Different Types Domains are Present in Complex I from Immature Seeds and of CA Adult Plants in Arabidopsis thaliana

Abstract

Mitochondrial Nicotinamide adenine dinucleotide (NADH) dehydrogenase complex is the first complex of the mitochondrial electron transfer chain. In plants and in a variety of eukaryotes except Opisthokonta, complex I (CI) contains an extra spherical domain called carbonic anhydrase (CA) domain. This domain is thought to be composed of trimers of gamma type CA and CA-like subunits. In Arabidopsis, the CA gene family contains five members (CA1, CA2, CA3, CAL1 and CAL2). The CA domain appears to be crucial for CI assembly and is essential for normal embryogenesis. As CA and CA-like proteins are arranged in trimers to form the CA domain, it is possible for the complex to adopt different arrangements that might be tissue-specific or have specialized functions. In this work, we show that the proportion of specific CI changes in a tissue-specific manner. In immature seeds, CI assembly may be indistinctly dependent on CA1, CA2 or CA3. However, in adult plant tissues (or tissues derived from stem cells, as cell cultures), CA2-dependent CI is clearly the most abundant. This difference might account for specific physiological functions. We present evidence suggesting that CA3 does not interact with any other CA family member. As CA3 was found to interact with CI FRO1 (NDUFS4) subunit, which is located in the matrix arm, this suggests a role for CA3 in assembly and stability of CI.

Accession numbers: Sequence data from this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: Arabidopsis CA1 (At1g19580), Arabidopsis CA2 (At1g47260), Arabidopsis CA3 (At5g66510), Arabidopsis FRO1 (At5g67590) and mitochondrial Arabidopsis ND6 (AtMg00270) and ND9 (AtMg00070).

Introduction

Mitochondrial complex I (CI also referred as Nicotinamide adenine dinucleotide (NADH) dehydrogenase or NADH: UbiquinoneOxidoreductase, EC 1.6.5.3) is the main entrance of electrons from metabolism to the mitochondrial electron transfer chain in the oxidative phosphorylation (OXPHOS) system. In all organisms, CI conforms an L-shaped structure with a peripheral arm towards the mitochondrial matrix and a membrane arm. In bacteria, the CI contains 14 subunits, called the core, which is conserved in all other organisms (Cardol 2011, Baradaran et�al. 2013, Berrisford et�al. 2016, Subrahmanian et�al. 2016). However, other eukaryotic organisms contain a variable number of accessory subunits (ranging from 30 to 36), some of them involved in the assembly process of the complex (Stroud et�al. 2016).

In plants and in a variety of eukaryotes except for Opisthokonta, CI contains an extra spherical domain named carbonic anhydrase (CA) domain, composed of putative gamma type CA subunits that are thought to be arranged in trimers (Dudkina et�al. 2005). In Arabidopsis, these subunits are CA1, CA2 and CA3, which contain almost all the aminoacids of the active site conserved respect to the active gamma type CA from Methanosarcina thermophila (Parisi et�al. 2004). In addition, other two subunits, CAL1 and CAL2 (for CA like) show less conservation (Perales et�al. 2004). In all plants analyzed so far both CAs and CALs are present in their genomes (Perales et�al. 2004, Peters et�al. 2008, Klodmann et�al. 2010).

The CA domain has been implicated in the assembly of CI (Perales et�al. 2005) which was experimentally demonstrated by biochemical decortications of the complex (Klodmann et�al. 2010) or by subcomplexes accumulation in CI mutants (Meyer et�al. 2011). Recently, it was shown that the CA domain represents the initial step of CI assembly (Ligas et�al. 2018). Overexpression of CA2 results in anther indehiscence, which has been related to low accumulation of reactive oxygen species (ROS) (Villarreal et�al. 2009). Even when the CA activity could not be demonstrated so far, bicarbonate/CO₂ binding was experimentally shown for recombinant CA2 (Martin et�al. 2009). Also, the CA domain has been associated with carbon recycling in the context of photorespiration. Mutants lacking CA2 and one of the CAL proteins show a photorespiratory phenotype (Soto et�al. 2015). Recently, the CA domain has been also linked to plant embryogenesis (C�rdoba et�al. 2016, Fromm et�al. 2016a, Fromm et�al. 2016b, Fromm et�al. 2016c, Ostersetzer-Biran 2016). Plants lacking two CA subunits (CA1 and CA2) exhibit embryo developmental defects although shrunken seeds are rescued in sucrose supplemented media. Unexpectedly, the fifth protein, CA3, had not been found in the 85 kDa CA domain (Klodmann et�al. 2010), although a recent work reported the presence of this subunit within the assembly intermediate of 85 kDa through new complexome profiling techniques (Ligas et�al. 2018).

In this work, we show that ca1ca3 double mutants are embryo or seedling lethal as the ca1ca2 knockout lines because CI is not assembled. Plants can be rescued using sucrose supplemented media, but they are severely affected. On the other hand, ca2ca3 double knockout mutants are able to complete embryogenesis and germinate normally, although growth of these mutant plants is strongly compromised. They show a severe dwarf phenotype that cannot be completely rescued by high CO₂ concentration or sucrose supplementation, suggesting problems in respiration. Indeed, CI activity is very low. Since CA1 and CA2 do not interact, a model involving a CA1 dependent CI and a CA2 dependent CI was suggested where CA1-CA1-CAL trimers or CA2-CA2-CAL trimers allow the assembly of the entire complex (C�rdoba et�al. 2016). In this work, we show evidence that each type of CI is differently present depending on the plant life cycle stage/tissue. Thus, we propose that CA1 and CA2 dependent complexes I are indistinctly assembled in immature embryos, while CA2-dependent CI is clearly the most abundant in leaves. CA3 seems to be a constitutively component, since ca3 mutant plants have lower NADH dehydrogenase (DH) activity than Col-0 in all stages analyzed. However, CA3 does not interact with any of the other four members of the family. In light of a recent publication where CA3 was found as a component of the 85 kDa intermediate (Ligas et�al. 2018), we propose that this subunit is possibly forming homotrimers and also contribute to the assembly of the complex, particularly in early embryos where the expression of the CA3 gene is increased. In addition, CA3 may interact with FRO1 (NDUFS4), a subunit present in the interface of the membrane and matrix arms. The fro1 mutant lines show a dwarf phenotype and are able to assemble only the membrane arm of CI with no NADH DH activity (K�hn et�al. 2015). Double mutant ca3fro1 shows a significant increase in the number of unviable seeds from 5% in the single mutant to 20% in the double mutant. Altogether, our results suggest that CA3 might contribute to the assembly and stability of CI.

