Author + information
- Received October 6, 2012
- Revision received February 20, 2013
- Accepted March 21, 2013
- Published online December 1, 2013.
- Raphaël Duivenvoorden, MD∗,
- Diederik van Wijk, MD∗,
- Michael Klimas, PhD, MBA†,
- John J.P. Kastelein, MD, PhD∗,
- Erik S.G. Stroes, MD, PhD∗ and
- Aart J. Nederveen, PhD‡∗ ()
- ∗Department of Vascular Medicine, Academic Medical Center, Amsterdam, the Netherlands
- †Merck Research Laboratories, West Point, Pennsylvania
- ‡Department of Radiology, Academic Medical Center, Amsterdam, the Netherlands
- ↵∗Reprint requests and correspondence:
Dr. Aart J. Nederveen, Department of Radiology, Academic Medical Center, Z0-120, Meibergdreef 9, 1105 AZ Amsterdam, the Netherlands.
Objectives This study hypothesized that 1H magnetic resonance spectroscopy (1H-MRS) can identify carotid plaque cholesteryl ester in vivo in humans.
Background Liquid phase cholesteryl ester comprises a major fraction of atherosclerotic plaque, and its abundance is associated with plaque rupture and atherothrombosis. A noninvasive imaging technique to detect liquid cholesteryl ester that has been applied ex vivo is now demonstrated in vivo.
Methods 1H-MRS scans were obtained of carotid plaques of 35 subjects at 3.0 T. Turbo spin echo, black blood, T1-weighted images were acquired for localization. Spectra were acquired using a 2-dimensional point resolved spectroscopy sequence: repetition time/echo time = 1,100/30 ms, 5-mm slice thickness, 8 × 8-cm field of view, 16 × 16 matrix size, and 13-min acquisition time. Saturation bands were placed around the artery. Resonance of methylene protons and allylic methylene protons were assigned to 1.2 ppm and 2.0 ppm. The 2.0:1.2 ppm ratio was calculated to reflect the ratio of the fatty acid composition of plaque cholesteryl ester to that of triglycerides of perivascular tissue. We obtained spectra of lipid standards as a reference.
Results Our 1H-MRS data showed typical spectra of cholesteryl ester mixed with triglycerides, with intense resonance from methylene (1.2 ppm) and allylic methylene (2.0 ppm) protons. The average 2.0:1.2 ppm ratio was 0.10 ± 0.03. The 2.0:1.2 ppm ratio correlated with the plaque tissue volume to perivascular tissue volume ratio (Spearman rho = 0.55, p = 0.02), suggesting that more 1H-MRS signal was obtained from cholesteryl ester when the 1H-MRS voxel comprised more plaque tissue. Repeat 1H-MRS scans in 4 subjects showed an intraclass correlation coefficient of 0.92 (95% prediction intervals: 0.40 to 0.99), indicating good reproducibility. Seventeen of the 35 1H-MRS spectra were of adequate quality for analysis.
Conclusions In vivo image-guided 1H-MRS for detection of liquid phase cholesteryl ester in carotid atherosclerotic plaques in humans is feasible.
Cholesteryl ester in the liquid phase comprises a major portion of the lipid-rich core of atherosclerotic plaques, and its abundance has been associated with plaque rupture and formation of atherothrombosis (1). The unique lipid composition of atheroma, with a high proportion of cholesteryl ester relative to other lipids, is mainly a result of the infiltration and retention of cholesterol carrying low-density lipoprotein that is transported through injured endothelium and penetrates the vascular tissue (2). Furthermore, macrophages ingest and digest lipoprotein cholesterol and store it as cholesteryl ester in their cytoplasm as droplets (1). Reversibility of atherosclerotic lesions is thought to be limited by removal of cholesteryl esters in the liquid phase, whereas dissolution of crystalline deposits are thought to be minimal (1).
