This association is independent of adiposity, and a potential underlying mechanism may be that lower physical fi tness increases systemic infl ammatory activity 1 .Improved physical fi tness increases insulin sensitivity but the mechanisms for these effects are still somewhat unclear.One possible pathway may be through anti-infl ammatory effects attributed to regular physical activity.
Whether the cytokine interleukin (IL)-6 has positive or negative effects on metabolic processes is a subject of continuing controversy 2 .Adipose tissue is an important source of IL-6, and increased adiposity is known to be associated with reduced physical fi tness 3 .Similarly, lower physical fi tness is associated with higher plasma (p) levels of IL-6 1 , and several prospective studies of exercise training have demonstrated reductions of p-IL-6.
Tumor necrosis factor (TNF)- is considered to contribute to insulin resistance by inhibiting the expression of genes essential for insulin signalling and adipocyte differentiation 4 .P-levels of TNF- do not refl ect regular physical activity in healthy non-diabetic women and are not associated with fi tness after adjusting for visceral adiposity 5 .
Studies on the effect of aerobic exercise on fi tness and adipocytokines in subjects with deteriorating p-glucose derived from other ethnic groups than Caucasians are lacking.We have previously shown that Amerindian women from the capital of Peru, Lima, are characterized by substantial hyperinsulinemia, increased body fat mass and waist circumference but lower p-glucose and leptin than in a matched white population 6 .Five years later the participants in this prospective study had developed signifi cantly (p < 0.0001) higher fasting p-glucose with unaltered p-insulin and leptin 7 .However, a six-months supervised aerobic training period resulted in an increase of physical fi tness (VO 2 max ), and decreased p-glucose 8 .
As regular aerobic exercise may improve insulin sensitivity by modulating plasma content and/ or function of adipocytokines, the present study focussed on potential effects of aerobic training on circulating p-levels of IL-6 and TNF-.
We investigated whether cardiorespiratory endurance (VO 2 max ) prior to and after aerobic training was associated with changes in p-IL-6 and TNF- independently of visceral adipose tissue accumulation, and whether p-IL-6, TNF-, glucose and fi tness prior to the exercise programme were determinants of p-glucose after the exercise period.
on topics related to urban poverty in the district of San Martin de Porres in Lima.Alternativa is working with community kitchens organized by women, facing nourishment problems by collective cooking of the main meal.
About 15 to 20 families constitute each kitchen in the district of San Martin de Porres.Women were invited from 14 community kitchens to the Department of Nutrition and Health Education of Alternativa.
The study population consisted of 182 Peruvian Amerindian women who took part in the baseline study in 1999, and of those two were excluded due to a baseline p-glucose level exceeding 7.0 mmol/L 6 .A migration antecedent from indigenous Andean communities in Peru was identifi ed in all participants.The age range was 20-59 years, with a mean age of 41 years.Fatty acid composition in serum was measured in 141 subjects 7 .This subgroup was invited to take part in the follow-up examination in 2004.A total of 83 women (59%) participated in the second examination.Of these, 76 consented to take part in the aerobic training study, and 59 completed it.Fitness was measured and data on p-TNF- and IL-6 were available from 44 of these participants 8 .These 44 women did not differ signifi cantly from the remaining participants regarding baseline measurements of body weight, body mass index (BMI), waist circumference, p-glucose, p-insulin and p-leptin.None of the participants had ongoing cardiovascular disease but two subjects had developed diabetic fasting p-glucose concentrations at the reexamination in 2004 7 .Approval for the original study had been given in 1998 and 1999 by the Ethics Committees of Lund University and San Martin University Hospital in Lima.Furthermore, the follow-up and training studies were separately approved by the Ethics Committe of San Martin University Hospital.The study was undertaken in accord with the Helsinki Declaration of 1975, as revised in 1983.

Methods
The examination procedures were the same at all three occasions.The examinations took place in the morning after an overnight fast at Alternativa.All subjects refrained from any severe physical activity 48 h before measurements.Body weight and height were measured with the subject wearing light clothes.Abdominal obesity was determined as waist circumference measured in a standing position.BMI was calculated in kg/m 2 .Cardiorespiratory fi tness was assessed as VO 2max (mL kg -1 min -1 ) indirectly after subjects ran around two cones spaced 20 m apart for 12 minutes, as previously described 9 .

Exercise training
Exercise sessions were undertaken once or three times per week 8 , and always supervised by a physiotherapist and took place outdoors on a square with a concrete surface.Of the 44 women presented in this paper, 23 exercised once a week, and 24 subjects attended three training sessions/ week.There was a close association between the number of training sessions and improvement in VO 2max (p = 0.007).A warm-up with stretching, light jogging and fl exibility movements was followed by aerobic dancing for a total of 60 minutes.The average metabolic equivalent (MET) score calculated for total physical activity performed per training session was 275 or 4.6 MET-minutes / session 10 .

Dietary intake
Twenty-four-hour recall was used to evaluate the composition of the diet at the baseline examination, and analyzed with the use of the Peruvian tables of food composition from the Ministry of Health (7 th edition), Lima, Peru, 1996.

