Ceramides: shared lipid biomarkers of cardiovascular disease and schizophrenia.

Abstract

Schizophrenia, although a debilitating mental illness, greatly affects the individuals’ physical health, as well. One of the leading somatic comorbidities associated with schizophrenia is cardiovascular disease, which has been estimated to be one of the leading causes of excess mortality in patients diagnosed with schizophrenia. Although the shared susceptibility to schizophrenia and cardiovascular disease is well established, the mechanisms linking these two disorders are not well understood. Genetic studies have hinted toward shared lipid metabolism abnormalities co-occurring in the two disorders, while lipid compounds have emerged as a prognostic markers for cardiovascular disease. In particular, three ceramide species in the blood plasma, Cer(d18:1/16:0), Cer(d18:1/18:0), and Cer(d18:1/24:1), have been robustly linked to the latter disorder. In this work, we assessed the alterations of these three ceramide species in schizophrenia patients compared to healthy controls. To this end, we measured the abundances of these three ceramide species in a cohort of 82 patients with schizophrenia and 138 controls without a psychiatric diagnosis, and validated the results using an independent cohort of 26 patients with schizophrenia, 55 control individuals, and 19 patients experiencing a first psychotic episode. We found significant alterations for all three ceramide species and a particularly strong difference in concentrations between psychiatric patients and controls for the ceramide species Cer(d18:1/18:0).

Full Text

Introduction.

Schizophrenia is a debilitating mental illness affecting the individual’s thinking, perception, emotional response, as well as daily functioning, in general. Individuals diagnosed with schizophrenia are also at increased risk of somatic comorbidities, such as cardiovascular disease (CVD), type 2 diabetes, and obesity 1,2. It has been estimated that patients with schizophrenia have a 2-fold increase in mortality compared to the general population 3.  Most of this excess mortality can be attributed to physical illnesses, with CVD being one of the leading causes of death 3–5.

Understanding the mechanisms of CVD development in individuals with schizophrenia is of extreme importance for the management and treatment of the disorder. While lifestyle and antipsychotic medication can affect the risk of CVD 5, there seems to be an intrinsic and complex relationship between this physical disorder and schizophrenia. Genetic pleiotropy has been reported for schizophrenia and cardiometabolic abnormalities 6–10, implying shared genetic risk factors for CVD and SCZ. Stratification by genetic susceptibility to cardiometabolic abnormalities has also been proposed to separate SCZ patients into subgroups with differing metabolic profiles 11. At the same time, even though the genetic studies hint toward a shared susceptibility toward both SCZ and CVD, the mechanisms linking these two disorders are difficult to untangle from genetic information alone. In this respect, quantifying metabolic profiles of compounds associated with CVD in individuals diagnosed with SCZ can deepen the understanding of this connection by identifying the abnormalities common to the two disorders. Lipids, in particular, have been shown to undergo substantial alterations in schizophrenia 12. Further, genetic variants exhibiting pleiotropic effects in SCZ and CVD are especially enriched in genes involved in lipid metabolism 9,10.

While classical lipid measurements, such as low-density lipoproteins (LDL), high-density lipoproteins (HDL), and total triglycerides are well known to be associated with both CVD 13 and SCZ 14–16, advances in lipidomics have provided the opportunity to expand the scope of metabolic profiling and deepen our understanding of lipid alterations in disease. Ceramides, a class of lipids that consist of a sphingosine backbone, connected by an amide bond to a fatty acid (FA) chain of varying length, have been shown to predict cardiovascular events more effectively than classical lipid measurements 17. In particular, three lipid species, Cer(d18:1/16:0), Cer(d18:1/18:0), and Cer(d18:1/24:1), were proposed as promising biomarkers of CVD 18. To assess whether disruption in metabolism of these ceramides is also characteristic of patients with SCZ, we investigated the levels of these three lipids in 82 patients diagnosed with SCZ in comparison to 138 control (CTL) individuals without a psychiatric diagnosis. We further validated the results on an independent dataset of 26 SCZ patients, and 55 CTL patients, as well as 19 patients exhibiting first episode of psychosis (FEP).



Results

We collected the blood plasma samples for the main cohort comprising 82 patients with schizophrenia (SCZ) and 138 control individuals (CTL) (SCZ: age 31 ± 8, 23% female; CTL: age 30 ± 8, 22% female). We further collected a validation cohort including samples from 26 SCZ patients, 19 patients experience a first psychotic episode (FEP), and 55 CTL individuals (SCZ: age 35 ± 12, 58% female ; FEP: age 27 ± 7, 53% female; CTL: age 32 ± 8, 35% female). We extracted lipids from the samples in a randomized order and measured lipid abundances using liquid chromatography coupled with mass spectrometry (LC-MS). We determined the molecular ions corresponding to Cer(d18:1/16:0), Cer(d18:1/18:0), and Cer(d18:1/24:1) based on their fragmentation profiles (LC-MS2).