Quantification of liquid cholesteryl ester content of plaque is therefore valuable for the assessment of plaque vulnerability and for monitoring the effects of antiatherosclerotic drugs. However, techniques that permit this type of quantification on a molecular level in vivo in humans are currently not available. Imaging techniques such as ultrasound and magnetic resonance imaging (MRI) are quite capable of assessing plaque size (3). In addition, MRI can identify plaque components, such as lipid-rich tissue, hemorrhage, and calcifications, through signal intensity differences elicited by varying contrast-weighted sequences (3). Nonetheless, the sensitivity of the detection of plaque cholesterol content is limited because lipid regions are inhomogeneous. Furthermore, lipid regions may contain crystals of cholesterol and are sometimes proximal to regions of calcification, both of which create a signal void in proton MRI.
Image-guided proton 1H magnetic resonance spectroscopy (1H-MRS) has the potential to detect liquid cholesteryl ester noninvasively without the need for radiation or additional contrast agents (4–7). 1H-MRS gives a spectrum of signals from all proton resonances in a given volume, thus affording detection of specific chemical components through their inherent frequency shift relative to water. On image-guided 1H-MRS, a magnetic resonance image can be used to image and localize plaque. Proton spectra can then be collected from these plaques, so that the specific proton resonances of lipid components in a liquid state, including cholesteryl ester, can be identified (4–7). A major advantage is that a change in the composition of the plaque could be evaluated without (and potentially preceding) changes in plaque volume.
The aim of our study was to develop a noninvasive image-guided 1H-MRS approach that enables quantification of plaque cholesteryl ester content. First, we assessed 1H-MRS signals from lipid standards containing triglycerides and cholesteryl ester to characterize their spectrum. We then performed in vivo 1H-MRS measurements of patients with atherosclerotic lesions in the carotid arteries.
1H-MRS spectra of pure cholesteryl ester and triglycerides as well as mixtures of both lipids were obtained. These data were acquired at Merck Research Laboratories, Rahway, New Jersey.
MRI (3.0-T) and 1H-MRS of atherosclerotic plaques in carotid arteries of 35 patients were performed at the Academic Medical Center, Amsterdam, the Netherlands. The study protocol was reviewed and approved by the institutional review board, and all subjects gave written informed consent. Inclusion criteria were 18 to 75 years of age and carotid artery stenosis between 30% and 70% on duplex ultrasound. Exclusion criteria were previous or planned carotid surgery or any contraindication to MRI. The presence of cardiovascular risk factors and the use of medication were assessed by questionnaire. Brachial artery blood pressures were measured using an oscillometric blood pressure device (Omron 705IT, Omron Healthcare Inc., Lake Forest, Illinois). Weight and length were measured to calculate body mass index. Venous blood samples were obtained after overnight fasting to assess plasma total cholesterol, high-density lipoprotein cholesterol, and triglyceride levels. Low-density lipoprotein cholesterol levels were calculated using the Friedewald equation.
Lipid standard 1H-MRS
To characterize the 1H-MRS spectra, image-guided 1H-MRS was performed on purified triglycerides (glyceryl trioleate), and cholesteryl ester (cholesteryl linoleate, Sigma-Aldrich, St. Louis, Missouri), as well as mixtures of both lipids. The lipid standards were scanned with an 11.7-T 500 wide-bore MRI system (Bruker NMR, Inc., Billerica, Massachusetts) with a vertical bore diameter of 89 mm. Spectra were acquired with a point resolved spectroscopy sequence. Scan parameters were repetition time (TR)/echo time (TE) = 5,000/10 ms, 2 × 2 × 2 to 2 × 4 × 2 mm3 voxel size, and 256 excitations. The line width was 100 Hz. No water suppression was used because there was no water content in the lipid standards.
In vivo carotid MRI and 1H-MRS
MRI scans were performed using a 3.0-T whole-body scanner (3.0 Tesla Intera, Philips Medical Systems, Best, the Netherlands), using an 8-element carotid coil (Philips Medical Systems). For localization of the carotid plaques, axial T1-weighted image stacks were acquired at late diastole (retrospective electrocardiographic gating). Sequence parameters were turbo spin echo, double inversion recovery black blood prepulse, 2 signal averages, 8-ms TE, TR according to heart rate (∼800 to 900 ms), 3-mm slice thickness, 240 matrix size, 60 × 60-mm field of view, 0.25 × 0.25-mm noninterpolated pixel size, and active fat suppression (spectral abiabatic inversion recovery). Acquisition time of the 1H-MRS spectra was approximately 13 min, and total scan time including the T1-weighted imaging was ∼20 min.
