3-Hydroxyisobutyrate Dehydrogenase Is Involved in Both, Valine and Isoleucine Degradation in Arabidopsis thaliana1[OPEN]

3-Hydroxyisobutyrate dehydrogenase, an enzyme of the branched-chain amino acid breakdown pathway, is involved in valine and isoleucine but not in leucine degradation in Arabidopsis. In plants, amino acid catabolism is especially relevant in metabolic stress situations (e.g. limited carbohydrate availability during extended darkness). Under these conditions, amino acids are used as alternative substrates for respiration. Complete oxidation of the branched-chain amino acids (BCAAs) leucine, isoleucine (Ile), and valine (Val) in the mitochondria efficiently allows the formation of ATP by oxidative phosphorylation. However, the metabolic pathways for BCAA breakdown are largely unknown so far in plants. A systematic search for Arabidopsis (Arabidopsis thaliana) genes encoding proteins resembling enzymes involved in BCAA catabolism in animals, fungi, and bacteria as well as proteomic analyses of mitochondrial fractions from Arabidopsis allowed the identification of a putative 3-hydroxyisobutyrate dehydrogenase, AtHDH1 (At4g20930), involved in Val degradation. Systematic substrate screening analyses revealed that the protein uses 3-hydroxyisobutyrate but additionally 3-hydroxypropionate as substrates. This points to a role of the enzyme not only in Val but possibly also in Ile metabolism. At4g20930 knockdown plants were characterized to test this conclusion. Root toxicity assays revealed increased root growth inhibition of the mutants if cultivated in the presence of Val or Ile but not in the presence of leucine. We conclude that AtHDH1 has a dual role in BCAA metabolism in plants.

Plants can synthesize all 20 proteinogenic amino acids. Their carbon skeletons and amino groups directly or indirectly derive from photosynthesis. Besides amino acid biosynthesis, plants also can break down all 20 amino acids. Amino acid catabolism is especially relevant in the context of germination (the conversion of storage proteins into carbohydrates), senescence (the recycling of energy-rich compounds), and in the context of stress reactions (Hildebrandt et al., 2015). The synthesis of some amino acids is massively induced upon drought and salt stress, most of all proline, because they can serve as compatible osmolytes (Szabados and Savouré, 2010). Upon stress release, these amino acids are rapidly degraded. Furthermore, amino acid catabolism is essential for respiration in low-light conditions or extended darkness. Upon light shortage, the availability of carbohydrates for respiration is limited, and amino acids can be used as alternative respiratory substrates (Araújo et al., 2011). Amino acid catabolism mainly takes place in the mitochondria of plants, but most enzymatic reactions so far have not been characterized (Hildebrandt et al., 2015). They may or may not resemble reactions taking place in mammalian cells, which have been investigated extensively (Harper et al., 1984).
The branched-chain amino acids (BCAAs) Leu, Ile, and Val have aliphatic and comparably short side chains. Their concentrations increase under various stress conditions (Zhao et al., 1998;Joshi et al., 2010). The complete oxidation of BCAAs in the mitochondria allows the generation of high amounts of ATP (Hildebrandt et al., 2015). Indeed, it has been shown that the degradation pathways for BCAAs are up-regulated in extended darkness (Däschner et al., 2001;Ishizaki et al., 2005;Araújo et al., 2010;Binder, 2010). The BCAA catabolic pathways are especially complicated and, so far, only fragmentarily understood in plants.
The first two steps in the degradation of Val, Ile, and Leu are identical (see Fig. 7, which summarizes the known and putative steps of BCAA catabolism in plants). After an initial transamination reaction, the BCAAs are decarboxylated. Several transaminases that use BCAAs as substrates have been identified in plants. As a result, branched-chain 2-oxoacids are produced (Angelovici et al., 2013). The decarboxylation of the BCAAs is carried out by the branched-chain a-ketoacid dehydrogenase complex (BCKDH), which is a very large complex and consists of multiple copies of the three enzymes E1, E2, and E3 (Mooney et al., 2002). The third enzyme in the degradation of Leu is the isovaleryl-CoA dehydrogenase (IVDH). This enzyme most likely also is involved in the parallel step of Val and Ile breakdown pathways. IVDH is a flavoenzyme that transfers electrons to the Electron Transfer Flavoprotein (ETF; Däschner et al., 2001;Araújo et al., 2010). The fourth biochemical step, a carboxylation reaction, is only known for Leu catabolism (Alban et al., 1993;Anderson et al., 1998). Starting from this point, further steps in the degradation of the three BCAAs are not known in plants. Only putative enzymes showing some degree of sequence similarity to already identified and characterized mammalian and bacterial enzymes have been reported (Hildebrandt et al., 2015). Most likely, an enoyl-CoA hydratase is involved in converting methyl-glutaconyl-CoA coming from Leu as well as 2-methylbutanoyl-CoA and 2-methylpropanoyl-CoA coming from Ile and Val catabolism, respectively. From this step, the three degradation pathways are predicted to differ. The final reaction in Leu catabolism is the conversion of 3-hydroxymethylglutaryl-CoA to acetyl-CoA and acetoacetate. In Ile catabolism, probably a 3-hydroxyacyl-CoA dehydrogenase and a 3-ketoacyl-CoA thiolase are involved in the formation of acetyl-CoA. The terminal steps of Val catabolism seem to be most complex. Probably, 3-hydroxyisobutyrate is formed from 3-hydroxyisobutyryl-CoA by a hydrolyzation reaction. Next, an oxidation takes place that is catalyzed by 3-hydroxyisobutyrate dehydrogenase. This step is special in the degradation of BCAAs because 3-hydroxyisobutyrate is not coupled to CoA. Different mammalian as well as bacterial 3-hydroxyisobutyrate dehydrogenases have been characterized in detail (Rougraff et al., 1988;Hawes et al., 1995Hawes et al., , 1996Chowdhury et al., 2003;Murín et al., 2008;Lee et al., 2014;Park et al., 2016). In contrast, this enzyme has not been characterized in plants so far. The product of the reaction catalyzed by 3-hydroxyisobutyrate dehydrogenase is methylmalonate semialdehyde. A final decarboxylase of the Val catabolic pathway catalyzes the conversion of methylmalonate semialdehyde into propionyl-CoA. The breakdown of propionyl-CoA in plants results in acetyl-CoA. It has been suggested that at least parts of this pathway in plants are carried out by the same set of enzymes that convert 3-hydroxyisobutyryl-CoA to propionyl-CoA (Hildebrandt et al., 2015).
We here describe the identification and characterization of a 3-hydroxyisobutyrate dehydrogenase (AtHDH1) in Arabidopsis (Arabidopsis thaliana). The enzyme was recombinantly expressed in Escherichia coli and affinity purified. The native state and the enzyme properties of AtHDH1 were determined. Besides 3-hydroxyisobutyrate, AtHDH1 converts 3-hydroxypropionate and methyl-3-hydroxy-2methylpropionate, as revealed by systematic substrate screening analyses. This points to a role of AtHDH1 not only in Val but also in Ile metabolism. Root toxicity assays using AtHDH1 knockdown plants were employed to evaluate the physiological role of AtHDH1 in plants. The results indicate a role of AtHDH1 in both Val and Ile catabolism but not in the breakdown of Leu.

