Alterations of Mitochondrial Protein Assembly and Jasmonic Acid Biosynthesis Pathway in Honglian (HL)-type Cytoplasmic Male Sterility Rice*

Background: Cytoplasmic male sterility (CMS) is associated with mitochondrial defects. Results: Reduction of assembled mitochondrial protein complexes and altered jasmonic acid pathways were observed in the sterile line. Conclusion: CMS may be related to a mitochondrial complex assembly defect and abnormal jasmonic acid pathway. Significance: Our discoveries suggest a novel mechanism for CMS. It has been suggested that the mitochondrial chimeric gene orfH79 is the cause for abortion of microspores in Honglian cytoplasmic male sterile rice, yet little is known regarding its mechanism of action. In this study, we used a mass spectrometry-based quantitative proteomics strategy to compare the mitochondrial proteome between the sterile line Yuetai A and its fertile near-isogenic line Yuetai B. We discovered a reduced quantity of specific proteins in mitochondrial complexes in Yuetai A compared with Yuetai B, indicating a defect in mitochondrial complex assembly in the sterile line. Western blotting showed that ORFH79 protein and ATP1 protein, an F1 sector component of complex V, are both associated with large protein complexes of similar size. Respiratory complex activity assays and transmission electron microscopy revealed functional and morphological defects in the mitochondria of Yuetai A when compared with Yuetai B. In addition, we identified one sex determination TASSELSEED2-like protein increased in Yuetai A, leading to the discovery of an aberrant variation of the jasmonic acid pathway during the development of microspores.

It has been suggested that the mitochondrial chimeric gene orfH79 is the cause for abortion of microspores in Honglian cytoplasmic male sterile rice, yet little is known regarding its mechanism of action. In this study, we used a mass spectrometry-based quantitative proteomics strategy to compare the mitochondrial proteome between the sterile line Yuetai A and its fertile near-isogenic line Yuetai B. We discovered a reduced quantity of specific proteins in mitochondrial complexes in Yuetai A compared with Yuetai B, indicating a defect in mitochondrial complex assembly in the sterile line. Western blotting showed that ORFH79 protein and ATP1 protein, an F 1 sector component of complex V, are both associated with large protein complexes of similar size. Respiratory complex activity assays and transmission electron microscopy revealed functional and morphological defects in the mitochondria of Yuetai A when compared with Yuetai B. In addition, we identified one sex determination TASSELSEED2-like protein increased in Yuetai A, leading to the discovery of an aberrant variation of the jasmonic acid pathway during the development of microspores.
The world's food supply depends on a few high yield crops, such as rice, corn, and wheat. Hybrid seed generation is vital for high yield crop production. Cytoplasmic male sterility (CMS) 4 is a maternally inherited phenomenon observed in the higher plant kingdom. CMS leads to pollen abortion but does not affect female fertility and vegetative growth, which form the key technical base for hybrid seed production (1,2). A number of mitochondrial genes associated with CMS have been identified in the past, but few mechanistic details are known for this phenomenon. Most CMS-associated mitochondrial genes are chimeric genes composed of a fragment of a normal mitochondrial gene, encoding small and low abundance mitochondrial membrane proteins (3).
Based on limited biochemical studies of the molecular mechanisms of CMS, two theories have been proposed (4). The first is the "gain-of-function" theory, which is supported by the finding that T-URF13, an inner mitochondrial membrane protein, may form a pore and interfere with membrane activity in the cytoplasm of maize (5,6). In addition, ORF138, which forms a complex of over 750 kDa, has been proposed to act as a pore in the inner mitochondrial membrane in a sterile line of rapeseed (7). The second theory, the "loss-of-function" theory, is supported by the characterization of ORF522, which shares a similar N-terminal amino acid sequence with ORFB but competes with ORFB in binding to the ATP synthase (complex V) subunit in sunflowers to reduce the stability and efficiency of this complex (8).
There are three widely distributed CMS rice varieties in China, including "Honglian" (HL), the "Wild Abortive," and "Chinsurah Boro II/Taichung native 65." These CMS lines have differences in inheritance, morphology of abortive pollens, and restoration-maintenance relationships (9). HL CMS rice was developed by backcrossing red-awned wild rice (Oryza rufipogon) with the indica variety Lian Tang-Zao in Hainan China in the 1970s. The sterility of Yuetai A (YtA) is maintained by backcrossing with its isogenic fertile line Yuetai B (YtB). Restriction fragment length polymorphism analysis and BAC library screening of the mitochondrial DNA from YtA and YtB revealed the chimeric gene orfH79 located downstream of the atp6 gene in the sterile line (10 -12). Transgenic analysis showed that the ORFH79 protein is responsible for CMS in the sterile line (13). However, it remains unclear how the aberrant ORFH79 affects organization and function in mitochondria.