Results

Double ca2ca3 mutants show low growth rate while double ca1ca3 mutants are embryo/seedling lethal

The first described CA mutant with biochemical alterations was a T-DNA insertional mutant in the CA2 gene of Arabidopsis thaliana showing a strong reduction of CI (Perales et�al. 2005). However, any other single mutant affecting each one of the rest of the CA/CAL proteins do not show any altered phenotype in comparison to Col-0 plants, suggesting at least partial functional redundancy. Thus, with the aim to elucidate the physiological role of the CA domain, we thus performed crosses involving ca3 mutants and other ca mutants in order to study its physiological role.

Double homozygous ca2ca3 mutants show normal embryo development and germinate within a normal time frame. Nevertheless, plants are significantly smaller than Col-0 (Fig.�1A). Under a 12/12 L/D photoperiod, ca2ca3 mutant plants flowered in average 9 d later than Col-0 (Fig.�1B), but showing the same number of rosette leaves at bolting. This phenotype could not be completely rescued by cultivation of plants under 2000 ppm CO₂ atmosphere or in MS medium supplemented with 3% w/v sucrose (Supplementary Fig. S2).

Fig. 1.

ca2ca3 double null mutant shows lower growth rate than Col-0 plants. Col-0 and ca2ca3 double mutant plants were cultivated under 12/12 h L/D light conditions. (A) Double mutant plants are, in average, smaller than Col-0 43 d after germination. (B) ca2ca3 shows lower growth rate and extended life cycle. Flowering takes 9 d more than in Col-0, but with the same number of leaves.

On the other hand, using the alleles described in M & M, double homozygous ca1ca3 mutants could not be found in the progeny of selfed heterozygous CA1ca1ca3ca3 plants. As this suggests a lethal combination, a silique analysis was performed. Indeed, siliques from CA1ca1ca3ca3 plants contain a high proportion of small, white, immature seeds that shrunk as the siliques matured (Fig.�2A). The proportion detected was 24.95% in still green siliques and 19.74% in matured, dry siliques (Supplementary Fig. S3), which suggests embryo lethality. Indeed, white (delayed) and green embryos collected from CA1ca1ca3ca3 siliques were genotyped, confirming double mutation in homozygosis (Supplementary Fig. S4). As occurred in ca1ca2, ca1ca3 double mutant embryos show delayed embryogenesis, which is detectable 60 hours after pollination (HAP) (Fig.�2B-VII) (C�rdoba et�al. 2016). At about 240 HAP, while Col-0 embryos are completely developed (Fig.�2B-V), ca1ca3 embryos are found at late torpedo/curled cotyledon stage without greening (Fig.�2B-X).

Fig. 2.

ca1ca3 double null mutation is embryo lethal. (A) ca1ca3 siliques show abnormal seeds. In immature siliques (I and III), double homozygous embryos are still pale (arrowheads). In mature siliques (II and IV), the mutant ones are shrunken (arrowheads). Bars: 0.5 mm. (B) Embryos within pale seeds are delayed in their development. Percentages indicate the proportion of observed embryos in that stage. Bars: 50 mm.

Double mutant ca1ca3 embryos show low mitochondrial membrane potential

As a low amount of CI negatively affects mitochondrial membrane potential (MMP) (C�rdoba et�al. 2016), MMP was measured using TetraMethylRodhaMine probe (TMRM) on developing mutant and Col-0 embryos. After controlled manual pollination, embryos were extracted at different developmental stages from Col-0 and CA1ca1ca3ca3 siliques and stained with TMRM. In normal active mitochondria, TMRM staining results in intense red fluorescence (Fig.�3A–D). When embryos from CA1ca1ca3ca3 siliques were analyzed around 60 HAP (Fig.�3F), 23.2% (N = 65) of embryos show evident low MMP levels (about 70% lower than Col-0). At later stages (around 90 HAP), embryos displaying low fluorescence show developmental delay (24.4% from total embryos in the siliques; N = 53 embryos; Fig.�3G). Notably, around 120 HAP, ca1ca3 embryos (which are still delayed) recover their MMP at levels comparable to the ones observed in Col-0 embryos (25.7%, N = 58; Fig.�3H). Normal MMP levels are maintained until the end of embryogenesis (240 HAP).

Fig. 3.

Delayed ca1ca3 embryos have low MMP. Col-0 and ca1ca3 embryos were incubated with tetra-methyl-rhodamine after different HAP as indicated. Embryos were then analyzed using confocal microscopy. Within mutant CA1ca1ca3ca3 siliques, there are around 25% of delayed embryos, which show low MMP. Membrane potential drops around 60 HAP in delayed embryos (F), but there are not differences with Col-0 at 120 HAP (H). Right column plots show red fluorescence relative to Col-0. Asterisks mean statistically significant difference analyzed by T-test (P < 0.05). Percentages indicate the proportion of embryos observed at each stage. Upper right pictures in C, D, F, G and H indicate bright field images and the region shown using confocal microscopy. Bars: 50 �m.

