Metabolic Phenotypes of Carotid Atherosclerotic Plaques Relate to Stroke Risk – An Exploratory Study
Panagiotis A Vorkas a, Joseph Shalhoub b, Matthew R Lewis a, KonstantinaSpagou a, Elizabeth J Want a, Jeremy K Nicholson a, Alun H Daviesb, Elaine Holmes,a,*
aSection of Biomolecular Medicine, Division of Computational & Systems Medicine, Department of Surgery & Cancer, Faculty of Medicine, Imperial College London, UK
bAcademic Section of Vascular Surgery, Division of Surgery, Department of Surgery & Cancer, Faculty of Medicine, Imperial College London, UK
* To whom correspondence should be addressed:
Imperial College London, South Kensington Campus, Imperial College Road, Sir Alexander Fleming Building, Biomolecular Medicine, SW7 2AZ, London, UK
Email:; Tel:+44 (0)20 7594 3220; Fax:+44 (0) 20 759 43226
Short title: Metabolic Signatures of Carotid Plaques
Original article
1
What this paper adds
We demonstrate that the metabolic signature of carotid plaque tissue from patients with cerebrovascular symptoms significantly differs from carotid plaque tissue derived from asymptomatic patients.This was achieved by a comprehensive metabolic profiling application utilising ultra performance liquid chromatography coupled to mass spectrometry. The enhanced downregulationof the β-oxidation pathwayin symptomatic plaques is demonstrated for the first time.Metabolites associated with cell death were unaffected. The metabolic signatures identified showpotential as differential diagnostic biomarkers for symptomatic plaques and may provide targets for pharmacotherapeutic intervention.
1
Abstract
Objectives: Stroke is a major cause of death and disability. The fact that three-quarters of stroke patients will never have previously manifested cerebrovascular symptoms demonstrates the unmet clinical need for new biomarkers able to stratify patient risk and elucidation of the biological dysregulations.In this study, we assess the utility of comprehensive metabolic phenotyping to provide candidate biomarkers that relate to stroke risk in stenosing carotid plaque tissue samples.
Design:Carotid plaque tissue samples were obtained from patients with cerebrovascular symptoms of carotid origin (n=5), and asymptomatic patients (n=5). Two adjacent biological replicates were obtained from each tissue.
Materials and Methods: Organic and aqueous metabolite extracts were separately obtained and analysed using two ultra performance liquid chromatography coupled to mass spectrometry metabolic profiling methods. Multivariate and univariate tools were utilised for statistical analysis.
Results: The two studied groups demonstrated distinct plaque phenotypes using multivariate data analysis. Univariate statistics also revealed metabolites that differentiated the two groups with a strong statistical significance (p=10-4-10-5). Specifically, metabolites related to the eicosanoid pathway (arachidonic acid and arachidonic acid precursors), and threeacylcarnitinespecies (butyrylcarnitine, hexanoylcarnitine and palmitoylcarnitine), intermediates of the β-oxidation, were detected in higher intensities in symptomatic patients. However, metabolites implicated in the process of cell death, a process known to be upregulated in the formation of the vulnerable plaque, were unaffected.
Conclusions: Discrimination between symptomatic and asymptomatic carotid plaque tissue is demonstrated for the first time using metabolic profiling technologies. Two biological pathways (eicosanoid and β-oxidation) were implicated and will be further investigated.These results indicate that metabolic phenotyping should be further explored to investigate the chemistry of the unstable plaque, in the pursuit of candidate biomarkers for risk-stratification and targets for pharmacotherapeutic intervention.