Plant mitochondria not only play a pivotal role in energy production but are also involved in multiple biological processes such as amino acid metabolism, biosynthesis of vitamins and lipids, and programmed cell death (14). Therefore, it is difficult to attribute a precise function to any specific CMS protein in mitochondria, especially when the CMS protein shares limited homology to any known protein. In searching for clues to the mechanism of CMS, past studies have used proteomics to obtain a global view of the composition and spatial distribution of mitochondrial proteins (8,15,16). In this study, quantitative proteomics was used to compare the mitochondrial proteome between YtA and its maintainer line YtB. Using blue native-PAGE (BN-PAGE) to separate protein complexes, we also quantitatively compared the protein composition of mitochondrial protein complexes. Compared with YtB, reduced quantities of proteins were detected in mitochondrial complexes in YtA, suggesting a defect in mitochondrial complex assembly. Western blotting indicated that ORFH79 protein location is associated with several large protein complexes, ranging from ϳ400 kDa to ϳ1.2 MDa. ATP synthase ␣ subunit (ATP1), an F 1 sector component of complex V, also belongs to protein complexes of similar size as ORFH79. Respiratory complex activity assays and transmission electron microscopy revealed functional and structural defects in the mitochondria of YtA when compared with YtB. Furthermore, our quantitative proteomic analyses revealed that one sex determination TASSELSEED2like protein was up-regulated in YtA, leading to the discovery that an aberrant jasmonic acid pathway existed in CMS rice.

EXPERIMENTAL PROCEDURES
Plant Materials-Seeds of Honglian CMS line rice YtA and its corresponding fertile maintainer line YtB were washed in 1% (v/v) bleach for 10 min and rinsed in distilled water. The seedlings were grown in vermiculite trays at 30°C, with the etiolated seedlings in the dark, and the green seedlings were grown in a 14-h light/8-h dark regime. The seedlings were watered daily and harvested at 10 days for experiments. YtA and YtB were grown in an experimental field of Wuhan University in the summer of 2009. Tassels at specific developmental stages were collected after plants had grown for ϳ3 months. Developmental stage of microspores was determined according to the method of Wan et al. (17).
Rice Mitochondrial Isolation-Isolation of rice mitochondria was performed by differential centrifugation followed by continuous Percoll gradients as described by Heazlewood et al. (18). The amount of mitochondrial protein was determined using the modified Lowry protein assay kit (Pierce).
Protein Digestion and Dimethyl Labeling-For mitochondrial total protein analysis, a fixed amount of enriched mitochondria (1 mg from YtA or YtB, digested separately) were lysed with buffer containing 25 mM Tris-HCl, pH 7.6, 150 mM NaCl, 1% Nonidet P-40, 1% sodium deoxycholate, 0.1% SDS, and protease inhibitor mixtures (Roche Applied Science). Three volumes of 50% acetone, 50% ethanol, 0.1% acetic acid were added to the lysates followed with 2 h of incubation on ice. The protein pellets were resuspended in 8 M urea, 0.2 M Tris, pH 8.0, 4 mM CaCl 2 and sonicated for several seconds to fully resuspend the proteins.
Samples were reduced with 10 mM DTT for 1 h and alkylated with 40 mM iodoacetamide in the dark for 30 min. After diluting the urea concentration to 2 M, trypsin (Promega) (1:50; trypsin/ mitochondrial proteins) was added, and the digestion was carried out at 37°C overnight. After digestion, urea and salts were removed by C 18 Sep-Pak columns (Waters).
For the stable isotope dimethyl labeling, the digested peptide samples from YtA and YtB were reconstituted separately with 200 l of CH 3 COONa, pH 5.9. Eight microliters of CH 2 O (light labeled) and CD 2 O (heavy labeled) were added to YtA and YtB samples, respectively. Eight microliters of 0.6 M NaBH 3 CN was added to both samples and incubated at room temperature for 1 h. To quench the reaction, 32 l of 1% (v/v) ammonia solution and 16 l of 5% formic acid were added to both samples on ice. Light labeled YtA and heavy labeled YtB samples were mixed at the ratio of 1:1 and desalted before strong cation exchange (SCX) fractionation.
