Uridine supplementation antagonizes zidovudine-induced mitochondrial myopathy and hyperlactatemia in mice




Zidovudine is an antiretroviral nucleoside analog reverse transcriptase inhibitor that induces mitochondrial myopathy by interfering with the replication of mitochondrial DNA (mtDNA). Because zidovudine inhibits thymidine kinases, the mechanism of mtDNA depletion may be related to an impairment of the de novo synthesis of pyrimidine nucleotides, which are required building blocks of mtDNA. This study was undertaken to determine whether mitochondrial myopathy is a class effect of antiretroviral nucleoside analogs, and whether the muscle disease can be prevented by treatment with uridine as a pyrimidine nucleotide precursor.


BALB/c mice were treated with zidovudine or zalcitabine. Some of the mice were cotreated with mitocnol, a dietary supplement with high uridine bioavailability. Mice hind limb muscles were examined after 10 weeks.


Zidovudine induced muscle fiber thinning, myocellular fat deposition, and abnormalities of mitochondrial ultrastructure. In mice treated with zidovudine, organelles contained low mtDNA copy numbers and reduced cytochrome c oxidase activity. The expression of the mtDNA-encoded cytochrome c oxidase I subunit, but not of nucleus-encoded mitochondrial proteins, was impaired. Zidovudine also increased the levels of myocellular reactive oxygen species and blood lactate. Uridine supplementation attenuated or normalized all pathologic abnormalities and had no intrinsic effects. Zalcitabine did not elicit muscle toxicity.


Our findings indicate that zidovudine, but not zalcitabine, induces mitochondrial myopathy, which is substantially antagonized by uridine supplementation. These results provide proof of the importance of pyrimidine pools in the pathogenesis of zidovudine myopathy. Since uridine supplementation is tolerated well by humans, this treatment strategy should be investigated in clinical trials.

Zidovudine is a nucleoside analog reverse transcriptase inhibitor (NRTI) that is widely used in the antiretroviral treatment of patients infected with the human immunodeficiency virus (HIV). A major limitation of long-term use of zidovudine is the occurrence of mitochondrial myopathy. Patients taking zidovudine may also experience an elevation in serum lactate, which may even serve as a noninvasive test for the presence of myopathy (1). A decreased intramuscular copy number of mitochondrial DNA (mtDNA) probably plays an important role in the pathogenesis of myopathy (2–4), but the exact mechanism of mtDNA depletion is under debate. Some NRTIs are thought to interfere with the mitochondrial metabolism by impairing the activity of polymerase-gamma, the enzyme which replicates mtDNA. Dideoxynucleoside analogs (stavudine, didanosine, and zalcitabine) are the drugs most commonly implicated (5), with zalcitabine being the strongest inhibitor of polymerase-gamma. Zidovudine, however, is only a weak inhibitor of polymerase-gamma (5).

Results of recent studies indicate that zidovudine may cause mtDNA depletion by a different mechanism, namely by inhibiting the intramitochondrial phosphorylation of thymidine, thus limiting the supply of thymidine triphosphate molecules as mtDNA building blocks (6–8). Zidovudine has also been shown to have many other effects; it inhibits adenylate kinase, impairs the mitochondrial ADP/ATP translocator, and can induce oxidative stress (5, 9). Severe mtDNA depletion and secondary respiratory chain dysfunction may contribute to the depletion of intracellular pyrimidines because a normal electron flux through the respiratory chain is required for the activity of dihydroorotate dehydrogenase (DHODH), an enzyme which is essential for pyrimidine de novo synthesis (10).

In vitro studies of models of steatohepatitis and lipoatrophy have shown that the mitochondrial toxicity of some antiretroviral agents can be antagonized by uridine supplementation (11, 12), and clinical trials have also yielded promising results with uridine for these indications (13, 14). Exogenous uridine supplementation may act by replenishing the diminished intracellular pyrimidine pools distal from DHODH through the salvage pathway and enhancing the synthesis of thymidine nucleotides (15). Alternatively, uridine or its metabolites may compete with pyrimidine analogs or their metabolites at various steps of mitochondrial pyrimidine import, degradation, or phosphorylation (15).

