Influence of Mitochondrial Genetics on the Mitochondrial Toxicity of Linezolid in Blood Cells and Skin Nerve Fibers

ABSTRACT

The antibiotic linezolid is a ribosomal inhibitor with excellent efficacy. Although the administration period has been reduced to 28 days, side effects, usually of hematologic or neuropathic origin, are still reported due to secondary inhibition of mitochondrial protein synthesis. Susceptibility to linezolid toxicity remains unknown. Therefore, the objective of this study was to gain an understanding of clinical heterogeneity in response to identical linezolid exposures through exhaustive examination of the molecular basis of tissue-dependent mitotoxicity, consequent cell dysfunction, and the association of mitochondrial genetics with adverse effects of linezolid administered for the recommended period. Peripheral blood mononuclear cells (PBMC) and skin nerve fibers from 19 and 6 patients, respectively, were evaluated before and after a 28-day linezolid treatment in order to assess toxic effects on mitochondria and cells. Mitochondrial DNA haplotypes and single nucleotide polymorphisms (SNPs) in ribosomal sequences where linezolid binds to mitochondrial ribosomes were also analyzed to investigate their genetic contributions. We found that linezolid reduced mitochondrial protein levels, complex IV activity, and mitochondrial mass in PBMC and was associated with a trend toward an increase in the rate of apoptosis. In skin tissue, mitochondrial mass increased within nerve fibers, accompanied by subclinical axonal swelling. Mitochondrial haplogroup U, mutations in 12S rRNA, and the m.2706A→G, m.3197T→C, and m.3010G→A polymorphisms in 16S rRNA showed a trend toward an association with increased mitochondrial and clinical adverse effects. We conclude that even when linezolid is administered for a shorter time than formerly, adverse effects are reported by 63% of patients. Linezolid exerts tissue-dependent mitotoxicity that is responsible for downstream cellular consequences (blood cell death and nerve fiber swelling), leading to adverse hematologic and peripheral nervous side effects. Multicentric studies should confirm genetic susceptibility in larger cohorts.

INTRODUCTION

Linezolid belongs to one of the newest classes of antibiotics, oxazolidinones, which exert their therapeutic activity by interfering with bacterial protein synthesis. Linezolid binds to 23S rRNA in the large subunit of the prokaryotic ribosome, preventing the binding of aminoacyl-tRNA for bacterial translation (1). This drug has shown excellent efficacy for the treatment of respiratory tract and skin infections caused by Gram-positive pathogens and is effective against pathogens resistant to other drugs, including methicillin-resistant Staphylococcus aureus, Nocardia spp., and multidrug-resistant Mycobacterium tuberculosis (2). Linezolid has 100% oral bioavailability and reaches high concentrations in most tissues, making it a good alternative for the long-term treatment of infections related to foreign bodies.

However, long-term therapy with linezolid has been associated with severe side effects, including hyperlactatemia, lactic and metabolic acidosis (3), myelosuppression leading to hematologic disturbances such as thrombocytopenia and anemia (4–6), gastrointestinal disturbances, and optic or peripheral neuropathy (7–11). Thus, treatment with this drug has been restricted to 28 consecutive days by the U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMEA) (9, 10).

Most of these adverse events could be related to the capacity of linezolid to interfere with mitochondrial function (12, 13). The endosymbiotic theory argues that mitochondria are ancient proteobacteria that were absorbed by the eukaryotic cell in order to obtain an energy supply from oxygen consumption. Thus, mitochondrial and bacterial ribosomes share evolutionary similarities, and consequently, mitochondria are sites at which interactions between ribosomal inhibitor antibiotics and the host can manifest with severe clinical consequences common to primary mitochondriopathies (3–11). It has been suggested that members of the oxazolidinone class of antibiotics, such as linezolid and eperezolid, and other ribosomal inhibitor antibiotics, such as chloramphenicol, tetracycline, erythromycin, and aminoglycosides, inhibit not only bacterial but also mitochondrial protein synthesis and that this inhibition is responsible for associated adverse events (14–16).

Experimental studies using human cell lines and rat tissues have demonstrated decreased rates of mitochondrial protein synthesis upon treatment with linezolid (15, 16).

Clinicians are aware of the need to monitor possible adverse effects of linezolid in humans. Severe secondary manifestations have been described in case reports (12, 14, 17, 18). However, little is known about the mechanisms involved in linezolid-associated mitochondrial toxicity in humans (13, 19–21). Previous studies reported partial measures of mitochondrial function in small cohorts of patients, usually treated for extended periods longer than 28 consecutive days (13, 19–21). To date, no study has evaluated the toxicity of linezolid administered for the recommended period or the potential causes of the clinical heterogeneity found in response to linezolid.

The administration of vitamin B₆ reverses linezolid-associated cytopenias but not peripheral neuropathy (11), suggesting that the molecular mechanisms involved in mitotoxicity may be organ specific. However, little is known about linezolid tissue specificity. Similarly, no studies have evaluated the downstream consequences of linezolid mitotoxic insult for exposed cells, which is especially important considering that mitochondria are key organelles triggering or amplifying signals of cell death through apoptosis. Such cellular consequences in different tissues may also affect the clinical outcome.

