Chemical Exchange Saturation Transfer MRI Identifies Abnormal Calf Muscle-Specific Energetics in Peripheral Artery Disease

Abstract

Background:

Peripheral artery disease (PAD) results in exercise-induced ischemia in leg muscles. ³¹Phosphorus (P) magnetic resonance spectroscopy (MRS) demonstrates prolonged phosphocreatine (PCr) recovery time constant after exercise in PAD, but has low signal to noise, low spatial resolution, and requires multi-nuclear hardware. Chemical exchange saturation transfer (CEST) is a quantitative MRI method for imaging substrate (CEST_asym) concentration by muscle group. We hypothesized that kinetics measured by CEST could distinguish between PAD patients and controls.

Methods:

PAD patients and age-matched normal subjects were imaged at 3T with a transmit-receive coil around the calf. Four CEST mages were acquired over 24 second intervals. The subjects then performed plantarflexion exercise on an MRI-compatible ergometer until calf exhaustion. Twenty five CEST images were obtained at end exercise. Regions of interest (ROI) were drawn around individual muscle groups and (CEST_asym ) decay times were fitted by exponential curve to CEST values. In 10 patients and 11 controls, ³¹P spectra were obtained 20 minutes later after repeat exercise. 5 patients and 5 controls returned at a mean of 1±1 days later for repeat CEST studies.

Results:

Thirty five PAD patients (31 male, age 66±8 yrs.) and 29 controls (11 male, age 63±8 yrs.) were imaged with CEST. The CEST_asym decay times for the whole calf (341±332s vs. 153±72s, p<0.03) as well as for the gastrocnemius and posterior tibialis were longer in PAD patients. Agreement between CEST_asym decay and PCr recovery time constant was good.

Conclusions:

CEST is a MRI method that can distinguish energetics in PAD patients from age-matched normal subjects on a per muscle group basis. CEST agrees reasonably well with the gold standard ³¹P MRS. Moreover, CEST has higher spatial resolution, creates an image, and does not require multi-nuclear hardware and thus may be more suitable for clinical studies in PAD.

Introduction

Peripheral arterial disease (PAD) is quite prevalent, affecting over 200 million worldwide, and is characterized by atherosclerotic lesions in the arteries supplying one or more limbs, leading to decreased perfusion in downstream tissue.¹ PAD leads to significant burden of morbidity with over 7% of the general population over 40 being affected in the US, and 29% of adults over 70 having PAD.² Symptoms such as claudication can severely limit mobility and, in 11% of PAD patients, it can progress to critical limb ischemia (CLI) with rest pain, ischemic ulceration, or gangrene.³

There is an unmet need to better understand the underlying mechanisms of PAD to enable improved treatments and outcomes.¹ The standard diagnostic test for identifying PAD is the ankle brachial index (ABI). This relatively straightforward test is 90% sensitive and 95% specific for PAD, ⁴ but decreasing ABI does not correlate well with worsening function or outcomes. Additionally, incompressible vessels due to vessel wall calcification leads to erroneously high ABIs.⁵

One technique to assess calf muscle physiology, specifically energetics, is phosphorus-31 (³¹P) magnetic resonance spectroscopy (MRS).⁶ Phosphocreatine (PCr) is used in highly oxidative tissue as a secondary energy source to rephosphorylate ADP after use without the need for aerobic respiration. After exertion, the phosphate group of PCr is expended and joins the free phosphate pool within the tissue until oxidative phosphorylation can occur in the mitochondria. Prolonged PCr recovery kinetics have been shown to be indicative of tissue ischemia in PAD.⁶ However, ³¹P MRS suffers from low signal-to-noise ratio (SNR) due to the low concentration of PCr and small gyromagnetic ratio leading to a low NMR sensitivity to ³¹P.⁶ Specialized imaging hardware is necessary for detecting nuclei other than hydrogen and is frequently unavailable in clinical settings. Additionally, MRS typically results in frequency spectra rather than traditional images and has limited spatial resolution.