Identification of a Putative 3-Hydroxyisobutyrate Dehydrogenase from Arabidopsis
To better understand amino acid degradation in plants, the Arabidopsis genome sequence was systematically searched for genes encoding proteins similar to known enzymes involved in amino acid catabolism in animals, fungi, or bacteria (Hildebrandt et al., 2015). Furthermore, genes encoding mitochondrial dehydrogenases were systematically searched based on sequence comparisons (Schertl and Braun, 2014). Finally, shotgun proteome data sets of mitochondrial fractions from Arabidopsis were searched for putative enzymes involved in amino acid catabolism (Schertl, 2015). One identified gene encodes a 6-phosphogluconate dehydrogenase family protein (annotation by TAIR; accession no. At4g20930). Besides resembling 6-phosphogluconate dehydrogenase family proteins, the protein was found to exhibit sequence similarity to mammalian 3-hydroxyisobutyrate dehydrogenases (HDH; Supplemental Fig. S1). The protein At4g20930 was identified previously in a mitochondrial fraction of Arabidopsis in the course of a 2D gel-based proteome project (Taylor et al., 2011;Supplemental Fig. S2). It has an apparent molecular mass of 34 kD. Furthermore, At4g20930 has been identified in a mitochondrial fraction of Arabidopsis by a complexome-profiling approach (Senkler et al., 2017).

Recombinant Expression and Purification of At4g20930
A full-length Gateway clone (G83598) encoding At4g20930 was ordered from the Arabidopsis Biological Resource Center (Yamada et al., 2003). The open reading frame, which encodes a protein of 37.4 kD, was amplified by PCR, but excluding the N-terminal 84 bp, which encode a predicted mitochondrial presequence of 28 amino acids (Supplemental Fig. S3). For the overexpression of At4g20930 in E. coli, the Gateway pDEST17 vector was used, which allows expression of the protein in frame with an N-terminal 63 His tag. The overexpressed fusion protein was partly soluble if E. coli cells were cultivated at 16°C (at higher temperatures, the protein formed inclusion bodies). The protein was successfully affinity purified, as documented by SDS-PAGE and immunoblotting using an IgG directed against the His tag ( Fig. 1). Furthermore, mass spectrometry (MS) analysis of the purified protein revealed that the generated transcripts are translated in the correct frame and that the protein is complete (Table  I; Table S1). The overexpressed protein is very pure (Fig. 1). MS analyses of two minor protein bands at 75 and 30 kD did not lead to the detection of any endogenic E. coli dehydrogenases (Supplemental Table S1).

