Plasma ceramides relate to mild cognitive impairment in middle-aged men: The Maastricht Study

Abstract

Introduction

There is an urgent need for biomarkers identifying individuals at risk of early-stage cognitive impairment. Using cross-sectional data from The Maastricht Study, this study included 197 individuals with mild cognitive impairment (MCI) and 200 cognitively unimpaired individuals aged 40 to 75, matched by age, sex, and educational level.

Methods

We assessed the association of plasma sphingolipid and ceramide transfer protein (CERT) levels with MCI and adjusted for potentially confounding risk factors. Furthermore, the relationship of plasma sphingolipids and CERTs with magnetic resonance imaging brain volumes was assessed and age- and sex-stratified analyses were performed.

Results

Associations of plasma ceramide species C18:0 and C24:1 and combined plasma ceramide chain lengths (ceramide risk score) with MCI were moderated by sex, but not by age, and higher levels were associated with MCI in men. No associations were found among women. In addition, higher levels of ceramide C20:0, C22:0, and C24:1, but not the ceramide risk score, were associated with larger volume of the hippocampus after controlling for covariates, independent of MCI. Although higher plasma ceramide C18:0 was related to higher plasma CERT levels, no association of CERT levels was found with MCI or brain volumes.

Discussion

Our results warrant further analysis of plasma ceramides as potential markers for MCI in middle-aged men. In contrast to previous studies, no associations of plasma sphingolipids with MCI or brain volumes were found in women, independent of age. These results highlight the importance of accounting for sex- and age-related factors when examining sphingolipid and CERT metabolism related to cognitive function.

BACKGROUND

Lipid metabolism plays a critical role in the development and progression of neurodegenerative diseases like Alzheimer's disease (AD) and has been linked to early dementia and mild cognitive impairment (MCI).¹ Sphingolipids, including ceramides, are particularly enriched in the central nervous system (CNS). These lipids have essential structural roles but also function as second messengers to modulate a wide variety of signaling events, including cellular differentiation and proliferation, apoptosis, cytokine production, and synaptic plasticity.² Ceramide consists of an N-acylated fatty acid attached to a sphingoid base of varying chain lengths, including long-chain (C16 and C18) and very long (> C22) acyl chain lengths, which can be either saturated or monounsaturated.³ The sphingoid backbone is generally comprised of sphingosine (d18:1) or its saturated analog sphinganine (d18:0). Although it has been suggested that ceramides with different acyl chain lengths and saturation levels have unique physiological actions,³ the exact role of specific ceramide moieties in diseases of the nervous system remains unclear. Nevertheless, lipidomic, metabolomic, and targeted approaches have identified pathways of lipids and enzymes of the sphingolipid metabolism that are altered early in the course of AD and contribute to the neuropathological alterations associated with AD.^1,4–6 Sphingolipid metabolites like sphingosine-1-phosphate (S1P) have additionally been implicated in neuroinflammation and neurodegeneration.¹ Along these lines, previous longitudinal studies examining blood sphingolipids reported that high levels of multiple ceramide species predicted cognitive impairment among cognitively unimpaired individuals,⁷ as well as memory decline, hippocampal volume loss,⁸ and faster rates of cognitive decline among AD patients.⁹ Although this suggests the potential use of blood sphingolipids as predictors of cognitive decline, studies focusing on blood sphingolipids in younger, middle-aged, age groups remain scarce. Additionally, it remains unclear to what extent the relationship of sphingolipids with cognitive decline is influenced by potentially mediating, moderating, or confounding factors like sex, ethnicity, cardiovascular risk factors, endothelial dysfunction, inflammation, and medication use.

Ceramides are produced in the endoplasmic reticulum of neurons and glia, and transported from the endoplasmic reticulum to the trans-Golgi regions of these cells by ceramide transporter proteins (CERTs).⁵ CERTs are ubiquitously expressed in the CNS, where they are involved in brain development and homeostasis and transport ceramide to membranes via its steroidogenic acute regulatory protein (StAR)-related lipid transfer domain.^2,5,10,11 Specifically, they are known to transfer ceramide C14, C16, C18, and C20 chains. In the brain, the long isoform (CERTL) containing an additional 26-amino acid domain has been identified in AD to partially co-localize with serum amyloid P component and amyloid beta in plaques.¹² In a transgenic mouse model of AD, CERTL levels are reduced in the AD brain and overexpression of CERTL in neurons affects plasma ceramide C16 levels.⁵ However, the relationship of plasma CERTs with levels of specific ceramide chain lengths in the blood is currently unknown, as well as the associations among plasma CERT levels, cognition, and brain atrophy.

In this study, we aimed to determine whether plasma sphingolipid and CERT levels differed between middle-aged individuals with MCI and cognitively unimpaired individuals. Individuals that were selected for plasma sphingolipid and CERT analyses were matched on age, sex, and educational level. Because cardiovascular risk factors like body mass index (BMI), hypertension, and endothelial dysfunction,^1,13–16 as well as inflammation and medication use^1,4,14 have been associated with the regulation of lipid levels, the effects of these factors were also evaluated. In addition, the association between plasma ceramide and CERT levels with hippocampal and cerebrospinal fluid (CSF) volumes, and white matter hyperintensities (WMHs) was investigated. Finally, because AD and MCI are age-related diseases and sex differences have been reported in both the rate of cognitive decline¹⁷ and plasma sphingolipid concentrations during aging,¹⁸ potential age and sex interaction effects on the association of sphingolipids and CERTs with MCI were examined.

