Introduction

Complementary factors 3 and 4 (C3, C4) are the major plasma proteins of the complement pathway of the immune system. The synthesis of these proteins increases in response to inflammation and infection, with a slower increase than the traditional acute-phase proteins [1–3]. Although hepatic production is the main source of these proteins, C3 and C4 are also expressed and secreted by other cells, for example, in adipose tissue [4, 5].

Individuals who are overweight or obese generally have raised concentrations of C3 and C4 [4–10], and it has been reported that C3 levels decrease after weight loss [6, 7]. The proteolytic fragment of C3, acylation-stimulating protein (ASP), is a paracrine metabolic factor that stimulates the uptake of glucose and fat storage in human adipose tissue [11–14]. Knock-out mice that are deficient for C3 and ASP show resistance to weight gain on a high-fat diet, despite increased food intake [11, 14]. However, no prospective study has investigated the relationship between C3 and weight gain in humans.

Previous studies from the present cohort have shown that raised concentrations of fibrinogen, haptoglobin, ceruloplasmin, orosomucoid and α1-antitrypsin are associated with weight gain [15]. However, the correlations between levels of C3 and levels of these inflammatory markers are small and we recently reported that C3 is associated with incidence of diabetes independently of these proteins [16]. Based on the findings from animal models, we explored the hypothesis that C3 could be associated with future weight increase independently of other inflammatory markers.

Subjects, materials and methods

Subject recruitment and study protocol

Between 1974 and 1984, 22,444 men participated in a screening programme for the detection of those at high risk for cardiovascular diseases. Complete birth cohorts from the city of Malmö were invited [17] and the participation rate was 71%. Plasma protein determination was part of the programme for 6,193 men who were selected at random from cohorts examined between 1974 and 1982. After the exclusion of men with diabetes or a history of myocardial infarction, stroke or cancer (according to questionnaire), 5,729 men remained.

A follow-up examination was performed after a mean follow-up of 6.1±0.93 years (range 3.0–9.0). Of the screened birth cohorts, men born between 1926 and 1931 and in 1938 (aged 38–50 years at baseline) were invited to the follow-up. A total of 2,822 participated, representing 81% of those who were alive in 1982 when the re-examinations started. Measurements of C3 and C4 were available for 2,706 and 2,704 of these men, respectively. The sample in the present study thus consists of 2,706 healthy men, aged 38–50 years at baseline, who were re-examined after a mean period of 6.1 years.

The health service authority of Malmö approved and funded the screening programme and all participants gave their informed consent.

Baseline examinations

Diabetes was defined as fasting whole blood glucose ≥6.1 mmol/l, 2-h glucose ≥10.0 mmol/l (glucose load: 30 g/m2 body surface area), or self-reported diabetes.

Information about fasting insulin was available in a subgroup of 1,907 men. Insulin (mIU/l) was measured with a non-specific radioimmunoassay [18]. The homeostasis model assessment (HOMA) formula according to Matthews et al. [19], i.e. (fasting insulin×fasting glucose)/22.5, was used to calculate a score for insulin resistance.

Subjects were placed in two categories: smokers and non-smokers. Physical inactivity was assessed in a questionnaire at baseline and at follow-up. Men who reported that they were mostly sedentary in their spare time were categorised as physically inactive. Some items of the questionnaire were changed before the end of the follow-up period. At the follow-up examination, physical inactivity was therefore defined either as those who were mostly sedentary in their spare time or those who reported that they did not perform physical activity in their spare time (e.g. walking, cycling) regularly every week. Alcohol consumption was assessed using the modified shortened version of the Michigan Alcoholism Screening Test [20]. Men with more than two (of nine) affirmative answers were considered to be high-consumers of alcohol.

Weight was measured in the morning with the subject wearing light indoor clothing and no shoes. Weight increase (in kg) was calculated as weight at baseline subtracted from the follow-up value. A large weight gain was defined as one on or above the 75th percentile (>3.8 kg) [15].

Plasma proteins

Electroimmunoassay was used to assess the concentrations of plasma proteins [21]. The coefficient of variation was <5% [21]. The analyses were performed consecutively at the study entry. C3 and C4 were originally expressed as the percentage of the mean values from a reference population of blood donors. The reference values for C3 were 70–130%, which corresponds to 0.67–1.29 g/l, and the reference values for C4 were 65–170% (0.13–0.32 g/l). To facilitate the interpretation of the C3 and C4 values, the percentages have been converted into g/l (C3:100%=0.98 g/l, C4:100%=0.20 g/l). The distributions of the other plasma proteins (fibrinogen, haptoglobin, ceruloplasmin, orosomucoid, α1-antitrypsin) have been presented elsewhere [22]. The correlation between C3 and C4 was r=0.47 and the correlations between C3 and the other proteins were all below 0.3 in this cohort [16].