We conducted the present study to explore tissue-specific toxicity, downstream cell consequences of mitochondrial protein synthesis inhibition, and individual susceptibility to the adverse effects of linezolid within the recommended 28-day period of administration. We investigated mitochondrial toxicity and cell damage in the peripheral blood mononuclear cells (PBMC) and skin nerve fibers (two of the tissues usually affected by linezolid through hematologic and peripheral neuropathy disturbances) of patients before and after a 28-day linezolid treatment.

Furthermore, we observed that following identical antibiotic exposures, patients may present a wide range of clinical severity, ranging from almost no symptoms to severe adverse reactions that may lead to death. Individual genetic susceptibility to the mitochondrial toxicity of other antibiotics has been described previously (22–26). Just as genetic variations in bacterial rRNA can induce structural changes in the bacterial ribosome leading to prokaryotic resistance to the ribosomal antibiotic (22), so genetic variations in human mitochondrial rRNA may be associated with an increase in sensitivity to linezolid in mitochondrial ribosomes, promoting hypersensitivity to the drug in patients (23–26). Consequently, mitochondrial DNA (mtDNA) mutations or polymorphisms, especially in the mitochondrial rRNA, where linezolid binds to the mitochondrial ribosome (16S rRNA), may be some of the molecular mechanisms leading to differences in clinical manifestations (26). Thus, to better understand individual susceptibility to linezolid toxicity, we conducted, in parallel, a mitochondrial genetic study in this cohort of patients receiving a 28-day linezolid treatment and presenting different degrees of clinical manifestations, in order to associate genetic variations with adverse mitochondrial, cellular, and clinical effects.

RESULTS

Clinical data.

Linezolid administration was reserved for severe infections caused by Gram-positive pathogens. Most of the patients were >50 years old, and the most common infection pathogens were staphylococci. The epidemiologic and infectious characteristics of the patients are summarized in Table 1.

TABLE 1.

Epidemiologic and infectious characteristics of the 19 patients studied^a

Code	Gender	Age (yr)	Type of infection	Microorganism(s)	Skin analysis performed
P1	Male	72	Knee prosthetic joint infection	Coagulase-negative staphylococci	−
P2	Male	71	Knee prosthetic joint infection	Coagulase-negative staphylococci	−
P3	Male	90	Knee prosthetic joint infection	Coagulase-negative staphylococci	−
P4	Male	52	Knee prosthetic joint infection	Staphylococcus aureus	−
P5	Male	51	Orthopedic implant infection	Coagulase-negative staphylococci	−
P6	Female	55	Tibial osteomyelitis	Coagulase-negative staphylococci	+
P7	Female	71	Spondylodiskitis	Negative culture	+
P8	Male	47	Hip prosthetic joint infection	Coagulase-negative staphylococci	+
P9	Female	80	Knee prosthetic joint infection	Coagulase-negative staphylococci	+
P10	Male	69	Knee prosthetic joint infection	Coagulase-negative staphylococci	−
P11	Male	25	Knee prosthetic joint infection	Enterococcus faecium	−
P12	Female	57	Orthopedic implant infection	Staphylococcus aureus	−
P13	Female	74	Hip prosthetic joint infection	Staphylococcus aureus	−
P14	Male	83	Knee prosthetic joint infection	Coagulase-negative staphylococci	−
P15	Female	85	Hip prosthetic joint infection	Negative culture	−
P16	Female	69	Osteomyelitis	Staphylococcus aureus	+
P17	Male	78	Cerebral abscess	Nocardia otitidiscaviarum	−
P18	Female	71	Knee prosthetic joint infection	Staphylococcus aureus, Enterococcus faecium	+
P19	Male	51	Septic arthritis	Staphylococcus aureus	−

^aMitochondrial analysis of blood cells was performed in all cases before and after the 28-day linezolid treatment. Additionally, mitochondrial analyses of skin biopsy specimens were performed for 6 patients before and after therapy, as indicated. Linezolid administration was reserved for severe infections, such as prosthetic joint infections, osteomyelitis, cerebral abscess, and septic arthritis caused by Gram-positive pathogens. Most of the patients were >50 years old, and the most common infection pathogens were staphylococci.

Linezolid causes mitochondrial and cell toxicity in PBMC.

A reduction of mitochondrial protein synthesis in PBMC by 55.72% ± 7.74% (P < 0.001) led to a decrease in mitochondrial respiratory chain complex IV (mtCIV) activity by 52.86% ± 5.34% (P < 0.001) and reduced mitochondrial mass by 26.23% ± 15.18% (P < 0.05) (Fig. 1). Interestingly, linezolid therapy significantly increased mtDNA content, by 59.92% ± 21.18% (P < 0.05). Expression of mitochondrial RNA (mtRNA) and ATPase-α levels was also elevated, although nonsignificantly (23.76% ± 19.21% and 41.14% ± 26.07%, respectively). These increments, which tended to overcome linezolid toxicity, were unable to prevent the nonsignificant increase in the rate of cell apoptosis (a 12.53% ± 14.56% increase in the ratio of caspase 3 levels to voltage-dependent anion channel protein [VDAC] levels) (Fig. 1).

FIG 1.