After normalizing the lipid abundances of the two cohorts using reference plasma samples, we compared the abundances of the three ceramide compounds between psychiatric and control groups within each cohort. In the main cohort, all three ceramides, Cer(d18:1/16:0), Cer(d18:1/18:0), and Cer(d18:1/24:1) displayed statistically significant abundance differences between SCZ and CTL, with all three compounds showing higher abundance in SCZ (Welch t-test, p = 4e-6, 3e-15, 7e-6 for Cer(d18:1/16:0), Cer(d18:1/18:0), and Cer(d18:1/24:1), respectively) (Figure 1; Table 1). Similarly, the intensities of all three ceramides were elevated in SCZ and FEP patients compared to CTL individuals in the validation cohort. Further, despite reduction in statistical power due to smaller sample size, abundance of all three ceramides differed significantly between FEP and CTL samples (Welch t-test, p = 0.0125, 5e-6 and 0.0255, respectively), as well as for Cer(d18:1/18:0) – between SCZ and CTL samples (Welch t-test, p = 0.0012; Figure 1; Table 1).

We next tested the possibility to discriminate between the two groups of individuals, psychiatric patients and controls, based on the abundances of each of the three ceramides. The abundances of each compound, in conjunction with an abundance threshold, represented a naive classifier, with individuals with abundances higher than the cutoff threshold classified as the psychiatric patient group, and individuals with abundances lower the cutoff – as CTL group. To assess the predictive power of these naive classifiers, we used Area Under the Curve of the Receiver Operator Characteristic (AUC ROC), which reflects, for the different cutoff thresholds, the relationship between the ratio of false positive and true positive classifications. A large AUC (with maximum 1) corresponds to a classifier for which, with an appropriate cutoff threshold, the ratio of true positives is higher, while the ratio of false positives remains low. Of the three ceramides, Cer(d18:1/18:0) showed best performance, indicated by its highest ROC AUC score (Cer(d18:1/16:0): AUC = 0.68; Cer(d18:1/18:0): AUC = 0.81; Cer(d18:1/24:1): AUC = 0.61; Figure 2; Table 1). Next, we assessed the performance of these same classifiers when applied to the classification of psychiatric patients versus CTL in the validation cohort. For the three ceramides, the separation of SCZ and CTL was slightly worse than in the main cohort (Cer(d18:1/16:0): AUC = 0.57; Cer(d18:1/18:0): AUC = 0.73; Cer(d18:1/24:1): AUC = 0.61; Figure 2; Table 1). The classification of FEP versus CTL individuals, however, displayed no accuracy decline compared to the main cohort performance (Cer(d18:1/16:0): AUC = 69; Cer(d18:1/18:0): AUC = 81; Cer(d18:1/24:1): AUC = 67; Figure 2; Table 1), indicating the good generalization capabilities of the separation models.

We next used the abundances of the three ceramides together to construct a logistic regression model separating psychiatric patients from CTL individuals. The model trained on the samples from the main cohort showed similar accuracy to the best-performing individual ceramide, Cer(d18:1/18:0), but its accuracy in the validation dataset was higher (main cohort, SCZ vs CTL: AUC = 0.82; validation cohort, SCZ vs CTL: AUC = 0.81; FEP vs CTL: AUC = 0.83).

Discussion

Schizophrenia (SCZ) is burdened by a range of metabolic abnormalities. In this work, we have assessed the abundance levels of three ceramide species, Cer(d18:1/16:0), Cer(d18:1/18:0), and Cer(d18:1/24:1), in the blood plasma of patients suffering from SCZ and control individuals with no psychiatric diagnosis. These ceramide species were chosen on account of their emerging role as biomarkers of cardiovascular disease (CVD) 18. Our results demonstrate that ceramide alterations might be an abnormality common to both SCZ and CVD.

While CVD progression can be affected by environmental factors such as lifestyle, weight, and smoking 5, there is evidence for a genetic basis of CVD 19. There are also reports of shared genetic risk factors for both SCZ and CVD 6–8, with lipid-related genes being highlighted in the literature 9,10. Likewise, while ceramides can be affected by lifestyle, diet, and other environmental factors, the human plasma lipidome, and ceramides in particular, were found to have a marked genetic component 20–22. The identifications of genes, such as SPTLC3, that show association with both CVD and blood plasma ceramide levels 20, but were also reported in connection to SCZ 23,24, demonstrates the possibility of an intrinsic connection between SCZ, CVD, and ceramide metabolism abnormalities beyond the superficial influences of environmental factors.