Measurements
Serum or plasma was separated from venous blood and stored within 1 hour at minus 20 °C and then brought to Sweden for analysis.Insulin was measured with double-antibody radioimmunoassay techniques with the use of guinea pig anti-human insulin antibodies and human insulin as the standard (Linco Research, St Charles, MO).Leptin was measured with a radioimmunoassay specifi c to human leptin (Linco Research).
P-glucose was measured by using the glucose oxidase procedure.P-TNF- and p-IL-6 were measured by ELISA using commercially available test kits (Pharmingen, San Diego, CA, USA.Detection limits were 0.12 pg/ml and 0.70 pg/ml 11 , intraassay coeffi cients of variation (CV) were 8.8% and 4.4%, and interassay CV were 16.7% and 6.4%, respectively.

Statistical analyses
Data are expressed as mean±SD.Comparisons between values at baseline and examinations before and after the training period were determined by use of Wilcoxon signed rank tests.The relation between waist circumference, VO 2max and infl ammatory markers were assessed by adjusted and unadjusted Spearman rank correlations.Results were declared signifi cant at p < 0.05.P-TNF-, IL-6 and glucose as well as VO 2max values were logarithmically transformed to reduce skewness.Multiple regression analysis was undertaken to examine which variables predicted p-g lucose concentration after the intervention period.The selected variables were forced into the model based on analysis (Table 2).All statistical analyses were conducted by using the statistical package Stat-View (version 5.0.1, for Macintosh; SAS Institute Inc, Cary, NC).

Anthropometric and biochemical characteristics
Table 1 presents anthropometric data, physical fi tness, and related biochemical characteristics at baseline and fi ve years later, both prior to and after the aerobic training program.An increase in aerobic fi tness by 12.7 % (from 18.55 to 20.91 mL .min -1 kg -1 ; p < 0.0001) was achieved during the aerobic training programme.Body weight and BMI were unchanged in contrast to waist circumference which decreased by 1.31 % (p = 0.0128).Exercise training did not infl uence fasting p-insulin and pleptin, although mean p-glucose was signifi cantly (p < 0.0001) lower after the intervention period.P-TNF- was higher (p = 0.027) after aerobic training, and was almost the same as baseline values fi ve years earlier.P-IL-6 before and after training were not signifi cantly different from each other (p = 0.152).

Associations between fi tness, p-TNF- and IL-6
Regression analysis showed that log TNF-was positively correlated to logVO 2max , both prior to (p=0.0024) and after (p=0.0197) the exercise programme, whereas there were no signifi cant associations between p-IL-6 and logVO 2max .

Associations of p-TNF- and IL-6 with abdominal adiposity and fi tness
Spearman correlation coeffi cients between waist circumference, cardiorespiratory physical fi tness (CRF) and p-TNF- and IL-6 prior and after the training period are presented in Table 2. Waist circumference correlated positively with p-IL-6 before and after adjustment for CRF, whe-reas no correlation was found between p-TNF- and waist circumfere nce.
CRF was positively correlated with p-TNF- before and after adjustment for waist circumference, whereas no relation was observed between CRF and p-IL-6.The correlation between CRF and p-TNF- was sig nifi cant both prior to and after the exercise training period.

Predictors of p-glucose at the end of the exercise training period
In a multiple regression analysis including p-glucose, p-TNF-, p-IL-6, and VO 2 max prior to the exercise training as independent variables, p-glucose and VO 2 max were determinants of pglucose after training (Table 3).

