Stem tissue of the CLS mutant carrying a T-DNA in the CLS gene and a complementing gene copy under an estradiol inducible promotor (Katayama and Wada, 2012) was used to establish cell cultures providing suitable amounts of plant material for organelle preparations. The site of T-DNA insertion is at the start of the fifth exon of the CLS gene (SALK_4984, At4g04870, Katayama and Wada, 2012; Pan et al., 2014). After 7 days of growth cell mass of the homozygous knock-out mutant approximately doubled while the WT cell line increased by factor three (data not shown). Complementation of the cls knock-out was induced by adding estradiol to fresh MS-medium to a final concentration of 4 μM prior to subculturing. Organelle preparations were performed after 6 or 7 days of growth.

To monitor the success of the T-DNA insertion as well as the complementation strategy in the cell culture system, targeted selected ion monitoring mass spectrometry (tSIM-MS) was performed for two selected tryptic peptides of CLS in isolated mitochondria of the knock-out mutant (cls), complemented mutant (cls/CLS), wild type (WT), and estradiol treated WT cell cultures (WT+). The peptides (LLQSATPLHWR and DLLHPGLVGIVLLR) were chosen for quantitation due to their traceability in shotgun MS/MS runs, their uniqueness within the Arabidopsis proteome, and the low amount of expected modifications due to the absence of cysteines and methionines. The mono-isotopic mass and the following two masses of the isotope distribution of triply charged peptides were taken for this analysis (Figure 1). Results indicate comparable amounts of CLS in WT and WT+ but higher average values for cls/CLS, concordant with a constitutive expression of the CLS gene in the presence of estradiol. The high level of variation observed for cls/CLS is striking and most likely the result of experimental deviations, such as age of estradiol solution. Combined abundances of both peptides reveal CLS abundance in cls not changing significantly. However, closer inspection reveals the peptide DLLHPGLVGIVLLR to be virtually absent in cls. This peptide is encoded by the end of exon 4 and the start of exon 5 (Figure 1, Katayama and Wada, 2012), spanning the insertion site proposed by Pan et al. (2014). The relatively high amounts of LLQSATPLHWR in cls in combination with the absence of DLLHPGLVGIVLLR suggest the synthesis of a truncated and therefore inactive form of the CLS protein in cls.

Cardiolipin deficiency was reported to have a strong impact on the respiratory capacity of yeast mitochondria but mitochondrial matrix metabolism may be affected as well via feed-back effects. Indeed, altered amounts of TCA-cycle intermediates under CL-deficiency were reported previously (Pineau et al., 2013). Photometric assays for seven TCA-cycle enzymes as well as the pyruvate dehydrogenase complex (PDC) and mitochondrial NAD malic enzyme (NAD-ME) were conducted. Most TCA-cycle enzymes show decreased (2-oxglutarate dehydrogenase complex, OGDC; mitochondrial malate dehydrogenase, mMDH) or severely decreased (PDC; citrate synthase, CS; isocitrate dehydrogenase, ICDH; fumarase, Fum, NAD-ME) capacities in cls (Figure 2). Hence, all NADH-forming enzymes possess lower capacities in the mutant. Interestingly, aconitase is the only TCA cycle enzyme with increased capacity in the mutant. This enzyme has been shown to be a major regulator of the TCA cycle and its increased capacity may therefore be aimed at increasing flux through this essential pathway (Araujo et al., 2012). Equally interesting is the response of OGDC to estradiol. In both, cls/CLS and WT+, capacities of these enzymes are increased considerably compared to the corresponding untreated lines. For mammals, estrogen has been reported to indirectly alter the phosphorylation pattern of OGDC, thereby impacting its function (Lagranha et al., 2010). If this also holds true for OGDC of plant mitochondria is currently unknown but the data presented here seem to support the existence of a similar mechanism. In summary, most TCA cycle components and enzymes associated with this pathway show reduced enzymatic capacities in the mutant. As such, mitochondrial capacity for the production of NADH is reduced under CL deficiency.

