Plasma Sphingolipids Mediate a Relationship Between Type 2 Diabetes and Memory Outcomes in Patients with Coronary Artery Disease Undertaking Exercise

Abstract

Background:

Exercise prevents recurrent cardiovascular events and it may combat cognitive decline in coronary artery disease (CAD); however, evidence in type 2 diabetes (T2DM) has been mixed. T2DM and memory decline have been associated with differences in the plasma sphingolipidome.

Objective:

Here, we will investigate whether T2DM-related sphingolipids predict less memory improvement over an exercise intervention for CAD.

Methods:

Among participants with CAD entering a 6-month exercise intervention, we matched 20 with T2DM to 40 without T2DM for age, sex, and body mass index. We assessed 45 sphingolipid species using high-performance liquid chromatography coupled electrospray ionization tandem mass spectrometry with multiple reaction monitoring. We assessed memory using the California Verbal Learning Test, 2nd Ed, and the revised Brief Visuospatial Learning Test.

Results:

Partial least squares discriminant analysis identified 8 species that distinguished T2DM from non-T2DM groups with 83% (95% confidence interval [70%, 95%]) accuracy in a receiver operator characteristic curve (validated by internal resampling, 1000 permutations, p = 0.01). At baseline, T2DM-associated sphingolipids (ceramide C22 : 0, monohexylceramide C16 : 1, and lactosylceramide C24 : 0) were associated with poorer memory, attention, and psychomotor processing speed performance. Among 50 completers, an indirect effect of T2DM on less improvement in verbal memory was mediated by monohexylceramide C16 : 1 (0.86 fewer words recalled, 95% bootstrap confidence interval [–2.32, –0.24]), and an indirect effect of T2DM on less visuospatial memory improvement was mediated by ceramide C22 : 0 concentrations (0.42 fewer points, 95% bootstrap confidence interval [–1.17, –0.05]).

Conclusions:

Ceramide species associated with T2DM predicted poorer cognitive responses to exercise in patients with CAD.

Keywords

Cognition coronary artery disease exercise memory sphingolipids type 2 diabetes

INTRODUCTION

Cognitive deficits, including memory impairments, are clinically important in people with type 2 diabetes (T2DM), especially because they interfere with diabetes self-management [1]. Vascular disease is increasingly appreciated as a contributor to cognitive decline in later life, [2] and specifically in people with T2DM [3]. Therefore, it is of interest to understand the mechanisms that influence cognition in people with T2DM, and cardiovascular complications such as coronary artery disease (CAD), for which T2DM is a prominent risk factor.

Physical exercise is one of the most promising interventions to forestall age-related cognitive decline, improve insulin sensitivity, and reduce the risk of CAD-related events; however, among those with T2DM, the evidence to support the cardiovascular and cognitive benefits of physical activity has been mixed. Notably, the Action for Health in Diabetes trial found no significant long-term differences in cardiovascular morbidity or mortality, and only small differences in cognition between those who received an intensive intervention involving exercise, and those who received standard of care [4, 5]. It is therefore of interest to explore potential biomarkers that can predict cognitive response to exercise, and more specifically, markers that might explain the lack of cognitive benefit observed in T2DM.

Widespread changes in lipid metabolism have been detected in T2DM; however, their impact on the brain is not well understood. Previously, long-chain (e.g., carbon chain lengths C16 to C20) and very long-chain (e.g., C22 to C26) ceramides have been associated with poorer metabolic health in animals [6], hippocampal volume loss over time in people with mild cognitive impairment [7], and with incident dementia [8], in longitudinal cohort studies. Higher very long-chain ceramides have been associated consistently with less improvement in verbal memory in CAD patients following cardiac rehabilitation [9 –11]. It has been shown that exercise can lower peripheral blood ceramide concentrations in healthy individuals [12]; however, these markers have not been examined in the context of exercise interventions in T2DM.

The present study compares peripheral blood sphingolipid concentrations between age-, sex-, and body mass index (BMI)-matched groups of CAD patients with and without T2DM. We aim to identify the sphingolipid species that differentiate these groups, to determine whether those species are associated with cognitive performance, and to determine if they can predict memory outcomes after 6 months of an exercise intervention. We hypothesize that ceramides will explain (mediate) the relationship between T2DM and poorer memory outcomes over 6 months.

METHODS

Study design and participants

This retrospective study examined associations between plasma sphingolipids and verbal memory changes among CAD patients, with and without T2DM undertaking a 6-month exercise program, matched individually 2 : 1 for age, sex, and BMI category. Patients were included if they had evidence of CAD (previous hospitalization for acute myocardial infarction, coronary angiographic evidence of ≥50% blockage in one or more major coronary artery, or prior revascularization). Patients were excluded based on previously diagnosed neurodegenerative illnesses or probable dementia (standardized Mini-Mental Status Examination score < 24), [13] active cancer, surgery planned within 12 months, schizophrenia, bipolar affective disorder, and substance abuse. Patients with a confirmed diagnosis of T2DM were identified from a clinical chart review. The participants included without T2DM did not have a clinical diagnosis of diabetes, were not using an antidiabetic medication, and had a baseline glycated hemoglobin A1c (HbA1C) of less than 0.065. This research was conducted in accordance with a study protocol reviewed and approved by the Research Ethics Boards of Sunnybrook Health Sciences Centre and the University Health Network, and in accordance with the ethical principles set for by these boards, International Conference on Harmonization-Good Clinical Practices and the Tri-Council Policy Statement: Ethical Conduct for Research Involving Human subjects. Prior to enrollment in the study, informed written consent was obtained from all participants.