CA1 and CA3 promoters are active during embryo development

Putative promoter regions of the genes of interest were cloned upstream the uidA (b-glucuronidase GUS) reporter gene to determine their transcriptional activity. A region covering 2,000 base pair (BP) upstream of the first translational ATG codon of CA1 was used. High promoter activities were found during pollen development while no activity was observed in embryo sac (Fig.�4A, B). Gus staining was evident in the developing embryo, from globular to heart stages (Fig.�4C–F). GUS activity was not detected in mature green (final stage) embryos although peripheral endosperm seems to have some CA1 promoter activity (Fig.�4H).

Fig. 4.

CA1 and CA3 promoter regions are mainly active during early stages of embryogenesis. Two thousand and 1047 base pairs upstream the first translational ATG codon from CA1 and CA3 genes (respectively) were transcriptionally fused to the GUS reporter gene and introduced into Col-0 plants by floral dip. Siliques at different stages were manually opened and incubated with X-gluc in an appropriate buffer overnight. After incubation, samples were mounted on slides and cleared with Hoyer’s solution. Stages: (A, I) anther showing stained microgametophytes; (B) ovules; (C, K) globular; (D) early heart; (E, L) heart; (F) late heart; (G, N) late torpedo (curled cotyledon); (H, O and P) mature green; (J) early globular; (M) torpedo. Bars: 50 �m.

To assay CA3 promoter activity, 1,470 base pairs upstream of the first translational ATG of CA3 were also fused to the uidA (GUS) reporter gene. Although there is no detectable activity at very early stages of embryo development (zygote to preglobular stages Fig.�4J), strong GUS activity was detected at late globular to heart stages (pale embryos, Fig.�4K, L).

Since the CA2 gene is highly expressed at the end of embryogenesis (green embryos, C�rdoba et�al. 2016), all three CA genes seem to have a complementary expression pattern during embryogenesis.

ca1ca3 seedlings do not survive under normal growth conditions

Shrunken seeds obtained from CA1ca1ca3ca3 self-pollinated plants germinate about 9 d after sowing in MS medium or soil, although seedlings die 2 d later. However, if these double mutant seeds are sown in MS medium supplemented with 1% w/v sucrose, double mutant ca1ca3 seedlings are able to survive. The same result was observed when seedlings were growing under long- or short-day conditions. However, 2 weeks after sowing (11 d if growing under long-day conditions) is necessary to transfer the plants to a medium containing 3% w/v sucrose for the plants to survive. Even when ca1ca3 double homozygous plants are able to grow on plates supplemented with sucrose, they are severely affected, looking smaller and showing bleached leaves (Fig.�5A). Fifteen days after growing in these conditions, plants were transferred to soil. After 21 d, ca1ca3 plants look healthier (Fig.�5B), although they show a notable developmental delay taking into account that Col-0 plants grown during the same period of time (47 d) under the same conditions are at ripening stage (Fig.�5B). Rescued plants were confirmed to be double homozygous for the mutated alleles by genotyping (Supplementary Fig. S5A). Further quantitative Reverse Transcriptase-Polymerase Chain Reaction (qRT-PCR) analyses show extremely low transcript levels of CA1 and CA3 genes in those plants (Supplementary Fig. S5B). Rescued plants are able to produce few seeds (5–10 seeds per silique); however, they are all shrunken and do not germinate, even in sucrose-supplemented media (Supplementary Fig. S5C). Moreover, alternative oxidases are induced in ca1ca3 rescued plants (Supplementary Fig. S6), and they also accumulate ROS as hydrogen peroxide and superoxide anion (Fig.�5C).

Fig. 5.

ca1ca3 rescued plants accumulate ROS. (A) Seeds from Col-0, ca1, ca3 and shrunken ca1ca3 were cultivated on 1% sucrose enriched MS medium. After 11 d, they were transferred to a 3% sucrose enriched medium and incubated 15 d more. Bars: 3 cm. (B) ca1ca3 rescued plants were transferred to soil. Twenty-one days later were photographed. Bars: 3 cm. (C) Plants growth on MS medium as described were exposed to DAB and NBT staining in order to detect hydrogen peroxide and superoxide, respectively. Bars: 5 mm (Col-0) and 2 mm (ca1ca3).

CI is not detected by Coomasie blue staining in the ca1ca3 mutants

Cell suspension cultures derived from ca1ca3 rescued plants were established. In order to study the OXPHOS composition of this mutant, mitochondria were isolated, and their membrane proteins were solubilized. Proteins were then separated in 2D Blue Native/sodium dodecyl maltoside-Polyacrylamide Gel Electrophoresis (BN/SDS-PAGE) gels. The comparison between Col-0 (Fig.�6A) and ca1ca3 (Fig.�6B) proteomes reveals that CI and I+III₂ supercomplex could not be detected in ca1ca3 mitochondria. None of the normal CI subunits were observed in the second dimension and NADH DH activity could not be detected in ca1ca3 mitochondrial protein complexes (Fig.�6B and Supplementary Fig. S7).

Fig. 6.

CI is not detected by Coomasie blue in ca1ca3 proteome. Mitochondria were isolated from cell cultures derived from Col-0 (A) and rescued ca1ca3 plants (B). After solubilization, membrane protein complexes were separated on Blue Native PAGE (lane on the top of the figure). Then, one lane of each sample was separated on a second, denaturing dimension (SDS-PAGE) in order to disaggregate these complexes in their proteins (middle square). After second electrophoresis, gels were stained by colloidal Coomasie blue or transferred to nitrocellulose membranes to perform Western blotting analysis (bottom panel). Antibodies raised against gamma CAs were used to detect gamma CA proteins. Identity of each protein complex and molecular weight in the second dimension are indicated. Red arrows signalize gamma-CA detection.