In this study, we investigated the in vivo effects of uridine supplementation on zidovudine myopathy and its associated hyperlactatemia. We also aimed to determine whether mitochondrial myopathy occurs only with zidovudine treatment, or if muscle abnormalities are also observed after treatment with zalcitabine, a strong polymerase-gamma inhibitor. Mice were treated with one of the NRTIs (either zidovudine or zalcitabine), and some mice were cotreated with mitocnol, a dietary supplement with high uridine bioavailability (16). After 10 weeks, hind limb muscles were harvested and analyzed.



Male BALB/c mice were purchased from Janvier (Le Genest St. Isle, France) and housed in a normal night–day rhythm under standard temperature and humidity conditions. At 7 weeks of age, the mice were divided into 6 groups. Group A (n = 10) received no treatment and served as the control group. Group B (n = 10) received 340 mg/kg/day of mitocnol (Pharma Nord, Vojens, Denmark) in the drinking water. This dosage of mitocnol corresponded on a per–body weight basis to a human dose of 24 gm/70 kg, and on a per–body surface area basis to 13 gm/m2. Groups C and D (n = 10 each) received zidovudine (100 mg/kg/day; MP Biomedicals, Strasbourg, France) in the drinking water. Groups E and F (n = 9 each) received 13 mg/kg/day of zalcitabine (Sigma, Taufkirchen, Germany). The daily dosages of zidovudine and zalcitabine corresponded to the human dosage adjusted for body area and the higher metabolic and drug disposal rate in rodents and were calculated based on daily liquid consumption of 5 ml (17–19). Groups D and F were cotreated with 340 mg/kg/day of mitocnol in the drinking water, whereas groups C and E received no mitocnol.

Mice were observed daily for fluid consumption, clinical signs, and mortality, and body weight was recorded weekly. All animals were killed by cervical dislocation at 16 weeks of age, immediately prior to organ collection and postmortem examination. Gastrocnemius and soleus muscles were snap-frozen and cryopreserved in liquid nitrogen until analyzed. Aliquots were fixed in glutaraldehyde (3%) for subsequent electron microscopy. All animal work was performed after animal welfare board approval and conformed to institutional guidelines as well as the policy of the National Institutes of Health (http://grants.nih.gov/grants/olaw/olaw.htm).

Histopathologic examination and mitochondrial ultrastructure.

Skeletal muscle fiber diameters were morphometrically quantified in all animals on a 0.09-mm2 area of 8-μm–thick, hematoxylin and eosin–stained sections of gastrocnemius muscle, using automated image analysis and processing software (Leica QWin Standard, version 2.7; Leica Microsystems Imaging Solutions, Cambridge, UK). Muscle fiber steatosis was assessed semiquantitatively by oil red O staining, and fiber degeneration was assessed by acid phosphatase staining. Two randomly selected muscle samples from each group were examined by electron microscopy as previously described (20), by an evaluator who was blinded with regard to treatment group.

Measurement of myocellular triglycerides and blood lactate.

Lipids were freshly extracted from the muscle using methanol–chloroform–water according to the Bligh-Dyer method and quantified spectrophotometrically using a sulfo-phospho–vanillin reaction on lipid standards (Sigma), as previously described (21). For lactate measurements, blood was collected immediately after cervical dislocation by puncturing the beating heart. The blood was then immediately added to ice-cold perchloric acid–EDTA solution for enzyme lactate measurement after removing proteins (19).

Measurement of enzyme activity.

The enzyme activity levels of cytochrome c oxidase, succinate dehydrogenase, and citrate synthase were measured spectrophotometrically in freshly prepared tissue extracts (22). Cytochrome c oxidase is a multisubunit respiratory chain complex encoded by both nuclear DNA and mtDNA, whereas succinate dehydrogenase is a respiratory chain enzyme encoded entirely by nuclear DNA. Citrate synthase is a nuclear DNA–encoded component of the Krebs cycle and is located in the mitochondrial matrix.

Quantification of mtDNA-encoded respiratory chain protein.

