Aims Cardiac energy requirement is met to a large extent by oxidative phosphorylation in mitochondria that are highly abundant in cardiac myocytes. Human mitochondrial thioredoxin reductase (TXNRD2) is a selenocysteine-containing enzyme essential for mitochondrial oxygen radical scavenging. Cardiac-specific deletion of Txnrd2 in mice results in dilated cardiomyopathy (DCM). The aim of this study was to investigate whether TXNRD2 mutations explain a fraction of monogenic DCM cases.
Methods and results Sequencing and subsequent genotyping of TXNRD2 in patients diagnosed with DCM (n = 227) and in DCM-free (n = 683) individuals from the general population sample KORA S4 was performed. The functional impact of observed mutations on Txnrd2 function was tested in mouse fibroblasts. We identified two novel amino acid residue-altering TXNRD2 mutations [175G > A (Ala59Thr) and 1124G > A (Gly375Arg)] in three heterozygous carriers among 227 patients that were not observed in the 683 DCM-free individuals. Both DCM-associated mutations result in amino acid substitutions of highly conserved residues in helices contributing to the flavin–adenine dinucleotide (FAD)-binding domain of TXNRD2. Functional analysis of both mutations in Txnrd2−/− mouse fibroblasts revealed that contrasting to wild-type (wt) Txnrd2, neither mutant did restore Txnrd2 function. Mutants even impaired the survival of Txnrd2 wt cells under oxidative stress by a dominant-negative mechanism.
Conclusion For the first time, we describe mutations in DCM patients in a gene involved in the regulation of cellular redox state. TXNRD2 mutations may explain a fraction of human DCM disease burden.
Dilated cardiomyopathy (DCM) is a frequent cause of congestive heart failure and is the most common diagnosis in patients undergoing heart transplantation. According to systematic studies, familial transmission is observed in about 20–30% of DCM cases.1–3 Rare mutations in more than 20 different disease genes, most of them encoding for structural proteins of cardiomyocytes, have been identified.4–7 Mutations in known disease genes, though, explain <20% of inherited cases.8
Thioredoxin reductases are essential components of the thioredoxin system9 and are therefore crucial for the control of cellular redox balance. They are selenocysteine (Sec)-containing, homodimeric flavoenzymes that maintain thioredoxins, small proteins that catalyse redox reactions, in their reduced state using the reducing power of NADPH.10 Three mammalian thioredoxin reductases exist; a cytosolic (TXNRD1), a mitochondrial (TXNRD2), and a testis-specific thioredoxin reductase (TXNRD3). The Sec-residue, encoded by a UGA codon, is the penultimate amino acid of the C-terminal catalytic centre of all thioredoxin reductases and is essential for enzyme activity. The premature termination of protein synthesis at the UGA codon is prevented by a stem-loop-like structure, the selenocysteine insertion sequence (SECIS) element located in the 3′ UTR.11,12 Cardiac energy requirement is met to a large extent by oxidative phosphorylation in mitochondria that are highly abundant in cardiac myocytes.13 The mitochondrial thioredoxin reductase (TXNRD2), along with mitochondrial thioredoxin and peroxiredoxins III and V, is of paramount importance for mitochondrial scavenging of reactive oxygen species (ROS).9,10
It is well established that excessive ROS causes oxidative stress and cell death.14 On the other hand, compelling evidence has established a role for ROS as modulators of intracellular signalling cascades.15 Beyond providing protection against ROS, thioredoxins are known to inhibit or activate apoptotic signalling molecules like apoptosis signal-regulating kinase 1 and Ras or transcription factors like NF-κB.16 We showed that glutathione (GSH) peroxidase 4, along with GSH, senses and translates oxidative stress into a distinct cell death signalling cascade involving the activation of 12/15-lipoxygenase and apoptosis-inducing factor.17
Recently, we generated and characterized transgenic mice deficient for Txnrd2.18 Ubiquitous inactivation resulted in embryonic death of anaemic embryos exhibiting marked thinning of the ventricular heart walls. Analysis of fibroblasts isolated from Txnrd2−/− embryos revealed a critical role for Txnrd2 in the removal of toxic ROS species. Heart-specific inactivation of Txnrd2 resulted in a phenotype reminiscent of human DCM with dilatation of heart chambers and thinning of ventricular walls and death shortly after birth.18 This study set out to search for and characterize rare TXNRD2 mutations associated with DCM in humans.
From January 1996 to July 2004, we collected blood from 227 consecutive patients with DCM at three participating centres in Germany: Deutsches Herzzentrum München, 1. Medizinische Klinik, Klinikum rechts der Isar, München, and Zentrum für Innere Medizin, Klinikum Garmisch-Partenkirchen. Cardiac catheterization and echocardiography was performed in all patients. Patients' charts were used as the main source for clinical information. Diagnosis of DCM was based on the ‘Guidelines for the study of familial dilated cardiomyopathies’.19 Patients were included if they had an ejection fraction of the left ventricle (LV) of <45% and an LV end-diastolic diameter of >117% of the predicted value corrected for age and body surface area according to the equation of Henry et al.20 Patients with coronary heart disease (>50% stenosis of at least one coronary artery or a major branch), a history of severe systemic arterial hypertension (arterial blood pressure >160/100 mmHg documented at repeated measurements), myocarditis (suspected or confirmed), persistent high-rate supraventricular arrhythmias, systemic disease, pericardial disease, congenital heart disease, or cor pulmonale were excluded. All patients had given written informed consent for participation in the study. The investigation conforms to the principles outlined in the Declaration of Helsinki and was approved by the institutional Ethics Committee.
General population control sample
Between 1999 and 2001, we conducted an epidemiological survey of the general population living in or near the city of Augsburg, Germany (KORA S4).21 This was the fourth in a series of population-based surveys originating from our participation in the World Health Organisation (WHO) Multinational MONItoring of trends and determinants in CArdiovascular disease (MONICA) project. The study population consisted of residents of German nationality born between 1 July 1925 and 30 June 1975 and identified through the registration office. A sample of 6640 subjects was drawn with 10 strata of equal size according to gender and age. Following a pilot study of 100 individuals, 4261 individuals (66.8%) agreed to participate in the survey, which were ethnic Germans with very few exceptions (>99.5%). During 2002 and 2003, we reinvestigated a subsurvey of 880 persons specifically for cardiovascular diseases. Seven hundred and two individuals from that subsurvey were selected as population-based controls. In nineteen (2.7%) individuals, an ejection fraction of the LV of <45% was determined by echocardiography or symptoms/signs of heart failure were observed. The remaining 683 individuals were used as a DCM and congestive heart failure-free control sample. Blood samples were drawn after informed consent had been obtained.
