Walking in the light: How history of physical activity, sunlight, and vitamin D account for body fat - a UK Biobank study

Abstract

Objective:

High prevalence of vitamin D deficiency and obesity drive the need for successful strategies that elevate vitamin D levels, prevent adipogenesis and stimulate lipolysis. We provide a theoretical model to evaluate how physical activity and sunlight exposure influence serum vitamin D levels and regional adiposity. We hypothesized posteriori that sunlight is associated with undifferentiated visceral adiposity by increasing the ratio of brown to white adipose tissue.

Methods:

Using ten-year longitudinal data, accelerometry, a sun exposure questionnaire, and regional adiposity quantified by DEXA imaging, a structural equation mediation model of growth curves was constructed with data-driven methodology.

Results:

Sunlight and physical activity conjointly increased serum vitamin D. Changes in vitamin D levels partially mediated how sunlight and physical activity impacted adiposity in visceral and subcutaneous regions within a subjective physical activity model. In an objective physical activity model, vitamin D was a mediator for subcutaneous regions only. Interestingly, sunlight was associated with less adiposity in subcutaneous regions, yet greater adiposity in visceral regions.

Conclusions:

Sunlight and physical activity may increase vitamin D levels. For the first time, we characterize a positive association between sunlight and visceral adiposity. Further investigation and experimentation are necessary to clarify the physiological role of sunlight exposure on adipose tissue.

INTRODUCTION

Despite a century of advancements in life expectancy, longevity in the U.S. has been on the decline since 2014 (1). Both an aging baby-boomer generation and an alarming acceleration in mortality among adults in midlife are responsible for this trend (2). While the nature of this bleak trajectory is multifactorial, the prevalence of obesity has consistently risen since the early 1990’s and has been shown to have a profoundly detrimental effect, not only on the prevalence of cardiovascular diseases but also for that of neurological disorders (3, 4).

Vitamin D serum levels are thought to operate as a central control mechanism of body composition, where lower systemic concentrations stimulate adipogenesis. This would, in turn, maximize survival in energy-scarce seasons including winter (5). For decades, however, vitamin D levels among infants and toddlers (6), adolescents (7) and adults (8) have remained chronically low in the general population. This chronic, multi-generational deficiency in vitamin D could compound or in part give rise to obesity and its sequelae, including gut inflammation, hyperglycemia, and progressive deficits in immuno-metabolic signaling in the brain that hinder cognitive function (9).

Physical activity has a myriad of positive health effects, including aiding in the reduction of adipose tissue volume while maintaining or increasing muscle mass (10, 11). Recent studies further suggest that engagement in greater levels of physical activity is associated with higher circulating vitamin D levels (12), even from excise done exclusively indoors (13). Ultraviolet radiation from sunlight exposure increases endogenous production of vitamin D with subsequent reduction in adiposity (14). In experimental mice models, exposure to summer-related UV wavelengths has been shown to enhance lipolysis of white adipose tissue and activate brown adipose tissue deposition (15).

Physical activity frequently coincides with increased sunlight exposure; however, it is difficult to disentangle independent and additive effects of these two variables on abdominal adipose mass. Although cross-sectional studies have observed greater mean levels of serum vitamin D with outdoor activities in contrast to indoor activities (16), it is important to examine further how vitamin D levels, physical activity, and sunlight exposure interact with one another to influence body fat composition. Few studies to-date have conjointly estimated the extent to which accelerometer-based physical activity and long-term sun exposure raise vitamin D levels and impact obesity from mid-to-late adulthood. It is also critical to understand these relationships for individuals located in the northern and southernmost regions of the world, where little UV-induced vitamin D occurs outside of the summer months, as well as in aged adults who synthesize vitamin D less efficiently compared to younger counterparts (17).

Our objectives in the present study were (1) to establish how physical activity and sunlight exposure are conjointly related to abdominal fat mass using dual energy x-ray absorptiometry (DEXA), as well as (2) characterize how serum vitamin D may underlie and describe these associations. We also considered subcutaneous fat mass and compared questionnaire-and accelerometry-based models. Potentially confounding dietary variables were controlled in order to ensure that effects from physical activity and sunlight were not stemming from correlations between these variables and diet.