Enzyme Properties of At4g20930
The substrate specificity of the purified protein was systematically tested by photometric activity assays using NAD + as a cosubstrate (Table II). Tested compounds were selected either because they were found to be substrates of previously characterized bacterial and/or mammalian 3-hydroxyisobutyrate dehydrogenases or because they exhibit structural similarities to 3-hydroxyisobutyrate. The substrate screening revealed highest activity of the purified enzyme for 3-hydroxyisobutyrate (4.61 mM mg 21 min 21 = 100%). To a minor extent, 3-hydroxypropionate also is converted by the enzyme (13% activity compared with 3-hydroxyisobutyrate). 3-Hydroxypropionate is an intermediate of the suggested conversion of propionyl-CoA into acetyl-CoA, which represents the final phase of the Val and Ile catabolic pathways. Furthermore, purified At4g20930 showed some low activity with methyl-3-hydroxy-2-methylpropionate as substrate (1.2%). Methyl-3-hydroxy-2-methylpropionate has an additional methyl group at the carboxyl group in comparison with 3-hydroxyisobutyrate and was reported to be one of the major volatile components in the mammee apple (Mammea americana) fruit (Morales and Duque, 2002). Remarkably, the R-enantiomer of 3hydroxyisobutyrate was only converted with 2% activity in comparison with the S-enantiomer. The R-enantiomer of methyl-3-hydroxy-2-methylpropionate was not converted at all by the enzyme. These results reflect a very high stereospecificity of the enzyme. Other tested compounds did not reveal any activity (Table II). For instance, glycerate, serine, as well as b-hydroxypyruvate are no substrates for the enzyme, although these compounds have similar structures to 3-hydroxyisobutyrate. We conclude that At4g20930 encodes a 3-hydroxyisobutyrate dehydrogenase. The enzyme was designated AtHDH1.
K m and K cat (turnover rate) values of AtHDH1 from Arabidopsis were determined using freshly purified protein, 3-hydroxyisobutyrate as substrate, and NAD + as cosubstrate. The calculated K m values were 686 and 350 mM for 3-hydroxyisobutyrate and NAD + , respectively (Table III). The turnover rate K cat was 3.43 s 21 for 3-hydroxyisobutyrate and 2.652 s 21 for NAD + . This implies a catalytic efficiency (K cat /K m ) of 4.993 and 7.571 for 3-hydroxyisobutyrate and NAD + , respectively. The reaction mechanism catalyzed by AtHDH1 is a ; -A, protein fraction of noninduced cell lysate (5 mg); AI, protein fraction after induction of the AtHDH1 gene by arabinose for 24 h at 16°C (5 mg); NB, soluble protein fraction not bound to Ni-NTA agarose beads (5 mg); W1, protein fraction of the first washing step (15 mL); W2, protein fraction of the second washing step (15 mL); E1, first protein fraction eluted from the Ni-NTA agarose beads (recombinant AtHDH1); E2, second protein fraction eluted from the Ni-NTA agarose beads (recombinant AtHDH1). B, Corresponding western blot. Recombinant AtHDH1 was detected using a horseradish peroxidase-coupled His antibody.  Fig. S4). sequential bibi mechanism common for most dehydrogenases (Supplemental Fig. S5). Both substrates (3hydroxyisobutyrate and NAD + ) have to bind before either product is released. The pH optimum of AtHDH1 is 8.5 ( Fig. 2A), which is close to the estimated pH value of 8.1 for the mitochondrial matrix in Arabidopsis as determined by a mitochondria-specific fluorescence pH sensor (Shen et al., 2013). Enzyme activity measurements at varying temperatures revealed a wide temperature range. The optimum activity is at about 40°C (Fig. 2B). Above 50°C, the enzyme activity declines. Arabidopsis AtHDH1 is highly specific for NAD + as a cofactor. NADP + is not a suitable cofactor for AtHDH1. Also, FAD, oxidized cytochrome c, as well as the artificial electron acceptors phenazine methosulfate (PMS) and dichlorophenol indophenol cannot be reduced by AtHDH1 (data not shown).
In parallel, another Arabidopsis protein, At4g29120, which slightly resembles AtHDH1, was overexpressed in E. coli and affinity purified (Supplemental Fig. S6). This protein did not exhibit any 3-hydroxyisobutyrate dehydrogenase activity. We conclude that HDH is encoded by single-copy gene. Furthermore, lack of activity in overexpressed At4g29120 fractions further excludes that the E. coli background of our protein expression system interfered with our biochemical characterization of overexpressed AtHDH1.