MATERIALS AND METHODS Study population

We used data from The Maastricht Study, an observational, prospective population-based cohort study. The rationale and methodology have been described previously.¹⁹ In brief, the study focuses on the etiology, pathophysiology, complications, and comorbidities of type 2 diabetes mellitus (T2DM), and it is characterized by an extensive phenotyping approach. Eligible for participation were all individuals aged between 40 and 75 years and living in the southern part of the Netherlands. Participants were recruited through mass media campaigns and from the municipal registries and the regional Diabetes Patient Registry via mailings. Recruitment was stratified according to known T2DM status, with an oversampling of individuals with T2DM, for reasons of efficiency. The present report includes cross-sectional data from the first 3451 participants who completed the baseline survey between November 2010 and September 2013. To select individuals for plasma sphingolipid and CERT analyses, 200 cases of MCI were identified and matched to cognitively unimpaired individuals based on sex, age, education level (low/middle/high),¹⁹ and available sample material, in a nested case–control setting. This resulted in 197 cases of MCI compared to 201 cognitively unimpaired individuals (Table 1). The examinations of each participant described above were performed within a time window of 3 months. Magnetic resonance imaging (MRI) measurements were implemented from December 2013 onward until February 2017. Of the 398 participants with plasma ceramide assays, 259 (65%) participants (127 MCI, and 132 cognitively unimpaired) had available MRI data. The lag time between MRI and assessment of neurological symptoms was 102 to 120 days. The study has been approved by the institutional medical ethical committee (NL31329.068.10) and the Ministry of Health, Welfare and Sport of the Netherlands (permit 131088-105234-PG). All participants gave written informed consent.

TABLE 1 General participant characteristics.

	CU (n = 200)	MCI (n = 197)
Characteristic	Mean (SD)	Mean (SD)	P-value
Demographics
Age (years)	60 (8.8)	60 (8.8)	0.839
Sex (% female)	47.5 (—)	47.8 (—)	0.966
Educational level—low/medium/high (%)	39.5/24.0/36.5 (—)	38.6/24.9/36.5 (—)	0.912
Smoking status—never/former/current (%)	33.5/51.0/15.5 (—)	35.5/46.2/18.3 (—)	0.916
Cardiovascular risk factors
Diabetes (%)	8.5 (—)	15.7 (—)	0.027
Body mass index (kg/m2)	25.8 (4.2)	27.0 (4.4)	0.006
Hypertension (%)	51.2 (0.5)	57.9 (0.5)	0.202
Markers of endothelial dysfunction
Soluble intercellular adhesion molecule-1 (ng/mL)	339.5 (79.2)	354.0 (92.1)	0.093
Soluble vascular cell adhesion molecule-1 (ng/mL)	416.0 (92.7)	424.9 (92.3)	0.339
E-Selectin (ng/mL)	106.7 (53.5)	122.1 (91.5)	0.041
Von Willebrand factor	128.1 (40)	132.0 (51.1)	0.401
Markers of inflammation
Serum amyloid A (μg/mL)	4.5 (6.6)	6.8 (17.2)	0.081
Human interleukin-6 (pg/mL)	0.8 (1.1)	0.9 (1.1)	0.227
Human interleukin-8 (pg/mL)	4.8 (3.6)	4.8 (2.8)	0.828
Human tumor necrosis factor alpha (pg/mL)	2.2 (0.5)	2.3 (0.6)	0.021
Medication use
Glucose-lowering medication (%)	6.5 (—)	14.2 (—)	0.011
Blood pressure lowering medication (%)	33.0 (—)	24.1 (—)	0.061
Lipid-modifying medication (%)	28.0 (—)	35.0 (—)	0.133
Cognitive function
Memory function	0.23 (0.86)	−0.92 (0.93)	0.000
Information processing speed	0.00 (0.68)	−0.48 (0.76)	0.000
Executive function & attention	0.14 (0.75)	−0.57 (0.88)	0.000
Overall cognitive functioning	0.12 (0.6)	−0.64 (0.56)	0.000
	CU (n = 132)	MCI (n = 127)
Brain volumes (%)
White matter hyperintensity volume	0.09 (0.27)	0.11 (0.34)	0.615
Hippocampus volume (average left/right)	0.55 (0.05)	0.55 (0.06)	0.823
Cerebrospinal fluid volume	17.70 (2.35)	18.54 (2.82)	0.071

Abbreviations: CU, cognitively unimpaired; MCI, mild cognitive impairment; SD, standard deviation. Bold: p < 0.05

