ARTHRITIS & RHEUMATISMVol. 63, No. 8, August 2011, pp 2172–2182DOI 10.1002/art.30395© 2011, American College of Rheumatology

Hypoxia Induces Mitochondrial Mutagenesis and Dysfunctionin Inflammatory Arthritis

Monika Biniecka,1 Edward Fox,2 Wei Gao,1 Chin Teck Ng,1 Douglas J. Veale,1

Ursula Fearon,1 and Jacintha O’Sullivan1

Objective. To assess the levels and spectrum ofmitochondrial DNA (mtDNA) point mutations in syno-vial tissue from patients with inflammatory arthritis inrelation to in vivo hypoxia and oxidative stress levels.

Methods. Random Mutation Capture assay wasused to quantitatively evaluate alterations of the syno-vial mitochondrial genome. In vivo tissue oxygen levels(tPO2) were measured at arthroscopy using a Licoxprobe. Synovial expression of lipid peroxidation (4-hydroxynonenal [4-HNE]) and mitochondrial cyto-chrome c oxidase subunit II (CytcO II) deficiency wereassessed by immunohistochemistry. In vitro levels ofmtDNA point mutations, reactive oxygen species (ROS),mitochondrial membrane potential, and markers ofoxidative DNA damage (8-oxo-7,8-dihydro-2�-deoxyguanine [8-oxodG]) and lipid peroxidation (4-HNE) were determined in human synoviocytes undernormoxia and hypoxia (1%) in the presence or absenceof superoxide dismutase (SOD) or N-acetylcysteine(NAC) or a hydroxylase inhibitor (dimethyloxalylglycine[DMOG]). Patients were categorized according to theirin vivo tPO2 level (<20 mm Hg or >20 mm Hg), andmtDNA point mutations, immunochemistry features,and stress markers were compared between groups.

Results. The median tPO2 level in synovial tissueindicated significant hypoxia (25.47 mm Hg). Higherfrequency of mtDNA mutations was associated with

reduced in vivo oxygen tension (P � 0.05) and withhigher synovial 4-HNE cytoplasmic expression (P �0.04). Synovial expression of CytcO II correlated with invivo tPO2 levels (P � 0.03), and levels were lower inpatients with tPO2 <20 mm Hg (P < 0.05). In vitro levelsof mtDNA mutations, ROS, mitochondrial membranepotential, 8-oxo-dG, and 4-HNE were higher in synovio-cytes exposed to 1% hypoxia (P < 0.05); all of theseincreased levels were rescued by SOD and DMOG and,with the exception of ROS, by NAC.

Conclusion. These findings demonstrate thathypoxia-induced mitochondrial dysfunction drives mi-tochondrial genome mutagenesis, and antioxidants sig-nificantly rescue these events.

Hypoxia is characterized by an inadequate supplyof molecular oxygen, of which mitochondria are impor-tant consumers and sensors (1). Mitochondrial compo-nents are highly susceptible to attack by reactive oxygenspecies (ROS) due to their close proximity to theelectron transport chain and the presence of polyunsat-urated fatty acid–rich membranes (2). Oxidative damageto mitochondrial DNA (mtDNA) itself can affect genesencoding respiratory chain complexes and transcription,which may lead to further mtDNA mutations (3). An-other form of DNA oxidative damage results in theformation of DNA adducts, such as 8-oxo-7,8-dihydro-2�-deoxyguanine (8-oxodG). The 8-oxodG adduct isformed by the reaction of the hydroxyl radical with theDNA guanine base and is a promutagenic lesion thatmispairs with adenine, leading to GC-to-TA transver-sion. Oxidative DNA damage induced by ROS couldpotentially be a major source of mitochondrial genomicinstability, leading to respiratory chain dysfunction (4,5).

Mitochondrial membrane lipids are highly sus-ceptible to oxidative damage, and lipid peroxidation cansuppress mitochondrial metabolism and dynamics (6).Lipid peroxides affect vital mitochondrial functions,

Supported by the Health Research Board of Ireland (grantsR10238 and JRFC-05-01).

1Monika Biniecka, PhD, Wei Gao, PhD, Chin Teck Ng, MD,Douglas J. Veale, MD, Ursula Fearon, PhD, Jacintha O’Sullivan, PhD:Dublin Academic Medical Centre, St. Vincent’s University Hospital,and The Conway Institute of Biomolecular and Biomedical Research,University College Dublin, Dublin, Ireland; 2Edward Fox, PhD: Uni-versity of Washington, Seattle.

Address correspondence to Jacintha O’Sullivan, PhD, Educa-tion and Research Centre, St. Vincent’s University Hospital, Elm Park,Dublin 4, Ireland. E-mail: jacintha.osullivan@ucd.ie.

Submitted for publication November 22, 2010; accepted inrevised form April 5, 2011.

2172

such as cellular respiration and oxidative phosphoryla-tion. In addition, they alter membrane barrier proper-ties, mitochondrial membrane potential, and Ca2� buff-ering capacity (5). Products of lipid peroxidation, such as4-hydroxynonenal (4-HNE), concentrate in biomem-branes, where proteins quickly react with 4-HNE, caus-ing biochemical and biophysical protein modifications(6). Reaction of 4-HNE with mitochondrial proteins,such as subunits of complex IV of electron transportchain, is known to inactivate their function in a time- andconcentration-dependent manner (7). Recently, exten-sive formation of the 4-HNE protein adducts was shownto cause damage to electron transport chain complexesin cardiomyocytes, resulting in inhibition of electrontransport (8). Moreover, 4-HNE induces intracellularproduction of ROS (9). This highly mutagenic aldehydecan also directly bind to DNA and induce clonal nucleargenome alterations (10).

Another inner mitochondrial membrane protein,cytochrome c oxidase (CytcO), is a multi-subunit en-zyme that acts as both an O2 sensor and a regulatoryenzyme that determines mitochondrial respiration ca-pacity, and its down-regulation is a primary response tohypoxia in vitro (11). Studies have shown correlationsbetween lipid peroxidation levels and CytcO activity(12), indicating that peroxidation of membrane lipids iscrucial in regulating CytcO functions.