DNA extraction and DNA sequencing
Peripheral venous whole blood samples were collected using EDTA-vials (Sarstedt, Numbrecht, Germany). DNA was extracted from these samples using the commercially available QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany).
A randomly selected subset of 96 patients of the entire DCM study population was subjected to capillary Sanger sequencing. Primers, flanking the 17 coding exons plus the 3′ SECIS (exon 18) element of TXNRD2, were designed using the ExonPrimer software based on Primer3 (http://primer3.sourceforge.net/) according to the human genome build hg17. Amplifications were conducted following standard polymerase chain reaction (PCR) protocol procedures. PCR products were purified with both Exonuclease I and Shrimp alkaline phosphatase (Fermentas, St Leon-Rot, Germany) to digest the remaining primers and to reduce the remaining dNTPs. Subsequent cycle sequencing was performed including the respective forward and reverse primers and BigDye-Terminator 3.1 (Qiagen). Thereafter, products were cleaned by DyeEx (Qiagen). Both forward and reverse strands were sequenced for all 18 TXNRD2 exons in the population of 96 patients. Sequencing was performed with an ABI 3730 Automated Sequencer (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's recommendations. Results of sequencing were analysed with Mutation Surveyor v3.00 (Softgenetics, State College, PA, USA).
Matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry genotyping
All synonymous and non-synonymous exonic sequence variants identified by direct sequencing in the population of 96 DCM patients were subsequently genotyped in all DCM patients (n = 227) and in the KORA controls (n = 683) using matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry (MALDI-TOF, Autoflex HT, Sequenom, San Diego, CA, USA) in a 384-well format using the standard protocols supplied by the manufacturer as described previously.22 Genotype-analyses were performed with the Sequenom MassARRAY Typer Analyzer software v18.104.22.168 (Sequenom).
Cloning of full-length wild-type Txnrd2, generation of both Txnrd2 mutants, and cloning into the lentiviral expression system
Vector pGEM-Teasy-Txnrd2 was a kind gift from Dr Antonio Miranda-Vizuete (Universidad Pablo de Olavide, Sevilla, Spain). The vector contained most parts of the 5′ region of murine Txnrd2 cDNA including the mitochondrial leader sequence; however, it lacked the 3′ UTR including the SECIS element. The missing sequence was amplified by PCR from mouse liver cDNA18 with the primers Oligo-SECIS-BglII-for (5′-TCTCAGAGATCTGAGAAGATGTGGATGGAAC-3′) and Oligo-SECIS-rev (5′-GTTTGAACCCCTGGCATTTCTAGAGCACT-3′). The resulting 160 bp PCR product was purified, cloned into pDrive via PaeI and BglII (Qiagen) and sequenced. The 3′ region of Txnrd2 was cloned via PaeI and BglII in pGEM-T-Easy-Txnrd2 (4702 bp), yielding pGEM-T-Easy-Txnrd2 [in the following referred to as wild-type (wt)].
Site-directed PCR mutagenesis using pGEM-T-Easy-Txnrd2 as a template was performed to generate the Txnrd2-A59T and Txnrd2-G375R mutant forms. Codons were chosen according to the highest murine codon usage of the respective amino acid. The Ala codon (GCC) at position 175 of Txnrd2 gene was replaced with Thr (ACC) using the primer pair Oligo-Txnrd2-A59T-for (5′-AGCGGGAATCGATTATAAAGAT-3′)/Oligo-Txnrd2-A59T-rev (5′-GCCAGAAGCTTTCTTGCCTTGATAGC-3′); the Gly codon (GGG) at position 375 was mutated to Arg (AGG) with the primer pair Oligo-Txnrd2-G375R-for 5′-GCTATCAAGGCAAGAAAGCTTCTGGC-3′)/Oligo-Txnrd2-G375R-rev (5′-GCCAGAAGCTTTCTTGCCTTGATAGC-3′). Mutations were verified by sequencing. All three forms were additionally tagged by a novel N-terminal TAPe (tandem affinity purification tag enhanced)23 for detection of the tagged proteins with a FLAG-specific antibody in immunocytochemistry.
The lentiviral expression vector 442L117 was used for efficient gene transfer and expression of wt Txnrd2 and the two mutant variants in mouse embryonic fibroblasts (MEFs). XhoI and EcoRI were used for digestion of 442L1 (7870 bp) and pcDNA3-NTAPe-Txnrd2 (1992 bp). Two shorter fragments carrying the mitochondrial NTAPe version of Txnrd2 were transferred into the backbone of 442L1, yielding the different NTAPe mitochondrial Txnrd2 forms. In any of these vectors, the VENUS nuclear membrane anchor protein in the bicistronic expression cassette was replaced by the puromycin acetyltransferase gene from the plasmid 442L1-NTAPe-Txnrd1 using BsrGI and SnaBI. This allowed stable expression of the various forms under puromycin selection.
Isolation, maintenance of mouse embryonic fibroblasts, and stable expression of the different Txnrd2 forms in mouse embryonic fibroblasts
Mouse embryonic fibroblasts were isolated from E12.5 Txnrd2−/− and Txnrd2+/+ embryos and cultured in DMEM supplemented with 10% FCS, 1% glutamine, 50 U/mL penicillin G, and 50 µg/mL streptomycin as described. For stable expression of wt Txnrd2, Txnrd2-A59T, and Txnrd2-G375R in wt and knockout cells in a bicistronic manner, a third-generation lentiviral expression system was utilized as described.17 Transduced cells were selected with puromycin (1 µg/mL) for at least 3 weeks prior to the start of the experiment.