A recently developed alternative is chemical exchange saturation transfer (CEST).⁷ This MRI methodology allows indirect detection of molecules that exchange protons with free water. The CEST effect from Cr can be isolated from PCr, adenosine triphosphate, and adenosine diphosphate due to differences in amine proton exchange rates. After an off-resonance saturation prepulse tuned to the resonant frequency of creatine, the creatine in the muscle post-exercise exchanges amine protons with water protons when being rephosphorylated by creatine kinase. The creatine concentration in the tissue is conserved before and after exercise, but the oxidative rephosphorylation removes a hydrogen from the magnetically saturated amine group. This method allows for three orders-of-magnitude higher SNR than ³¹P MRS.^{8, 9} While ³¹P MRS directly measures the relative amounts of phosphorus present in different molecular configurations using a specialty coil, CEST images are made with ¹H coils imaging water molecules that have exchanged magnetization with the metabolite of interest. CEST could address many of the pitfalls of ³¹P MRS as it creates images where signal is proportional to substrate concentration in each voxel, allowing for muscle group-specific study of metabolic energetics at high spatial resolution.

This study was aimed to demonstrate the ability of CEST to distinguish the muscle-group specific energetics of PAD patients from age-matched controls. CEST energetics in PAD and controls were compared to the current gold standard, ³¹P MRS. We hypothesized that CEST is a reliable and repeatable method for assessing energetics on a muscle-wise basis in the setting of PAD and could be useful in future clinical studies.

Methods

The data that support the findings of this study are available from the corresponding author upon reasonable request. Patients between the ages of 35 to 85 years with symptoms of intermittent claudication or critical limb ischemia and a documented ankle brachial index (ABI) of <0.9 were eligible for this study. Exclusion criteria included lower extremity vascular surgery or percutaneous intervention <3 months prior to enrollment, gangrene or a non-healing wound in the foot of the leg of interest, need for urgent revascularization, active coronary artery disease or recent (<2 months) myocardial infarction, body mass index of >40, or a known contraindication to MRI. Normal human subjects without risk factors for atherosclerosis were recruited from the community to serve as age-matched control subjects. Subjects with risk factors were eligible only with normal ABIs. All data from consecutively recruited patients and control subjects were included in the manuscript. The protocol was approved by the Human Investigation Committee at the University of Virginia, and all participants signed informed consent. CMK had access to all data and takes responsibility for its integrity and the data analysis.

CEST images were obtained on a 3T MRI scanner (Prisma, Siemens Medical Solutions, Erlangen, Germany) using a 15-channel transmit-receive extremity coil centered at mid-calf. Measurements were acquired using a pulse sequence developed at University of Pennsylvania.^{8, 9} Water saturation shift referencing (WASSR) and B1 maps were collected for B0 and B1 correction, respectively. Six images were acquired over 24 s intervals with saturation frequency offsets of ±1.3, ±1.8, and ±2.3 ppm. The CEST effect reduces the signal at +1.8 ppm compared to the reference at −1.8 ppm, with this asymmetry quantified as CEST_asym. A 500ms saturation pulse train was applied consisting of five 99.6ms Hanning-windowed pulses with 150 Hz peak B1 amplitude and 0.4ms inter-pulse delay. Imaging parameters were single-shot spoiled gradient-echo readouts with centric encoding, fat saturation, flip angle 10°, field of view (FOV) 160×160 mm, matrix 144×144, in-plane resolution 1.1 mm, repetition time (TR) 6.0ms, echo time (TE) 2.9ms, and slice thickness 10 mm.

Subjects were placed supine, feet first, into the scanner, with the foot of the leg of interest placed in an MR-compatible plantar flexion ergometer (Trispect, Ergospect GmbH, Innsbruck, Austria). A reference image and B1, B0, and WASSR maps were collected. This was followed by 4 baseline CEST images obtained prior to exercise.⁹ Then subjects were asked to begin plantarflexion in time with a metronome set at 60 bpm until claudication symptoms or calf exhaustion. This was chosen to ensure all subjects reached a state of muscle ischemia. The ergometer resistance was adjusted for the subjects’ ability to pedal uniformly against the resistance level. Immediately post exercise, 25 images were obtained. This was followed by repeat B0, B1, and WASSR maps. Five subjects in each group returned for studies of test-retest reliability.