(a) Mitochondrial and cell toxicity analysis in PBMC of patients after 28 consecutive days of linezolid treatment. Results are expressed as the percentage of variation of each parameter after linezolid treatment with respect to the baseline. Prot, protein. *, P < 0.05; **, P < 0.001. Linezolid treatment significantly reduced mitochondrial protein levels in PBMC by 55.72% ± 7.74% (P < 0.001), mtCIV activity by 52.86% ± 5.34% (P < 0.001), and mitochondrial mass by 26.23% ± 15.18% (P < 0.05). The unexpected increments in mtDNA content (59.92% ± 21.18%) (P < 0.05), mtRNA levels (23.76% ± 19.21%), and ATPase-α levels (41.14% ± 26.07%) may indicate homeostatic compensatory mechanisms to overcome linezolid toxicity. Despite the development of these strategies, linezolid therapy trended to impair cell viability, though not significantly, by enhancing apoptosis and increasing the caspase 3/VDAC ratio by 12.53% ± 14.56%. (b) Western blot quantification of mitochondrial protein synthesis (COXII) with respect to cytoplasmic protein VDAC in PBMC before and after linezolid treatment.

Changes in mitochondrial protein levels in PBMC significantly and positively correlated with changes in mtCIV activity (R² = 0.190; P < 0.05 [Fig. 2, top]), demonstrating that optimal mitochondrial function depends on efficient translation of the mtDNA. Additionally, differences in ATPase-α content significantly and negatively correlated with both mtCIV activity (R² = 0.010; P < 0.05 [Fig. 2, center]) and the apoptotic rate (R² = 0.321; P < 0.05 [Fig. 2, bottom]), suggesting that enhanced energetic mechanisms may be compensating for compromised mitochondrial function to prevent apoptosis.

FIG 2.

Correlation between mitochondrial and cell toxicity parameters in PBMC of linezolid-treated patients. Results are expressed as the percentage of variation of each parameter after linezolid treatment with respect to the baseline. Changes in mitochondrial protein levels in PBMC were significantly and positively correlated with changes in mtCIV activity, demonstrating that proper mitochondrial function depends on efficient translation of the mitochondrial genome. Additionally, differences in ATPase-α contents significantly and negatively correlated with both mtCIV activity and the apoptotic rate, suggesting the existence of upregulatory energetic mechanisms designed to compensate for compromised mitochondrial function and prevent the development of apoptosis.

Linezolid treatment also results in mitochondrial and cell toxicity in skin nerve fibers.

Antibiotic treatment did not significantly impair mtRNA content in skin nerve fibers (a 14.27% ± 14.88% decrease [Fig. 3a]), intraepidermal nerve fiber density (IENFD) (an 11.28% ± 11.77% increase [Fig. 3a and b]), or fiber area (a 16.33% ± 21.13% increase [Fig. 3a]). However, the mitochondrial area increased dramatically, by 139.67% ± 78.74%, and consequently, the ratio of mitochondrial to fiber area increased significantly, by 160.12% ± 75.36% (P < 0.05) (Fig. 3a), probably to counteract mitochondrial lesions. Additionally, axonal swelling of terminal nerve endings became evident after linezolid treatment, suggesting the development of subclinical small-fiber damage (Fig. 3b).

FIG 3.

Analysis of mitochondrial and nerve fiber toxicity in the skin tissue of patients after 1 month of linezolid treatment. (a) Results are expressed as the percentage of variation of each parameter after linezolid treatment with respect to the baseline. Linezolid treatment did not significantly impair mtRNA content (decreased by 14.27% ± 14.88%), IENFD, or fiber area (increased by 11.28% ± 11.77% and 16.33% ± 21.13%, respectively). However, after linezolid treatment, the mitochondrial area increased dramatically, by 139.67% ± 78.74%, and consequently, the ratio of the mitochondrial area to the fiber area showed a significant increase of 160.12% ± 75.36% (*, P < 0.05). The increase in the mitochondrial area within nerve fibers may be interpreted as a homeostatic mechanism to compensate for the reduction in mitochondrial function. (b) Histochemical staining of small-fiber axons with PGP 9.5 as a marker (indicated by arrows), using a bright-light microscope, to calculate IENFD. With the same biopsy specimen, an immunofluorescent confocal microscope allowed the colocalization of mitochondria through an OxPhos marker for subunit I of mitochondrial complex IV within nerve fibers stained by the PGP 9.5 marker, to calculate areas corresponding to mitochondrial localization, fiber localization, or localization of mitochondria with nerve fibers. −Line, not treated with linezolid; +Line, treated with linezolid. After linezolid treatment, all skin biopsy specimens showed the presence of morphological abnormalities characterized by axonal swelling of terminal nerve endings (white arrowheads), suggesting the development of subclinical small-fiber damage.

Linezolid-derived mitochondrial toxicity in PBMC correlates with mitotoxicity in skin nerve fibers.

Due to the paucity of the skin biopsy specimens and the different natures of the tissues, mitochondrial toxicity and cell toxicity were differentially evaluated in blood and skin cells. Despite technical difficulties (the paucity of skin biopsies and different parameters evaulated in PBMC and skin biopsies), evidence of mitochondrial and cell damage was detected in both tissues (Fig. 4; Tables 2 and 3). Mitochondrial function in PBMC correlated negatively and significantly with the density of nerve fibers (IENFD) (R² = 0.41; P < 0.05). Accordingly, the mtRNA content in PBMC correlated positively and significantly with the mitochondrial content within dermic nerve fibers (R² = 0.71; P < 0.05), suggesting correlations in the severity of toxicity and in mitochondrial and cell lesions between PBMC and skin.