METHODS

Cohort

Participants were a part of the UK Biobank study (18). This prospective cohort study collected baseline data in a half million individuals from 22 assessment centers located in the United Kingdom north of the 53° latitude, starting in 2006. Each participant had baseline measurements taken between 2006 and 2010, when genetic, behavioral, and biological data were collected. A visit to the assessment center involved six consecutive steps: 1) consent, 2) touchscreen questionnaire, 3) verbal interview, 4) eye measures, 5) physical measures and 6) blood/urine sample collection. The touchscreen questionnaire collected sociodemographic, occupation, lifestyle, early life exposure, cognitive function, and family history of illness data. Informed consent to participate was given at baseline. Longitudinal assessments are ongoing in a subset of participants. The UK Biobank protocol was approved by the North West MultiCentre Research Ethics Committee. Due to power and generalizability considerations, only subjects of European ancestry were considered. As noted in Figure S1, a total sample of 1,853 participants were available for self-reported assessment of physical activity (i.e. the subjective model), and a subset of 1,353 participants for accelerometry-based assessment of physical activity (i.e., the objective model). As noted in Figure S2, participants were aged 48 to 80 years old at the completion of this study.

Measures

A timeline of all measures and their assessment dates is illustrated in Figure S3.

Body Composition

A subset of participants had body composition imaging data collected in 2015–2016. Compartment measurements of body composition – lean muscle mass (LMM) in kilograms (kg), subcutaneous adipose mass (SAM) in kg, visceral adipose mass (VAM) in kg, and bone density in g/cm² – were determined by a trained radiographer delivering a five-minute, full-body DEXA (General Electric Lunar iDXA, Madison, WI) to each participant while they laid supine (19). To enhance the value of DEXA as a prospective outcome measure, imaging started in May 2014 after collection of the lifestyle data.

Subjective Physical Activity Levels

Subjective physical activity was assessed using adapted questions from the validated short International Physical Activity Questionnaire (IPAQ), (20) which covered the frequency, intensity and duration of moderate and vigorous activity. Values were quantified in mean minutes/day. Data processing rules published by IPAQ were followed (21).

Objective Physical Activity Levels

Invitations were mailed to 236,519 eligible participants to participate in the sub-study of objective physical activity (PA). Participants in the northwest region who had been involved in other sub-studies were not invited for accelerometer measurement due to potential participant burden. A total of 1,422 participants had complete data. Eligible participants were mailed a tri-axial accelerometer (Axivity AX3, Newcastle upon Tyne, UK), which was set to capture three-dimensional acceleration at 100 Hz with a dynamic range of ±8g in order to quantify PA in m/s². Instructions on proper use of the accelerometer were provided. Consenting participants wore the device on their dominant wrist for seven consecutive days. Device programming automatically started and stopped recording at predefined times. A pre-paid envelope was provided to return the equipment after use. Non-wear was identified as time periods of > 60 minutes where standard deviations of all three axes were < 13.0 milli-g (1 milli-g=0.001 milli-g). Participants with < 72 h of wear time (n=69) were excluded from the analyses. Of the remaining 1,353, the device was worn for an average of 6.69 days. To describe the overall level and distribution of PA intensity, the sample level data was combined into five-second epochs for summary data analysis, maintaining the mean vector magnitude value over the epoch. To represent the distribution of time spent by an individual in different levels of PA intensity, an empirical cumulative distribution function from all available 5-second epochs was generated (22). Data processing has been described in detail elsewhere (23).

Sunlight Exposure and Sun Protection

Participants answered the question, “In a typical day in summer, how many hours do you spend outdoors?” as part of the touch-screen questionnaire at three separate occasions. Responses were recorded as integers between 0 and 12. As there are approximately six hours maximum of meaningful UV radiation at the peak of summer in the UK, values above 6 were set to 6. Participants also answered the question, “Do you wear sun protection (e.g. sunscreen lotion, hat) when you spend time outdoors in the summer?” Responses were recorded as one of four ordinal categories (“Never/rarely”, “Sometimes”, “Most of the time”, “Always”).

Serum Biomarker Levels

At two separate visits (2006–2010 and 2012–2013), plasma samples were collected in 4mL EDTA vacutainers and analyzed within 24 hours of sampling utilizing four Beckman Coulter LH750 instruments (Sheard et al., 2017). Serum samples were analyzed for levels of vitamin D (nmol/L) and 24 other biomarkers, as listed in Text S1. All serum biomarkers were tested in a backwards elimination approach to reduce bias (see statistical analysis section). Vitamin D serum levels were corrected for seasonal effects based on the time of the year. The serum samples were collected in spring, summer, autumn, or winter.