Studies were analyzed blinded to subject group. CEST_asym decay times were generated from ROIs including the entirety of the tibialis anterior, posterior compartment, gastrocnemius, and the whole calf (excluding bones) for each subject in MATLAB (MathWorks, Natick, Massachusetts). The soleus muscle was not included in analysis, as it is not activated as uniformly as the other muscles during plantarflexion, and it contains several large vessels that are difficult to exclude during analysis. Decay times were generated from the individual ROIs by fitting using a least-squares regression fit to a monoexponential and recorded for each subject (Figures 1 and 2).

Figure 1.

Calf muscle maps, CEST_asym Maps, and Regions of Interest with decay curves (AT=Anterior Tibialis m.; PT=Posterior Tibialis m.; Soleus=Soleus m.; Gastro=Gastrocenmius m.; V=arterial and venous vessels; T=Tibia; F=Fibula)

Figure 2.

Example CEST images and corresponding CEST_asym decay curves from a normal control (top) and a PAD patient. The first 4 images are pre-exercise and the subsequent images begin immediately post-exercise. Note the red color depicting increased CEST_asym signal immediately post-exercise that resolves quickly in the normal subject and takes significantly longer to normalize in the PAD subject. This is reflected in the much longer CEST_asym decay constant in the PAD patient (bottom right).

For ³¹P MRS studies, subjects were positioned in the scanner with the same foot on the ergometer pedal with a multinuclear surface coil against their gastrocnemius muscle, and 5 pre-exercise spectra were acquired. After subjects performed plantar flexor ergometry to claudication or calf exhaustion, 25 signal averages were acquired after 4 preparation pulses at a repetition time of 550 ms for a total acquisition time of 16 s per spectrum, and 18 spectra were acquired at the cessation of exercise.⁶

Statistical Analysis

Subject characteristics were summarized as mean and standard deviation for continuous variables and by frequencies and percentage for categorical variables. Characteristics were compared between control subjects and patients using the Student t-test or Wilcoxon rank-sum test for continuous variables and the Pearson χ² test or Fisher’s exact test for categorical variables. Exercise times and CEST_asym decay were summarized as median and interquartile range (IQR). Pearson correlation coefficients were calculated for CEST_asym decay times after log transformation compared to demographic variables. Pearson correlation was performed comparing CEST decay constant and phosphocreatine recovery time constant. For all comparisons, p<0.05 was considered significant. Agreement between PCr recovery by ³¹PMRS and CEST_asym decay was analyzed using the method of Bland and Altman. Test-retest reliability for CEST_asym decay was also analyzed in 5 subjects who underwent testing twice at different time points using Bland-Altman. Analysis was performed using SAS, version 9.4 (SAS Institute, Cary, North Carolina).

Results

A total of 64 subjects were studied; 35 subjects with PAD and 29 controls. The mean age was similar between groups (Table 1), but there were fewer women in the PAD group. The mean ABI for the PAD group was 0.65 ± 0.11. Demographics are shown in Table 1. A majority of PAD patients had at least one of the following conditions: hypertension, hyperlipidemia, or a smoking history. Approximately a third of PAD patients had coronary artery disease. Very few of the normal subjects had risk factors for atherosclerosis.

Table 1:

Demographics of the Study Population

	Controls	PAD	p*
Number of subjects, n	29	35	-
Women, n (%)	17 (59)	4 (11)	< 0.001
African-American, n (%)	1 (3)	7 (20)	0.063
Mean Age, mean ± SD (years)	65±8	66±8	0.103
Diabetes mellitus, n (%)	3 (10)	14 (30)	0.007
Hypertension, n (%)	5 (17)	29 (83)	<0.001
Hyperlipidemia, n (%)	7 (24)	30 (86)	<0.001
Coronary Artery Disease, n (%)	4 (14)	13 (37)	0.035
Smoking, n (%)	2 (7)	31 (89)	< 0.001
ABI of leg studied, mean ± SD	-	0.65±0.11	-

^*p was generated by using the Pearson χ² test or Fisher’s exact test for categorical variables, and the Student t-test for normally distributed continuous variables.

