Arabidopsis thaliana alternative dehydrogenases: a potential therapy for mitochondrial complex I deficiency? Perspectives and pitfalls

For the evaluation of the above-described therapeutic strategy on cellular models, we focused on control and CI defective human fibroblasts.

The control fibroblasts (NDHF) were purchased from Lonza (Cat. No. CC-2509). Patients’ fibroblasts were obtained from skin biopsies of patients with signed informed consent. The CI defective cell line (79787) belongs to a patient affected with Leigh syndrome carrying the homozygous frameshift mutation c.462delA (p.Lys154fs) within NDUFS4, located in 5q11 and encoding for a CI subunit close to the catalytic region of NADH-quinone oxidoreductase. The mutation is predicted to result in the synthesis of a truncated protein. Indeed, the absence of NDUFS4 protein was previously reported in fibroblasts derived from patients harbouring the same NDUFS4 homozygous mutation [21].

Skin fibroblast cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) with Glutamax +/− 4.5 g/L Glucose, supplemented with 10% fetal calf serum (FBS), 2.5 mM Pyruvate and maintained in a 5% CO2 incubator at 37 °C. Patients’ fibroblasts were obtained from skin biopsies of patients and signed informed consent. Selective growth of transfected cells was maintained by adding blasticidin 5 μg/ml to DMEM.

Control and patient fibroblasts were transfected with constructs containing the four NDH-2 genes of interest (AtNDA1, AtNDA2, AtNDB4 and ScNDI1) fused with human mitochondrial targeting signal (MTS) and a blasticidin resistance sequence (Additional file 1: Supplementary Methods). Transfection has been performed using a lentiviral vector from Invitrogen™ (ViraPower™ HiPerform™) according to Kremer and Prokisch [22]. Assessment of transduction efficacy and selection of transfected cell lines were performed using the results of qPCR (not shown) and oxygen consumption analysis (Fig. 2) as previously described [22].

Collection and permeabilization of fibroblasts were performed as previously described [23].

Spectrophotometric analysis of NADH:quinone oxidoreductase specific activity was performed on a Cary 60 spectrophotometer equipped with a18-cell holder maintained at 37 °C.

Measures of NADH:quinone oxidoreductase specific activity were performed in buffer A containing 10 mM KH₂PO₄, pH 7.2 and 1 mg/mL BSA at wavelengths of 340 nm–380 nm to assess NADH oxidation using an extinction coefficient of 4.87 as previously described [23, 24].

The sample compartment was kept open in order to allow manual stirring of the cuvette content after each addition. For K_M determination, samples (8–20 μL) were added to water, incubated for 1 min before mixing with buffer A. Rotenone (8 μM), KCN (650 μM), DCQ (50 μM) were sequentially added to the cuvettes before starting the reaction with the substrate NADH (at concentrations ranging from 0.3 to 150 μM) and following the reaction kinetics. A comparative assay was performed without rotenone in order to quantify the amount of rotenone resistant NADH:quinone oxidoreductase activity. All the measurements were performed at least in triplicates.

K_M and Vmax were estimated using an online available tool (http://www.ic50.tk/K_Mvmax.html) using the Michaelis-Menten model.

Proteins were measured according to Bradford [25].

SOD activity was measured according to Stefan L. Marklund following the described method of pyrogallol auto-oxidation inhibition. One unit of SOD inhibits 50% of pyrogallol autoxidation, measured at 420 nm [26].

Subconfluent fibroblasts (75 cm² flask) were trypsinized and the pellet was washed once with 1 mL PBS. The oxygen uptake was measured with an optic fiber equipped with an oxygen-sensitive fluorescent terminal sensor (Optode device: FireSting O₂, Bionef, Paris, France). The optic fiber was fitted to a printed cap ensuring the closure of the quartz-cell yet allowing micro-injections (0.6 mm hole diameter) for concurrent measurement of oxygen uptake with mitochondrial potential (determined by the fluorescence change of 100 nM rhodamine). Cells were added to 750 μL of buffer consisting of 0.25 M sucrose, 15 mM KCl, 30 mM KH₂PO₄, 5 mM MgCl₂, 1 mM EGTA, pH 7.4, followed by the addition of rhodamine (100 nM), BSA 1 mg /ml and 0.01% w/v digitonin. The permeabilized cells were successively added followed by the addition of mitochondrial substrates (6.25 mM glutamate/malate or 6.25 mM succinate) and two consecutive injections of ADP (40 nmol each) to ensure state 3 (phosphorylating) conditions or ATP (40 nmol) in order to estimate ATP recycling due to ATPases activity. The reaction was followed until state 4 (the respiratory rate after all the ADP has been phosphorylated to form ATP) was reached back and maintained. Respiratory rates during state 3 and state 4 were estimated as the speed of oxygen consumption (nmol/min) adjusted to protein concentration (μg) in each cuvette. Respiratory control index was later calculated as the ratio between state 3 to state 4 respiratory rates. P/O values (corresponding to the number of ATP molecules produced for every oxygen atom consumed) were also measured as the ratio between concentration (in nmol) of ADP (or ATP) added to the cuvette and the amount of oxygen atoms (nmol of molecular oxygen*2) consumed during state 3 to state 4 transition. All the assays were repeated at least three times. Protein content was measured according to Bradford [25].