The mtDNA-encoded subunit I of cytochrome c oxidase was quantified by immunoblot analysis (23). The cytochrome c oxidase I signal was normalized to the expression of cytochrome c oxidase IV, which is encoded by nuclear DNA. Blots were also probed with a third antibody (Research Diagnostics, Flanders, NJ) against GAPDH, an enzyme that is encoded entirely in the nucleus.

Determination of mtDNA copy number.

Total DNA was extracted with the QIAamp DNA isolation kit (Qiagen, Hilden, Germany). Mitochondrial DNA and nuclear DNA copy numbers were determined by quantitative polymerase chain reaction, as previously described (17). Amplifications of mitochondrial and nuclear products were performed in triplicate. Absolute mtDNA and nuclear DNA copy numbers were calculated using serial dilutions of plasmids with known copy numbers.

Lipid peroxidation.

Malondialdehyde (MDA) is one of the end products of lipid peroxidation and an indicator of free radical production and oxidative stress. MDA was spectrophotometrically quantified in tissue using an assay for material reactive with thiobarbituric acid (24).

Statistical analysis.

Group means were compared by analysis of variance, Mann-Whitney test, unpaired t-test, or Wilcoxon analysis, as appropriate. Regressions were computed by nonlinear exponential regression analysis. All graphics and calculations were performed using SigmaPlot 2000, version 8.0 (SPSS, Chicago, IL) and SigmaStat, version 3.1 (Jandel Corporation, San Rafael, CA).


Macroscopic and microscopic pathologic examination.

Daily fluid consumption and body weight were unaffected by zidovudine treatment (data not shown). Zidovudine substantially reduced muscle fiber diameters, leading to an increased number of fibers per microscopic area (P < 0.001) (Table 1 and Figure 1). Cotreatment with mitocnol significantly attenuated zidovudine-induced fiber thinning. Fiber splitting was not observed. Zalcitabine, in contrast, did not significantly alter muscle fiber diameters. Among animals receiving no NRTI, there was a slight, but significant increase in muscle fiber diameter in the mitocnol group compared with controls (P = 0.02). Such an intrinsic effect of mitocnol was not seen among animals receiving zalcitabine.

Table 1. Effect of uridine supplementation with mitocnol on hind limb muscle histology and mitochondria in mice treated with zidovudine or zalcitabine*
 Treatment group
Control (n = 10)Mitocnol 340 mg/kg/day (n = 10)Zidovudine 100 mg/kg/day (n = 10)Zidovudine 100 mg/kg/day + mitocnol 340 mg/kg/day (n = 10)Zalcitabine 13 mg/kg/day (n = 9)Zalcitabine 13 mg/kg/day + mitocnol 340 mg/kg/day (n = 9)
  • *

    Except where indicated otherwise, values are the mean ± SD. mtDNA = mitochondrial DNA; MDA = malondialdehyde.

  • P < 0.05 versus controls.

  • P < 0.001 versus controls.

  • §

    P < 0.001 versus mice treated with zidovudine alone.

  • P < 0.05 versus mice treated with zidovudine alone.

Muscle fiber diameter, μm18.2 ± 1.220.2 ± 0.513.3 ± 1.317.1 ± 1.2§18.2 ± 1.518.5 ± 1.8
Muscle fiber number per 0.09 mm2129 ± 19131 ± 12171 ± 20151 ± 19148 ± 26125 ± 21
Muscle fiber degeneration, level of acid phosphatase staining++++
Muscle fiber steatosis, level of oil red O staining++++
Muscle lipids, mg/gm of tissue0.7 ± 0.20.8 ± 0.12.5 ± 0.61.5 ± 0.2§0.8 ± 0.10.9 ± 0.2
No. of mtDNA copies in the myonucleus587 ± 58632 ± 129472 ± 88574 ± 91577 ± 77579 ± 81
Cytochrome c oxidase activity, μmoles/minute/gm of protein29 ± 426 ± 1415 ± 823 ± 727 ± 524 ± 4
Ratio of cytochrome c oxidase to succinate dehydrogenase, % of control100 ± 25102 ± 1151 ± 34108 ± 1097 ± 8103 ± 16
Citrate synthase activity, μmoles/minute/gm of protein2,274 ± 4782,385 ± 7352,197 ± 3742,033 ± 4562,309 ± 6682,405 ± 488
Ratio of cytochrome c oxidase I to cytochrome c oxidase IV, % of control100 ± 2492 ± 1076 ± 28103 ± 19105 ± 1795 ± 15
Ratio of cytochrome c oxidase IV to GAPDH, % of control100 ± 11102 ± 22103 ± 19111 ± 23111 ± 22116 ± 23
MDA, μmoles/gm of tissue45 ± 2644 ± 1691 ± 5552 ± 2148 ± 2438 ± 31
Figure 1.

