Metabolic Phenotypes of Carotid Atherosclerotic Plaques Relate to Stroke

Risk – An Exploratory Study

Panagiotis A Vorkas a, Joseph Shalhoub b, Matthew R Lewis a, Konstantina Spagou a,

Elizabeth J Want a, Jeremy K Nicholson a, Alun H Davies b, Elaine Holmes, a,*

a Section of Biomolecular Medicine, Division of Computational & Systems Medicine,

Department of Surgery & Cancer, Faculty of Medicine, Imperial College London, UK

b Academic 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: elaine.holmes@imperial.ac.uk; Tel: +44 (0) 20 7594 3220; Fax: +44 (0) 20 759 43226

Short title: Metabolic Signatures of Carotid Plaques

Original article

1

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

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 downregulation of the β-oxidation pathway in symptomatic plaques is

demonstrated for the first time. Metabolites associated with cell death were unaffected. The

metabolic signatures identified show potential as differential diagnostic biomarkers for

symptomatic plaques and may provide targets for pharmacotherapeutic intervention.

2

1

2

3

4

5

6

7

8

9

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 three

acylcarnitine species (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

3

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

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

4

1

2

3

4

5

6

7

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,1 which 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.3

A recent metabolic profiling study in blood plasma demonstrated promising results for

identifying patients with stroke recurrence.4 A subsequent study utilising a lipidomic

workflow to profile plaques reported successful discrimination between lipid signatures of

the stable and unstable parts of the same plaque tissue, but not between plaque 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 wide metabolome

coverage, multiple methods or techniques are required.6-8 Subsequent deconvolution 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 as tissue

homogenisation13 and metabolite extraction.14 Methods with the ability to handle intact tissue

5

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

function complementary to tissue extraction workflows15 and are preferred in translational

clinical settings.16

We hypothesised that stenosing carotid plaque tissue will exhibit a different metabolic

signature according to patient symptomatic status. Herein, we describe a pilot study

employing comprehensive untargeted metabolic phenotyping methodologies 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, 17 Implicated mechanistic processes

and candidate diagnostic signatures or metabolites, could function towards generating

hypotheses and candidate biomarkers relating to plaque rupture and stroke risk.

6

1

2

3

4

5

6

7

8

9

10

11

12

13

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 symptoms pertaining to the anterior circulation of

the cerebral hemisphere ipsilateral to the index carotid stenosis within the 6 months prior to

carotid endarterectomy. The patients with asymptomatic carotid stenosis in this study had

never experienced focal neurological symptoms at any time 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.

7

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

UPLC-MS Analyses Data 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.6 These two UPLC-MS methods combined can cover analytes in a range of

physicochemical properties, maximising metabolome coverage.6 Samples 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 functional groups which 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 provide a 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 quality molecules18 and

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 significant metabolites was conducted by matching mass

measurements to theoretical values from online databases: LipidMaps (www.lipidmaps.org),

Metlin (metlin.scripps.edu/index.php) and HMDB (www.hmdb.ca), and combining

information from: isotopic patterns, MSE spectra,19 in-house developed libraries, and

8

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

matching of experimental MS/MS spectra to MS/MS spectra from the Metlin database and

published literature.

9

1

2

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 3rd

principal 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 were PC(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 by palmitoylcarnitine 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 PCA as can be visualized in the scores

plots shown in Figure 2.C and D. For positive mode, separation was achieved in the 1st and

2nd principal components (Figure 2.C), while for negative mode in the first three principal

10

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

components (Figure 2.D). Representative chromatograms of methods used can be viewed in

the Supplementary Material Figure IV and Figure V. Model variation was 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).

11

1

2

3

4

5

6

7

8

9

10

11

Discussion

Here we describe a pilot study demonstrating potential in discriminating 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, 21 The 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. One issue is the lack of follow-up data, since the plaque can no longer be

responsible for any adverse health events following its removal at endarterectomy.

Additionally, detected metabolic alterations of the unstable plaque could be debated as being

as much the cause of instability manifestation as the effect. Nonetheless, characterizing

discriminant metabolites as an effect of intra-plaque haemorrhage and subsequent stabilising

wound healing, could still prove valuable in stratifying patients at 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,

12

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

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 also utilised as diagnostic biomarkers for

in 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 dysregulation prior to hypothesising the role of AA.

Nonetheless, these findings are in agreement with literature, as reviewed by Libby et al.22

An additional 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, unesterified carnitine was unaffected. Mitochondrial dysregulation is known to be

involved in atherosclerosis.23 Moreover, AcCs have been previously demonstrated having

significantly altered levels in stenosing atherosclerotic plaques, although not in a

symptomatic against asymptomatic setting and with a different pattern.11 However, this is to

our knowledge the first time AcCs have been connected to plaque rupture risk.

Cell death is the physiological cell process. Manifestation of amplified cell death processes

has been proposed as a contributor towards the formation of the advanced atherosclerotic

vulnerable plaque.24 Cell death related lipid species, such as ceramides,25 were amongst the

13

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

detected molecules in the analysed samples. Ceramides were 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 ceramides nor their direct

products sphingomyelins contributed to the separation between the symptomatic and

asymptomatic groups in the PCA. This observation requires further investigation in order to

clarify the 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.

14

1

2

3

4

5

6

7

8

9

10

11

12

13

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.