RNA sequencing was performed as described [27]. Briefly, RNA was isolated from whole-cell lysates using the AllPrep RNA Kit (Qiagen) and RNA integrity number (RIN) was determined with the Agilent 2100 BioaAnalyzer (RNA 6000 Nano Kit, Agilent). For library preparation, 1 μg of RNA was poly(A)- selected, fragmented and reverse transcribed with the Elute, Prime and Fragment Mix (Illumina). End repair, A-tailing, adaptor ligation and library enrichment were performed as described in the Low Throughput protocol of the TruSeq Stranded mRNA Sample Prep Guide (Illumina). RNAcDNA libraries were assessed for quality and quantity with the Agilent 2100 BioaAnalyzer and quantity using the Quant-iT PicoGreen dsDNA Assay Kit (Life Technologies). RNA libraries were sequenced as 150 bp paired-end runs on an Illumina HiSeq4000 platform. The STAR aligner* (v 2.4.2a) with modified parameter settings (−-twopassMode = Basic) was used for split-read alignment against the human genome assembly hg19 (GRCh37) and UCSC knownGene annotation. Prior alignment, the reference genome sequence was augmented by two new contigs, one for each plant gene (NDA2 and NDB4, respectively). The nucleotide sequences of these two genes corresponded to the transgenic constructs cloned in the lentiviral vector (see Additional file 1: Supplementary Methods). To quantify the number of reads mapping to annotated genes we used HTseq-count (v0.6.0). FPKM (Fragments Per Kilobase of transcript per Million fragments mapped) values were calculated using custom scripts.

Preliminary assays on different CI defective fibroblast cell lines had showed the ability of several NDH-2 to rescue respiration defect (Fig. 2b-d). We decided to focus our subsequent analysis on three NDH-2: ScNDI1, the internal NDH-2 of Saccharomyces cerevisiae; AtNDB4, an Arabidopsis thaliana NDH-2 localized to the external side of inner mitochondrial membrane (IMM); AtNDA2, another Arabidopsis thaliana NDH-2 localized to the internal side of IMM.

As expected, transfection of control fibroblasts with AtNDA2, AtNDB4, and ScNDI1 led to rotenone-resistance without any significant effect on overall respiration rate (Fig. 2a). In order to examine the rescuing efficiency of plant NDH-2 in more depth, we chose to focus on fibroblasts carrying a pathogenic homozygous mutation in the nuclear gene NDUFS4, as a well-established cellular model of complex I deficiency. Indeed, consequences of deleterious mutations affecting NDUFS4 have been thoroughly studied on several patient cell lines and on whole-body and tissue specific knockout mice [28].

Therefore, we verified and confirmed that besides conferring rotenone resistance, all the aforementioned NDH-2 (ScNDI1, AtNDA2 and AtNDB4) were able to restore respiration when expressed in NDUFS4 deficient fibroblasts, nearly reaching control levels. (Fig. 2c).

ScNDI1, AtNDA2, and AtNDB4 transfected control fibroblasts (NHDF) and CI-defective fibroblasts (NDUFS4) exhibited comparable growth rates when compared to the corresponding untransfected control, both in glucose (4.5 g/L) and glucose-deprived media (not shown).

We further confirmed the observed rescue by measuring NADH:quinone oxidoreductase specific activity by spectrophotometry in control cells and NDUFS4 mutated fibroblasts before and after transfection by ScNDI1, AtNDA2, and AtNDB4, show that all three dehydrogenases were able to rescue the CI defect (Table 1). We could also observe that, while AtNDA2 and AtNDB4 restored CI activity to levels comparable to those observed in control cells, ScNDI1 transfected cells exhibited levels of NADH:quinone oxidoreductase activity much higher than untransfected cells (Table 1).