Immunoblotting and production of an antibody directed against murine Txnrd2
Cell lysis, protein determination, SDS–PAGE, and protein transfer were performed as described.17 A peptide sequence next to the C-terminus of Txnrd2 (VKLHISKRSGLEPTVTG) lacking the three C-terminal amino acids Cys, Sec, and Gly was used to raise monoclonal antibodies against Txnrd2 in rats. The peptide was obtained from Peptide Specialty Laboratories (Heidelberg, Germany) and was coupled to ovalbumin (OVA) or bovine serum albumin (BSA) at the C-terminus (peptide-OVA/KLH).
Immunocytochemistry and confocal microscopy
Mouse embryonic fibroblasts were seeded onto cover slips in six-well cell culture dishes at ∼45 000 cells per well, cultured for 24 h in standard DMEM and then fixed for 15 min in 2% PFA solution. Cells were washed twice with PBS and then permeabilized by treatment with 0.15% Nonidet P-40 (NP-40) in PBS for 3 min. Unspecific antibody binding was minimized by washing three times in PBS+ [PBS containing 1% (w/v) BSA, 0.15% (w/v) glycine]. After incubation of cells over night with the primary antibody (FLAG; Sigma-Aldrich, Deisenhofen, Germany; F1804 and Prx III; LabFrontier; LP-PA0030) at 4°C in a humidified chamber, cells were washed with PBS, treated twice with PBS/NP-40 (0.15%), and transferred to PBS+. Cells were then incubated with fluorophore-labelled secondary antibodies (anti-Mouse-Alexa Fluor 568; A-10037 and anti-Rabbit-Alexa Fluor 488 A-21441; Invitrogen, Karlsruhe, Germany) in the dark for 45 min. Cells were then washed twice in PBS/NP-40 (0.15%) and twice in PBS. Finally, DAPI solution (1:10 000 in PBS) was briefly added, and cells were washed twice in PBS and mounted in mounting medium (Dako, Glostrup, Denmark). Cover slips were sealed with nail polish to prevent drying and they were kept overnight at 4°C in the dark. Confocal microscopy was performed with a Leica DM IRBE microscope (500 nm excitation and 550 nm emission), a Leica TCS SP2 scanner, and Leica Confocal software (Leica Microsystems GmbH, Wetzlar, Germany). Pictures were processed by Adobe Photoshop CS3. Processing included the removal of unspecific background and addition of false colours (blue for DAPI, green for Alexa488, and red for Alexa568) as well as the assembly of an overlay of a representative optical section from each cell. Scale bars included represent 10 µm.
Treatment of mouse embryonic fibroblasts with l-buthionine sulfoximine and determination of cell viability
Cells were treated with increasing concentrations of l-buthionine sulfoximine (BSO; Sigma-Aldrich). The cell number was determined 48 h later by trypan blue exclusion as described previously.18
Transmission electron microscopy
Cells were harvested and fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4; Electron Microscopy Sciences, Hatfield, PA, USA) and embedded in epoxy resin (epon 812; Electron Microscopy Sciences). Ultrathin sections were examined with an EM 10 CR transmission electron microscope (Carl Zeiss, Inc.). For image acquisition, a MegaView III camera system (Olympus) was used.
Statistical analysis of patients and controls was performed with the STATA SE 8.0 Statistics/Data analysis package (StataCorp LP, TX, USA). Genotypes were compared between patients and controls with two-sided χ2 test. Statistical analysis of the cellular studies was performed using SigmaStat 3.1 (Systat Software GmbH, Erkrath, Germany). Within each panel, wt and knockout cells expressing empty vector (mock), wt Txnrd2, and the different mutants were compared using all pairwise multiple comparison procedures (Holm–Sidak method). P-values <0.05 were considered as significant.
We modelled the structure of human TXNRD2 including FAD and NADPH according to the crystal structure of human TXNRD1 (monomers C and D, 22)24 and murine Txnrd2,25 using the Swiss-Model automated comparative protein modelling server (http://swissmodel.expasy.org/).
To identify disease-associated TXNRD2 mutations, we studied a cohort of 227 DCM patients and 683 individuals from a general population sample. The baseline characteristics of patients (n = 227) and the general population sample (n = 683) are shown in Table 1. The mean (±SD) ejection fraction in DCM patients was 26.9 ± 9.5 vs. 65.5 ± 11.9% in the general population sample. Sequencing of all exons and the SECIS region of TXNRD2 in a subset of the DCM patients (n = 96) yielded seven synonymous and seven non-synonymous TXNRD2 variants (Table 2). Genotype distributions of non-synonymous variants in the DCM patients and the general population sample were not significantly different (Table 3).
Clinical characteristics of dilated cardiomyopathy patients and control individuals from the general population sample
DCM patients (n = 227)
Controls (n = 683)
59.7 ± 12.8
57.4 ± 12.4
26.9 ± 9.5
65.5 ± 11.9
67.3 ± 8.3
47.5 ± 6.4
Family history for DCM (+/−/unknown)
Left bundle branch block
Previous pace maker implantation
Previous ICD implantation
Previous heart transplantation
Data are presented as mean ± standard deviation or number of patients (%). DCM, dilated cardiomyopathy; EF, ejection fraction; LVEDD, left ventricular end-diastolic diameter; ICD, implantable cardioverter defibrillator.
Genotype distributions of synonymous and non-synonymous exonic TXNRD2 variants derived from forward and reverse sequencing of 96 of the 227 dilated cardiomyopathy patients
175G > A
GG 95 AG 01 AA 00
177C > T
CC 20 CT 57 TT 19
196G > T
GG 41 GT 46 TT 09
762C > T
CC 94 CT 02 TT 00
816C > T
CC 94 CT 02 TT 00
858G > C
GG 95 GC 01 CC 00
895A > C
AA 56 AC 38 CC 02
933C > T
CC 95 CT 01 TT 00
1077C > T
CC 95 CT 01 TT 00
1101G > A
GG 95 AG 01 AA 00
1109C > T
CC 56 CT 37 TT 03
1124G > A
GG 95 AG 01 AA 00
1150G > A
GG 92 AG 04 AA 00
1206G > A
GG 65 AG 26 AA 05
Detected variants in n = 96 DCM patients with nucleotide and exon position within the TXNRD2 gene. Genotype distributions of the detected variants showed no significant deviation from the Hardy–Weinberg equilibrium (P > 0.05 for all variants). Genotypes are highlighted in bold letters in the table.