Dietary and Alcohol Consumption

To ensure that associations of physical activity and sunlight were not confounded by correlations with diet, we covaried several aspects of dietary composition. This analysis focused on total dietary composition, rather than specific nutrients like protein, crude fiber, or moisture. Participants completed a Food Frequency Questionnaire (24), including 18 questions about commonly eaten food groups, as part of the touch-screen questionnaire at three separate assessments. More information is available in Text S2.

Covariates

Covariates included sex, age, education, socioeconomic status, and tobacco smoking. Age was measured in years at baseline. A categorical variable was used to capture education level at baseline. Education categories were considered sequentially and included the following: College or other higher-level qualification; post-secondary or vocational; secondary; or none of the previous education levels listed. Socioeconomic status was considered sequentially and based on the participant’s average total household income between 2006 and 2014. Responses were recorded as one of five ordinal categories in British Pounds (“Less than 18,000”, “18,000 to 30,999”, “31,000 to 51,999”, “52,000 to 100,000”, “Greater than 100,000”), where the lowest two categories were classified as lower class, the next two categories were classified as middle class, and the greatest category was classified as upper class. Tobacco smoking indicated who has never smoked, used to smoke, and who is currently a smoker.

Statistics

Longitudinal Modeling

For longitudinally observed variables, using difference equations we computed individual, across-time averages and non-linear changes over time to enhance model fit (25). We then integrated these values to derive average levels and the sum of changes for each longitudinally assessed variable (26). Serum biomarkers were computed as average individual levels over two visits spanning four years. As we have recently demonstrated (27), this method produces superior goodness-of-fit, while increasing testing power and elucidating relationships between variables more robustly by capturing both within-(28) and between-subject variation over time (29, 30).

Outlier Analysis

In order to ensure that our models were generalizable to at least 99.9% of the sample population, 0.1% quantiles were computed, and 68 participants beyond 99.9% of the sample distribution of the mean among any variable were removed from further analysis.

Structural Equation Modeling

Structural equation modelling (SEM) was done using R 3.4.1 (31). Graphs were prepared in ggplot2 3.1.1 (32). SEM was initially used to determine model fit of two separate physical activity constructs: subjective activity using self-report, and objective activity use accelerometry data. SEM-based mediation has more statistical power than the standard regression procedure (33). SEM has the additional benefit of easily extending to longitudinal data within a single framework (30), as done in this report.

Variable Selection

An empirical model-building approach was employed to select the most salient variables that predicted serum biomarker levels and visceral or subcutaneous adipose outcomes. In this backward elimination approach, a full, “all variables in” model was constructed, and the least significant variable was removed one at a time. The model was then recomputed until all variables remaining reached p<.050.

Parameter estimation, ANOVA, uncertainty analysis, and mediation

To establish predictors that significantly explained regional adiposity outcomes, a structural equation model was used to comprehensively fit the covariance structure (34) of regional adiposity with vitamin D and 24 other available serum biomarkers, physical activity, sunlight exposure, and self-report consumption of whole foods in the diet. See Figure 1 for a conceptual representation of the model. Standardized parameter estimates (β) were computed using maximum likelihood and interpreted as the mean potential effects of a variable. ANOVA is reported as the overall portion of variation in visceral or subcutaneous adiposity that is explained (R²). Uncertainty analysis relied upon standard errors and p-values, where results were considered significant at p<.001 (***), p<.01 (**), p<.05 (*) and trending at p<.10 (#). Mediation tested if serum biomarker levels due to variation in physical activity and sunlight exposure explained visceral or subcutaneous adiposity associations. Specifically, parameter decomposition was used to distinguish indirect (λ) from direct (β) effects (35). In order to maintain an empirical data-driven analysis and ensure robustness, only participants with no data missingness were considered.

Figure 1.