As a direct effect of defective CI activity and consequent increase of ROS production, SOD activity is shown to be significantly higher in NDUFS4 mutated patient’s cell (Fig. 3). Transfection with AtNDA2 and AtNDB4 but not with ScNDI1 was almost able to decrease SOD activity to levels observed in control fibroblasts (Fig. 3).

We evaluated by RNA sequencing the expression levels of CI subunits, AtNDA2, and AtNDB4 in control cells before and after the stable transduction with the plant genes AtNDA2 and AtNDB4 (RNA sequencing was not performed on NDUFS4 deficient fibroblasts due to scarcity of material). We evaluated the FPKM values for CI subunits before and after transduction. The median FPKM of CI subunits was similar in all cell lines, indicating that the transduction with the plant genes did not affect the expression level of CI subunits (median FPKM between 30 and 35, Table 2). AtNDA2 had an expression level of 25 FPKM, which falls in the range of CI subunit expression, while AtNDB4 had an expression level of 127 FPKM, much higher than the median expression level of CI subunits (Table 2). In A. thaliana the endogenous expression of NDA2 and NDB4 is significantly lower than the expression of CI subunits, in all parts of the plant (flower, root, leaf, and fruit). The expression of NDA2 is 10 times lower than the median expression of CI subunits, while NDB4 is almost 500 times lower than the median expression of CI subunits [29] (Additional file 1: Table S1).

Thereafter, in view of its expression profile in A. thaliana (Add file 1), we selected the AtNDA2 to verify the lack of competition between CI and plant NDH-2 when expressed in control fibroblasts. We first studied the P/O ratio with different substrates, the latter being supposedly decreased if NADH, normally oxidized by the proton-motive CI, is diverted to AtNDA2. P/O calculation is usually performed on isolated mitochondria in order to eliminate cytosolic ATPases activity. ATPases increase ADP recycling, allowing for a continuous stimulation of the mitochondrial ATP synthase and respiration, thus affecting state 4 establishment. However, taking into consideration the scarcity of material and the slow growth rate of fibroblasts, we performed the assays using permeabilized cells. As expected, the observed P/O values were underestimated comparing to the ones measured on purified mitochondria (about 2.5 for NADH related substrates and 1.5 for succinate, respectively)-see Hinkle et al. [30] for a complete review on this topic. Nevertheless, using this approach, we were able to measure the P/O ratio (Fig. 4). Unexpectedly, we evidenced a competition between AtNDA2 and functional mitochondrial respiratory chain CI during glutamate/malate oxidation in the control cell line expressing AtNDA2 (Fig. 4). Transfected cells showed P/O values lowered by half (0.43 ± 0.08) in comparison to untransfected cells (0.9 ± 0.1). Moreover, respiratory control index, calculated as the ratio between state 3 and state 4, which represents a numeric estimation of mitochondrial coupling efficiency, was also clearly lowered in transfected cells under glutamate/malate stimulation (Fig. 5).

A further validation of these results came from P/O data obtained with succinate, the substrate of CII. Indeed, although P/O ratios for succinate were not significantly lower in transfected cells, 0.43 ± 0.05 and 0.39 ± 0.02, respectively, we observed a reduction of respiratory control index in transfected cells (Fig. 5). This is likely due to the metabolic conversion of a fraction of succinate to glutamate, which later enters the oxidation machinery proceeding through CI and AtNDA2.

When ATP instead of ADP was used, we observed only a very low OXPHOS stimulation in both transfected and untransfected cell lines, providing an additional confirmation that our measurments were not significantly affected by ATPase mediated ATP recycling to ADP (not shown).

A further evidence of the likely competition between CI and AtNDA2, came from calculation of K_M for NADH in both transfected and untransfected cell lines (Fig. 6). CI affinity for NADH was estimated in untransfected control cells considering only rotenone sensitive internal NADH:quinone oxidoreductase activity, while, to evaluate AtNDA2 affinity for NADH, we exclusively analyzed rotenone insensitive activity in AtNDA2-transfected control cells. Our estimation of CI and AtNDA2 K_M gave values of 2.7 ± 0.4 μM and 9.7 ± 3.3 μM, respectively. Hence, when transfected in human cells, K_M of AtNDA2 for NADH appear to be about only 3 folds higher than K_M of CI for NADH, i.e. in the same order of magnitude; this gap is probably insufficient to prevent competition for the substrate within physiological range of NADH concentration inside the mitochondria, therefore indirectly confirming that upon these experimental conditions, it seems likely that CI and AtNDA2 compete for the oxidation of NADH.