Assessment of mild cognitive impairment

Cognitive impairment was assessed as described previously.¹⁹ In short, a 30-minute neuropsychological test battery consisting of six cognitive tests was used. For conceptual clarity, test scores were standardized and divided into three cognitive domains (i.e., memory function, information processing speed, and executive function and attention), as described previously.¹⁹ Tests included the Visual Verbal Word Learning Task,²⁰ the Stroop Color-Word Interference Test,²¹ the Concept Shifting Test,²² the Letter-Digit Substitution Test,²³ the Verbal Fluency Test,²⁴ and the Mini-Mental State Examination (MMSE).²⁵ Raw scores from each test (except MMSE) were transformed into a standardized z score. Z scores from tests included in each compound performance index were then averaged. When a cognitive disorder was suspected (i.e., MMSE < 24 or > 1/6 of the core tests described above in clinical range; ← 1.5 standard deviations below their norm-based expected score in any domain), participants were determined to have MCI.

RESEARCH IN CONTEXT

Literature Review: A literature review was conducted using the Google Scholar and PubMed databases to evaluate the research concerning the role of plasma sphingolipid dysregulation in early dementia and the potential of plasma sphingolipids and ceramide transfer proteins (CERTs) as (sex-specific) biomarkers. Limited research has explored sex differences in cross-sectional lipidomic studies related to cognitive decline and mild cognitive impairment (MCI) at a relatively early age.

Interpretation: This study investigated plasma sphingolipids and CERT levels in relation to MCI and brain volumes. The findings revealed a significant interaction of sex, with higher ceramide levels associated with MCI in middle-aged men but not women. These findings emphasize the importance of considering sex as a significant moderator in lipid–MCI associations.

Future Directions: Further research is needed to validate these findings and elucidate the underlying mechanisms. Exploring the potential of lipids as sex-specific biomarkers opens new avenues for early biomarker discovery, not only in MCI or Alzheimer's disease, but also in other forms of dementia. Understanding the implications of sex differences in lipidomic profiles may lead to improved diagnostic and therapeutic approaches.

Plasma extraction

Plasma was collected in the morning in a fasting state and immediately stored at the biobank of Maastricht University at −80°C. The plasma samples were put through no more than two freeze–thaw cycles before experimental use.

Brain imaging

Brain images were acquired on a 3T clinical magnetic resonance scanner (MAGNETOM Prismafit, Siemens Healthineers GmbH) located at a dedicated scanning facility (Scannexus, Maastricht, The Netherlands) using a head/neck coil with 64 elements for parallel imaging. The MRI protocol included a three-dimensional (3D) T1-weighted (T1w) magnetization prepared rapid acquisition gradient echo (MPRAGE) sequence (repetition time/inversion time/echo time [TR/TI/TE] 2300/900/2.98 ms, 176 slices, 256 × 240 matrix size, 1.0 mm³ reconstructed voxel size); and a fluid-attenuated inversion recovery (FLAIR) sequence (TR/TI/TE 5000/1800/394 ms, 176 slices, 512 × 512 matrix size, 0.49 × 0.49 × 1.0 mm reconstructed voxel size). The protocols for MRI acquisition and analysis are in line with current Standards for Reporting Vascular Changes on Neuroimaging 2 (STRIVE-2) imaging standards.²⁶ To assess white matter, gray matter, and CSF volumes, T1 weighted images and FLAIR images were analyzed using an ISO-13485:2021 certified, automated method (which included visual inspection).²⁷ Intracranial volume was calculated as the sum of white matter, gray matter, and CSF, for which the WMHs and volumes of the hippocampus were standardized. Furthermore, brain segmentation was performed with FreeSurfer v6.0²⁸ as described previously²⁹ to extract CSF, WMHs, and left/right volumes of the hippocampus, using T1w and FLAIR images as input, which were then averaged. The optional arguments “FLAIRpial” and “3T” were used to optimize segmentation quality. In addition, segmentations underwent quality control through visual inspection.²⁹

Plasma sphingolipid measurements

Plasma ceramide analyses were performed using ultra performance liquid chromatography (UPLC) mass spectrometry at the Metabolomics Core of the Mayo Clinic, Rochester, Minnesota, USA. Ceramides were extracted from 100 uL of plasma as described previously⁹ after the addition of internal standards. The extracts were measured against a standard curve on a Thermo TSQ Quantum Ultra mass spectrometer. The following carbon chain length species were used in the analyses: C8:0, C14:0, C16:0, C18:0, C18:1, C20:0, C22:0, C24:0, and C24:1. Furthermore, three additional sphingolipids (sphingosine, sphinganine, and S1P) were measured.