Top panel: A conceptual diagram of associations and their directionality between physical activity and sunlight exposure, serum vitamin D, and regional adipose tissue mass outcomes. Model is shown as exposure -> mediator -> outcome. Lower panels: These graphical equations algebraically summarize the models illustrated in Figures 2, 3, and 4. Because vitamin D is an outcome that is also nested within the prediction of regional adiposity, the equations that describes regional adiposity are computed sequentially after the vitamin D-related equation. The symbol Δ₁ is defined as the average level and totality of changes observed over time, and Δ₂ is defined as average level over time. Each coefficient represents the change expected in the outcome if an individual progressed halfway (1.5 standard deviations) along the distribution of that predictor. To offer context for each coefficient, the range of each outcome is listed below the left-hand side of the equation.

Post-hoc Analyses

To assess whether any observations were driving the results, Cook’s Distance was calculated for each observation in the Subjective PA and Objective PA models, and a threshold value of 1.0 was set. The largest value in the Subjective PA model was 0.022, and in the Objective PA model we observed 0.024. As no values >1.0 were noted, we made no further adjustments to the models.

Sensitivity Analysis

Given the sample size of 1,853, 30 predictors (the most observed here), and no a priori hypotheses (“two-tails”), GPower 3.1 estimated the smallest effect which could be detected, while minimizing type I error to α=.05, and type II error to β=.20, β=.10, β=.05, or β=.01, the size of effect was f=0.004, f=0.006, f=0.007, or f=0.010, respectively.

RESULTS

Data Summary

Demographics and summary data are listed in Table 1 and Table S1. As illustrated in Figure 1, two equivalent structural equation models were determined to link subjective or objective physical activity and sun exposure measures with regional adiposity. These models examined associations with (1) visceral and subcutaneous fat, (2) sunlight exposure and physical activity, (3) vitamin D and (4) 24 serum biomarkers currently available in UK Biobank. In the first model, subjective, questionnaire-based physical activity was assessed. In the second model, objective, accelerometer-based physical activity was assessed.

Table 1.

Demographics and Data Summary – Participants with Subjective PA

Data, measurement unit	Women	Men
Sample size (n)	761	1,092
Age (y)	63 ± 7.4	65 ± 7.2
Education Level (number and sample %)
College/other higher level	523 (67.9%)	775 (70%)
Post-secondary/vocational	108 (14.0%)	196 (17.7%)
Secondary	116 (15.0%)	73 (6.6%)
Other	23 (2.9%)	60 (5.4%)
Social Class (number and sample %)
Lower	331 (42.9%)	429 (38.8%)
Middle	396 (51.4%)	618 (55.9%)
Upper	43 (5.5%)	57 (5.1%)
Muscle Mass (kg)	39 ± 4.5	55 ± 6.2
Visceral Adipose Mass (kg)	0.67 ± 0.51	1.59 ± 0.09
Subcutaneous Adipose Mass (kg)	24 ± 8.1	22 ± 7.4
Bone Mineral Density (g/cm)	1.13 ± 0.11	1.30 ± 0.11
Serum Vitamin D (nmol/L)	53.4 ± 19.5	53.3 ± 19.4
Subjective Moderate Physical Activity (m/d)	56 ± 47.9	59.6 ± 53.4
Subjective Vigorous Physical Activity (m/d)	38 ± 29.5	41 ± 33.6
Sunlight Exposure (h/d)	3.3 ±1.6	3.8 ± 1.7

The values of this demographics table are Mean ± SD unless stated otherwise. Years is abbreviated y, kilograms is abbreviated kg, grams is abbreviated g, centimeters is abbreviated cm, nanomoles is abbreviated nmol, liters is abbreviated L, minutes is abbreviated m, hours is abbreviated h, and days are abbreviated d.

Subjective Physical Activity Model: Physical Activity and Adiposity Mass

This model focused on questionnaire-based, subjective total (or moderate / vigorous) physical activity, or “Subjective PA” (Table 2, Figure 2). Here, aging was associated with more visceral fat mass (β=.049, p=.009), while increasing vitamin D levels over time corresponded with less fat mass (β_c=−.051, p=.001). Regarding subcutaneous adiposity (Table 2, Figure 3), aging was associated with less fat mass (β=−.076, p=.001) and so were higher vitamin D levels over time (β_c=−.078, p<.001).

Table 2.