Rheumatoid arthritis (RA) and psoriatic arthritis(PsA) are among the most common inflammatory ar-thritides, typical features of which are joint inflamma-tion, synovial tissue hypertrophy, joint effusions, anddegradation of articular cartilage and bone (13). Severalstudies have suggested that genetic and environmentalfactors can trigger joint inflammation (14,15). Using anovel technique, we have recently demonstrated thehypoxic nature of synovial tissue in inflammatory arthri-tis, and we reported that in vivo synovial tissue partialoxygen pressure (tPO2) measurements were stronglyassociated with levels of joint inflammation, vascularity,blood vessel stability, and disease activity (16,17). Fur-thermore, our studies showed that levels of oxidativedamage in patients with inflammatory arthritis werestrongly associated with tPO2 in vivo, disease activity, andsecretion of angiogenic markers (18).

Mitochondrial DNA point mutations are randomsingle-base changes, found in many human diseases (3).These mutations may occur in all copies of mtDNAwithin a cell (homoplasmy) or there can be mixture ofmutated mtDNA and wild-type mtDNA within a singlecell (heteroplasmy). Mitochondrial DNA point muta-tions can be accumulated clonally in certain cells. It is

known that a critical threshold of mutated mtDNA mustbe reached before tissue dysfunction appears, althoughthis mechanism is still not fully understood. Clonalexpansion of mtDNA mutations has been demonstratedin aging human tissue from a number of anatomicsources (19). A recent study showed an associationbetween the presence of clonally expanded mtDNApoint mutations in colonic crypt stem cells and reducedexpression of multiple electron transport chain com-plexes, suggesting point mutation as a potential triggerfor respiratory chain alterations (20). Clonally expandedmtDNA mutations have also been detected in patientswith RA, in whom an increased incidence of thesemutations was found in synoviocytes and synovial tissue(21). However, due to technical limitations, the fre-quency and spectrum of random mitochondrial muta-tions that may be crucial in disease progression have notyet been examined in RA synovial tissue.

The aim of this study was to assess the levels andspectrum of mtDNA mutations in synovial tissue frompatients with inflammatory arthritis and to correlatethese with in vivo tissue hypoxic status, lipid peroxida-tion, and CytcO expression levels. The effect of antiox-idant treatment on the above processes was also examined.

PATIENTS AND METHODS

Patient recruitment. Twenty patients with active in-flammatory arthritis (14 with RA and 6 with PsA) who had notstarted biologic treatment were recruited from outpatientclinics at the Department of Rheumatology, St. Vincent’sUniversity Hospital. All patients fulfilled the American Col-lege of Rheumatology criteria for RA (22) or published criteriafor PsA (23). At baseline, 50% of the patients had not takendisease-modifying antirheumatic drugs (DMARDs) or ste-roids. Those who were receiving DMARDs were taking meth-otrexate (MTX) alone (35%), MTX plus sulfasalazine (10%),or hydroxychloroquine (5%). All patients, including thosetaking DMARDs, had active disease with inflammation of oneor both knees. Clinical disease activity was assessed with the28-joint Disease Activity Score (24) using the C-reactiveprotein level (DAS28-CRP). Ethical approval to conduct thisstudy was granted by St. Vincent’s Healthcare Group MedicalResearch and Ethics Committee.

Arthroscopy, oxygen partial pressure measurements,and sample collection. Under local anesthesia, each patientunderwent arthroscopy of the inflamed knee, performed usinga Wolf 2.7-mm needle arthroscope. A Licox combined PO2 andtemperature probe (Integra Life Sciences) was used to deter-mine oxygen partial pressure as previously described (18).Synovial membrane biopsy samples were obtained from thesite of the oxygen tension measurement and immediatelyembedded in mounting media for immunohistochemical ana-lysis or snap frozen in liquid nitrogen for mitochondrialmutagenesis analysis.

HYPOXIA INDUCES MITOCHONDRIAL MUTAGENESIS 2173

Mitochondrial Random Mutation Capture assay. Tocharacterize the frequencies of random mutations in synovialbiopsy samples, we used the mitochondrial Random MutationCapture assay as described previously (25). This quantitativepolymerase chain reaction (PCR)–based approach enablesprecise determination of mutation frequencies following ex-haustive digestion of all wild-type (nonmutant) sequences bythe restriction enzyme Taq I. This methodology allows forexact determination of mutation frequencies in high-throughput screens that interrogate millions of basepairs si-multaneously. With the Random Mutation Capture assay,random mitochondrial point mutations in the gene encodingthe 12S ribosomal RNA (rRNA) subunit (bp 1215–1218) arescreened and detected. This single-basepair change in one ofthe rRNA-encoding genes may impair protein translation ofsome or all of the mtDNA-encoded subunits of the electrontransport chain.

All biopsy specimens were analyzed under blindedconditions. Mitochondrial DNA was extracted using a previ-ously reported protocol (25). Following extraction, 10 �g ofmtDNA was digested with 100 units of Taq �I restrictionenzyme (New England Biolabs), 1� bovine serum albumin,and a Taq �I–specific digestion buffer (10 mM Tris HCl, 10mM MgCl2, 100 mM NaCl [pH 8.4]) for 10 hours, with 100units of Taq �I added to the reaction mixture every hour. PCRamplification was performed in 25-�l reaction mixtures con-taining 12.5 �l 2� SYBR Green Brilliant Mastermix (Strat-agene), 0.1 �l uracil DNA glycosylase (New England Biolabs),0.7 �l forward and reverse primers (10 pM/�l; IDT), and 6.7 �lH2O. The samples were amplified using a Roche LightCycler480, according to the following protocol: 37°C for 10 minutesand 95°C for 10 minutes, followed by 45 cycles of 95°C for 15seconds and 60°C for 1 minute. Samples were kept at 72°C for7 minutes and, following melting-curve analysis, immediatelystored at �80°C. The primer sequences used were as follows:for mtDNA copy number 5�-ACAGTTTATGTAGCTTA-CCTCC-3� (forward) and 5�-TTGCTGCGTGCTTGATGC-TTGT-3� (reverse); for random mutations 5�-CCTCAACAG-TTAAATCAACAAAACTGC-3� (forward) and 5�-GCGCTT-ACTTTGTAGCCTTCA-3� (reverse). All PCR products weresequenced to identify the mutation at the Taq �I recognitionsite (High-Throughput Sequencing Facility, University ofWashington).