For the SCX fractionation, the mixed peptides were resuspended in buffer A containing 5 mM KH 2 PO 4 , 20% acetonitrile, pH 2.7. SCX was performed on a polysulfoethyl column (2.1 ϫ 50 mm, 5 m ϫ 200 Å) using a KCl gradient from 0 to 0.5 M to fractionate the peptides. Nine fractions were collected and desalted with C 18 ZipTip (Millipore) before MS analysis.
For protein in-gel digestion, BN gel lanes containing complexes above 100 kDa were divided into six fractions. Each fraction was cut into ϳ1-mm 3 pieces. The gel pieces were washed with 50% acetonitrile, 100 mM NH 4 HCO 3 , pH 8.0, three times, reduced with 10 mM DTT, alkylated with 55 mM iodoacetamide in the dark, and digested in-gel with trypsin. The tryptic peptides from YtA and YtB were dimethyl-labeled as mentioned above, and the corresponding fractions were mixed together. Finally, the samples were desalted by C 18 Sep-Pak columns (Waters).
LC-MS/MS and Data Analysis-A QSTAR ELITE mass spectrometer (Applied Biosystems) coupled with a nanoflow HPLC system (Tempo TM , Applied Biosystems) was used for relative quantitation of mitochondrial proteins from YtA and YtB rice. The LC-MS method was the same as used by Chen et al. (20). Raw data from QSTAR ELITE were analyzed with MASCOT Daemon software (Version 2.2.2, Matrix Science) using a local MASCOT engine. The data were searched against a rice database (Release 6.1, The Institute for Genomic Research, Rice Genome Annotation Project), including mitochondrial and plastid proteins plus ORFH79 (21) with added dimethyl masses. Carbamidomethylation of cysteine was set as a fixed modification, and oxidation of methionine and phosphorylation of serine, threonine, and tyrosine were set as variable modifications. Peptide mass tolerance was set as 200 ppm and 0.4 Da. The peptide charge was set to 2ϩ and 3ϩ, allowing for up to two missed cleavages, and the significance threshold was set at p Ͻ 0.05.
After the Mascot search, the rov data obtained from the database search were opened by Mascot Distiller (Version 2.3.2.0) for quantitation. For the quantitation analysis, we set fraction, correlation, and standard error at 0.5, 0.9, and 0.2, respectively. The false-positive rates of peptide spectrum matches were determined by data search against a decoy database. The peptide ratios were calculated as weighted average ratios (ion intensity versus ratio) if several spectra for the same peptide were available. The median of all quantitation data from all peptides was used to normalize the peptide ratios. Protein ratios were also calculated as the geometric mean of the peptide ratios for a protein. Student's t test were performed, and p values were calculated using software SPSS 13.0.
Antibodies-For ORFH79 antibody production, the sequence of orfH79 lacking the 5Ј-terminal 66-bp region homologous to coxI was amplified by PCR using primers 5Ј-CCG-GAATTCTTCGGTGTCGTTGTAGGA-3Ј (the underlined letters show the EcoRI site) and 5Ј-ACGCGTCGACACCACT-GTCCTGTCTTTCTT-3Ј (the underlined letters show the SalI site). This PCR fragment was digested with EcoRI and SalI and subcloned into the EcoRI and SalI sites of pET-28a (Merck) for bacterial expression. The recombinant protein was purified by affinity chromatography and then injected into rabbits. The anti-ORFH79 serum was generated by NewEast TM Biosciences. ATP1 antibodies were a gift from Dr. T. Elthon, and NAD9 antibodies were a gift from Prof. J. M. Grienenberge. PHB2, HSP60, and CYTC antibodies were from Proteintech Group Inc. ORFB antibodies were from Beijing Protein Institute.
Immunoprecipitation and Immunoblotting-Mitochondrial membrane proteins (1 mg) were resuspended at a concentration of 2.5 mg of protein/ml in lysis buffer for BN-PAGE and solubilized with DDM at a protein/detergent ratio of 1:2 (g/g). After centrifugation, the supernatant was used for immunoprecipitation with the co-immunoprecipitation kit (Pierce Lot. 26149) according to the manual.