FIG 4.

Correlations between mitochondrial and cell toxicity markers in peripheral blood mononuclear cells (PBMC) and skin (dermic) biopsy specimens of 6 linezolid-treated patients. Results are expressed as the percentage of variation in each parameter after linezolid treatment with respect to the baseline. Mitochondrial and cell toxicities found in PBMC from linezolid-treated patients correlated positively with alterations in skin biopsy specimens.

TABLE 2.

Changes in mitochondrial and cell toxicity markers in the peripheral blood mononuclear cells of the six patients whose blood and skin nerve fibers were analyzed simultaneously

^aEach individual result is shown as a longitudinal percentage of variation after linezolid treatment with respect to the baseline. Negative values indicate reductions; positive values indicate increases.

^bMt, mitochondrial.

^cFor each marker, the mean value for the complete cohort of patients whose blood was analyzed (n = 19) is shown as a reference value. Mitochondrial parameters that were significantly altered after treatment and with respect to the baseline in the whole cohort are marked with asterisks.

TABLE 3.

Changes in mitochondrial and cell toxicity markers in the peripheral nervous systems of the six patients whose blood and skin nerve fibers were analyzed simultaneously

^aEach individual result is shown as a longitudinal percentage of variation after linezolid treatment with respect to the baseline. Negative values indicate reductions; positive values indicate increases.

^bMt, mitochondrial.

^cFor each marker, the mean value for the complete cohort of patients whose skin biopsy specimens were analyzed (n = 6) is shown as a reference value.

Mitochondrial genetic variations and potential susceptibility to mitochondrial and clinical toxicity resulting from linezolid treatment.

Despite identical antibiotic exposures, clinical manifestations of linezolid toxicity ranged from asymptomatic individuals to subjects presenting as many as 5 secondary effects, including asthenia (often accompanied by hyperlactatemia and lactic acidosis), hematologic toxicity (anemia or thrombocytopenia), or gastrointestinal disturbances (abdominal pain, nausea, and vomiting) (Table 4). None of the patients showed signs of peripheral neuropathy, which is usually documented after extended periods of treatment (>28 days) (8, 10).

TABLE 4.

Clinical and genetic characteristics of patients included in the study sorted according to clinical severity and percentage of variation in mitochondrial protein synthesis in PBMC^a

Code	Clinical manifestation(s)	Polymorphism(s) in:		Haplotype	% change in PBMC mitochondrial protein synthesis
		12S rRNA	16S rRNA
P1	Asymptomatic			H*	−75.80
P2	Asymptomatic			H*	−67.00
P3	Asymptomatic	m.961T→G		H*	−88.48
P4	Asymptomatic	m.709G→A		H*	−17.74
P5	Asymptomatic	m.1189T→C	m.1811A→G, m.2706A→G	UK	−90.06
P6	Asymptomatic		m.2706A→G, m.3010G→A	No N	−54.63
P7	Asymptomatic	m.1598G→A	m.1703C→T, m.1719G→A, m.2706A→G	N*	−58.18
P8	Anemia			H5	−99.13
P9	Hyperlactatemia			H*	−78.08
P10	Hyperlactatemia		m.2098G→A	H*	−84.89
P11	Hyperlactatemia		m.2706A→G, m.3197T→C	U5	−1.96
P12	Hyperlactatemia	m.980T→C	m.1811A→G, m.2523C→T, m.2706A→G	U1811/U*	−47.19
P13	Hyperlactatemia, asthenia	m.709G→A	m.1888G→A, m.2706A→G, m.2850T→C	T (JT, no J)
P14	Hyperlactatemia, gastrointestinal discomfort	m.1189T→C	m.1811A→G, m.2706A→G	UK	−54.74
P15	Vomiting, anemia		m.2092C→T	H*
P16	Sickness and vomiting, asthenia, anemia	m.709G→A, m.930G→A	m.1888G→A, m.2706A→G	T (JT, no J)	−12.29
P17	Hyperlactatemia, anemia, asthenia		m.3010G→A	H1	−10.60
P18	Thrombocytopenia		m.1700T→C, m.2706A→G, m.3197T→C	U5
P19	Lactic acidosis, gastrointestinal discomfort, sickness and vomiting, anemia, thrombocytopenia	961insC	m.2706A→G, m.3197T→C	U5	−50.70

^aIn addition to the polymorphisms and mutations depicted here, all the patients presented m.750A→G and m.1438A→G polymorphisms with respect to the mitochondrial DNA Cambridge reference sequence containing these 2 rare variations. Despite identical antibiotic exposures for all the patients, clinical manifestations of linezolid toxicity ranged from asymptomatic individuals to patients presenting as many as 5 secondary effects. None of them showed clinical signs of peripheral neuropathy, which has usually been described upon linezolid treatment for extended periods, longer than the currently approved 28-day period. The distributions of mitochondrial haplogroups and polymorphisms in the cohort of linezolid-treated patients resembled the standard distributions observed in European populations.

