Assessment of In Vivo Mitochondrial Metabolism by Magnetic Resonance Spectroscopy

Abstract

Magnetic resonance spectroscopy (MRS), a companion technique to the more familiar MRI scan, has emerged as a powerful technique for studying metabolism noninvasively in a variety of tissues. In this article, we review two techniques that we have developed which take advantage of the unique characteristics of ³¹P and ¹³C MRS to investigate two distinct parameters of muscle mitochondrial metabolism; ATP production can be estimated by using the ³¹P saturation-transfer technique, and oxidation via the TCA cycle can be modeled from ¹³C MRS data obtained during the metabolism of a ¹³C-labeled substrate. We will also examine applications of the techniques to investigate how these parameters of muscle mitochondrial metabolism are modulated in insulin resistant and endurance trained individuals.

1. Introduction

The prevalence of type 2 diabetes has dramatically increased over the past 25 years, and has been predicted to maintain this expansion over the next two decades (Wild et al., 2004). Although causal mechanisms remain to be elucidated, the development of insulin resistance assumes a key, early role in the pathogenesis of most cases of diabetes (Haffner et al., 1990; Lillioja et al., 1987; Warram et al., 1990). Elevated, fasting plasma fatty acid concentrations are a hallmark of insulin resistance in obese and diabetic individuals (Boden and Shulman, 2002), and an inverse relationship between insulin sensitivity and circulating fatty acid levels has also been observed in normal-weight offspring of type 2 diabetic patients (Perseghin et al., 1997). Skeletal muscle triglyceride (TG) content, estimated from biopsy samples, was found to be negatively correlated with insulin sensitivity (Pan et al., 1997) and more recently, several studies have demonstrated an even stronger association with intramyocellular lipid (IMCL) content, estimated using magnetic resonance (MR) techniques (Krssak et al., 1999; Perseghin et al., 1999; Szczepaniak et al., 1999), in healthy, as well as insulin resistant individuals. This relationship is conserved in other insulin-responsive tissues, for example, liver (Petersen et al., 2002, 2005; Seppala-Lindroos et al., 2002), suggesting that dysregulation of lipid metabolism may contribute to insulin resistance in these individuals. Substantiated by numerous animal studies, mechanisms for lipid-induced insulin resistance have been proposed (Morino et al., 2006; Savage et al., 2007; Shulman, 2000), that attribute impaired insulin action in muscle and liver to the presence of lipid intermediates, in particular diacylglycerol (DAG), rather than intracellular TG per se, which interfere with the insulin-signaling cascade.

1.1. Mitochondrial fatty acid oxidation and insulin resistance

The net accumulation of intracellular lipid metabolites in muscle could either be caused by increased delivery and uptake of fatty acids, by decreased utilization, or a combination of these factors (Shulman, 2000). Intriguingly, a number of muscle biopsy studies demonstrated that the occurrence of obesity and/or diabetes was accompanied by a decrease in the content and activity of mitochondrial enzymes (Colberg et al., 1995; He et al., 2001; Kelley et al., 1999; Ritov et al., 2005; Simoneau and Kelley, 1997; Simoneau et al., 1999) suggesting that impaired fatty acid oxidation may contribute to the pathogenesis of these conditions. Mitochondrial density was also found to be reduced in the muscle of obese and type 2 diabetics accompanied by smaller and morphologically damaged mitochondria (Kelley et al., 2002) and analysis of gene-chip datasets indicated that down-regulation of genes associated with oxidative metabolism may occur in a coordinate manner in obese, type 2 diabetic individuals (Mootha et al., 2003; Patti et al., 2003) and overweight, nondiabetic subjects with a family history of diabetes (Mootha et al., 2003; Patti et al., 2003). However, one limitation of these muscle biopsy studies is that they provide an index of mitochondrial capacity but may not represent function. We therefore sought to implement techniques by which muscle mitochondrial function could be determined in vivo.

2. In Vivo Magnetic Resonance Spectroscopy

Since its initial demonstration as a viable methodology for in vivo research (Hoult et al., 1974), Magnetic resonance spectroscopy (MRS) has emerged as a powerful technique for studying metabolic function in a variety of tissues in both animals and humans (Dobbins and Malloy, 2003; Prompers et al., 2006; Shulman et al., 1996). MRS is a companion technique to the more familiar magnetic resonance imaging (MRI) scan, which is in widespread use providing clinical anatomical images. Unlike MRI, which measures the spatial distribution of water in the region of interest, MRS determines the content of MR visible nuclei. Isotopes of many biologically relevant elements are MR visible, including hydrogen, carbon, and phosphorus. Each has a characteristic resonance frequency in the presence of a strong static magnetic field, i.e., within the MR scanner. What makes MRS of particular utility is that the chemical environment of a nucleus also influences its precise frequency, a phenomenon which gives rise to the MRS spectrum. Therefore, within a tissue of interest, the content of specific compounds and their constituent chemical groups can be resolved and determined using in vivo MRS. Figure 21.1 illustrates examples of in vivo spectra obtained from muscle showing two nuclei of metabolic relevance, phosphorus (³¹P) and carbon (¹³C).