Transmission Electron Microscopy-Leaf squares cut from etiolated seedlings at 10 days were fixed, embedded, cut, and soaked as described by Sosso et al. (23). Microscopy used an H-8100 electron microscope (Hitachi H-8100) at an accelerating voltage of 150 kV.
Quantitation of Jasmonic Acid and Its Biosynthetic Intermediates-Quantitation of the jasmonic acid pathway intermediates was performed by a recently described method (24). Rice seedlings or tassels (0.25 g) were frozen in liquid nitrogen, ground to a fine powder, and extracted with 1.25 ml of 80% (v/v) methanol at 4°C for 12 h. Then 9,10-dihydrojasmonic acid (50 ng/g), d 2 -indole-3-acetic acid (50 ng/g), and d 6 -abscisic acid (50 ng/g) were added to plant samples as internal standards, prior to grinding. After centrifugation (30,

Mitochondrial Proteome and Protein Complex Analysis-
Our proteomics strategy is outlined in Fig. 1. We purified rice mitochondria from YtA and YtB using differential centrifugation and a Percoll gradient. Mitochondrial samples were further separated into two groups. Each group contained 1 mg of a fixed amount of mitochondrial protein. One group was used in quantitative mitochondrial total proteome analysis, which included both mitochondrial membrane proteins and the soluble proteins. The other group containing 1 mg of intact mitochondria was used to analyze mitochondrial protein complexes after BN-PAGE.
In the total mitochondrial proteome group, nine SCX fractions were collected and labeled with stable isotope dimethyl labels for quantitative LC-MS/MS analysis. After data search-ing against the rice database from The Institute for Genomic Research, a total of 522 proteins were identified in two biological replicates (supplemental Table S1, with duplicated data alignment shown in supplemental Fig. S1). The false-positive rates of peptide-spectrum matches determined by data search against a decoy database were 0.92 and 1.96%, respectively, for the two biological replicates. Among these 522 proteins, 245 proteins were confirmed as mitochondrial components through comparison with a list of 322 nonredundant rice mitochondrial proteins previously compiled ( Fig. 2A; a list of these proteins is shown in supplemental Table S1A) (16).
In the mitochondrial complexes group, starting from 1 mg of isolated mitochondria, protein complexes were extracted and separated on BN gels. Protein bands above 100 kDa were divided into six fractions, and labeled as "BN1" to "BN6", respectively (Fig. 2B). In-gel tryptic digestion was performed on each fraction, and peptides were labeled with either the light or heavy dimethyl group and mixed for LC-MS/MS analysis. In all the BN fractions combined, we identified 191 proteins with 138 proteins confirmed as mitochondrial proteins ( Fig. 2A; a list of these proteins is shown in supplemental Table S2A, with duplicated data alignment shown in supplemental Fig. S1). The falsepositive rates of peptide spectrum matches determined by data search against a decoy database were 0.71 and 0.72%, respectively, for the two mitochondrial protein complex analysis data sets. Among the mitochondrial proteins in the two groups, 122 proteins were in common ( Fig. 2A).
To establish a structural linkage between the protein subunits and their relevant complexes, a heat map was generated showing the number of identified peptides from each protein and their corresponding BN gel location (supplemental Fig. S3). Proteins of mitochondrial complex I and complex III 2 were concentrated in two fractions, BN1 and BN3. Therefore, we concluded that the relative protein ratio in a BN fraction should correlate to the ratio of their corresponding complexes. Mitochondrial complex V was identified primarily in fractions BN2, BN3, and BN4. This distribution pattern is consistent with previous reports that there are three forms for this complex (19,25).
The function of all the identified proteins was assigned to 12 functional categories based on functional assignments previously established (Fig. 2C) by Huang et al. (16). From the number of proteins identified in each category, as expected, mitochondrial complex associated proteins or membrane proteins ("oxidative phosphorylation complexes" and "carriers and transporters") were well represented in data from BN analysis. Although some proteins normally not considered as assembled complex members were also identified from BN-PAGE fractions, their distributions were not BN fraction-specific, indicating these proteins may be loosely attached to mitochondrial complexes.
Mitochondrial Protein Assembly Analysis-A biological process is usually not performed by a single protein. Multiple proteins, either as an assembled complex or working together, are often needed to perform biological functions. A good example is the mitochondrial oxidative phosphorylation system that consists of a series of protein complexes and super complexes (26). Therefore, in addition to the total mitochondrial proteome analysis and comparison, we also examined how the mitochondrial protein complexes were formed and whether there were any differences between the YtA and the YtB samples.