Cardiac rehabilitation program

The cardiac rehabilitation program consisted of supervised aerobic walk or walk/jog and resistance training once a week for 24 weeks. Participants were also asked to exercise at home for a minimum of 4 times per week. The aerobic exercise prescription was based on 80% of the peak oxygen consumption (VO_2Peak) achieved during the baseline fitness assessment. The VO_2peak, which has been suggested to be the most reproducible and reliable indicator of cardiopulmonary fitness, [14] was determined using a cycle ergometer (Ergoselect 200P, Ergoline, Bitz, Germany) symptom-limited graded exercise test.

Demographics and clinical characteristics

Participant interviews were used to collect demographic information, and medical history, including cardiac diagnoses, concomitant cardiac and diabetes medications, and cardiovascular risk factors (i.e., details of hyperlipidemia, hypertension, etc.). Anthropometrics and fitness characteristics were assessed by the exercise team at the cardiac rehabilitation site, including height, weight, BMI and waist circumference. Standard labs (HbA1c, total cholesterol, high-density lipoprotein [HDL], low-density lipoprotein [LDL], and triglyceride concentrations) were assayed according to standard clinical protocols from blood samples obtained during the participant interview on the day of cognitive assessments. The presence of an apolipoprotein E (APOE) ɛ4 allele was determined from whole blood using restriction fragment length polymorphism polymerase chain reaction analysis as a potential cognitive confounder [15].

Cognitive testing

Cognition was assessed by a trained researcher at baseline and at the completion of the exercise program (6 months). Participants did not consume caffeine for at least 4 h prior to testing. The California Verbal Learning Test 2nd Edition (CVLT-II) was used to assess verbal memory. The long-delay free recall outcome (LDFR) was recorded as recall of the word list after 20 min without cueing [16]. Visuospatial memory was assessed using the Brief Visuospatial Memory Test-Revised (BVMT-R) recall of 6 figures after 25 min without any memory cues (BVMT-DR) [17]. The Digit Symbol Substitution Test (DSST) [18] and the Trail Making Test Part A (TMT-A) were used to assess attention and psychomotor processing speed while the Trail Making Test Part B (TMT-B) [19] and Stroop (Victoria version) were used to assess executive function. Longer times to complete the DSST, TMT-A, TMT-B and the Stroop reflect poorer performance.

Sphingolipid measurements

Fasting blood was drawn directly before the cognitive assessment (0900 h±30 min) and centrifuged at 4C for 10 min at 1000 rpm. Plasma was isolated and stored at –80C immediately after. High performance liquid chromatography coupled electrospray ionization tandem mass spectrometry (LC/MS/MS) using multiple reaction monitoring and subsequent processing by the Analyst 1.4.2 software package was used to quantify 45 individual sphingolipid species (see Supplementary Table 1). A modified Bligh and Dyer procedure (Avanti Polar Lipids, Alabaster, AL, USA) was used to perform a crude lipid extraction from the plasma as described previously [20]. Before analysis, the plasma extracts were dried and resuspended in methanol. For the temporal resolution of compounds, high performance liquid chromatography (PerkinElmer, MA, USA) with a phenomenex luna 100 x 2 mm C18 column, coupled with guard column containing identical packing material (Phenomenex, Torrance, CA, USA), was used. For a typical run, the LC column was first pre-equilibrated for 0.5 min with the first mobile phase consisting of 85% methanol, 15% H₂O, and 5 mM ammonium formate. The column was then eluted with the second mobile phase consisting of 99% methanol, 1% formic acid, and 5 mM ammonium formate at the flow rate of 100.0 ml /min. Eluted samples were then injected into an electrospray ion source coupled to a triple-quadrupole mass spectrometer (API3000, AB Sciex Inc, Thornhill, Ontario, Canada) and analysis was conducted by multiple reaction monitoring mode monitoring the parent compound and products by ion scan. The internal standard for sphingomyelins was sphingomyelin C12 : 0, and for ceramide, ceramide C12 : 0. For each sphingolipid of interest, 8-point calibration curves (0.1–1000ng/mL) were constructed by plotting area under the curve with each calibration normalized to the internal standard. Identified species were fit to standard curves based on acyl chain length to determine sphingolipid concentrations (ng/mL). Analyst 1.4.2 and MultiQuant software (AB Sciex Inc, Thornhill, Ontario, Canada) determined instrument control and quantification of spectral data.

Statistical analysis

Participant characteristics were compared using Pearson chi-squared, or univariate analyses of variance (ANOVA), as appropriate. ANOVA models were used to determine the plasma sphingolipid species that significantly differed between the groups. Sphingolipid species significantly associated with antidiabetic medication use were also determined and removed from subsequent analyses. To determine the T2DM-related sphingolipid species that are best able to discriminate the groups, a Partial Least Squares Discriminant Analysis (PLS-DA) was used. Plasma sphingolipid concentrations were log transformed and Pareto scaled prior to PLS-DA. The predictive accuracies of the signature were visualized in variable importance plots, and tested in multivariate receiver operator characteristic (ROC) curves. The results were cross-validated in permutation (n = 1000) testing (metaboanalyst.ca) [21].

To examine the effects of T2DM and sphingolipids on cognition over time, a longitudinal mediation analysis, implemented in the PROCESS macro for SPSS, was used [22]. Specifically, we hypothesized an indirect effect of T2DM on verbal memory at 6 months that is mediated by ceramide concentrations (computed as the product of the linear regression coefficients a×b; see Fig. 3). The model controlled for the baseline measure, age and sex to account for possible residual confounding, regressed at once on the mediator and the outcome [23]. This mediation effect is referred to as “indirect” to distinguish it from the direct effect (c’), which is the effect that remains between the independent variable and the outcome after accounting for the effect of the mediator. An inferential bootstrapping procedure with 10,000 permutations was used to compute 95% confidence intervals (CI), and the result was considered to be significant if the bias-corrected bootstrap confidence interval did not cross 0. Post-hoc models controlled for potential confounders identified in univariate analyses.