Figure 21.1.

Typical ³¹P (A) and ¹³C (B) MRS spectra acquired from the soleus-gastrocnemius muscle complex in vivo.

The advantages of MRS as a tool for investigating muscle metabolism are considerable. It is safe, noninvasive, involves no ionizing radiation and permits repeated measures of metabolites during a single session, or over an extensive study. It is, therefore, ideal for monitoring dynamic metabolic processes. Multinuclear capabilities enable a plethora of information to be collected and provide the flexibility to study independent parameters of metabolism during the same experiment. However, MRS is a relatively insensitive technique compared to other imaging/spectrometry modalities, for example, PET (Positron Emission Tomography) or mass spectrometry, and metabolite concentrations in the millimolar range are required to be observed in vivo. Specialized equipment is necessary, although spectroscopy packages are now available for the new generation of clinical MR scanners.

As can be seen in panel A, a ³¹P spectrum obtained from muscle is characterized by peaks due to phosphocreatine (PCr), the three (α, β, and γ) phosphate moieties of adenosine tri-phosphate (ATP), unbound inorganic phosphate, and mono- and di-phosphoesters (PME, PDE). With the exception of adenosine di-phosphate (ADP), which is present in too low a concentration to be observed in resting muscle, the principal metabolites involved in energy transduction can be observed which makes in vivo ³¹P MRS particularly useful for studying muscle energetics.

Since all organic compounds contain carbon atoms, ¹³C MRS can provide information on a huge number of compounds involved in muscle metabolism. High resolution ¹³C spectra of muscle extracts contain a myriad of overlapping peaks corresponding to different metabolites and their constituent carbons. In vivo, this situation is somewhat simplified, due to the low natural abundance of ¹³C (1.1% of total carbon) and the high concentration of lipids compared to other metabolites. An in vivo muscle ¹³C spectrum (Fig. 21.1B) is dominated by peaks due to the fatty acid and glycerol components of lipid, accompanied by peaks of other metabolites present in relatively high concentrations. However, this potential drawback can be turned to an advantage by infusing a ¹³C-labeled compound and using ¹³C MRS to monitor the metabolism of this tracer.

This article focuses on two techniques we have developed which take advantage of the unique characteristics of ³¹P and ¹³C MRS to investigate different facets of human muscle mitochondrial metabolism. The terminal stage of mitochondrial energy production can be estimated by using a ³¹P magnetization-transfer technique to determine the unidirectional flux of P_i → ATP (“ATP synthesis”), and oxidation via the TCA cycle can be assessed by applying a metabolic model to ¹³C MRS data obtained during the oxidation of a ¹³C-labeled substrate. After providing a thorough description of the implementation of the techniques, we will describe how these parameters of muscle metabolism are modulated in insulin resistant and endurance trained individuals.

2.1. Examining exchange reactions by magnetization-transfer MRS

Although ATP can be synthesized from several sources, the principal synthetic reaction is from the combination of ADP and P_i which occurs primarily in mitochondria. ATP synthesis is in equilibrium with ATP hydrolysis as shown in Eq. (21.1), forming a phosphate exchange reaction with forward and reverse rate constants, k_f and k_r. The rate of ATP synthesis (V_ATP) is therefore equivalent to the forward rate constant multiplied by the concentrations of ADP and P_i, Eq. (21.2). Under steady-state conditions, where the concentration of ADP can be assumed to remain constant, this relationship can be further simplified such that V_ATP is the product of the pseudo-first-order rate constant (for P_i), $k_{\mathrm{f}}^{\prime }$, and the P_i concentration, Eq. (21.3).

$$\mathrm{ADP}+{\mathrm{P}}_{\mathrm{i}}\underset{{\mathrm{k}}_{\mathrm{r}}}{\overset{{\mathrm{k}}_{\mathrm{f}}}{\rightleftarrows }}\mathrm{ATP}$$ (21.1)

$$V_{\mathrm{ATP}}=k_{\mathrm{f}}\left[\mathrm{ADP}\right]\left[{\mathrm{P}}_{\mathrm{i}}\right]$$ (21.2)

$$V_{\mathrm{ATP}}=k_{\mathrm{f}}^{\prime }\left[{\mathrm{P}}_{\mathrm{i}}\right]$$ (21.3)

The unidirectional fluxes involved in an exchange reaction can be studied using MRS since the magnetic equilibrium of the participant metabolites can be perturbed and monitored, while the chemical equilibrium remains intact. This effect can be accomplished using magnetization-transfer techniques (Forsen and Hoffman, 1963; Hoffman and Forsén, 1966). As mentioned above, a ³¹P MRS spectrum detects signals from the “high-energy phosphates” involved in cellular energetics. Therefore, the ATP synthesis flux (P_i → ATP) can be estimated, since both the ATP and P_i metabolite pools are visible.

