1 INTRODUCTION

In 1975, <1% of children and adolescents (aged 5–19 years) had obesity (4% if including children and adolescents with overweigh as well) globally.¹ In 2016, approximately 40 years later, the prevalence rate of obesity had increased to 6% for girls and 8% for boys, while 18% of girls and 19% of boys were overweight.¹ Overweight and obesity are major risk factors for cardiovascular diseases, the main cause of death worldwide in 2017,² especially for women with insulin resistance.³ In addition, for adolescents with higher body mass index (BMI) or obesity, the risk for several forms of cancer such as leukemia and colorectal cancer or adult diabetes type II, respectively, is also increased.⁴

Even though overweight and obesity are dependent on lifestyle factors, such as diet and physical activity, obesity is a heritable trait with known genetic influence.⁵ Genome-wide association studies have identified more than 100 loci that are robustly associated with BMI.^6-8 The fat mass and obesity-associated gene (FTO; FTO alpha-ketoglutarate-dependent dioxygenase) is the gene with the strongest influence on BMI.⁹ In 2007, FTO was determined to be one of the contributing factors to early-onset obesity and severe obesity.^{10, 11} Since then, knowledge regarding the association between FTO and overweight and obesity has increased.^12-17 FTO, located on chromosome 16, encodes a 2-oxoglutarate-dependent nucleic acid demethylase,⁹ which is expressed in all tissues, but its highest levels are observed in brain areas that control food intake.^{18, 19} FTO also plays an important role in adipogenesis.²⁰ Single nucleotide polymorphisms (SNPs) in FTO appear to regulate the expression of other genes in a tissue-specific manner that varies depending on the developmental stage of the tissue.²¹ The causal variant rs1421085 disrupts a binding site for a gene repressor, which leads to overexpression of Iroquois homeobox 3 and Iroquois homeobox 5, which in turn changes these cells to energy-storing white adipocytes and increases lipid accumulation.²² This SNP is in high linkage disequilibrium with many other SNPs in the genetic region, including rs9939609 (distance 19,573 bp, D' = 0.98, and R² = 0.92 in the CEU population, which includes Utah residents with Northern and Western European ancestry, according to the LDlink tool²³), which, therefore, can act as a proxy SNP.

The A allele of rs9939609 in the intron of FTO has been associated with higher risk for obesity^{12, 17} and type 2 diabetes.¹² Several studies have found that this association changed throughout the life course.^{16, 24} Hardy et al. found a stronger association during childhood and adolescence with a peak at 20 years, after which the association declined again during adulthood.¹⁶ Sovio et al. found that the relationship between the A allele and the BMI was inversed for children below 2.5 years of age.²⁴ However, some did not find any influence by genotype,²⁵ while others found it only in girls.²⁶

Because obesity is difficult to reverse once established,²⁷ it is important to identify early risk factors as well as protective factors for a healthy weight pattern. One factor that occurs during the child's earliest life and may have an impact on obesity is breastfeeding. The World Health Organization recommend exclusive breastfeeding for up to 6 months of age with continued breastfeeding for up to 2 years of age or beyond.²⁸ The Swedish National Board of Health and Welfare and the National Food Agency have adapted this recommendation to Swedish conditions promoting exclusive breastfeeding for up to 6 months of age and continued breastfeeding for up to 1 year or beyond.^{29, 30} Despite these breastfeeding promoting recommendations,²⁹ breastfeeding is declining in Sweden.³¹ A dose–response relationship between breastfeeding duration and future overweight risk has been observed.³² Compared with breastfeeding, formula-feeding is associated with higher levels of protein and energy intake³³ as well as higher plasma insulin levels that promote fat deposition.³⁴ In contrast, breastfeeding is associated with higher levels of infant control of feeding and self-regulation.³³ Altogether, formula-feeding is associated with a more rapid weight gain³⁴ through overnutrition.³² This is considered as one of the main risk factors in the first 1000 days since birth for the development of childhood obesity.³⁵

Earlier attempts at investigating a possible interaction effect between breastfeeding and FTO rs9939609 variants on future BMI have revealed inconclusive results. In a Chinese study on 1149 children ( = 10.8 years) analyzing BMI or body fat percentage, no significant interaction effect was found.³⁶ In contrast, a study on 959 children from Australia showed that a longer period of exclusive breastfeeding could reverse the increased BMI otherwise associated with the A allele.³⁷ Consistent with this, results from a study on British children revealed that at least 5 months of exclusive breastfeeding had a substantial impact on BMI growth trajectories.³⁸ Dedoussis et al. reported an interaction effect in two Greek samples (FTO rs17817449 was analyzed in one of the samples; N = 1138, mean age = 11.2 years and N = 2374, mean age = 41.4 months). However, their findings could not be replicated in a British sample of 4325 participants (mean age = 11.7 years).³⁹ Further studies are needed to establish a consensus regarding these relationships. To our knowledge, these associations are yet to be investigated in a Nordic sample.