Plasma CERT measurements

Plasma CERT levels were measured by enzyme-linked immunosorbent assay (ELISA) at the School for Mental Health and Neuroscience at Maastricht University, the Netherlands. CERTs were measured in plasma, and recombinant CERT proteins were used as a standard to assess protein concentration as described previously⁶ (inter-assay coefficient of variability [CV; n = 10] = average of high and low control CV = 12.9%; intra-assay CV [n = 397] = average % CV = 7.56%). In brief, ELISA plates were coated overnight at 4°C with polyclonal antibody (pAb) rabbit anti-CERT (Rb01) in 50 μl of coating buffer containing 50 mM sodium21 carbonate, pH 9.6. After washing five times with 0.05% Tween-20 in phosphate-buffered saline (PBS) and blocking with 4% non-fat dry milk powder, plates were incubated in sample incubation buffer (10% bovine serum albumin [BSA], 0.02% Tween20 in PBS, pH 7.4) for 1 hour at 37°C containing the plasma samples and serial dilutions of known concentrations of human recombinant CERTs produced in-house as a standard as described previously.¹² Subsequently, the wells were washed and incubated with a biotinylated second pAb rabbit anti-CERT (Rb02) in incubation buffer (1% BSA, 0.02% Tween20 in PBS, pH 7.4), followed by incubation with horseradish peroxidase-conjugated streptavidin. Absorption was measured in a microplate reader @450 nm (Perkin Elmer). Assays were sensitive to measure > 10 ng/mL of CERTs in plasma.

Covariates

Because The Maastricht Study is oversampled for T2DM, this variable was included as a covariate in the analyses. Diabetes was assessed by an oral glucose tolerance test.³⁰ Furthermore, the potential effects of cardiovascular risk factors (BMI; hypertension, for which office blood pressure was measured¹⁹) and markers of endothelial dysfunction related to cognitive decline (soluble intercellular adhesion molecule-1, soluble vascular cell adhesion molecule, e-Selectin, von Willebrand factor),³¹ low-grade inflammation (serum amyloid A, interleukin-6, interleukin-8, tumor necrosis factor alpha [TNF-α]), and medication use (glucose-, lipid-modifying–, and blood pressure-lowering medication) were assessed by a medication interview in which dose, frequency, and generic name were registered¹⁹ (see Table 1).

Statistical analyses

First, a normality check was performed on the data from the sphingolipid and CERT measurements using the Shapiro–Wilk method. Data from one cognitively unimpaired participant was eliminated from further analysis because no plasma lipid measurement data was available. Independent samples t tests were performed to compare participant characteristics by MCI status. To standardize the b-coefficients and compare the blood lipids, z scores were calculated for sphingolipid and CERT data and included covariates. Z scores of individual attributes related to markers of endothelial dysfunction, low-grade inflammation, and medication use were then combined into their class using a weighted average.

The cross-sectional associations of the blood sphingolipids and CERTs (determinant) and MCI (outcome) were assessed by multivariable logistic regression models (model 1) reported as odds ratios (ORs) and 95% confidence intervals (CIs). The variables sex, age, and educational level were included as covariates in Model 2 together with T2DM when non-matched variables were analyzed (i.e., plasma sphingolipids with CERT level associations [Figure S1 in supporting information], MRI brain volume analyses, and age- and sex-stratified associations). To identify additional potential confounders, we univariately assessed the association of each covariate with each sphingolipid species and MCI. The following factors were found to be potential confounders and their combined z scores were included in Model 3, along with the variables in Model 2: cardiovascular risk factors, endothelial dysfunction and inflammatory markers, and medication use.

A ceramide risk score was calculated as described, based on the included ceramide species (i.e., C16:0, C18:0, C20:0, C22:0, C24:0, and C24:1) and their ratios to C24:0 (i.e., C16:0/24:0, C18:0/24:0, C20:0/24:0, C22:0/224:0, and C24:1/24:0), resulting in a ceramide risk score of 0 to 22. Because AD and MCI are age-related diseases and sex differences in plasma sphingolipids have been reported,^32–34 an interaction term of age and sex by sphingolipid level was also included. When the interaction term was significant at trend level (P < 0.1),³⁵ analyses were stratified by age or sex.

Linear regression analyses were performed to assess the associations between plasma sphingolipid or CERT levels with neuroimaging volumes (CSF and WMH volumes corrected for intracranial volume, and volumes of the averaged left/right hippocampus corrected for total brain volume), adjusted for MCI status. Because the brain volume analyses were performed separately from the association of plasma lipid or CERT protein levels with MCI in a subgroup of individuals with neuroimaging volumes, model 3 was not included. However, because sex, age, educational level, and T2DM were non-matched for the subgroup of participants, these covariates were included in the regression model. In additional analyses, a receiver operating characteristic (ROC) curve and the area under the ROC curve (AUC) were generated to assess the diagnostic values of the ceramide risk score.

The assumptions of both the linear and logistic regression models were checked. All analyses were performed using IBM SPSS Statistics for Windows, version 26.0 (IBM Corp). According to the Rothman principle,³⁶ to minimize the chance of type II errors of the interpretations, no multiple-testing corrections are reported. Nevertheless, a Benjamini–Hochberg analysis of the multivariable-adjusted associations of plasma ceramide species and CERT levels with MCI was performed (RStudio, Build 492) to assess which significant associations (P < 0.05) remain robust after reduction of the false discovery rate (FDR; Table S1 in supporting information).

