Cell Physiology

31P-NMR observation of free ADP during fatiguing, repetitive contractions of murine skeletal muscle lacking AK1

Chad R. Hancock, Jeffrey J. Brault, Robert W. Wiseman, Ronald L. Terjung, Ronald A. Meyer


Metabolic control within skeletal muscle is designed to limit ADP accumulation even during conditions where ATP demand is out of balance with ATP synthesis. This is accomplished by the reactions of adenylate kinase (AK; ADP+ADP ↔ AMP+ATP) and AMP deaminase (AMP+H2O → NH3+IMP), which limit ADP accumulation under these conditions. The purpose of this study was to determine whether AK deficiency (AK−/−) would result in sufficient ADP accumulation to be visible using 31P-NMRS during the high energy demands of frequent in situ tetanic contractions. To do this we examined the high-energy phosphates of the gastrocnemius muscle in the knockout mouse with AK1−/− and wild-type (WT) control muscle over the course of 64 rapid (2/s) isometric tetanic contractions. Near-complete depletion of phosphocreatine was apparent after 16 contractions in both groups. By ∼40 contractions, ADP was clearly visible in AK1−/− muscle. This transient concentration of the NMR visible free ADP was estimated to be ∼1.7 mM, and represents the first time free ADP has been directly measured in contracting skeletal muscle. Such an increase in free ADP is severalfold greater than previously thought to occur. This large accumulation of free ADP also represents a significant reduction in energy available from ATP, and has implications on cellular processes that depend on a high yield of energy from ATP such as calcium sequestration. Remarkably, the AK1−/− and WT muscles exhibited similar fatigue profiles. Our findings suggest that skeletal muscle is surprisingly tolerant to a large increase in ADP and by extension, a decline in energy from ATP.

  • muscle energetics
  • muscle relaxation
  • magnetic resonance spectroscopy

in normal muscle during exercise the demand for ATP is adequately balanced by ATP supply from oxidative phosphorylation and glycolysis, coupled with the ATP buffering capacity of the creatine kinase reaction, such that ATP is not significantly depleted. Even when ATP demand does exceed ATP supply, and a decline in ATP is observed, the decline in ATP is not matched by a stoichiometric accumulation of ADP because of the action of adenylate kinase (AK), which transfers a phosphate from one ADP to another ADP resulting in ATP and AMP formation. In skeletal muscle the AMP formed from this reaction is rapidly deaminated by AMP deaminase (AMPD) resulting in the formation of inosine 5′-monophosphate (IMP). Thus, when the rate of ATP hydrolysis exceeds the ATP supply capacity, there is a stoichiometric accumulation of IMP, due to the coupled reactions of AK and AMPD. Therefore, at high energy demands, the reactions of AK and AMPD both serve to limit ADP accumulation.

The importance of limiting ADP accumulation in skeletal muscle is implied by the apparent lack of such a limit on another product of ATP hydrolysis, inorganic phosphate (Pi). The accumulation of Pi during high-energy demands is close to being proportional to the decline in phosphocreatine (PCr), and can be as great as 40 mM (24). In contrast, ADP accumulation in healthy muscle has been estimated to rise to no greater than ∼300 μM even during intense, fatiguing contraction regimens (4). The physiological consequences of ADP accumulation are related to a direct effect of ADP on the kinetics of various enzymes. For example, millimolar concentrations of ADP inhibit the peak rate of myosin cross-bridge cycling (6), and may cause calcium leakage from the sarcoplasmic reticulum (SR) (17). In addition, ADP accumulation will cause a decline in the free energy available from the hydrolysis of ATP (ΔGATP) because ΔGATP is a function of the ratio of ADP and Pi to ATP as defined below. Math A large decline in the ΔGATP could impact enzymatic reactions that function near the thermodynamic limit of energy available from ATP hydrolysis. One candidate is calcium sequestration by the SR. The SR Ca2+-ATPase (SERCA) is thought to pump calcium against a gradient of ∼1:10,000, at a high energetic cost (5, 10). Thus a decline in ΔGATP is expected to impair the function of SERCA and the maintenance of this gradient, and consequently impair muscle relaxation because of the high cost of maintaining the large gradient of calcium.