Dalle Molle et al. suggested that the differential susceptibility to environment model; that individuals vary in their susceptibility to environmental influences and plasticity,⁴⁰ is “highly applicable to the study of genetic and environmental influences on feeding behaviour and obesity risk” as well.⁴¹ In other words, the incorporation of both positive and negative environmental influences in these studies is essential. In this case, shorter or longer breastfeeding duration could be viewed as negative or positive environmental influence, respectively. Due to the lack of studies that apply this theory on obesity research, we consider this a gap in this research field.

This study hypothesized that breastfeeding duration plays a role in weight regulation through links to feeding behavior as well as through the nutritional value and that individuals differ in their susceptibility to this influence due to genetic variants. The aim of this study was to investigate the possible effect of childhood breastfeeding duration on the association between FTO rs9939609 and BMI and overweight in adolescence and young adulthood in a Caucasian sample, taking the levels of physical activity into account.

2 MATERIALS AND METHODS

2.6 Statistical analyses

Using χ² tests for categorical variables and the Mann–Whitney U test for continuous variables, we analyzed whether there were any sex differences between rs9939609 allele variants, breastfeeding duration, age, physical activity, and overweight. In addition, we analyzed rs9939609 allele variants using χ² tests for associations with BMI categories (normal weight and overweight) at the three time points. Rs9939609 allele variants were also analyzed for any effect of breastfeeding duration and physical activity using the Mann–Whitney U test.

The moderating effect of parental reported breastfeeding duration in months on the relationship between FTO rs9939609 variants and BMI was analyzed using the PROCESS macro tool (version 3.5⁴⁴). We analyzed the PROCESS “Model 1” (Figure 1) separately for each time point using FTO rs9939609 as X; BMI as Y; breastfeeding duration in months as moderator (W); and sex, age, and physical activity as covariates. The confidence interval was set to 95% with 5000 bootstrap samples using no centering of the variables. We probed the interactions if p < 0.05 and conditioning values were set to 16th, 50th, and 84th percentiles. To determine the value(s) of the moderator where p = 0.05, we used the Johnson–Neyman significance output, which also defines the regions of significance.

To assess the possible interaction between rs9939609 variants and parental reported breastfeeding duration for risk of overweight, we included rs9939609 and breastfeeding duration categories (Model 1) and, additionally, an interaction term between rs9939609 and breastfeeding duration (Model 2) in three separate binary logistic regression models using overweight at each time point as the outcome. These analyses were controlled for age, sex, and physical activity at the time. Data were analyzed using IBM SPSS Statistics (version 24 for Windows; IBM SPSS).

3 RESULTS

3.2 Breastfeeding as a moderator

Process analyses revealed a significant moderation effect of parental reported breastfeeding on the relationship between rs9939609 and BMI at W1 and W2 but not W3 (Table 2). R² for the model summary was 0.054 (p < 0.001) for W1 and 0.026 (p < 0.001) for W2. The R² change for the interaction term was 0.004 (p = 0.029) for W1 and W2. Figure 2A,B displays plots for the conditional effect of the focal predictor using data from the PROCESS output at W1 and W2. Figure 2A shows that for TT individuals, the BMI slope starts at a lower level than for AT and AA individuals, and then, it increases slightly with longer breastfeeding duration. For AT individuals, the slope starts at a level between those for TT and AA individuals, and then, it declines with longer breastfeeding duration. The slope for AA individuals starts at the highest level and has the steepest decrease as the breastfeeding duration increases. In Figure 2B, the slopes follow the same pattern as in Figure 2A, but the slopes for AT and AA overlap. The Johnson–Neyman output revealed that values <1.64 and >28.05 for the moderator were statistically significant at W1 (Figure 3A). For W2, values >19.85 for the moderator were statistically significant (Figure 3B). We did not find any statistically significant interaction effect for W3 BMI. The binary logistic regression analyses showed that parental reported breastfeeding for 7–12 months was associated with lower odds ratio (OR) for overweight at W1 compared with shorter breastfeeding (Model 1, Table 3). The combination of rs9939609 TA and breastfeeding for >12 months was associated with a lower OR than the reference combination of rs9939609 TT and breastfeeding for 0–6 months (Table 3). Using FTO rs9939609 as a continuous variable in the regression model for W1 overweight, we found that for each additional A allele combined with breastfeeding for >12 months, the OR decreased significantly (OR = 0.41, 95% confidence interval: 0.19–0.88). All variables in the binary logistic regression models for W2 and W3 were nonsignificant (data were not shown). The complete output from the analyses can be obtained from the main author upon reasonable request.