RESULTS

Patient characteristics

Characteristics of the 20 CAD patients with T2DM, and 40 age-, sex-, and BMI category-matched CAD patients without T2DM, are reported in Table 1. Those with T2DM had significantly lower HDL concentrations and higher triglyceride concentrations compared to those without T2DM controls. VO_2peak was also significantly lower, and the frequency of calcium channel blocker use higher, in CAD patients with T2DM compared to without T2DM. As expected, HbA1c concentrations and antidiabetic use were significantly higher in CAD patients with T2DM compared to CAD patients without T2DM.

Table 1

Characteristics of T2DM and non-T2DM cardiac rehabilitation participants at baseline

Characteristic	Non-T2DM (n = 40)	T2DM (n = 20)	Test statistic,
	Mean±SD or n (% )	Mean±SD or n (% )	p-value
Sociodemographic
Age, y	63±6	63±7	F<0.005, p = 0.99
Sex, male	32 (80)	16 (80)	χ ² < 0.005, p > 0.99
Ethnicity, Caucasian	35 (88)	14 (70)	χ ² = 2.73, p = 0.10
Education, y	16±63	15±3	F = 2.45, p = 0.12
ApoE ɛ4 allele carrier	13 (33)	6 (30)	χ ² = 0.04, p = 0.84
Depression	6 (15)	1 (5)	χ ² = 1.29, p = 0.26
Lipid profile and HbA1c
LDL, mmol/L	1.59±0.57	1.29±0.72	F = 3.13, p = 0.08
HDL, mmol/L	1.38±0.40	1.17±0.29	F = 4.39, p = 0.04^*
Total cholesterol, mmol/L	3.48±0.73	3.12±0.89	F = 2.89, p = 0.09
Triglycerides, mmol/L	1.07±0.49	1.45±0.71	F = 5.85, p = 0.02^*
HbA1c	0.056±0.003	0.069±0.009	F = 65.25, p < 0.005^*
Vascular risk factors
Hypertension	38 (95)	20 (100)	χ ² = 1.03, p = 0.31
BMI, kg/m²	30.8±4.7	32.7±6.7	F = 1.67, p = 0.20
Smoking History, ever smoker	23 (58)	12 (60)	χ ² = 0.03, p = 0.85
Hypercholesterolemia	40 (100)	20 (100)	χ ² < 0.005, p > 0.99
Cardiac history
MI	19 (48)	11 (55)	χ ² = 0.30, p = 0.58
CABG	11 (28)	9 (45)	χ ² = 1.84, p = 0.18
Stent	23 (58)	11 (55)	χ ² = 0.03, p = 0.85
Angina	6 (15)	0 (0)	–
Cardiopulmonary fitness
VO_2peak	20.8±05.2	17.2±5.2	F = 6.21, p = 0.02^*
Medications
β-blockers	34 (85)	17 (85)	χ ² < 0.005, p = 1.00
Calcium channel blockers	5 (13)	7 (35)	χ ² = 4.22, p = 0.04^*
Diuretics	8 (20)	5 (25)	χ ² = 0.20, p = 0.67
Antihypertensives	27 (68)	18 (90)	χ ² = 3.60, p = 0.06
Hypoglycemic agents	0 (0)	17 (85)	–
Metformin	0 (0)	14 (70)	–
Insulin	0 (0)	7 (35)	–
Platelet inhibitors	38 (95)	19 (95)	χ ² < 0.005, p = .99
Statins, high dose	22 (55)	13 (65)	χ ² = 0.77, p = 0.38

At follow-up, data were available for 50 of the 60 participants. Participants who did not complete the study were younger (59±6 versus 64±6 years; F = 6.09, p = 0.02), had fewer years of education (14±4 years versus 16±3 years; F = 4.32, p = 0.04), had higher baseline triglycerides (1.65±0.94 versus 1.11±0.46 mmol/L; F = 7.59, p = 0.008), and they were more likely to be ApoE ɛ4 allele carriers (60% versus 26% ; χ ² = 4.45, p = 0.04).

A sphingolipid signature of T2DM

Of the 45 sphingolipid species assessed, ceramide C22 : 0 (F_1,59 = 9.74, p = 0.003), C24 : 0 (F_1,59 = 4.59, p = 0.04), C22 : 1 (F_1,59 = 5.46, p = 0.02), monohexylceramide C16 : 1 (F_1,59 = 8.74, p = 0.004), and lactosylceramide C24 : 0 (F_1,59 = 4.50, p = 0.04) were significantly higher in CAD patients with T2DM versus patients without T2DM. Sphingomyelin C24 : 1 (F_1,59 = 9.05, p = 0.004), monohexylceramide C18 : 0 (F_1,59 = 5.43, p = 0.02) and lactosylceramide C18 : 1 (F_1,59 = 4.69, p = 0.03) were significantly lower in CAD patients with T2DM versus those without. These species were entered into a PLS-DA model to determine the sphingolipid signature associated with T2DM in CAD. Overall, the sphingolipid signature associated with T2DM among CAD patients accounted for 40% of the variance between the samples (Fig. 1). The signature of 8 species produced a ROC with an area under the curve of 83% (95% CI [70% –95%]), which cross-validated in permutation testing (1000 permutations, p = 0.01; Fig. 2). The species with the highest capacity to discriminate between those with and without T2DM were ceramide C22 : 0 and monohexylceramide C16 : 1.