Subjective Physical Activity Model

	Mechanism of Action	Effect Size	% of Total Effect
1. Visceral Adiposity ~ Vigorous PA		βtotal=−055***
	(1) Unspecified:	βa=−.039**	71%
	(2) ALT levels:	γ=−.016**	29%
2. Subcutaneous Adiposity ~ Vigorous PA		βtotal=−.074***
	(1) Unspecified:	βa=−.056***	76%
	(2) ALT levels:	γ=−.017**	23%
	(3) Vitamin D levels:	γb,c=−.004*	5%
	(4) GGT levels:	γ=.002*	3%
3. Visceral Adiposity ~ Sunlight Exposure		βtotal=.023#
	(1) Unspecified:	βe=.029*	126%
	(2) Vitamin D levels:	γd,c=−.006**	26%
4. Subcutaneous Adiposity ~ Sunlight Exposure		βtotal=−.009***
	(1) Vitamin D levels:	γd,c=−.009***	100%
5. Subcutaneous Adiposity ~ Moderate PA		βtotal=008*
	(1) Creatinine	γ=.008*	100%

All models are shown as Outcome ~ Exposure. The percentage of effect for the mediation effects (γ) are with respect to the total effect that they compose. If mediation effects for one total oppose each other, percentages will not necessarily equal 100. The table does not include non-significant γeffects. PA is abbreviated to stand for physical activity. Totals do not include non-significant γpathways. Alanine aminotransferase is abbreviated ALT and γ-glutamyltransferase is abbreviated GGT.

Figure 2.

This snippet of the subjective structural equation model illustrates which factors explain subcutaneous adipose mass. The model examined: (1) metabolic biomarkers of the mechanisms of action which explain associations between physical activity and sunlight exposure; and (2) the remaining, unspecified effects (also called ‘direct effects’) of physical activity and sunlight exposure. The Delta symbol (Δ) is defined as the average level and totality of changes in that variable observed over 6 years. The standardized β reflects the average effect size of each path and each path is denoted with a subscript. Each λ reflects the mediation effect resulting from the path analysis and is subscripted to illustrate the paths which compose it. p<.001=***, p<.010 =**, and p<.050=*. The scatterplots illustrate intercept and slope among women (in red) and men (in blue), and the grey band represents the conditional standard error of the mean.

Figure 3.

This snippet of the subjective structural equation model illustrates which factors explain visceral adipose mass. The model examined: (1) metabolic biomarkers of the mechanisms of action which explain associations between physical activity and sunlight exposure; and (2) the remaining, unspecified effects (also called ‘direct effects’) of physical activity and sunlight exposure. The Delta symbol (Δ) is defined as the average level and totality of changes in that variable observed over 6 years. The standardized β reflects the average effect size of each path and each path is denoted with a subscript. Each λ reflects the mediation effect resulting from the path analysis and is subscripted to illustrate the paths which compose it. p<.001=***, p<.010=**, and p<.050=*. The scatterplots illustrate intercept and slope among women (in red) and men (in blue), and the grey band represents the conditional standard error of the mean.

For vigorous Subjective PA, higher self-reported levels were associated with less visceral fat (p=.009). Serum biomarkers were able to account for 34% of the variance in this model. Specifically, vitamin D levels mediated 5% of Subjective PA’s effect on visceral adiposity (p=.052), due to more Subjective PA predicting more vitamin D (β_b=.054, p=.019). Higher levels of the liver enzyme alanine aminotransferase (ALT) separately explained of 28% the variance in this model.

In contrast, greater vigorous Subjective PA predicted less subcutaneous fat (p=.001) (Figure 3). Vitamin D levels explained 5% of this effect (p=.026).

Subjective Physical Activity Model: Sunlight Exposure and Adiposity Mass

For hours per day spent in sunlight, greater exposure was associated with more visceral fat mass (p=.014). Due to more sun exposure being related to higher vitamin D levels (p<.001), vitamin D reduced by the association between greater sunlight and more visceral adiposity by 26% (p=.014). No other serum markers were found to be statistically significant.

In contrast, more sunlight exposure was related to less subcutaneous fat mass (p=.001), where vitamin D levels accounted for 100% (p=.001) of the effects of sunlight on subcutaneous fat mass.

Objective Physical Activity Model: Physical Activity and Adiposity Mass

This model focused on accelerometry-based, objective physical activity, or “Objective PA” (Table 3). For visceral adiposity, aging was not a significant covariate. Higher levels of Objective PA strongly coincided with less visceral fat (p<.001). Many serum markers, particularly lipid fraction and transport proteins, accounted for 35% of the variance in this model.