Plantarflexion exercise times tended to be longer in the normal subjects as would be expected, but this difference did not reach statistical significance due to the wide variation in exercise times in both groups. (Table 2) The CEST_asym decay times for the overall calf as well as the individual gastrocnemius and posterior tibialis in the PAD group were increased compared to the age-matched control group. (Table 2) CEST_asym decay times were similar between groups in the anterior tibialis. The CEST_asym levels at end-exercise was 178±58% of baseline in PAD subjects and 180±33% of baseline in control subjects, p=0.86.

Table 2.

CEST_asym results

	Controls n=29	PAD N=35	p
Exercise time (s) Median (IQR)	251 (538)	125 (87)	0.19
Calf CESTasym decay constant (s) Median (IQR)	152 (84)	276 (329)	<0.03
Tibialis anterior CESTasym decay constant (s) Median (IQR)	126 (182)	139 (246)	0.36
Tibialis posterior CESTasym decay constant (s) Median (IQR)	125 (142)	186 (347)	<0.02
Gastrocnemius CESTasym decay constant (s) Median (IQR)	178 (127)	267 (300)	<0.03

A history of hypertension was associated with prolonged whole calf CEST_asym decay times in the entire group of 64 subjects (r=0.32, p<0.01) as was history of smoking (r=0.26, p<0.04). There was no association between ABI and CEST_asym decay times. Age tended to be associated with the CEST_asym decay times in the anterior tibialis (r=0.24, p<0.06), but not in other muscle groups. A history of CAD was associated with CEST_asym decay times in the anterior tibialis (r=0.30, p<0.02) and posterior tibialis (r=0.32, p<0.01).

Test-retest reliability was performed on 5 normal subjects and 5 PAD patients. Measurements were obtained at a mean time of 1±1 days after the initial scan. Exercise times for the 10 subjects were similar between the 2 time points (168±137s at time 1 and 246±287s at time 2, p=0.45). Results for agreement are displayed using Bland Altman analysis (Figure 3) showing very good agreement with 1 data point lying outside the 2 standard deviations.

Figure 3.

Bland-Altman plot comparing test-retest values in a total of 10 subjects (5 with PAD and 5 controls). There is one data point that does not lie within 2 standard deviations of the difference between the 2 studies.

Thirteen PAD patients and 11 normal patients were imaged by CEST and underwent ³¹P MRS. pH at end-exercise was not different between groups (7.02±0.05 in PAD and 7.07±0.08 in controls, p=NS). Table 3 demonstrates the data from these studies. Figure 4 displays the Bland Altman analysis between the 2 groups showing good agreement. Figure 5 shows a Pearson correlation plot showing 3 patients with high PCr recovery time constant yet relatively preserved CEST decay and resultant lack of correlation.

Table 3.

Data from PAD and controls undergoing both CEST and ³¹P MRS post-exercise.

	Controls N=11	PAD N=13	P*
Exercise time, median (IQR) (s)	224 (450)	125(199)	0.192
Calf CESTasym decay constant, median (IQR) (s)	134(115)	134 (330)	0.524
Phosphocreatine recovery time constant, median (IQR) (s)	46(80)	199(492)	0.007

^*p was generated using the Wilcoxon rank-sum test for non-normally distributed continuous variables.

Figure 4.

Bland-Altman plot comparing PCr recovery time constant by ³¹P MRS and CEST_asym decay constant for 13 PAD patients and 11 normal controls. Agreement is excellent.

Figure 5.

Pearson correlation plot of CEST decay constant vs. phosphocreatine recovery time constant. p = 0.66, r = 0.09.