RESULTS General characteristics of the participants

The participant characteristics are shown by MCI status in Table 1. Matched covariates sex, age, and education level were similar between cognitively unimpaired and MCI together with smoking status. As expected, participants with MCI displayed significantly lower measurements of memory function, information processing speed, executive functioning, and overall cognitive functioning. Additionally, compared to cognitively unimpaired individuals, those with MCI had a higher mean BMI and waist circumference, higher plasma levels of eSelectin and TNF-α, and were prescribed a greater number of medications.

Interestingly, equality testing of survival distributions to analyze the cumulative incidence of MCI for men and women showed a higher risk of MCI for women around middle age (X² [1, N = 197[ = 12.2, P < 0.0005; Figure 1).

FIGURE 1. The cumulative incidence of MCI for men and women. Sex differences in the risk of MCI are seen primarily around middle age. MCI, mild cognitive impairment

Additionally, higher plasma ceramide C18:0 levels were significantly associated with higher CERT levels (β = 0.10, P = 0.044), which remained significant after controlling for age, sex, educational level, T2DM (β = 0.13, P = 0.016), and the addition of covariates of cardiovascular risk factors, inflammatory markers, endothelial markers, medication use (β = 0.12, P = 0.021). No significant associations were found between other plasma ceramide chain lengths and CERT levels.

Associations of plasma sphingolipids and CERTs with mild cognitive impairment

Because plasma concentrations of the sphingolipids sphinganine, sphingosine, ceramide C8:0, C14:0, and C18:1 were below detection level, data from these lipid species were eliminated from the analyses. Each standard deviation increase in S1P (OR = 1.26, 95% CI [1.02, 1.54]) and ceramide species C18:0 (OR = 1.27, 95% CI [1.04, 1.56]) and C20:0 (OR = 1.23, 95% CI [1.00, 1.50]) was associated with increased odds of MCI (Table 2a). The addition of the covariates cardiovascular risk factors, inflammatory markers, and medication use attenuated the results, although the ORs changed only slightly. A higher ceramide risk score was not associated with higher odds of MCI (Table 2b). Similarly, plasma CERT levels were not associated with MCI (Table 2c).

TABLE 2 Multivariable-adjusted associations of plasma ceramide species and CERT levels with MCI.

a.
	OR	(95% CI)	P-value
Model
S1P
Model 1	1.26	(1.02; 1.54)	0.030
Model 2	1.26	(1.02; 1.55)	0.030
Model 3	1.21	(0.98; 1.49)	0.081
Ceramide C16:0
Model 1	1.16	(0.95; 1.41)	0.153
Model 2	1.18	(0.96; 1.44)	0.113
Model 3	1.22	(0.98; 1.51)	0.072
Ceramide C18:0
Model 1	1.27	(1.04; 1.56)	0.019
Model 2	1.23	(1.00; 1.51)	0.054
Model 3	1.18	(0.95; 1.46)	0.132
Ceramide C20:0
Model 1	1.23	(1.00; 1.50)	0.046
Model 2	1.20	(0.98; 1.47)	0.079
Model 3	1.18	(0.95; 1.45)	0.130
Ceramide C22:0
Model 1	1.11	(0.90; 1.35)	0.327
Model 2	1.11	(0.91; 1.36)	0.307
Model 3	1.11	(0.90; 1.37)	0.326
Ceramide C24:0
Model 1	1.09	(0.89; 1.33)	0.403
Model 2	1.14	(0.93; 1.39)	0.218
Model 3	1.19	(0.96; 1.47)	0.108
Ceramide C24:1
Model 1	1.20	(0.98; 1.47)	0.076
Model 2	1.20	(0.98; 1.47)	0.072
Model 3	1.23	(0.99; 1.52)	0.057
b.
Model	OR	(95% CI)	P-value
CERT
Model 1	1.07	(0.88; 1.31)	0.50
Model 2	1.10	(0.90; 1.34)	0.35
Model 3	1.09	(0.89;1.34)	0.39
c.
Model	OR	(95% CI)	P-value
Ceramide risk score
Model 1	1.05	(0.99; 1.10)	0.099
Model 2	1.04	(0.99; 1.10)	0.104
Model 3	1.05	(0.99; 1.10)	0.106

Notes: Model 1: Crude. Model 2: Model 1 + type 2 diabetes. Model 3: Model 2 + BMI, hypertension, inflammatory markers, endothelial markers, medication use.

Abbreviations: BMI, body mass index; CERT, ceramide transfer protein; CI, confidence interval; MCI, mild cognitive impairment; OR, odds ratio. Bold: p < 0.05

Age- and sex-stratified associations of plasma ceramide species with mild cognitive impairment

No interaction effect of age was observed for the relationship of plasma ceramides with MCI. Additionally, no significant differences in baseline plasma ceramide or CERT levels were found between men and women. However, a trend-level (P < 0.1) or stronger interaction of sex was observed in the associations of ceramide C18:0 (P = 0.058), C24:1 (P = 0.012), and the ceramide risk score with MCI (P = 0.059; Table 3). Among the measured ceramide species, sex-stratified analyses revealed that higher C18:0 and C24:1 ceramides levels and a higher ceramide risk score were significantly associated with MCI in men (Figure 2). Although these associations were confounded by BMI in men (C18:0 [OR = 1.11, 95% CI (1.01, 1.22), P = 0.035]; C24:1 [OR = 1.13, 95% CI (1.03, 1.24), P = 0.013]; ceramide risk score [OR = 1.11, 95% CI (1.01, 1.22), P = 0.035]), the associations between these ceramide species and the ceramide risk score with MCI remained significant after controlling for BMI and other potentially confounding factors. No significant associations of any ceramide species or ceramide risk score with MCI were found in women. An ROC curve and AUC were generated to assess the diagnostic value of the ceramide risk score indicating MCI for men (AUC = 0.631, P < 0.001) and women (AUC = 0.504, P = 0.467; Figure S1).

