Quantitative Assessment of Peripheral Oxidative Metabolism with a new Dynamic ¹H MRI technique: a Pilot Study in People with and without Diabetes Mellitus

Abstract

Background:

Type II Diabetes Mellitus (T2DM) is linked to impaired mitochondrial function. Chemical exchange saturation transfer (CEST) MRI is a gadolinium-contrast-free ¹H method to assess mitochondrial function by measuring low-concentration metabolites. A CEST MRI-based technique may serve as a non-invasive proxy for assessing mitochondrial health.

Hypothesis:

A ¹H CEST MRI technique may detect significant differences in in vivo skeletal muscle phosphocreatine ($\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$) kinetics between healthy volunteers and T2DM patients undergoing standardized isometric exercise.

Study Type:

Cross-sectional study.

Subjects:

7 subjects without T2DM (T2DM−) and 7 age, sex, and BMI-matched subjects with T2DM (T2DM+).

Field Strength/Sequence:

Single-Shot Rapid Acquisition with Refocusing Echoes (RARE) and Single-Shot Gradient-Echo sequences, 3T.

Assessment:

Subjects underwent a rest-exercise-recovery imaging protocol to dynamically acquire $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ maps in calf musculature. Medial gastrocnemius (MG) and soleus $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentrations were plotted over time, and $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ recovery time, $\tau$, was determined. Mitochondrial function index was calculated as the ratio of resting $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ to $\tau .$ Participants underwent a second exercise protocol for imaging of skeletal muscle blood flow (SMBF), and its association with $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ was assessed.

Statistical Tests:

Unpaired t-tests and Pearson correlation coefficient. A p value < 0.05 was considered statistically significant.

Results:

$\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentrations in MG and soleus displayed expected declines during exercise and returns to baseline during recovery. $\tau$ was significantly longer in the T2DM+ cohort (MG 83.5±25.8 vs 54.0±21.1, Soleus 90.5±18.9 vs 51.2±14.5). The mitochondrial function index in the soleus was signficantly lower in the T2DM+ cohort (0.33±0.08 vs 0.66±0.19). SMBF was moderately correlated with the $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ in T2DM−; this correlation was not significant in T2DM+ (r = −0.23, p = 0.269).

Conclusion:

The CEST MRI method is feasible for quantifying $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ in peripheral muscle tissue. T2DM+ individuals had significantly lower oxidative capacities than T2DM− individuals. In T2DM, skeletal muscle metabolism appeared to be decoupled from perfusion.

Introduction

Type II Diabetes Mellitus (T2DM) is a progressive metabolic disorder characterized by resistance to insulin leading to hyperglycemia [1]. Progression of T2DM may lead to neuropathy, peripheral artery disease (PAD), chronic kidney disease, among other complications [2]. T2DM is linked to impaired mitochondrial function in muscle tissue which may precede micro-vascular disorders, evidenced by mitochondrial morphologic changes, reduced density, and decreased mRNA expression in genes playing a role in oxidative phosphorylation [3]. Non-invasive diagnostic methods to characterize mitochondrial function in lower extremity tissue are desirable to inform long-term care.

Measurements of mitochondrial function in humans has focused on techniques that are able to quantify metabolic substrates. Mitochondria are primarily responsible for oxidizing metabolic energy through the Krebs cycle to produce adenosine triphosphate (ATP) to fuel other cellular processes [4]. While ATP is the primary energy currency of muscle cells, creatine (Cr) is a physiologic compound crucial to accommodating muscle contraction. Cr may accept inorganic phosphate (P_i) from ATP under conditions of ATP surplus, producing phosphocreatine (PCr). PCr may then donate P_i to ADP under conditions of ATP depletion for rapid replenishment mediated enzymatically by Creatine Kinase [5]:

$$\mathrm{C}\mathrm{r}+\mathrm{A}\mathrm{T}\mathrm{P}\leftrightarrow \mathrm{P}\mathrm{C}\mathrm{r}+\mathrm{A}\mathrm{D}\mathrm{P}+{\mathrm{H}}^{+}$$ (1)

Thus, imaging methods sensitive to Cr and/or PCr can be used as a proxy for visualizing the energy use of tissue in place of ATP. In particular, methods of analyzing the kinetics of PCr depletion and recovery are desirable to provide insight regarding mitochondrial function and health [6, 7].