During an MRS experiment, the signal of a particular metabolite can be expressed using the Bloch equations (Bloch, 1946), which describe the variation of the MRS signal over time. These can be modified to incorporate chemical exchange (McConnell, 1958). The response of the P_i signal involved in P_i ↔ ATP exchange is shown in Eq. (21.4):

$$\mathrm{d}{M}_{\mathrm{Pi}}∕\mathrm{d}t=(M_0^{\mathrm{Pi}}-M_{\mathrm{Pi}})∕T_1^{\mathrm{Pi}}-k_{\mathrm{f}}^{\prime }\left[M_{\mathrm{Pi}}\right]+k_{\mathrm{r}}\left[M_{\mathrm{ATP}}\right]$$ (21.4)

where M_Pi is the P_i magnetization, ${\mathrm{M}}_0^{\mathrm{Pi}}$ is the equilibrium P_i magnetization, M_ATP is the γ-ATP magnetization, and $T_1^{\mathrm{Pi}}$ is the longitudinal relaxation time of P_i.

During a magnetization-transfer experiment, the γ-phosphate peak of ATP (i.e., that which participates in exchange) can be selectively irradiated using a long-duration, frequency-selective, saturation pulse to equalize its spin-states and eliminate its signal, i.e., M_ATP = 0, which simplifies the Bloch equation to Eq. (21.5).

$$\mathrm{d}{M}_{\mathrm{Pi}}∕\mathrm{d}t=(M_0^{\mathrm{Pi}}-M_{\mathrm{Pi}})∕T_1^{\mathrm{Pi}}-k_{\mathrm{f}}^{\prime }\left[M_{\mathrm{Pi}}\right]$$ (21.5)

Effectively this states that under conditions of γ-ATP saturation, the signal of P_i will decrease due to P_i → ATP flux. A solution to this equation is, Eq. (21.6):

$$M_{\mathrm{Pi}}^{\prime }=M_0^{\mathrm{Pi}}∕\left(1+k_{\mathrm{f}}^{\prime }{T}_1^{\mathrm{Pi}}\right)-M_0^{\mathrm{Pi}}\left\{1-1∕\left(1+k_{\mathrm{f}}^{\prime }{T}_1^{\mathrm{Pi}}\right)\right\}\exp \left\{-\left(k_{\mathrm{f}}^{\prime }+1∕T_1^{\mathrm{Pi}}\right)t\right\}$$ (21.6)

Steady-state saturation is attained following a long duration saturation pulse. Under these conditions t → ∞, the second term in Eq. (21.6) vanishes, and it simplifies to:

$$M_{\mathrm{Pi}}^{\prime }=M_0^{\mathrm{Pi}}∕\left(1+k_{\mathrm{f}}^{\prime }{T}_1^{\mathrm{Pi}}\right)$$ (21.7)

The T₁ relaxation times of P_i under γ-ATP saturated (${\mathrm{T}}_1^{\prime }$ and non-saturated $\left(T_1^{\mathrm{Pi}}\right)$ conditions are related by the expression: $1∕T_1^{\prime }=1∕T_1^{\mathrm{Pi}}+k_{\mathrm{f}}^{\prime }$. Substitution into Eq. (21.7) and rearranging gives the classic equation of saturation transfer, Eq. (21.8).

$$k_{\mathrm{f}}^{\prime }=\left(1-M_{\mathrm{Pi}}^{\prime }∕M_0^{\mathrm{Pi}}\right)∕T_1^{\prime }$$ (21.8)

Therefore, the pseudo-first-order rate constant for ATP synthesis is equivalent to the change in the P_i signal from equilibrium conditions (without saturation, $M_0^{\mathrm{Pi}}$) to γ-ATP saturated conditions ($M_{\mathrm{Pi}}^{\prime }$) expressed in units of the T₁ of P_i under conditions of γ-ATP saturation ($T_1^{\prime }$).