Analyses of the respiratory capacities of isolated mitochondria by oxygen electrode assays are preferentially conducted using substrates aimed at the internal production of NADH in intact organelles. However, lowered NADH-production within the mutant is expected to interfere with O₂-consumption rates in the mutant when using such substrates. Hence, the alternative approach of feeding gently disrupted mitochondria with externally applied NADH was chosen to ensure investigation of the true enzymatic capacities of the mitochondrial electron transfer chain (mETC) in cls, cls/CLS, and WT. Using osmotic disruption, gently lysed organelles isolated from cell cultures of cls, cls/CLS, WT, and WT+ were investigated by means of a Clark type oxygen electrode (Figure 3). Using the electron donor NADH the electron transfer capacity of cls was found to be less than 20% of its WT counterpart (5.6 nmol × min^-1 × mg_Prot^-1/37.1 nmol × min^-1 × mg_Prot^-1). In contrast, cls/CLS showed values intermediate to those of cls and the WT (16.5 nmol × min^-1 × mg_Prot^-1), indicating a partial recovery of respiratory capacity following the estradiol-triggered complementation of the CLS knock-out. However, given the high abundance of CLS in this line (Figure 1) respiratory capacity is unexpectedly low; suggesting that estradiol itself may negatively impact respiratory capacity. Treatment of the WT cell line with this hormone (WT+) also leads to reduced respiratory capacities (26.2 nmol × min^-1 × mg_Prot^-1) compared to untreated WT, thereby supporting the notion of estradiol inhibiting respiration. Indeed, estradiol was previously reported to have detrimental effects on mitochondrial transporters and respiratory chain components in rat liver mitoplasts (Thiede et al., 2012). Hence, respiration rates determined for the estradiol treated lines most likely underestimate their true enzymatic capacities.

Densitometric quantitation of digitonin solubilized I_1, III₂, and the supercomplex consisting of these two individual protein complexes (I₁III₂, Figures 4 and 5) on 1D BN gels of digitonin treated cls, cls/CLS, WT, and WT+ mitochondria were performed. In cls this supercomplex is decreased to approximately 20% of the WT level while abundances of individual complexes I₁ and III₂ are less affected (Figure 5). Unfortunately, II₁ and IV₁ could not be quantified accurately from 1D BN gels as mass spectrometry (MS) revealed strong background of other proteins in the regions of their respective bands on the gel (Figure 6, Supplementary Table 1). To allow quantitation of II₁ without the risk of including non-II₁ proteins, spot volumes of the majority of its subunits were obtained from 2D BN/sodium dodecyl sulfate (SDS) gels of WT and cls using the Delta2D gel analysis software (Decodon, Figure 7A, Supplementary Table 2). Background subtracted volumes of those II₁ spots detected in six gels (3x WT, 3x cls) were then summed up for an estimation of complex abundance and used for the calculation of relative complex abundance. Interestingly, II₁ does not follow the trend observed for I₁ and III₂. On average, abundances of its subunits detectable on BN/SDS gels are higher in cls and no subunit of this complex was detected showing a significantly reduced abundance in cls. In combination, this may seem indicative of a higher abundance of the assembled complex but statistical analysis does not support this notion since multiplicity adjusted p-values equals one suggesting an unchanged abundance (Table 1). To further analyze II₁ abundance in the four lines immunoblots employing an antibody directed against the flavoprotein subunit of II₁ (SDH1-1, AT5G66760, Peters et al., 2012) were produced using cell extracts of cls, cls/CLS, WT, and WT+ cell cultures (Figure 7B). As can be seen on the Coomassie stained control gel SDH1-1 signals are higher in the cls and cls/CLS lines but this may be due to higher protein loads of these lines. In summary, the abundance of this complex is best described as being unchanged in cls and cls/CLS. It follows that the observed reduced enzymatic capacity of this complex in cls cannot be attributed to a decreased abundance. Using BN/SDS analyses for I₁, III₂, and I₁III₂ the changes in protein amount largely match the values obtained from 1D BN gels (Table 1). However, multiplicity adjusted p-values for I₁ and III₂ are notably higher than 0.05. Interestingly, the proportions of the proteins showing a reduction in cls for I₁ and III₂ are 0.55 and 0.62, respectively. This can most likely be attributed to a compromised spot detection and spot matching by the analysis software which was originally designed for the true round protein spots found on 2D isoelectric focusing (IEF)/SDS gels. Gel spots on BN/SDS differ considerably in shape since they have a rather ellipsoid shape and this compromises spot detection. Particularly I₁, which is characterized by a rather diffuse band in 1D BN gels and has matching oval spots for its subunits on BN/SDS gels is affected by this. Hence, multiplicity adjusted p-values are worse for this complex. No quantitation could be obtained for IV₁ since this enzyme complex migrates in more than one band on BN/SDS gels and partly overlaps with the TOM-complex (Millar et al., 2004). Its subunits are therefore not confidently detected on 2D BN/SDS gels under the conditions applied here, preventing a meaningful comparative analysis. With the exception of II₁, data obtained from 1D BN gels are therefore used preferentially over those originating from 2D BN/SDS gels.