4-HNE and CytcO subunit II immunohistochemistryanalysis and scoring. Immunohistochemistry analysis was per-formed using 7-�m cryostat synovial tissue sections and theDako ChemMate Envision Kit. Tissue sections were fixed inacetone for 10 minutes and air-dried. Nonspecific binding wasblocked using 10% casein. 4-HNE mouse monoclonal antibody(Genox) was diluted 1:40 in antibody diluent (Dako) andincubated on sections for 2 hours at room temperature. CytcOII rabbit monoclonal antibody (Abcam) was diluted 1:250 inantibody diluent and incubated on sections overnight at 4°C.An IgG1 control antibody was used as a negative control.Following primary antibody incubation, endogenous peroxi-dase activity was blocked using 0.3% H2O2. Slides wereincubated for 1 hour with horseradish peroxidase–conjugatedsecondary antibody (Dako). Color was developed in diamino-benzidine solution (1:50; Dako) and counterstained with hema-toxylin. Slides were mounted in Pertex media and analyzedusing an established and validated semiquantitative scoring

method (26). For 4-HNE, both lining and sublining layer cellswere assessed for percentage of positive cytoplasmic staining.For CytcO II, percentage of positive cytoplasmic staining wasscored in the whole synovial tissue. Percentage positivity wasgraded on a 0–4 scale, where 0 � no stained cells, 1 � 1–25%stained cells, 2 � 25–50% stained cells, 3 � 50–75% stainedcells, and 4 � 75–100% stained cells.

K4 cell line culture and cell treatments. An immortal-ized normal human synovial cell line (K4) (a kind gift from Dr.Evelyn Murphy, University College Dublin) was established;this cell line was characterized by Dr. Hermann Eibel, Univer-sity Hospital, Freiburg, Germany (27). K4 cells between pas-sages 35 and 38 were used to perform in vitro analyses. Toassess whether hypoxia drives mitochondrial mutagenesis, cellswere seeded into T25 culture flasks at a density of 2 �106cells/flask. To examine ROS production and mitochondrialmembrane potential, cells were seeded onto 96-well plates at adensity of 5,000 cells/well. RPMI 1640 culture medium con-taining 10% fetal bovine serum (FBS) was supplemented withsuperoxide dismutase (SOD) (75 �g/ml; Sigma),N-acetylcysteine (NAC) (5 mM; Sigma), or dimethyloxalylgly-cine (DMOG) (1 mM; Cayman Chemical) and cells wereexposed to normoxia (21% O2, 5% CO2) or hypoxia (3% O2,5% CO2, 90% N2 [3% hypoxia] or 1% O2, 5% CO2, 90% N2[1% hypoxia]) for 24 hours. In the control group, cells werecultured with RPMI 1640 culture media containing 10% FBSwithout further supplementation and exposed to normoxia or3% or 1% hypoxia for 24 hours.

In vitro mitochondrial dysfunction. After 24-hour in-cubation, mitochondrial mutagenesis, ROS production, andmitochondrial membrane potential were assessed. To deter-mine the frequency of mtDNA point mutations, K4 pelletswere obtained and processed using the Mitochondrial RandomCapture assay as described above. To measure the level ofROS, cells were washed twice with buffer (130 mM NaCl, 5mM KCl, 1 mM Na2HPO4, 1 mM CaCl2, 1 mM MgCl2, 25 mMHEPES [pH 7.4]). Cells were loaded with 5 �M 2�,7�-dichlorofluorescein diacetate (DCF-DA; Invitrogen) in theabove buffer for 40 minutes at 37°C. To measure mitochon-drial membrane potential, cells were washed using the abovebuffer and loaded with 5 �M rhodamine-123 (Sigma) in thebuffer for 40 minutes at 37°C. After 40-minute incubation,ROS and mitochondrial membrane potential probes wereremoved and cells were washed with the buffer and analyzedusing the SpectraMax Gemini System (Molecular Devices);DCF-DA and rhodamine-123 were excited at 485 nm, andfluorescence emission at 538 nm was recorded. Mean fluores-cence values from 4 wells for each condition were obtained.

In vitro oxidative stress markers. In vitro secretionlevels of oxidative DNA damage (8-oxodG) and lipid peroxi-dation (4-HNE) were determined in K4 cell culture superna-tants after 24-hour incubation under conditions of normoxiaand hypoxia. Levels of 8-oxodG and 4-HNE were also exam-ined in cell culture supernatants after stimulation with SOD,NAC, or DMOG. Competitive enzyme-linked immunosorbentassays (Gentaur) were used to measure 8-oxodG and 4-HNE,as previously described (18).

Statistical analysis. In vivo data are presented as themedian and interquartile range and were assessed by Wilcox-on’s signed rank test or Spearman’s rank correlation as appro-priate, using SPSS. In vitro data are presented as the

2174 BINIECKA ET AL

mean � SEM and were assessed by one-way analysis ofvariance. All P values are 2-sided; P values less than or equalto 0.05 were considered significant.

RESULTS

In vivo levels of PO2 and random mtDNA pointmutations in synovial tissue. In vivo synovial tPO2measurements were obtained in the 20 patients. Themedian PO2 level in the synovial tissue was 25.47 mm Hg(range 3.19–54.11), equivalent to an ambient oxygentension of 3.3% (range 0.42–7.11%). Low in vivo tPO2was associated with increased disease activity as assessedby the DAS28-CRP. There was a significant inversecorrelation between tPO2 levels and the DAS28-CRP(r � �0.44, P � 0.04) (data not shown), indicating thatas the joint becomes less hypoxic, disease activity isdecreased.