The stable isotope labeling technique we employed provided relative quantitative information. The light/heavy (L/H) ratio from each peptide can provide information about protein levels. We analyzed the L/H ratio (or YtA/YtB ratio) for both the total and assembled form of mitochondrial proteins. The num- ber of identified proteins that fall into each 0.25-interval of log 2 (L/H) was displayed in Fig. 3, A and B. In the group of total mitochondrial proteins, proteins exhibiting positive log 2 (L/H) (L,YtA; H, YtB) was almost the same number as those having negative values (Fig. 3A). Most were localized between Ϫ1.0 and 1.0, indicating that the sterile line (YtA) contained a mitochondrial proteome similar to the fertile line, which was confirmed by the total mitochondria SDS-PAGE assay (supplemental Fig. S2). However, the proteins in assembled form showed a strong deviation in the negative direction, most of which fell into the section between Ϫ3.5 and 0 (Fig. 3B). This difference suggested that the total amount of mitochondrial proteins were nearly the same between the sterile line (YtA) and the fertile line (YtB) but that much less functional mitochondrial complexes were assembled in YtA.
To identify proteins which were selectively reduced in mitochondrial assembled protein complexes of YtA, we created a scatter plot based on the log 2 (L/H) (or YtA/YtB) values of 122 proteins shared by both total mitochondrial proteins (y axis) and assembled mitochondrial proteins (x axis; Fig. 3C). In this plot, if a protein is found along the diagonal line (y ϭ x line), it would suggest that the protein has the same relative ratio distribution in the total mitochondrial protein and in the assembled mitochondrial complexes. Instead, we found 57% of these data points fell into quadrant 2 (Q2), which has a positive log 2 (L/H) in total mitochondrial proteins but a negative log 2 (L/H) in assembled complexes. This indicated that these proteins were increased in total mitochondrial form in the sterile line (YtA) but decreased in the assembled form. For example, for the complex V, 11 subunits were identified in both experiments but exhibited a lower ratio as assembled mitochondrial proteins in different BN fractions (Fig. 3C).
Statistic analysis of the YtA/YtB ratio indicated that ϳ12% of mitochondria total protein were considered as significantly different from a ratio of 1, but ϳ63% of mitochondria assembled protein were considered as significantly different from ratio of 1 (supplemental Tables S1 and S2). This provided additional evidence suggesting that although the total amount of mitochondrial proteins was similar between the sterile line (YtA) and the fertile line (YtB), assembled mitochondrial protein complexes were reduced in YtA.
Discovery of ORFH79 Complexes-Previous studies have clearly linked the CMS protein ORFH79 to YtA, but we failed to identify the ORFH79 protein in our proteomics analyses described above. Because ORFH79 was localized to the mitochondrial membrane (13), it was possible that ORFH79 was in low abundance and below the limits of MS detection. Therefore, in searching for the ORFH79 protein, we performed Western blotting, which is normally more sensitive (Fig. 4, A-C). Based upon the Western blot data, the ORFH79-containing complexes were estimated to be ϳ350, 600, and 900 kDa and 1.1 MDa (Fig. 4A). Use of Western blotting and two-dimensional BN-SDS-PAGE showed ORFH79 protein matches in the vertical location (a potential indication of the same complex) spanning from the BN1 band to the BN4 band.
Based on our proteomics results, many well known protein complexes were identified from BN1 to BN4 fractions, including the prohibitin (PHB) complex, complex I, the heat shock protein (HSP) complex, complex V, complex III 2 , and the F 1 part of complex V (7,19,25). It was possible that ORFH79 was co-localized with these complexes. ATP1 protein, an F 1 sector component of complex V, is often used as a marker for complex V (27). We unexpectedly found by Western blotting that there was a correlation of ATP1 with that of ORFH79 on the BN gel ( Fig. 4, B and C). Western blot analysis after two-dimensional BN-SDS also showed minor differences between YtA and YtB on ATP1 staining (Fig. 4, B and C), suggesting possible structural differences of complex V between YtA and YtB.