Direct quantification of the ΔGATP is impossible because the ADP present in muscle is largely bound to proteins and not metabolically active or free to participate in enzymatic reactions (18). The myofibrillar proteins actin and myosin both bind ADP (2) and account for most of the ADP measured in acid extracts of muscle tissue, making a direct measurement of free ADP (20–300 μM) from the total, chemically measured ADP (>600 μM) unreliable. Thus the free ADP in skeletal muscle is typically calculated from the stoichiometry and assumed equilibrium of the creatine kinase reaction (16). In principle, free ADP can be directly measured by high-resolution in vivo 31P-NMR spectroscopy (31P-NMRS; e.g., Ref. 7) because the ADP bound to macromolecules is substantially line broadened and not observed in high-resolution spectra. However, phosphorus NMR is a relatively insensitive technique, and the small ADP accumulation expected in normal skeletal muscle has not been directly observed. Therefore, the limit of free ADP accumulation normally observed in skeletal muscle has prevented any direct assessment of free ADP.

A transgenic knockout model of AK deficiency (AK1−/−) has been recently developed (12), and provides a tool to investigate the metabolic and functional consequences of AK deficiency in skeletal muscle. Because AK is an important part of the metabolic circuit involved in the management of ADP, consequences of an inordinate accumulation of ADP would be expected if energy demands could be sufficiently elevated. The basal metabolic phenotype of AK1−/− gastrocnemius muscle was reported to be similar to the wild-type (WT) controls, although an increase in mitochondrial markers and mitochondrial volume has been reported in a small superficial portion of the gastrocnemius muscle (9, 13). The tetanic contractile performance of AK−/− muscle was also examined at extreme energy demands, and whereas the overall pattern of muscle fatigue was similar to WT controls, there was a marked slowing of muscle relaxation in AK−/− muscles (9). Remarkably, AK1−/− muscle exhibited an accumulation of up to 1.5 mM ADP (as measured in perchloric acid extracts) during energetically demanding, repetitive tetanic contractions, whereas the WT control muscles exhibited no such accumulation of ADP (9). An accumulation of free ADP of this magnitude would be expected to severely impair muscle relaxation (7), and slow cross-bridge cycling (6, 34). The main purpose of this study was to determine whether the increase in ADP observed in contracting AK1−/− muscle is in fact free, representing an energetic challenge to muscle larger than previously considered possible in intact muscle. This was accomplished using 31P-NMR because only the free ADP is discernible using this technique. A secondary objective was to determine whether the reported elevations in mitochondrial content in AK1−/− would result in an enhanced rate of PCr recovery after demanding contractions. The results indicate that free ADP accumulation observed by 31P-NMRS is transiently as high as 1.7 mM in contracting AK1−/− gastrocnemius muscle, consistent with the amount of ADP accumulation estimated from the chemical measurements reported previously (9). In addition, the results show that the PCr recovery rate constant is significantly faster in muscle of AK1−/− compared with WT control mice, consistent with increased mitochondrial capacity.


Animal care.

Adult AK1−/− mice (12) and C57Bl6 control mice were housed up to five per cage, fed ad libitum, and kept on a 12:12-h light:dark cycle. Seventeen AK1−/− and sixteen WT animals were used for the experiments in this study. Animals were anesthetized with pentobarbital sodium (∼70 mg/kg ip) injection. All mouse care and experimental protocols were approved by the All University Committee on Animal Use and Care at Michigan State University.

In situ contractions.

Muscle contractions of the mouse hindlimb muscles were elicited via direct electrical stimulation of the sciatic nerve, accessed by an incision on the lateral portion of the right leg, near the hip. Electrodes were placed in contact with the nerve and shielded from surrounding tissue to prevent extraneous stimulation. The sciatic nerve was crushed by tying a knot above the point of contact with the electrodes to eliminate antidromic propagation. The plastic shielding of the electrode wires was secured to the mouse limb with adhesive to prevent any movement of the electrodes. Stimulation voltage was determined to be supramaximal by test stimuli before the experiment. Tetanic contractions were elicited via 100-ms trains of 0.05-ms pulses at a frequency of 150 Hz, applied at a rate of 2 tetani/s for 32 s. This is the same frequency and duration of contraction that previously elicited large changes in total chemically measured ADP (9).

The mouse was secured in the head-down position, such that the lower right limb could be positioned adjacent to the NMR coil in a custom-built probe. Movement of the lower limb was minimized by tying the patellar tendon to Plexiglas supports. The achilles tendon was tied to a wheatstone load cell for recording muscle tension. The muscle was then stretched by adjusting the position of the load cell to a length that yielded maximum tetanic tension.