In spite of these results, Western blotting analysis using anti-CA antibody reveals an extremely low signal on the lane corresponding to CI subunits from ca1ca3 sample (Fig.�6B, lower panel). This result suggests that there could be traces of assembled CI that are not detectable by colloidal Coomasie blue staining.

CI NADH DH activity level in CA mutants is tissue-dependent

In order to determine the relative contribution of each CA protein to CI activity, we isolated total membrane proteins from immature (pale, from globular to torpedo stages), mature (green) embryos and leaves of Col-0, ca1, ca2 and ca3 mutants. Proteins were then separated in a Blue Native PAGE and NADH DH activity of mitochondrial CI was measured (Fig.�7).

Fig. 7.

CI activity in CA mutants is different among tissues. Membrane proteins from pale (A) and green (B) seeds, and leaves (C) corresponding to Col-0, ca1, ca2 and ca3 genotypes, were isolated and solubilized with digitonin. After Blue Native PAGE, in-gel NADH DH activity was determined. Representative gels are shown. Upper arrows indicate supercomplex I+III₂ position. Lower arrows indicate CI position. Numbers below indicate the quantification of CI by band density (in %) relative to loading taking from independent gels using the Image J software. Duplicates of each gel were stained with colloidal Coomasie blue G250 to use as loading controls (lower panel).

When CI NADH DH activity was analyzed in pale embryos, all three single mutants show a reduction of about 40%. At these stages of embryogenesis, ca1 mutants seem to have more supercomplex activity (Fig.�7A). When green embryos were analyzed, ca1 and ca3 single mutants show a reduction of about 40% in comparison to Col-0, while ca2 present a stronger reduction of around 60%. Unexpectedly, ca1 and ca3 mutants seem to have more activity of I+III₂ supercomplex than Col-0 (Fig.�7B).

NADH DH activity of mitochondrial CI determined in leaves from plants grown in solid MS medium show a strong reduction (around 80%) in ca2 mutants while ca1 mutants show normal CI activity and ca3 a mild reduction (around 35%) compared to Col-0 (Fig.�7C). These results confirm published data (Perales et�al. 2005, Meyer et�al. 2011, Soto et�al. 2015, C�rdoba et�al. 2016) for leaves and cell suspension.

Thus, we conclude that the contribution of each CA protein to the assembly of an active CI differ depending the tissue or life cycle stage. While seems to be indistinctly during seed development, the contribution of the CA2 subunit is clearly the more important from green embryos on during embryogenesis and especially in leaves of adult plants.

CA3 can form homodimers and interacts with FRO1 in vivo

In the Arabidopsis Interactome Mapping (Braun et�al. 2011) is suggested that the CA3 subunit was able to interact with some CA subunits. In order to test CA3 interaction partners, we performed Bimolecular Fluorescence Complementation (BiFC) assays in Nicotiana benthamiana epidermal cells. The CA3 coding sequence and those of all other CA family members were cloned into pH7FGW2 in order to make fusions to the N- or C-terminal fragments of Yellow Fluorescent Protein (YFP) (see M & M). The CA3-nYFP construct was transiently expressed in leaf cells of N. benthamiana alone or together with the tested proteins fused to the cYFP fragment by Agrobacterium tumefaciens-mediated leaf infiltration. BIFC signal was visualized by confocal microscopy 72 h after infiltration. When CA3nYFP-CA3cYFP was assayed, YFP fluorescence was detected following a punctuated pattern consistent with an interaction (Fig.�8A). However, no BIFC signal was observed in leaves in which CA3-nYFP was co-expressed with CA1, CA2 or CALs-cYFP proteins (Supplementary Fig. S9). As expected, BIFC signal was observed for CA1-CA1 interaction (Supplementary Fig. S8; see C�rdoba et�al. 2016 for two-hybrid analysis). These experiments revealed that CA3 is able to interact with itself, but it is unable to interact with any other member of the CA family.

Fig. 8.

CA3 protein is able to form homodimers and interacts FRO1 CI subunit in vivo. BiFC analysis of the interaction between CA3, CA1, ND6 and FRO1 proteins. Reconstituted YFP fluorescence or bright field images are shown in N. benthamiana epidermal leaf cells co-infiltrated with A. tumefaciens harboring the indicated constructs. (A) Representative images from confocal microscopy showing the interactions of CA3 with CA3, (B, C) interaction of CA3 with FRO1, in both directions, respectively. (D, E) Representative images from confocal microscopy showing that CA1 does not interact with CA3 nor FRO1 proteins. Bars: 50 �m.

In the same in silico approach, it is also predicted that CA3 protein may interact with FRO1 (At5g67590), ND9 (AtMg00070) and ND6 (AtMg00270) among other CI subunits (Braun et�al. 2011). To further test this possibility, a BIFC assay was performed. For the mitochondrial encoded genes, ND6 and ND9, a mitochondrial signal peptide that was proved to address proteins to plant mitochondria (Hernould et�al. 1993, G�mez-Casati et�al. 2002, Busi et�al. 2006) was used in their N-termini. For these mitochondrial proteins we were unable to detect expression after infiltration using similar constructs but fused to Green Fluorescent Protein (GFP) at their C-termini, thus we were unable to test these interactions. However, FRO1 and CA3 are able to interact in vivo, as BIFC signals were detected in both combinations (FRO1-cYFP with CA3-nYFP and FRO1-nYFP with CA3-cYFP) (Fig.�8B, C, respectively). Interestingly, the almost identical protein, CA1, neither interacts with CA3 nor with FRO1 (Fig.�8D, E), indicating that the CA3/FRO1 interaction is specific to CA3.