TABLE 2. Breastfeeding duration in months as a moderator of the relationship between FTO rs9939609 variants and BMI

	W1 BMI (N = 1221)						W2 BMI (N = 1120)						W3 BMI (N = 771)
	Coeff	SE	t	p	LLCI	ULCI	Coeff	SE	t	p	LLCI	ULCI	Coeff	SE	t	p	LLCI	ULCI
FTO rs9939609	0.49	0.24	2.07	0.04	0.03	0.96	0.5	0.28	1.78	0.08	−0.05	1.04	0.45	0.34	1.33	0.19	−0.22	1.12
Breastfeeding	0.01	0.02	0.57	0.57	−0.03	0.06	0.02	0.03	0.63	0.53	−0.03	0.07	−0.02	0.03	−0.67	0.51	−0.08	0.04
Int_1aa	−0.05	0.02	−2.19	0.03	−0.1	−0.01	−0.06	0.03	−2.18	0.03	−0.12	−0.01	−0.04	0.03	−1.21	0.23	−0.11	0.03
Sex	0.08	0.2	0.39	0.7	−0.31	0.46	−0.34	0.23	−1.49	0.14	−0.79	0.11	−0.23	0.29	−0.79	0.43	−0.81	0.34
Age	0.69	0.09	7.37	<0.001	0.5	0.87	0.5	0.11	4.58	<0.001	0.29	0.71	0.12	0.13	0.86	0.39	−0.15	0.38
Physical activity	−0.16	0.06	−2.54	0.01	−0.28	−0.04	0.1	0.07	1.41	0.16	−0.04	0.23	0.003	0.08	0.04	0.97	−0.16	0.16
Constant	10.94	1.42	7.71	<0.001	8.16	13.72	13.8	1.98	6.96	<0.001	9.91	17.69	21.45	2.81	7.62	<0.001	15.93	26.98

• Note: PROCESS procedure using W1 BMI (Y1), W2 BMI (Y2), and W3 BMI (Y3) as outcomes, including FTO rs9939609 (X) and breastfeeding (W) with sex, age at the time, and physical activity at the time as covariates. Significant results are reported in the bold font.
• Abbreviations: BMI, body mass index; Coeff, unstandardized coefficient; LLCI, lower level of confidence interval; ULCI, upper level of confidence interval.
• ^a Int_1a: Interaction term FTO rs9939609 × breastfeeding in months.

TABLE 3. Binary logistic regression models investigating the association between FTO, breastfeeding, and their interaction for overweight including obesity at W1

	Model 1		Model 2
	p	OR (95% CI)	p	OR (95% CI)
Sex (1)	0.02	0.69 (0.51–0.93)	0.01	0.69 (0.51–0.93)
Age	0.99	1 (0.87–1.15)	0.98	1 (0.86–1.15)
Physical activity	<0.001	0.85 (0.78–0.93)	<0.001	0.85 (0.78–0.93)
Breastfeeding	0.05		0.04
Breastfeeding (1)	0.02	0.67 (0.49–0.93)	0.07	0.6 (0.35–1.04)
Breastfeeding (2)	0.28	0.78 (0.49–1.23)	0.31	1.4 (0.73–2.66)
FTO rs9939609	0.94		0.87
FTO rs9939609 (1)	0.93	0.99 (0.71–1.37)	0.71	1.1 (0.68–1.79)
FTO rs9939609 (2)	0.78	1.07 (0.69–1.66)	0.63	1.17 (0.62–2.2)
FTO rs9939609 × breastfeeding			0.1
FTO rs9939609 (1) by breastfeeding (1)			0.66	1.18 (0.57–2.41)
FTO rs9939609 (1) by breastfeeding (2)			0.03	0.31 (0.11–0.87)
FTO rs9939609 (2) by breastfeeding (1)			0.7	1.2 (0.47–3.07)
FTO rs9939609 (2) by breastfeeding (2)			0.14	0.28 (0.05–1.51)
Constant	0.64	0.6	0.61	0.58

• Note: Binary logistic regression (N = 1221), method enter, using overweight including obesity at W1 as outcome, including in Model 1: sex (male = 0), age, breastfeeding 0–6 months (0)/7–12 months (1)/>12 months (2), FTO rs9939609 TT (0)/TA (1)/AA (2), W1 physical activity >30 min and in Model 2, additional interaction term FTO × breastfeeding. Significant results are reported in the bold font.
• Abbreviations: CI, confidence interval; OR, odds ratio.