Genotype distributions of non-synonymous TXNRD2 variants in dilated cardiomyopathy patients and controls
DCM patients, n = 227 (CR)
Controls, n = 683 (CR)
175G > A
GG 222 AG 2 AA 0 (98.7%)
GG 674 AG 0 AA 0 (98.6%)
1124G > A
GG 216 AG 1 AA 0 (95.6%)
GG 670 AG 0 AA 0 (98.1%)
196G > T
GG 101 GT 100 TT 26 (100%)
GG 282 GT 314 TT 76 (98.4%)
858G > C
GG 224 GC 2 CC 0 (99.6%)
GG 677 GC 2 CC 0 (99.4%)
895A > C
AA 149 AC 70 CC 8 (100%)
AA 450 AC 200 CC 26 (99.0%)
1109C > T
CC 135 CT 81 TT 11 (100%)
CC 353 CT 233 TT 34 (90.8%)
1150G > A
GG 216 GA 6 AA 0 (97.8%)
GG 543 GA 5 AA 0 (80.2%)
Detected amino acid exchanging variants in the entire DCM population (n = 227) and n = 683 controls from a general population sample (KORA S4). CR, call rate of MALDI-TOF genotyping. n.d., not determined. Genotype distributions of the detected variants showed no significant deviation from the Hardy–Weinberg equilibrium in the controls (P > 0.05 for all variants). RefSNP accession IDs (rs numbers) of previously described variants: 1rs5748469, 2rs5992495, 3rs2073752. Genotypes are highlighted in bold letters in the table.
Mutations in dilated cardiomyopathy cases
We identified two novel amino acid residue-altering TXNRD2 mutations [175G > A (Ala59Thr = A59T) and 1124G > A (Gly375Arg = G375R)] in three heterozygous carriers of the 227 patients (3/227 = 1.3%). Both A59T and G375R were not observed in the general population KORA S4 sample. The first mutation (A59T) results in a substitution of alanine by threonine at residue 59 of the major splice isoform (isoform 1) of TXNRD2 and was identified twice (2/227 = 0.88%) in the 227 DCM patients. Figure 1A shows a sequence electropherogram of one of the A59T mutation carriers. Both patients carrying this mutation were not knowingly related. The second mutation (G375R) results in a substitution of glycine by arginine at residue 375 of the major splice isoform (isoform 1) of TXNRD2 and was found in one of the 227 DCM patients (1/227 = 0.44%). Figure 1B shows a sequence electropherogram of the G375R mutation carrier. Clinical characteristics of the three patients with novel TXNRD2 mutations that were only found in DCM patients are displayed in Table 4. We obtained all available information about the three index patients and their families by contacting the patients and their relatives by phone and by scheduling visits for patients and relatives in our outpatient clinic. In addition, genetic analyses in search of the mutation carried by the index case in relatives were performed whenever possible: Patient 1 (A59T mutation carrier) was childless and died at the age of 68. The mother of Patient 1 died at the age of 75 due to congestive heart failure. The father had a negative history for cardiovascular disease and died in World War II. The only sister of Patient 1 had atrial fibrillation and died of sudden cardiac death at the age of 69. No family members were available for clinical assessment and genetic analyses. Patient 2 (A59T mutation carrier) had an entirely negative family history of DCM by hearsay. The patient was childless, had no siblings, and died at the age of 65. No family members were available for clinical assessment and genetic analyses. Patient 3 (G375R mutation carrier) was childless as well and also had a negative family history for DCM. He died at the age of 83. Two half-sisters and two daughters of the half-sisters were available for clinical assessment including physical examination, electrocardiogram, and ultrasound echocardiography. Clinical assessment, in particular LV size and function, was normal in all four relatives. Genetic analyses showed that the four relatives do not carry the G375R mutation. The discovery of two heterozygous mutations in three DCM patients raised the questions whether the mutations were functionally silent or important, i.e. abolishing or impairing the function of the enzyme. In case the variants were functionally important, the question would have to be answered, whether the mutations act in a dominant-negative fashion, thus impairing the function of the heterozygous wt allele.
Identification of dilated cardiomyopathy-associated TXNRD2 mutations (marked with arrows). (A) Sequence electropherogram from Patient 2, who was a heterozygous carrier of the 175G > A (Ala59Thr, A59T) mutation. He was also a heterozygous carrier of the synonymous variant 177C > T (Ala59Ala) (asterisks). (B) Sequence electropherogram from Patient 3, who was a heterozygous carrier of the 1123G > A (Gly375Arg, G375R) mutation.
The protein structure of TXNRD2 and the FAD-binding domain
To gain better insight into a possible adverse role of the mutations for TXNRD2 function, we followed two complementary approaches: structural modelling of TXNRD2 and integrated analysis of the evolutionary conservation of FAD-binding domains in thioredoxin reductases and GSH reductases (GR), which are evolutionary highly related enzymes. Although the crystal structure of murine Txnrd2 as well as rat and human Txnrd1 have been resolved,24–26 the molecular nature of the FAD-binding domain has remained poorly defined in thioredoxin reductases. The FAD-binding domain is essential for enzyme function since the reducing equivalents from NADPH are first transferred to FAD, from where they are passed on to the N-terminal redox-reactive centre within the same molecule and eventually to the Sec-containing C-terminal catalytic site of the second monomer.27 Close inspection of the structure revealed that the human TXNRD2-binding pocket for FAD is formed by the N-terminal parts of four helices (h1: 49–61, h2: 92–106, h4: 228–240, and h6: 368–382), the N-terminally adjacent amino acids, and eight non-contiguous, non-helical short stretches of amino acids. The mutations A59T and G375R are located in helices 1 and 6, respectively, that both contribute to the formation of the FAD-binding pocket (Figure 2A).