Discussion and conclusions
The main objective of the present study was to investigate the effect of aerobic training on p-TNF- and IL-6 in a cohort of sedentary Amerindian women characterized by deteriorating p-glucose and low cardiorespiratory fi tness 8 .Furthermore, we wanted to assess the respective contribution of abdominal adiposity and cardiorespiratory fi tness to variation of p-TNF- and p-IL-6 prior to as well as after the aerobic training.The six-month exercise training programme resulted in an increase of subject´s VO 2max by an average of 12.7 %, and promoted a signifi cant (p = 0.0266) increase of p-TNF- simultaneously with a reduction (p < 0.0001) in p-glucose (Table 1).P-TNF- was positively associated with VO 2max, prior to as well after the aerobic training period, a relation that was independent of the infl uence of abdominal adiposity (Table 2).P-IL-6 was associated with waist circumference but not with fi tness.
The strong association between p-IL-6 and waist circumference, as well as the fact that this relation is independent of the infl uence of fi tness are both in agreement with other studies 5 , whereas the positive relation between p-TNF- and VO 2max is a novel observation.
Weight reduction in obese subjects following a hypocaloric diet for 12 months resulted in signifi cant decreases of p-TNF- and p-IL-6 12 , whereas in contrast p-TNF- concentrations were not infl uenced by exercise training or induced weight losses in non-obese 13,14 and obese pre- 15 and postmenopausal 16 women.In the present study exercise training did not change body weight, p-leptin or p-insulin, whereas mean elevated p-glucose was signifi cantly lower after the intervention period.Thus, it is likely that the improved insulin sensitivity was not a consequence of changed body fat mass, but instead associated with increased fi tness and p-TNF-.
TNF- is secreted from a variety of cells including macrophages, muscle, and adipose tissue.Most of the TNF- secreted by adipose tissue is derived from macrophage infi ltration or other nonfat cell sources.Subcutaneous adipose tissue has been claimed to produce more TNF- than visceral 17 , but in contrast to for IL-6 no net release of TNF- to plasma has been detected from this source 18 .As p-levels of TNF- were not associated with gene expression of TNF- in subcutaneous adipose tissue, neither at the beginning nor at the end of three months aerobic training 15 , the most likely explanation is that the elevated p-TNF- levels after long term aerobic training program-mes are derived from other sources than adipose cells per se.
In the present study mean p-TNF- increased 28% after aerobic training, and reached the same values that had been observed fi ve years earlier when all women had normal fasting glucose concentrations.A likely explanation for this discrepancy from other investigations resulting in reductions or moderate increases of mean p-TNF- concentrations [13][14][15][16] , is that the food consumption pattern in our study population is characterized by a low percentage of calories from fat, 19 %, Table 1, in comparison with reports from industrialized countries 7 .Thus, the observation by Meksawan et  al. that the magnitude of the increase in p-TNF- concentration at post-exercise on a 19% fat diet is greater than that on 30% or 50% fat diets 19 may in part explain our fi ndings.
To reconcile the apparently confl icting data on whether changed p-TNF- concentrations at muscular work are deleterious or benefi cial to insulin sensitivity, recent evidence suggesting that infl ammation may have two different roles in the regulation of metabolism are of importance 20 .The fi rst may be inhibition of insulin sensitivity (a negative effect), as suggested by many studies of low grade infl ammation 4,21,22 , and the second may be induction of energy expenditure (a positive effect) 20 .
Our study population is characterized by obesity, a sedentary lifestyle and low cardiorespiratory fi tness 7 .Only 36% of the women had a normal BMI (< 25 kg/m 2 ) when the training period started.That is comparable to fi gures reported by Jacoby et al 23 in a health survey performed in six Peruvian cities.It is most likely that the increases of body weight and waist circumference during the 5 years non-intervention period prior to the aerobic training were due to reduced energy expenditure.In transgenic mice with defi ciencies in TNF-, IL-1, and IL-6 levels or their receptors, body weight gain is enhanced 24 .In contrast, when cytokine activity is stimulated, energy expenditure is increased and body weight gain is attenuated 25 .
VO 2max, evaluated prior to the start of the intervention programme was an independent predictor of post-exercise p-glucose, in contrast to p-TNF- and p-IL-6 (Table 3).
In a large prospective study p-TNF- was not an independent determinant of future type 2 diabetes development 26 .This is also in agreement with the observation of no effect of circulating TNF- on whole-body insulin sensitivity in predominantly nondiabetic subjects 27 .The effect of p-TNF- on insulin resistance is therefore questionable, and the association with improved p-glucose in the present study may be indirect through the benefi cial impact of physical fi tness on glucose tolerance.Our study did not include a control group, and this issue needs to be studied under controlled conditions with greater power in larger materials.Furthermore, it is apparent that the present study provides no evidence that IL-6 is directly involved in exercise related improvements of glucose toleAt the baseline examination all Amerindian women had normal fasting p-glucose but increased p-insulin in comparison to Caucasian subjects 6 , and already before any exercise intervention we found disassociations between p-TNF- and p-IL-6 in relation to changes of body weight and waist circumference in our material 28 .The extent to which our fi ndings may be generalized to subjects of other ethnic backgrounds is not known, however.Aboriginal Canadian women with normal fasting p-glucose showed no increased p-TNF- values compared to white women, despite an increased insulin resistance 29 .
In conclusion, the present study indicates that p-concentrations of TNF-, but not of IL-6, are positively associated with cardiorespiratory fi tness in Amerindian women prior to as well as after sixmonths of endurance physical exercise, a relation that is independent of the infl uence of abdominal adiposity.Improved aerobic fi tness and elevated p-TNF- concentrations are associated with lower fasting p-glucose values when these women participate in a long-term aerobic exercise programme.
TNF- and fitness after aerobic exercise in Amerindian women -F.Lindgärde et al

Table 1 .
Data from baseline and 5 years later prior to and after exercise intervention in 44 Peruvian Amerindian women Mean±SD.No intervention took place between baseline and the training programme.ND=not done, P=plasma, TNF=tumor necrosis factor.TNF- and fitness after aerobic exercise in Amerindian women -F.Lindgärde et al Rev Med Chile 2011; 139: 998-1005

Table 3 .
Multiple regression with plasma (p)-tumor necrosis factor (TNF)-α, interleukin-6, glucose and VO 2max prior to exercise intervention as independent variables and p-glucose measured after six-months exercise as dependent variable in 44 Peruvian Amerindian women

Table 2 .
Unadjusted and adjusted Spearman correlations between waist circumference, cardiorespiratory fitness (CRF), and inflammatory markers tumor necrosis factor (TNF)-α and interleukin (IL)-6 in 44 Peruvian Amerindian women before and after exercise intervention