Despite the small number of patients treated, when we analyzed mitochondrial dysfunction (inhibition of mitochondrial protein synthesis) in relation to the number of clinical manifestations, nonstatistical trends toward association were observed (Fig. 5). Interesting results were observed when we analyzed mitochondrial dysfunction (inhibition of mitochondrial protein synthesis) and clinical events (number of clinical reactions) with a specific genetic variation relative to the rest of the cohort. Decreased mitochondrial protein synthesis and increased numbers of clinical symptoms were considered suggestive of increasing linezolid toxicity. There was a trend toward reduced toxicity in patients harboring haplogroup H (with 26.11% enhanced mitochondrial protein synthesis and 40.67% fewer clinical manifestations [P, nonsignificant {NS}] [Table 5]). In contrast, patients carrying haplogroup U showed trends toward increased mitotoxicity (30.49% decreased mitochondrial protein synthesis and 67% more clinical events [P, NS]). The rest of the haplotypes were scarcely represented. Thus, the results should be taken with caution, but none of them affected the mitochondrial or clinical outcome.

FIG 5.

Correlation between mitochondrial protein synthesis inhibition in peripheral blood mononuclear cells and number of clinical symptoms. Mitochondrial results are expressed as the percentage of variation after linezolid treatment with respect to the baseline. There is a trend toward an increased number of clinical symptoms in those patients who have strongly reduced mitochondrial protein synthesis after linezolid treatment.

TABLE 5.

Mitochondrial and clinical severity of symptoms of patients after linezolid treatment in relation to genetic data^a

Haplogroup or polymorphismb	No. of carriers vs noncarriers	% of change in carriers from the rest of the cohortc
		Mitochondrial protein synthesis	No. of clinical symptoms
Haplogroups
H	9 vs 10	26.11	−40.67
U	6 vs 13	−30.49	67
No N, N*, T	4 vs 15	9.30	3.92
Polymorphisms
12S rRNA
m.709G→A	3 vs 16	26.58	47.79
m.1189T→C	2 vs 17	5.18	−19.35
961insC	1 vs 18	−46.31	400
m.1598G→A	1 vs 18	−31.12	−100
m.980T→C	1 vs 18	NA	−18.03
m.930G→A	1 vs 18	NA	170.27
m.961T→G	1 vs 18	5.35	−100
16S rRNA
m.2706A→G	10 vs 9	−16.73	68.54
m.3197T→C	3 vs 16	−25.61	133
m.1811A→C	3 vs 16	5.18	−20
m.3010G→A*	2 vs 17	−19.23	27.12
m.1888G→A	2 vs 17	40.51	135.85
m.2850T→C	1 vs 18	40.51	70.94
m.2523C→T	1 vs 18	NA	−18.03
m.2098G→A	1 vs 18	−34.95	−18.03
m.2092C→T	1 vs 18	57.34	70.94
m.1719G→A	1 vs 18	−31.12	−100
m.1703C→T	1 vs 18	−31.12	−100
m.1700T→C	1 vs 18	1.16	−18.03

^aIn addition to the polymorphisms and mutations depicted here, all the patients presented the m.750A→G and m.1438A→G polymorphisms with respect to the mitochondrial DNA Cambridge reference sequence containing these 2 rare variations. There is evidence of increased mitochondrial protein synthesis inhibition and increased severity of clinical symptoms in patients harboring mtDNA haplogroup U, mutations in 12S rRNA, and some polymorphisms in the 16S rRNA sequence. With regard to mitochondrial 12S rRNA, such a trend was relevant only for one patient carrying a mutation consisting of a cytosine insertion in position 961 of mtDNA, but no increased toxicity was associated with any of the polymorphisms detected. In contrast, no mutations were found in mitochondrial 16S rRNA, but the presence of the m.2706A→G, m.3197T→C, or m.3010G→A polymorphism was, per se, responsible for an enhanced trend toward toxicity.

^bGenetic variants with increased mitochondrial and clinical toxicity are shown in boldface. Genetic variants with reduced mitochondrial and clinical adverse manifestations are shown in italics. *, 0.05 < P < 0.12.

^cEach result is expressed as the mean percentage of variation in mitochondrial protein synthesis or in the number of clinical symptoms after linezolid treatment with respect to the baseline for one specific genetic variant relative to the rest of the cohort (those not carrying that variant). Patients presenting multiple genetic variants were analyzed for each mutation or polymorphism. Negative values indicate reductions; positive values indicate increases. NA, not applicable.

Interestingly, there were trends toward differential mitochondrial and clinical manifestations according to mutations or polymorphisms in mitochondrial 12S and 16S rRNA (Table 5). With regard to mitochondrial 12S rRNA, only the presence of a mutation consisting of a cytosine insertion in position 961 of mtDNA (961insC) was associated with decreased mitochondrial translation (−46.31%) and increased clinical impairment (400%). Given that linezolid binds to 16S rRNA, we hypothesized that a high number of polymorphisms, not necessarily mutations, may be associated with differential mitochondrial and clinical toxicity. Interestingly, the m.2706A→G, m.3197T→C, and m.3010G→A polymorphisms in 16S rRNA induced similar mitochondrial and clinical toxicity (changes in mitochondrial protein synthesis and numbers of clinical symptoms, −16.73% and 68.54%, −25.61% and 133%, and −19.23 and 27.12%, respectively) than mutations in 12S rRNA, an effect that was especially enhanced for the m.3010G→A polymorphism.