2.2. Measuring muscle ATP synthesis in vivo by ³¹P saturation-transfer MRS

The application of the ³¹P saturation-transfer technique to a biological system was initially used to measure the kinetics of the bacterial ATPase (Brown et al., 1977), and has since been implemented in a variety of in vitro and in vivo systems (Alger et al., 1982; Brindle et al., 1989; Lei et al., 2003; Matthews et al., 1981). We have successfully used the technique to investigate the synthesis of ATP in resting human muscle using 2.1 T and 4 T whole-body MR systems. Experimentally, our methodology incorporates three separate elements: (1) calculation of the saturation-transfer effect (i.e., $M_{\mathrm{Pi}}^{\prime }∕M_0^{\mathrm{Pi}}$) caused by γ-ATP saturation on P_i; (2) calibration of the T₁ of P_i under conditions of γ-ATP saturation (i.e., $T_1^{\prime }$), together these parameters can be inserted into Eq. (21.8) to derive the rate constant for ATP synthesis; and (3) determination of the in vivo P_i concentration (i.e., [P_i]), which enables the unidirectional P_i → ATP flux to be calculated by solving Eq. (21.3).

The saturation-transfer effect can be determined by using a simple pulse-acquire MRS sequence preceded by a long, low power, frequency-selective saturation pulse (Fig. 21.2A). In vivo, muscle ³¹P spectra are typically acquired using surface coils, therefore, adiabatic (B₁ insensitive) pulses are preferred for optimal signal excitation (de Graaf, 2008). Pairs of spectra are acquired with the saturation pulse centered on the γ-ATP resonance or at a downfield frequency equidistant from P_i, to partially compensate for off-resonance effects of the saturation pulse, which can be significant, particularly at lower field strengths (Kingsley and Monahan, 2000a,b). Precise acquisition parameters are system dependent; at 4 T, we use a 9 cm surface coil for ³¹P spectroscopy of the soleus/gastrocnemius muscle complex, an adiabatic half passage (90°) pulse for excitation and a 15 s “soft” pulse for saturation. A typical pair of saturation-transfer spectra acquired with these parameters is shown in Fig. 21.3. The T₁ of P_i under conditions of γ-ATP saturation is measured by an 8-point inversion-recovery calibration using a customized sequence (Fig. 21.2B), with γ-ATP saturation not only prior to the adiabatic full passage (180°) inversion pulse, but also during the variable inversion delay to maintain steady-state saturation of the γ-ATP spins. Inversion recovery data is fitted by a 3-parameter model (MacFall et al., 1987), optimized using a nonlinear least squares algorithm. A fully relaxed ³¹P spectrum acquired without saturation is used to determine intracellular [P_i], assuming a constant muscle ATP concentration of 5.5 mmol/kg (Kemp et al., 2007).

Figure 21.2.

Pulse sequences for the ³¹P saturation-transfer experiment. (A) Pulse-acquire sequence with RF saturation. A long, low power, frequency-selective saturation pulse precedes signal excitation/detection. Adiabatic excitation pulses are optimal for in vivo MRS using surface coils. (B) Inversion-recovery with RF saturation for $T_1^{\prime }$ calibration. Frequency-selective saturation of the γ-ATP peak is applied prior to the adiabatic inversion pulse, and during the variable inversion-delay to maintain steady-state spin saturation throughout the inversion-recovery measurement.

Figure 21.3.

In vivo muscle ³¹P spectra acquired from the soleus-gastrocnemius muscle complex during a saturation-transfer experiment. Spectra are acquired with (black spectrum) and without (gray spectrum) frequency-selective irradiation of the γ-ATP peak. The rate constant of exchange $\left(k_{\mathrm{f}}^{\prime }\right)$ is equivalent to the decrease in the P_i peak upon γATP saturation $(\Delta {\mathrm{P}}_{\mathrm{i}}\Rightarrow 1-M_{\mathrm{Pi}}^{\prime }∕M_0^{\mathrm{Pi}})$ divided by the T₁ relaxation time $\left(T_1^{\prime }\right).V_{\mathrm{ATP}}=k_{\mathrm{f}}^{\prime }\left[{\mathrm{P}}_{\mathrm{i}}\right]$.

2.3. Monitoring metabolic fluxes using ¹³C MRS

As mentioned previously, ¹³C MRS offers a method to examine the metabolism of carbon-based compounds in vivo. The metabolic fate of a ¹³C-labeled substrate can be monitored in real time, providing that it is metabolized sufficiently rapidly to induce a significant, dynamic enrichment in one of its metabolic products, that this product can be resolved from other metabolites with ¹³C MRS and that it is of high enough concentration to be detected in vivo. For mitochondrial metabolism, the TCA cycle, which participates in the oxidation of lipids and carbohydrate, is ideally suited for examination by this technique.