Keywords: Embolic stroke; Metabolomics;Metabonomics; Metabolic profiling; Metabolic phenotyping;Lipidomics, Lipid profiling; Mass spectrometry
1
Introduction
According to the World Health Organization stroke is a major cause of death and disability. Patients with cerebrovascular symptoms of carotid origin are at high risk of a subsequent imminent life-threatening stroke,1which declines with time after symptom onset. However, three-quarters of stroke patients will have been previously asymptomatic.2 There is, therefore, an on-going clinical need to identify biological markers that can stratify plaque rupture risk.3A recent metabolic profiling study in blood plasma demonstrated promising results for identifying patients with stroke recurrence.4 A subsequent study utilising a lipidomicworkflow to profile plaques reported successful discrimination between lipid signatures of the stable and unstable parts of the same plaque tissue, but not betweenplaque tissues obtained from symptomatic and asymptomatic patients.5
Metabolic phenotyping relies on the use of modern chemical analytical instrumentation to detect metabolic alterations in a biological system. In order to achieve a widemetabolome coverage, multiple methods or techniques are required.6-8Subsequentdeconvolution and interpretation is conducted through data processing algorithms,9 statistical analysis and modelling,10, 11 followed by molecular structure assignment and biological pathway mapping.6, 11, 12
Analysis of tissue samples can provide candidate biomarkers for in vivo imaging and guide further targeted biomarker discovery studies in matrices such as blood and urine. Most importantly – in contrast to blood plasma/serum samples which provide a more systemic view – tissue samples can provide clear,disease-specific insight regarding biological mechanistic dysregulations.11However, the use of tissue for metabolic phenotyping can be challenging due to the additional steps required prior to analysis, such astissue homogenisation13 and metabolite extraction.14Methods with the ability to handle intact tissue function complementary to tissue extraction workflows15 and are preferred in translationalclinical settings.16
We hypothesised that stenosing carotid plaque tissue will exhibit a different metabolic signature according to patient symptomaticstatus. Herein, we describe a pilot study employing comprehensive untargeted metabolic phenotypingmethodologies in order to explore the ability to reveal metabolic signatures in stenosing carotid plaque tissue samples. Samples were obtained from patients who had recently (≤ 12 days) presented with a cerebrovascular event (high risk/symptomatic group), and of asymptomatic patients as the control group (low risk/asymptomatic group). Ultra performance liquid chromatography coupled to mass spectrometry (UPLC-MS) was the technique of choice utilised for the untargeted comprehensive metabolic profiling analysis.6, 17Implicated mechanistic processes and candidate diagnostic signatures or metabolites, could function towards generating hypotheses and candidate biomarkersrelating to plaque rupture and stroke risk.
1
Material and Methods
Patients
Atherosclerotic plaques were obtained from consenting patients and after research ethics committee approval (08/H0706/129), at the time of carotid endarterectomy surgery: 5 recently (≤ 12 days) symptomatic of cerebrovascular symptoms occurring in the territory of the ipsilateral carotid circulation and 5 asymptomatic. Patients were considered asymptomatic if they did not have any focal neurological symptomspertaining to the anterior circulation of the cerebral hemisphere ipsilateral to the index carotid stenosis within the 6 monthsprior to carotid endarterectomy. The patients with asymptomatic carotid stenosis in this study had never experienced focal neurological symptoms at anytime point prior to their carotid endarterectomy. There was no post-operative mortality amongst the patients enrolled. One symptomatic patient developed a post-operative haematoma which required operative evacuation on the first post-operative day. Patients’demographics can be found in Supplementary Material Table I.
Sample Preparation
Three transverse segments of stenosing carotid plaque tissue were obtained from each sample. The central slice was stored for imaging and staining purposes. The two slices flanking the central slice were placed into separate bead beating tubes, for tissue lysis and metabolite extraction. Two consecutive extractions were performed: for polar compounds (aqueous extracts), and lipophilic compounds (organic extracts).6 A detailed description of sample preparation is presented in Supplementary Methods. A schematic illustration of the sample preparation procedure is demonstrated in Figure 1.
UPLC-MS AnalysesData Processing and Statistics
An untargeted lipidomics reversed phase (RP)-UPLC-MS analysis was applied on the organic extracts.6 Respectively, an untargeted polar metabolic phenotyping method was employed for analysing the aqueous extracts using hydrophilic interaction liquid chromatography (HILIC)-UPLC-MS.6These two UPLC-MS methods combined can cover analytes in a range of physicochemical properties, maximising metabolome coverage.6Samples were analysed in both positive and negative electrospray ionization (ESI) modes. The two polarity modes generate complementary information due to preferential ionisation of metabolites (diminished ionisation can reduce sensitivity) according to their functionalgroupswhich carry the charge of the molecule. Data were processed using the XCMS package.9 The resulting feature intensities were normalised and imported into SIMCA-P+ 12.0.1 software (Umetrics, Sweden) for multivariate data analysis (MVDA). Principal components analysis (PCA) was used as an unsupervised MVDA method to visualize data. PCA can providea simplified overview of all the features detected for each sample and therefore uncover differential metabolic patterns. Additionally, all features were subjected to a 2-tailed t-test, assuming unequal variance, and were considered statistically significant for p<0.0001. The p-value cut-off was calculated based on the number of unique and of sufficient qualitymolecules18and after Bonferroni correction. Further information on UPLC-MS analyses and data processing can be found in Supplementary Methods.
Metabolite structural assignment
Structural assignment of statistically significantmetabolites was conducted by matching mass measurements to theoretical values from online databases: LipidMaps ( Metlin (metlin.scripps.edu/index.php) and HMDB ( and combining information from: isotopic patterns, MSE spectra,19 in-house developed libraries, and matching of experimental MS/MS spectra to MS/MS spectra from the Metlin database and published literature.