In order to gain information on the enzymatic capacities of respiratory complexes in gel activity assays for digitonin solubilized I₁, II₁, and IV₁ separated on 1D BN gels were performed (Figure 8A). All three complexes display reduced staining intensities in cls. Additionally, II₁ and IV₁ are characterized by higher electrophoretic mobilities in cls. This hints toward lower molecular masses of these complexes. Solubilization of mitochondrial membrane protein complexes with DDM, a stronger detergent than digitonin, reduces mutant protein complex abundances to a higher degree than in the digitonin treated samples (Figure 8B). In gel activity assays of DDM solubilized protein complexes only worked satisfactorily for I₁ and residual I₁III₂. Here too, reductions are more apparent than in the digitonin solubilized samples. CL deficiency in combination with harsher solubilization therefore suggest a reduced stability of respiratory complexes in the absence of CL.

For a more quantitative assessment of the effect of CL-deficiency on respiratory enzymes their capacities in cls, cls/CLS, WT, and WT+ were deduced from photometric assays (Birch-Machin et al., 1994; Zhou et al., 2003, Figure 5). In cls, all four assays revealed reduced enzymatic capacities by 40–60% in comparison to the WT. For complexes I₁ and III₂ reductions closely match the changes in protein complex abundances when intensities of the supercomplex bands (I₁III₂) are split according to the relation of molecular masses of I₁ and III₂ (1000 kD: 500 kD = ²/₃: ¹/₃) and added to the abundances of the individual complexes I₁ and III₂ (Figure 5). Data on IV₁ abundance are not available and no comment can be made on the correlation of enzyme capacity and protein complex abundance. Interestingly, II₁ activity is also reduced considerably in cls despite the unaltered abundance of this enzyme.

Considering that the transfer of electrons along the mitochondrial respiratory chain is a sequential process and that each of the individual enzymes (protein complexes, supercomplexes) may act as bottlenecks reductions in enzymatic activity of individual complexes do not sufficiently explain the mutant’s strong reduction in respiratory capacity. The electron flow in the mutant’s respiratory chain must therefore face additional restrictions. The I₁III₂ supercomplex in the mutant accounts to only 20% of that of the WT (Figure 5) and this reduction could have an effect on respiratory capacity.

In addition to the membrane embedded protein complexes of the respiratory chain, the abundance of the mobile electron carrier cytochrome c was analyzed by immunoblotting (Figure 9, Welchen et al., 2012). Cytochrome c is reduced to approximately 15% of WT level in cls which closely matches the reduction in respiratory capacity. The cytochrome c contents of cls/CLS and WT+ are close to WT level, which is not exactly mirrored by their respective respiration rates. However, as mentioned earlier, respiration of cls/Cls and WT+ is potentially negatively affected by the estradiol added to the cell culture medium.