Substrates derived from either source enter the cycle via the common intermediate, acetyl CoA, and are metabolized through a series of condensation, isomerization and oxidation (decarboxylation) reactions in the mitochondrial matrix ultimately producing carbon dioxide, NADH and FADH₂, the latter of which are oxidized by the electron transport chain. The ¹³C label can be introduced into the cycle as shown in Fig. 21.4 by infusing [2-¹³C]-labeled acetate as an MR-visible tracer, which is rapidly taken up by muscle and other tissues and converted into [2-¹³C]-acetyl-CoA that enters the TCA cycle by condensing with oxaloacetate to form [4-¹³C]-citrate. The position of the ¹³C label is conserved through the initial steps of the TCA cycle, labeling α-ketoglutarate at the C₄ position. The principal TCA-cycle intermediates (TCAI) are present in concentrations that are too low to be observed in vivo using ¹³C MRS. However, glutamate, which is present in millimolar concentrations in muscle, is in relatively rapid exchange with α-ketoglutarate and will also become ¹³C enriched at C₄. As the TCA cycle progresses, the ¹³C label becomes scrambled between the C₂ and C₃ positions due to the symmetry of the succinate molecule; a second turn of the TCA cycle yields [2-¹³C] and [3–¹³C]-glutamate.

Figure 21.4.

Oxidation via the TCA cycle: substrate derived from carbohydrate and lipid enters the cycle via the common intermediate acetyl CoA. Infused plasma [2-¹³C]-acetate is converted to acetyl CoA, and the ¹³C label (denoted by *) becomes incorporated into the muscle TCA-cycle intermediate and glutamate pools forming [4-¹³C] glutamate on the first turn of the TCA cycle. The ¹³C label becomes scrambled between the C₂ and C₃ positions due to the symmetry of the succinate molecule; a second turn of the cycle forms [2-¹³C] and [3-¹³C]-glutamate. Reaction rates incorporated into the metabolic model to calculate the TCA cycle flux (V_TCA) are shownand described in the text. Adapted and reprinted from Befroy et al. (2008a); © 1993–2008 by The National Academy of Sciences of the United States of America, all rights reserved.

Unlike brain, where natural abundance [4-¹³C]-glutamate can be resolved in an in vivo ¹³C MRS spectra, the C₄ glutamate peak at 34.3 ppm in muscle is obscured by overlapping lipid resonances as shown in Fig. 21.1. The contribution from sub-cutaneous fat can be eliminated by using a localized MRS sequence to selectively acquire signal from muscle only. The IMCL contribution is also significant, but can be minimized by an inversion-null strategy which takes advantage of the more rapid T₁ relaxation times of lipid moieties compared to other metabolites as shown in Fig. 21.5. At 4 T, we use an adiabatic ¹H–¹³C polarization-transfer (PT) sequence (Burum and Ernst, 1980; Shen et al., 1999) for ¹³C acquisition with WALTZ 16 decoupling. PT echo-times are optimized for C₄ glutamate detection. Spectra are acquired using a 9 cm diameter ¹³C surface coil positioned underneath the soleus-gastrocnemius muscle complex of the calf, with twin, orthogonal 13 cm diameter ¹H surface coils arrayed in quadrature for decoupling and excitation/inversion on the ¹H channel. Voxel localization is achieved using 2-dimensional outer volume suppression (OVS) to select a ~90 cm³ volume within the calf muscles. OVS enables the maximal ¹³C metabolite signal to be acquired, although an ISIS-based localization scheme could also be implemented. IMCL is suppressed using adiabatic T₁-selective nulling, as described above, optimized for the lipid resonance at 34.7 ppm. Previously, we have used direct ¹³C detection with decoupling at 2.1 T (Lebon et al., 2001); however, the acquired ¹³C signal is enhanced approximately 2-fold in vivo by using a PT sequence, and localization and lipid suppression can be performed on the ¹H channel where there is less error due to chemical shift dispersion.

Figure 21.5.

Natural abundance ¹³C spectra acquired from the soleus-gastrocnemius muscle complex demonstrating the use of lipid suppression techniques to minimize the contribution of subcutaneous fat and IMCL resonances to the spectrum.

To determine the TCA cycle flux, the time course of ¹³C incorporation into C₄-glutamate is monitored during a 120-min infusion of 99% enriched [2-¹³C]-acetate (350 mmol/l sodium salt) at a rate of 3.0 mg/kg/min. ¹³C spectra are acquired with ~5 min time resolution for 20 min prior to the infusion to establish a baseline, and continuously until the end of the experiment. Plasma samples are obtained at 10 min intervals throughout the experiment for the measurement of plasma acetate concentration and fractional enrichment by gas chromatography/mass spectrometry (Befroy et al., 2007). Steady-state enrichment of plasma acetate is attained within 5 min of originating the infusion and reaches ~85%; with the concentration of plasma acetate approximately 0.80 mM, consistent with functioning as a tracer-level substrate. To eliminate any residual contribution of lipid, the increment in the C₄-glutamate peak during the infusion is determined from difference spectra created by subtracting averaged baseline spectra from the ¹³C spectrum for each time-point. The maximal enrichment in the [4-¹³C]-glutamate peak is calculated relative to that of the C₂-glutamate peak assuming partial (5%) dilution of the C₂ pool due to anaplerosis. The absolute enrichment of the C₂-glutamate peak at the end of infusion is determined relative to the natural abundance enrichment (1.1%) of the baseline spectra.