4 DISCUSSION

This study indicates that rs9939609 AA is a plasticity variant, which is receptive to both negative and positive environmental conditions in relation to BMI. In this case, its plastic nature was demonstrated by the effect of parental reported breastfeeding duration on its relationship with BMI among the adolescents in this sample. Parental reported breastfeeding was also a moderator of the relationship between rs9939609 variants and overweight among the younger adolescents.

The Johnson–Neyman output revealed that rs9939609 AA individuals were more susceptible than AT and TT individuals for both short, associated with higher BMI, and long durations, associated with lower BMI, of breastfeeding at W1, which is in line with the differential susceptibility hypothesis. However, the limits were very short and very long breastfeeding duration, where the former likely can be associated with other factors that was not evaluated, and the latter is uncommon in general. For W2 BMI, only long breastfeeding duration was important. It is noteworthy that these values depend on the range of the moderator values that are included in the analyses. In another sample, with shorter or longer breastfeeding duration, the region of significance will likely be different. In agreement with Horta et al.⁴⁵ Wu et al.³⁸ Abarin et al. with findings on girls,³⁷ and Dedoussis et al. with findings on pre-schoolers³⁹ long breastfeeding duration was found to counteract the effect of the rs9939609 risk allele on BMI. However, Jiang et al. could not find any significant interaction effect in their study on Chinese participants.³⁶ Authors investigated a binary measurement of exclusive breastfeeding duration with a cutoff at 4 months and analyzed differences between TT and A carriers, that is, combined TA and AA into one group.³⁶ Because any specific biological event occurring at exactly 4 months cannot be recognized, the cutoffs that represent the breastfeeding recommendations were investigated. Variations in the methods might be the main reason for the discrepancies. Four of the mentioned studies^{36, 38, 39, 45} used a binary measurement of breastfeeding, which makes it impossible to detect regions of significance for any interaction effects. Moreover, Abarin et al.³⁷ found a decrease in the BMI of 0.12 and 0.18 kg/m² per additional month of breastfeeding in the AT and AA groups, respectively, which also indicates that AA is more susceptible than AT. If rs9939609 AA indeed is a plasticity variant as suspected, the main effects would vary depending on the non-measured factors among the study participants, as discussed previously regarding serotonin transporter polymorphisms.⁴⁶ The interaction effect found in this study was statistically significant for the younger adolescents but declined in the older adolescents and was absent for the young adults. This might in part be caused by lower power for W2 and W3 analyses due to a loss of participants at the follow-up. An addition of factors such as possible injuries that affect mobility and eating habits and that influence BMI throughout a person's life will also dilute the effect of breastfeeding. Some findings suggest an age-dependent relationship between the FTO rs9939609 variants and BMI.²⁴

Parental reports of longer breastfeeding duration were associated with lower risk for overweight among the younger adolescents. The findings indicate a protective effect from longer breastfeeding among the AA individuals, but the small sample of participants with overweight and breastfeeding for >12 months has to be considered. Wu et al.³⁸ showed that only AA participants with exclusive breastfeeding for <3 months had an upper quartile BMI value above the overweight line. Future studies with a larger sample size of individuals with long breastfeeding duration are warranted.

Only 13% of the participants had a breastfeeding duration >12 months. Data have shown that high-income countries have shorter breastfeeding duration than low- and middle-income countries.⁴⁷ Even in the latter, only 37% of the infants aged <6 months are exclusively breastfed as recommended.⁴⁷ Previous findings indicate that breastfeeding (partial or exclusive) is associated with a slower growth rate during infancy,^{38, 48} which is a major determinant of overweight and obesity later.³⁸

Given the present findings, it is plausible that rs9939609 A is a plasticity allele. Breastfeeding duration might be one small piece in the large puzzle that encompasses the increasing obesity trend. Indeed, there are many other aspects related to obesity, such as built environment, screen time, and agroalimentary environment,⁴⁹ that have not been considered in this study.

5 CONCLUSION

FTO rs9939609 AA might be a plasticity variant, with differential susceptibility to environmental influences. Breastfeeding duration may act as one of many moderators of the relationship between FTO rs9939609 and BMI among Caucasian adolescents. Future studies on this research topic should consider the differential susceptibility hypothesis.

CONFLICT OF INTEREST

Sofia H. Kanders, Kent W. Nilsson, and Cecilia Åslund have no conflict of interest.

AUTHOR CONTRIBUTIONS

All authors contributed to the study design. Kent W. Nilsson and Cecilia Åslund provided the data. Sofia H. Kanders analyzed the data. All authors interpreted the data. Sofia H. Kanders drafted the manuscript. All authors contributed to critical revisions and approved the final manuscript.