³¹P magnetic resonance spectroscopy (MRS) offers a direct method of measuring PCr and ATP concentrations in tissue, and normalized PCr peaks throughout an exercise-recovery protocol can provide dynamic measurements of PCr kinetics with good temporal resolution [6 – 8]. However, additional hardware requirements impede the clinical utility of ³¹P MRS; additionally, this method does not provide spatial resolution for region of interest assessment [6]. Methods for dynamic ³¹P imaging of PCr are emerging but given the low concentration of PCr relative to water, the spatial resolution is limited by the need to achieve adequate temporal resolution and SNR [8]. Muscle biopsies may be performed but this approach is inherently invasive, subject to sampling errors, challenged by delayed metabolic arrest in biopsied tissue, and limited to tissue close to the skin [7].

Chemical Exchange Saturation Transfer (CEST) imaging is a relatively new ¹H MRI technique whereby compounds with hydrogen atoms that undergo kinetic exchange with water can be measured with high spatial resolution [9]. Effective CEST imaging in muscle tissue has been demonstrated using either Cr or PCr as the compound of interest; PCr was chosen for the current study as its chemical shift – approximately 2.6 ppm – is further from water than Cr (~ 2 ppm), which may enable more reliable detection of the PCr spectrum, especially in a clinical 3.0 T MRI system [10, 11]. The 2 ppm chemical shift of Cr may also be more strongly impacted by labile peptides, which can limit the efficacy of signal interpretation [12].

Thus the aim of this pilot study was to develop a ¹H-based MRI method to dynamically quantify PCr concentrations in a clinical 3.0 T MRI system with relatively high spatial resolution and then to assess the kinetics of skeletal muscle PCr ($\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$) as a non-invasive proxy for the assessment of mitochondrial health in individuals with and without T2DM.

Methods

Theory

¹H-based $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ imaging is developed from the MRI CEST technique. MRI CEST contrast occurs when ¹H atoms bound to a molecule with a Larmor frequency distinct from free water’s frequency are saturated by a B₁ radiofrequency (RF) pulse-train [9]. The spontaneous transfer of saturated ¹H atoms between the target molecules and water allows the saturation to be transferred to bulk water in tissue. With an appreciable RF irradiation saturation time, the number of saturated ¹H in the bulk water pool may be orders of magnitude larger than the number of ¹H directly being saturated by B₁, leading to appreciable signal intensity reduction [9]. Magnetization recovery during CEST is characterized mathematically by the Bloch-McConnell equations [13].

For a system of three ¹H pools of distinct Larmor frequencies, where pool “a” may exchange with pools “b” and “c”, the Bloch-McConnell equations in a rotating frame are given by [13]:

$$\frac{\mathrm{d}\stackrel{⃑}{\mathrm{M}}}{\mathrm{d}\mathrm{t}}=\mathbf{A}{\stackrel{⃑}{\mathrm{M}}}_0+\stackrel{⃑}{\mathrm{B}}$$ (2)

where,

$$\overrightarrow{\mathrm{M}}=\left[\begin{array}{ccccccccc}{\mathrm{M}}_{\mathrm{x}}^{\mathrm{a}}& {\mathrm{M}}_{\mathrm{y}}^{\mathrm{a}}& {\mathrm{M}}_{\mathrm{z}}^{\mathrm{a}}& {\mathrm{M}}_{\mathrm{x}}^{\mathrm{b}}& {\mathrm{M}}_{\mathrm{y}}^{\mathrm{b}}& {\mathrm{M}}_{\mathrm{z}}^{\mathrm{b}}& {\mathrm{M}}_{\mathrm{x}}^{\mathrm{c}}& {\mathrm{M}}_{\mathrm{y}}^{\mathrm{c}}& {\mathrm{M}}_{\mathrm{z}}^{\mathrm{c}}\end{array}\right]$$

$$\overrightarrow{{\mathrm{M}}_0}=\left[\begin{array}{ccccccccc}0& 0& {\mathrm{M}}_0^{\mathrm{a}}& 0& 0& {\mathrm{M}}_0^{\mathrm{b}}& 0& 0& {\mathrm{M}}_0^{\mathrm{c}}\end{array}\right]$$