MRI analysis was performed using semiautomated measurement software (VesselMass, Leiden University Medical Center, Leiden, the Netherlands). We drew the lumen and outer wall boundaries in the 2 slices (each with a 3-mm slice thickness) that were located where the 1H-MRS voxel was located. The localization of the 1H-MRS voxel was based on the axial T1-weighted image stacks, which enabled us to identify which T1-weighted slices were located in the 1H-MRS voxel. We measured the lumen area (mm2) and wall area (mm2), which we multiplied by 5 mm (the slice thickness of the MRS voxel) to calculate lumen volume (mm3) and wall volume (mm3). Subsequently, perivascular tissue volume (mm3) was calculated by subtracting the lumen volume and wall volume from the volume of the MRS voxel (which was 500 mm3). After the 1H-MRS scan was performed, we obtained another T1-weighted slice located in the 1H-MRS voxel to assess whether the subject had moved during the 1H-MRS scan.
Image-guided 1H-MRS acquisition was performed on a 3.0-T whole-body scanner using an 8-element carotid coil (3.0 Tesla Intera, Philips Medical Systems). Spectra with and without water suppression were obtained. Two-dimensional chemical shift imaging data were collected using a point resolved spectroscopy sequence with the following parameters: TR = 1,100 ms, TE = 30 ms, 5-mm slice thickness, 16 × 16 matrix size, 8 × 8-cm field of view, 13-min acquisition time, 1 acquisition. Each 1H-MRS voxel was 5 × 5 × 5 mm (125 mm3). For all analyses, 4 voxels (total volume = 500 mm3) covering the carotid artery plaque were selected for further analysis (Fig. 1). To minimize patient movement during the 1H-MRS scan, we instructed the subjects not to move during the scan, and we reminded them of this by repeating this instruction just before starting the MRS sequence. The 4 voxels (total volume = 500 mm3) covering the carotid artery plaque that were selected for further analysis (Fig. 1) were processed using the freely available 3DiCSI package (version 1.9.11, Columbia University, New York, New York). Chemical shifts were reported using water as the internal standard at 4.65 ppm. Average spectra were then processed using specialized computer software (jMRUI 2.2, MRUI, Leuven, Belgium) (8). Three peaks were line fitted: methyl (CH3) at 0.8 ppm, methylene (CH2) at 1.2 ppm, and allylic methylene (CH2-CH=CH-CH2) at 2.0 ppm (Fig. 1). We used water-suppressed spectra to measure the 0.8-ppm, 1.2-ppm, and 2.0-ppm peaks. In a subset of 4 subjects, we performed initial and repeat 1H-MRS scans with at least 1 week between scans to assess interscan reproducibility.
Continuous variables are expressed as mean ± SD, unless otherwise specified. Spearman rho values were calculated to assess the correlation between the allylic methylene (2.0 ppm) to methylene (1.2 ppm) ratio and serum total cholesterol serum triglycerides, as well as the plaque volume to perivascular tissue volume ratio. Spearman rho was used because the data was not normally distributed. The agreement between initial and repeat 1H-MRS scans within patients was assessed by the intraclass correlation coefficients (r), the SD of the paired differences, and the repeatability coefficient. All statistics were performed with SPSS version 19.0 (SPSS Inc., Chicago, Illinois).
Lipid phantom results
1H-MRS spectra of lipid standards were obtained of 5 different mixtures of triglycerides and cholesteryl ester. The allylic methylene (CH2-CH=CH-CH2 at 2.0 ppm) to methylene (CH2 at 1.2 ppm) ratio shows an increase as the amount of cholesteryl ester increases relative to triglycerides (Fig. 2). Note that this increase remains present if the peaks of 2.0 ppm and 2.2 ppm are merged together as will be the case at lower field strength at lower spectral resolution. The absolute values of the 2.0 ppm to 1.2 ppm ratio of the lipid phantoms cannot be compared with the absolute values of the in vivo 2.0 ppm to 1.2 ppm ratio because different TEs were used in vivo and ex vivo and pure glyceryl trioleate and cholesteryl linoleate were used, whereas, in vivo, a mixture of different fatty acid chains exists.