TABLE 3 Sex-stratified associations of MCI with plasma ceramide species.

	Sex interaction term P-value	Men OR	(95% CI)	P-value	Women OR	(95% CI)	P-value
Model
Ceramide C18:0	0.058
Model 1		1.50	(1.13; 2.01)	0.006	1.07	(0.81; 1.43)	0.627
Model 2		1.52	(1.13; 2.06)	0.006	1.00	(0.73; 1.37)	0.994
Model 3		1.43	(1.05; 1.96)	0.024	0.98	(0.71; 1.36)	0.908
Ceramide C24:1	0.012
Model 1		1.52	(1.14; 2.02)	0.004	0.92	(0.69; 1.23)	0.574
Model 2		1.60	(1.19; 2.15)	0.002	0.92	(0.68; 1.24)	0.583
Model 3		1.75	(1.27; 2.41)	0.001	0.92	(0.67; 1.25)	0.587
Ceramide risk score	0.059
Model 1		1.09	(1.03; 1.14)	0.002	1.00	(0.95; 1.05)	0.978
Model 2		1.09	(1.03; 1.15)	0.001	0.99	(0.94; 1.05)	0.756
Model 3		1.09	(1.03; 1.15)	0.004	0.99	(0.93; 1.04)	0.650

Notes: Model 1: Crude. Model 2: Model 1 + age, educational level, type 2 diabetes. Model 3: Model 2 + BMI, hypertension, inflammatory markers, endothelial markers, medication use.

Abbreviations: BMI, body mass index; CI, confidence interval; MCI, mild cognitive impairment; OR, odds ratio. Bold: p < 0.05

FIGURE 2. Higher ceramide risk scores increase the risk for MCI in men. MCI, mild cognitive impairment

Associations of plasma sphingolipids and CERT levels with MRI brain volumes

Multivariable-adjusted associations of plasma sphingolipids and CERTs with brain volumes showed that higher levels of ceramide species C20:0, C22:0, and C24:1 were associated with larger volumes of the hippocampus after adjustment for age, sex, education level, and T2DM (Table 4). Adjustments for markers for cardiovascular risk factors, inflammation, endothelial function, or medication use did not substantially change these results. Results were similar for those with and without MCI, and no significant interaction by age or sex was observed. No associations were identified for CERT levels with brain WMHs, hippocampal, or CSF volumes.

TABLE 4 Multivariable-adjusted associations of sphingolipids and CERT with MRI brain volumes.

	β	(95% CI)	P-value
S1P
Hippocampus	0.05	(−0.07; 0.17)	0.381
White matter hyperintensities	0.05	(−0.08; 0.18)	0.459
Cerebrospinal fluid	–0.02	(−0.13; 0.08)	0.641
Ceramide C16:0
Hippocampus	0.10	(−0.01; 0.20)	0.080
White matter hyperintensities	0.04	(−0.08; 0.16)	0.480
Cerebrospinal fluid	–0.05	(−0.14; 0.05)	0.305
Ceramide C18:0
Hippocampus	0.08	(−0.03; 0.20)	0.143
White matter hyperintensities	0.11	(−0.01; 0.23)	0.079
Cerebrospinal fluid	–0.03	(−0.13; 0.07)	0.531
Ceramide C20:0
Hippocampus	0.12	(0.01; 0.23)	0.030
White matter hyperintensities	0.05	(−0.07; 0.17)	0.383
Cerebrospinal fluid	–0.04	(−0.14; 0.06)	0.408
Ceramide C22:0
Hippocampus	0.12	(0.01; 0.23)	0.030
White matter hyperintensities	0.02	(−0.10; 0.14)	0.742
Cerebrospinal fluid	–0.05	(−0.14; 0.05)	0.340
Ceramide C24:0
Hippocampus	0.11	(0.00; 0.22)	0.054
White matter hyperintensities	0.04	(−0.08; 0.16)	0.555
Cerebrospinal fluid	-0.03	(−0.13; 0.06)	0.488
Ceramide C24:1
Hippocampus	0.14	(0.02; 0.24)	0.016
White matter hyperintensities	0.06	(−0.06; 0.17)	0.370
Cerebrospinal fluid	0.00	(−0.10; 0.09)	0.926
Ceramide risk score
Hippocampus	0.09	(−0.02; 0.20)	0.123
White matter hyperintensities	0.07	(−0.05; 0.19)	0.242
Cerebrospinal fluid	–0.02	(−0.11; 0.08)	0.740
CERT
Hippocampus	–0.03	(−0.14; 0.07)	0.540
White matter hyperintensities	–0.04	(−0.15; 0.08)	0.519
Cerebrospinal fluid	0.06	(−0.03; 0.15)	0.217

Note: Model corrected for sex, age, educational level, type 2 diabetes.