Statistics

One-way analysis of variance and the Mantel–Haenzel χ 2 test were used to compare the risk factors by quartiles of C3. Pearson’s correlation was used for two continuous variables. A general linear model was used to compare weight at follow-up in quartiles of C3, with adjustments for confounding factors. Logistic regression was used to study the proportion of men with large weight gain in relation to plasma protein levels. The relationships between C3 and weight gain were adjusted for initial weight in all calculations. Two-tailed p values <0.05 were considered statistically significant.

Results

The baseline and follow-up characteristics by quartiles of C3 are presented in Table 1. BMI and weight were substantially higher in men with high C3. The proportion of smokers was somewhat lower in men with high C3 (Table 1). However, C3 showed no association with smoking after adjustments for BMI (data not shown).

Table 1 Description of the study cohort at baseline and at follow-up in relation to quartiles of C3 at baseline
Full size table

Weight at baseline was inversely correlated with weight increase during the follow-up (r=−0.17, p<0.0001). The proportion with a large weight gain (>3.8 kg) was 24.1% in men with low concentrations of C3 (1st quartile), and 25.9% in men with high concentrations of C3 (4th quartile). However, when weight at baseline, age, height and follow-up time were taken into account, a large weight gain was significantly associated with C3 at baseline (Table 2). This relationship remained significant after further adjustments for lifestyle factors (Model 2: +smoking, physical activity, alcohol consumption), metabolic factors (Model 3: +glucose, cholesterol, triglycerides), and inflammation-sensitive proteins (ISPs) (Model 4: +fibrinogen, ceruloplasmin, α1-antitrypsin, orosomucoid, haptoglobin). The results also remained significant after adjustments for C4. The relationships were similar when C3 was used as a continuous variable (per SD) (Table 2).

Table 2 Weight gain in relation to quartiles of C3
Full size table

Weight at follow-up showed similar associations with C3 after adjustments for initial weight (Table 2). C3 was significantly associated with weight at follow-up after adjustments for inflammatory markers (Model 4). After further adjustments for C4, this association was of borderline significance (p=0.06).

Weight loss >1.2 kg (lowest quartile of weight gain) was not associated with C3 after adjustment for initial weight (Table 2).

The results were identical if weight increase (initial weight subtracted from weight at follow-up) was used instead of weight at follow-up. The results were also identical if BMI was used to assess weight and weight increase (not shown). The results were essentially unchanged after exclusion of 117 subjects who developed diabetes during the follow-up period (not shown).

The relationships between C3 and weight gain were largely consistent over the quartiles of BMI. The relationships were somewhat weaker in smokers than in non-smokers or in smokers who gave up smoking during the follow-up period (Table 3).

Table 3 The odds of a large weight gain (>3.8 kg) in relation to C3, by smoking and quartiles of BMI
Full size table

Subgroup with HOMA-values

Information about HOMA was available for a subgroup of 1,907 men. The relationship with large weight gain was studied in this subgroup, with and without adjustments for HOMA (log-transformed). After adjustments for initial weight, age, height, follow-up time, lifestyle, and metabolic factors (Model 3), the odds ratios for a large weight gain were 1.00, 1.1 (0.8–1.6), 1.2 (0.9–1.7) and 1.56 (1.1–2.2) respectively for C3 levels in the first, second, third and fourth quartiles (p for trend=0.008). Odds ratio (OR) per 1 SD increase in C3 was 1.17 (1.04–1.32, p=0.007). When HOMA was added to the model, the ORs were 1.00, 1.1 (0.8–1.6), 1.2 (0.9–1.7), and 1.54 (1.1–2.1), respectively (p for trend: 0.009). OR per 1 SD was 1.17 (1.04–1.31, p=0.01). These relationships remained significant after further adjustments for ISPs (Model 4).

The relationship between C3 and weight at follow-up was significant after adjustments for lifestyle, metabolic factors (Model 3) and HOMA (adjusted mean+standard error, 78.1+0.20, 78.3+0.22, 78.3+0.20 and 78.8+0.21 kg, respectively, for the quartiles of C3 concentration, p for trend <0.05). After further adjustments for inflammatory proteins, this association did not reach significance in this subgroup (p for trend=0.11).

ISPs and C4

Table 4 presents the relationships between a large weight gain and the number of ISPs in the top quartile. The risk of a large weight gain was significantly associated with the number of elevated ISPs, and this relationship remained significant after adjustments for C3 and C4.

Table 4 Weight gain in relation to inflammation-sensitive plasma proteins (ISPs)
Full size table

Table 5 presents the relationship between C4 and a large weight gain. C4 was associated with weight gain after adjustments for initial weight, follow-up time and lifestyle factors. However, there was no association between C4 and weight gain after adjustments for C3. This relationship was similar when C4 was used as a continuous variable (not shown).