In vivo 1H-MRS results
Carotid MRI and 1H-MRS scans were obtained of 35 subjects with carotid atherosclerosis, defined as 30% to 70% carotid stenosis on ultrasound duplex. Of the 35 1H-MRS spectra, 17 spectra (49%) were of adequate quality for analysis. Patient characteristics of the 17 patients are shown in Table 1. The methyl (0.8 ppm), methylene (1.2 ppm), and allylic methylene (2.0 ppm) peaks were line fitted. The methyl (0.8 ppm) peak proved to be intricate to fit properly. Because of the limited spectral resolution the methyl peak (0.8 ppm) often partially blended into the methylene (1.2 ppm) peak and because of this, these fits were unreliable. Of the allylic methylene (2.0 ppm) and methylene (1.2 ppm) peaks, we were able to obtain good quality fits in 17 of 35 spectra. The median allylic methylene (2.0 ppm) to methylene (1.2 ppm) ratio was 0.10 (interquartile range: 0.08 to 0.12).
The volume of tissue from which we obtained the spectra was composed of blood in the lumen, plaque, and perivascular tissue. We assessed whether serum lipids influenced our 1H-MRS measurement. We did not find a correlation between the allylic methylene (2.0 ppm) to methylene (1.2 ppm) ratio and serum total cholesterol (Spearman rho = 0.43, p = 0.10) or serum triglycerides (Spearman rho = 0.27, p = 0.31), indicating that serum lipids did not influence our 1H-MRS spectra. We also assessed whether our 1H-MRS measurement was influenced by the amount of plaque volume (cholesteryl ester–rich tissue) relative to the perivascular tissue volume (triglyceride-rich tissue). Figure 3 shows that there is a significant correlation between the allylic methylene (2.0 ppm) to methylene (1.2 ppm) ratio and plaque volume to perivascular tissue volume ratio.
Seven subjects were willing to visit us twice, and in these subjects, initial and repeat 1H-MRS scans were obtained, of whom 4 subjects had analyzable data of both the initial and repeat scan. In this subset of 4 subjects we assessed the agreement between initial and repeat 1H-MRS scans within patients, and the intraclass correlation coefficient (r) was 0.92 (95% prediction intervals: 0.40 to 0.99), the SD of the paired differences was 0.011, and the reproducibility coefficient was 0.022 (95% limits of agreement: −0.017 to 0.029), for the allylic methylene (2.0 ppm) to methylene (1.2 ppm) ratio, indicating good reproducibility.
The present study shows for the first time that image-guided 1H-MRS of carotid atherosclerotic plaques is feasible in vivo in humans. We successfully quantified the allylic methylene (2.0 ppm) to methylene (1.2 ppm) ratio, which reflects the ratio of the fatty acid composition of plaque cholesteryl ester to that of perivascular tissue triglyceride. Furthermore, the allylic methylene (2.0 ppm) to methylene (1.2 ppm) ratio was shown to have good reproducibility. 1H-MRS of plaques still has a limited spatial resolution and is technically challenging, illustrated by the fact only 49% of the obtained 1H-MRS spectra was of adequate quality for analysis. The present data imply that in vivo in humans, 1H-MRS of carotid artery plaques offers a promising and valuable tool that can specifically identify liquid cholesteryl ester through its specific proton resonances.