The Native State of AtHDH1
The native state of AtHDH1 from Arabidopsis was tested by analyzing the purified recombinant protein using blue native PAGE in combination with a 3-hydroxyisobutyrate dehydrogenase in-gel activity assay (Fig. 3). Due to a high Coomassie Blue background in the low-molecular-mass region of the gel, the activity of the monomer could not be seen. Activity signals were visible at about 150, 187, 200, and 224 kD. Since the overexpressed protein is of high purity, we conclude that AtHDH1 can form homooligomeric protein complexes. Dimeric and tetrameric forms of HDH1 were described previously for several organisms (Rougraff et al., 1988;Lokanath et al., 2005). The 150-kD band nicely corresponds to the expected mass of a tetramer. The 187-and 224-kD bands could represent pentamers or hexamers. The identity of the 200-kD band is not clear. Table II. Enzymatic activities of recombinant AtHDH1 using different substrates and NAD + as electron acceptor Activity was measured using the standard conditions described in "Materials and Methods." The activity of the substrate S-3-hydroxyisobutyrate was defined to be 100%. Kinetic data were fitted by using nonlinear regression analysis. The values represent means 6 SE of three independent enzyme preparations.

Characterization of Arabidopsis AtHDH1 Knockdown Lines
Arabidopsis lines carrying T-DNA insertions in the gene encoding At4g20930 were ordered to investigate the physiological role of AtHDH1. Overall, six mutant lines are available, termed DHDH1-1 to DHDH1-6 ( Fig.  4). The exact positions of the insertions were determined by DNA sequencing for lines DHDH1-2, DHDH1-4, and DHDH1-6. Line DHDH1-4 carries an insertion 471 bp upstream of the transcription initiation site, which is usually considered to be outside the promotor region in Arabidopsis (Kleinboelting et al., 2012;Shahmuradov et al., 2017). Insertions in DHDH1-1 and DHDH1-5 are even farther upstream (Fig. 4). All three lines were discarded for further analyses because the insertions probably do not affect the expression of the AtHDH1 gene.
The T-DNA insertion of another line (DHDH1-3) could not be confirmed by the GABI-Kat consortium. For these reasons, all further experiments were carried out using lines DHDH1-2 (GK-911G06) and DHDH1-6 (GK-710H08). Line DHDH1-6 carries the insertion within the 59 untranslated region (9 bp upstream of the transcript initiation site) and line DHDH1-2 within intron 6 ( Fig. 4). Both lines are homozygous with respect to the T-DNA insertion, as confirmed by PCR (Supplemental Fig. S7). HDH1 expression is reduced by around 50% in the two mutant lines (Fig. 5). We conclude that both lines represent knockdown lines. HDH1 knockout lines possibly are not viable because disruption of the hdh1 gene causes an embryo-lethal defect.

DHDH1 Knockdown Lines Are Deficient in Val and Ile But Not in Leu Breakdown
Plants of the Arabidopsis lines DHDH1-2 and DHDH1-6 had no visible phenotypes under the conditions tested. Plants were cultivated in long-day conditions and shortday conditions. Even upon extended darkness for 7 d, no phenotypic differences were visible between wild-type plants and the knockdown mutants (Supplemental Fig.  S8). The progress of chlorosis was similar in all lines. We conclude that reduction in AtHDH1 expression is not strong enough to cause a visible phenotype. Therefore, root toxicity assays were carried out to use a more sensitive experimental system for monitoring phenotypic effects of the T-DNA insertion line.
Wild-type plants and plants of the AtHDH1 knockdown lines DHDH1-2 and DHDH1-6 were cultivated under sterile conditions on agar plates for 4 d. Subsequently, plant cultivation was continued on agar plates, which were supplemented with either 250 and 500 mM Val or with 50 and 100 mM Leu/Ile. After 10 d of further Figure 2. pH and temperature optima of overexpressed AtHDH1 from Arabidopsis. A, pH optima. The activity of AtHDH1 was measured in a buffer containing 250 mM Tris-HCl, pH 7 to 9.5, 5 mM NAD + , and 5 mM 3-hydroxyisobutyrate. B, Temperature optima of AtHDH1. The activity of AtHDH1 was measured at 20°C to 65°C in a buffer containing 250 mM K 2 HPO 4 3 KH 2 PO 4 , pH 7.5, 5 mM NAD + , and 10 mM 3-hydroxyisobutyrate. cultivation, roots were scanned and root length was calculated using the AxioVision Microscope Software (Carl Zeiss). The cultivation of plants in the presence of excess Val and Ile clearly increased root growth inhibition in the mutant with respect to wild-type plants (Fig. 6). In contrast, root growth in the presence of Leu did not cause a decrease in root growth in the mutant line compared with the wild-type line. The observed root inhibition effects in the AtHDH1 knockdown line were of moderate intensity, which probably is caused by the remaining AtHDH1 expression (Fig. 5). We conclude that AtHDH1 is involved in the breakdown of Val and Ile but not of Leu.