Discussion

This study examined differences in CEST_asym between normal subjects and patients with PAD using CEST MRI. We found that patients with PAD demonstrated a significant increase in CEST_asym decay times in the entire calf compared to their normal age-matched controls. In addition, the difference in CEST_asym decay was also able to be isolated to the specific muscle groups of the calf including the gastrocnemius and posterior tibialis muscles. When compared with normal subjects, there was an almost two-fold increase in the CEST_asym decay time in PAD patients. This is the inverse of the increase in phosphocreatine recovery times seen in the patients who underwent ³¹P-MRS and in our previous studies. However, the ability to isolate specific muscle groups is an advantage of CEST_asym compared to ³¹P-MRS and raises the potential for utilizing this technique to evaluate therapies that might improve perfusion in specific muscle groups. These results indicate that CEST_asym can be used to differentiate CEST_asym kinetics in PAD and normal patients with excellent spatial resolution. Test-retest reliability of the CEST_asym decay measures was very good.

Measuring tissue energetics in exercising skeletal muscle in PAD has been performed with ³¹P MRS for over 25 years.¹⁰ Chronic limitations in blood flow can lead to maladaptive changes within mitochondria that lead to delayed phosphocreatine recovery,⁶ although there is not a direct correlation between perfusion and energetics in the later stages of PAD.¹¹ The present study significantly extends prior studies using the relatively new CEST technique which were primarily aimed at proof of concept. For example, prior studies were performed with few subjects (n =3 PAD patients and 3 controls)⁷ and one study showing reasonable agreement between CEST_asym and ³¹P-MRS was performed in only 6 volunteers, not in PAD.⁸ Feasibility and reproducibility of these measurements has been shown, again with small numbers of volunteers, at higher field strength (7T).¹² The present study sets the stage for clinical applications of the technique.

A major challenge for the development of new therapies in PAD is the lack of quantitative measures for measuring tissue physiology such as energetics and perfusion. Large vessel inflow as measured by ABI and angiography does not reflect the tissue effects of therapies. CEST MRI offers such promise as an attractive target for assessing benefits of revascularization, other novel therapies, and clinical outcomes in PAD. ³¹P MRS has been available since the 1980’s for evaluation of PAD, but has not been utilized clinically due in part to lack of availability due to lack of multi-spectral capabilities at many centers. Strengths of ³¹P MRS include the extensive literature supporting its application in PAD, its reproducibility, and robustness. CEST is an imaging technique that does not require multispectral hardware and thus could theoretically be applied on any high field clinical scanner with the appropriate pulse sequence and thus has far broader clinical potential than ³¹P-MRS. In addition, the ability to localize abnormal energetics to individual muscle groups could allow the development of vessel and thus muscle-specific revascularization therapies.

Differences between PAD subjects and controls were seen in each muscle group except for the anterior tibialis. The anterior tibialis muscle contains more fast twitch muscle fibers than the rest of the calf, allowing it to function better in hypoxic conditions.¹³ This may account for the similar results between groups. In addition, plantar flexion exercise appears to trigger preferential use and higher perfusion of the anterior tibialis¹⁴ and this may also account for less of a difference between PAD and normal subjects. Hypertension and a history of smoking were associated with prolonged CEST_asym decay times. Future studies may determine the mechanisms underlying these risk factors effects on calf muscle energetics and whether reduced perfusion plays a role.

Our group has developed novel non-contrast methods of measuring tissue perfusion in PAD subjects using arterial spin labeling (ASL) MRI after exercise¹⁴ or with thigh cuff occlusion/hyperemia.¹⁵ Similarly to CEST, perfusion can be measured by ASL on a per muscle group basis. Combining these non-contrast measures of perfusion and energetics could offer significant potential for understanding the physiologic effects of revascularization and novel medical therapies on specific muscle groups in the calf. Perfusion as measured with single photon emission tomographic techniques has been shown to be an important marker of prognosis, predicting amputation-free survival in PAD.¹⁶ Future studies will help determine whether perfusion, energetics, or the combination are the best predictors of PAD outcome.