31P-NMR spectroscopy.

31P-NMR spectra [162 MHz, 10,000 Hz sweep width, 1,024 complex points, repetition time (TR), 1 s] were acquired on a Bruker AM400 spectrometer via a 0.5 × 0.8-cm saddle-shaped surface coil positioned over the gastrocnemius muscle. Spectra acquired at rest and after stimulation were the average of 128 and 32 scans, respectively. To minimize motion artifacts during the 2 tetani/s stimulation, acquisition of single scans was triggered by the stimulator to occur 200 ms after the start of every other tetanus (i.e., TR 1 s). These scans were retrospectively added in blocks of eight, yielding a time resolution of 8 s (16 contractions) during the stimulation. All spectra were filtered by a 25-Hz exponential and zero filled to 2,048 points before Fourier transformation and manual phase correction. Finally, spectra from 8–9 animals at corresponding time intervals during the stimulation were added together. Four complete series of combined spectra during stimulation with 8–9 animals each were thus obtained. All spectra are presented with a common (−2.52 ppm) shift for the PCr peak. The time course of PCr recovery after stimulation was determined in individual animals. For this analysis, PCr concentrations were estimated using the method of natural line shapes (11). The time constant for PCr recovery (τ) was estimated from a single exponential fit (25) immediately after 64 contractions at 120 tetani/min. Intracellular pH was estimated from the chemical shift of Pi, as done previously (23). The concentrations of ATP, ADP, IMP, and PCr were calculated using the known content of ATP measured from chemical extracts in the gastrocnemius muscle of these mice [5.84 μmol/g wet wt (9)]. The peak integral of the β-ATP peak at rest, and the sum of the peak integrals of the β-ATP, IMP, and β-ADP peaks during contractions, were considered quantitatively equivalent to the chemically measured concentration of ATP (5.84 μmol/g wet wt), assuming that no substantial loss of nucleotides occurred during contractions (21). Thus metabolite concentrations were estimated on the basis of relative peak integrals.

Analysis of the amount and location of 31P-NMRS of ADP in muscle may potentially be confounded by the changes in the chemical shift of the ADP peak with pH, and by differences in T1, and hence in the relative signal saturation that occurs as a result of the short TR interval used in this study (1 s). Therefore, a series of 31P-NMR spectra (242 MHz, sweep width 48,000 Hz, 1,024 complex points, TR 0.2–8 s, 1,024 scans/spectrum) were acquired on a Varian Unity 600 MHz spectrometer from solutions maintained at 37°C containing the following constituents (in mM): 8 ATP, 0.5 ADP, 1 Pi, 25 PCr, and 1 Mg, at an ionic strength of 0.25 M; ionic strength was balanced using acetate as the anion and Tris as the cation. The change in pH was buffered by (in mM) 50 MES, pH 6 and 6.5, 50 MOPS, pH 7, and 50 TES, pH 7.5 and 8. The apparent T1 of each peak was estimated from the exponential rise in peak intensity with increased TR. In solution, and at the higher magnetic field strength, we were able to maximize the signal-to-noise ratio to examine both pH and the pattern of relative saturation of ATP and ADP.

Statistical analysis.

Student's t-test was used to determine significant differences in performance at 60 contractions and the time constants for PCr recovery between AK1−/− and WT control groups. A P value of <0.05 was considered significant.


In vitro 31P-NMR of ADP.

Figure 1 shows spectra (Fig. 1A) and chemical shifts (Fig. 1B) of the ATP- and ADP-containing solutions from pH 6 to 8. These data show that the chemical shift difference between the γ-ATP and β-ADP resonances under conditions mimicking the ionic strength and free magnesium in muscle cytoplasm is near 1 ppm at pH 7, and increases as pH declines. On the other hand, the chemical shift difference between α-ATP and α-ADP peaks is only 0.3 ppm at pH 7, and this difference decreases as pH is lowered. Thus, given the linewidth of the ATP peaks typically achieved in 31P-NMR studies of mouse skeletal muscle [∼1 ppm (29)], one might expect to resolve the β-ADP, but not the α-ADP peak in in vivo muscle spectra.

Fig. 1.