As an alternative strategy, anti-ORFH79 serum was used to immunoprecipitate mitochondrial membrane proteins, and Western blotting was used to detect PHB2, HSP60, NAD9, ATP1, and CYTC (members of the PHB complex, the HSP complex, complex I, complex V, and complex III 2 , respectively). As expected, ORFH79 protein was immunoprecipitated and detected in YtA but not in YtB. Unfortunately, none of the other proteins tested were found to be YtA-specific, with a number of proteins detected in the pre-immune serum IP lane (supplemental Fig. S5). Therefore, no ORFH79-specific protein partner information could be derived from these data.
Enzyme Activities of Several Mitochondrial Respiratory Complexes-Enzyme activities of several mitochondrial respiratory complexes were compared in-gel and in-solution between the fertile and sterile lines. The activity of oxidative phosphorylation complexes, particularly complex V, was reduced in the FIGURE 3. Comparison of the ratio correlation between total and assembled mitochondrial proteins. A, protein L/H ratio (or YtA/YtB ratio) distribution for the total mitochondrial proteins. y axis shows the number of proteins identified that fall into each 0.25-interval of log 2 (L/H). B, protein L/H ratio (or YtA/YtB ratio) distribution for the assembled form of mitochondrial proteins. y axis shows the number of proteins identified that fall into each 0.25-interval of log 2 (L/H). C, scatter plot based on the log 2 (L/H) (or YtA/YtB) value of 122 proteins shared by both total mitochondrial proteins (y axis) and assembled mitochondrial proteins (x axis). Each spot represents a protein, and its assigned functional category was indicated. Four quadrants (Q1-Q4) were assigned as indicated. sterile line (Fig. 5). This observation suggested that reduction in enzyme activity was accompanied by a defect in complex assembly in the sterile line (YtA), which provided a correlation between enzyme activity and the quantitative proteomic analyses.
Additional assays were performed with mitochondria purified from transgenic callus carrying the orfH79 gene. Silver staining indicated little or no difference between total proteins of the orfH79 transgenic line (Trans79) and the blank vector transgenic line (TransBL). Western blotting showed equal amounts of ATP1 in both transgenic lines. However, the in-gel enzyme activity assay showed a reduced activity of complex V in Trans79 (supplemental Fig. S4). Taken together, the results suggested that the assembly defect is directly linked to the presence of ORFH79.
Altered Mitochondrial Morphology in the Sterile Line-It has been reported that mitochondrial ultrastructure is closely related to its function (28). The results of this study indicated that mitochondria of the sterile line (YtA) contained reduced functional protein complexes, while gaining an additional ORFH79-containing complex. Transmission electron microscopy was used to compare mitochondrial ultrastructure between YtA and YtB in etiolated seedlings at 10 days (Fig. 6). No significant differences in the number of mitochondria per cell were observed. However, most of the mitochondria in YtA displayed a spherical morphology with reduced cristae, whereas in the control YtB, sausage-shaped mitochondria were observed with normal cristae. The morphology of etioplasts in YtA and YtB were similar (Fig. 6). These results suggested that complex assembly defects in the sterile line may have influenced its morphology.
Abnormal Jasmonic Acid Biosynthesis in the Sterile Line-From total mitochondrial proteomic analyses, only two proteins were identified with a greater than 2-fold increase in the sterile line YtA (supplemental Table S1A). These include a protein of unknown function (Os05g51540.1, 2.5-fold), and sex determination protein TASSELSEED-2 (Os07g46920.1, 3.2fold) (16,21). Mass spectra for the identification and quantitation of sex determination protein TASSELSEED-2 is shown in supplemental Fig. S6. A homolog of this sex determination protein TASSELSEED-2 in maize, TASSELSEED2 (TS2), was reported to be involved in floral organ determination (29) and may participate in JA biosynthesis (30,31).
We hypothesized that the up-regulation of the TS2 homolog in YtA may influence the JA biosynthesis in this sterile line (Fig.  7A). Therefore, the abundances of major components in the JA pathway were compared between the etiolated seedlings of YtA and YtB. Although the JA abundance was too low to compare, we observed significant up-regulation of several of its precursors such as hydroxyoctadecatrienoic acid and cis-(ϩ)-12-oxophytodienoic acid in the sterile line YtA compared with YtB (Fig. 7B). In green seedlings, there were also various alterations in JA pathway intermediates between the YtA and YtB lines (Fig. 7B).