Limitations

Gender differences could in theory account for the differences in CEST_asym decay between normal and PAD patients as there was a significantly higher number of women in the normal group. In addition, the present study did not include patients with non-compressible vessels with ABIs >1.3 that may have symptoms of PAD. CEST_asym decay times are significantly longer than PCr recovery time constant. Based on chemical principles, the rate of CEST_asym decay should parallel phosphocreatine recovery. However, ³¹P MRS is not spatially localized and thus is likely preferentially measuring gastrocnemius energetics due to the location of the ³¹P surface coil. As shown in Table 2, gastrocnemius decay times appear to be somewhat longer than other muscle groups and this may account for some of the difference. Additionally, at 3T the resulting CEST_asym signal likely receives interference from tissue acidification, particularly lactic acid buildup from anaerobic muscle metabolism due to ischemia.¹⁷ Thus, the CEST_asym signal is likely not purely from creatine. Other potential reasons for the longer CEST_asym decay include signal to noise issues, patient motion, nonlinearity, or relative insensitivity to creatine levels from a technical standpoint. A major advantage ³¹P has over this fast time-course CEST is that relative quantities of phosphorous metabolites (ATP, PCr, Pi and pH) can be imaged at once. Doing so using ¹H CEST would require finely sampling the z-spectra through multiple excitations, making the resulting time-course too long to assess kinetics.

In addition, CEST and ³¹P-MRS are dependent on mitochondrial function, which has shown to be altered in patients with heart failure.¹⁸ As many patients with PAD have some degree of CAD and may have resultant reduced left ventricular function, there is a risk for confounding due to this. However, the PAD cohort in this study had a mean ejection fraction of 60% ± 4% with only one patient having a clinical diagnosis of heart failure, so it is unlikely that heart failure is contributing significantly.

Lastly, with regards to the technique itself, one of the difficulties is high sensitivity to patient motion during scanning. The patients were instructed to lie as still as possible in the scanner during imaging. Another potential difficulty is with fitting the CEST_asym decay curves as they may be altered due to combination of abnormal CEST_asym kinetics, motion artifact, and interference from vessel and bony structures when drawing regions of interest. Given its size relative to other muscles in the calf, the gastrocnemius muscle is the easiest to identify on the CEST maps, and therefore, the most consistent to draw the ROI. This may explain, in part, why it demonstrated the greatest difference in CEST_asym decay between the PAD patients and normal subjects.

Future Directions

Next steps will include correlation of energetics as measured with CEST with muscle-specific perfusion as measured by ASL in PAD patients. Additional studies will examine how revascularization alters both CEST_asym decay and perfusion and the utility of these measures in determining long-term prognosis.

Clinical Perspective.

Peripheral arterial disease (PAD) affects approximately 8.5 million patients in the United States. The evaluation and management of patients with PAD is largely based on the use of the ankle-brachial index (ABI) and vessel imaging such as duplex ultrasonography or computed tomography which primarily identify alterations in bulk flow in the peripheral arteries. New diagnostic techniques are necessary given the limitations of these techniques and the need for development of new therapies which require greater understanding of the pathophysiology of PAD. One focus is skeletal muscle metabolism, specifically the conversion of phosphocreatine to creatine that occurs with exercise. Using phosphorus-31 (³¹P) magnetic resonance spectroscopy (MRS), PAD patients have shown decreased rates of phosphocreatine recovery following exercise compared to control patients. Although promising, this use ³¹P MRS is limited by poor spatial resolution and a reliance on specific hardware that restricts its use to specialized centers. These limitations led to the development of a novel MRI-based technique termed chemical exchange saturation transfer (CEST) which examines the rate of CEST_asym decay following exercise. The present study shows in an age-matched population that patients with PAD have a decreased rate of CEST_asym decay following exercise compared to normal controls with excellent correlation with previously validated techniques. This study supports the use of CEST in future clinical studies in PAD patients to aid in the role in the development of new therapies and can be combined with MRI methods of measuring calf blood flow to more comprehensively evaluate the physiologic effects of PAD.

Abbreviations:

ABI

ankle brachial index

CEST

chemical exchange saturation transfer

MRI

magnetic resonance imaging

PAD

peripheral artery disease

³¹P MRS

31phosphorus magnetic resonance spectroscopy

PCr

phosphocreatine

ROI

region of interest

WASSR

water saturation with shift reference