To determine the relationship between hypoxia

and frequency of random mitochondrial mutations, pa-tients were categorized into a high hypoxia group (invivo tPO2 �20 mm Hg [n � 13]) and a low hypoxia group(tPO2 �20 mm Hg [n � 7]); selection of 20 mm Hg as thecutoff for high hypoxia was based on previously pub-lished criteria (28). The mean level of tPO2 differedsignificantly between the high hypoxia and low hypoxiagroups (P � 0.0001). A statistically significant increasein the frequency of point mutations was detected insynovial biopsy specimens from patients with tPO2 levelsof �20 mm Hg compared to patients with tPO2 levels of�20 mm Hg (P � 0.05) (Figure 1A). Figure 1B illus-trates the number and spectrum of different types ofsynovial point mutations. Transitions, i.e., AT�GC andCG�TA, were observed most frequently. Interestingly,no difference in the mutation spectrum was foundbetween the tPO2 �20 mm Hg and tPO2 �20 mm Hggroups. Figure 1C indicates that higher in vivo tPO2 was

Figure 1. Random mitochondrial point mutations in synovial tissue. A, Significantly higherfrequency of mitochondrial DNA (mtDNA) point mutations in patients with synovial tissue partialoxygen pressure (tPO2) levels of �20 mm Hg compared to patients with tPO2 levels of �20 mm Hg.Data are presented as box plots, where the boxes represent the 25th to 75th percentiles, the lineswithin the boxes represent the median, and the lines outside the boxes represent the 10th and 90thpercentiles. B, The spectrum of mutations at the Taq I site. Solid bars represent transitions, andopen bars represent transversions. C, Significant negative correlation between the frequency ofmtDNA point mutations and the degree of hypoxia in the joint. D, Significant positive correlationbetween the frequency of mtDNA point mutations and the activity of inflammatory arthritis asassessed by the 28-joint Disease Activity Score using the C-reactive protein level (DAS28-CRP).

HYPOXIA INDUCES MITOCHONDRIAL MUTAGENESIS 2175

significantly associated with a decrease in the frequencyof mtDNA point mutations (r � �0.38, P � 0.05), andFigure 1D shows a significant increase in the frequencyof point mutations with higher disease activity (r � 0.52,P � 0.01).

Synovial tissue mitochondrial point mutationand oxidative stress. Lipid peroxidation was assessedbased on levels of 4-HNE (predominantly cytoplasmic),measured in both the lining layer and the sublining layerof synovial tissue. Figures 2A–C show representativeimages of 4-HNE expression levels in a control sampleand in samples from patients with low and high frequen-cies of mtDNA point mutation. There was a significantpositive correlation between the frequency of mtDNApoint mutations and 4-HNE cytoplasmic expression inthe lining and sublining layers of synovial tissue (r �0.46, P � 0.04 and r � 0.44, P � 0.03, respectively)(Figures 2D and E).

Cytochrome c oxidase subunit II deficiency inhypoxic synovial tissue. Figures 3A–C show representa-tive images of CytcO II expression in a control sample

and in samples from patients with tPO2 levels of �20 mmHg and levels of �20 mm Hg. A low percentage ofpositively stained cells was detected in patients with tPO2levels of �20 mm Hg, while significantly higher expres-sion of CytcO II was found in patients with tPO2 levels of�20 mm Hg (P � 0.05) (Figure 3D). Higher numbers ofCytcO II–positive cells were significantly correlated withhigher in vivo tPO2 levels (r � 0.51, P � 0.03) (Figure3E).

Effect of hypoxia on mitochondrial dysfunction.To determine the direct consequences of hypoxia onmitochondrial function, K4 cells were exposed to nor-moxia and hypoxia (3% or 1%), and changes in thefrequency of point mutations, ROS production, andmitochondrial membrane potential were measured.In addition, we evaluated whether rescue of mito-chondrial dysfunction using antioxidants (SOD andNAC) and a hydroxylase inhibitor (DMOG) could res-cue random mitochondrial mutation, correct defects inmitochondrial membrane potential, and reduce ROSproduction.

Figure 2. Frequency of mitochondrial DNA (mtDNA) point mutations and oxidative stress inhypoxic synovial tissue. A–C, A synovial biopsy sample stained with IgG1 as a negative control (A),a sample with a low frequency of mtDNA point mutations, exhibiting weak staining for4-hydroxynonenal (4-HNE) (B), and a sample with a high frequency of mtDNA point mutations,exhibiting strong staining for 4-HNE (C). Representative images are shown. Original magnifica-tion � 20. D and E, Significant positive correlation between the frequency of mtDNA pointmutations and the degree of 4-HNE cytoplasmic expression in both the lining layer (D) and thesublining layer (E) of synovial tissue.

2176 BINIECKA ET AL

Figure 4A shows the frequency of mitochondrialpoint mutations in cells exposed to normoxia or 1%hypoxia and stimulated with SOD, NAC, and DMOG.Hypoxia (1%) significantly increased the frequency ofmitochondrial mutation events in comparison to nor-moxia (1.35-fold increase; P � 0.05). Administration ofSOD, NAC, or DMOG significantly reduced the fre-quency of point mutations under 1% hypoxia (1.65-, 1.8-,and 2.65-fold, respectively; all P � 0.0001 versus 1%hypoxia control). Intracellular ROS levels were signifi-cantly increased under 1% hypoxia (1.5-fold increase),and SOD and DMOG significantly reduced ROS pro-duction (1.55- and 1.75-fold, respectively, both P � 0.05)(Figure 4B). Mitochondrial membrane potential wassignificantly (1.6-fold) increased under 1% hypoxia (P �0.01), and treatment with SOD, NAC, or DMOG signif-icantly attenuated mitochondrial membrane hyperpolar-ization (2.15-, 1.75- and 1.65-fold, respectively; all P �0.0001) (Figure 4C). K4 cells exposed to 3% hypoxia,

which reflects the in vivo joint environment, also showedincreased levels of mitochondrial point mutations, ROS,and mitochondrial membrane potential (data notshown). However, induction of these mitochondrial dys-functions was greater when cells were exposed to 1%hypoxia.

