Rice FLOURY SHRUNKEN ENDOSPERM 5 Encodes a Putative Plant Organelle RNA Recognition Protein that Is Required for cis-Splicing of Mitochondrial nad4 Intron 1

The sequences of several important mitochondrion-encoded genes involved in respiration in higher plants are interrupted by introns. Many nuclear-encoded factors are involved in splicing these introns, but the mechanisms underlying this splicing remain unknown. We isolated and characterized a rice mutant named floury shrunken endosperm 5 (fse5). In addition to having floury shrunken endosperm, the fse5 seeds either failed to germinate or produced seedlings which grew slowly and died ultimately. Fse5 encodes a putative plant organelle RNA recognition (PORR) protein targeted to mitochondria. Mutation of Fse5 hindered the splicing of the first intron of nad4, which encodes an essential subunit of mitochondrial NADH dehydrogenase complex I. The assembly and NADH dehydrogenase activity of complex I were subsequently disrupted by this mutation, and the structure of the mitochondria was abnormal in the fse5 mutant. The FSE5 protein was shown to interact with mitochondrial intron splicing factor 68 (MISF68), which is also a splicing factor for nad4 intron 1 identified previously via yeast two-hybrid (Y2H) assays. Fse5 which encodes a PORR domain-containing protein, is essential for the splicing of nad4 intron 1, and loss of Fse5 function affects seed development and seedling growth.


Background
According to endosymbiotic theory, mitochondria originated from α-proteobacteria ancestors (Gray 1999). Throughout the long-term evolutionary history, most mitochondrial genes have been transferred to the nuclei of their host cells (Timmis et al. 2004). Only a minor portion of mitochondrial genes are retained in angiosperms, and those genes are involved in the electron transport system or encode ribosomal proteins and tRNAs (Kubo and Newton 2008). After these mitochondrial genes are transcribed, their transcripts are processed mainly by nuclear-encoded proteins. This posttranscriptional processing generally includes intron splicing, RNA editing, cleavage, and maturation (Small et al. 2013;Hammani and Giegé 2014).
According to their structural features and splicing mechanisms, mitochondrial introns in flowering plants can be divided into group I members and group II members, with the latter being more predominant (Bonen 2008). Group II introns are involved in both cis-splicing and trans-splicing. Cis-splicing occurs within one pre-mRNA molecule, whereas trans-splicing occurs between two pre-mRNA molecules (Sharp 1987;Lasda and Blumenthal 2011).
Members of another protein family characterized by their possession of the plant organelle RNA recognition (PORR) domain play an important role in intron splicing in plant organelles. The PORR domain was previously referred to as the "domain of unknown function 860" (DUF860) (http://pfam.xfam.org/family/PF11955) but was later renamed the PORR domain (Kroeger et al. 2009).
To date, several PORR domain-containing proteins have been identified in plants. AtRPD1, a member of the PORR/DUF860 family, has a role in prearranging the maintenance of active cell proliferation during root primordium development. Disruption of the ROOT PRIMORDIUM DEFECTIVE 1 (RPD1) gene causes embryogenesis arrest at the globular to transition stages. In silico structural characterization of RPD1 and RPD1-like proteins implied their possible involvement in various regulatory functions through DNA binding, RNA binding, and protein-protein interactions (Konishi and Sugiyama 2006). At4g08940, which encodes a PORR domain-containing protein, is expressed in response to oxidative stress. Compared with wild-type (WT) plants, transgenic At4g08940overexpressing plants were more tolerant to paraquat and cold and less tolerant to t-butyl hydroperoxide and salinity (Luhua et al. 2008). AtWTF9, a mitochondrion-localized PORR domain-containing protein, is required for splicing introns in ribosomal protein L2 (rpl2) and cytochrome c biogenesis F C (ccmF C ). T-DNA insertion alleles wtf9-1 and wtf9-2 caused severely stunted shoots and roots; both homozygous mutants were shown to survive to flowering, but the flowers were small and produced only a few milky aborted seeds (Colas des Francs- . A recent study showed that mitochondrial HSP60s interact together with WTF9 to regulate intron splicing of ccmF C and rpl2 (Hsu et al. 2019).