The identified mutations localize to highly conserved amino acid residues in the FAD-binding domain of TXNRD2. (A) Overview of one TXNRD2 molecule (grey). The surfaces of FAD and NADP are shown in yellow and blue, respectively. Amino acid variants [red (observed in patients); blue and green (observed in patients and controls, not tolerated or tolerated according to SIFT analysis, respectively)] and relevant amino acids (grey) are shown as a ball and stick model. N- and C-termini are marked by N (grey) and C (orange). Molecular graphics images were produced using the UCSF Chimera package. (B) Alignment between human thioredoxin reductase 2 and human glutathione reductase. The alignment was performed using NCBI Blast. The numbering of amino acids is taken from the protein structures of mouse Txnrd2 (1ZDL) and human glutathione reductase (3DJG). Amino acids shown in bold are conserved to more than 80% in 46 FAD- and NADPH-binding oxidoreductases (36 thioredoxin reductases and 10 glutathione reductases) across a large number of species (see also Supplementary material online, Figure S1). Helices in TXNRD2 are underlined in green and helices in glutathione reductases in yellow. Amino acids in close contact with FAD as revealed by analysis of the modelled hTXNRD2 structure are shown in magenta. Corresponding amino acids forming the FAD-binding domain of human glutathione reductase28 are underlined in cyan. The binding pocket for FAD is formed by four conserved helices, amino acids N-terminally adjacent to these helices, and eight non-helical, non-contiguous stretches of highly conserved amino acids. In glutathione reductases, a fifth helix participates in the formation of the FAD-binding pocket. This helix is somewhat disturbed in TXNRD2 (amino acids 208–212), but the corresponding conserved residues contribute to the binding pocket in a similar manner. The mutations G375R and A59T are shown as red letters and the non-synonymous variants observed in patients as well as in controls in blue letters. Note that mutations G375 and A59T are located in helices contributing to the formation of the FAD-binding pocket. Of the non-synonymous variants, only I370T is located in a helix contributing to FAD binding, but threonine at this position is structurally tolerated and represents the evolutionary more ancient allele (see also Supplementary material online, Figure S1, Figure S2, and Table S1).
As FAD binding is common to thioredoxin reductases and GR, we reasoned that amino acids conserved among both types of enzymes would participate in FAD binding. To address this, we made two sequence alignments, one comparing human thioredoxin reductase 2 with human GR (Figure 2B), and a second that included 36 thioredoxin reductases (including 13 thioredoxin reductase 2 genes) and 10 GR genes (see Supplementary material online, Figure S1and Table S1). Amino acids shared among all or almost all enzymes across a large number of species were defined and marked in the alignment of TXNRD2 and GR. The overlay of the structural and evolutionary analysis clearly demonstrated that the regions participating in FAD binding are evolutionary highly conserved. We then compared the FAD-binding domain, which had been previously determined for GR,28 with the structural information obtained from the analysis of the modelled TXNRD2 structure. Notably, the FAD-binding domain of GR as defined by Schulz et al.28 is virtually identical to that of TXNRD2 as defined by inspection of the TXNRD2 structure (Figure 2B). The four helices as well as the non-contiguous, non-helical stretches of amino acids are highly conserved in evolution. In GR, a fifth helix participates in the formation of the FAD-binding pocket. This helix is somewhat disturbed in TXNRD2 (amino acids 208–212), but the corresponding conserved residues contribute to the binding pocket in a similar manner.
Implications of the mutations for TXNRD2 protein structure
G375R is located in the middle of helix 6. Not only is arginine much larger than glycine, it is also a polar amino acid. Our model indicates that the charged side chain points into the core of the enzyme (Figure 2A). This makes it likely that the bulky charged side chain disrupts the hydrophobic interaction with the neighbouring helix 1 and thus destroys the FAD-binding pocket. A59T (Figure 2A) is located at the end of helix 1 and is thus not directly involved in FAD binding. Our sequence alignments (see Supplementary material online, Figure S1) including 46 enzymes revealed only alanine and valine at this position. Two explanations may be given that are not mutually exclusive. First, an unpolar amino acid at this position may be required in this region. In our model, the side chain of threonine causes clashes with A55, located on helix1 or with V65 in the adjacent β-sheet (Figure 2A). Proper backfolding of this β-sheet towards FAD is required for hydrogen bonding of D69 to the ribose of FAD (as shown for E50 in GR28) and for bringing helix 2 into the proper position relative to the other helices forming the FAD-binding pocket. Furthermore, the polar threonine may disturb the movement of the second redox centre, which is located on the flexible C-terminal part of the second subunit.
In addition, we took a close look at the position of the five non-synonymous variants that were identified in patients as well as in controls. Four of them (A66S, R286S, S299R, and G384S) are located more in peripheral regions of TXNRD2 facing the solvent and are neither involved in FAD/NADPH binding nor enzymatic function (Figure 2A). The fifth non-synonymous variant I370T is located at the beginning of helix 6, but it points away from FAD and is thus not involved in FAD binding. It points into a pocket formed by residues P365, L367, M390, Y392, and V395 of one subunit and V494’ of the other. All amino acids of this pocket are strictly conserved in mouse Txnrd2, which harbours threonine at this position. This shows that threonine is tolerated. Moreover, amino acid sequence alignment of orthologous TXNRD2 sequences revealed that 8 of 11 species harbour threonine at this position, providing evidence that threonine is the evolutionary older variant.
Taken together, G375 and A59 are highly conserved across a wide range of species, whereas the five other non-synonymous variants are conserved to a much lower degree in evolution and can be predicted not to interfere with FAD binding (for a summary, see Supplementary material online, Table S2).
Both mutations abolish the function of Txnrd2
We reasoned that if the mutations were functionally silent and did not impact on Txnrd2 function, they would be able to rescue the phenotype of Txnrd2−/− cells in a manner similar to the wt Txnrd2 gene. To address this question experimentally, we cloned mouse wt Txnrd2 and the two mutants Txnrd2-A59T and Txnrd2-G375R into a bicistronic lentiviral vector and expressed them stably in primary Txnrd2−/− MEFs. Immunoblotting of cellular lysates with a Txnrd2-specific antibody showed that all three variants were expressed in knockout MEFs, albeit, to varying extent that was highly reproducible (Figure 3A). Double immunocytochemical staining of these cells with a FLAG-specific antibody for the reconstituted wt Txnrd2 and a peroxiredoxin III (Prx III)-specific antibody for mitochondria29 revealed the expected mitochondrial localization of wt Txnrd2 in Txnrd2−/− MEFs (Figure 3B). Likewise, localization of the A59T mutant was mitochondrial in Txnrd2−/− MEFs as shown by co-staining of Prx III and the TAPe-tagged mutant (Figure 3B). The G375R mutant, however, was barely detectable immunocytochemically in accordance with the data obtained by immunoblotting (compare Figure 3A), rendering unequivocal assignment to a distinct cellular compartment difficult. In Txnrd2−/− cells, this mutant also appeared to be localized solely or predominantly in mitochondria (Figure 3B). Next, we studied the function of both Txnrd2 mutants in Txnrd2−/− MEFs. To this end, we used an assay that is based on the fact that Txnrd2 is indispensible for cell survival of fibroblasts when the cells are depleted of GSH by treatment with the γ-glutamyl-cysteine-synthetase inhibitor BSO.18 As illustrated in Figure 3C, reconstitution of wt Txnrd2 in Txnrd2−/− cells fully rescued cell death induced by GSH depletion. In contrast, cells expressing the mutants A59T and G375R died at very low BSO concentrations similarly to mock-transduced Txnrd2−/− cells (Figure 3C). We conclude from this experiment that both mutations abolish the function of Txnrd2.