In vitro hypoxic accumulation of oxidative stressmarkers. To determine the direct consequences of hyp-oxia on secretion of markers of oxidative stress, levels of8-oxodG and 4-HNE were examined in cell culturesupernatants. Compared to control K4 cell culture su-pernatants exposed to normoxia, cells exposed to 1%hypoxia exhibited significantly elevated accumulation of8-oxodG (1.25-fold increase and 3-fold increase, respec-tively; both P � 0.0001) (Figures 5A and B). Stimulationwith SOD, NAC, or DMOG significantly suppressed thehypoxic secretion of oxidative stress markers (2.75-,1.65-, and 4.4-fold, respectively, for 8-oxodG and 4.15-,3.9-, and 4.95-fold, respectively, for 4-HNE; both P �

Figure 3. Expression of cytochrome c oxidase subunit II (CytcO II) in synovial tissue. A–C, Asynovial biopsy sample stained with IgG1 as a negative control (A), a sample with a tissue partialoxygen pressure (tPO2) level of �20 mm Hg, exhibiting a low number of CytcO II–positive cells (B),and a sample with a tPO2 level of �20 mm Hg, exhibiting a high number of CytcO II–positive cells(C). Representative images are shown. Original magnification � 20. D, CytcO II expression inpatients with tPO2 levels of �20 mm Hg compared to patients with tPO2 levels of �20 mm Hg. Dataare presented as box plots, where the boxes represent the 25th to 75th percentiles, the lines withinthe boxes represent the median, and the lines outside the boxes represent the 10th and 90thpercentiles. E, Significant positive correlation between the in vivo tPO2 level and the degree ofCytcO II expression.

HYPOXIA INDUCES MITOCHONDRIAL MUTAGENESIS 2177

0.0001). This is consistent with our results showing thatwhile hypoxia increases ROS levels in vitro, SOD, NAC,and DMOG are able to decrease ROS production andsubsequently reduce DNA and lipid oxidative damage.

DISCUSSION

Hypoxia has been implicated in driving the pro-inflammatory response (17,29), although its role inrelation to mitochondrial genome dysfunction in inflam-

Figure 4. Mitochondrial DNA (mtDNA) mutations, reactive oxygenspecies (ROS) production, and mitochondrial membrane potential(MMP) in K4 cells exposed to normoxia (open bars) or 1% hypoxia(solid bars) and stimulated with superoxide dismutase (SOD),N-acetylcysteine (NAC), or dimethyloxalylglycine (DMOG). Frequen-cies of mtDNA mutations (A), levels of ROS production (B), andfrequency of mitochondrial membrane potential (C) were all signifi-cantly increased in control cells exposed to 1% hypoxia compared tocells exposed to normoxia. SOD, NAC, and DMOG significantlyreduced the frequency of mtDNA point mutations and the hyperpo-larization of mitochondrial membrane potential under 1% hypoxiacompared to control 1% hypoxia (A and C), and SOD and DMOG alsosignificantly reduced hypoxia-induced ROS production (B). Values arethe mean � SEM.

Figure 5. Levels of 8-oxo-7,8-dihydro-2�-deoxyguanine (8-oxodG)and 4-hydroxynonenal (4-HNE) in K4 cells exposed to normoxia (openbars) or 1% hypoxia (solid bars) and stimulated with SOD, NAC, orDMOG. Levels of 8-oxo-dG (A) and 4-HNE (B) were significantlyincreased in control cells exposed to 1% hypoxia compared to cellsexposed to normoxia. SOD, NAC, and DMOG significantly reducedthe secretion of both oxidative stress markers under 1% hypoxiacompared to control 1% hypoxia. Values are the mean � SEM. SeeFigure 4 for other definitions.

2178 BINIECKA ET AL

matory arthritis is less well understood. In this study, wehave shown for the first time that levels of randommitochondrial mutations significantly correlate with invivo PO2 levels in synovial tissue; a higher frequency ofrandom mitochondrial mutations was significantly asso-ciated with reduced in vivo oxygen levels, as well as withhigher synovial lipid peroxidation and loss of cyto-chrome c oxidase expression. We have also mechanisti-cally shown in vitro that hypoxia drives the accumulationof these random mitochondrial mutations and altersROS, mitochondrial membrane potential, 8-oxodG, and4-HNE levels, all of which can be rescued by antioxi-dants (SOD and NAC) and a hydroxylase inhibitor(DMOG).

This is the first reported study in which altera-tions of the mitochondrial genome induced by hypoxia insynovial tissue have been quantitatively evaluated usingthe newly validated mitochondrial Random MutationCapture assay. This methodology relies on single-molecule amplification to detect rare mutations amongmillions of wild-type bases and analyzes mitochondrialmutagenesis at single-basepair resolution (25). Unlikeother techniques used to date to identify mutations inRA (21), the mitochondrial Random Mutation Captureassay can screen a large number of mtDNA moleculesfor the presence of unexpanded random mutationswhich may be crucial in driving disease progression.With the assay’s unprecedented sensitivity, 1 mutationcan be detected among 10 million wild-type bases.

We detected a wide range of tPO2 levels withinthe hypoxic joint and showed that these levels wereassociated with the frequency of mitochondrial pointmutations. The mutations detected were mainly transi-tions, characteristic of mutation following oxidativestress (30). Interestingly, the mutation spectrum wasindependent of the in vivo hypoxia level. Hypoxia hasbeen associated with increases in free radical generation,and our study demonstrated a strong link between ROSlevels and the frequency of mitochondrial mutagenesisevents. Our data complement in vitro studies that haveshown that ROS levels are a primary source of mito-chondrial mutagenesis (31). Other work has demon-strated that H2O2 induces double-strand breaks whichcan mediate mtDNA deletions (32). Moreover, trans-genic mice that overexpress mitochondrially targetedcatalase were found to have reduced H2O2 productionand low numbers of clonal mutations and deletions (33).We found an increase in mitochondrial mutagenesisassociated with lower tPO2 levels and higher diseaseactivity, suggesting that accumulation of random mito-chondrial mutations driven by hypoxia can result in a

mitochondrial mutator phenotype and mtDNA altera-tion may be implicated in the pathogenesis of inflamma-tory arthritis.

Previous studies using a DuraScript reversetranscription–PCR system demonstrated that mtDNAsomatic mutations are present in high frequency insynovial tissue of RA patients (21). These mutationswithin protein-encoding genes of mtDNA may be rec-ognized by the immune system, and mtDNA alterationscould result in increased expression of major histocom-patibility complex. Other reports have described pro-inflammatory features of oxidatively damaged mtDNA,which increased NF-�B activity and tumor necrosisfactor � production and resulted in induction of arthritisin mice (34).