DISCUSSION

Linezolid is an antibiotic with widespread use in the treatment of severe infections, and its high efficiency has prompted physicians to use it for prolonged periods (27). However, a major concern is its safety profile due to secondary inhibition of mitochondrial protein synthesis (12–16), which has been reported to be reversible after linezolid withdrawal (19). Currently, FDA and EMEA guidelines recommend linezolid treatment for 10 to 14 days and restrict its administration to 28 days to avoid adverse events (9, 10).

In this study, we evaluated mitochondrial and clinical toxicity in patients treated with linezolid for 28 consecutive days. Despite the reduction of the treatment period to 28 days, 63% of patients presented a wide panoply of secondary effects ranging from hyperlactatemia to severe gastrointestinal discomfort or thrombocytopenia, among others. Thus, we exhaustively examined the molecular basis of tissue-dependent toxicity, mitotoxicity-derived cell dysfunction, and the association of mitochondrial genetics with a diversity of adverse effects.

Given that most adverse effects of linezolid therapy involve the development of hematologic abnormalities and peripheral neuropathy, we analyzed markers for mitotoxicity in the PBMC and skin nerve fibers of treated patients. We first demonstrated that linezolid decreased mitochondrial protein levels in PBMC, leading to decreases in mitochondrial function and content. Surprisingly, we observed an increase in the mitochondrial content of nucleic acid species upstream of the protein blockage (both mtDNA and mtRNA) together with an increased quantity of ATPase, suggesting bioenergetic compensatory mechanisms to prevent further damage. These findings are consistent with previous studies reporting enhanced mitochondrial transcription (19, 28) and stability of nucleus-encoded respiratory gene transcripts (29) to counteract mitochondrial dysfunction.

Linezolid has 100% oral bioavailability and reaches high concentrations in all tissues studied (9). Interestingly, we found that the severity of mitochondrial toxicity in PBMC correlated with that in skin nerve fibers. In skin biopsy specimens, mitochondrial contents increased significantly within nerve fibers after linezolid treatment, probably to counteract mitochondrial toxicity and prevent extended damage. Diverse mitotoxic fingerprints in different tissues and the presence of mitochondria in almost all cells of the organism may explain the pleiotropic nature and systemic range of clinical reactions in treated patients.

In this study, we also measured downstream cell dysfunction after linezolid intake. Blood cells from treated patients showed partially increased apoptosis, affecting 12% of PBMC. This is a strikingly relevant percentage when one takes into account the short life span and high replacement rate of blood cells, and the strict quality control of blood cells performed by the spleen. Molecular associations between mitochondrial and apoptotic parameters in PBMC demonstrate that increased apoptosis downstream may be promoted by linezolid mitotoxicity. Linezolid also induced axonal swellings at terminal nerve endings (considered an initial marker of nerve damage [21, 30]), even in the absence of clinical symptoms of polyneuropathy. Therefore, the mitotoxicity of linezolid has different clinical consequences depending on the tissue considered: cell decay leading to myelosuppression (PBMC) or axonal damage leading to further neuropathy (nerve fibers). In both cases, mitochondrial lesions and consequent cell lesions would lead to linezolid-associated clinical toxicity.

The therapeutic activity and toxicity of linezolid are time and concentration dependent. However, upon identical antibiotic exposures, patient outcomes range from a total lack of symptoms to severe life-threatening toxicity. We hypothesized that mitochondrial genetics may modulate the severity of linezolid toxicity, as demonstrated previously for other ribosome-targeting antibiotics, including aminoglycosides (25). Like linezolid, aminoglycosides exert antibacterial activity by blocking prokaryotic protein synthesis, and hypersusceptibility to aminoglycoside-induced deafness has been demonstrated in patients who are carriers of the m.1494C→T and m.1555A→G mtDNA mutations, due to increased binding of the antibiotic to the mitoribosome (25). With linezolid administration, our exploratory study suggests increased mitotoxicity and clinical symptoms in patients harboring mtDNA haplogroup U, mutations in 12S rRNA, or, especially, polymorphisms in the 16S rRNA sequence. With regard to mitochondrial 12S rRNA, such a trend was relevant only for one patient carrying a mutation consisting of a cytosine insertion in position 961 of mtDNA, but no increased toxicity was associated with any of the polymorphisms detected. In contrast, no mutations were found in mitochondrial 16S rRNA, but the sole presence of the m.2706A→G, m.3197T→C, or m.3010G→A polymorphism was associated with higher levels of toxicity. Hypersensitivity to linezolid toxicity in patients harboring polymorphisms in 16S rRNA is consistent with the evidence of linezolid binding to the mitoribosome through this rRNA sequence, especially considering that the patient carrying the 961C insertion in 12S rRNA also presented m.2706A→G and m.3197T→C polymorphisms in 16S rRNA, which may be responsible, per se, for increased toxicity.