Fig.1

Variable importance plot of T2DM-related sphingolipids showing predictive accuracies of discrimination between CAD and CAD and T2DM patient groups. The species with the highest capacity to discriminate between those with and without T2DM were ceramide C22 : 0 and monohexylceramide C16 : 1. Overall, the sphingolipid signature associated with T2DM in CAD accounted for 40% of the variance between the samples. c, ceramide, MHxC, monohexylceramide, LacC lactosylceramide, SM, sphingomyelin.

Fig.2

Multivariate receiver operator characteristic (ROC) curve analysis assessing the ability of the T2DM-related sphingolipid signature to discriminate between CAD and CAD and T2DM patient groups. The signature of 8 species produced a ROC with an area under the curve of 83% (95% CI [70% –95%]), which cross-validated in permutation testing (p = 0.01).

Cognitive performance in CAD patients with and without T2DM

There were no significant differences in cognitive test scores between patients with and without T2DM at baseline (Table 2). Verbal memory (CVLT-II, LDFR), attention/psychomotor processing speed (DSST, symbols coded) and executive function (time to complete the Stroop) scores improved significantly over 6 months of exercise intervention in the whole group (Table 3).

Table 2

Cognitive performance in CAD patients with and without T2DM at baseline

Cognitive Assessments	Total	CAD	CAD+T2DM	F	p
	(n = 60)	(n = 40)	(n = 20)
Digit Symbol Substitution Test
Symbols	59±14	61±14	55±13	2.80	0.10
Z score	0.08±0.82	0.21±0.89	–0.17±0.62
Trail Making Test Part A
Seconds	36±12	36±12	36±12
Z Score	–0.13±0.85	–0.12±0.88	–0.15±0.80	0.01	0.91
Stroop Color-Word Interference Test
Seconds	31±11	30±10	32±12	0.17	0.68
Z score	0.13±1.08	0.18±1.04	0.05±1.16
Trail Making Test Part B
Seconds	87±36	85±36	91±38
Z Score	–0.07±0.98	–0.02±1.00	–0.17±0.94	0.29	0.59
CVLT-II — Long Delay Free Recall
Words	11.4±3.8	11.4±3.8	11.3±3.5	0.39	0.54
Z score	0.74±1.11	0.68±1.19	0.87±0.93
BVMT-R — Long Delay Free Recall
Raw Score	9.0±2. 8	9.2±2.6	8.5±3.2	0.93	0.34
Z score	0.28±1.18	0.38±1.11	0.07±1.32

^*Statistical tests were performed on age, sex and education adjusted Z-scores using analyses of covariance.

Table 3

Cognitive performance in CAD patients with and without T2DM after 6 months of exercise

Cognitive Assessments	Total	CAD	CAD+	Time
	(n = 50)	(n = 34)	T2DM (n = 16)	F	p
Digit Symbol Substitution Test
Symbols	64±14	66±14	60±13
Z score	0.43±0.83	0.13±0.70	0.57±0.86	8.26	0.01^*
Trail Making Test Part A
Seconds	33±11	32±10	36±11
Z Score	0.10±0.84	0.21±0.84	–0.15±0.82	0.96	0.33
Stroop Color-Word Interference Test
Seconds	29±8	27±8	29±8
Z score	0.40±1.00	0.46±1.03	0.27±0.94	7.23	0.01^*
Trail Making Test Part B
Seconds	80±32	78±34	83±29
Z Score	0.17±1.00	0.24±1.01	0.001±0.98	3.48	0.07
CVLT-II — Long Delay Free Recall
Words	13.1±3.3	13.4±2.8	12.4±4.2
Z score	1.29±0.97	1.35±0.81	1.16±1.26	5.43	0.02^*
BVMT-R — Long Delay Free Recall
Raw Score	8.4±2.6	8.2±2.1	8.7±2.7
Z score	0.03±1.13	–0.04±1.17	0.17±1.06	2.68	0.11

^*Statistical tests were performed on age, sex and education adjusted Z-scores using repeated measures analyses of covariance.

Associations between T2DM related sphingolipids and cognition at baseline

Of the T2DM related sphingolipids, higher ceramide C22 : 1 was significantly associated with poorer visuospatial memory (BVMT-DR) (r = –0.27, p = 0.04) and DSST (r = –0.27, p = 0.04) scores and higher monohexylceramide C16 : 1 (r = 0.26, p = 0.05) and lactosylceramide C24 : 0 (r = 0.26, p = 0.05) were associated with longer time to complete TMT-A in bivariate correlations in all participants at baseline. Sphingomyelin C24 : 1, ceramides C22 : 0 and C24 : 0, monohexylceramide C18 : 0, and lactosylceramide C18 : 1 were not correlated with cognitive scores at baseline.

When assessing the associations between species with the highest capacity to discriminate between those with and without T2DM and baseline cognitive performance in the whole study cohort, baseline ceramide C22 : 0 (β= 0.16, p = 0.28) and monohexylceramide C16 : 1 (β= 0.07, p = 0.66) concentrations were not associated with LDFR in a linear regression model controlling for age, sex, years of education, and diabetes. Ceramide C22 : 0 concentrations were also not associated with baseline BVMT-DR (β= –0.13, p = 0.45), DSST (β= –0.04, p = 0.83), TMT-A (β= –0.22, p = 0.17), TMT-B (β= 0.08, p = 0.67) or Stroop (β= 0.12, p = 0.49) scores in linear regression models. Higher monohexylceramide C16 : 1 concentrations were associated with longer time to complete the TMT-A (β= 0.46, p = 0.005) but not BVMT-DR (β= 0.19, p = 0.28), DSST (β= 0.05, p = 0.76), TMT-B (β= –0.07, p = 0.68) or Stroop (β= –0.12, p = 0.50) scores at baseline in linear regression models.