Solution analysis of the chemical shift with changes in pH and progressive signal saturation. A: stack plot of ATP, ADP, phosphocreatine (PCr), and inorganic phosphate (Pi) over the pH range of 6–8. Inset: portion of the spectra at pH 6.5 showing both peaks of the ADP phosphates. A reference peak from a separate capillary containing 85% H3PO4 is also included. B: plots of the chemical shift of the phosphate groups of ATP, ADP, and Pi with changes in pH.

Figure 2 illustrates the effect of variations in TR on the relative amplitudes of the ADP and ATP peaks. As expected, the fitted apparent T1s are somewhat longer for the terminal phosphates compared with the α-phosphates in both ADP and ATP, and to the β-phosphate of ATP. The T1 of the β-ATP (1.7 s) is somewhat longer than that of β-ADP (0.87 s). However, at TR of 1 s, the effect of this difference is relatively minor, leading to at most a 24% underestimation of ADP compared with ATP, assuming that the relaxation times are similar in vivo as in these solutions (see discussion).

Fig. 2.

The T1 recovery plots of the phosphate groups of ATP and ADP.

In vivo 31P-NMRS.

There were no significant differences between relative peak areas and intracellular pH (6.98 ± 0.06 and 6.97 ± 0.04 in AK1−/− and WT, respectively) in spectra of AK1−/− compared with WT muscle before stimulation (Fig. 3), which is consistent with previous chemical measures of metabolites in gastrocnemius muscles of the same mouse lines (9). Spectra acquired during contractions in the four combined groups are presented in Fig. 4, A and B. Spectra from both WT and AK1−/− groups were initially divided into groups of 8–9 animals each to verify the reproducibility of the findings. From Fig. 4, the following points are apparent: 1) the fall in PCr occurs similarly in the AK1−/− muscle and in the WT groups, and is almost completely depleted by contractions 17–32; 2) less phosphomonester (presumptive IMP) accumulation is apparent in the AK1−/− muscle compared with the WT group; 3) the chemical shift in Pi with contractions indicates that the decline in pH was similar [minimum 6.50 ± 0.03 and 6.48 ± 0.04 (x ± SE, n = 16) in AK1−/− and WT, respectively, based on analysis of the first postcontractile spectra in individual animals]; and 4) the β-ADP peak is clearly resolved 0.98 ppm from the γ-ATP peak in both AK1−/− groups in the spectra of contractions 33-48. There is no corresponding peak in the spectra of WT muscles at any time point. The spectra in Fig. 5A are the combined spectra for all AK1−/− and WT muscles (i.e., Fig. 4, A and B, combined). Figure 5B shows the spectra from contractions 33-48 in AK−/− and WT, along with the difference spectrum between spectra over this time block. In addition to the β-ADP peak, the difference spectrum reveals a corresponding peak at the chemical shift of α-ADP, which is not resolved from α-ATP peak in the unsubtracted spectra.

Fig. 3.

Representative 31P-NMR spectroscopy (31P-NMRS) spectra of unstimulated gastrocnemius muscle from both adenylate kinase-deficient (AK1−/−) and wild-type (WT) animals [128 scans each, repetition time (TR) = 1 s].

Fig. 4.

31P-NMRS spectra of contracting WT and AK1−/− hindlimb muscle. Each spectrum is the summed average of 8 animals over an 8-s (16 contractions) window. Two different series (A and B) of eight animals are shown for both WT and AK1−/− animals. The β-ADP peaks are clearly discernible in the AK1−/− muscle spectra after the first 16 contractions with the largest β-ADP peak observed in the spectra from contractions 33-48. The spectra in this figure are displayed with subsequent contraction sequences having a slight rightward shift.

Fig. 5.

A: summed spectra of 16 animals for WT and AK1−/− animals. B: difference between the spectra from AK1−/− and WT of contractions 33-48 is illustrated and highlights the two ADP peaks present in the AK1−/− muscle.

Ignoring the minor effects due to T1 relaxation differences, the magnitude of the free ADP accumulation in Figure 5B can be estimated by assuming that any decline in ATP must be accounted for by an increase in IMP (19) and/or ADP (free AMP is quantitatively insignificant). Therefore, if the sum of integrated area of IMP, β-ATP, and β-ADP peak is assumed to stay constant, the relative area of ADP (17.3%) times the ATP content of muscle before stimulation yields a quantitative estimate of free ADP present in the intact muscle. The ATP content of the mixed gastrocnemius of the AK1−/− and WT mice was measured from perchloric acid extracts previously [5.84 ± 0.14 and 5.95 ± 0.18 μmol/g wet wt, respectively (9)]. Thus we estimate free ADP in Fig. 5B to be 1.01 μmol/g (∼1.7 mM assuming 60% intracellular water). This is consistent with the increase in total ADP measured chemically at the same time period in the previous study (1.5 mM; see Ref. 9), indicating that the increase in ADP is almost entirely NMR-observable, metabolically active free ADP.