ZmWTF1, a chloroplast-targeted PORR domaincontaining protein, is required for splicing chloroplastencoded introns. wtf1-1, a relatively weak mutant, exhibited a pale green phenotype, whereas mutants wtf1-3 and wtf1-4 with null alleles were albino (Kroeger et al. 2009). ZmEMP6, a PORR domain-containing protein located in the mitochondria, is required for both endosperm and embryo development, but the target intron is unclear (Chettoor et al. 2015). To date, no PORR domain-containing protein has been identified in rice.
Here, we report the isolation of Fse5 in rice though map-based cloning. The Fse5 allele encodes a mitochondrion-targeted PORR domain-containing protein that is expressed constitutively in various tissues. Cis-splicing of nad4 intron 1 is abolished in the fse5 mutant, and mitochondrial function and structure are consequently disrupted. Like in other mutants whose mutations cause defective splicing of mitochondrial intron(s), fse5 seed development and seedling growth were affected. Mature fse5 seeds exhibited a floury, shrunken phenotype and either failed to germinate or produced weak seedlings that died within 1 month. Taken together, these results indicated that Fse5 plays an essential role in seed development and subsequent seedling growth.

fse5 Seeds Have a Floury, Shrunken Phenotype
An N-methyl-N-nitrosourea (MNU)-treated population of WT japonica cultivar (cv) W017 was generated to study rice endosperm development. A mutant with floury, shrunken endosperm was selected and named floury shrunken endosperm 5 (fse5) . In addition to the abnormal phenotype, the seeds of the fse5 mutants failed to germinate or produced seedlings which grew slowly and died within 1 month. Homozygous fse5 seeds are presumably reproduced by heterozygous plants (+/fse5) whose progenies have approximately one-quarter seeds with a mutant phenotype (Additional file 1: Figure S1 and Table S1). Compared with those of the WT W017 seeds, both the 1000-grain weight and grain thickness of mature fse5 seeds were significantly reduced, but there was no difference in grain length or grain width (Additional file 1: Figure S2A-S2D). In the fse5 mutant, seed starch and amylose contents were lower (Additional file 1: Figure S2E-S2F), whereas the protein and lipid contents were higher than in the WT (Additional file 1: Figure S2G-S2H). These data showed that the synthesis of seed storage products, especially starch, was significantly affected in fse5 seeds.
To evaluate defects at the cellular level, observations were made on mature and immature endosperm cells of the WT and fse5 seeds. Scanning electron microscopy showed that starch grains of mature fse5 seeds were smooth and loosely packed, whereas those of WT seeds had sharp edges and were tightly assembled ( Fig. 1e-h). Compared with those in the WT seeds, starch grains in the fse5 seeds were irregular and loosely assembled in the developing endosperm cells (Fig. 1i-l). Taken together, these data indicated that the fse5 mutation affected starch grain development.

The fse5 Mutation Causes Embryo and Seedling Death
Detailed examination of the germination of the mutant and WT seeds showed that fse5 seeds with hulls could not germinate, so dehulled seeds were used in further experiments. Compared with that of the WT seeds, the germination percentage of the fse5 seeds was significantly lower (Additional file 1: Figure S2I), and the plumules and radicles of the latter grew more slowly ( Fig. 2a-b). Compared with 94.0% of the WT seeds, only 35.7% of the mutant seeds had produced seedlings by 9 days after sowing (DAS) in the soil (Additional file 1: Figure S2J). The height of fse5 seedlings reached only one-fourth of that of the WT seedlings (Fig. 2c, Additional file 1: Figure S2K). Moreover, all the fse5 seedlings were dead by 30 DAS (Fig. 2d). Taken together, these data showed that the fse5 mutants exhibited seedor seedling-lethal phenotypes.