Expression of wild-type (wt) Txnrd2 and the novel mutations (A59T and G375R) in Txnrd2−/− cells. Immunoblotting using a Txnrd2-specific antibody (A) and confocal microscopy of immunocytochemical staining of transduced cells with an anti-Prx III-specific antibody (Alexa488) for mitochondria, an anti-FLAG-specific antibody (Alexa568) for reconstituted wt Txnrd2 and the A59T and G375R mutants and a nuclear counterstain with DAPI (B) revealed similar expression of wt Txnrd2 and A59T in both wild-type and knockout cells, whereas the G375R mutant was only weakly expressed inTxnrd2−/− cells. (B) Wild-type Txnrd2 and the A59T mutant predominantly localized to mitochondria [co-localization (yellow) of Prx III (green) and TAPe-tagged protein (red)] in knockout cells as shown in the false colour merge. The G375R mutant was only weakly expressed [compare (A)], but appeared also to localize to mitochondria. Shown are optical sections acquired by confocal microscopy of transduced cells stained with a nuclear DAPI counterstaining (blue), an anti-Prx III-specific antibody (Alexa488 shown in green) and an anti-FLAG-specific antibody for TAPe-tagged Txnrd2 (Alexa568 shown in red). Co-localization is indicated by a yellow colour arising from Prx III (green) and TAPe-tagged Txnrd2 (red) staining of knockout and wild-type cells as shown in the false colour merge. Mock = empty virus-transduced cells. Scale bars, 10 µm. (C) The novel mutations did not restore the function of Txnrd2 in Txnrd2−/− fibroblasts under oxidative stress. Both transduced mutants A59T and G375R failed to rescue cell death of Txnrd2−/− cells under glutathione depletion induced by increasing l-buthionine sulfoximine concentrations. Txnrd2−/− cells transduced with the mutant forms were even more sensitive to glutathione depletion than mock-transduced Txnrd2−/− cells. Cell numbers were determined by trypan blue exclusion 48 h after treatment with the indicated l-buthionine sulfoximine concentrations. Shown are results of one out of three independent experiments with similar results. Cell numbers are presented as relative to the number of mock-transfected Txnrd2+/+ cells in the absence of l-buthionine sulfoximine. Bars represent means and error bars standard deviations derived from cell counting of three different wells for each condition. P-values <0.001 were considered as highly significant and are marked with two asterisks. Values between mock-transfected Txnrd2+/+ cells (black bars) and Txnrd2−/− cells (empty bars) treated with different concentrations of l-buthionine sulfoximine were always highly significant (not shown for a simpler illustration).
Both mutants impair the function of endogenous wild-type Txnrd2
Since Txnrd2 is enzymatically active as a homodimer and since all three patients were heterozygous for the mutations, it was important to see whether the mutant allele also impacts on the function of the wt enzyme allele by a dominant-negative mechanism. To address this question, the lentiviral vectors encoding wt as well as the A59T and G375R Txnrd2 mutants were transduced into Txnrd2+/+ cells. Subcellular localization and cell survival in the presence of BSO were monitored in the same manner as described above for lentivirally transduced Txnrd2−/− cells. Although wt Txnrd2 and the A59T mutant were expressed at high level in Txnrd2+/+ cells, the mutant G375R was barely detectable by western blotting (Figure 4A) as well as immunofluorescent staining (Figure 4B). Like in Txnrd2−/− cells, wt Txnrd2 and the A59T mutant localized to mitochondria in Txnrd2+/+ cells, whereas the low expression of mutant G375R in Txnrd2+/+ cells precluded any conclusion regarding its subcellular localization (Figure 4B). Although mock-transduced and wt Txnrd2-transduced cells were resistant to GSH depletion over a wide range of BSO concentrations, cell survival of A59T- and G375R-transduced Txnrd2+/+ cells was significantly impaired in the presence of increasing BSO concentrations (Figure 4C). The fact that both mutants had a similar impact on cell survival despite different levels of expression suggests that mutant G375R is more toxic and that a non-toxic threshold of expression is selected for upon retroviral transduction. Notably, the introduction of the mutant alleles into Txnrd2+/− cells would have been closer to the in vivo situation in the patients. But, as it is impossible to adjust the expression level to that of the endogenous Txnrd2 alleles in the patients in vivo (which has been unknown anyway), it was of lesser importance whether one or two copies of the wt allele have resided in the cells. Apparently, it is more difficult to detect a dominant-negative effect in cells with two wt Txnrd2 copies left than in cells with one copy. Yet, the dominant-negative effect was already visible in an unequivocal manner in cells harbouring two wt alleles precluding the necessity of further breeding and of establishing cell lines heterozygous for Txnrd2.