DISCUSSION
Plants can generate ATP from ADP and inorganic phosphate in various ways. In the light, ATP mainly comes from photophosphorylation taking place in the chloroplasts. At the same time, as in heterotrophic eukaryotes, ATP is generated in the mitochondria by oxidative phosphorylation. The latter process is the main ATP-generating process in plants in the absence of light (e.g. at night) or in nongreen tissue (e.g. roots). For oxidative phosphorylation, plants have to fuel the electron transport chain with electrons, which mainly come from the oxidation of carbohydrates via glycolysis and the citric acid cycle. However, under special abiotic as well as biotic stress conditions, carbohydrates can become limiting and oxidation of other compounds becomes important for providing electrons to the mitochondrial electron transport chain. The oxidation of BCAAs was found to be very important for the respiration of plant cells under carbon starvation conditions (Ishizaki et al., 2005(Ishizaki et al., , 2006Araújo et al., 2010). In BCAA catabolism, BCKDH generates NADH. We here report the identification of a second NADH-generating enzyme of the BCAA degradation pathway, AtHDH1. The suggested role of the enzyme in BCAA catabolism is summarized in Figure 7. Besides BCKDH and AtHDH1, the existence of a third enzyme, methylmalonate-semialdehyde dehydrogenase, is predicted to produce NADH. The formed NADH can transfer electrons to the NADH dehydrogenase complex (complex I) or to one of the alternative NAD(P)H dehydrogenases forming part of the respiratory chain in plants (Rasmusson et al., 2004). Furthermore, IVDH, another enzyme of the BCAA catabolic pathway, contributes electrons to the respiratory chain. However, this enzyme directly transfers electrons via the ETF/ETFQO system to ubiquinone. Finally, BCAA catabolism leads to the formation of acetyl-CoA, which is a substrate of the citric acid cycle. Complete oxidation of the acetyl group generates another three NADHs and one FADH 2 . All experiments of our study were carried out using lines DHDH1-2 and DHDH1-6 (insertions indicated by blue triangles). TIS, Transcription initiation site. Figure 5. Relative AtHDH1 expression in DHDH1-2 and DHDH1-6 knockdown lines. Relative expression was determined by reverse transcription-quantitative PCR (n = 3). Actin was used for normalization. Wt, Wild type. Asterisks show significant differences compared with the wild type according to Student's t test (P , 0.05).
In summary, BCAA degradation can greatly contribute to oxidative phosphorylation at carbohydrate limitation. Nevertheless, Leu, Ile, and Val catabolism has not been much studied in plants so far. We here report the characterization of one of the so-far unknown enzymes of the BCAA degradation pathway in plants, AtHDH1 (At4g20930). A second putative HDH gene was recently predicted, At4g29120 (Hildebrandt et al., 2015). The amino acid sequence identity between At4g29120 and At4g20930 is 32%. We also recombinantly expressed At4g29120 in E. coli (Supplemental Fig. S6). However, the purified protein did not exhibit 3-hydroxyisobutyrate dehydrogenase activity (Supplemental Fig. S6). We cannot exclude that the overexpressed protein was inactive due to misfolding or aggregation. However, our results here rather indicate that HDH is encoded by a single-copy locus in Arabidopsis. In contrast, HDH isoforms have been described for mammalian mitochondria (Loupatty et al., 2006).
AtHDH1 was shown to be a highly active 3hydroxyisobutyrate dehydrogenase. In addition, 3-hydroxypropionate and methyl-3-hydroxy-2methylpropionate also are substrates for AtHDH1.
No activity could be measured using glycerate or serine as substrate. This is in contrast to HDH from mammalian mitochondria. The enzyme from rat is active with a series of 3-hydroxyacid substrates, including S-and R-hydroxyisobutyrate, L-glycerate, and L-Ser (Hawes et al., 1996). In our study, AtHDH1 shows high specificity for S-enantiomers, as reported for the majority of HDHs from microorganisms and mammals (Hasegawa, 1981). AgNO 3 is a strong inhibitor of AtHDH1 activity (data not shown). This points to the presence of sulfhydryl groups close to the 3-hydroxyisobutyrate-or NAD + -binding sites. The optimal pH for AtHDH1 is around 8.5. Also, HDH from different mammals show a basic pH optimum (Robinson and Coon, 1957;Tasi et al., 2013). The temperature optimum of AtHDH1 is around 40°C, and the enzyme shows a wide activity range between 20°C and 50°C, which reflects the conditions of natural Arabidopsis habitats. The AtHDH1 knockdown plants used in our study had no visible phenotype under the conditions tested, even under extended darkness, which most likely is due to the remaining AtHDH1 expression. However, more sensitive root toxicity assays revealed a role of AtHDH1 in BCAA catabolism. The 3-hydroxyisobutyrate dehydrogenase from Thermus thermophilus has been crystallized and shown to be a homotetrameric enzyme (Lokanath et al., 2005). Similarly, AtHDH1 might form homooligomeric complexes. Activity signals of the recombinantly expressed protein could be detected in the molecular mass range from 150 to 225 kD upon blue native PAGE (Fig. 3). This might indicate that AtHDH1 forms tetrahomomeric to hexahomomeric complexes. However, these results have to be taken with caution, because artificial aggregations of AtHDH1 overexpressed in E. coli cannot be excluded.
According to currently available information, the first four enzymatic steps of Val, Ile, and Leu degradation are carried out by the same set of enzymes in plants (Hildebrandt et al., 2015), with an enoyl-CoA hydratase representing the last enzyme acting in all three degradation pathways (Fig. 7). In Val degradation, hydroxyisobutyrate is produced from hydroxyisobutyryl-CoA. This step distinguishes the Val catabolic pathway from Leu and Ile degradation. In Leu as well as Ile degradation, no CoA liberation takes place at this step. AtHDH1 converts 3-hydroxyisobutyrate to methylmalonate semialdehyde. Root toxicity experiments with mutant lines show the involvement of AtHDH1 in Val degradation. Cultivation of AtHDH1 knockdown mutants in the presence of excess of Val inhibits root growth stronger than in wild-type plants (Fig. 6). In the next step, CoA is again attached. This step is carried out by the methylmalonate Figure 7. BCAA catabolism in plants (modified from Hildebrandt et al., 2015). Enzymes are underlaid by gray circles (BCAT, branched-chain amino acid transaminase; MCCase, methylcrotonyl-CoA carboxylase; ECH, enoyl-CoA hydratase; HL, hydroxmethylglutaryl-CoA lyase; HCDH, 3-hydroxyacyl-CoA dehydrogenase; KCT, 3-ketoacyl-CoA thiolase; HADH, hydroxyacyl-CoA hydrolase; MMSDH, methylmalonatesemialdehyde dehydrogenase; ACOX, acyl-CoA oxidase; ACAT, acetyl-CoA acetyltransferase). Some key substrates, cosubstrates, and products are underlaid in blue (Leu, Ile, and Val). The two enzymatic steps catalyzed by HDH are underlaid in green. semialdehyde dehydrogenase. The mammalian enzyme is unique among known aldehyde dehydrogenases because it is CoA dependent (Kedishvili et al., 1992). It is assumed that propionyl-CoA is produced in Val as well as in Ile degradation (Fig. 7). Finally, it has been suggested that at least five enzymes are required to convert propionyl-CoA into acetyl-CoA (Hildebrandt et al., 2015). We present evidence that AtHDH1 is among these five enzymes. AtHDH1 is capable of converting 3-hydroxypropionate into malonate semialdehyde (Table II). Indeed, root toxicity experiments using the AtHDH1 knockdown mutants revealed that growth inhibition increased not only upon cultivation of plants in the presence of Val but also in the presence of Ile (Fig.  6). This double role of HDH also has be reported for 3-hydroxyisobutyrate dehydrogenase from microorganisms (Yao et al., 2010). In contrast, according to available experimental data, HDH seems to be involved only in Val degradation in animals. Further investigation of BCCA catabolism in plants will be required to better understand the plant-specific features and the physiological function of BCAA catabolism in the context of photoautotrophic life.