To determine the cause of seed or seedling lethality, the embryo structures were observed. Longitudinal sections of seeds after 24 h of imbibition with water revealed clear differentiation in the WT embryos but little change in the fse5 mutant embryos (Fig. 2e-f). Triphenyl tetrazolium chloride (TTC) staining indicated strong red staining of the WT embryo region, indicating active dehydrogenase activity (Brown et al. 1987), but comparatively little staining of the same tissues in fse5 seeds ( Fig.   Fig. 1 Phenotypes of mature and developing seeds of the WT and fse5 mutant. a and b Mature WT (a) and fse5 (b) seeds. Bars, 1 mm. c and d Transverse sections of mature WT (c) and fse5 (d) seeds. Bars, 0.5 mm. e-h Scanning electron microscopy of transverse sections of mature WT (e, g) and fse5 (f, h) seeds. Bars, 100 μm in (e) and (f); 10 μm in (g) and (h). i-l Iodine-stained semithin endosperm sections of wt (i, k) and fse5 (j, l) seeds at 12 DAF. Bars, 15 μm in (i) and (j); 7.5 μm in (k) and (l) 2g-h). These results indicated defective embryo differentiation and low seed vigor in the mutant.

Positional Cloning of Fse5
An F 2 population originating from a cross between a heterozygous (+/fse5) plant and Nanjing 11 (indica) was used to map the Fse5 locus. Based on the data from 92 F 2 individuals, the locus was mapped to a 7.2 cM region flanked by RM242 and RM257 on the long arm of chromosome 9. Nine hundred and seventy-eight F 2 individuals were used to further delimit the Fse5 locus to a 54 kb region between insertion/deletion (In/Del) markers wi-10 and wi-11. Nine open reading frames (ORFs) were predicted in this region ( Fig. 3a) (http:// rice.plantbiology.msu.edu). Sequence analysis revealed a 408 bp deletion in LOC_Os09g29760 in the fse5 mutant compared with the WT. The deletion caused a truncation of 136 amino acids (aa) in the fse5 mutant transcript (Fig. 3b). Thus, LOC_Os09g29760 was considered to be the candidate gene for Fse5.
To further confirm these results, a vector containing the native promoter and WT coding DNA sequence (CDS) of LOC_Os09g29760 was constructed and introduced into the fse5 mutant. Quantitative RT-PCR (qRT-PCR) showed that the expression level of LOC_ Os09g29760 was rescued in the positive transgenic lines (Fig. 3c). The dehulled seeds of these lines were transparent and plump, similar to those of the WT (Fig. 3d). Moreover, seedlings grown from these seeds exhibited normal phenotypes (Fig. 3e). Further validation was acquired from LOC_Os09g29760 knockout (KO) lines produced by CRISPR/Cas9 (Additional file 1: Figure S3A-S3B). The dehulled seeds from four independent KO lines segregated for both normal and mutant seeds in a pattern similar to that of the original mutant (Additional file 1: Figure S3C). Therefore, LOC_Os09g29760 was demonstrated to be the Fse5 allele.

Fse5 Encodes a PORR Domain-Containing Protein
Sequence analysis showed that the cloned Fse5 allele contains a single exon (Fig. 4a) and encodes a putative protein of 399 aa (http://rice.plantbiology.msu.edu). The protein was annotated as containing a PORR domain (https://www.ncbi.nlm.nih.gov; https://ricexpro.dna.affrc. go.jp), and the truncation of 136 aa encoded by the 408 bp fragment deleted in the fse5 allele was within this domain (Fig. 4b). FSE5 was named OsPORR1 because it was the first PORR domain-containing protein identified among the 17 putative PORR domain-containing proteins in rice (Kroeger et al. 2009;Colas des Francs-Small et al. 2012;Chettoor et al. 2015) (Fig. 4c). Multiple sequence alignment showed that 6 aa (Leu 172 , Glu 305 , Leu 307 , Phe 344 , Tyr 345 , and Leu 357 ) are highly conserved among all 17 members of the rice family of PORR domain-containing proteins, and all six aa are within the PORR domain, suggesting that they are essential for the function of PORR domain-containing proteins (Fig. 4d). Five of these aa (Leu 172 is excluded) are present within The orthologs of OsPORR1 in maize and Arabidopsis are GRMZM2G048392 (ZmEmp6) and At4g08940, respectively (Kroeger et al. 2009;Chettoor et al. 2015). The mutation of ZmEmp6 causes early arrest in grain development, and compared with WT plants, At4g08940-overexpressing transgenic plants are more tolerant to paraquat and cold and less tolerant to t-butyl hydroperoxide and salinity (Luhua et al. 2008;Chettoor et al. 2015). However, the mechanisms underlying these effects remain unknown.