Double mutants fro1ca3 show a stronger phenotype

To obtain independent data on the putative interaction between FRO1 and CA3, double fro1ca3 knock out mutants were obtained. Single knock out fro1 mutants show a number of unviable seeds while survival seedlings show a characteristic dwarf phenotype until maturity. It was previously shown that this particular mutant present traces of NADH DH activity and accumulates the membrane arm of CI (K�hn et�al. 2015). In the double fro1ca3 mutant, the number of unviable seeds increased significantly (from around 5% in the single fro1 mutant to ∼20% in the double mutant, n = 150). Moreover, the size of survival seedlings is slightly smaller than single fro1 mutants (Supplementary Fig. S9). This result suggests that the lack of CA3 further destabilizes the subcomplex present in the fro1 single mutant strengthening its already altered phenotype.

Discussion

Earlier reports showed specific roles for different members of the CA domain, suggesting their involvement during embryogenesis (C�rdoba et�al. 2016, Fromm et�al. 2016a) and their roles in a photorespiratory context (Soto et�al. 2015). In this work, we provide evidence about the different composition of complexes I through development and the role of the CA3 protein subunit in the assembly of CI. Taken together our results suggest that the first intermediate formed (i.e. the CA domain, Ligas et�al. 2018) in the way of CI varies during the plant life cycle, presenting specific conformations along the different developmental stages. In pale embryos, CI may be formed from trimers containing indistinctly CA1-CA1-CAL, CA2-CA2-CAL or homotrimers of CA3. However, from green embryo stage on to adult plants, the CA2-CA2-CAL trimer becomes the more abundant (80% in the adult).

Complete lack of CI leads to embryo lethality

Many mutants affect CI assembly, directly by lacking an essential subunit or assembly factor or indirectly by lacking proteins important for maturation of mitochondrial RNAs coding for nad subunits. In Arabidopsis, the first mutant reported was ca2 (Perales et�al. 2005), which shows a reduction of fully assembled CI estimated in 80%. Also ca3 mutant was affected, although with less severity. Other mutants like ndufs4 (fro1) and ndufv1 (Meyer et�al. 2009, K�hn et�al. 2015) accumulate assembly intermediates of CI. While in fro1, traces of fully assembled CI with NADH DH activity are found, in ndufv1, NADH DH activity is not detected, although traces of a nearly complete assembled CI is also present (Ligas et�al. 2018). Consequently, germination in the absence of sugar was not much affected in ndufs4 mutants, while it is strongly affected and most seedlings died only in ndufv1 mutants (K�hn et�al. 2015). Some mutants lacking Pentatricopeptide Repeat Proteins, which are RNA binding proteins, with defects in splicing or stability of nad transcripts exhibit severely impaired CI assembly and activity, and show visual phenotypes such as reduced germination, retarded growth or curled leaves and in some cases embryo arrest phenotypes (OTP43, Falcon de Longevialle et�al. 2007, PPR19, Lee et�al. 2017, SLO3, Hsieh et�al. 2015, MTSFs, Wang et�al. 2017, Wang et al. 2018) depending on the remnant fully assembly CI present. Other CI mutants in nuclear maturase genes, like nmat1 (Keren et�al. 2012), nmat2 (Keren et�al. 2009) and nmat4 (Cohen et�al. 2014), all show a strong decrease in CI levels, although active traces are present and consequently mild phenotypes in all three mutants. Also in maize, four mutants, emp8 (EMPTY PERICARP 8), emp11, dek2 (DEFECTIVE KERNEL2) and dek37 are strongly defective in seed development, which correlates with a defect in the splicing of some nad transcripts and strong decrease of fully assembled CI (Qi et�al. 2017, Ren et�al. 2017, Dai et�al. 2018, Sun et�al. 2018). Thus, only Arabidopsis mutants completely lacking fully assembled CI, like ca1ca2 (C�rdoba et�al. 2016, Fromm et�al. 2016a), ca1ca3—this work—otp43 (Falcon de Longevialle et�al. 2007) and ndufv1 (K�hn et�al. 2015, Ligas et�al. 2018) do not survive on soil and require sucrose to germinate. Traces of CI might be enough in Arabidopsis but not in maize to achieve complete embryogenesis. These results indicate that the response of maize and Arabidopsis to altered CI assembly might be similar but with some variations. Maize embryos do not survive with traces of CI as Arabidopsis embryos do. An extreme opposite example constitutes Viscum album, which has a highly unusual OXPHOS system. It has recently been reported that the European mistletoe mitochondria completely lack CI and have greatly reduced amounts of complexes II and V. However, it only survives as an obligate semi-parasite, living on branches of trees. At the same time, the complexes III and IV form remarkably stable respiratory supercomplexes (Maclean et�al. 2018, Senkler et�al. 2018).

Double homozygous ca1ca3 and ca2ca3 mutants are severely affected

In our work, the alleles used to generate the double homozygous ca1ca3 show seedling lethality as was previously shown for ca1ca2 (C�rdoba et�al. 2016). Although it was previously published that the double homozygous ca1ca3 mutant does not show any phenotypical alteration compared to Col-0 plants (Wang et�al. 2012, Senkler et�al. 2017a), this discrepancy can be explained by the use of different mutant lines. In the first publication, the mutant alleles are not detailed; the origin or stock number of the mutant lines used is not specified as it is not shown whether they are null alleles.