Longitudinal mediation effect of the T2DM-related sphingolipids on verbal memory changes with exercise

In a mediation analysis controlling for age, sex, and baseline LDFR scores, higher baseline monohexylceramide C16 : 1 mediated a significant indirect effect of T2DM on poorer LDFR scores at 6 months (–0.86, 95% bootstrap CI [–2.32, –0.24] words, n = 50; Fig. 3). In post-hoc mediation models that included years of education (–0.54, 95% bootstrap CI [–1.68, –0.11]), baseline HDL concentrations (–0.92, 95% bootstrap CI [–2.37, –0.29]), baseline triglyceride concentrations (–0.85, 95% bootstrap CI [–2.40, –0.24]), HbA1c (–1.05, 95% bootstrap CI [–3.20, –0.25]), VO_2peak (–0.98, 95% bootstrap CI [–2.84, –0.30]), and calcium channel blocker use (–0.85, 95% CI [–2.40, –0.22]), the indirect effect persisted.

Fig.3

Mediation analysis of effects of T2DM and sphingolipids on cognition over time. Higher baseline monohexylceramide C16 : 1 mediated a significant indirect effect (computed as the product of coefficients a×b) of T2DM on poorer LDFR scores at 6 months (–0.86, 95% bootstrap CI [–2.32, –0.24] words, n = 50). An inferential bootstrapping procedure with 10,000 permutations was used to compute 95% confidence intervals, and the result was considered to be significant if the bias-corrected bootstrap confidence interval did not cross 0. The model controlled for age, sex and baseline verbal memory.

An indirect effect of T2DM on follow-up LDFR scores mediated by ceramide C22 : 0 concentrations was not significant (–0.69, 95% bootstrap CI [–2.48, 0.04]).

Exploratory associations between the T2DM sphingolipid signature and other cognitive domains

In exploratory mediation analyses, higher baseline ceramide C22 : 0 mediated a significant indirect effect of T2DM on poorer BVMT-R delayed recall scores at 6 months (–0.42, 95% bootstrap CI [–1.17, –0.05] point, n = 50) in a model controlling for age, sex, and the baseline BVMT-R delayed recall score (Supplementary Table 2). Baseline concentrations of monohexylceramide C16 : 1 and ceramide C22 : 0 did not mediate effects of T2DM on changes in other cognitive domains (Supplementary Table 2).

DISCUSSION

Among CAD patients in an exercise program, this study identified 8 plasma sphingolipids that differed between those with and without comorbid T2DM. This 8-species signature was able to distinguish CAD patients with and without T2DM with 83% accuracy. Although this result cross-validated in permutation testing, it should be replicated in independent cohorts. The most important species in discriminating the two groups were elevated concentrations of ceramide C22 : 0 and monohexylceramide C16 : 1, although species with chain lengths 18 (monohexyl and lactosyl ceramide species lower) and 24 (one ceramide higher, one sphingomyelin lower) were also involved. The results complement a previous study in subjects without CAD, finding higher concentrations of total ceramide, and specifically C18 : 0, C20 : 0, C24 : 1, in people with T2DM compared to those without T2DM [24]. Since CAD morbidity and mortality are important complications of T2DM, these sphingolipid differences may be important heterogeneity factors in the presentation of CAD. Recently, ceramides C16 : 0 and C24 : 1, and ratios of C16 : 0, C18 : 0 and C24 : 1 to C24 : 0, were associated with cardiac mortality in patients with stable CAD, independent of the risk associated with LDL concentrations and other vascular risk factors [25]. The specific species associated with cardiovascular mortality in T2DM might be explored in future studies.

In this population with CAD, the ceramide species that differed between the patients with and without T2DM were also associated with poorer cognitive performance at baseline, specifically on tests of psychomotor processing speed and attention [7 , 26]. These domains are affected in vascular cognitive impairment, [16] and such deficits often reflect vascular insults to the frontal lobes or cerebral white matter [27]. White matter changes have been shown previously to predict cognitive non-response to exercise among CAD patients, [28] and they have been linked to very long ceramides including ceramide C22 : 0, C22 : 1, and C24 : 1 in cognitively normal older adults [29]. These results suggest that ceramides may mediate an effect of T2DM on vascular cognitive impairment.

After 6 months of exercise, participants experienced an improvement in the DSST, the Stroop test, and verbal memory (Table 3). These results are consistent with previous findings suggesting that certain cognitive domains, such as processing speed, executive function and memory are amenable to exercise effects [30 –33]. While adaptations to exercise are incompletely understood, exercise is associated with improved insulin sensitivity, which may be relevant in this population [34].

The novel T2DM-associated lipid monohexylceramide C16 : 1 predicted less improvement in verbal memory in all CAD patients. The function and role of glucosylceramides in cognitive decline is not well known. Previously, higher peripheral blood lactosylceramide concentrations were associated with a 10-fold increase in the risk of AD among elderly women [35]. Higher concentrations of glucosylceramides were also found in the AD brain postmortem, [36] and in a cellular model of AD [37]. Recent findings suggest a link between glucosylceramides, decreased insulin sensitivity and neuronal viability in models of AD in vitro and in vivo. Glucosylceramides may be especially relevant to cognitive changes in at-risk populations, such as the present cohort of those with CAD and diabetes.