For complexes I₁ and III₂, reductions of enzymatic capacities closely follow decreased abundances of these complexes on BN-gels in the mutant, indicative of unchanged enzymatic properties of these complexes under CL deficiency. Most likely, CL can be replaced by other lipids in these complexes under the applied growth conditions without negatively affecting enzymatic capacities. Considering the use of NADH in the I₁ assay, it should be noted that this also serves as substrate for non-energy conserving alternative NAD(P)H dehydrogenases and that the capacity for NADH oxidation deduced from this assay therefore is not exclusively attributable to singular or supercomplex bound complex I. However, since the assay buffer did not contain calcium which stimulates activity of most NAD(P)H dehydrogenases (Rasmusson et al., 2004) the contribution of alternative dehydrogenases toward the assay results is judged to be negligible. The most abundant plant mitochondrial supercomplex, I₁III₂ is of decreased abundance in cls and this result is in concordance with previously published data (Pineau et al., 2013). Interestingly, the loss of I₁III₂ extends far beyond the reductions observed for the complexes I₁ and III₂. In baker’s yeast (which lacks I₁) abundances of respiratory supercomplexes III₂IV₁ and, particularly, III₂IV₂ were reduced in a CLS knock-out mutant (Zhang et al., 2002) emphasizing the importance of CL for the abundance of respiratory supercomplexes. While complex IV containing supercomplexes were reported for potato and spinach mitochondria, they are unknown for Arabidopsis (Eubel et al., 2004; Krause et al., 2004; Chaban et al., 2014). As such, a direct transfer of results from yeast to Arabidopsis is difficult. In combination with the data shown here it nevertheless suggests that CL is more important for the association of respiratory complexes than for the abundance of their building blocks, the individual protein complexes.

It is currently not clear if the lowered amounts of I₁III₂ supercomplex in cls negatively impact mitochondrial respiration. A reduced level of substrate channeling in the mutant would be the most obvious explanation. However, given that substrate channeling requires a separate, supercomplex bound ubiquinone pool whose existence remains doubtful in the light of recent findings (summarized by Acin-Perez and Enriquez, 2014), it seems unlikely that the association of I₁ and III₂ is accountable for the mutant’s lowered respiratory capacity. Hence, other alterations in the CL deficient mETC are expected to be responsible for this, particularly the reduction of mitochondrial cytochrome c content.

The loss of cytochrome c in the mutant is expected to be the result of missing cytochrome c: CL interactions tying the protein to the inner membrane (Kagan et al., 2009). Some of the changes observed for mitochondrial protein complexes and supercomplexes in the mutant may actually not be caused primarily by the loss of CL but by secondary effects. Reduced cytochrome c content was reported to affect the abundance of IV₁ but to have little or no implications for I₁III₂, I₁ or III₂ (Welchen et al., 2012). Quantities of IV₁ could not be established confidently in this study but have been described to remain unchanged by CL deficiency (Pineau et al., 2013). However, IV₁ enzymatic capacity is markedly decreased in cls. In contrast to IV₁, II₁ was reported to be slightly increased in cytochrome c knock-down plants and average II₁ amounts are also higher in cls but variance is too high to confirm this (Figure 5).

Once mitochondrial cytochrome c content is reduced to a degree which negatively affects respiration and, consequently, the transport of protons across the inner membrane, the pH of the crista lumina increases. As a result, the binding strength of cytochrome c to anionic phospholipids via exposed lysine residues of the protein decreases. This stabilizes its tertiary structure and lowers the peroxidative capacity of cytochrome c (Rytömaa et al., 1992; Kagan et al., 2009). In turn, cytochrome c, when released from mitochondria, also serves as an antioxidant. The observed reduction in cytochrome c content in cls therefore also affects the mitochondrial capacity for scavenging and reducing the production of reactive oxygen species (ROS; Skulachev, 1998; Kagan et al., 2009). Moreover, cytochrome c also is an electron acceptor in the final step of ascorbate synthesis in mitochondria via the L-galactono-1,4-lacton dehydrogenase (GLDH, Schertl et al., 2012; Schimmeyer et al., 2016). Reduction in cytochrome c levels as observed for the mutant may therefore negatively impact the mutant’s ability to detoxify mitochondrial ROS.