2.4. Metabolic modeling of the TCA cycle flux

The TCA cycle flux (V_TCA) is determined by using CWave software to model the incorporation of ¹³C label from plasma [2-¹³C]-acetate into the muscle [4-¹³C]-glutamate pool. The CWave model consists of isotopic mass balance equations shown below that describe the metabolic fate of the plasma [2-¹³C]-acetate. The corresponding reaction rates are noted in Fig. 21.4. The ¹³C label enters the TCA cycle following the conversion of [2-¹³C]-acetate into [2-¹³C]-acetylCoA (V_AC). Entry of unlabeled substrates via acetyl CoA is incorporated into the model as a separate reaction (V_{PDH + FA}). CWave determines V_TCA as the rate of total carbon flow from acetyl CoA to α-ketoglutarate using a non-linear least-squares algorithm to fit the curve of C₄-glutamate enrichment. The rate of α-ketoglutarate/glutamate exchange has been determined in previous studies and is fixed at 150 nmol/g(muscle)/min which is significantly faster than the TCA cycle flux, and therefore non-limiting. In this model, V_TCA is equal to the sum of (V_AC + V_{PDH + FA}) but is independent of the absolute fractional enrichment at C₄-glutamate (V_AC/V_{PDH + FA}). The intracellular concentration of glutamate has been determined by enzymatic analysis of muscle biopsy samples from healthy individuals, and is assumed to be 2.41 mmol/l in all subject groups (Lebon et al.,2001). In our in vivo studies to date, the ratio of natural abundance C₂-glutamate/C₂-creatine is unaffected by the study population indicating that the intracellular concentration of glutamate is stable (Befroy et al., 2007, 2008a). The concentrations of the other TCAI are very low and therefore not critical to the accuracy of the model and have been taken from the literature (Hawkins and Mans, 1983). The curve of muscle [4-¹³C]-glutamate enrichment for a typical healthy individual is shown in Fig. 21.6 with the fit of the CWave metabolic model of these data superimposed, along with the calculated V_TCA.

Figure 21.6.

Time course of muscle [4-¹³C]-glutamate enrichment measured by ¹³C MRS in a typical healthy individual during a [2-¹³C]-acetate infusion. The fit of the data using the CWave metabolic model of the TCA cycle is superimposed along with the calculatedTCA cycle flux (V_TCA).

2.4.1. Isotopic mass balance equations

2.4.1.1. Mass balance

$$\begin{array}{cc}\hfill & \mathrm{d}\left(\text{Citrate}\right)∕\mathrm{d}t=V_{\mathrm{TCA}}-V_{\mathrm{TCA}}\hfill \\ \hfill & \mathrm{d}\left(\alpha \mathrm{KG}\right)∕\mathrm{d}t=V_{\mathrm{TCA}}+V_{\mathrm{X}}-(V_{\mathrm{TCA}}+V_{\mathrm{X}})\hfill \\ \hfill & \mathrm{d}\left(\text{Glutamate}\right)∕\mathrm{d}t=V_{\mathrm{X}}-V_{\mathrm{X}}\hfill \\ \hfill & \mathrm{d}\left(\mathrm{AcetylCoA}\right)∕\mathrm{d}t=V_{\mathrm{AC}}+V_{\mathrm{PDH}+\mathrm{FA}}-V_{\mathrm{TCA}}\hfill \end{array}$$