Ceramides are structural elements of the eukaryotic cell membranes. They also have signaling functions, being involved in processes such as apoptosis and inflammation 25,26. Moreover, ceramides are enriched in neural tissues and are vital for the normal functioning of brain cells, in particular 27–29. While the exact role of ceramides is not fully elucidated, perhaps the most convincing evidence of their important functional role are the numerous reports of ceramide alterations in various disorders besides CVD, such as diabetes, insulin resistance, neurodegenerative disorders, and multiple sclerosis 30,31. Similarly to CVD, specific ceramide lipid species quantified in blood plasma, including the Cer(d18:1/18:0) compound assessed in this study, were proposed as a biomarker of depression 32. In schizophrenia, ceramide alterations have been reported in the brain 33, and mechanisms linking metabolic abnormalities and SCZ through sphingolipids have been proposed 12,34.

While ceramide in the blood plasma have been extensively studied as biomarkers of depression 32, surprisingly, we haven’t found studies that focused on ceramides in the blood plasma of SCZ patients explicitly. Nevertheless, alterations in blood plasma of SCZ patients were investigated and reported for other lipid classes 35–47. In our work, we demonstrated that the abundance of one ceramide previously reported as reliable depression and CVD biomarker  18,32 – Cer(d18:1/18:0) – was sufficient to distinguish SCZ from CTL with an AUC higher than 0.8, the result further reproduced in an independent cohort of first psychotic episode patients. We also found that the discriminatory power of the other two assessed ceramides, Cer(d18:1/16:0) and Cer(d18:1/24:1) was lower, but their abundance differences between SCZ and CTL groups nonetheless statistically significant. We found that the combination of abundance values for the three ceramides in a single model produced moderate improvement in the classification performance compared to the single compound Cer(d18:1/18:0). The model performance level achieved in this study is not, however, the best reported for SCZ and CTL group separation – models based on other lipid classes allegedly reached AUC = 0.98 43, indicating extensive disruption of lipid metabolism in SCZ.

Interplay between SCZ and metabolic abnormalities, such as CVD, remains poorly understood. Genetic studies have hinted toward a connection between these disorders through common lipid metabolism alterations 9,10. Specific ceramide species have been proposed as biomarkers of CVD, and we found these same ceramides to be altered in SCZ, as well. While the mechanism of ceramide alterations in disease is not clear, the systematic study of lipid and other metabolic disruptions that co-occur with SCZ and CVD might help elucidate the connection between these disorders in the future.

 

Methods.

 

Main sample cohort

Subjects included in this cohort were inpatients and recruited from of the Mental Health Research Center, Moscow. The cohort included adult (≥18 years) participants with a diagnosis of SCZ (n = 82; age 31.2 ± 8.4; 23% female), BPD (n = 36; age 28.2 ± 9.1; 67% female) according to the ICD-10 criteria. Sample collection was performed at the Neuroimmunology Laboratory of the Mental Health Research Center, Moscow. The control group consisted of healthy volunteers from Mental Health Research Centre who had no signs of psychiatric disorders and met the same exclusion criteria  (age<18 years, family history of any psychiatric disorder, severe somatic and neurological illness, recent surgery, pregnancy, substance and alcohol abuse).  (n = 138; age 29.5 ± 8.3; 22% female). Patient evaluation was performed by board-certified psychiatrists of the same center. The study was approved by the local ethics committee of the Mental Health Research Center  (Protocol No. 281; 05/05/2016). Informed consent was obtained from all participants. The entire study was conducted in line with the World Medical Association Declaration of Helsinki formulating ethical principles for medical research involving human subjects.

Validation sample cohort

Participants were inpatients, diagnosed with a first episode of psychosis (n=19, age 27 ± 7, 53% female), and schizophrenia (n=26, age 35 ± 12, 58% female ) recruited at Mental-health Clinic No. 1 named after N.A. Alexeev. All participants underwent complete diagnostic evaluation and individuals with intellectual disabilities, substance abuse and dependence and any comorbid severe somatic or neurological disorder were excluded. The control group (n=55; age 32 ± 8, 35% female) consisted of healthy volunteers who had no signs of psychiatric disorders and met the same exclusion criteria  (age<18 years, family history of any psychiatric disorder, severe somatic and neurological illness, recent surgery, pregnancy, substance and alcohol abuse). Informed consent was obtained from all participants. The protocol of this study was approved by the Interdisciplinary Ethics Committee, Moscow (22/07/2017).

 

Although the FEP patients used in our study were not drug-naïve, they had been receiving antipsychotic medication for a short period of time, from several days to a week.