$$\mathbf{A}=\left[\begin{array}{ccccccccc}-{\mathrm{k}}_{2\mathrm{a}}& -\left({\mathrm{\omega }}_{\mathrm{a}}-\mathrm{\omega }\right)& 0& {\mathrm{C}}_{\mathrm{b}}& 0& 0& {\mathrm{C}}_{\mathrm{c}}& 0& 0\\ \left({\mathrm{\omega }}_{\mathrm{a}}-\mathrm{\omega }\right)& -{\mathrm{k}}_{2\mathrm{a}}& -{\mathrm{\omega }}_1& 0& {\mathrm{C}}_{\mathrm{b}}& 0& 0& {\mathrm{C}}_{\mathrm{c}}& 0\\ 0& {\mathrm{\omega }}_1& -{\mathrm{k}}_{1\mathrm{a}}& 0& 0& {\mathrm{C}}_{\mathrm{b}}& 0& 0& {\mathrm{C}}_{\mathrm{c}}\\ {\mathrm{C}}_{\mathrm{a}\mathrm{b}}& 0& 0& -{\mathrm{k}}_{2\mathrm{b}}& -\left({\mathrm{\omega }}_{\mathrm{b}}-\mathrm{\omega }\right)& 0& 0& 0& 0\\ 0& {\mathrm{C}}_{\mathrm{a}\mathrm{b}}& 0& \left({\mathrm{\omega }}_{\mathrm{b}}-\mathrm{\omega }\right)& -{\mathrm{k}}_{2\mathrm{b}}& -{\mathrm{\omega }}_1& 0& 0& 0\\ 0& 0& {\mathrm{C}}_{\mathrm{a}\mathrm{b}}& 0& {\mathrm{\omega }}_1& -{\mathrm{k}}_{1\mathrm{b}}& 0& 0& 0\\ {\mathrm{C}}_{\mathrm{a}\mathrm{c}}& 0& 0& 0& 0& 0& -{\mathrm{k}}_{2\mathrm{c}}& -\left({\mathrm{\omega }}_{\mathrm{c}}-\mathrm{\omega }\right)& 0\\ 0& {\mathrm{C}}_{\mathrm{a}\mathrm{c}}& 0& 0& 0& 0& \left({\mathrm{\omega }}_{\mathrm{c}}-\mathrm{\omega }\right)& -{\mathrm{k}}_{2\mathrm{c}}& -{\mathrm{\omega }}_1\\ 0& 0& {\mathrm{C}}_{\mathrm{a}\mathrm{c}}& 0& 0& 0& 0& {\mathrm{\omega }}_1& -{\mathrm{k}}_{1\mathrm{c}}\end{array}\right]$$

$$\stackrel{⃑}{\mathrm{B}}=\left[\begin{array}{ccccccccc}0& 0& \frac{{\mathrm{M}}_0^{\mathrm{a}}}{{\mathrm{T}}_{1\mathrm{a}}}& 0& 0& \frac{{\mathrm{M}}_0^{\mathrm{b}}}{{\mathrm{T}}_{1\mathrm{b}}}& 0& 0& \frac{{\mathrm{M}}_0^{\mathrm{c}}}{{\mathrm{T}}_{1\mathrm{c}}}\end{array}\right]$$

${\mathrm{M}}_0^{\mathrm{i}}$ is the initial magnetization of ¹H in a pool $\mathrm{i}$ prior to RF irradiation, ${\mathrm{C}}_{\mathrm{a}\mathrm{b}}$ and ${\mathrm{C}}_{\mathrm{a}\mathrm{c}}$ are the rate constants for ¹H exchanging out of pool a, ${\mathrm{C}}_{\mathrm{b}}$ and ${\mathrm{C}}_{\mathrm{c}}$ are the rate constants for ¹H exchanging out of pools $\mathrm{b}$ and $\mathrm{c}$ respectively, and ${\mathrm{\omega }}_{\mathrm{i}}$ is the frequency offset of ¹H in a given pool from water. $\mathrm{\omega }$ is the RF frequency offset, and ${\mathrm{\omega }}_1$ is the RF amplitude. Additionally,