Table 5 Weight gain in relation to quartiles of C4
Full size table

Discussion

Cross-sectional studies have shown that obesity is associated with elevated plasma levels of C3 [4, 6–10], but the causal and temporal relations of this association have been unclear. The present study showed that C3 was strongly associated both with weight at baseline and weight at follow-up. When adjusting for initial weight, C3 was significantly associated with an increased risk of developing a large weight gain. To our knowledge, this is the first study linking C3 to future weight gain in humans.

Although the causes of obesity are multifactorial, increasing evidence links inflammation and infections to obesity or weight gain [23–28]. Seropositivity for Chlamydia pneumoniae has been associated with high BMI in humans [28] and in mice virus infections have been associated with weight gain [27]. Elevated levels of fibrinogen, white blood cells, von Willebrand factor or factor VIII, i.e. four putative markers of inflammation, have been associated with large weight gain in a large cohort study [26]. A study from this cohort showed that high levels of fibrinogen, ceruloplasmin, α1-antitrypsin, orosomucoid and haptoglobin were associated with weight gain [15]. However, C3 shows only modest correlations with these inflammatory markers (all r<0.3) [16]. The relationship between C3 and weight gain was essentially unchanged after adjustment for inflammatory proteins and, conversely, the number of elevated ISPs was associated with weight gain even after adjustments for C3. Hence, the relationship between C3 and weight gain cannot be explained by inflammation as measured by these acute-phase proteins. Thus C3 and the ISPs could represent different pathways of weight gain.

Another question is whether the relationships are specific for C3 or whether other complement components are also associated with weight gain. Data on C4 were available in this study and the correlation between C3 and C4 was approximately r=0.5. The relationship between C3 and weight gain was slightly reduced after adjustments for C4. In contrast, the relationship between C4 and weight gain became clearly non-significant after adjustment for C3. Hence, the association between C4 and weight gain seems to be explained by its relationship with C3.

ASP is a possible link between C3 and weight gain. This hormone has autocrine or paracrine effects and is produced in the adipose tissue by conversion from C3 [11]. It has been shown that ASP stimulates the storage of fat in human adipocytes, inhibits intracellular lipolysis and increases plasma clearance of triglycerides [11–13]. ASP also stimulates the disposal of glucose by increasing the release of insulin [29]. Mice with C3 and ASP deficiencies show resistance to weight gain despite increased food intake [14]. These animals also have increased energy expenditure [30]. However, it is unclear to what extent circulating C3 reflects the concentrations in the adipose tissue. As local effects in the adipose tissue are probably most important, circulating levels of C3 could be a crude proxy for the local effects.

The hepatic synthesis of C3 is stimulated by various cytokines, e.g. IL-1 and TNF-α [31]. These cytokines are also involved in the regulation of metabolism and food intake. For example, TNF-α and IL1-β decrease the production and expression of leptin, an anorexigenic hormone formed in the adipose tissue [32]. Polymorphisms of the genes encoding TNF-α receptor have been associated with leptin-resistance and obesity [33, 34]. In studies of rodents, injection of TNF-α and IL-1 initially reduces food intake. However, tolerance soon develops after repetitive administration of TNF-α [35]. The relation between plasma C3 and weight gain could hypothetically reflect the effects of these cytokines.

We had no information on diet in this cohort. It is unclear whether the relations between weight and C3 could reflect dietary factors that increase C3. It has been reported that intake of lipids and proteins could induce oxidative stress [36], which hypothetically could increase inflammation. On the other hand, studies of C3 levels before and after food intake have not shown any consistent differences. Two studies reported no postprandial changes [37, 38], while another study showed a small increase [39]. Since the plasma samples were drawn after an overnight fast, this probably reduced the effects of food intake.

The participant rates were high and the procedures were identical at both examinations. As smoking cessation is associated with weight gain, smoking is a potential confounder. However, smoking was not associated with C3 after body weight had been taken into account and the relation between C3 and weight gain persisted after adjustments for current and former smoking and smoking cessation during follow-up. Physical activity is another factor that could affect weight gain. The results were adjusted for physical inactivity both at baseline and at follow-up. However, only two categories of physical activity were used and it is possible that the variables did not pick-up all effects of physical activity.

Another limitation is that only men were included in the study. Apart from cytokine stimulation of C3 synthesis, C3 concentrations are determined by immune-complex deposition, genetic factors and female sex hormones [40]. It is not known whether the relation between C3 and weight gain is similar in women.

High levels of C3 are a risk factor for incurring large weight gain. The reasons for this association remain to be explored.