Figure 6 shows the time course of PCr recovery in muscles after the stimulation in AK−/− and WT mice. The fitted time constant (τ) for PCr was significantly faster in AK−/− (1.9 ± 0.2 min) compared with WT (2.7 ± 0.2 min) (P < 0.05). The recovery rate of PCr is an index of the oxidative capacity of muscle (25); therefore, this result is consistent with previous reports of increased intermyofibrillar mitochondria (12, 13) and increased citrate synthase activity (9) in the superficial gastrocnemius of AK−/− mice.

Fig. 6.

The recovery of PCr after contractions in AK1−/− (○) and WT (▪) control gastrocnemius. Plots are means ± SE.

Contractile performance.

The overall pattern of fatigue was not different between groups (Fig. 7) and was similar to previous reports of tetanic contractile performance at this intensity (9, 12).

Fig. 7.

The pattern of fatigue in AK1−/− (○) (n = 7) and WT (▪) (n = 6) control gastrocnemius contracting at 120 tetanic contractions/min. Plot is the mean ± SE of each contraction.


The most novel and striking finding of this study is that free ADP accumulation in intact skeletal muscle of AK1−/− mice can be driven to ∼1.5 mM. To our knowledge, this is the first condition in which ADP has been observed by 31P-NMR in intact skeletal muscle and represents a decline in the ΔGATP previously deemed inconsistent with competent SERCA function and complete muscle relaxation (7). In normal skeletal muscle the accumulation of free ADP is limited by several processes involved in ATP synthesis (glycolysis and oxidative phosphorylation) and ADP buffering (CK). Even when energy demands exceed the capacity for ATP synthesis such that a decline in ATP is observed, ADP removal is facilitated by the coupled reactions of AK and AMPD and the accumulation of IMP accounts for nearly the entire decline in ATP (20). The objective of this study was to determine whether the increase in ADP measured from muscle extracts under similar conditions in these mice was metabolically active free ADP. The observation of free ADP by 31P-NMRS confirms that the increase in total ADP during intense contractions measured chemically in AK1−/− muscle (9) is indeed free ADP.

Accurate in vivo quantitation of ADP by 31P-NMRS is difficult for the following reasons: 1) the mouse muscle is very small and the signal available from 31P is not very sensitive, 2) the ADP accumulation is transient, so the signal-to-noise ratio cannot be increased by averaging the signal from an individual experiment over long periods of time, and 3) the longitudinal relaxation constant, or T1, of ADP is unknown in the intracellular environment and impossible to directly determine, given the transient nature of ADP accumulation. To overcome the problem of the low signal-to-noise ratio and transient nature of ADP accumulation, we added spectra from multiple animals at the same time point. Because the calculations of ADP content are dependent on the relative peak integrals of β-ATP and β-ADP, and these peak integrals could be influenced by the relative T1 relaxation rates, in vitro experiments were performed using a series of different TRs. This revealed that the peak difference between β-ATP and β-ADP was 24% at 1 s (cf. Fig. 2). In other words, a comparison of the two peaks would underestimate ADP by 24%. However, the difference in vivo is likely less or even completely eliminated for two reasons. First, the T1 relaxation constants in vitro were measured at a higher field strength (600 MHz) than the experiments in vivo (400 MHz). Relatively shorter T1 relaxation constants are expected at lower field strengths (14). Second, relaxation rates in vivo are expected to be shorter than in vitro because of interactions with intracellular proteins. Therefore, in the present experiments, the peak integrals were not adjusted for apparent differences in T1 relaxation rates. Nevertheless, the calculated free ADP content using 31P-NMRS (1.7 mM) is in agreement with the ADP accumulation measured previously from muscle extracts (1.5 ± 0.1 mM; Ref. 9) and supports the accuracy of our calculations.

Free ADP.

