Associations between outdoor temperature and bright sunlight with metabolites in two population-based European cohorts

Context: Outdoor temperature and bright sunlight may directly and/or indirectly modulate systemic metabolism. Objective: We assessed the associations between outdoor temperature and bright sunlight duration with metabolomics. Design: meta-analysis of two cross-sectional studies. Setting: Two population-based European cohort studies. Patients or other participants: Non-diabetic individuals from the Oxford BioBank (OBB; N=6,368; mean age 47.0 years, males 44%) and the Netherlands Epidemiology of Obesity (NEO; N=5,916; mean age 55.6 years, males 43%) studies. Intervention(s): Data on mean outdoor bright sunlight and temperature collected from local weather stations in the week prior to blood sampling. Main Outcome Measure(s): Serum levels of 148 metabolites measured using NMR spectroscopy, including 14 lipoprotein subclasses. Statistical analyses: Multivariable linear regression analyses adjusted for age, sex, body mass index, season and either outdoor temperature or bright sunlight. Summary statistics from the OBB and NEO cohorts were combined using fixed-effect meta-analyses. Results: A higher mean outdoor temperature was associated with increased concentrations of lipoprotein (sub)particles and certain amino acids such as phenylalanine and leucine. In contrast, longer mean hours of bright sunlight were specifically associated with lower concentrations of very low density lipoprotein (sub)particles. The direction of effects was consistent between the OBB and NEO, although effect sizes were generally larger in the OBB. Conclusions: Increased bright sunlight duration is associated with an improved metabolic profile whilst higher outdoor temperature may adversely impact cardiometabolic health.


Introduction
Outdoor temperature and sunlight intensity and duration affect our daily activities and may consequently have an impact on metabolic health status. Human and animal studies have also highlighted that environmental temperature and sunlight exposure may directly impact systemic metabolism e.g. by modulating brown adipose tissue (BAT) activity and influencing circadian rhythms.
[1] Consistent with these notions several epidemiological studies have identified associations between outdoor temperature and prevalence of type 2 diabetes mellitus (T2D) and cardiometabolic diseases. [2][3][4] For example, Blauw et al. showed that the incidence of diabetes was greater in U.S. states with a higher average annual temperature. [4] Similarly, by investigating the associations between outdoor weather conditions and metabolic traits in two European population-based cohorts, namely the Oxford BioBank (OBB) and Netherlands Epidemiology of Obesity (NEO), we have recently shown that increased bright sunlight exposure was associated with a lower Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) and reduced triglyceride levels. [5] In contrast, no associations between mean outdoor temperature and glucose or lipid metabolism were detected. These results indicate that outdoor sunlight is specifically associated with a more beneficial cardiometabolic risk profile, although further studies are necessary to confirm these findings and determine the underlying biological mechanism(s) underpinning these associations.
Recently, high-throughput metabolite profiling has emerged as a powerful tool for the exploration of disease mechanisms and the identification of novel therapeutic targets [6][7][8][9], especially in the context of (cardio)metabolic disorders. [10,11] Based on our earlier study [5], we hypothesized that outdoor bright sunlight would be associated with a more favourable metabolite profile. In this study, we performed a cross-sectional analysis investigating the association between outdoor bright sunlight and environmental temperature 5 with plasma levels of 148 lipids and metabolites as determined by NMR spectroscopy in a combined sample of over 12,000 middle-aged population-based subjects, without pre-existing diabetes mellitus, from the OBB and NEO study cohorts. 6

Study design
The OBB is a population-based cohort of randomly selected healthy participants aged 30 to 50 years from Oxfordshire (UK). Individuals with a history of myocardial infarction, diabetes mellitus, heart failure, untreated malignancy, other ongoing systemic diseases or ongoing pregnancy were not eligible for study inclusion. Participants were included between 1999 and May 2015. The OBB cohort comprises 7,185 individuals (4,054 women and 3,131 men). A more detailed description of the study recruitment criteria and population characteristics is reported elsewhere. [12] The NEO study is population-based prospective cohort study of men and women aged between 45 and 65 years with an oversampling of individuals with a BMI of 27 kg/m 2 or higher, living in the greater area of Leiden (in the West of the Netherlands). In addition, all inhabitants aged between 45 and 65 years from one municipality (Leiderdorp) were invited in the study irrespective of their BMI, to allow for a reference distribution of BMI. Between September 2008 and September 2012, 6,671 individuals (3,505 women and 3,168 men) were included in the study. Detailed information about the study design and data collection has been described previously [13].
In both cohorts, participants were invited for a detailed baseline assessment, conducted after an overnight fast, which included blood sampling and anthropometry. Both studies were approved by local ethics committees, and written informed consent was obtained from all study participants.