The type of lipids and their physical properties in atherosclerotic lesions have been well investigated over the past decades (1). Cholesterol, cholesteryl ester, and phospholipids are known to accumulate in fatty streaks and plaques, whereas triglyceride accumulation does not occur (1). As early as the 1960s, it was recognized that a large portion of lipids in atherosclerotic lesions accumulate in the form of intracellular droplets. Lang and Insull (9) showed that the droplets contained about one-half of the lipids in the lesions. Furthermore, the composition of the lipids in the droplets was remarkably uniform and comprised 94.9% cholesteryl esters, 1.7% free cholesterol, 1.0% phospholipids, and 2.4% triglycerides. Thus, cholesteryl ester in the liquid phase in intracellular droplets makes up a large portion of the plaque lipid content. Moreover, the reversibility of atherosclerotic lesions likely depends on removal of these liquid cholesteryl esters, in contrast to crystalline cholesterol deposits, which are thought to be trapped within the plaque (1). Therefore, a technique that enables direct and unequivocal identification of cholesteryl ester in the liquid state in plaque can provide important information on plaque phenotype and would enable specific investigation of the effect of antiatherosclerotic interventions aimed at removing cholesteryl esters from plaques.
Based on the findings by Lang and Insull (9), the 1H-MRS spectra of atherosclerotic plaque are expected to be derived for 95% of cholesteryl esters and for only a small amount of triglycerides (phospholipids are known not to show resonance). The latter contrasts with our observation. Based on the characteristics of the lipid phantoms and our in vivo spectra, we estimate that in addition to cholesteryl ester, the signal from triglycerides does contribute significantly to the in vivo spectra. This apparent discrepancy most likely pertains to the fact that the volume of tissue from which we obtained the in vivo spectra does not only contain plaque, but also a substantial amount of perivascular tissue. Perivascular tissue is known to be composed mainly of adipocytes in which triglyceride droplets are stored (10). The fatty acyl chains of triglycerides in adipose tissue are 50% comprised of oleate, 12% of linoleate, 25% of palmitic acid, and to a lesser extent, of other fatty acyl chains (11). In contrast, the fatty acyl chains of the cholesteryl esters in plaques comprise ∼35% oleate, 35% linoleate, 14% palmitic acid, and to a lesser extent other fatty acyl chains (12). Oleate (the carbon-to-double bond ratio of 18:1) consists of a single allylic methylene (CH2-CH=CH-CH2) and 14 methylenes (CH2). Linoleate (18:2) consists of 2 allylic methylenes and 12 methylenes. Palmitic acid (16:0) has 14 methylenes but does not have an allylic methylene group. The markedly different composition of fatty acyl chains of plaque cholesteryl ester and perivascular triglycerides results in a lower allylic methylene-to-methylene ratio in perivascular triglycerides as opposed to plaque cholesteryl ester.
The contribution of serum lipids to the spectra is minimal. We estimate that in the lumen volume included in the 1H-MRS voxel, the amount of lipids approximates 0.5 mg total serum cholesterol and 0.3 mg serum triglycerides. This is negligible compared with the amount of cholesteryl ester in plaques, which exceed an order of magnitude of 4 to 28 mg (9). Therefore, we expect the impact of serum lipids on the 1H-MRS spectra to be minimal. Furthermore, we did not find a relationship between serum lipid levels and the allylic methylene (2.0 ppm) to methylene (1.2 ppm) ratio.
MRS studies of ex vivo atherosclerotic lesions have been performed with both 13C-MRS and 1H-MRS (4–7,13–16). 13C-MRS studies showed that the spectra obtained from plaque reflect almost exclusively cholesteryl ester in the liquid state (13–16). Lipids in liquid crystalline and crystalline phases, such as phospholipids and cholesterol were detected by magic angle spinning 13C-MRS (13–16). Furthermore, 4 studies have performed 1H-MRS of plaques ex vivo (4–7). Zajicek et al. (4) performed 1H-MRS at 8.481-T on 8 ex vivo plaques that were scanned at 37°C and compared the spectra with plaque lipid content assessed by quantitative thin-layer chromatography, gas-liquid chromatography, and histology. They showed that the 1H-MRS spectra largely reflect the presence of cholesteryl esters in the isotropic liquid phase. Maynor et al. (5) reported a 1H-MRS study at 7.0-T using chemical shift imaging on 13 atherosclerotic lesions and also showed that plaques reveal a typical cholesteryl ester spectrum that differed from the triglyceride spectrum of adipose tissue.