Prevention of fatty degeneration of muscle and increase in muscle fiber diameters in mice with zidovudine myopathy treated with uridine supplementation (mitocnol). A–I, Photomicrographs of representative sections from untreated mice (A, D, and G), mice treated with zidovudine alone (B, E, and H), and mice treated with zidovudine plus mitocnol (C, F, and I). Bars = 100 μm. J, Relationship between muscle fiber number and muscle fiber diameter in untreated mice (left), mice treated with zidovudine alone (middle), and mice treated with zidovudine plus mitocnol (right).

Acid phosphatase staining in the zidovudine group, but not the zalcitabine group, showed increased reactivity in numerous fibers distributed at times in patches throughout all muscle sections (Table 1 and Figure 1). This sign of enhanced lysosomal activity and fiber degeneration was almost completely prevented by cotreatment with mitocnol. Examination of the ultrastructure showed that myofibers from the zidovudine-only group contained increased numbers of intermyofibrillar mitochondria (Figure 2). The organelles were enlarged and their crystal architecture was lost and disrupted. Subsarcolemmal mitochondria were also increased in number (results not shown). Uridine supplementation largely abrogated this ultrastructural damage (Figure 2). No ultrastructural damage was observed in the mitochondria in any other group (results not shown).

Figure 2.

Representative electron photomicrographs, demonstrating myofiber degeneration with enlargement of mitochondria (arrows) and disrupted crystal architecture of the organelles in mice treated with zidovudine alone (A and C), and a reduction in the ultrastructural abnormalities of mitochondria (arrows) in mice treated with zidovudine plus mitocnol (B and D). Bars in A and B = 2.5 μm; bars in C and D = 0.6 μm.

Muscle lipid levels.

Zidovudine induced a myocellular deposition of lipids, evidenced by oil red O staining (Figure 1) and biochemical quantification (Table 1). Mitocnol alone had no effect on muscle lipid levels, but when given with zidovudine, it reduced the zidovudine-induced gain of myocellular lipids by 59% (P < 0.001). None of the mice receiving zalcitabine showed any evidence of muscle steatosis. Among all animals, muscle lipid levels were inversely correlated with fiber muscle diameter (r = −0.89, P < 0.001).

Blood lactate levels.

Compared with control mice, blood lactate levels were significantly elevated in zidovudine-treated mice (mean ± SD 1.43 ± 0.33 mM versus 1.64 ± 0.30; P < 0.05). Mitocnol alone did not produce any change in blood lactate levels (1.39 ± 0.18 mM). When mice were treated with mitocnol plus zidovudine, their blood lactate level was similar to that observed in controls (1.48 ± 0.22 mM). No blood was drawn from zalcitabine-treated mice for lactate measurements.

Mitochondrial DNA copy number.

In mice treated with zidovudine alone, the mean myocellular mtDNA copy numbers were reduced by 20% compared with control values (P = 0.003). Uridine supplementation fully normalized mtDNA copy numbers in mice treated with zidovudine. Zalcitabine, in contrast, did not cause mtDNA depletion in muscle. Among all mice, the myocellular mtDNA copy numbers were positively correlated with mean muscle fiber diameter (r = 0.53, P < 0.001) (Figure 3A) and inversely correlated with mean myocellular lipid content (r = −0.66, P < 0.001) (Figure 3B).