Table 3.

Objective Physical Activity Model

	Mechanism of Action	Effect Size	% of Total Effect
1. Visceral Adiposity ~ Physical Activity		βtotal=−.234***
	(1) Unspecified:	β=−.155***	66%
	(2) HDL levels:	γ=−.052***	22%
	(3) ApoA levels:	γ=.031***	13%
	(4) Urate levels:	γ=−.017***	7%
	(5) ALT levels:	γ=−.016**	7%
	(6) Cystatin C levels:	γ=−.012***	5%
	(7) SHBG levels:	γ=.010**	4%
	(8) Triglycerides:	γ=−.004*	2%
2. Visceral Adiposity ~ Sunlight Exposure		βtotal=.032*
	(1) Unspecified:	β=.032*	100%
3. Subcutaneous Adiposity ~ Physical Activity		βtotal=−.367***
	(1) Unspecified:	βa=−.264***	72%
	(2) HDL levels:	γ=−.043***	12%
	(3) ApoA levels:	γ=.025***	7%
	(4) SHBG levels:	γ=−.022***	6%
	(5) Cystatin C levels:	γ=−.021***	6%
	(6) Urate levels:	γ=−.017***	5%
	(7) ALT levels:	γ=−.014**	4%
	(8) CRP levels:	γ=−.008**	2%
	(9) Vitamin D levels:	γb,c=−.007*	2%
	(10) GGT levels:	γ=.004*	1%
4. Subcutaneous Adiposity ~ Sunlight Exposure		βtotal=−.004*
	(1) Vitamin D levels:	γd,c=−.004*	100%

All models are shown as Outcome ~ Exposure. The percentage of effect for the mediation effects (γ) are with respect to the total effect that they compose. If mediation effects for one total oppose each other, percentages will not necessarily equal 100. The table does not include non-significant γeffects. PA is abbreviated to stand for physical activity. Totals do not include non-significant γpathways. High density lipoprotein is abbreviated HDL, apolipoprotein A is abbreviated ApoA, alanine aminotransferase is abbreviated ALT, sex hormone binding globulin is abbreviated SHBG, c-reactive protein is abbreviated CRP, and γ-glutamyltransferase is abbreviated GGT.

For subcutaneous adiposity (Table 3, Figure 4), less fat mass was observed with age in years (β=−.131, p<.001). Higher levels of Objective PA (p<.001) and increased vitamin D (β_c=−.048, p=.006) were related to less subcutaneous fat. The negative effect of physical activity on subcutaneous fat (β_b=.152, p<.001) was partially mediated by its impact on vitamin D (λ_b,c=−.007, p=.014).

Figure 4.

This snippet of the objective structural equation model illustrates which factors explain subcutaneous adipose mass. The model examined: (1) metabolic biomarkers of the mechanisms of action which explain associations between physical activity and sunlight exposure; and (2) the remaining, unspecified effects (also called ‘direct effects’) of physical activity and sunlight exposure. The Delta symbol (Δ) is defined as the average level and totality of changes in that variable observed over 6 years. The standardized β reflects the average effect size of each path and each path is denoted with a subscript. Each λ reflects the mediation effect resulting from the path analysis and is subscripted to illustrate the paths which compose it. p<.001=***, p<.010=**, and p<.050=*. The scatterplots illustrate intercept and slope among women (in red) and men (in blue), and the grey band represents the conditional standard error of the mean.

Objective Physical Activity Model: Sunlight Exposure and Adiposity Mass

More sunlight exposure predicted more visceral fat mass (β_total=.032, p=.018). Unlike the Subjective PA model, vitamin D levels did not mediate the relationship with visceral fat among our sample.

Conversely, more sun exposure predicted less subcutaneous fat (β_total=−.004, p=.046). Vitamin D levels fully mediated the effects from sunlight (λ_d,c=−.004, p=.046) due to the impact it had on serum levels (β_d=.080, p=.004).