$${\mathrm{k}}_{1\mathrm{a}}=\frac{1}{{\mathrm{T}}_{1\mathrm{a}}}+{(\mathrm{C}}_{\mathrm{a}\mathrm{b}}+{\mathrm{C}}_{\mathrm{a}\mathrm{c}})$$

$${\mathrm{k}}_{2\mathrm{a}}=\frac{1}{{\mathrm{T}}_{2\mathrm{a}}}+{(\mathrm{C}}_{\mathrm{a}\mathrm{b}}+{\mathrm{C}}_{\mathrm{a}\mathrm{c}})$$

with ${\mathrm{k}}_{1\mathrm{b}},{\mathrm{k}}_{2\mathrm{b}},{\mathrm{k}}_{1\mathrm{c}}$, and ${\mathrm{k}}_{2\mathrm{c}}$ following analogous form.

In our application, the three ¹H pools ($\mathrm{a},\mathrm{b}$, and $\mathrm{c}$) represent water, $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$, and magnetic transfer contrast (MTC), a saturation exchange effect between water and macromolecules, respectively.

CEST imaging can be performed at a series of offset frequencies to allow for a Z-spectrum to be constructed at each pixel in the set of images, representing how the signal intensity changes with offset frequency and B₀ field inhomogeneity [9]. These Z-spectra can then be fitted to the Bloch-McConnell equations to quantitatively determine the concentration of the target molecule – $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ in the current study [14, 15]. Initial conditions for concentrations in this study were normalized to the concentration of protons in the water pool; the concentration of water in muscle tissue was estimated as 75% water by mass [16]. As the fractional concentration of protons in pools relative to water is proportional to the magnetizations in $\overrightarrow{{\mathrm{M}}_0}$, the concentration of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ is then calculated from $\overrightarrow{{\mathrm{M}}_0}$ as:

$$\left[SMPCr\right]=\left(\frac{M_{SMPCr}}{M_{H2O}}\right)\left({\left[H^{+}\right]}_{H_{2}O}\right)$$ (3)

In vivo MRI

This study was approved by the local institutional review board, and written consent was obtained from each participant prior to the study. Seven subjects without T2DM (T2DM−) and seven age, sex, body-mass-index (BMI) matched subjects with T2DM (T2DM+) were recruited. The T2DM− group had no symptomatology or documented evidence of T2DM, smoking, peripheral neuropathy, or underlying PAD. The T2DM+ group had T2DM as determined by their treating physician following well-defined national guidelines, and had no history of PAD, smoking, or peripheral neuropathy [17]. Exclusion criteria for both groups were contraindications to MR scanning (intracranial vascular clips, pacemaker, or intraocular metal), pregnancy, and comorbidities that severely limited the patient’s ability to perform a modest calf-contraction test.

The in vivo imaging sessions were performed on a 3T Prisma Siemens whole-body MR system (Siemens Healthcare, Malvern, PA). A commercial flexible surface coil (receive-only) was used for calf imaging. Each participant underwent a rest-exercise-recovery imaging protocol (3-min rest, 4-min exercise, and 6-min recovery) within the MR scanner bore. A custom-made ergometer was used in this study for standardized isometric plantarflexion contraction [18]. During the exercise each participant plantarflexed their right ankle, pressing onto a pressurized air bulb that was connected through a polyethylene tube to a pressure regulator. The latter has one end connected to a custom-made pressure-voltage transducer that converts voltages to digits for real time display of the pressure forces. The maximal voluntary contraction (MVC) of the plantarflexion was determined for each participant prior to each exercise session. The participant was instructed to perform the isometric exercise for 4 min at approximately 40% MVC force level. All subjects were able to perform the 40% MVC isometric exercise without complaints of pain or discomfort.

The MRI sequence for the CEST scan was a single-slice, continuous wave saturation pulse prepared single-shot rapid acquisition with refocusing echoes (RARE) sequence [19]. The saturation power was 0.6 μT with a saturation time of 800 ms achieved by using ten 80 ms Gaussian shaped pulses. The area of largest calf cross section was first identified on scout images for selection of axial calf slice for CEST imaging. A total of 31 frequency offsets were acquired to generate a Z-spectrum from 1.3 to 3.5 ppm (0.1 ppm increment) at a smaller increment around 2.5 ppm (0.05 ppm increment) to facilitate capturing of the $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ peak in the Z-spectrum. Other imaging parameters for the CEST scan were TR/TE = 1860 ms/7.5 ms, field-of-view (FOV) = 220 × 220 mm², matrix = 76 × 76, slice thickness = 8 mm, and data acquisition time of 60s for one $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ map. Prior to CEST imaging, Water Saturation Shift Referencing (WASSR) images were obtained in order to correct for B₀ inhomogeneity occurring in a clinical 3T scanner, with offset frequencies ranging from −1.2ppm to 1.2ppm [20]. The $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ imaging continued for 13 min to cover the entire rest-exercise-recovery protocol time.