1
Results
Lipidomic analysis of Plaque Tissue Extracts
The PCA scores plots of both ionization modes showed group discrimination of symptomatic and asymptomatic samples in the 2nd principal component (Figure 2.A) and 2nd and 3rdprincipal components (Figure 2.B), for positive and negative modes, respectively.Representative chromatograms of methods used can be viewed in the Supplementary Material Figure I and Figure II. The features driving model variation were identified from loadings plots (Supplementary Material Figure III and Table II). These included phosphatidylcholines (PC), lysoPCs, phosphatidylethanolamines (PE), ceramides (Cer), sphingomyelins (SM), oxidised cholesterol esters (oxCE), triglycerides (TG), diglycerides (DG) and fatty acids.A panel of five metabolites appeared to be the major drivers of separation between the two groups based on the loadings of the PCA model (Supplementary Material Figure III.C). These werePC(16:0/20:4), PC(16:0/18:1), PE(18:1/18:0), arachidonic acid (AA), and an as yet unassigned feature, the levels of which were significantly higher in the symptomatic group.
Independent from MVDA, univariate statistics were applied to all features. Features with high statistical significance are presented in Figure 3 and Supplementary Material Table III. The highest statistical significance were presented bypalmitoylcarnitine and TG(58:6), with p=10-5 and p=7x10-5, respectively, and fold-changes of 2.5 and 3.1.
Polar Metabolic Phenotyping of Aqueous Plaque Extracts
The two disease groups showed discrimination with PCAas can be visualized in the scores plots shown inFigure 2.C and D. For positive mode, separation was achievedin the 1st and 2ndprincipal components (Figure 2.C), while for negative modein the first threeprincipal components (Figure 2.D).Representative chromatograms of methods used can be viewed in the Supplementary Material Figure IV and Figure V. Model variationwas induced by short-chain acylcarnitines (AcC), carnitine, lysoPCs, PCs, glycerophosphocholine, glycerophosphoethanolamine, glycerophosphoinositol, adenosine, inosine and uridine (Supplementary Material Figure VI and Table IV). The (iso-)butyrylcarnitine, lysoPC(O-16:0) and an unassigned feature, were the metabolites responsible for driving the separation of the groups.
Univariate statistics (Figure 3 and Supplementary Material Table III) detected two features with significantly higher intensities in the symptomatic group: hexanoylcarnitine (p=3x10-4; fold-change 1.9) and an unassigned feature eluting at 8.14 min with m/z of 645.3829(p=4x10-4; fold-change 3.5) (Figure 3).
1
Discussion
Here we describe a pilot study demonstrating potential indiscriminating between symptomatic and asymptomatic carotid atherosclerotic plaque tissue for the first time using a metabolic phenotyping strategy. Compared with asymptomatic individuals, patients with focal cerebrovascular symptoms (symptomatic) are at considerably higher risk of experiencing a stroke in the immediate period following symptom initiation.20, 21The use of tissue provides the advantage of disease specificity, which in turn can facilitate hypothesis generation.However, the use of plaque tissue in such a prognostic setting comes with several challenges. Oneissue is the lack of follow-up data, since the plaque can no longerbe responsible for any adverse health events following its removal atendarterectomy. Additionally, detected metabolic alterations of the unstable plaque could be debated as being as much the cause of instability manifestation as the effect.Nonetheless, characterizingdiscriminant metabolitesas an effect of intra-plaque haemorrhage and subsequent stabilising wound healing, could still prove valuable in stratifying patientsat risk of future stroke.
The current feasibility study will provide the necessary assurance and framework in order to invest in larger studies, preferably using biofluids(blood and urine) to obtain the necessary patient follow-up. Moreover, the information obtained from the current study could be used as guidance for targeting specific pathways hypothesised as being involved in plaque instability. This study is based on a relatively small numbers of patients and,although statistical analysis is clear, biological interpretation is made with caution. Nonetheless, additional confidence was provided by the fact that the metabolites which deferred between the groups were identified as being members of biological pathways recognised for their involvement in plaque rupture. Up to 50 features were structurally assigned to their corresponding metabolite.A number of them were driving the variation in the PCA models, but were not related to discrimination between the two phenotypic groups.