Plasma ceramide C22 : 0 concentrations were also higher among CAD patients with T2DM compared to those without T2DM, and this difference mediated a negative effect of T2DM on visuospatial memory improvement. This adds to previous data implicating ceramides, particularly the very long-chain ceramides C22 : 0 and C24 : 0, in interfering with cognitive effects of exercise in people with CAD [9]. Previous studies have also implicated C22 : 0 and C24 : 0 in declines in memory and hippocampal volume in patients with mild cognitive impairment [7]. The present study suggests that these same species are associated with vascular and metabolic comorbidities, which are risk factors for dementia. As such, sphingolipid pathways may indicate important targets for cognitive interventions or preventative strategies. Future preclinical studies might explore the relationships between ceramide-related pathways [39] (e.g., endoplasmic reticulum stress, mitochondrial dysfunction, pro-apoptotic signaling, neuroinflammation, etc.) to identify potential underlying mechanisms or downstream molecular targets. Long-chain and very long-chain ceramides and glucosylceramides are elevated in patients with diabetes [24 , 41]. Animal studies also suggest that these lipids may be involved in the development of brain insulin resistance, which could contribute to neurodegeneration, [42] but these relationships have not been well established in people.

A limitation of this study is that it is a small single-site study of mostly older men. It was not powered to detect a main effect of diabetes on cognition; however, due to adequate matching of T2DM and non-T2DM patients, it was sufficiently powered to detect effects of T2DM on the sphingolipidome, and of ceramides on cognition. Future studies could validate the findings in independent populations, and determine generalizability to other cohorts, including those that include more women. The majority of patients with T2DM were using antidiabetic medications, most commonly oral metformin, a biguanide that acts by increasing insulin uptake. Accordingly, this study was not able to differentiate effects of metformin from those of T2DM, and therefore, species associated with metformin use were removed from the models to account conservatively for any potential confounding effects on the sphingolipidome. Future studies might examine the sphingolipidome before and after the prescription of metformin, and of other antidiabetic medications, to determine their specific effects. Lipid metabolic differences have been associated with obesity and T2DM, including ceramide alterations in peripheral blood, notably higher C18 : 0 and C24 : 0, and lower ceramide C20 : 0 and C22 : 0 [43]. This study addressed obesity as a potential confounder by matching patient groups for BMI category. Differences in HDL and triglycerides persisted between groups; however, the main findings were independent of these lipid parameters in post-hoc models. As a potential methodological limitation, PLS-DA may be subject to overfitting and other pitfalls [44]; therefore, the PLS-DA findings should be viewed as exploratory, mainly providing insight into the discriminant species to be used in further analyses. Finally, the physiological sources of the lipid species identified in peripheral blood are unknown, limiting inferences regarding their precise relationships with neural and/or vascular mechanisms of cognitive impairment.

The sphingolipidome might be an important source of heterogeneity in the presentation of CAD that can indicate contributions of T2DM to cognitive decline. Some of the specific sphingolipids associated with T2DM here have been previously associated with cognitive decline in other populations, and, if replicated, they might be developed to monitor and predict vascular cognitive impairment in those with T2DM. Future work might seek to determine the underlying mechanisms, and to determine if these sphingolipid pathways might be targets for intervention to protect against cognitive decline in T2DM, or targets for co-intervention with exercise to restore the cognitive benefits of exercise in T2DM.

Footnotes

ACKNOWLEDGMENTS

This work was supported by grants from the Diabetic Complications Consortium (16GRU3711), the University of Toronto Centre for Collaborative Drug Research, The Heart and Stroke Foundation Canadian Partnership for Stroke Recovery, the National Sciences and Engineering Research Council (RGPIN-2017-0692), the Canadian Institutes of Health Research (MOP-114913 [KLL] and PJT-159711 [WS]), and by the Alzheimer’s Association Research Grant Program (USA) in partnership with Brain Canada (AARG501466).

The authors would like to thank the staff and participants of the UHN Toronto Rehab Cardiac.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Munshi

(2017) Cognitive dysfunction in older adults with diabetes: What a clinician needs to know, Diabetes Care 40, 461–467.

Deckers

, Schievink

SHJ

, Rodriquez

MMF

, van Oostenbrugge

, van Boxtel

MPJ

, Verhey

FRJ

, Kohler

(2017) Coronary heart disease and risk for cognitive impairment or dementia: Systematic review and meta-analysis, PLoS One 12, e0184244.

Ruis

, Biessels

, Gorter

, van den Donk

, Kappelle

, Rutten

(2009) Cognition in the early stage of type 2 diabetes, Diabetes Care 32, 1261–1265.

Look

ARG

, Wing

, Bolin

, Brancati

, Bray

, Clark

, Coday

, Crow

, Curtis

, Egan

, Espeland

, Evans

, Foreyt

, Ghazarian

, Gregg

, Harrison

, Hazuda

, Hill

, Horton

, Hubbard

, Jakicic

, Jeffery

, Johnson

, Kahn

, Kitabchi

, Knowler

, Lewis

, Maschak-Carey

, Montez

, Murillo

, Nathan

, Patricio

, Peters

, Pi-Sunyer

, Pownall

, Reboussin

, Regensteiner

, Rickman

, Ryan

, Safford

, Wadden

, Wagenknecht

, West

, Williamson

, Yanovski

(2013) Cardiovascular effects of intensive lifestyle intervention in type 2 diabetes, N Engl J Med 369, 145–154.