Expression of wild-type (wt) Txnrd2 and the novel mutations (A59T and G375R) in Txnrd2+/+ cells. Immunoblotting using a Txnrd2-specific antibody (A) and confocal microscopy of immunocytochemical staining of transduced cells with an anti-Prx III-specific antibody (Alexa488) for mitochondria, an anti-FLAG-specific antibody (Alexa568) for reconstituted wild-type (wt) Txnrd2 and the A59T and G375R mutants and a nuclear counterstain with DAPI (B) revealed comparable expression of wt Txnrd2 and A59T in wild-type cells, whereas the G375R mutant was barely detectable in Txnrd2+/+ cells. (B) Wild-type Txnrd2 and the A59T mutant predominantly localized to mitochondria in Txnrd2+/+ cells as shown in the false colour merge. The low expression of mutant G375R in Txnrd2+/+ cells, however, precluded any conclusion regarding its subcellular localization (for details regarding staining and illustration, see Figure 3 and the Methods section). Mock = empty virus-transduced cells. Scale bars, 10 µm. (C) Cell survival of Txnrd2+/+ cells transduced with A59T or G375R was impaired in response to increasing l-buthionine sulfoximine concentrations, indicating that both forms harbour a dominant-negative function. The fact that both mutants had a similar impact on cell survival despite different levels of expression suggests that mutant G375R is more toxic and that a non-toxic threshold of expression is selected for upon retroviral transduction Bars represent means and error bars standard deviations derived from cell counting of three different wells for each condition. P-values <0.05 were considered as significant and are marked with an asterisk, and P-values <0.001 were considered as highly significant and are marked with two asterisks (for experimental details and statistical analysis, see the legend of Figure 3).
Ultrastructural analysis of cells expressing the Txnrd2 mutants A59T and G375R
We then performed ultrastructural analyses to address the questions (i) whether morphological differences in mitochondria can be observed in knockout when compared with wt fibroblasts, similarly to those observed in heart-specific Txnrd2−/− cardiomyocytes ex vivo,18 and (ii) whether the add-back of wt Txnrd2 or of the different mutants would affect mitochondrial morphology (Figure 5). We found that the size of mitochondria varied to a greater extent in knockout (Figure 5B) when compared with wt mitochondria (Figure 5A). Ultrastructural changes, ranging from swelling and loss of cristae to deterioration of matrix and mitochondrial membranes, were sometimes detectable in knockout mitochondria (Figure 5B). The add-back of wt Txnrd2 and of the different mutants, however, did not induce major changes in the overall morphology of mitochondria (Figure 5A and B). From this, we conclude that the functional parameters described in Figures 3 and 4 are more sensitive and more readily quantifiable than ultrastructural parameters.
Ultrastructural analysis of Txnrd2+/+ and Txnrd2−/− cells expressing wt Txnrd2 and the different Txnrd2 mutants. (A) Empty virus-transduced (mock), wt Txnrd2-, and A59T- and G375R-transduced Txnrd2+/+ mouse embryonic fibroblasts did not reveal major differences in mitochondrial structures. (B) Although the submitochondrial structures (cristae) were less distinct in Txnrd2−/− mouse embryonic fibroblasts compared with wild-type cells, the add-back of the different Txnrd2 variants did not grossly change the overall mitochondrial structure. The scale bar represents 500 nm.
For the first time, we describe rare mutations in human TXNRD2 in patients with DCM. Both identified mutations result in amino acid substitutions in the highly conserved and functionally essential FAD-binding domain of the enzyme. A major influence of mutations located in FAD-binding domains on enzyme activity of oxidoreductases has been repeatedly demonstrated.30–32 Functional analysis of the two identified mutants reconstructed in murine fibroblasts revealed that both forms do not rescue Txnrd2−/− cells from cell death induced by GSH depletion and that they exerted a dominant-negative effect when expressed in Txnrd2+/+ cells.
Regarding a possible causative role of the mutations for DCM, two questions had to be answered: first, are the mutations rare polymorphisms that are functionally silent or do they abolish or impair the function of Txnrd2? And second, as the mutations were heterozygous in all three patients, do they exert a dominant-negative function on the wt enzyme? Both questions raise fundamental issues regarding the enzymatic function of Txnrd2 that is shared among different cell types and tissues and should therefore be independent of the cellular model system. Apparently, the first question could only be addressed in Txnrd2−/− cells, thus precluding the use of cardiomyocytes. We reasoned that if the mutant alleles could rescue the Txnrd2−/− phenotype in a manner similar to the wt allele, the mutations would be functionally silent. Conversely, a functional significance of the mutations could be disclosed if the mutant alleles were not able to rescue the phenotype of Txnrd2−/− cells as the wt Txnrd2 gene does. The phenotype of Txnrd2−/− cells is defined as high sensitivity to GSH depletion, a phenotype that is highly sensitive and well quantifiable. The necessity to assess the function of Txnrd2 quantitatively in a homogeneous population of cells precluded the use of primary cells. Given these experimental constraints, we considered the question whether conclusions drawn from fibroblasts might be irrelevant or relevant for cardiomyocytes. At least two points suggest that a phenotype observed in fibroblasts will also be relevant for cardiomyocytes. First, cardiomyocytes are extremely dependent on energy production by mitochondria, probably more so than fibroblasts. Findings made in fibroblasts are therefore likely to impact also on cardiomyocytes. Secondly, fibroblasts express Txnrd1 as well as Txnrd2, whereas cardiomyocytes only show low expression of Txnrd133 and depend on Txnrd2 for reduction of thioredoxin.18 If there is some redundancy between the thioredoxin-1 and thioredoxin-2 system, which is suggested by the phenotype of mouse knockouts in both systems, cardiomyocytes should be functionally affected by mutations at least as severely as fibroblasts.
We did not reconstruct or functionally test any of the five non-synonymous variants identified in both patients and controls to rule out a major pathological effect by them. Structural considerations combined with evolutionary studies on 46 aligned FAD-binding oxidoreductases (Figure 2A and B; see Supplementary material online, Figure S1) including 13 Txnrd2 genes revealed that A59 and G375 are located in two helices contributing to the formation of the FAD-binding pocket. The mutations A59T and G375R are predicted not to be tolerated (SIFT; see Supplementary material online, Table S3) and to disrupt the formation of the FAD-binding pocket. G375 is conserved in all FAD-binding oxidoreductases across evolution that we have studied. In all but two oxidoreductase genes, A59 is conserved and only two Drosophila thioredoxin reductases harbour valine as another hydrophobic amino acid at this position. Both mutations A59T and G375R stand out in that they are unique across all in silico analyses performed which discriminates them from the other five variants and supports their causality (see Supplementary material online, Table S3). None of the five non-synonymous variants observed in patients and controls are conserved in evolution. Four of them are located at the surface of the molecule and are thus irrelevant for NADPH/FAD binding and enzymatic activity. The fifth I370T is located in the same helix as G375R, but the side chain points away from FAD. Most importantly, T370 is present at this position in mouse Txnrd2 used for modelling human TXNRD2, and sequence alignments have furthermore provided conclusive evidence that threonine at this position is the evolutionary ancient allele.