Expression and Purification of Recombinant AtHDH1
The nucleotide sequence encoding Arabidopsis (Arabidopsis thaliana) AtHDH1 was amplified from a Gateway clone (G83598; Arabidopsis Biological Resource Center). The first 84 bp of the open reading frame were excluded because it encodes the N-terminal presequence as predicted by MitoFates Fukasawa et al. 2015 (http://mitf.cbrc.jp/MitoFates/cgi-bin/top.cgi; Supplemental Fig. S3). The forward PCR primer contained the sequence CACC at the 59 end to enable directional TOPO cloning. The following primers were used: AtHDH1_FWD_w/ o_Mito (59-CACCTCTTCGTCTCAAAATTCAA-39) and AtHDH1_RV_w/o_Mito (59-TCAGACCTCATCTTTCCCATTG-39). PCR was carried out using Phusion High-Fidelity DNA Polymerase (Thermo Fisher Scientific) to create blunt-end PCR products. The following conditions were applied: initial denaturation at 95°C for 5 min; followed by 37 cycles of denaturation at 95°C for 30 s, annealing at 67.5°C for 30 s, and elongation at 72°C for 90 s; and a final elongation step at 72°C for 5 min. The obtained DNA fragment was cloned into the pENTR/D-TOPO vector using the pENTR Directional TOPOCloning Kit (Thermo Fisher Scientific). The resulting plasmid was called pENTR-D-TOPO-AtHDH1-w/o-28-AA. After transformation of the plasmid in One Shot Chemically Competent Escherichia coli cells (contained in the pENTR Directional TOPOCloning Kit) and overnight culturing on Luria-Bertani (LB) agar plates supplemented with 50 mg mL 21 kanamycin, single colonies were picked and propagated overnight at 37°C in LB medium at 170 rpm. Afterward, the plasmid DNA was isolated using the GeneJET Plasmid Miniprep Kit (Thermo Fisher Scientific). The plasmid was sequenced by SEQLAB (SEQLAB Sequence Laboratories). Afterward, the LR recombination reaction using the Gateway LR Clonase II Enzyme Mix (Thermo Fisher Scientific) was performed to clone AtHDH1 into the Gateway pDEST17 vector so that AtHDH1 can be expressed in frame with an N-terminal 63 His tag. The resulting expression plasmid was designated pDest17-AtHDH1-w/o-28-AA. Next, the plasmid was transformed into NEB 5-alpha (New England Biolabs) competent cells and grown overnight on LB agar plates supplemented with 50 mg mL 21 carbenicillin. Single colonies were picked and grown in LB medium at 37°C and shaking at 170 rpm in the presence of 50 mg mL 21 carbenicillin. The correct sequence of plasmids was checked by sequencing (SEQLAB Sequence Laboratories).
For AtHDH1 expression, pDest17-AtHDH1-w/o-28-AA was transformed in BL21-AI One Shot Chemically Competent E. coli cells (Thermo Fisher Scientific). In these cells, the T7-RNA polymerase gene is under the control of the araBAD promoter. Transformed cells were grown in LB medium supplemented with 50 mg mL 21 carbenicillin at 37°C until the OD 600 reached 0.6 to 1. AtHDH1 expression was induced using 0.05% (w/v) L-arabinose. Starting from this time point, bacteria were transferred to 16°C for 24 h in order to decrease the risk of inclusion body formation. Bacteria were lysed according to the manufacturer's instructions using BugBuster Master Mix (Merck Millipore). The soluble fraction was used for affinity purification using immobilized metal ion chromatography on Ni-NTA agarose (Qiagen). NaCl (1 M) and 20 mM imidazol as final concentrations were added to the supernatant to prevent the unspecific binding of non-His-tagged proteins. The beads were equilibrated and washed with 33 mM NaH 2 PO 4 , 2 M NaCl, and 53 mM imidazol. The protein was eluted using 50 mM NaH 2 PO 4 , 300 mM NaCl, and 400 mM imidazol. Desalting was achieved by using Microcon-30kDa centrifugal filters (Merck Millipore). After 10 min of centrifugation at 14,000g, 200 mL of 20 mM K 2 HPO 4 , pH 8, was added and again centrifuged for 10 min at 14,000g. The freshly desalted protein was immediately used to determine K m and K cat or 20% (v/v) glycerol was added for protein storage at 220°C for a short term.