Hayden

, Baker

, Bray

, Carvajal

, Demos-McDermott

, Hergenroeder

, Hill

, Horton

, Jakicic

, Johnson

, Neiberg

, Rapp

, Wadden

, Miller

(2018) Long-term impact of intensive lifestyle intervention on cognitive function assessed with the Toolbox: The Look AHEAD study, Alzheimers Dement (Amst) 10, 41–48 National Institutes of Health.

Xia

, Holland

, Kusminski

, Sun

, Sharma

, Pearson

, Sifuentes

, McDonald

, Gordillo

, Scherer

(2015) Targeted induction of ceramide degradation leads to improved systemic metabolism and reduced hepatic steatosis, Cell Metab 22, 266–278.

Mielke

, Haughey

, Ratnam

Bandaru VV

, Schech

, Carrick

, Carlson

, Mori

, Miller

, Ceritoglu

, Brown

, Albert

, Lyketsos

(2010) Plasma ceramides are altered in mild cognitive impairment and predict cognitive decline and hippocampal volume loss, Alzheimers Dement 6, 378–385.

Mielke

, Bandaru

, Haughey

, Xia

, Fried

, Yasar

, Albert

, Varma

, Harris

, Schneider

, Rabins

, Bandeen-Roche

, Lyketsos

, Carlson

(2012) Serum ceramides increase the risk of Alzheimer disease: The Women’s Health and Aging Study II, Neurology 79, 633–641.

Saleem

, Ratnam

Bandaru VV

, Herrmann

, Swardfager

, Mielke

, Oh

, Shammi

, Kiss

, Haughey

, Rovinski

, Lanctot

(2013) Ceramides predict verbal memory performance in coronary artery disease patients undertaking exercise: A prospective cohort pilot study, BMC Geriatr 13, 135.

10.

Chan

, Saleem

, Herrmann

, Mielke

, Haughey

, Oh

, Kiss

, Lanctot

(2018) Ceramide accumulation is associated with declining verbal memory in coronary artery disease patients: An observational study, J Alzheimers Dis 64, 1235–1246.

11.

Saleem

, Herrmann

, Dinoff

, Mielke

, Oh

, Shammi

, Cao

, Venkata

SLV

, Haughey

, Lanctot

(2017) A lipidomics approach to assess the association between plasma sphingolipids and verbal memory performance in coronary artery disease patients undertaking cardiac rehabilitation: A C18:0 signature for cognitive response to exercise, J Alzheimers Dis 60, 829–841.

12.

Kasumov

, Solomon

, Hwang

, Huang

, Haus

, Zhang

, Kirwan

(2015) Improved insulin sensitivity after exercise training is linked to reduced plasma C14:0 ceramide in obesity and type 2 diabetes, Obesity (Silver Spring) 23, 1414–1421.

13.

Cacciatore

, Abete

, Maggi

, Luchetti

, Calabrese

, Viati

, Leosco

, Ferrara

, Vitale

, Rengo

(2004) Disability and 6-year mortality in elderly population. Role of visual impairment, Aging Clin Exp Res 16, 382–388.

14.

Milani

, Lavie

, Mehra

, Ventura

(2006) Understanding the s of cardiopulmonary exercise testing, Mayo Clin Proc 81, 1603–1611 basic.

15.

Zivelin

, Rosenberg

, Peretz

, Amit

, Kornbrot

, Seligsohn

(1997) Improved method for genotyping apolipoprotein E polymorphisms by a PCR-based assay simultaneously utilizing two distinct restriction enzymes, Clin Chem 43, 1657–1659.

16.

Hachinski

, Iadecola

, Petersen

, Breteler

, Nyenhuis

, Black

, Powers

, DeCarli

, Merino

, Kalaria

, Vinters

, Holtzman

, Rosenberg

, Dichgans

, Marler

, Leblanc

(2006) National Institute of Neurological Disorders and Stroke-Canadian Stroke Network vascular cognitive impairment harmonization standards, Stroke 37, 2220–2241.

17.

Benedict

BVMT-R Brief Visuospatial Memory Test-Revised, Ann Arbor Publishers.

18.

Joy

, Kaplan

, Fein

(2004) Speed and memory in the WAIS-III Digit Symbol–Coding subtest across the adult lifespan, Arch Clin Neuropsychol 19, 759–767.

19.

Gaudino

, Geisler

, Squires

(1995) Construct validity in the Trail Making Test: What makes Part B harder?, J Clin Exp Neuropsychol 17, 529–535.

20.

Bandaru

, Mielke

, Sacktor

, McArthur

, Grant

, Letendre

, Chang

, Wojna

, Pardo

, Calabresi

, Munsaka

, Haughey

(2013) A lipid storage-like disorder contributes to cognitive decline in HIV-infected subjects, Neurology 81, 1492–1499.

21.

Xia

, Sinelnikov

, Han

, Wishart

(2015) MetaboAnalyst 3.0–making metabolomics more meaningful, Nucleic Acids Res 43, W251–257.

22.

Cogo-Moreira

, Swardfager

(2019) On mediation models in clinical neurology studies, JAMA Neurol 76, 116–117.

23.

Preacher

, Hayes

(2004) SPSS and SAS procedures for estimating indirect effects in simple mediation models, Behav Res Methods Instrum Comput 36, 717–731.

24.

Haus

, Kashyap

, Kasumov

, Zhang

, Kelly

, Defronzo

, Kirwan

(2009) Plasma ceramides are elevated in obese subjects with type 2 diabetes and correlate with the severity of insulin resistance, Diabetes 58, 337–343.

25.