Abbreviations: CERT, ceramide transfer protein; CI, confidence interval; MCI, mild cognitive impairment; OR, odds ratio. Bold: p < 0.05

DISCUSSION Study overview

Our study aimed to assess the association of plasma sphingolipids and CERT protein levels with MCI and their association with brain volumes after adjustment for cardiovascular and inflammatory risk factors and medication use. The association of ceramide levels with MCI differed between men and women; ceramide levels were positively associated with MCI in men, while no association was found in women. In addition, higher levels of ceramide species C20:0, C22:0, and C24:1 were associated with larger volumes of the hippocampus in multivariable models.

Sex differences in plasma ceramide levels and confounding factors

Among men, plasma ceramide levels and the ceramide risk score were positively associated with MCI. Although sex differences have been described in the risk of developing MCI and AD, and the importance of attention to these differences is increasingly advocated,¹⁷ few studies assessed sex differences in cross-sectional lipidomic studies related to cognitive decline. Nevertheless, baseline blood sphingolipids and their associations with AD are also known to be differentially expressed in men and women.^32,33,37,38 However, clear evidence of sex as a robust significant moderator of the relationship between plasma sphingolipids and cognitive decline in middle age has not yet been described. In our study, we observed a significant interaction effect of sex on the association of plasma ceramide levels with MCI in middle-aged individuals. Because early discovery is essential for favorable outcomes, our results show the importance of considering sex differences in early-stage dementia research and may potentially have a substantial impact on sex-specific biomarker development and early diagnosis. Our results further indicate that cardiovascular risk factors like BMI may underlie the observed sex differences in middle age. Although associations between plasma ceramides with MCI were robust after adjustment for covariates, the associations of all analyzed ceramide species and the ceramide risk score were partially dependent on BMI in men. Indeed, BMI is known to be related to sphingolipid metabolism, as well as increased inflammation and cognitive decline.^16,39 Additionally, obesity and associated metabolic diseases have shown apparent sex differences.³⁸ Therefore, BMI, in addition to the fact that the current study assesses ceramides as markers for MCI in a population-based sample rather than evaluating their potential as progression markers to probable AD, has to be considered when comparing the results of our study to previous longitudinal research that indicated that plasma ceramides levels might predict cognitive impairment in women.⁹ Furthermore, previous studies examining plasma lipid relationships with cognitive decline generally included individuals in a relatively older cohort, starting from age 60 and over (e.g., Mielke et al.^7–9 and Lim et al.³⁷). Although no significant interaction effect of age was found in our study, an interplay of various factors may lead to an altered lipid profile in a relatively younger cohort compared to an older cohort.

In addition, our study showed that higher levels of ceramide species C20:0, C22:0, and C24:1 were found to be associated with higher volumes of the hippocampus, independent of age, sex, education level, T2DM, and MCI status, whereas previously reported data indicated that elevated plasma sphingolipid levels are related to hippocampal volume loss.⁸ Furthermore, although it has been shown that higher levels of ceramides in the plasma were related to WMHs,⁴⁰ no such associations were found in our study. Because volumes of the hippocampus and WMHs are both known to be strongly dependent on age and changes are particularly observed around middle age,^41–43 a potentially changing lipid profile in this age group, together with more subtle changes in brain volumes, may underlie the observed results. Potential confounding factors like risk for cardiovascular diseases (CVDs) may cause an altered peripheral lipid profile in younger men. The risk of developing CVDs is known to differ considerably between the sexes—especially before menopause under the age of 50—when the risk of cardiovascular conditions in women is significantly lower than in men.⁴⁴ After menopause, however, the incidence of CVDs increases substantially in women. Additionally, sphingolipids and ceramide risk scores have been consistently associated with CVDs.⁴⁵ It has recently been suggested that ceramides may even be involved in the causality of heart dysfunction by driving a variety of pathophysiological processes.⁴⁶ Because BMI is also known to be associated with a markedly increased risk of developing CVDs at an earlier age,⁴⁷ this factor should be considered when investigating plasma sphingolipids in men and women < 60 years of age. Moreover, differences in demographic cohort characteristics should be considered.

Ceramide transfer proteins

Higher plasma ceramide C18:0 levels were significantly associated with higher CERT levels. Although little research is performed on plasma CERT in relation to ceramide levels, our findings are in line with previous reports showing upregulation of ceramide C18:0 after CERTL overexpression by adeno-associated virus (AAV)-mediated gene delivery in the cortex.⁵ Plasma CERT levels were not associated with MCI in either men or women. Similarly, CERT levels did not seem to be associated with WMHs, CSF, or hippocampal brain volumes. Although this could indicate that the ceramide transporter is minimally involved in the pathway regulating ceramide levels in MCI, methodological differences should additionally be considered.