Both CA double mutants (ca1ca2 and ca1ca3) are lethal but can be rescued if grown in a medium enriched in sucrose. However, they show slight differences. Although developmental delay during embryogenesis is detected about 60 HAP in both mutants, MMP drops earlier in the ca1ca3 genotype. In addition, ca1ca3 embryos recover their mitochondrial potential from torpedo stage on, even when embryogenesis continues to be affected. We suggest that the apparent recovery of MMP could be due to a high induction of alternative electronic transport pathways (as alternative NADH DHs).

Both ca1ca2 and ca1ca3 mutant seedlings die soon after germination when not supplemented with sucrose (C�rdoba et�al. 2016, Fig.�5). Germination and seedling establishment are high energy demanding processes that depend only on compounds accumulated along embryogenesis, especially after embryo greening. As embryo development is delayed, ca1ca2 and ca1ca3 mutants never arrive to the greening stage. Consequently, it is possible that they might not be able to synthesize (or accumulate) enough lipids to sustain the entire process (C�rdoba et�al. 2016). Although ca1ca2 (Fromm et�al. 2016a) and ca1ca3 adult plants can survive on soil after being rescued on sucrose, they do not have active CI and their energy transformation (from carbon to Adenosine Triphosphate (ATP)) efficiency is probably very low. Indeed, rescued plants are severely affected, showing high ROS accumulation (Fig.�5) and induction of alternative oxidases (Supplementary Fig. S7), responses that are very well known markers of a deficiency of CI activity (Falcon de Longevialle et�al. 2007, Garmier et�al. 2008, Keren et�al. 2012, Hsu et�al. 2014, K�hn et�al. 2015, Fromm et�al.2016c).

The differences observed between ca1ca2 and ca1ca3 double mutants could be explained by the presence of extremely low levels of CI in the latter. Indeed, traces of putative assembled CI were detected by Western blot after 2D BN/SDS-PAGE in ca1ca3 mitochondria (Fig.�6). This suggested that although the assembly of CI might initiate with a CA domain composed by CA2 and CAL1/2 proteins, it would not be stable enough in a ca3 mutant background. Since the transcript level of CA2 is almost unaffected (Supplementary Fig. S5B), we conclude that if translation is not affected, the leftover CA2 should be degraded.

To elucidate the basis of ca2ca3 growth deficiencies, mutant plants were grown under a high CO₂ atmosphere (which is sufficient to rescue ca2cal photorespiratory mutants, Soto et�al. 2015) or supplemented with sucrose (as lethal ca1ca2-Fromm et�al. 2016a—and ca1ca3 mutants—this work are rescued). However, these treatments were not enough to rescue the ca2ca3 phenotype, suggesting that the severe growth deficiencies observed in ca2ca3 mutants are at least not only due to photorespiratory or respiratory problems.

Expression of gamma CA members is tissue-specific

The promoter activity for each gamma CA member was analyzed using fusions with a reporter gene (C�rdoba et�al. 2016 and this work). Along embryogenesis, CA2 promoter was strongly active at late stages (mature green), while CA1 and CA3 promoter regions are more active in globular to heart stages. These differences might indicate a tissue-specific expression pattern and specific physiological roles of each gamma CA protein along embryo development.

It was previously suggested that the CA2 subunit may act as a CI assembly factor, essential for the initial steps of the assembly process (Meyer 2012). Thus, ca2 null mutants show a strong reduction in CI abundance and NADH DH activity. However, this assumption was originated from results obtained from leaves or from cell suspensions derived from stem cells (Perales et�al. 2005, Soto et�al. 2015).

In a previous work (C�rdoba et�al. 2016), it was suggested that the formation of CI was dependent on CA2 or CA1. These proteins could form trimers with CAL proteins such as CA2-CA2-CAL (80%) or CA1-CA1-CAL (20%), because the CA3 subunit had not been detected in the 85 kDa CA domain extracted with low SDS (Klodmann et�al. 2010). In light of recent works where CA3 was found first in a 130 kDa intermediate (Senkler et�al. 2017b) and then in the 85 kDa intermediate (the first proposed intermediate for CI assembly, i.e. the CA domain) (Ligas et�al. 2018), these ideas should be reinterpreted considering a proportion of CA3 dependent CI. Taking into account our results of interactions, where CA1 and CA2 proteins do not interact among them (C�rdoba et�al. 2016), the trimers might be CA2-CA2-CAL, CA1-CA1-CAL or homotrimers of CA3, because this last subunit does not interact with any other member of the CA family (this work). Furthermore, we show that NADH DH activity assays show that each single CA mutant is differently affected depending on the tissue and life cycle stage assayed. According to our results, during embryogenesis, assembly of CI seems to depend indistinctly on CA1, CA2 or CA3, which contrasts to the situation observed in mature plants where CA2 is the dominant subunit accounting for 80% of the trimers and consequently CI abundance. Surprisingly, ca3 mutants show lower NADH DH activity in every tissue and stage compared to Col-0 (around 60%). These results suggest that gamma CA3 subunit is important for CI activity along all stages of the plant life cycle, probably contributing to CI stability as well.

The gamma CA3 subunit as a CI stabilization factor

In vivo interaction analysis revealed that the CA3 subunit was unable to interact with any other member of the CA family. An interpretation to explain these results is that a proportion of CI is dependent on homotrimers composed only by CA3 subunits. In that case, this proportion would be low, because ca1ca2 and cal1cal2 double mutants show embryo lethal phenotypes (Wang et�al. 2012, C�rdoba et�al. 2016, Fromm et�al. 2016a) and, at least in rescued ca1ca2 lines, no traces of CI were observed. It is possible that the amount of CA3 dependent CI is might not be enough to sustain the embryogenesis and, in that context, be degraded. Even that, the difference between NADH DH activity seen in ca3 leaves respect to Col-0 together with an increase in I+III₂ supercomplex activity, point to a stabilization role.