In addition to $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ imaging, participants also underwent a second exercise protocol for dynamic imaging of skeletal muscle blood flow (SMBF) by utilizing an established contrast free arterial spin labeling (ASL) method [21]. This approach utilized a Flow-sensitive-Alternative-Inversion-Recovery type of ASL to acquire slice-selective (SS) and nonselective (NS) inversion-recovery (IR) prepared images. Multiple single-shot gradient-echo data acquisitions were executed to obtain images at different T₁. The SMBF was calculated as [22]:

$$SMBF=\lambda \frac{T_{1,ns}}{T_{1,blood}}(\frac{1}{T_{1,ss}}-\frac{1}{T_{1,ns}})$$ (4)

Where $\lambda$ is a constant blood-tissue partition coefficient of water (0.92 ml/g), $T_{1,\mathrm{n}\mathrm{s}}$ is the ${\mathrm{T}}_1$ of skeletal muscle obtained with the nonselective inversion recovery (IR) pulse; ${\mathrm{T}}_{1,\mathrm{B}\mathrm{l}\mathrm{o}\mathrm{o}\mathrm{d}}$ is the ${\mathrm{T}}_1$ of blood pool and set as 1871 msec at 3T [23]; and ${\mathrm{T}}_{1,\mathrm{S}\mathrm{S}}$ is the ${\mathrm{T}}_1$ of skeletal muscle obtained with the slice-selective IR pulse.

The ASL sequence parameters for calf imaging were: eight single-shot gradient-echo acquisitions after each IR pulse, TR/TE = 4.0 msec/1.3 msec; TI = 220, 720, 1220, …., 3720 msec, fat saturation; flip angle = 5°; image resolution = 1.7 × 1.7 × 8 mm³; FOV = 220 mm x 220 mm; matrix size = 128 × 128; slice thickness = 8 mm; average number = 4, temporal resolution = 1 min.

Image Analysis

Lorentzian fitting of the WASSR spectrum was first conducted pixel-by-pixel on WASSR image sets, and the offset of minima were used to generate a B₀ inhomogeneity map. The map was then used to correct the Z-spectra, addressing shifts in the $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ peak. Corrected Z-spectra were fitted to the Bloch-McConnell equations following a linear 3-pool system model to produce PCr images. Identical initial conditions, upper, and lower bounds were used in fitting Z-spectra for all patients and healthy volunteers: T_1PCr = 50ms (20ms – 80ms), T_2PCr = 20ms (0ms – 50ms), [PCr] = 0mM (0mM – 200mM), k_PCr-H2O = 200s⁻¹ (100s⁻¹ – 300s⁻¹), and ppm_PCr = 2.5 (2.4 – 2.6). An example of a B₀ inhomogeneity map and Z-spectrum with Bloch-fitting are presented in Figure 1. Computations were performed by utilizing the facilities of the University’s Center for High Performance Computing, including 4 CPU nodes on an Intel 16-Core Xeon Gold 6226R 2.9GHz processor. The SMBF maps were created using a previously established method [21].

Figure 1.

B₀ inhomogeneity map of the calf obtained from WASSR images (a), and a sample Z spectrum with associated Bloch-Fitting (b).

Two regions-of-interest (ROIs) were drawn on the medial gastrocnemius and soleus muscles on the maps of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ and SMBF. An anatomic image obtained prior to the rest-exercise-recovery protocol was used as reference for polygonal ROIs to be drawn; deformities of the musculature during plantarflexion slightly changed their relative position during exercise.