15

1

2

3

4

5

6

7

8

9

10

11

12

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

8. Saric J, Want EJ, Duthaler U, Lewis M, Keiser J, Shockcor JP, et al. Systematic evaluation of extraction methods for multiplatform-based metabotyping: Application to the fasciola hepatica metabolome. Analytical chemistry. 2012;84:6963-6972

9. Smith CA, Want EJ, O'Maille G, Abagyan R, Siuzdak G. Xcms: Processing mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and identification. Analytical chemistry. 2006;78:779-787

10. Trygg J, Holmes E, Lundstedt T. Chemometrics in metabonomics. Journal of proteome research. 2007;6:469-479

11. Vorkas PA, Shalhoub J, Isaac G, Want EJ, Nicholson JK, Holmes E, et al. Metabolic phenotyping of atherosclerotic plaques reveals latent associations between free cholesterol and ceramide metabolism in atherogenesis. Journal of proteome research. 2015;14:1389-1399

12. Vorkas PA, Isaac G, Holmgren A, Want EJ, Shockcor JP, Holmes E, et al. Perturbations in fatty acid metabolism and apoptosis are manifested in calcific coronary artery disease: An exploratory lipidomic study. International journal of cardiology. 2015;197:192-199

13. Geier FM, Want EJ, Leroi AM, Bundy JG. Cross-platform comparison of caenorhabditis elegans tissue extraction strategies for comprehensive metabolome coverage. Analytical chemistry. 2011;83:3730-3736

14. Anwar MA, Vorkas P, Li JV, Want E, Davies AH, Holmes E. Optimization of metabolite extraction of human vein tissue for ultra performance liquid chromatography-mass spectrometry and nuclear magnetic resonance-based untargeted metabolic profiling.

16

1

23456789

1011121314151617181920212223242526272829303132333435363738394041424344

Analyst. 201515. Beckonert O, Coen M, Keun HC, Wang Y, Ebbels TM, Holmes E, et al. High-resolution

magic-angle-spinning nmr spectroscopy for metabolic profiling of intact tissues. Nat Protoc. 2010;5:1019-1032

16. Anwar MA, Shalhoub J, Vorkas PA, Lim CS, Want EJ, Nicholson JK, et al. In-vitro identification of distinctive metabolic signatures of intact varicose vein tissue via magic angle spinning nuclear magnetic resonance spectroscopy. European journal of vascular and endovascular surgery : the official journal of the European Society for Vascular Surgery. 2012;44:442-450

17. Want EJ, Wilson ID, Gika H, Theodoridis G, Plumb RS, Shockcor J, et al. Global metabolic profiling procedures for urine using uplc-ms. Nat Protoc. 2010;5:1005-1018

18. Mahieu NG, Huang X, Chen YJ, Patti GJ. Credentialing features: A platform to benchmark and optimize untargeted metabolomic methods. Analytical chemistry. 2014;86:9583-9589

19. Plumb RS, Johnson KA, Rainville P, Smith BW, Wilson ID, Castro-Perez JM, et al. Uplc/ms(e); a new approach for generating molecular fragment information for biomarker structure elucidation. Rapid Commun Mass Spectrom. 2006;20:1989-1994

20. Chatzikonstantinou A, Wolf ME, Schaefer A, Hennerici MG. Asymptomatic and symptomatic carotid stenosis: An obsolete classification? Stroke Res Treat. 2012;2012:340798

21. Thapar A, Jenkins IH, Mehta A, Davies AH. Diagnosis and management of carotid atherosclerosis. BMJ. 2013;346:f1485

22. Libby P. Current concepts of the pathogenesis of the acute coronary syndromes. Circulation. 2001;104:365-372

23. Dromparis P, Michelakis ED. Mitochondria in vascular health and disease. Annu Rev Physiol. 2013;75:95-126

24. Tabas I. Macrophage apoptosis in atherosclerosis: Consequences on plaque progression and the role of endoplasmic reticulum stress. Antioxid Redox Signal. 2009;11:2333-2339

25. Taha TA, Mullen TD, Obeid LM. A house divided: Ceramide, sphingosine, and sphingosine-1-phosphate in programmed cell death. Biochim Biophys Acta. 2006;1758:2027-2036

17

123456789

101112131415161718192021222324252627282930313233

34

Figure Legends:

Figure 1: A schematic showing how the carotid plaque specimen was sectioned for analysis.

Segments A and B underwent separate metabolite extraction and analysis. The yellow colour

represents the stenosing plaque. UPLC: ultra performance liquid chromatography, MS: mass

spectrometry.

Figure 2: Principal component analysis scores plots of atherosclerotic plaque tissue extracts

showing differentiation of symptomatic and asymptomatic plaque tissue samples for:

Lipidomic analysis of (A) positive and (B) negative ionization mode. Polar metabolic

phenotyping of aqueous extracts, using hydrophilic interaction liquid chromatography

coupled to mass spectrometry (HILIC-MS) in (C) positive and (D) negative ionization mode.

The quality control (QC) samples are denoted in green and present a good indication of the

reproducibility of the methodology and stability of the specific run. Samples obtained from

the same plaque tissue are denoted by the same alphanumeric. Sample γ is represented by

only one biological replicate.

Figure 3: Box plots of statistically significant metabolites as demonstrated by univariate

statistics, from data of both organic and aqueous atherosclerotic plaque extracts. Two-tailed t-

tests were conducted, assuming unequal variance.

18

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19