Laaksonen

, Ekroos

, Sysi-Aho

, Hilvo

, Vihervaara

, Kauhanen

, Suoniemi

, Hurme

, Marz

, Scharnagl

, Stojakovic

, Vlachopoulou

, Lokki

, Nieminen

, Klingenberg

, Matter

, Hornemann

, Juni

, Rodondi

, Raber

, Windecker

, Gencer

, Pedersen

, Tell

, Nygard

, Mach

, Sinisalo

, Luscher

(2016) Plasma ceramides predict cardiovascular death in patients with stable coronary artery disease and acute coronary syndromes beyond LDL-cholesterol, Eur Heart J 37, 1967–1976.

26.

Mielke

, Bandaru

, Haughey

, Rabins

, Lyketsos

, Carlson

(2010) Serum sphingomyelins and ceramides are early predictors of memory impairment, Neurobiol Aging 31, 17–24.

27.

Santiago

, Herrmann

, Swardfager

, Saleem

, Oh

, Black

, Lanctot

(2015) White matter microstructural integrity is associated with executive function and processing speed in older adults with coronary artery disease, Am J Geriatr Psychiatry 23, 754–763.

28.

Santiago

, Herrmann

, Swardfager

, Saleem

, Oh

, Black

, Bradley

, Lanctot

(2018) Subcortical hyperintensities in the cholinergic system are associated with improvements in executive function in older adults with coronary artery disease undergoing cardiac rehabilitation, Int J Geriatr Psychiatry 33, 279–287.

29.

Gonzalez

, Venkatraman

, An

, Landman

, Davatzikos

, Ratnam Bandaru

, Haughey

, Ferrucci

, Mielke

, Resnick

(2016) Peripheral sphingolipids are associated with variation in white matter microstructure in older adults, Neurobiol Aging 43, 156–163.

30.

Angevaren

, Aufdemkampe

, Verhaar

, Aleman

, Vanhees

(2008) Physical activity and enhanced fitness to improve cognitive function in older people without known cognitive impairment, Cochrane Database Syst Rev, CD005381.

31.

Renaud

, Maquestiaux

, Joncas

, Kergoat

, Bherer

(2010) The effect of three months of aerobic training on response preparation in older adults, Front Aging Neurosci 2, 148.

32.

Langlois

, Vu

, Chasse

, Dupuis

, Kergoat

, Bherer

(2013) Benefits of physical exercise training on cognition and quality of life in frail older adults, J Gerontol B Psychol Sci Soc Sci 68, 400–404.

33.

Bherer

(2015) Cognitive plasticity in older adults: Effects of cognitive training and physical exercise, Ann N Y Acad Sci 1337, 1–6.

34.

Dube

, Allison

, Rousson

, Goodpaster

, Amati

(2012) Exercise dose and insulin sensitivity: Relevance for diabetes prevention, Med Sci Sports Exerc 44, 793–799.

35.

Mielke

, Bandaru

, Haughey

, Xia

, Fried

, Yasar

, Albert

, Varma

, Harris

, Schneider

, Rabins

, Bandeen-Roche

, Lyketsos

, Carlson

(2012) Serum ceramides increase the risk of Alzheimer disease: The Women’s Health and Aging Study II, Neurology 79, 633–641.

36.

Grimm

, Hundsdorfer

, Grosgen

, Mett

, Zimmer

, Stahlmann

, Haupenthal

, Rothhaar

, Lehmann

, Patzold

, Zinser

, Tanila

, Shen

, Muller

, Grimm

, Hartmann

(2014) PS dependent APP cleavage regulates glucosylceramide synthase and is affected in Alzheimer’s disease, Cell Physiol Biochem 34, 92–110.

37.

Noel

, Ingrand

, Barrier

(2017) Ganglioside and related-sphingolipid profiles are altered in a cellular model of Alzheimer’s disease, Biochimie 137, 158–164.

38.

Herzer

, Meldner

, Rehder

, Grone

, Nordstrom

(2016) Lipid microdomain modification sustains neuronal viability in models of Alzheimer’s disease, Acta Neuropathol Commun 4, 103.

39.

Stiban

, Caputo

, Colombini

(2008) Ceramide synthesis in the endoplasmic reticulum can permeabilize mitochondria to proapoptotic proteins, J Lipid Res 49, 625–634.

40.

Bergman

, Brozinick

, Strauss

, Bacon

, Kerege

, Bui

, Sanders

, Siddall

, Kuo

, Perreault

(2015) Serum sphingolipids: Relationships to insulin sensitivity and changes with exercise in humans, Am J Physiol Endocrinol Metab 309, E398–408.

41.

Bergman

, Brozinick

, Strauss

, Bacon

, Kerege

, Bui

, Sanders

, Siddall

, Wei

, Thomas

, Kuo

, Perreault

(2016) Muscle sphingolipids during rest and exercise: A C18:0 signature for insulin resistance in humans, Diabetologia 59, 785–798.

42.

la Monte SM

, Re

, Longato

, Tong

(2012) Dysfunctional pro-ceramide, ER stress, and insulin/IGF signaling networks with progression of Alzheimer’s disease.S, J Alzheimers Dis 30(Suppl 2), 217–229.

43.

Hanamatsu

, Ohnishi

, Sakai

, Yuyama

, Mitsutake

, Takeda

, Hashino

, Igarashi

(2014) Altered levels of serum sphingomyelin and ceramide containing distinct acyl chains in young obese adults, Nutr Diabetes 4, e141.

44.

Brereton

, Lloyd

(2014) Partial least squares discriminant analysis: Taking the magic away, J Chemom 28, 213–225.