Lipid measurements

Plasma was obtained from peripheral venous blood in the morning from individuals that underwent an overnight fast. Plasma samples were collected in 4 ml Vacutainer tubes containing the chelating agent ethylenediaminetetraacetic acid (EDTA) (BD Vacutainer, Franklin Lakes, NJ, USA). Tubes were centrifuged at 4°C at 1100g for 15 min. The supernatant was stored immediately in 500μl aliquots at -80°C.

Lipids abundance were measured using LC-MS in negative ionization mode for the main dataset and validation dataset in two distinct experimental runs. Sample processing, LC-MS measurements, and data preprocessing was performed as described in 48.

Ceramide species were annotated according to tandem mass spectrometry experiment on pooled plasma samples, and ceramide species were validated by the presence of characteristic sphingoid base fragments.

To normalize lipid abundances between the main and validation datasets, 20 reference samples from the main dataset were re-measured together with the validation dataset. The mean abundances of these samples in the validation and main datasets were used to normalize the abundances of the validation dataset: for each lipid, the differences between these means were substracted from the abundances of each sample in the validation dataset to produce normalized abundances.

 

Statistical analysis

Statistical tests were performed on the base-two log transformed abundances to ensure normal distribution. Welch t-test of unequal variance was used to assess statistical differences. However, bar plots were visualized in the original scale.

 Analysis, including statistical analysis, performance calculation for naïve classifiers, and logistic regression, was performed using python scipy and sklearn libraries. The prediction models were trained on the main dataset, and tested on both the main and validation datasets.

 

Figure 1. The median and interquartile ranges in sample groups for each of the ceramides’ abundances: Cer(d18:1/16:0), Cer(d18:1/18:0), and Cer(d18:1/24:1).
The median abundances in each sample group correspond to bar heights, and the interquartile range is indicated by the error bars. For each of the ceramide, sample groups are indicated on the bottom, from left to right: CTL main cohort, SCZ main cohort, CTL validation cohort, SCZ validation cohort, FEP validation cohort. The p-values of the Welch t-test and the fold-changes in abundances for the comparisons between psychiatric and CTL groups are indicated on the plots.

 

Figure 2. The receiver operating characteristic (ROC) curves for the naïve classifiers that were based on each of the ceramides’ abundances: Cer(d18:1/16:0), Cer(d18:1/18:0), and Cer(d18:1/24:1).
Each ceramide species, in conjunction with an abundance threshold, represents a naive classifier, with individuals with abundances higher than the cutoff threshold classified as the psychiatric patient group, and individuals with abundances lower the cutoff – as CTL group. For each ceramide. the ratio of the true positive classifications is plotted against the ratio of false positive classifications for the different cutoff thresholds. Colors correspond to the different classification tasks: main cohort SCZ vs CTL (dark red), validation cohort FEP versus CTL (red), validation cohort SCZ versus CTL (orange). The corresponding areas under the curves (AUCs) are indicated on the plots.

 

Table 1

P-values, fold-changes, as well as ROC AUC values for the comparisons between psychiatric groups and CTL

×

About the authors

Anna Tkachev

Author for correspondence.
Email: anna.tkachev@skolkovotech.ru
Russian Federation

Elena Stekolshchikova

Email: E.Stekolschikova@skoltech.ru
Russian Federation

Anna Morozova

Email: hakurate77@gmail.com
Russian Federation

Nickolay Anikanov

Email: koenzyme@mail.ru
Russian Federation

Yana Zorkina

Email: zorkina.ya@serbsky.ru
Russian Federation

Polina Alekseeva

Email: p249703a@yandex.ru
Russian Federation

Elena Khobta

Email: dr.khobta@gmail.com
Russian Federation

Denis Andreyuk

Email: Denis.s.andreyuk@yandex.ru
Russian Federation

Svetlana Zozulya

Email: s.ermakova@mail.ru

Aleksandra Barkhatova

Email: abarkhatova@yandex.ru
Russian Federation

Tatiana Klyushnik

Email: klushnik2004@mail.ru
Russian Federation

Alexander Reznik

Email: a.m.reznik1969@gmail.com
Russian Federation

Georgiy Kostyuk

Email: kgr@yandex.ru
Russian Federation

Philipp Khaitovich

Email: khaitovich@eva.mpg.de
Russian Federation

Supplementary files

There are no supplementary files to display.

Statistics

Views

Abstract: 76

PDF (English): 7

Article Metrics

Metrics Loading ...

Dimensions

PlumX


Copyright (c) Tkachev A., Stekolshchikova E., Morozova A., Anikanov N., Zorkina Y., Alekseeva P., Khobta E., Andreyuk D., Zozulya S., Barkhatova A., Klyushnik T., Reznik A., Kostyuk G., Khaitovich P.

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.

This website uses cookies

You consent to our cookies if you continue to use our website.

About Cookies