We have previously observed synovial tissue lipidperoxidation (4-HNE) in patients with inflammatoryarthritides and found no significant differences in4-HNE levels between RA patients and PsA patients(18), which may suggest that oxidative stress mechanismsare similar in the two diseases. We demonstrated that invivo PO2 levels were inversely correlated with lipidperoxidation in synovial tissue and synovial fluid but notin matched serum, suggesting that local levels of hypoxiaand oxidative stress may be early markers of diseaseprogression. In the present study we have significantlyextended these findings, demonstrating for the first timea significant positive association between frequencies ofmitochondrial mutations and 4-HNE expression in boththe lining and sublining layers of synovial tissue. Thissuggests that hypoxia-induced lipid peroxides may occursimultaneously with higher numbers of mtDNA pointmutations.

These findings are consistent with those of otherstudies that have suggested a substantial role of oxida-tive damage in mitochondrial mutagenesis. Levels oflipid peroxidation correlated positively with mtDNAdeletions in human tissue during aging (35), and humanretinal epithelial cells exposed to chemical oxidantsrapidly accumulated mtDNA mutations (36). Experi-ments using genetic mutants lacking mitochondrial SODhave provided direct evidence that oxidative mtDNAdamage can be a major contributor to mitochondrialgenomic instability (4); however, our study is the first toshow that hypoxia induces random mitochondrial ge-nome alterations in the synovial tissue as demonstratedusing the Random Mutation Capture assay system.

Cytochrome c oxidase deficiency is the mostcommonly recognized respiratory chain defect (37).Suppression of CytcO II has been detected in patientswith juvenile idiopathic arthritis but not in healthyindividuals (38). In the present study, we demonstrated

HYPOXIA INDUCES MITOCHONDRIAL MUTAGENESIS 2179

abundant CytcO II expression in patients with tPO2levels of �20 mm Hg, whereas in more hypoxic syno-vium, a loss of CytcO II was observed. Therefore,hypoxia decreased CytcO II expression, which is consis-tent with a corresponding increase in random pointmutations of mtDNA. Previous studies (11,39,40) haveshown that low oxygen levels reduce the enzyme kineticsof cytochrome c oxidase in cardiac myocytes and hepa-tocytes and are responsible for the increase in ROSsignaling during hypoxia, suggesting that this enzymefunctions as an oxygen sensor.

Functionally, we demonstrated that hypoxia (1%level) induces mtDNA point mutation, coupled withROS overproduction, increased oxidative stress, andmitochondrial membrane hyperpolarization. These find-ings strengthen our in vivo correlations and furthersupport the hypothesis that an increase in mutationfrequency is a consequence of elevated oxidative dam-age to the mitochondrial genome.

Hypoxia has been shown to be a stress factor thatincreases mitochondrial membrane potential and stimu-lates ROS overproduction through activation of themitochondrial ATP–sensitive potassium channel (41).Angiogenesis and inflammation are regulated by redox-sensing transcription factors. Therefore, manipulation ofoxidative stress levels is likely to alter blood vesselstability and inflammatory response. Recently, a novellink between oxidative stress, angiogenesis, and inflam-mation has been demonstrated. That study showed thatproducts of lipid oxidation induced angiogenesis inde-pendently of vascular endothelial growth factor/vascularendothelial growth factor receptor 2 signaling, and thisprocess was mediated by activation of Toll-like receptor2 (42).

In the present study, we further examinedwhether rescuing mitochondrial defects may reducemitochondrial mutagenesis events. Antioxidants such asSOD and NAC have been used to inhibit oxidative stressby acting as ROS scavengers. Moreover, NAC caninfluence NF-�B activation and modulates the inflam-matory response. We found that mitochondrial altera-tions are significantly reduced in the presence of SODand NAC, coupled with a reduction in mitochondrialgenome mutagenesis, a decrease in oxidative stresslevels, and altered mitochondrial membrane potential.Previous studies using mouse models of arthritis showedreduced levels of proinflammatory cytokines after treat-ment with SOD (43). Furthermore, the finding of in-creased numbers of blood vessels in SOD-1�/� mice (44)suggests that ROS play a role in regulation of angiogen-esis. Further, NF-�B down-regulation induced by NAC

has been associated with decreased release of proinflam-matory cytokines (45) and angiogenesis (44).

We used DMOG as an inhibitor of prolyl hydrox-ylase domain (PHD) enzymes. Inhibition of these en-zymes enables activation of hypoxia-inducible factor 1�(HIF-1�)–dependent gene expression and therefore islikely to mimic the effects of hypoxia. Previous studieshave shown that HIF-1� activation has both prosurvivaland proapoptotic effects on cells exposed to hypoxia,which may be related to a posttranslational HIF-1�modification. While the phosphorylated HIF-1� form isa positive regulator of cell survival, the dephosphory-lated form acts as a pro-death factor that stabilizes p53tumor suppressor protein (46).

We have demonstrated that DMOG preventedhypoxia-induced ROS production and accumulation ofoxidative stress markers, and subsequently reduced thelevel of mtDNA mutations. This is consistent with thefindings of a previous genetic study showing lower levelsof ROS production and mitochondrial oxidative damagein muscle cells from PHD-deficient mice (47). In inflam-matory bowel disease and chronic kidney failure,DMOG reduced secretion of proinflammatory cyto-kines, macrophage infiltration, and oxidative stress(48,49). Furthermore, in PHD2-haplodeficient mice,HIF-1 activity normalized blood vessel maturation andoxygen supply (50).

In conclusion, the present findings clearly dem-onstrate for the first time that greater mitochondrialmutation burden in synovial tissue is associated withhigher hypoxia levels in vivo and in vitro, and antioxi-dants can rescue these significant mitochondrial genomealterations. Additionally, these data implicate mitochon-drial genome alterations as an acute consequence ofsevere hypoxia in the synovial joint.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising itcritically for important intellectual content, and all authors approvedthe final version to be published. Dr. O’Sullivan had full access to allof the data in the study and takes responsibility for the integrity of thedata and the accuracy of the data analysis.Study conception and design. Veale, Fearon, O’Sullivan.Acquisition of data. Biniecka, Fox, Gao, Ng, Veale, O’Sullivan.Analysis and interpretation of data. Biniecka, Fox, Gao, Ng, Veale,Fearon, O’Sullivan.