2.4.1.2. Isotope balance

$$\begin{array}{cc}\hfill \mathrm{d}\left({\mathrm{C}}_4\text{-Citrate}\right)∕\mathrm{d}t=& V_{\mathrm{TCA}}({\mathrm{C}}_2\text{-}\mathrm{AcetylCoA}∕\mathrm{AcetylCoA})-V_{\mathrm{TCA}}({\mathrm{C}}_4\text{-Citrate}∕\text{Citrate})\hfill \\ \hfill \mathrm{d}\left({\mathrm{C}}_4\text{-}\alpha \mathrm{KG}\right)∕\mathrm{d}t=& V_{\mathrm{TCA}}({\mathrm{C}}_4\text{-Citrate}∕\text{Citrate})+V_{\mathrm{X}}({\mathrm{C}}_4\text{-Glutamate}∕\text{Glutamate})-(V_{\mathrm{TCA}}+V_{\mathrm{X}})({\mathrm{C}}_4\text{-}\alpha \mathrm{KG}∕\alpha \mathrm{KG})\hfill \\ \hfill \mathrm{d}\left({\mathrm{C}}_4\text{-Glutamate}\right)∕\mathrm{d}t& =V_{\mathrm{X}}({\mathrm{C}}_4\text{-}\alpha \mathrm{KG}∕\alpha \mathrm{KG})-V_{\mathrm{X}}({\mathrm{C}}_4\text{-Glutamate}∕\text{Glutamate})\hfill \\ \hfill \mathrm{d}\left({\mathrm{C}}_2\text{-}\mathrm{AcetylCoA}\right)∕\mathrm{d}t=& V_{\mathrm{AC}}({\mathrm{C}}_2\text{-Acetate}∕\text{Acetate})+V_{\mathrm{PDH}+\mathrm{FA}}({\mathrm{C}}_0\text{-}\mathrm{FFA}∕\mathrm{FFA})-V_{\mathrm{TCA}}({\mathrm{C}}_2\text{-}\mathrm{AcetylCoA}∕\mathrm{AcetylCoA})\hfill \end{array}$$

2.4.1.3. Muscle metabolite concentrations

$$\begin{array}{c}\hfill \mathrm{AcetylCoA}=0.05\mathrm{umol}∕\mathrm{g}\hfill \\ \hfill \text{Citrate}=0.02\mathrm{umol}∕\mathrm{g}\hfill \\ \hfill \text{Glutamate}=2.41\mathrm{umol}∕\mathrm{g}\hfill \\ \hfill \alpha \mathrm{KG}=0.05\mathrm{umol}∕\mathrm{g}\hfill \end{array}$$

This relatively simple metabolic model, which incorporates only a single turn of the TCA cycle, and requires only plasma C₂-acetate and muscle C₄-glutamate enrichment, performs well in resting muscle. The advantage of the single-turn model is that the ¹³C MRS methods can be optimized to detect [4-¹³C]-glutamate with the maximum possible sensitivity and time-resolution and therefore obtain the most accurate time course of enrichment. This scheme generates the same rates of V_TCA as a more robust 2-turn model that also requires data for muscle C₂-glutamate which is enriched during a second turn of the TCA cycle. We have previously used this more complex model to estimate V_X which confirmed our assumption that V_X > V_TCA (Befroy et al., 2007; Lebon et al., 2001). We also investigated the influence of anaplerosis on the calculated rate of TCA cycle flux by using a model that incorporated an anaplerotic flux via pyruvate carboxylase (V_ana). The calculated V_TCA is unaffected by incremental increases in V_ana up to 20% of V_TCA (Befroy et al., 2007), which is significantly greater than is likely in resting muscle. We have also found that the absolute fractional enrichment of C₂-glutamate at the end of the [2-¹³C]-acetate infusion is not different between subject groups in our studies, indicating that rates of anaplerosis in quiescent muscle are similar between study populations (Befroy et al., 2007, 2008a). However, this condition may not be preserved for all subject cohorts or in other tissues.

3. Insulin Resistance and Resting Mitochondrial Metabolism

Using the techniques outlined above, we have examined how these parameters of resting mitochondrial metabolism are affected in two different cohorts of insulin-resistant individuals. Both V_TCA and V_ATP were decreased by ~40% in the muscle of insulin-resistant elderly subjects (mean age ~70 years) compared with activity-matched young adults (mean age ~27 years) (Petersen et al., 2003), demonstrating that these older individuals had decreased resting mitochondrial function. Similarly, both parameters were decreased ~30% in insulin-resistant offspring of type 2 diabetic patients compared with insulin-sensitive controls (Befroy et al., 2007; Petersen et al., 2004), consistent with the ~40% reduced mitochondrial density observed in a similar cohort of subjects (Morino et al., 2005). Average curves of muscle [4-¹³C]-glutamate enrichment for each group are shown in Fig. 21.7, with the fits of the CWave model superimposed, demonstrating the difference in the rates of enrichment and the resultant TCA cycle flux between these cohorts. It is worth noting that these studies were performed in lean, otherwise healthy individuals and the subject groups were matched for body composition, to prevent the confounding factor of obesity, and to try to isolate effects occurring early in the pathogenesis of insulin resistance. Activity levels were also matched, since many of the age-related effects on muscle metabolism can be attributed to the decrease in physical activity that typically occurs in older individuals (Chilibeck et al., 1998; Kent-Braun and Ng, 2000). These spectroscopy techniques are also independent of the mass of tissue from which the MRS signal is detected and are therefore not directly affected by the muscle atrophy that is commonly associated with aging (Frontera et al., 1991).

Figure 21.7.