Enzyme Activity Assays and Substrate Screening
The standard 3-hydroxyisobutyrate dehydrogenase assay took place in a buffer composed of 250 mM Tris-HCl, pH 8.5, 5 mM NAD + , 200 ng of AtHDH1 (from at least a minimum of three different enzyme batches from elution fraction 2), and 5 mM 3-hydroxyisobutyrate. Elution fraction 2 was used in order to have a very pure fraction for experiments. Assays were carried out in a total volume of 300 mL in the wells of a 96-well plate. The production of NADH (« 340 = 6.22 mM 21 cm 21 ) was measured at 340 nm using an Epoch Microplate Spectrophotometer (Biotek). Enzyme assays were carried out at 25°C. Kinetic values were corrected by values obtained in parallel control experiments (assays without substrate or added protein). Protein quantification was carried out using the Pierce Coomassie (Bradford) Protein Assay (Thermo Fisher Scientific).
The K m and V max values were determined at pH 8.5 by varying the concentration of the substrate while keeping the concentrations of the other components constant. At least three different enzyme batches were used, and kinetic constants were calculated using nonlinear regression (Graph Pad Prism 7; GraphPad Software).
To determine the pH optimum of AtHDH1, a buffer mixture was used containing 250 mM Tris-HCl, pH 7 to 9.5, 5 mM NAD + , and different amounts of recombinantly expressed AtHDH1 The reaction was started with 5 mM 3-hydroxyisobutyrate and took place at a final volume of 300 mL.
The temperature optimum of AtHDH1 was tested using a buffer including 250 mM K 2 HPO 4 3 KH 2 PO 4 , pH 7.5 (since the pH of a Tris-HCl buffer changes with different temperatures, we decided to use a potassium phosphate buffer).

SDS-PAGE and Western Blotting
Protein concentration was determined using the Pierce Coomassie (Bradford) Protein Assay (Thermo Fisher Scientific). SDS-PAGE was performed using precast gels from Bio-Rad. Gel runs were performed according to the manufacturer's instructions. After completion of the gel run, the gel was either stained using the Coomassie Blue colloidal procedure (Neuhoff et al., 1988) or electroblotted onto a nitrocellulose membrane for immunoblotting. Western blots were incubated with a horseradish peroxidase-conjugated 63 His, His-Tag Mouse Monoclonal Antibody (Proteintech Europe) and developed using nickel chloride combined with 3,39-Diaminobenzidine (DAB) according to Coventry et al. (1995). The Amersham ECL Rainbow Marker-High Range was used (GE Healthcare).