REFERENCES

1. Bell EL, Emerling BM, Chandel NS. Mitochondrial regulation ofoxygen sensing. Mitochondrion 2005;5:322–32.

2. Wallace DC, Fan W. Energetics, epigenetics, mitochondrial genet-ics. Mitochondrion 2010;10:12–31.

2180 BINIECKA ET AL

3. Taylor RW, Turnbull DM. Mitochondrial DNA mutations inhuman disease. Nat Rev Genet 2005;6:389–402.

4. Doudican NA, Song B, Shadel GS, Doetsch PW. Oxidative DNAdamage causes mitochondrial genomic instability in Saccharomy-ces cerevisiae. Mol Cell Biol 2005;25:5196–204.

5. Ott M, Gogvadze V, Orrenius S, Zhivotovsky B. Mitochondria,oxidative stress and cell death. Apoptosis 2007;12:913–22.

6. Poli G, Biasi F, Leonarduzzi G. 4-hydroxynonenal-protein ad-ducts: a reliable biomarker of lipid oxidation in liver diseases. MolAspects Med 2008;29:67–71.

7. Musatov A, Carroll CA, Liu YC, Henderson GI, Weintraub ST,Robinson NC. Identification of bovine heart cytochrome c oxidasesubunits modified by the lipid peroxidation product 4-hydroxy-2-nonenal. Biochemistry 2002;41:8212–20.

8. Hill BG, Dranka BP, Zou L, Chatham JC, Darley-Usmar VM.Importance of the bioenergetic reserve capacity in response tocardiomyocyte stress induced by 4-hydroxynonenal. Biochem J2009;424:99–107.

9. Nakamura K, Miura D, Kusano KF, Fujimoto Y, Sumita-Yo-shikawa W, Fuke S, et al. 4-hydroxy-2-nonenal induces calciumoverload via the generation of reactive oxygen species in isolatedrat cardiac myocytes. J Card Fail 2009;15:709–16.

10. Kowalczyk P, Ciesla JM, Komisarski M, Kusmierek JT, Tudek B.Long-chain adducts of trans-4-hydroxy-2-nonenal to DNA basescause recombination, base substitutions and frameshift mutationsin M13 phage. Mutat Res 2004;550:33–48.

11. Chandel NS, Budinger GR, Choe SH, Schumacker PT. Cellularrespiration during hypoxia: role of cytochrome oxidase as theoxygen sensor in hepatocytes. J Biol Chem 1997;272:18808–16.

12. Paradies G, Ruggiero FM, Petrosillo G, Quagliariello E. Peroxi-dative damage to cardiac mitochondria: cytochrome oxidase andcardiolipin alterations. FEBS Lett 1998;424:155–8.

13. Firestein GS. Evolving concepts of rheumatoid arthritis. Nature2003;423:356–61.

14. Karouzakis E, Neidhart M, Gay RE, Gay S. Molecular and cellularbasis of rheumatoid joint destruction. Immunol Lett 2006;106:8–13.

15. Tak PP, Zvaifler NJ, Green DR, Firestein GS. Rheumatoidarthritis and p53: how oxidative stress might alter the course ofinflammatory diseases. Immunol Today 2000;21:78–82.

16. Kennedy A, Ng CT, Biniecka M, Saber T, Taylor C, O’Sullivan J,et al. Angiogenesis and blood vessel stability in inflammatoryarthritis. Arthritis Rheum 2010;62:711–21.

17. Ng CT, Biniecka M, Kennedy A, McCormick J, FitzGerald O,Bresnihan B, et al. Synovial tissue hypoxia and inflammation invivo. Ann Rheum Dis 2010;69:1389–95.

18. Biniecka M, Kennedy A, Fearon U, Ng CT, Veale DJ, O’SullivanJN. Oxidative damage in synovial tissue is associated with in vivohypoxic status in the arthritic joint. Ann Rheum Dis 2010;69:1172–8.

19. Fayet G, Jansson M, Sternberg D, Moslemi AR, Blondy P, LombesA, et al. Ageing muscle: clonal expansions of mitochondrial DNApoint mutations and deletions cause focal impairment of mito-chondrial function. Neuromuscul Disord 2002;12:484–93.

20. Greaves LC, Barron MJ, Plusa S, Kirkwood TB, Mathers JC,Taylor RW, et al. Defects in multiple complexes of the respiratorychain are present in ageing human colonic crypts. Exp Gerontol2010;45:573–9.

21. Da Sylva TR, Connor A, Mburu Y, Keystone E, Wu GE. Somaticmutations in the mitochondria of rheumatoid arthritis synovio-cytes. Arthritis Res Ther 2005;7:R844–51.

22. Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF,Cooper NS, et al. The American Rheumatism Association 1987revised criteria for the classification of rheumatoid arthritis.Arthritis Rheum 1988;31:315–24.

23. Veale D, Rogers S, FitzGerald O. Classification of clinical subsetsin psoriatic arthritis. Br J Rheumatol 1994;33:133–8.

24. Prevoo ML, van ‘t Hof MA, Kuper HH, van Leeuwen MA, van dePutte LB, van Riel PL. Modified disease activity scores thatinclude twenty-eight–joint counts: development and validation in aprospective longitudinal study of patients with rheumatoid arthri-tis. Arthritis Rheum 1995;38:44–8.

25. Vermulst M, Bielas JH, Loeb LA. Quantification of randommutations in the mitochondrial genome. Methods 2008;46:263–8.

26. Youssef PP, Kraan M, Breedveld F, Bresnihan B, Cassidy N,Cunnane G, et al. Quantitative microscopic analysis of inflamma-tion in rheumatoid arthritis synovial membrane samples selectedat arthroscopy compared with samples obtained blindly by needlebiopsy. Arthritis Rheum 1998;41:663–9.