DISCUSSION

Typical changes in physique associated with aging, such as greater visceral fat, are associated with many negative health outcomes, like all-cause mortality, metabolic syndrome and type II diabetes, cardiovascular disease, sarcopenia, and osteoporosis (36). Regular physical activity helps in preventing or minimizing weight gain and obesity (11). While sunlight exposure often coincides with physical activity, much less is known about its potential health effects (37). To the best of our knowledge, few studies have examined the extent to which accelerometer-based physical activity and history of sun exposure affect vitamin D levels and impact regional adiposity from mid-to-late adulthood, particularly at latitudes further from the equator. Briefly, we observed that higher levels of physical activity consistently corresponded to decreased adiposity. Interestingly, sunlight exposure was related to less subcutaneous fat, but more visceral fat, which may be due to compensatory behaviors such as feeling more hunger after a workout and present weight status (38) or differential impact of sunlight on the mass of different adipose tissue types (5, 15).

Our analysis of metabolic biomarkers suggests that physical activity and sunlight exposure are associated with regional adiposity partially due to serum vitamin D levels. Vitamin D levels may regulate body composition, where adipogenesis is stimulated during times when levels are depleted (5). Similar to Wanner et al. (12), we observed that serum vitamin D levels responded significantly from both physical activity and sunlight exposure. Although sunlight exposure had comparatively twice the effect on increasing vitamin D, importantly, physical activity further influences vitamin D above and beyond sun exposure that often coincides with it.

Physical activity levels may independently increase serum vitamin D levels. Our results showed that physical activity was associated with greater vitamin D levels, which remained significant even after considering sunlight exposure in our analysis. One study showed that each hourly increase in the level of moderate to vigorous physical activity engaged in per week was associated with a 1.54 ng/mL increase in 25(OH)D concentration, which is the predominant circulating, active form of vitamin D in the body and, consequently, the most common biomarker utilized to measure serum vitamin D levels (39). Some studies have shown that serum vitamin D levels were increased in more physically active individuals compared to sedentary individuals, regardless of whether the physical activity was done indoors or outdoors (12, 16). The association between greater physical activity and increased serum vitamin D levels independent of sunlight exposure could be elucidated by prior findings of increased plasma concentrations of vitamin D subsequent to engagement in physical activity. Physical activity has been found to alter serum nutrient concentrations that relate to vitamin D status including alterations in serum phosphate and ionized calcium levels (40, 41). This suggests that physical activity may directly elevate circulating vitamin D metabolites in the blood (42).

Furthermore, we observed that liver enzymes and lipid transport proteins mediated a significant portion of the effect of physical activity. A 2015 meta-analysis by Shephard and Johnson (43) stated that “the effect of physical activity on circulating aminotransferases is unclear”. In our model, which examined long-term effects and not acute changes, two liver enzymes levels, alanine transaminase (ALT) and gamma-glutamyl transferase (GGT), were lower when physical activity was greater, and higher levels of ALT was the strongest biomarker predictor of greater fat mass in visceral regions. ALT recycles nitrogen and carbon, particularly from muscle, while GGT catalyzes the synthesis of glutathione and plays a role in xenobiotic detoxification. Lawlor et al. (44) similarly reported in 2005 that physical activity decreased levels of ALT. Liver function has previously been associated with the distribution of regional fat (45). Consistent with earlier studies, we also found that changes in the lipid transport proteins apolipoprotein A and high-density lipoproteins (HDL) cholesterol from physical activity (46, 47) were partially responsible for reductions in adipose tissue.

Unexpected and discordant relationships between sunlight exposure and regional adipose distribution were detected. The results from our model are consistent with previous work from Kim et al. (48) who proposed that UV light reduces lipid synthesis in subcutaneous fat by altering the transcription of lipogenic enzymes and the production of cytokines within keratinocytes and fibroblasts. However, our model is the first to characterize a positive relationship between sunlight exposure and visceral adiposity. It is surprising that contrasting associations would be detected between regions, yet we consider Leitner et al. (49) who showed that lean men had greater brown adipose tissue than men who were obese, and that regions of visceral fat housed on average 70% of total brown adipose tissue. Comparatively, subcutaneous fat regions may contain as little as 10–15% of the total brown adipose tissue mass. Nayak et al. (15) demonstrated in mouse models that adipocytes contain Opsin-3 receptors which convert white adipose tissue into brown adipose tissue in response to summer UV wavelengths. Additionally, Foss discussed the hypothesis that vitamin D levels act as a central regulator of body fat composition by keeping the body informed about the time of year (5). Thus, there may be separate photic-and vitamin D-related effects. Given prior findings and our data-driven results, we propose that sunlight was associated with undifferentiated visceral adiposity by increasing the ratio of brown to white adipose tissue in that region. If sunlight exposure and higher levels of serum vitamin D from physical activity could truly enhance adaptive thermogenesis, the implications here could be significant for the obesity epidemic because brown adipose tissue is associated with leaner body composition (50).