Spatiotemporal Expression and Subcellular Localization of OsPORR1
qRT-PCR analysis showed that OsPORR1 was expressed in the roots, stems, leaves, leaf sheaths, panicles, developing seeds and seedlings according to the multiple phenotypic characteristics of the fse5 mutant (Fig. 5a). A Fig. 3 Positional cloning of the Fse5 locus. a The Fse5 locus was restricted to a 54 kb region flanked by markers wi-10 and wi-11 on chromosome 9 (Chr. 9) and included nine ORFs. Both markers and numbers of recombinants are shown. b Mutation site in LOC_Os09g29760 in the fse5 mutant. A deleted 408 bp fragment near the end of the exon leads to a 136 aa truncation in the putative encoded protein in the fse5 mutant. The red letters and dotted lines represent the deletion in LOC_Os09g29760 or its encoded protein in the fse5 mutant. The yellow and orange boxes indicate the 5′ untranslated region (UTR) and single exon of LOC_Os09g29760, respectively. c Expression of LOC_Os09g29760 in WT, fse5 mutant and complemented (COM) lines. The primer pair designed within the 408 bp deletion was used in qRT-PCR. Three biological replicates were subjected to qRT-PCR. OsActin1 was used as an internal control for data normalization. The values are the means ± SDs. ND, not detected. d Phenotypes of seeds (upper panel) and transverse sections (lower panel) of WT, fse5 mutant and COM seeds. Bars, 1 mm. e Nine-day-old seedlings of WT, fse5 mutant and COM lines. Bar, 5 cm vector construct containing the β-glucuronidase (GUS) reporter gene driven by the OsPORR1 promoter was transformed into rice cv Nipponbare (japonica) to confirm the expression of OsPORR1; GUS staining was observed in the abovementioned tissues (Additional file 1: Figure S4). Both experiments suggested that OsPORR1 was constitutively expressed in different tissues and at different growth stages.
OsPORR1 was predicted to localize to the mitochondria based on TargetP analysis (http://www.cbs.dtu.dk/ services/TargetP). A vector construct in which a green fluorescent protein (GFP) was fused to the full-length OsPORR1 sequence was transformed into rice cv Nipponbare. The GFP signals in protoplasts isolated from positive transgenic seedlings showed a punctate pattern that overlapped with the orange fluorescence of the mitochondrial indicator Mito Tracker Orange (Fig. 5b).
Taken together, these data showed that OsPORR1 was a mitochondrion-targeted protein.

OsPORR1 Functions in Splicing Mitochondrial nad4 Intron 1
Previous studies have shown that PORR domaincontaining proteins usually function in intron splicing of chloroplast-or mitochondrion-encoded genes (Kroeger et al. 2009;Colas des Francs-Small et al. 2012). Based on this premise and their subcellular localization, transcripts of 9 mitochondrion-encoded genes with introns were analyzed. The results showed an absence of the mature nad4 transcript in the fse5 mutant, but the mutant had a larger nad4 transcript than the WT did (Fig. 6a). These results suggested that an intron-splicing defect of nad4 had occurred in the fse5 mutant. To further confirm the nonspliced intron, each intron of nad4 was amplified by specific primers in the WT and fse5 mutant. The results indicated that the splicing of nad4 intron 1 was abolished and that no mature nad4 transcript was detected in the fse5 mutant (Fig. 6b). All mitochondrial introns in rice were examined via qRT-PCR in conjunction with two groups of specific primers. One group was used to amplify the spliced introns, whereas the other group was used for nonspliced introns. Only nad4 exon 1-exon 2 (spliced fragment) was not detected (ND) in the fse5 mutant, although its precursor fragment was present (Additional file 1: Figure  S5). These data indicated that OsPORR1 specifically affected the splicing of nad4 intron 1.