Anterior muscle groups such as tibialis anterior were not included in the analysis. The exercise protocol design was optimized to test the plantarflexors in the posterior compartment, and the isometric plantarflexion exercise performed by the participants would not reliably engage the tibialis anterior. The absolute $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentration was plotted as a function of scanning time after the start of the exercise-recovery protocol. The kinetics of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ recovery is modeled as first-order, as used in dynamic ³¹P MRS [6]:

$$\left[\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}\right](\mathrm{t})={\left[\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}\right]}_{\mathrm{m}\mathrm{i}\mathrm{n}}+({\left[\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}\right]}_{\mathrm{m}\mathrm{a}\mathrm{x}}-{\left[\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}\right]}_{\mathrm{m}\mathrm{i}\mathrm{n}})\left(1-{\mathrm{e}}^{-\frac{\mathrm{t}}{\mathrm{\tau }}}\right)$$ (5)

Rest $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ data were averaged and appended to 6 minutes after the recovery data as one additional data point to facilitate the fitting process. Time constants ($\tau$) following exercise were calculated by fitting recovery data to first order kinetics Eq.(3) with a least-squares regression algorithm. This $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ recovery time-constant $\tau$ is a metric that inversely describes mitochondrial oxidative phosphorylation function [30]. The maximum rate of ATP oxidative synthesis, ${\mathrm{Q}}_{\mathrm{m}\mathrm{a}\mathrm{x}}$, a biomarker for mitochondrial function of tissue can be estimated as the product of the $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ recovery rate constant ($k=\frac{1}{\tau }$) and rest $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ [23]. Exercise images of either $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ or SMBF with severe motion artifacts, for example bloch-fitting to boundary conditions throughout an image, were excluded for further analysis.

Statistical Analysis

Analyses were focused on assessing differences in $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ and SMBF data sets between T2DM− and T2DM+ groups. A double sided Student’s t-test was used to compare the means of two groups under assumption of a normal distribution. Correlation analysis was then performed to assess associations between SMBF and $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ for the entire rest-exercise-recovery period using Pearson correlation coefficient. The level of statistical significance was set at $p<0.05$. Statistical analyses were performed with MedCalc Statistics for Biomedical Research Version 18.2.1 (MedCalc Software, Ostend, Belgium).

Results

Dynamic $\mathit{S}\mathit{M}\mathit{P}\mathit{C}\mathit{r}$ findings

A summary of demographic data is presented in Table 1. In a total of 28 exercise images (4 time points per subject) for either $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ or SMBF in each group, 4 total images in T2DM− across 4 subjects and 4 total images in T2DM+ across 2 subjects were excluded from analysis due to artifacts. An example of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentration maps is shown in Figure 2.

Table 1:

Demographics of the volunteers with and without Type 2 Diabetes Mellitus

	T2DM+	T2DM−	p value
Male	4	4	NS
Female	3	3	NS
Age	63.4 (51, 79)	56.4 (43, 74)	NS
BMI	28.5 (20.5, 35.9)	26.5 (22.4, 30.4)	NS

T2DM, Type 2 diabetes mellitus, +with, -without, BMI, body mass index; NS, not significant (between T2DM− and T2DM+). Full range presented in brackets.

Figure 2.

Spatially resolved $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ maps in time series of rest-exercise-recovery in a subject without T2DM (a) and a subject with T2DM (b). Anatomic images have ROIs for the Medial Gastrocnemius (MG) and soleus muscles labelled in green and blue respectively. Rows are separated by which phase in the exercise-recovery protocol the data was acquired. A depiction of the plantarflexion and imaging protocol is depicted below.

ROI analysis was conducted by two individuals, one with five years of experience and one with <1 year experience. The intraclass correlation coefficient (ICC) was 0.73 ($\mathrm{p}<0.0001$), indicating overall good correlation between individuals.

The average trend throughout the rest-exercise-recovery protocol in medial gastrocnemius and soleus muscle groups is displayed in Figure 3 as absolute concentrations of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ in mM. $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentrations over time throughout the rest-exercise-recovery protocol decreased as exercise was performed, with a recovery to baseline after the cessation of the exercise. $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ recovery time $\mathrm{\tau }$ in MG and soleus muscles is presented in Figure 4, demonstrating a generally higher $\mathrm{\tau }$ and lower ${\mathrm{Q}}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ in the T2DM+ cohort. This is further analyzed in Table 2, which provides mean values of the $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentration at rest, $\mathrm{\tau }$, and ${\mathrm{Q}}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ for both the T2DM− and T2DM+ cohorts. Compared to the T2DM− cohort, the resting $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentrations were signficantly lower and recovery time $\mathrm{\tau }$ was significantly prolonged in the T2DM+ cohort, in both the MG and soleus muscles. However, ${\mathrm{Q}}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ was significantly lower only in the soleus muscle in T2DM+.