A comparison of the Subjective PA and Objective PA models showed important differences. Whereas the Subjective PA model showed that vitamin D levels were related to both visceral and subcutaneous adiposity, the Objective PA model only showed an association with subcutaneous adiposity. The effect of sunlight on vitamin D levels was stronger than effects from physical activity in the Subjective PA model. This relationship was then reversed when accelerometry was used in the Objective PA model. Lastly, several new relationships between physical activity and serum biomarkers were observed with accelerometry measures.

Our study had strengths to note. We examined quantified measures of adiposity in visceral and subcutaneous compartments by DEXA, rather than relying on less accurate anthropometric measures like Body Mass Index. We furthermore covaried the effect of dietary intake, which can confound associations with physical activity and vitamin D levels. We also compared Subjective vs. Objective PA, to determine the degree to which results did or did not agree between self-report data, which is often inaccurate compared to objective accelerometer data. Potential mechanisms of action via vitamin D were examined in this study and differed depending on visceral vs. subcutaneous adipose mass.

Several limitations of the study should also be noted. While serum vitamin D and PA levels were obtained longitudinally, DEXA-based body morphometry was cross-sectional. It is therefore unclear if changes-over-time in PA, sunlight exposure, or vitamin D correspond to changes-over-time in adipose mass. Consequently, we are unable to show the order of cause-to-effect relationships between the variables in our analyses. DEXA cannot distinguish between white vs. brown adipose mass, which may explain discrepant findings with sunlight exposure. Exposure to sunlight was measured subjectively in a recall questionnaire and would be more accurately measured using objective measurements, such as dosimetry and surface area of exposed skin. We furthermore acknowledge that hours spent outdoors during the summer are more so a proxy of sunlight exposure, for instance, because of time spent in shade. However, the estimates of sunlight exposure in our study gives additional insight into the body mass-sunlight relationship and provides framework for future studies among human subjects to expand upon our findings. The mediational effects of vitamin D were in some cases modest, despite testing for it and 24 other serum biomarkers currently present in UK Biobank. In this analysis, it was not possible to eliminate participants who were taking vitamin D supplementation. It is possible that participants may have been taking unreported occasional or habitual vitamin D supplements during the times of data collection, which may account for some vitamin D-linked variation not explained by our model.

Despite these limitations, our data provide evidence and a theoretical model for the physiological roles of physical activity and sunlight in the aging body and describe vitamin D’s biological mechanisms of action. Physical activity levels and sunlight exposure conjointly influenced serum vitamin D levels in our sample population. Interestingly, although our results showed that sunlight exposure was associated with increased visceral adiposity, our data suggests that vitamin D may provide protection against visceral adiposity with greater exposure to sunlight. Future studies that can utilize objective measures of sunlight exposure may be more sensitive to the detection of the influences of exposure on the amount and distribution of adipose tissue. Additional future directions include examining whether sunlight influences the brown vs. white ratio of adipose tissue in both visceral and subcutaneous fat masses.

Supplementary Material

1

Figure S1. A flow chart diagram indicating the progressive exclusion of participants to derive a final sub-sample with no missingness among data of interest.

Figure S2. A histogram illustrating the distribution of age among women (in red) and men (in blue).

Figure S3. A timeline illustrating when each time of data was observed.

STUDY IMPORTANCE.

• Obesity and vitamin D deficiency are prevalent among most age cohorts.

• Both physical activity and sunlight exposure are independently linked to increased vitamin D and reduced obesity.

• When tested conjointly, physical activity was associated with elevated vitamin D levels above and beyond the effects of sunlight.

• Greater sunlight exposure was associated with more visceral fat mass, but not subcutaneous fat mass.

• Mediation results suggest separate photic and vitamin D-related effects on adipose tissue from exposure to sunlight.

• Due to correlations between sunlight and physical activity, it may be best to consider them together in future studies.

• If there are separate vitamin D and photic effects, supplementing vitamin D alone may not adequately replace effects from sunlight exposure.