27. Haas C, Aicher WK, Dinkel A, Peter HH, Eibel H. Characteriza-tion of SV40T antigen immortalized human synovial fibroblasts:maintained expression patterns of EGR-1, HLA-DR and somesurface receptors. Rheumatol Int 1997;16:241–7.

28. Oliver KM, Garvey JF, Ng CT, Veale DJ, Fearon U, Cummins EP,et al. Hypoxia activates NF-�B-dependent gene expressionthrough the canonical signaling pathway. Antioxid Redox Signal2009;11:2057–64.

29. Roiniotis J, Dinh H, Masendycz P, Turner A, Elsegood CL, ScholzGM, et al. Hypoxia prolongs monocyte/macrophage survival andenhanced glycolysis is associated with their maturation underaerobic conditions. J Immunol 2009;182:7974–81.

30. Vermulst M, Bielas JH, Kujoth GC, Ladiges WC, Rabinovitch PS,Prolla TA, et al. Mitochondrial point mutations do not limit thenatural lifespan of mice. Nat Genet 2007;39:540–3.

31. Shokolenko I, Venediktova N, Bochkareva A, Wilson GL, Alex-eyev MF. Oxidative stress induces degradation of mitochondrialDNA. Nucleic Acids Res 2009;37:2539–48.

32. Srivastava S, Moraes CT. Double-strand breaks of mouse musclemtDNA promote large deletions similar to multiple mtDNAdeletions in humans. Hum Mol Genet 2005;14:893–902.

33. Schriner SE, Linford NJ, Martin GM, Treuting P, Ogburn CE,Emond M, et al. Extension of murine life span by overexpressionof catalase targeted to mitochondria. Science 2005;308:1909–11.

34. Collins LV, Hajizadeh S, Holme E, Jonsson IM, Tarkowski A.Endogenously oxidized mitochondrial DNA induces in vivo and invitro inflammatory responses. J Leukoc Biol 2004;75:995–1000.

35. Wei YH, Kao SH, Lee HC. Simultaneous increase of mitochon-drial DNA deletions and lipid peroxidation in human aging. AnnN Y Acad Sci 1996;786:24–43.

36. Mambo E, Gao X, Cohen Y, Guo Z, Talalay P, Sidransky D.Electrophile and oxidant damage of mitochondrial DNA leadingto rapid evolution of homoplasmic mutations. Proc Natl Acad SciU S A 2003;100:1838–43.

37. Clark KM, Taylor RW, Johnson MA, Chinnery PF, Chrzanowska-Lightowlers ZM, Andrews RM, et al. An mtDNA mutation in theinitiation codon of the cytochrome C oxidase subunit II generesults in lower levels of the protein and a mitochondrial encepha-lomyopathy. Am J Hum Genet 1999;64:1330–9.

38. Ishikawa S, Mima T, Aoki C, Yoshio-Hoshino N, Adachi Y,Imagawa T, et al. Abnormal expression of the genes involved incytokine networks and mitochondrial function in systemic juvenileidiopathic arthritis identified by DNA microarray analysis. AnnRheum Dis 2009;68:264–72.

39. Chandel N, Budinger GR, Kemp RA, Schumacker PT. Inhibitionof cytochrome-c oxidase activity during prolonged hypoxia. Am JPhysiol 1995;268:L918–25.

40. Duranteau J, Chandel NS, Kulisz A, Shao Z, Schumacker PT.Intracellular signaling by reactive oxygen species during hypoxia incardiomyocytes. J Biol Chem 1998;273:11619–24.

41. Hu HL, Zhang ZX, Chen CS, Cai C, Zhao JP, Wang X. Effects ofmitochondrial potassium channel and membrane potential onhypoxic human pulmonary artery smooth muscle cells. Am JRespir Cell Mol Biol 2010;42:661–6.

HYPOXIA INDUCES MITOCHONDRIAL MUTAGENESIS 2181

42. West XZ, Malinin NL, Merkulova AA, Tischenko M, Kerr BA,Borden EC, et al. Oxidative stress induces angiogenesis by acti-vating TLR2 with novel endogenous ligands. Nature 2010;467:972–6.

43. Iyama S, Okamoto T, Sato T, Yamauchi N, Sato Y, Sasaki K, et al.Treatment of murine collagen-induced arthritis by ex vivo extra-cellular superoxide dismutase gene transfer. Arthritis Rheum2001;44:2160–7.

44. Kubota M, Shimmura S, Kubota S, Miyashita H, Kato N, Noda K,et al. Hydrogen and N-acetyl-L-cysteine rescue oxidative stress-induced angiogenesis in a mouse corneal alkali-burn model. InvestOphthalmol Vis Sci 2011;52:427–33.

45. Sakurada S, Kato T, Okamoto T. Induction of cytokines andICAM-1 by proinflammatory cytokines in primary rheumatoidsynovial fibroblasts and inhibition by N-acetyl-L-cysteine andaspirin. Int Immunol 1996;8:1483–93.

46. Suzuki H, Tomida A, Tsuruo T. Dephosphorylated hypoxia-

inducible factor 1� as a mediator of p53-dependent apoptosisduring hypoxia. Oncogene 2001;20:5779–88.

47. Aragones J, Schneider M, van Geyte K, Fraisl P, Dresselaers T,Mazzone M, et al. Deficiency or inhibition of oxygen sensor Phd1induces hypoxia tolerance by reprogramming basal metabolism.Nat Genet 2008;40:170–80.

48. Cummins EP, Seeballuck F, Keely SJ, Mangan NE, Callanan JJ,Fallon PG, et al. The hydroxylase inhibitor dimethyloxalylglycine isprotective in a murine model of colitis. Gastroenterology 2008;134:156–65.

49. Song YR, You SJ, Lee YM, Chin HJ, Chae DW, Oh YK, et al.Activation of hypoxia-inducible factor attenuates renal injury inrat remnant kidney. Nephrol Dial Transplant 2010;25:77–85.

50. Mazzone M, Dettori D, Leite de Oliveira R, Loges S, Schmidt T,Jonckx B, et al. Heterozygous deficiency of PHD2 restores tumoroxygenation and inhibits metastasis via endothelial normalization.Cell 2009;136:839–51.

2182 BINIECKA ET AL