Mutation in OsPORR1 Affects Mitochondrial Structure and Function
Nad4 is an important subunit of complex I in the electron transfer chain of mitochondria. Complex I assembly and activity in seedlings were analyzed by blue native polyacrylamide gel electrophoresis (BN-PAGE). The results showed that the accumulation of complex I in the fse5 mutant was much lower than that in the WT, and NADH dehydrogenase activity was almost completely abolished (Fig. 7a). The ATP content and respiration rate were also reduced in the mutant compared with the WT (Fig. 7b-c). Complex I is an important component of the mitochondrial inner membrane. Observations were made of the mitochondrial inner structure in developing endosperm cells of the WT and fse5 mutant. The inner envelope cristae of mitochondria were well organized and surrounded by a dense stroma in the WT, whereas cristae development was poor and the stroma was thin in the fse5 mutant (Fig. 7d). Previous studies have shown that compromises to the main respiratory chain cause changes in the expression levels of alternative oxidase (AOX)-and NADH dehydrogenaseencoding genes (Toda et al. 2012;Li et al. 2013). Therefore, the transcript levels of AOX-encoding genes (OsAOX1a, OsAOX1b and OsAOX1c) and alternative NADH dehydrogenase-encoding genes (OsNDA1, OsNDA2, OsNDB1, OsNDB2, OsNDB3, and OsNDC1) Fig. 5 Spatiotemporal expression and subcellular localization of OsPORR1. a Expression of OsPORR1 in various plant organs, developing seeds and seedlings at different times. DAG, days after germination. Three biological replicates were subjected to qRT-PCR. OsActin1 was used as an internal control for data normalization. The values are the means ± SDs. b Subcellular localization of OsPORR1. GFP served as a control. GFP signals in protoplasts isolated from positive transgenic seedlings were observed via confocal laser scanning microscopy. Mito Tracker Orange was used to identify mitochondria. Bars, 10 μm were measured via by qRT-PCR. The expression of OsAOX1a, OsNDB2 and OsNDB3 was higher in the fse5 mutant than in the WT, but there were no differences in expression of the other genes (Fig. 7e). Taken together, these data showed that the defective splicing of nad4 intron 1 was associated with changes in mitochondrial structure and function.

Discussion
The fse5 mutant was isolated via artificial mutagenesis. Like that of other floury endosperm mutants such as floury endosperm2 (flo2) Figure S6) were reduced in the fse5 seeds compared with the WT W017 seeds. Unlike most other floury endosperm mutants, the fse5 mutant was embryo-or seedling-lethal. OsPORR1 is involved in the splicing of nad4 intron 1, which is indispensable for normal mitochondrial Fig. 6 Splicing of nad4 intron 1 was abolished in the fse5 mutant. Total RNA was extracted from nine-day-old seedlings, and random hexamer primers were used to synthetize first-strand cDNA. a, Transcript analysis of mitochondrial intron-containing genes in the WT (left) and fse5 mutant (right). OsActin1 was used as the internal control. Specific primers were used to amplify 9 transcripts . b, Schematic of nad4 precursor mRNA and RT-PCR analysis of three introns of nad4. nad4 precursor mRNA consists of 4 exons (E1-E4) and 3 introns. The black boxes indicate exons, and the curved lines represent introns. a, b, c, and m are fragments amplified by exon-exon flanking primers (Chen et al. 2017). Numbers of nucleic acids included by the introns and fragments are noted. M, DNA marker structure and function. Defects in the splicing of mitochondrion-encoded introns, especially in NADH dehydrogenase subunit (nad) genes, usually cause aborted seeds, slow-growing seedlings or small plants (Falcon de Longevialle et al. 2007;Koprivova et al. 2010;Liu et al. 2010;Colas des Francs-Small et al. 2014;Yang et al. 2014;Hsieh et al. 2015;Haïli et al. 2016;Xiu et al. 2016;Cai et al. 2017;Chen et al. 2017;Qi et al. 2017;Ren et al. 2017;Dai et al. 2018;Sun et al. 2018;Sun et al. 2019;Wang et al. 2018;Ren et al. 2019;Wu et al. 2019;Zhu et al. 2019;Ren et al. 2020;Yang et al. 2020).