We suggest that the lack of CA3 leads to a destabilization of a proportion of CI, which is in turn detected as lower NADH DH activity in every tissue compared to Col-0, around 60%. In agreement with this idea, the absence of one of its putative interaction partners, FRO1, causes extremely low CI activity and accumulation of the membrane arm, as CI is not fully assembled (Meyer et�al. 2009, Meyer et al. 2011). The phenotype of double mutants fro1ca3 is stronger than the single fro1 mutant, showing a 15% increase of unviable seeds and a slightly delay in the growing of the survival plants suggesting that the putative interaction between these two proteins might be important for CI assembly. An interpretation of these results might be that the proposed CA3 homotrimers should be able to interact with FRO1 in the matrix arm. However, as mentioned before, if that were the case, the proportion of CA3 homotrimers would be low, as ca1ca2 and cal1cal2 double mutants show embryo lethal phenotypes (Wang et�al. 2012, C�rdoba et�al. 2016, Fromm et�al. 2016a) and ca1ca2 lines show no traces of CI. A second interpretation could be to consider the existence of a homotrimer of CA3 located outside the CA domain, interacting with at least FRO1 and having a role of complex and supercomplex stability. This putative homotrimer of CA3 might have a similar size to the CA domain and if it is not stably bonded, might be detached during purification, migrating at a similar position to the 85 kDa intermediates when analyzed on gels (Ligas et�al. 2018). The fact that a complex CA3-FRO1 was not detected in the mass spectrometry experiments of Ligas et�al. (2018) or Senkler et�al. (2017b) points to an instability of this putative interaction. This interpretation may explain the similar decrease of NADH DH activity observed regardless on the tissue and the increase of supercomplex (I+III₂) detected in the ca3 single mutant, however, these results based on BIFC experiments should be taken with care, as some in vivo interactions may not be visualized using this method. In spite of the high sequence similarity among gamma CA members, especially between CA1 and CA3 (77% aminoacid sequence identity), the differences found in this work point to a functional and/or physiological specialization of each member of this protein family in A. thaliana.

Materials and Methods

Plant material and cell cultures

Long-day conditions were established at 16:8 h (L:D), 120 �E light intensity and 22�C. Short-day conditions were 10:14 h (L:D), 120 μE light intensity and 22�C. Arabidopsis thaliana Ecotype Columbia-0 and the mutants were growth on soil or in plate (with Murashige and Skoog medium supplemented with Gamborg’s Vitamins (Sigma) and with or without sucrose as indicated). For selection of transgenic lines, Hygromycin (Sigma) 15 μg/ml was added to MS medium. Seeds cultivated on plates were previously surface sterilized with 50% v/v ethanol during 10 min and then with isopropanol, 30 min with constant shaking. After this time, they were washed five times with sterile water. Double mutants were obtained by cross pollination between single null mutants: ca1 (Salk_109391), ca2 (Salk_010194), ca3 (CS69934), fro1 (SAIL_596_E11). Cell cultures were established as described in Perales et�al. (2005). See Supplementary Fig. S1 for alleles and the structure of genes used in this work and positions of T-DNA insertions.

The ca3 mutant was complemented using a construct with a 1,500 bp ca3 promoter and the Coding sequence (CDS) of the ca3 gene to distinguish from the WT sequence. Since ca3 mutants show lower NADH DH activity (around 65%), we measured this activity in complemented lines. Restored NADH DH activity was seen in complemented lines. Supplementary Fig. S10 shows a representative complemented line compared to Col-0 and ca3 mutant.

Genotyping of mutant lines

T-DNA insertional mutants or derived cell suspension cultures used in this work were genotyped by detection of mutated alleles. For Salk lines, a combination of LBb1.3 oligo (5′-ATTTTGCCGATTTCGGAAC-3′) and corresponding Fw (CA1 Fw: 5′-ATCTCGAGATGGGAACCCTAGGCA-3′; CA2 Fw: 5′-CACTCGAGTGGGAACCCTAGGA-3′). For ca3 mutation, SynLB1 oligo (5’- GCCTTTTCAGAAATGGATAAATAGCCTTGCTTCC-3′) and CA3 Rv (5′-ATCTCGAGTCAAGCTGCTTTTGGT-3′) were used. Wild type alleles were detected by combination of CA1Fw/CA1Rv (5′-CACACCACCTCCAATACTTTCAGCC-3′), CA2Fw/CA2Rv (5’- TCAGAGTAGGTAGAACCTTGCCA-3′) and CA3Fw (5’- GCCTCGAGATGAATGTTTTTGAC-3′)/CA3Rv.

Phenotype analysis

Visualization of ca1ca3 seed sets, embryo stages along embryogenesis taken by Differential interference contrast (DIC) optics and MMP status detected by staining with TetraMethylRhodamine (TMRM) were conducted as previously described (C�rdoba et�al. 2016). Quantification of red fluorescence intensity was measured following standardized protocols (Martin et�al. 2014, C�rdoba et�al. 2016).

Promoter activity

Promoter regions taken as sequences upstream first translational first codon (ATG) codon were cloned into pMDC162 vector (Curtis and Grossniklaus 2003). Col-0 plants were transformed with Promoter::GUS constructs by floral dip method (Clough and Bent 1998). Two thousand basis pair for CA1 and 1,470 bp for CA3 were analyzed by GUS staining as described in C�rdoba et�al. (2016) for CA2 promoter activity.