Figure 3.

Absolute concentrations of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ throughout the exercise-recovery protocol in Medical Gastrocnemius (MG) (a) and soleus (b) muscles. The region marked with a red overlay corresponds to the period of isometric plantarflexion. There is a decline in the $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentration throughout the duration of exercise with a return to baseline during recovery.

Figure 4.

Time constants ($\tau$) for the recovery of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ in T2DM− and T2DM+ groups for the medial gastrocnemius (MG) and soleus (a) muscles. Calculated ${\mathrm{Q}}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ (metric to estimate ATP synthesis rate) in the T2DM+ and T2DM− groups for the MG and soleus muscles (b).

Table 2:

Quantitative results from SMPCr measurements in the MG and Soleus muscles in volunteers with and without Type 2 Diabetes Mellitus

	Rest SMPCr (mM)	τ (s)	Qmax (mM/s)
T2DM− in MG	32.3 ± 3.1	54.0 ± 21.1	0.70 ± 0.33
T2DM+ in MG	29.7 ± 2.7**	83.5 ± 25.8*	0.41 ± 0.20
T2DM− in Soleus	31.5 ± 2.8	51.2 ± 14.5	0.66 ± 0.19
T2DM+ in Soleus	29.3 ± 2.5*	90.5 ± 18.9**	0.33 ± 0.08**

^*p < 0.05

^**p < 0.01, comparison between T2DM− and T2DM+;

SMPCr, skeletal muscle phosphocreatine; T2DM, Type 2 diabetes mellitus, +with, -without; MG, medial gastrocnemius; Q_max, metric for ATP synthesis rate

Correlation of $\mathit{S}\mathit{M}\mathit{P}\mathit{C}\mathit{r}$ and SMBF

The normalized SMBF and $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ relative to the resting values in MG muscle throughout the protocol are presented in Figure 5, showing slower recovery of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ and lower exercise SMBF magnitude in the T2DM+, compared to T2DM−. The correlation of absolute $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ and SMBF in MG muscle during exercise is presented in Figure 6, demonstrating a significant modest negative correlation in the T2DM− cohort (r = −0.47), but no significant correlation in the T2DM+ cohort (r = −0.23, p = 0.269).

Figure 5.

Dynamic display of normalized SMBF curves in Medial Gastrocnemius (MG) (a) and soleus (b) muscles, as well as normalized $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentration curves in MG (c) and soleus muscle (d) throughout the exercise-recovery protocol. Regions marked with a red overlay correspond to the period of isometric plantarflexion.

Figure 6.

Correlation of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ and SMBF in the medial gastrocnemius (MG) of T2DM− (a) and T2DM+ (b) subjects during exercise.

Discussion

The current study demonstrates the feasibility of noncontrast ¹H MRI imaging of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ in peripheral muscle tissue to assess differences in muscle metabolism between heathy controls and patients with diabetes. Imaging was performed by utilizing a clinical 3T MRI scanner with no exogenous contrast agent or additional hardware. The resting $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentrations in T2DM− were approximately 30 mM, consistent with the $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentrations in previous studies from a variety of methods, including ³¹P MRS [25–28]. The kinetics of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ depletion and recovery throughout the rest-exercise-recovery protocol agreed with those detected by ³¹P MRS, and the depletion of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ negatively correlated with the increased SMBF during exercise. The $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ recovery time $\mathrm{\tau }$ in T2DM+ was prolonged, as also demonstrated in previous studies employing ³¹P MRS data acquisition methods [27, 28]. The mitochondrial function (${\mathrm{Q}}_{\mathrm{m}\mathrm{a}\mathrm{x}}$), an empirical index of maximal rate of mitochondrial ATP synthesis, in T2DM+ was significantly lower in the soleus muscle, but this difference did not reach statistical significance in MG. This observation in MG muscle may be partially explained by relatively low exercise strength (40% MVC) and a small sample size (measurement and subject variability).