Figure 3.

Correlations between the number of mitochondrial DNA (mtDNA) copies and muscle fiber diameter (A) and muscle lipid levels (B), between the ratio of cytochrome c oxidase (COX) to succinate dehydrogenase (SDH) and the number of mtDNA copies (C), muscle lipid levels (D), and malondialdehyde (MDA) levels (F), and between the ratio of cytochrome c oxidase I to cytochrome c oxidase IV and muscle lipid levels (E).

Enzyme activity.

The mean enzyme activity level of cytochrome c oxidase, a respiratory chain enzyme which is composed of 3 mtDNA-encoded subunits, was depressed in the skeletal muscle of mice in the zidovudine group (52% of control activity, P < 0.001) (Table 1). Whereas uridine supplementation per se did not alter cytochrome c oxidase activity, it attenuated respiratory dysfunction when given in conjunction with zidovudine. Similar results were also obtained when cytochrome c oxidase activity was normalized to the activity of succinate dehydrogenase, a respiratory chain enzyme that does not require an intact mitochondrial genome. The mean ratio of cytochrome c oxidase to succinate dehydrogenase was reduced in the muscle of mice in the zidovudine-only group compared with the control group (P = 0.002), but unchanged in all other groups. The enzyme activity of citrate synthase, a nuclear DNA–encoded mitochondrial matrix enzyme of the Krebs cycle, did not differ significantly between any of the groups, indicating that the enzyme defect is specifically restricted to mtDNA-encoded activity.

Among all mice, absolute cytochrome c oxidase activity correlated inversely with muscle lipid levels (r = −0.58, P < 0.001), but positively with fiber diameter (r = 0.41, P < 0.002). The ratio of cytochrome c oxidase to succinate dehydrogenase was also positively correlated with mtDNA copy number (r = 0.49, P < 0.001) (Figure 3C) and fiber diameter (r = 0.61, P < 0.001), and inversely correlated with muscle lipid levels (r = −0.72, P < 0.001) (Figure 3D).

Mitochondrial DNA–encoded respiratory chain subunits.

The mean ratio of cytochrome c oxidase I to cytochrome c oxidase IV was reduced in the zidovudine group and remained essentially unchanged in all other groups (Table 1). In contrast, the ratio of cytochrome c oxidase IV to GAPDH was not decreased among zidovudine-treated mice. The ratio of cytochrome c oxidase I to cytochrome c oxidase IV was positively correlated with absolute cytochrome c oxidase activity (r = 0.26, P = 0.048), and with the ratio of cytochrome c oxidase to succinate dehydrogenase (r = 0.55, P < 0.001). These results indicate a link between reduced cytochrome c oxidase activity and defective synthesis of mtDNA-encoded cytochrome c oxidase subunits. The ratio of cytochrome c oxidase I to cytochrome c oxidase IV was positively correlated with muscle fiber diameter (r = 0.36, P = 0.03), and inversely correlated with muscle lipid levels (r = −0.61, P < 0.001) (Figure 3E).

MDA levels and lipid peroxidation.

MDA levels are an indirect indicator of lipid peroxidation due to reactive oxygen species (ROS). Compared with controls, MDA levels were increased by a factor of 2 in muscle exposed to zidovudine, but not in muscle exposed to zalcitabine (Table 1). Uridine supplementation normalized myocellular MDA formation. The intramuscular MDA content was inversely correlated with muscle fiber diameter (r = −0.71, P < 0.001), cytochrome c oxidase activity (r = −0.44, P = 0.003), the ratio of cytochrome c oxidase to succinate dehydrogenase (r = −0.68, P < 0.001) (Figure 3F), the ratio of cytochrome c oxidase I to cytochrome c oxidase IV (r = −0.71, P < 0.001), and mtDNA content (r = −0.73, P < 0.001). MDA content was positively correlated with intramuscular lipid levels (r = 0.76, P < 0.001).