A number of assigned metabolites were shown driving the separation and being discriminatory between the two groups. Discriminating metabolites in tissue, further to their potential for bio-mechanistic implication, could be alsoutilised as diagnostic biomarkers forin vivo imaging.One of the most important differences identified between groups was the higher intensities of AA and PC(16:0/20:4), an AA bearing phosphatidylcholine, in symptomatic atherosclerosis. The PC(16:0/20:4) can release AA after being hydrolysed by the phospholipase A2 enzyme. The AA functions as precursor molecule of a wide spectrum of the inflammation-related eicosanoids. Eicosanoids, although structurally and biologically related, may have opposing inflammatory functions. It is therefore important to first elucidate the downstream implications of this dysregulationprior to hypothesising the role of AA. Nonetheless, these findings are in agreement with literature, as reviewed by Libby et al.22
Anadditional pathway detected significantly different as compared to the control asymptomatic group, is that of β-oxidation. Specifically, a complement of AcCs, namely (iso-)butyrylcarnitine, hexanoylcarnitine and palmitoylcarnitine -which can function as β-oxidation intermediates - were detected at higher intensities in symptomatic patients.On the contrary, unesterifiedcarnitinewas unaffected. Mitochondrial dysregulation is known to be involved in atherosclerosis.23Moreover, AcCs have been previously demonstrated having significantly altered levels in stenosing atherosclerotic plaques, although not in a symptomatic against asymptomatic settingand with a different pattern.11 However, this is to our knowledge the first timeAcCs have been connected to plaque rupture risk.
Cell death is the physiological cell process. Manifestation of amplified cell deathprocesses has been proposed as a contributor towards the formation of the advanced atherosclerotic vulnerable plaque.24Cell deathrelated lipid species, such as ceramides,25 were amongst the detected molecules in the analysed samples. Ceramideswere driving the biochemical variation in the PCA fitted models as demonstrated by the model loadings plots (Supplementary Material Figure III and Table II). However, neither ceramidesnor their direct products sphingomyelinscontributed to theseparation between the symptomatic and asymptomatic groups in the PCA. This observation requires further investigation in order to clarifythe origins of the variation induced by the levels of these lipid species and relevance to the process of cell death in the carotid stenosing plaque.
The findings demonstrated here, provide evidence of the potential metabolic profiling can offer in order to discriminate, in vitro, stable asymptomatic from unstable symptomatic carotid atherosclerosis. Results from these analyses support further use of the described methodologies in the context of a larger study of biological samples (biofluids and tissue) from symptomatic and asymptomatic patients, as well as hypothesis driven and targeted approaches.
1
Acknowledgements
PAV acknowledges the Royal Society of Chemistry for supporting his PhD studentship. JS acknowledges the Royal College of Surgeons of England Research Fellowship Scheme, Circulation Foundation, Rosetrees Trust, Graham Dixon Trust and Peel Medical Research Trust for supporting his PhD studentship. EJW would like to acknowledge Waters Corporation for her funding.
Funding
This study was supported by the Royal Society of Chemistry (Grant number: 09/G31C). Additional support was received by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Imperial College Healthcare NHS Trust and Imperial College London. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.
References
1.Ois A, Cuadrado-Godia E, Rodriguez-Campello A, Jimenez-Conde J, Roquer J. High risk of early neurological recurrence in symptomatic carotid stenosis. Stroke; a journal of cerebral circulation. 2009;40:2727-2731
2.Halliday A, Harrison M, Hayter E, Kong X, Mansfield A, Marro J, et al. 10-year stroke prevention after successful carotid endarterectomy for asymptomatic stenosis (acst-1): A multicentre randomised trial. Lancet. 2010;376:1074-1084
3.Shalhoub J, Sikkel MB, Davies KJ, Vorkas PA, Want EJ, Davies AH. Systems biology of human atherosclerosis. Vascular and endovascular surgery. 2014;48:5-17
4.Jove M, Mauri-Capdevila G, Suarez I, Cambray S, Sanahuja J, Quilez A, et al. Metabolomics predicts stroke recurrence after transient ischemic attack. Neurology. 2015;84:36-45
5.Stegemann C, Drozdov I, Shalhoub J, Humphries J, Ladroue C, Didangelos A, et al. Comparative lipidomics profiling of human atherosclerotic plaques. Circ Cardiovasc Genet. 2011;4:232-242
6.Vorkas PA, Isaac G, Anwar MA, Davies AH, Want EJ, Nicholson JK, et al. Untargeted uplc-ms profiling pipeline to expand tissue metabolome coverage: Application to cardiovascular disease. Analytical chemistry. 2015;87:4184-4193
7.Ivanisevic J, Zhu ZJ, Plate L, Tautenhahn R, Chen S, O'Brien PJ, et al. Toward 'omic scale metabolite profiling: A dual separation-mass spectrometry approach for coverage of lipid and central carbon metabolism. Analytical chemistry. 2013;85:6876-6884