Nearly half of the Western European population harbors the m.2706A→G polymorphism, a definer of haplogroup H (31). Polymorphism m.3197T→C defines haplogroup U5, present in 8% of European individuals (31). Polymorphism m.3010G→A defines subhaplogroups H1 and J1, encompassing approximately 25% of European populations (26). Although m.2706A→G, m.3197T→C, and m.3010G→A are frequent polymorphisms of mitochondrial 16S rRNA in all populations, they may modulate linezolid toxicity with deleterious effects similar to those of rare mutations in alternative mitochondrial rRNA sequences. This is the first report of the potential association of linezolid toxicity with a cytosine insertion in position 961 or a m.3197T→C polymorphism in mtDNA. Interestingly, the m.2706A→G and m.3010G→A polymorphisms have been suggested to increase linezolid-mediated mitochondrial toxicity in previous case report studies (12, 32, 33) and in in vitro studies using transmitochondrial cell lines containing different haplogroups and polymorphisms (26). However, unambiguous elucidation of specific genotype-phenotype correlations and establishment of the influence of the mtDNA haplotype on the development of side effects of linezolid require further studies with larger cohorts, probably through multicentric approaches, which, additionally, could correlate toxic markers with drug levels. Indeed, although this is the largest cohort of patients treated with linezolid to be studied to date, the sample size is the main limitation of the present study, hindering our ability to draw significant conclusions. However, increasing the number of patients treated with linezolid remains challenging for several reasons. First, linezolid administration is restricted to severe infections. Second, patients with infectious systemic inflammatory response syndrome were excluded, because linezolid has been suggested previously to cause mitochondrial dysfunction (34). Third, patients receiving concomitant treatment with antibiotics other than rifampin (recommended to pharmacokinetically potentiate the therapeutic capacity of linezolid) were also excluded to avoid confounders of toxicity. In fact, rifampin has been reported to cause mitochondrial toxicity in rat tissues (35). Consequently, the concomitant use of rifampin with linezolid may be a source of interference. However, in the PBMC of treated patients, we did not observe the characteristic toxic profile of rifampin reported previously in animal models (consisting of decreases in mtRNA and ATP synthesis), suggesting that rifampin may be playing a secondary role in the observed molecular and clinical findings, in accordance with a lesser mitotoxic capacity of rifampin than of linezolid, reported previously in an in vitro study (36). However, the potential interference of rifampin in an observed mitotoxic phenotype cannot be completely excluded. Additionally, due to the ex vivo design of the present study, which hampers direct measurement of the functionalism of mitochondrial machinery, steady-state levels of molecules were measured instead. This approach makes the demonstration of causality difficult. Consequently, we cannot confirm that the mitotoxic phenotype was due to the interference of linezolid with mitochondrial protein synthesis, but previous studies in experimental cell and animal models strengthen this hypothesis (15, 16).

In summary, despite these considerations, the present findings confirm that linezolid mitotoxicity in different tissues of treated patients is correlated and is associated with downstream cell impairment, demonstrating the molecular basis of systemic side effects. Additionally, mitochondrial genetics may affect linezolid toxicity, especially for carriers of haplotype U and the m.3010G→A variant in the 16S rRNA sequence. Thus, although sensitivity to antibiotics is a multifactorial trait, prolonged courses of linezolid should be avoided, and, with short exposures, mitochondrial antibiograms should be implemented and genetic variations in mitochondrial ribosome RNA sequences should be considered as prognostic factors to predict off-target effects of linezolid.

MATERIALS AND METHODS

Design.

Observational and longitudinal follow-up of patients was conducted before and after intervention.

Patients.

Nineteen patients receiving linezolid for 1 month at a standard oral dosage (600 mg/12 h) to treat bacterial infection were prospectively included in this study in the Hospital Clinic of Barcelona (Barcelona, Spain). Patients received rifampin (600 mg/day) concomitantly, according to international clinical guidelines. Linezolid-related adverse events were monitored and registered. Hemograms were performed weekly, and levels of blood lactate (as a surrogate marker of mitochondrial toxicity) were determined when hyperlactatemia was suspected based on clinical findings.

To avoid confounders of mitochondrial toxicity, critically ill patients and those taking other drugs known to be toxic for mitochondria (antiretrovirals, antipsychotics, or statins) were excluded. The patients included took no medication other than linezolid and rifampin except for paracetamol, in some cases. Linezolid was withdrawn on the presentation of a severe adverse event (platelet count of <100,000 cells/mm³, hemoglobin level of <9 g/liter, or severe gastrointestinal manifestations).

All the individuals included in this protocol provided written consent to participate in the study, which was previously approved by the Ethical Committee of the Hospital Clinic of Barcelona (Barcelona, Spain), in accordance with the Helsinki Declaration.

Sample collection.

Twenty milliliters of blood was collected in Vacutainer tubes with EDTA before and after linezolid treatment. PBMC were isolated by use of a Ficoll gradient (37), and protein content was measured using the Bradford method (38).

For 6 of the patients, skin biopsy specimens were obtained with a disposable 3-mm circular punch from the lateral side of the distal leg, 10 cm above the lateral malleolus, before and after linezolid treatment. The skin tissues obtained were divided into two sections, one for mtRNA quantification and the other for histologic analysis. The latter section was fixed with 2% paraformaldehyde-lysine-periodate for 24 h and was cut into 50-μm slices in a microtome (21).

Mitochondrial protein levels in PBMC.

We assessed the mitochondrial protein content of cytochrome c oxidase subunit II (COXII, or p.MT-CO2) by Western blotting and normalized the value by the content of voltage-dependent anion channel protein (VDAC) to establish the relative COXII abundance per mitochondria (COXII/VDAC) (19).

mtCIV function in PBMC.