ROS detection

In order to detect hydrogen peroxide and hydrogen superoxide, plant leaves were stained with diaminobenzidine (DAB) and NitroblueTetrazolium (NBT), respectively. For DAB staining, leaves were incubated in total darkness with 1 mg/ml solution in (pH 3 water) along 12 h. After that, samples were washed with distilled water and chlorophyll were extracted by discoloration in 96% v/v ethanol, in constantly shaking. Then, leaves were washed with water. In situ brown staining was observed. NBT staining was conducted by incubation of leaves in 10 mM phosphate buffer pH 8, 10 mM NaN₃ and 0.1% NBT (Sigma), along 12 h at 37�C. Washing and chlorophyll discoloring were performed as described by DAB. In situ superoxide was detected by violet staining.

RT-qPCR transcription measurements

Total RNA from Col-0 and ca1ca3 double homozygous mutants was isolated by TRIzol reagent (Invitrogen) and used in retrotranscriptase reactions (with Promega enzymes and random hexamers). cDNA was then used in qPCR reaction using Power SyBR mix (AppliedBiosystems) in a StepOnethermocycler. qPCRoligos pairs were used as in Soto et�al. (2015). Transcript levels were normalized against ACT2 and UBQ5 as housekeeping genes.

2D Blue Native/SDS-PAGE

Isolation of mitochondria from cell cultures was carried out following the protocol described in Perales et�al. (2005). Mitochondrial membrane protein complexes were solubilized in digitonin. Ten micrograms of mitochondria (around 1 mg mitochondrial proteins) were mixed and incubated with 2.5 mg of digitonin in 100 �l (Eubel et�al. 2003) previous to electrophoresis separation. One-dimension blue native electrophoresis was performed as described in Wittig et�al. (2006), in large, polyacrylamide gradient (4 to 16% p/v) gels. After 18 h running at 500V, 15 mA, lanes were stained overnight in colloidal Coomasie blue G-250. Duplicated lanes were properly cut and mounted horizontally in a 16% polyacrylamide gel (Wittig et�al. 2006). Protein complexes were thus separated in their proteins after 20 h running at 500 V, 25 mA in denaturing conditions (SDS). After running, proteins spots were detected by staining in colloidal Coomasie blue G-250, overnight.

Mitochondrial membrane protein complexes from embryos or plants (leaves and roots) were isolated as is outlined in Pineau et�al. (2008). After that, proteins were separated as described above.

NADH DH activity staining

Separation of protein complexes under native conditions allows in-gel activity staining. NADH DH activity was assayed incubating the gels with 0.14 mM NADH (sigma) and 1 mg/ml NBT in 0.1 M Tris–HCl pH 7.4 until violet band is clearly viewed. After that, gels were fixed in 40% v/v methanol, 10% v/v acetic acid.

Protein immune detection

We used an anti-CAs serum to detect γCA1, γCA2 and γCA3 proteins. Mitochondrial proteins separated by 2D BN/SDS-PAGE were transferred to nitrocellulose membranes (GE Healthcare) and incubated with these primary antibody and then with an anti-rabbit secondary antibody (Sigma) conjugated with alkaline phosphatase, following standard protocols. Final development was carried out with NBT/BCiP (Promega).

Immune detection of alternative oxidase isoforms was performed on proteins obtained from Col-0 and ca1ca3 (rescued) plants. After mitochondrial membrane protein complexes solubilization in digitonin, proteins were treated under denaturing conditions (5 min at 98�C heating and β-mercaptoethanol), and then, separated in 12% polyacrylamide SDS-PAGE. Protein transfer and immunodetection were performed as described above. Anti-Alternative Oxidase (AOX) primary polyclonal antibodies are from Agrisera (Sweden; Catalog AS04 054).

BiFC in vivo interactions

In vivo analysis of protein interaction was performed by Bimolecular Fluorescence Complementation (BiFC) assay. The cDNA sequences of CA1, CA2, CA3, CAL1, CAL2, FRO1 (At5g67590, other name NDUFS4) and ND6(AtMg00270) were PCR amplified using the following primers: CA1 (5′-CACCATGGGAACCCTAGGCAGA-3′/5′-GTTCACATTAGAAGGACGCTT-3′), CA2 (5′-CACCATGGGAACCCTAGGACGA-3′/5′-GAAGTACTGGATAGACGGAAC-3′), CA3 (5′-CACCATGGGAACAATGGGTAAA-3′/5′-AGGTGCTTTTGGTACATTCTC-3′), CAL1 (5′-CACCATGGCGACTTCGATAGCT-3′/5′-TAAACGGCGATCCCAAGGGAC-3′), CAL2 (5′-CACCATGGCGACTTCGTTAGCA-3′/5′-GATGGCGATTCCAAGGGATTT-3′), FRO1 (5′-CACCATGGCGCTTTGTGCT-3′/5′-GTTTTCTGGTTGAGGATT3′) and ND6 (5′-CACCATGATACTTTCTGTT-3′/5′-GTAGATCGTTGGGTC-3′). The PCR product were cloned into pENTR TOPO plasmids using Gateway technology (Invitrogen), sequenced and recombined through BP reaction into pH7FGW2 destination vector (half YFP C-t; Grefen et�al. 2010). The binary plasmids were then transformed into A. tumefaciens strain GV3101 by electroporation. Split tagged protein -nYFP and tagged protein -cYFP pairs were co-expressed in young (4–5 weeks) N. benthamiana leaves by A. tumefaciens-mediated inoculation. Plant leaves were examined 72 h post-infiltration with a confocal microscope (Nikon Eclipse C1 Plus).

Funding

Agencia Nacional de Promoci�n Cient�fica y Tecnol�gica (ANPCyT), Consejo Nacional de Investigaci�n Cient�fica y Tecnol�gica (CONICET) and Howard Hughes Medical Institute (HHMI).