Muscle [4-¹³C]-glutamate enrichment (averaged data) measured by ¹³C MRS in a group of healthy, insulin-sensitive control subjects and insulin-resistant (IR) offspring of type 2 diabetic patients. CWave fits of the data are superimposed demonstrating the difference in the rates of enrichment and the resultant TCA cycle flux between the groups. Adapted and reprinted from Befroy et al. (2007); © American Diabetes Association (ADA).

We have also investigated how resting mitochondrial metabolism is modulated in endurance trained muscle, which is characterized by increased mitochondrial number and content of oxidative enzymes. Although V_TCA was increased in the muscle of trained individuals, V_ATP was unaltered (Befroy et al., 2008a) suggesting a decrease in the efficiency of energy production at rest. This elevated “idling” rate of the mitochondria may enable a more rapid response to the increased energy demand at the onset of exercise. We did not observe any significant effect on whole-body energy expenditure demonstrating that whole-body indirect calorimetry techniques are too insensitive to detect significant changes in muscle oxidation rates.

Comparison of the absolute rates of V_ATP, calculated by the ³¹P-saturation-transfer technique, and of oxygen consumption, estimated by a number of techniques (including V_TCA by ¹³C MRS), give rise to P:O ratios in resting muscle which are higher than the theoretical maximum (Brindle et al., 1989; Kemp, 2008). This is in contrast to other tissues which have higher respiration rates, for example, kidney (Freeman et al., 1983) and brain (Du et al., 2008; Lei et al., 2003), where good agreement is observed between these parameters. The total P_i → ATP flux from all cellular contributions (oxidative + glycolytic) is evaluated using the saturation-transfer technique; although the glycolytic contribution in resting muscle is low (Jucker et al., 2000); and unpublished data) it has been postulated that in quiescent muscle reversible P_i → ATP exchange via the mitochondrial F₁F₀ ATPase may contribute to the elevated absolute V_ATP (Brindle and Radda, 1987; Campbell-Burk et al., 1987). It is worth noting that even if exchange is the dominating contribution, it is likely to be proportional to the F₁ ATPase activity. We have successfully miniaturized the saturation-transfer technique to study mouse muscle and have found that the P_i → ATP flux responds as expected to perturbations in the expression of mitochondrial genes in transgenic mice including UCP3 knock-out (Cline et al., 2001) and UCP3 overexpression (Befroy unpublished data) and PGC1α overexpression (Choi et al., 2008) confirming that the rate of ATP synthesis determined by the saturation-transfer technique is an effective biomarker of mitochondrial function.

3.1. Resting mitochondrial metabolism versus maximal capacity

In counterpoint to our in vivo MRS studies investigating resting muscle mitochondrial metabolism, other groups have observed that mitochondrial oxidative capacity, estimated in vivo from PCr kinetics during exercise, was unaffected by aging (Chilibeck et al., 1998; Lanza et al., 2005) when activity patterns were accounted for, and was unchanged throughout the development of insulin-resistance and diabetes in the ZDF rat (De Feyter et al., 2008). Muscle biopsy studies also suggest that mitochondrial capacity is increased in Asian-Indians despite severe insulin resistance (Nair et al., 2008). The disparity between these studies highlights the fact that capacity and resting function are different parameters of mitochondrial metabolism, likely to be regulated or limited by different factors, and that the term mitochondrial “dysfunction,” which has become commonly used in the field and the source of some controversy, is misleading since it has no precise definition and could refer to reduced capacity, function or a combination of both. Furthermore, the findings of these studies will also have been influenced by the precise characteristics of the populations studied and the muscle chosen for analysis. Different muscle groups exhibit distinct metabolic and functional phenotypes (Essen et al., 1975; Pette and Spamer, 1986) and respond differently to aging (Houmard et al., 1998) and potentially disease. Therefore, it is highly likely that there are muscle-group specific differences in mitochondrial function and capacity in healthy subjects and in the modulation of energy production by aging/disease. With this issue in mind, we have recently implemented the saturation-transfer methodology in a localized manner to obtain muscle-group specific rates of P_i → ATP flux (Befroy et al., 2008b).

4. Conclusions

Because of its distinct characteristics, MRS is a unique tool for investigating metabolism across a wide range of systems, encompassing human, animal and cellular models, a wide array of tissues, and in vivo as well as in vitro applications. Multinuclear capabilities further extend these possibilities and since the technique is non-invasive, dynamic processes can be studied using repeated measurements in addition to static estimates of metabolite concentrations. By taking advantage of these unique properties, we have developed techniques that allow us to monitor muscle mitochondrial metabolism in vivo and investigate how it is modulated in health and disease. The implementation of spectroscopy packages on clinical MR scanners will allow these methodologies to reach a wider audience of investigators. With the progression to higher-field MR systems, which increase spectral resolution and sensitivity, and the continued development of enhanced detection techniques, there are a multitude of possibilities for future research.