Study population
In the OBB, we excluded individuals with missing data on mean outdoor temperature and/or bright sunlight in the week preceding the centre visit, body composition, and fasting 7 metabolomics (missing in 817 individuals). Consequently, data from 6,368 individuals were used for the present analyses. From NEO, we excluded individuals with both treated and diagnosed diabetes, as well as subjects with a fasting glucose concentration above 7.0 mmol/L (N = 749) in order to have a uniform population as that of the OBB regarding glycaemic status. Additionally, we excluded participants who were nonfasting, or had missing data on mean outdoor temperature and/or bright sunlight in the week preceding the centre visit, body composition, and fasting metabolomics (missing in 6 individuals). As a result, a total of 5,916 individuals were used for the analyses presented in this study.

Data collection on outdoor temperature and bright sunlight
Data on the mean temperature and hours of bright sunlight (defined as global radiation >120 W/m 2 ) were collected from the weather station that was located closest located to either Oxfordshire or Leiden. Based on these data, we estimated the mean outdoor temperature and bright light over the week prior to the date of the blood sampling. For the OBB data were obtained from the Radcliffe Meteorological Station (Woodstock Road, Oxford, UK). For the NEO study we obtained data from a measurement station from the Koninklijk Nederlands Meteorologisch Instituut (Royal Dutch Meteorological Institute).

NMR-based metabolic biomarker profiling
We used a high-throughput proton NMR metabolomics platform [14]

Covariates
Height and weight were measured by research nurses at the OBB and NEO study centres.
BMI was calculated by dividing the weight in kilograms by the height in meters squared.
Season was derived from the date of the blood sampling (winter: December -February, spring: March -May, summer: June -August, autumn: September -November). In both cohorts, use of lipid-lowering medication was determined by medication inventory.

Statistical analysis
In the NEO study, participants with a BMI of 27 kg/m 2 or higher are oversampled. To correctly represent associations for the general population [16], we corrected for oversampling of participants with a BMI ≥ 27 kg/m 2 , which was done by weighting individuals towards the BMI distribution of participants from the Leiderdorp municipality [17], whose BMI distribution was similar to the BMI distribution of the general As most of the metabolic outcome variables were not normally distributed, we logtransformed all these variables prior to standardization to a standard normal distribution (mean = 0, s.d. = 1) to be able to better compare the effect sizes of the different study outcomes. Associations of mean bright sunlight and temperature with lipid and metabolite concentrations were examined using multivariable linear regression analyses for the combined population of the OBB and NEO study populations. Estimates retrieved from the analyses were subsequently meta-analysed using fixed-effects meta-analysis as implemented in the rmeta statistical package in R. For presentation purposes, we analysed the data per 5 degrees Celsius increase in outdoor temperature and per hour increase in bright sunlight exposure. Consequently, results can be interpreted as the difference in standard deviation per 1 0 unit increase in either outdoor temperature (5 degrees Celsius) and bright sunlight exposure (1 hour).
To study the impact of adjustment of several of the covariates in the multivariable linear regression analyses, we considered 3 different statistical models. Model 1 was adjusted for age, sex and BMI (given our earlier observation that the weather exposures were associated with BMI). [5] Model 2 was additionally adjusted for season. Model 3 was additionally adjusted either for the mean temperature or mean hours of bright sunlight to fully dissect the two weather exposures in the present study, which were moderately correlated with each other. [5] In a sensitivity analysis, we additionally excluded individuals who used cholesterol-lowering treatment. To test the consistency of the results between the OBB and the NEO cohorts, a plot was constructed with beta estimates of the identified metabolites in both cohorts visualized against each other for outdoor temperature and bright sunlight, using the R-package ggplot2. [19] Given the high number of statistical tests performed in the present study, we corrected for multiple testing. The metabolic biomarkers used for the present study are correlated with each other, and therefore, conventional correction for multiple testing (e.g., Bonferroni) is too stringent. To obtain the number of independent metabolic biomarkers, we used the method as described by Li et al., [20] which takes the correlation between the different metabolic biomarkers into account. Based on this method, we found 37 independent metabolic markers. For this reason, associations were considered to be statistically significant in case the p-value was below 0.00134 (i.e. 0.05/37).

Study population characteristics
The total study population (N = 12,284) comprised of 6,368 individuals from the OBB and 5,916 individuals from the NEO study (see Table 1). Compared to participants from the NEO cohort, OBB volunteers were younger (mean age 47.0 vs. 55.6 years) and had a lower mean BMI (25.9 vs 26.1 kg/m 2 ). Mean outdoor temperature during the week prior to blood sampling was similar between the two cohorts (10.7 degrees Celsius in both cohorts) although mean hours of bright sunlight were higher in Leiden than in Oxfordshire (3.8 vs 5.0 hours).

Outdoor temperature and serum metabolites
A high-throughput proton NMR metabolomics platform [14] was utilized to quantify 148 lipid and metabolite concentrations in serum samples obtained following an overnight fast.
Associations between outdoor temperature and circulating levels of these metabolites following adjustment for age, sex, BMI, season and hours of bright sunlight in the combined study population are presented in Figure 1. In total we identified 27 metabolites whose concentration was associated with outdoor temperature after correction for multiple testing (p< 1.34e -3 ). For example, a higher mean outdoor temperature in the week preceding blood sampling was associated with an increased concentration of total cholesterol (β (SE) = 0.0642 (0.0184) SD per 5 degrees Celsius, p=5.03e -4 ). Lipoprotein subfraction analysis revealed that higher outdoor temperature was most strongly associated with raised serum very small VLDL

Bright sunlight and serum metabolites
Associations between outdoor bright sunlight and NMR spectroscopy measured metabolites following adjustment for age, sex, BMI, season and outdoor temperature in the combined OBB and NEO population are presented in Indeed, bright sunlight duration was generally inversely associated with the circulating levels of all the amino acids quantified by the NMR metabolomics platform utilised although these associations were not significant after correction for multiple testing. Figure 4, with the notable exception of leucine, glucose, glycoprotein, acetyl and total phosphoglyceride levels the effects of bright sunlight on the plasma metabolome were directionally identical in the OBB and NEO datasets.

Discussion
We compiled data from two population-based European cohorts, comprising a combined sample size of more than 12,000 non-diabetic individuals, to investigate the associations between mean outdoor temperature and bright sunlight and serum metabolites. Our results suggest that whilst higher outdoor temperature is associated with an unfavourable metabolic profile [21-23], prolonged exposure to bright sunlight is in contrast characterized by lower plasma lipoprotein concentrations and potentially decreased circulating levels of BCAAs.
Our findings build on our previous study in which we reported that bright sunlight, but not outdoor temperature was associated with enhanced glucose and lipid metabolism, as determined by a lower HOMA-IR and decreased plasma triglycerides. [5] Extending these findings we now show that prolonged mean hours of bright sunlight are correlated with lower levels of serum cholesterol, ApoB and the branched-chain amino acids (BCAAs) isoleucine, leucine and valine. The latter result has to be taken with caution as the effect sizes of the associations between bright sunlight and BCAAs differed considerably between the OBB and the NEO cohorts, and had opposite directions in case of leucine. Furthermore, we demonstrate that the association between bright sunshine and lower serum cholesterol concentration was primarily driven by a reduction in VLDL particle numbers, reflecting either decreased VLDL production or increased VLDL clearance.
[24] Consistent with our earlier study [5] a negative association was also detected between bright sunlight and plasma triglycerides (p-value<0.05) although this did not survive correction for multiple testing.
Notably, with a few exceptions, associations between outdoor temperature and circulating metabolites were directionally opposite to those detected with bright sunlight, suggesting that increased environmental temperature is associated with an unfavourable cardiometabolic profile. In this respect, higher circulating concentrations of BCAAs and aromatic amino acids, total VLDL, total cholesterol and ApoB have been associated with an increased risk of A particular strength of our study is the combination of two large independent cohorts originating from different countries and comprising more than 12,000 participants to study the effects of temperature and bright sunlight on metabolism. A limitation is the inherent heterogeneity in the two cohorts, which has to be taken into account in interpreting the results. These differences might have been driven by differences in lifestyle between the United Kingdom and The Netherlands. Unfortunately, detailed data on lifestyle such as habitual food intake and physical activity were not available in both cohorts. Consequently, we are also not able to investigate whether these lifestyle factors potentially mediated the association between either outdoor temperature and/or bright sunlight and metabolite concentrations. Nevertheless, additional adjustment for season, which would partially adjust for variations in lifestyle factors, did not materially alter our findings.

7
In summary, extending our earlier work [5] we provide further evidence that increased bright sunlight is associated with improved cardiometabolic health. In contrast increased outdoor temperature may be associated with an adverse cardiometabolic profile.
Future research is required to elucidate the mechanistic basis of the observed associations.   Supplementary Table 1.
Results of the NEO study population are weighted towards the body mass distribution of the general population.