The $\mathrm{\tau }$ values in T2DM− were higher than those in previous studies, measured by using ³¹P MRS [27, 28]. For example, Lanza et al. reported an average $\mathrm{\tau }$ of 29 s in a healthy cohort with an average age of 38 years, and Menon et al. reported an average $\mathrm{\tau }$ of 26 s in a healthy cohort with an average age of 25 years [28, 29]. The calculated ${\mathrm{Q}}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ in T2DM− was also lower than those reported in previous studies [28, 30, 31]. These discrepancies may be partially attributed to the difference in ages of subjects tested, in addition to healthy status of individuals (e.g, fitness) and potential measurement errors due to the limited temporal resolution. A previous study in muscle function and aging has found that $\mathrm{\tau }$ significantly increases with age, possibly due to changes in resting muscle perfusion leading to altered intramuscular metabolism [32].

Targeting creatine as an endogenous contrast agent (CrCEST) can also provide information regarding ATP metabolism and mitochondrial function, following an inverse relationship with PCr. CrCEST has been employed to quantify reaction kinetics in a variety of studies focusing on plantarflexion exercise and calf musculature in healthy volunteers [10, 33]. DeBrosse et al studied reaction kinetics following plantarflexion in both healthy volunteers as well as patients with primary genetic mitochondrial disorders [34]. PCr recovery kinetics of healthy volunteers in our study agreed with the kinetics of Cr depletion during recovery in healthy volunteers in these previous studies.

Previous work conducted by Kogan et al demonstrated that, under cuff-induced leg hypoxia and hyperemia in 5 healthy volunteers, modulated blood flow alone after the leg hypoxia may not appreciably impact $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ concentrations measured by CrCEST or ³¹P MRS methods [35]. In the current study, we further explored this relationship during plantarflexion exercise, rather than after the exercise. As the $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ decreased during the exercise portion of the protocol, SMBF increased proportionally in the T2DM− cohort. Similar findings were reported in electrically stimulated muscle of rats [36] and in the calf muscle of healthy volunteers during a 3-min plantarflexion exercise [37]. However, in the T2DM+ cohort this relationship was less clear. It could be speculated that this lack of significant correlation between SMBF and $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ may reflect an uncoupling of calf muscle metabolism and perfusion in the T2DM+ cohort. A similar observation has been reported in a prior study conducted using ³¹P MRS in patients with PAD following treadmill exercise, which indicated that mitochondrial dysfunction may be independent from tissue oxygenation [30]. Nevertheless, a large sample size is needed to confirm these findings to draw solid conclusion.

Future study employing this technique may include longitudinal evaluation of patients, e.g, individuals with PAD and DM, heart failure, etc, to better characterize their muscle tissue metabolism through a variety of interventions such as exercise programs, or procedures such as stenting or bypass.

Limitations

First, this study had a very limited number of participants. Second, as discussed in Methodology, each $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ map was calculated from 31 CEST images, which resulted in a net temporal resolution of 1 min. This resolution makes capturing a recovery time constant on the order of 30s challenging, with a tendency to overestimate and have higher variance. The use of convolutional neural networks may help to reduce the number of offset frequencies needed for $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ map production and improve the temporal resolution. This is an ongoing effort in our laboratory. Third, the exercise portion of the protocol could result in motion artifacts which present a major challenge in producing precise Z-spectra for fitting. Producing concentration maps requires that the CEST images from different frequency offsets physically align precisely with one another. Motion artifacts therefore greatly hinder this. It is expected that improved temporal resolution may reduce motion artifacts. An assessment of test-retest reproducibility by having subjects undergo additional $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ and SMBF imaging to quantify the consistency of this methodology was not conducted in this pilot study. A rigorous study for reproducibility measurements of $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ and the validation against reference method 31P MRS will be the task in near future in our laboratory.

Conclusion

This study presents a technique for the quantitative assessment of mitochondrial dysfunction through $\mathrm{S}\mathrm{M}\mathrm{P}\mathrm{C}\mathrm{r}$ measurements in the calf musculature of diabetic patients using ¹H MRI. Large scale of studies will be warranted to better define the clinical applications of this technique in metabolic disease.