The present study analyzed the effects of uridine on NRTI-mediated mitochondrial toxicity in skeletal muscle. Our findings confirm that zidovudine induces myopathy in which mtDNA-encoded respirator chain deficiency plays an important role. The reduction of cytochrome c oxidase activity and cytochrome c oxidase I expression observed only in zidovudine-treated animals strongly suggests that mtDNA depletion is the origin of mitochondrial dysfunction, rather than mtDNA depletion being a nonspecific consequence of muscle degeneration due to another cause (25). It is likely that increased ROS production originates, at least in part, from the respiratory chain, because the decreased activity of respiratory chain complexes partially blocks the flow of electrons, allowing them to react with oxygen (26).

Our results also demonstrate that myocellular toxicity is not an in vivo consequence of all members of the NRTI class of antiretroviral agents, despite the fact that zalcitabine is a strong inhibitor of polymerase-gamma and a myotoxin in vitro (27). The failure of zalcitabine to induce myopathy in vivo is not likely a function of the dose administered, because the same strain of mice treated with zalcitabine in a previous study developed steatohepatitis (17). A similar discrepancy between in vitro and in vivo muscle toxicity has also been suggested for didanosine (27, 28).

In the case of pyrimidine analogs, organ specificity may be related to cellular differences in the interaction of NRTIs with thymidine kinases (TKs). The cytoplasmic TK1, for example, is less abundant in mitotically quiescent cells (5). Recent evidence suggests that mtDNA depletion by zidovudine may result from the inhibition of mitochondrial TK2, with consecutive depletion of normal deoxypyrimidine pools as mtDNA building blocks (7, 29). This mechanism then operates predominantly in postmitotic cells in which TK1 activity as an alternate pathway of thymidine phosphorylation is down-regulated (6, 7).

Importantly, our study also demonstrates that uridine supplementation is able to attenuate all aspects of zidovudine myopathy. The exact mechanism of the beneficial effect of uridine is not fully delineated, but it is conceivable that uridine itself or its metabolites disinhibit mtDNA replication by competing with zidovudine either at TK2 or at other steps of intracellular NRTI transport or phosphorylation (11, 15). Uridine may alternatively correct an intracellular pyrimidine deficit which in itself induces cell cycle arrest and apoptosis, as demonstrated by experimental work with direct DHODH inhibitors (30, 31). To elucidate the mechanism of action of uridine, we recently treated human liver cells with an inhibitor of uridine de novo synthesis. Our results showed that pyrimidine depletion promotes the mitochondrial toxicity of thymidine analogs. Moreover, we have demonstrated that uridine supplementation totally abrogates mtDNA depletion and all other toxicities of both the NRTI and the pyrimidine-depleting drug (Walker UA: unpublished observations).

Our work lends important support to the notion of an intermediary role of pyrimidine depletion in the pathogenesis of zidovudine myopathy, and at the same time may open a new avenue for therapy. Numerous studies have examined uridine supplementation in humans for indications not related to HIV and have demonstrated the safety of uridine supplementation (15, 32). There is a theoretical concern that high cytoplasmic uridine levels may compete with NRTIs at HIV–reverse transcriptase and thereby compromise their antiretroviral efficacy. Phenotypic HIV-resistance assays and clinical trials, however, have not demonstrated a negative effect of uridine on viral suppression (14, 33–35). Aside from its bitter taste, uridine is well tolerated. The good tolerability of uridine supplementation and the promising clinical results for other NRTI-related organ toxicities (13, 14, 35) suggest that this intervention should be examined in a carefully monitored clinical trial in HIV-infected patients who have zidovudine myopathy and who cannot be switched to other antiretroviral agents.


Dr. Walker had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Lebrecht, Deveaud, Beauvoit, Bonnet, Walker.

Acquisition of data. Lebrecht, Deveaud, Beauvoit, Bonnet, Kirschner.

Analysis and interpretation of data. Lebrecht, Deveaud, Beauvoit, Bonnet, Kirschner, Walker.

Manuscript preparation. Lebrecht, Kirschner, Walker.

Statistical analysis. Lebrecht, Deveaud, Beauvoit, Bonnet, Walker.


We thank Karin Sutter and Carmen Kopp for expert technical assistance.