LIMITATIONS

In our current CERT assay, we were able to measure > 10 ng/mL of CERT proteins in plasma. Because our data shows that CERT levels ranged from 0 to 20 to 30 ng/mL for both MCI and cognitively unimpaired individuals, it is possible that subtle differences in plasma CERT protein levels were not identified by the current ELISA set-up, and assays with a higher sensitivity are needed. Additionally, the differentiation between the isoforms of CERT and CERTL may be necessary to identify differences in plasma CERT protein levels. Furthermore, because it has recently been reported that CERT plays a role in regulating the sphingolipid composition and biogenesis of extracellular vesicles (EVs),¹¹ and EVs enriched for neuronal origin may be extracted from peripheral blood and used as neuronal markers,⁴⁸ future studies may focus on detecting CERT levels associated to blood-derived EVs and their potential role in dementia. Nevertheless, because CERT is known to have various biological functions,⁴ additional research is needed to explain the relationship between plasma ceramide and CERT levels and their association with the brain lipidome.

While our research focused on the examination of plasma sphingolipids and ceramides in particular, other plasma lipid species may also be altered at an early age and predict cognitive decline. Recently, it was found that a variety of different plasma lipid species were either positively or negatively associated with AD, many of which demonstrated significant interactions with age and sex.³⁷ Furthermore, various sex-specific lipid changes related to age and BMI have been demonstrated.³⁸ While early stages of cognitive decline were not evaluated in these studies, these data show that the association of plasma lipids with AD may be heavily dependent on individual lipid species, sex, age, and age-related confounding factors like obesity and BMI.

Although MCI was psychometrically assessed in our study, the definition of MCI was not based on clinical criteria. Additionally, biomarker confirmation of AD is not incorporated in the present cohort. Therefore, the etiology of MCI cannot be attributed to potential underlying AD pathology. Moreover, although apolipoprotein E (apoE) is a protein that plays a key role in lipid metabolism and APOE ε4 carrier status has been described as an important factor in both AD and early MCI,^49,50 APOE genotype analyses were not included in our study. Markers of AD pathology, APOE haplotype, or genotype carrier status might be particularly relevant in the association of plasma ceramides with MCI and should be considered in future research.

Finally, although men with MCI showed significantly higher plasma ceramide risk scores compared to cognitively unimpaired men, the diagnostic value of the ceramide risk score as a marker for MCI showed modest discriminative power in ROC analyses and potentially confounding factors might need to be considered to improve its predictive potential. Additionally, longitudinal studies are needed to assess the effects of plasma sphingolipid and ceramide transfer protein levels over time, including potentially long-term confounding factors like age-related cardiovascular characteristics, dietary patterns, and exercise.

In summary, we found that plasma ceramide levels were positively associated with MCI in middle-aged men but not in women and higher levels of various ceramide species were associated with higher volumes of the hippocampus independent of sex, age, education level, or T2DM. However, plasma CERT levels were not found to be related to MCI or brain volumes in this study. These results underline the importance of considering sex- and age-related factors in studies evaluating sphingolipid and CERT metabolism changes in relation to cognitive decline. Additional studies with larger sample sizes are needed to further evaluate the association between blood sphingolipids across age groups, especially at younger ages, and the association with MCI after adjusting for potentially confounding factors. Furthermore, longitudinal studies could assess the predictability of early-life plasma sphingolipids and the progression of at-risk individuals to AD-related dementia.

ACKNOWLEDGMENTS

The authors have nothing to report. This work was supported by the Galen and Hilary Weston Foundation (project number NR160111). The Maastricht Study was supported by the European Regional Development Fund via OP-Zuid, the Province of Limburg, the Dutch Ministry of Economic Affairs (grant 31O.041), Stichting De Weijerhorst (Maastricht, the Netherlands), the Pearl String Initiative Diabetes (Amsterdam, the Netherlands), the Cardiovascular Center (CVC, Maastricht, the Netherlands), CARIM School for Cardiovascular Diseases (Maastricht, the Netherlands), CAPHRI Care and Public Health Research Institute (Maastricht, the Netherlands), NUTRIM School for Nutrition and Translational Research in Metabolism (Maastricht, the Netherlands), Stichting Annadal (Maastricht, the Netherlands), Health Foundation Limburg (Maastricht, the Netherlands), and by unrestricted grants from Janssen-Cilag BV (Tilburg, the Netherlands), Novo Nordisk Farma BV (Alphen aan den Rijn, the Netherlands), and Sanofi-Aventis Netherlands BV (Gouda, the Netherlands).

CONFLICT OF INTEREST STATEMENT

Dr Mielke has consulted for Biogen and Labcorp. All other authors have nothing to disclose. Author disclosures are available in the supporting information.

ETHICAL STANDARDS

The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures involving human subjects were approved by the institutional Medical Ethical Committee (NL31329.068.10) and the Ministry of Health, Welfare and Sport of the Netherlands (Permit 131088-105234-PG).

CONSENT STATEMENT

All participants provided informed consent. Study information including an informed consent form was sent to all participants by mail prior to the study visits. An oral explanation on study procedures is provided and the informed consent form is signed.