Protein Identification by MS
For identification, protein bands were cut out from polyacrylamide gels, and included proteins were fragmented into peptides by trypsin treatment as described previously (Klodmann et al., 2011). Peptide mixtures were purified, and peptides were separated by liquid chromatography and analyzed by tandem MS using a micrOTOF Q-II mass spectrometer (Bruker) as outlined before (Klodmann et al., 2011).

Blue Native PAGE and In-Gel Activity Staining
Blue native PAGE was performed employing a precasted 4-16% Bis-Tris Native Gel (Thermo Fisher Scientific). The blue native cathode buffer (50 mM Tricine, 15 mM BisTris, and 0.02% [w/v] Coomassie Blue G250, pH 7) and the blue native anode buffer (50 mM BisTris, pH 7) were used. The Coomassie Bluecontaining cathode buffer was replaced after half of the run by a cathode buffer without Coomassie Blue in order to strongly reduce excess Coomassie Blue. The gel run was carried out according to the manufacturer's instructions at 4°C. After the gel run was completed, in-gel activity assays were carried out by incubating gels in the dark at room temperature with 20 mL of staining solution containing 250 mM Tris-HCl, pH 8.5, 0.4 mg mL 21 nitroblue tetrazolium, 0.2 mM PMS, 2 mM NAD + , and 5 mM 3-hydroxyisobutyrate. As a native marker, the Amersham High M r Calibration Kit for native electrophoresis was used (GE Healthcare).

Gene Expression Analyses
To determine gene expression in lines DHDH1-2 and DHDH1-6, quantitative PCR analysis was carried out according to Heimann et al. (2013). First, total RNA was isolated from 100 mg of leaves using the GeneJET Plant RNA Purification Kit (Thermo Fisher Scientific). Subsequently, RNA was converted into first-strand cDNA. cDNA synthesis was performed with approximately 1 mg of total RNA and 50 pmol of random nonamer primer. Reactions were incubated for 5 min at 70°C and cooled down on ice before adding 200 units of Moloney murine leukemia virus reverse transcriptase (Promega) and 1 mM deoxyribonucleotide triphosphates in reaction buffer, as specified by the manufacturer. Quantitative PCR was performed on an ABI PRISM 7300 sequence detection system (Thermo Fisher Scientific) using SYBR Green fluorescence (Platinum SYBR Green QPCR Mix; Thermo Fisher Scientific). The following primers were used: DHDH1_qPCR_FW_2 (59-CTAGTGGCCGTTGTTGGAGCAGTGA-39) and DHDH1_qPCR_RV_2 (59-TCAGCTGAGGCTGCTGCGAGGT-39). Amplification conditions were 2 min of initial denaturation at 95°C, followed by 40 cycles of 15 s at 95°C and 1 min at 60°C. Afterward, a melting curve was recorded. General reaction conditions were 3 mM MgCl 2 and 200 nM of each oligonucleotide. The sizes of the amplified molecules were confirmed by gel electrophoresis. At least three biological replicates were carried out for each measurement.

Cultivation of Plants on Agar Plates and Root Toxicity Assays
Seeds of DHDH1-2, DHDH1-6, and Columbia-0 were surface sterilized and sown on one-half-strength Murashige and Skoog agar plates. After 2 d at 4°C, plates were transferred to a climate chamber (16/8 h of light/dark) for continuation of plant cultivation at 22°C and 85 mmol s 21 m 22 light. After 4 d in the climate chamber, 10 seedlings with approximately the same size and the same root length were transferred to one-half-strength Murashige and Skoog agar plates, which were supplemented with 250 and 500 mM Val or 50 and 100 mM Leu/Ile. Control plates did not contain any amino acids. After 10 d, plates were scanned, and the root lengths were determined with the AxioVision Microscope Software (Carl Zeiss). The results are derived from three different biological experiments each with 20 plants (two plates) per treatment (in total, 60 plantlets were measured per treatment and concentration).

Supplemental Data
The following supplemental materials are available.
Supplemental Figure S1. Alignment of At4g20930 and human 3hydroxyisobutyrate dehydrogenase using ClustalV.
Supplemental Figure S2. HDH1 has been identified in a mitochondrial fraction of Arabidopsis.
Supplemental Figure S4. Sequence coverage of HDH1 from Arabidopsis by peptides identified by MS.
Supplemental Figure S5. The reaction mechanism catalyzed by AtHDH1 is a sequential bibi mechanism.
Supplemental Figure S6. Affinity purification of recombinant At4g29120 and activity with 3-hydroxyisobutyrate as substrate Supplemental Figure S7. Identification of Arabidopsis lines homozygous for T-DNA insertions in At4g20930.
Supplemental Table S1. Proteins identified in the 75-, 37-, and 30-kD bands of the SDS gel shown in Figure 1 by MS using the SwissProt database.