The free ADP in tissue has normally been calculated from the known stoichiometry and assumed equilibrium of the CK reaction (16). The ADP measured from muscle extracts is mostly metabolically inactive because it is bound to intracellular proteins such as actin and is not free to participate in biochemical reactions (33). The difficulty in measuring free ADP directly with 31P-NMRS is not due to the presence of bound ADP, because the 31P-NMRS signal from bound ADP would be very broad and indistinguishable from background noise (22). The primary problems with observing free ADP with 31P-NMRS are that it never accumulates to levels necessary to be distinguishable from background noise (0.5–1 mM) as well as the lack of resolution between the chemical shift from ATP and ADP (Fig. 1). Direct measurement of free ADP by 31P-NMRS has been performed in a hypoxic coronary artery model using 31P-NMRS; however, this requires scans to be averaged over a 30-min time period to maximize the signal-to-noise ratio (8). Free-ADP accumulation in skeletal muscle has been suggested to rise to ∼0.3 mM (4) under high-energy demands, but this accumulation is transient due to the effectiveness of ATP synthesis pathways in skeletal muscle and to the rapid decline in energy demands due to fatigue. Thus this represents the first report of directly resolved free ADP by 31P-NMRS in the intact contracting skeletal muscle.

ADP and free energy of ATP.

The degree of ADP accumulation we report here represents, at the very least, a significant transient energetic challenge in the AK1−/− muscle compared with the WT control muscle that is apparent from the calculated ΔGATP [−46 and −53 kJ/mol, respectively (9)]. A decline in the energy available from ATP would be expected to affect ATPases that depend on a large portion of the total energy available from ATP. The management of calcium in skeletal muscle is an energetically expensive process due to the high concentration of calcium inside the SR (∼1 mM) and the relatively low concentration in the cytosol (∼50–100 nM at rest). The energy required to support this calcium gradient is defined as −2RT ln ([Ca2+]sr/[Ca2+]cyt), where [Ca2+]sr is the SR-free [Ca2+] and [Ca2+]cyt is the cytosolic [Ca2+], and at resting concentrations it is approximately −51 to −48 kJ/mol. At resting adenine nucleotide, and Pi concentrations, the ΔGATP is approximately −65 kJ/mol in the AK1−/− and WT muscle. Thus, if there were no limit on SR calcium loading, and the cytosolic calcium were maintained at resting levels, the resting ΔGATP of −65 kJ/mol would be sufficient to support SR free calcium concentration of ∼15 mM (Fig. 8), which is ∼15 times what the physiological SR free-calcium load is estimated to be (5, 10, 26, 30). Thus at rest there is an excess of energy available to maintain the expected calcium gradient between the cytosol and the SR. However, if the minimum cost to maintain the resting calcium gradient is −48 kJ/mol, the ΔGATP of −46 kJ/mol measured in AK1−/− muscle suggests that there is insufficient energy available to support this gradient. Research on the consequences of a depressed energy state on muscle calcium management has suggested that if the ΔGATP dropped below this limit, the failure of normal calcium management would lead to a lack of muscle relaxation (7), or in the case of cardiac muscle, impaired contractile reserve (31, 32). Remarkably, we (9) observed similar tetanic force production in the AK1−/− muscles compared with WT muscle, albeit with delayed but nearly complete muscle relaxation. We believe that there are a few possible reasons that full muscle relaxation can occur when the energy available is presumed insufficient to maintain the resting calcium gradient. First, a ΔG of −43 kJ/mol is required to maintain a ratio of ∼1:4,000 between the cytosol and the SR, which works out to be 250 nM [Ca2+]cyt to 1 mM [Ca2+]sr (Fig. 8). An increase in the [Ca2+]cyt of this magnitude may not result in any significant force production and thus allow full muscle relaxation (35). Furthermore, if a decline in the SR calcium occurred, the energy available would support lower cytosolic calcium concentrations. A decline in the free calcium content of the SR may occur by calcium phosphate precipitation during fatigue; however, this process likely requires ∼1–2 min to occur, and therefore cannot explain our findings (1). Another possible mechanism to reduce cytosolic calcium and allow full relaxation is an increased calcium uptake by other organelles in the cell, such as the mitochondria, to effect a lower cytosolic-to-[Ca2+]sr ratio (3, 15). However, due to the relatively small amount of calcium taken up by the mitochondria, this does not appear to be a very reasonable explanation (15).

Fig. 8.