Interestingly, and in contrast to all other respiratory chain components II₁ abundance does not alter between cls and WT while its enzymatic capacity is reduced by approximately the same degree seen in I₁ and III₂ (Figure 5). This connotes a reduction in the activity of each enzymatic unit. Since the flavoprotein subunit (SDH1) depends on activation by ATP (Huang et al., 2010) whose production is compromized in cls, in vivo activity of mutant II₁ may be reduced even further in the mutant. CL was found to stabilize yeast II₁ by strengthening the interaction of the membrane hetero-dimer with the catalytic heterodimer but stability of II₁ is not affected in Arabidopsis. One possible explanation for this may originate from the different solubilization approaches between the two studies: while DDM was chosen for the yeast study the milder digitonin was chosen for protein solubilization here. As suggested for III₂ DDM may remove potential CL substituting lipids more efficiently than digitonin, thereby decreasing stability of the complexes to a higher degree. Another reason for the stability of II₁ in Arabidopsis may be the higher number of subunits (eight in Arabidopsis compared to four in yeast; Eubel et al., 2003; Millar et al., 2004), which may provide a stronger connection between the membrane anchor and the catalytic subunits.

Reduced NADH oxidation capacity via the respiratory chain must be compensated by deceleration of its production in the mitochondrial matrix in order to limit ROS-production and prevent damage to the organelle and the cell as a whole. This, in turn, requires some sort of feedback inhibition of matrix enzymes. Since in heterotrophic cell culture mitochondria the TCA cycle is understood to operate largely in the cyclic mode, this would require regulation of PDC, ICDH, OGDC, and mMDH, (Lee et al., 2011). Enzymatic assays for these enzymes conducted during the course of this study strongly support this notion. Together with the respiration data this implies a general reduction of main mitochondrial functions in the mutant but further studies will be required to complete the picture of changes induced in plant mitochondria in response to a lack of CL.

A previous study using the same CLS knock-out line revealed similar results in respect to I₁III₂ abundance (Pineau et al., 2013). However, I₁ abundance in the mutant increased when leaf tissue was treated with digitonin thus contradicting its reduction in cls cell culture mitochondria. Interestingly, leaf and cell culture mitochondria display similar effects on I₁ abundances upon solubilization with DDM since in both tissues, abundances as well as in gel activities of I₁ are higher in the WT line. These tissue specific differences imply a lower stability of I₁ in the cell culture upon digitonin solubilization, potentially caused by differing lipid compositions of the inner mitochondrial membranes resulting in differences of the types and amounts of CL substituting lipids.

Seedlings of cls-2/cls-2 harboring pER8:CLS plants (here referred to as cls, Katayama and Wada, 2012) grown on Murashige and Skoog medium (Murashige and Skoog, 1962) containing 1% (w/v) sucrose and 3% (w/v) phytagel (Sigma–Aldrich, P8169) under continuous light (30 μmol photons m^-2 s^-1) at 23^oC for 1–2 weeks were used for the production of callus materials. When seedlings were large enough to handle, their stems were cut into small pieces, transferred to callus inducing medium [3.1 g L^-1 Gamborg’s B-5 basal salt mixture (Gamborg et al., 1968; Sigma–Aldrich, G5893), 2% [w/v] glucose, 0.5 g L^-1 MES pH 5.7 (KOH), 0.1 g L^-1 myo-inositol, 20 mg L^-1 thiamine-HCl, 1 mg L^-1 pyridoxine-HCl, 1 mg L^-1 nicotinic acid, 0.5 mg L^-1 2,4-dichlorophenoxyacetic acid, 0.05 mg L^-1 kinetin, 1 mg L^-1 biotin, 0.5% [w/v] phytagel (Sigma-Aldrich, P8169)], and incubated at 23°C to generate callus. Calli were then transferred to liquid medium and maintained as described previously (Van Gestel and Verbelen, 2002). Expression of the transgenic copy of CL synthase gene was induced by addition of 17 β-estradiol to fresh MS medium to a final concentration of 4 μM. Cells for mitochondrial preparations were harvested after 6–7 days of sub-culturing.