Pearlman et al. (6) performed 1H-MRS of ex vivo human plaques at 6.342-T, 8.481-T, and 11.75-T and investigated the relationship of temperature and the spectral resonances of plaque lipids. They found that the 1H-MRS signals are derived of a mixture of cholesteryl esters, whose liquid crystalline to isotropic fluid phase transition is near body temperature. Ruberg et al. (7) acquired 1H-MRS at 11.7-T of 10 human carotid endarterectomy specimens. They obtained MRS spectra from 1-mm3 voxels, localized to plaque regions that were either lipid rich or lipid poor and showed that the spectra from lipid-rich areas showed much more intense cholesteryl ester resonances than did lipid-poor regions. Our in vivo findings corroborate the previous ex vivo 1H-MRS studies, as our data also showed typical spectra of mixed cholesteryl ester and triglyceride, with resonance from the methyl (0.8 ppm), methylene (1.2 ppm), and allylic methylene (2.0 ppm) protons. Furthermore, Ruberg et al. (7) also performed 1H-MRS on lipid standards of pure cholesteryl ester, pure triglycerides, and pure phospholipids and showed identical proton spectra compared with our pure cholesteryl ester, pure triglyceride spectra. No resonances from the phospholipids were observed, which was expected because of their highly restricted and anisotropic motion in the liquid crystalline multilayer.
First, with our image-guided 1H-MRS approach, spectra were obtained of a volume that contained blood in the lumen of the vessel, plaque tissue, and perivascular tissue. Due to limited resolution and the point spread function of the chemical shift imaging sequence, the proton spectra could not be obtained of only plaque tissue. Nonetheless, plaque tissue is known to have abundant cholesteryl ester content and very little triglycerides, whereas the opposite is true for perivascular tissue. Thus, the contribution of the fatty acids of cholesteryl ester to the signal is likely to be derived specifically from plaque tissue. Second, the 1H-MRS spectra do not permit an absolute quantification of the amount of cholesteryl ester to be determined. Instead 1H-MRS spectra provide the amount of cholesteryl ester in relation to the amount of triglycerides. Because this relies on the amount of plaque tissue and perivascular tissue included in the volume from which the spectrum is obtained, comparison between subjects is cumbersome. Nonetheless, our current approach is suitable for longitudinal comparison within subjects, as long as the volume from which the 1H-MRS spectrum is obtained is located in the same location. The fact that our measurement showed good reproducibility shows that reproducible localization of image-guided 1H-MRS is feasible.
The third limitation is that it proved to be difficult to obtain good quality 1H-MRS spectra. About one-half of the spectra were of poor quality. This could be due to various factors, such as separating the signal of methylene and allylic methylene protons, unmet shimming requirements, signal loss due to the deep localization of the carotid arteries, and patient movement. For future studies we need to optimize the 1H-MRS methods to improve the quality of the spectra.
The present study shows for the first time that image-guided 1H-MRS of carotid atherosclerotic plaques is feasible in vivo in humans and can possibly detect cholesteryl ester in the liquid phase. The ratio of the resonance of the allylic methylene (2.0 ppm) to methylene (1.2 ppm) protons reflects the fatty acid composition of plaque cholesteryl ester relative to that of perivascular triglycerides in the scanned volume. Our image-guided 1H-MRS method showed good reproducibility. Future studies are needed to further address the limitations of the technique, improve the quality of the 1H-MRS spectra, and enable quantification of liquid-phase cholesteryl ester.
The authors thank A.M. van den Berg for assistance with data acquisition.
The study was supported by an educational research grant of Merck USA. The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- 1H magnetic resonance spectroscopy
- magnetic resonance imaging
- echo time
- repetition time
- Received October 6, 2012.
- Revision received February 20, 2013.
- Accepted March 21, 2013.
- American College of Cardiology Foundation
- Small D.M.,
- Shipley G.G.
- Ruberg F.L.,
- Viereck J.,
- Phinikaridou A.,
- et al.
- Kokatnur M.G.,
- Oalmann M.C.,
- Johnson G.T.,
- et al.
- Hamilton J.A.,
- Cordes E.H.,
- Glueck C.J.
- Guo W.,
- Morrisett J.D.,
- DeBakey M.E.,
- et al.
- Peng S.,
- Guo W.,
- Morrisett J.D.,
- et al.