In summary, two mechanisms are likely to account for the observed phenotype in the human DCM patients: first, increased oxidative stress due to reduced expression of fully functional TXNRD2. Even though heterozygosity of Txnrd2 (Txnrd2+/−) or of mitochondrial thioredoxin (Txn2+/−) has no gross effect on mouse development or maturation,18,34 gene dosage may contribute to the capacity to cope with oxidative stress under challenge. Secondly, increased oxidative stress due to a dominant-negative effect of the toxic mutants impairing the function of the wt protein that is plausible because thioredoxin reductases act as homodimers.27 All three patients were heterozygous carriers of the two TXNRD2 mutations. Although having severely impaired LV function when diagnosed, they died at rather advanced age. This suggests that heterozygous carriage of the mutations has resulted in a moderate phenotype and accumulation of cardiac tissue damage during life that has long been compensated. Similar to previous observations for other already established DCM causing mutations in different genes,4–7 TXNRD2 mutations described here explain only a small fraction of overall disease burden. As TXNRD2 is a Sec-containing enzyme, the presence of selenium is essential for proper enzyme function. In this context, TXNRD2 in addition to GPX1 may represent the link between selenium deficiency and Keshan's disease, an endemic cardiomyopathy prevalent in China.18,35,36
The following limitations of our study must be acknowledged: due to the fact that mostly rare mutations in more than 20 different disease genes have been described in inherited cases of DCM, we are unable to fully exclude the presence of already known mutations in our study cohort. In addition, due to their advanced age in none of the three patients with novel amino acid residue-altering TXNRD2 mutations, parents were available for study. Thus, the question of an inherited vs. a de novo mutation and cosegregation within the family cannot be resolved in either of them. The fact that we observed only three TXNRD2 mutation carrying patients precludes to reliably describing a TXNRD2 mutant DCM subphenotype. The answer to this question will have to await the identification of further TXNRD2 mutation carrying patients in other cohorts and the present study may provide the rationale for further investigations. Experimental limitations of our study include the fact that cardiomyocytes would have been the optimal model system. Yet, there are experimental constraints regarding the use of cardiomyocytes and therefore we had to choose the MEFs as the experimental system.
Our data underline the essential role of TXNRD2 for mitochondrial redox homeostasis in cardiomyocytes and normal heart function18 and point to a novel pathophysiological mechanism for heart failure: failure of the cellular mechanisms that counterbalance the production of ROS. For the first time, we describe mutations in DCM patients in a gene involved in the regulation of cellular redox state. TXNRD2 mutations may explain a fraction of human DCM disease burden and further studies are needed to corroborate the present results.
This work has been supported by a grant from the Technische Universität München (KKF 69-04) to N.B. and A.P. as well as by grants from the German Research Foundation (DFG) Priority Programme 1087 ‘Selenoproteins’ to G.W.B. and M.C. The KORA research platform was initiated and financed by the Helmholtz Center Munich, German Research Center for Environmental Health, which is funded by the German Federal Ministry of Education and Research and by the State of Bavaria. The work of KORA is supported by the German Federal Ministry of Education and Research (BMBF) in the context of the German National Genome Research Network (NGFN-2 and NGFN-plus). Our research was also supported within the Munich Center of Health Sciences (MC Health) as part of LMUinnovativ. Additional funding was obtained by A.P. from the German National Genome Research Network NGFN 01GR0803 and the BMBF German Federal Ministry of Research BMBF 01EZ0874, T.M. German National Genome Research Network NGFN and S.K. from German National Genome Research Network NGFN BMBF 01GS0838; Leducq Foundation 07-CVD 03, LMU Excellence Initiative. The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.
Conflict of interest: none declared.
We thank Dr Antonio Miranda-Vizuete (Universidad Pablo de Olavide, Sevilla, Spain) for providing the plasmid encoding the murine Txnrd2 cDNA and Luise Jennen for excellent technical assistance. We are most grateful to Stefan Rahlfs for his suggestion to include glutathione reductases in the definition of the FAD-binding domain in TXNRD2.
. Progress with genetic cardiomyopathies: screening, counseling, and testing in dilated, hypertrophic, and arrhythmogenic right ventricular dysplasia/cardiomyopathy. Circ Heart Fail 2009;2:253-261. doi:10.1161/CIRCHEARTFAILURE.108.817346.
. Thiol-based mechanisms of the thioredoxin and glutaredoxin systems: implications for diseases in the cardiovascular system. Am J Physiol Heart Circ Physiol 2007;292:H1227-H1236. doi:10.1152/ajpheart.01162.2006.
. Regulation of human thioredoxin reductase expression and activity by 3′-untranslated region selenocysteine insertion sequence and mRNA instability elements. J Biol Chem 1999;274:25379-25385. doi:10.1074/jbc.274.36.25379.
. Mitochondrial thioredoxin-2 has a key role in determining tumor necrosis factor-alpha-induced reactive oxygen species generation, NF-kappaB activation, and apoptosis. Toxicol Sci 2006;91:643-650. doi:10.1093/toxsci/kfj175.
. Guidelines for the study of familial dilated cardiomyopathies. Collaborative Research Group of the European Human and Capital Mobility Project on Familial Dilated Cardiomyopathy. Eur Heart J 1999;20:93-102. doi:10.1053/euhj.1998.1145.
. Crystal structures of oxidized and reduced mitochondrial thioredoxin reductase provide molecular details of the reaction mechanism. Proc Natl Acad Sci USA 2005;102:15018-15023. doi:10.1073/pnas.0504218102.
. Three-dimensional structure of a mammalian thioredoxin reductase: implications for mechanism and evolution of a selenocysteine-dependent enzyme. Proc Natl Acad Sci USA 2001;98:9533-9538. doi:10.1073/pnas.171178698.
. Structure and mechanism of mammalian thioredoxin reductase: the active site is a redox-active selenolthiol/selenenylsulfide formed from the conserved cysteine-selenocysteine sequence. Proc Natl Acad Sci USA 2000;97:5854-5859. doi:10.1073/pnas.100114897.