To investigate whether these JA precursor differences are also present in reproductive organs, tassels from meiosis in the bicellular pollen stage were collected, and the JA-related components were analyzed (Fig. 7C). Indole acetic acid and abscisic acid, two plant hormones not related to JA pathways, were included as controls (Fig. 7C). In the indole acetic acid and abscisic acid samples, only minor differences were observed between the YtA and YtB lines in the four stages examined. In contrast, significant differences among all the JA precursors in all the stages except the bicellular pollen stage were detected. In the meiosis state, excess of hydroxyoctadecatrienoic acid and cis-(ϩ)-12-oxophytodienoic acid accumulation were found in the YtA line compared with the YtB line. In the tetrad stage, OPC-6:0, the last precursor leading to JA, was elevated in YtA. In the uninucleate stage, most of the JA precursors except OPC-6:0 were nearly undetectable in YtB but not in YtA. In the final bicellular pollen stage, all the JA precursors were at similar levels in both lines. Taken together, the results clearly indicated that the JA biosynthetic pathway is differentially regulated in the sterile line.
Additional Altered Mitochondrial Proteins-Electron transfer flavoprotein subunit ␤ (Os04g10400.1) and superoxide dismutase (Os05g25850.1), two proteins that involved in reactive oxygen species related function, are also found as being up-reg-

DISCUSSION
How CMS-associated proteins cause male sterility in plants is an important but difficult mechanism to characterize, mainly due to the technical problems of obtaining sufficient quantities of viable reproductive tissue in a short time (3,32). Studies using mitochondria isolated from vegetative tissues that constitutively express CMS-associated proteins are therefore an excellent system to study the mechanism of aborted pollen development (7,32,33). In this study, a quantitative proteomics approach was used to compare intact mitochondrial proteins and mitochondrial protein complexes between the HL-type CMS rice YtA and its isogenic fertile line YtB. A major finding of this study is that, compared with YtB, YtA shows decreases in the ability to assemble mitochondrial complexes.
The finding that the ORFH79 protein can be a member of protein complexes ranging from ϳ400 kDa to ϳ1.2 MDa was unexpected. Another CMS protein, ORF138 in rapeseed, was found as part of a 750-kDa complex but not associated with a wide range of complexes (7). A previous study showed that the membrane potential is decreased in YtA (13). Whether this membrane potential decrease is related to ORFH79 remains unclear. When mitochondrial lysates of YtA were analyzed by SDS-PAGE in nonreducing conditions, no evidence of oligomerization could be found (data not shown). It is therefore unlikely that the ORFH79 protein acts as an uncoupling protein similar to ORF138 in rapeseed and T-URF13 in T-maize (5,7).
Several independent reports have suggested a linkage between ORFH79 and complex V. First, a phenotype of programmed cell death in microspores was observed in YtA (12). It is possible that ORFH79 acts in a similar manner as ORF522, which is thought to compete with normal subunits of complex V and result in programmed cell death in tapetal cells in PET1-CMS sunflower (8,34). Second, the amount and activities of each form of complex V were reduced in the sterile line (Figs. 3 and 5). Third, our BN-PAGE and two-dimensional BN-SDS-PAGE analyses showed that ORFH79-containing complexes co-migrated with a different ATP1 protein, which is a marker for the F 1 sector of complex V (Fig. 4). Fourth, complex V is one of the factors that could affect different mitochondrial morphologies, such as onion-shaped and reduced cristae (35). YtA mitochondria are mostly spherical shaped and hold reduced cristae, which implies that ORFH79 might change the ultrastructure of mitochondria by interfering with complex V. These results are suggestive, but further biochemical and structural studies are needed fully resolve whether ORFH79 proteins are involved in the assembly of complex V.
The up-regulation of one short chain alcohol dehydrogenase, which was annotated as sex determination protein TASSEL-SEED-2, was another important contribution in understanding the mechanism of HL-CMS. Furthermore, the JA biosynthetic pathway has been shown to be altered in YtA. JA biosynthesis is critical for pollen development and anther dehiscence, so mutation of several enzymes involved in this pathway could lead to male sterility (36).
In summary, our proteomic analyses of HL-type CMS rice mitochondria has provided important new information regarding the biochemical mechanisms underlying the CMS phenotype. From ORFH79 to alterations in the mitochondrial protein complex and to changes in JA production, it is likely that HLtype CMS is the consequence of multiple alterations in biological processes. Further study is needed to connect these processes together, to determine whether there is a single underlying genetic cause, or multiple genetic defects involved in this disorder.