The enzymatic activity of mitochondrial respiratory chain complex IV (mtCIV, or COX) was measured spectrophotometrically (19, 39) and was expressed in nanomoles per minute per milligram of cell protein.

Mitochondrial mass in PBMC.

Mitochondrial mass was measured by Western blot analysis of VDAC content, which was normalized by β-actin levels (VDAC/β-actin) (19).

Mitochondrial DNA content in PBMC.

Total DNA was extracted by the phenol-chloroform procedure. Fragments of the mitochondrial ND2 gene and the nuclear 18S rRNA gene were amplified by quantitative PCR (qPCR) (19) to quantify mtDNA levels (expressed as the ratio of ND2 mtDNA to 18S nuclear DNA).

Mitochondrial RNA content in PBMC.

Total RNA was isolated by an affinity column-based procedure (RNeasy; Qiagen), reverse transcribed to cDNA using random primers (Applied Biosystems), and quantified using previously described targets for mtDNA (19). Levels of mtRNA were expressed as the ratio of ND2 mtRNA to 18S nuclear RNA.

ATPase-α levels in PBMC.

We assessed the mitochondrial protein content of the ATPase-α subunit by Western blotting, normalizing this expression by the VDAC content (ATPase-α/VDAC) (19).

Apoptotic rate in PBMC.

We assessed apoptosis in leukocytes by Western blotting to determine the levels of cleaved caspase 3 (the active form of this proapoptotic protein), normalizing this expression by the VDAC content (34) (caspase 3/VDAC).

Mitochondrial RNA content in skin biopsy specimens.

Total RNA was obtained after homogenizing skin tissue in TriPure reagent (Roche Diagnostics, Mannheim, Germany) and was quantified and expressed as described previously (19).

Nerve fiber and mitochondrial quantification in skin biopsy specimens.

Intraepidermal nerve fiber density (IENFD) was measured, and axonal swellings were counted (magnification, ×40) (Bristol-Myers-Squibb, Germany) by bright-light microscopy after staining with the panaxonal antibody PGP 9.5 (diluted 1:800; AbD Serotec, Oxford, United Kingdom), by following the free-floating protocol (40–42). Three random skin sections (43) were analyzed, and a 1.5-μm-long marker was used to check the swelling diameters (30).

In parallel, double fluorescent immunostaining was performed on three skin sections with the PGP 9.5 marker and the OxPhos IV mitochondrial antibody (mtCIV subunit I; diluted 1:500; Molecular Probes Inc., OR, USA) to simultaneously quantify mitochondria and localize them with cutaneous nerve fibers. Confocal images were acquired using a Leica TCS-SL laser scanning spectral microscope (Leica Microsystems, Mannheim, Germany).

The results of both bright-light and confocal microscopy were quantified using ImageJ software and were expressed as IENFD, mitochondrial area (mtCIV or mtCIV-ir [where ir means immunoreactive]), fiber area (PGP 9.5 or PGP-ir), and the area of mitochondria within nerve fibers (mtCIV-ir/PGP-ir) (21).

Mitochondrial genetic characterization of haplogroups in PBMC.

Mitochondrial haplogroups were determined by qPCR (44). Three single nucleotide polymorphisms (SNPs), defining haplogroups JT (at nucleotide position 4216), H (position 7028), and U (position 12308), were genotyped. Fourteen other mtDNA SNPs (at positions 1811, 3010, 4336, 4580, 4769, 9477, 10873, 13708, 14766, 14793, 14798, 15218, 15257, and 15693) were characterized by following a phylogenetic approach to confirm particular haplogroups. The human revised Cambridge reference sequence (NCBI Reference Sequence Database accession number NC_012920) was used to locate polymorphisms.

Mitochondrial genetic characterization of 12S and 16S rRNA sequences in PBMC.

Sanger methodology was used to sequence the MT-RNR1 and MT-RNR2 genes, encoding, respectively, mitochondrial 12S rRNA and 16S rRNA (nucleotide positions 650 to 3229 from NCBI Reference Sequence Database accession number NC_012920). The primers used were as follows: forward (a), 5′-CTCCTCAAAGCAATACACTG-3′; reverse (a), 5′-TGCTAAATCCACCTCCGACC-3′; forward (b), 5′CGATCAACCTCACCACCTCT3′; reverse (b), 5′-TGGACAACCAGCTATCACCA-3′; forward (c), 5′-GGACTAACCCCTATACCTTCTGC-3′; reverse (c), 5′-GGCAGGTCAATTTCACTGGT-3′; forward (d), 5′-AAATCTTACCCCGCCTGTTT-3′; and reverse (d), 5′-AGGAATGCCATTGCGATTAG-3′. An automatic ABI Prism genetic analyzer (Applied Biosystems, CA, USA) and the human revised Cambridge sequence (accession number NC_012920) were used to detect individual genetic variants.

Statistical analysis.

Results are shown as percentages of change after linezolid treatment (mean percentage and standard error of the mean) from values before treatment. Nonparametric tests were used to assess longitudinal differences due to intervention or association between experimental measures and clinical findings. Statistical analysis was performed with the SPSS program, version 22.00, and significance was set at a P value of <0.05.