The energy cost (ΔGCa2+) of maintaining the gradient across the sarcoplasmic reticulum (SR). The x-axis is the SR free calcium content and the ordinate is the energy cost in kJ/mol. Three different cytosolic concentrations are plotted, illustrating the cost of maintaining a given cytosolic [Ca2+] over a range of SR calcium loads. A line is drawn at 1 mM SR calcium to represent the common estimate of the SR intracellular calcium content (5, 10, 26, 30). Lines are drawn to indicate the energy cost to maintain the specified [Ca] against 1 mM SR [Ca] (−43, −48, −53, and −65 kJ/mol), as mentioned in the discussion.

The estimations mentioned above presume a 100% efficient use of the energy from ATP for the maintenance of the calcium gradient, and a constant tight coupling of 2 Ca2+-ATP. If the pump does not function at 100% efficiency, then more energy from ATP must be required for SERCA function. If this were true, another explanation would be required for the full muscle relaxation observed in AK1−/− (in light of the fact that the ΔGATP in AK1−/− was determined to be insufficient for the maintenance of the normal resting calcium gradient). One possibility is increased calcium leak from the SR, effectively reducing the Ca2+-ATP coupling. Recent work by Macdonald and Stephenson (17) suggests that ADP accumulation on the order of 1 mM would likely result in significant calcium leak through the SERCA pump. This would likely decrease the ratio of 2 Ca2+-ATP that is commonly assumed for the SERCA pump when ADP levels rise. If the coupling were reduced by half, such that SERCA functioned at a ratio of only 1 Ca2+-ATP, then the energy required from each ATP molecule to maintain the resting gradient would be half of what it is with a coupling of 2 Ca2+-ATP. However, if this altered coupling occurred, it would also result in twice the normal ATP consumption by SERCA. The ATP consumption by SERCA for the maintenance of calcium accounts for roughly 30% of the total energy expenditure during contractions, and the vast majority of the rest of the ATP consumption is due to the contractile ATPase (27, 28). An altered Ca2+-ATP coupling would require an increase in the proportion of ATP consumption by SERCA to maintain a functional calcium gradient. Clearly, the simplest explanation for sustained muscle function is an elevation in cytosolic calcium concentration below significant force-generating concentrations.

It is worth mentioning that the increase in ADP, and hence the energetic challenge discussed above is likely very transient in nature. As contractions proceed, the force of contractions fall until a force is achieved that can be supported by the rate of ATP synthesis, and ADP accumulation would likely fall. The transient nature of this large ADP accumulation is even evident from an apparent reduction in the signal from β-ADP observed from contractions 34-48 to contractions 49-64 in Fig. 5A. Thus, whereas the energetic challenge of calcium sequestration illustrated in Fig. 8 should still be valid even for the transiently high ADP, we do not know what the consequences of a more sustained ADP accumulation would be. It is also interesting to consider that the direct measure of free ADP observed in this study is an average of the free ADP present in the whole gastrocnemius muscle within the field of the NMR coil, and not necessarily representative of the highest free-ADP accumulation in select muscle fibers. Thus the energetic impairment in certain sections of the muscle due to an inordinate ADP accumulation may in fact be more dramatic than what we discuss here.

Enhanced mitochondrial capacity.

In addition to the findings regarding free ADP, the faster rate of PCr recovery in AK−/− muscles after the extreme energy demands used in this study indicate that AK−/− resulted in an adaptive increase in mitochondrial capacity. This is in line with reports of increased citrate synthase activity (9) and mitochondrial volume by EM-morphometric analysis (13). While the increase in oxidative capacity was not evident in significantly improved force production with fatigue in this study (Fig. 7) or in others (9, 12), enhanced fatigue resistance may be more clear in much less demanding conditions.

In summary, we have demonstrated by 31P-NMRS that the chemically measured ADP accumulation of ∼1.5 mM in AK1−/− muscle represents free and metabolically active ADP. This is the first direct observation of free ADP in contracting skeletal muscle and is severalfold greater than the normal limit of ADP accumulation. Finally, we estimate that the energetic consequence of this large ADP accumulation on calcium uptake may be that cytosolic calcium is elevated but still maintained below force-generating levels, thus permitting full muscle relaxation.


This work was supported by National Institutes of Health Grants AR-21617 and AR -043903, National Biomedical Research Institute Grant MA-00210, and Michigan State University Grant IRPG 41006.


We thank Dr. Bé Wieringa and Dr. Edwin Janssen for providing the AK1−/− mouse model used in this study and Dr. P. Bryant Chase for calculations regarding model solutions.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract