Sex differences in the association of sphingolipids with age in Dutch and South-Asian Surinamese living in Amsterdam, the Netherlands

Background Men have a higher risk for cardiovascular disease (CVD) early in life, while women have a higher risk later in life. The sex-related differences in CVD risk, especially by age, could be related to sphingolipid metabolism. We compared plasma sphingolipid concentrations and its increase by age in men and women. Methods Plasma concentrations of 13 types of sphingolipids were measured by liquid chromatography-tandem mass spectrometry in a random subsample of 328 men and 372 women of Dutch and South-Asian Surinamese ethnic origin, participating in the HELIUS study. Sphingolipid concentrations were compared between men and women by age group (18–39, 40–55, and 56–70 years). Multiple linear regression was used to determine sex differences in age trends in sphingolipids stratified by ethnicity. Analyses were performed without adjustment and adjusted for body mass index (BMI) and waist circumference. Results At age 18–39 years, sphingolipid concentrations were lower in women than those in men, but at age 56–70 years this was reversed. At higher age, women showed higher concentrations than men. In line, we observed a more rapid increase of sphingolipid concentrations by age in women than in men. The observed sex differences were not explained by BMI or waist circumference. Patterns of sex differences were similar across ethnic groups, although the strength of associations differed. Conclusions Mean sphingolipid concentrations increase more rapidly with age in women than in men. Therefore, plasma lipid concentrations of sphingolipids, although lower in women than in men at younger age, are higher in women than in men at older age. Supplementary Information The online version contains supplementary material available at 10.1186/s13293-020-00353-0.

135 anthropometric measures were taken in duplicate, and 136 the mean was used in the analyses. If the discrepancy be-137 tween the duplicate measures differed more than 0.5 cm 138 for height, more than 0.5 kg for weight, or more than 1 139 cm for waist circumference, a third measurement was 140 taken. The two measures which were most similar were 141 used to calculate the mean. 142 The total reported fat intake and total energy intake 143 were derived from an ethnic-specific food frequency 144 questionnaire (FFQ) which was taken among a sub-145 sample of the HELIUS cohort, as described in detail 146 elsewhere [33]. The FFQ data were available for 259 par-147 ticipants of our study sample, of whom 58 participants 148 were Dutch men, 47 South-Asian men, 67 Dutch 149 women, and 87 South-Asian women. Menopause was 150 derived from the questionnaire based on lack of men-151 struation for a year or longer (not for reasons such as 152 pregnancy, breastfeeding, or using birth control). 153 Blood was collected after a fasting period of at least 154 10 h. Sphingolipids were measured in plasma by liquid 155 chromatography-tandem mass spectrometry (LC-tMS) 156 as described previously [23]. We adjusted for amino 157 acids in sensitivity analyses. These were determined in 158 plasma by LC-tMS as described previously [34]. 159 Statistical analyses 160 First, the normal distribution of variables was checked by 161 plotting histograms and evaluating skewness and kurtosis. 162 Baseline characteristics and sphingolipid concentrations 163 were examined among men and women stratified by eth-164 nicity. We calculated means and standard deviations (SD) 165 for continuous normally distributed variables, medians, 166 and interquartile ranges for continuous non-normally dis-167 tributed variables and numbers of observations and per-168 centages for categorical variables. Baseline characteristics 169 were not tested for statistical differences [35]. Waist cir-170 cumference was missing for one participant and imputed 171 with an expectation-maximization algorithm. Sex differ-172 ences in sphingolipid concentrations stratified by age (cat-173 egories 18-39, 40-55, and 56-70 years) were studied by 174 multiple linear regression within each age group. 175 Second, we analyzed the association of metabolites with 176 age. We checked the linearity of the association by plotting 177 scatterplots in the total population and stratified by ethni-178 city. The multiplicative interaction of age with ethnicity was 179 checked by adding an interaction term between age and 180 ethnicity with sphingolipids as the outcome. This was done 181 because BMI may reflect different levels of intra-abdominal 182 fat storage in European than South-Asian populations [36], 183 which may also have implications for the use of non-184 oxidative pathways. A multiplicative interaction between 185 age and ethnicity was observed for five of the thirteen in-186 cluded sphingolipids (GlcCer(d18:2), GlcCer(d18:1), Lac-187 Cer(d18:2), LacCer(d18:1), and Cer(d18:1)). Analyses were 188 thus stratified by ethnicity in all analyses. A multiplicative 189 interaction term between age and sex was used to investi-190 gate whether the association between sphingolipids and age 191 differed by sex.

192
All models were run both unadjusted and adjusted for 193 measures of body fat distribution (BMI and waist cir-194 cumference). We adjusted for cholesterol-and blood 195 pressure-lowering medication in sensitivity analyses, as 196 the use may affect sphingolipid concentrations [29,37]. 197 We also checked whether the amount of substrate avail-  Finally, we excluded participants with CVD at baseline.

203
In post-hoc analyses, we adjusted for menopause.

204
All analyses were conducted using IBM SPSS Statistics 205 23. Graphs were plotted in RStudio version 3.6.1 using 206 the visreg package. Tests were two-sided, and p values < 207 0.05 were considered statistically significant. Analyses 208 were not adjusted for multiple testing as our study was 209 of exploratory nature [38], but the consistency of find-210 ings was considered to avoid chance findings.

212
Mean age in the 18-39 year group ranged from 28.4 (SD  (Table   T1 1). Mean waist circum-222 ference was lowest among Dutch women aged 18-39 223 years old with a mean of 79.5 (SD 9.0) and highest in 55-224 70-year-old men with a mean of 97.6 (SD 11.3).

225
Dihydroceramide (Cer(d20:1), Cer(d18:2), Cer(d18:1), 226 Cer(d18:0), Cer(d17:1), and Cer(d16:1)) concentrations 227 were generally lower in women than in men in the 18-39 228 years age group, although mostly not statistically signifi-229 cantly different (Table   T2 2 Cer(d18:1), CTH(d18:1), and CTH(d20:1)) showed similar 239 patterns, but were already higher in women in the 18− 39 240 years age groups in the South-Asian Surinamese with a 241 further increase in the difference with men in older age 242 groups. Patterns of sex differences in mean sphingolipid 243 concentrations remained similar after adjustment for BMI 244 and waist circumference.   Table   T3 3). Cer(d18:1) for instance in-247 creased with 52.28 nmol/L (95% CI 26.56; 78.00) per year 248 in Dutch men. However, no clear trends were observed 249 for CTH(d18:1) and LacCer(d18:1). Most plasma con-250 centrations of sphingolipids increased more with age in 251 women than in men, although only statistically signifi-252 cantly differed for GlcCer(d18:2), LacCer(d18:2), and 253 Cer(d18:2). Figure 1 shows that plasma concentrations 254 of sphingolipids are generally lower in young adult 255 women than in men, but higher in women than in men         296 and showed that ceramide and dihydroceramides con-297 centrations were higher in women than in men and in-298 creased more strongly in women than in men by age 299 [30]. Although the study by Mielke et al. was limited to 300 participants over 55 years of age [30], this finding is in 301 line with our study. We added to these findings that the 302 slope of the association is such that in younger age 303 groups sphingolipid concentrations may be higher 304 among men than women. Moreover, we are the first to 305 report on sex differences in 1-deoxyceramides, 306 glucosylceramides, and globotriaosylceramides, for 307 which patterns of sex differences and age trends were 308 similar to the dihydroceramides.

309
The higher levels of sphingolipids in (especially older) 310 women when compared to men may partly be explained   mation of plaques, pro-inflammatory properties, and f1:1 Fig. 1 Sphingolipids which concentrations increase faster by age in women than men. The sphingolipids are shown of which the plasma f1:2 concentrations increase faster by age in women than in men. A significant interaction by sex means a statistically significant multiplicative f1:3 interaction between age and sex with sphingolipid concentrations as the outcome at a P value < 0.05. Analyses were stratified by ethnicity. Red f1:4 asterisk denotes statistically significant association between age and sphingolipid in women at a P value < 0.05. Blue asterisk denotes statistically f1:5 significant association between age and sphingolipid in men a P value < 0.05 f1:6 t3:1 lyses that also excluded participants with baseline CVD, 399 also more prevalent among those of South-Asian descent 400 than in the majority Dutch, did not affect our results.

401
Next, the cross-sectional design of our study is a limita-402 tion. We did not follow participants over time, but 403 cross-sectionally grouped our study population by age.

404
The results may thus reflect a cohort effect. Characteris-

530
Plasma ceramides predict cardiovascular death in patients with stable 531 coronary artery disease and acute coronary syndromes beyond LDL-