Emp6, an ortholog of OsPORR1 in maize, is also located in mitochondria. Mutant emp6 kernels show obvious defects in embryos and throughout the endosperm, similar to those of the fse5 mutant seeds (Chettoor et al. 2015); thus, ZmEMP6 probably functions in grain development in a manner similar to OsFSE5. Multiple sequence alignment showed that the sequences of the PORR domain in ZmWTF1 and AtWTF9 were 49.24% and 38.23% similar, respectively, to that of OsPORR1 (Additional file 1: Figure S7). Both ZmWTF1 and its PORR domain have RNA-binding activity, and ZmWTF1 promotes the splicing of multiple group II introns in chloroplasts (Kroeger et al. 2009). AtWTF9 can also bind RNA and is required for splicing introns of rpl2 and ccmF C in mitochondria (Colas des Francs- . OsPORR1 is probably involved in the splicing of nad4 intron 1 by binding to RNA fragments. Nad4 is an essential subunit of complex I in the mitochondrial respiratory chain. The splicing of nad4 intron 1 is required for mitochondrial function. Like OsPORR1, ZmDEK35, a mitochondrion-targeted PPR protein, is especially essential for the splicing of nad4 intron 1. A mutation in Dek35 affected mitochondrial complex I assembly and NADH dehydrogenase activity, and the dek35 mutant displayed defective seed development similar to that of the fse5 mutant (Chen et al. 2017). Intriguingly, small subcomplexes generally considered to be intermediates of complex I were observed by BN-PAGE of the dek35 mutant but not the fse5 mutant . A possible explanation is that different tissues of maize and rice were used for BN-PAGE. . The values are the means ± SDs. **, P < 0.01, Student's t test. d Mitochondria in developing endosperm cells of the WT and fse5 seeds at 9 DAF. Bars, 0.5 μm. e Expression of AOX-encoding genes (AOX1a, AOX1b and AOX1c) and alternative NADH dehydrogenase-encoding genes (NDA1, NDA2, NDB1, NDB2, NDB3, and NDC1) in nine-day-old seedlings. Specific primers were used for qRT-PCR (Toda et al. 2012;Li et al. 2013). The data are based on three biological replicates. OsActin1 was used as an internal control for data normalization. The values are the means ± SDs Homozygous immature dek35 kernels were used to extract crude mitochondria for BN-PAGE, whereas homozygous seedlings or calli (−/−) are usually used in the case of rice Wu et al. 2019;Zheng et al. 2021). In this study, homozygous fse5 seedlings were used for BN-PAGE. Homozygous immature dek35 kernels (−/−) are born on heterozygous maternal tissues (+/−) and can acquire maternal nutrients and energy. In those kernels, the defective splicing of nad4 intron 1 led to dramatically decreased complex I (Chen et al. 2017). The accumulation of subcomplexes may be a response to decreased complex I under coregulation of maternal tissues and mutational kernels. Without support from heterozygous maternal tissues, the response in homozygous seedlings or calli is perhaps weakened or abolished. Thus, small subcomplexes were not observed in the fse5 mutant or rice ppr mutants such as fgr1, flo10 and osppr939 Wu et al. 2019;Zheng et al. 2021).
In addition to ZmDEK35, other PPR proteins such as AtMISF68 , ZmEMP8 (Sun et al. 2018), ZmEMP602 ) and ZmDEK43 (Ren et al. 2020), and the DEXH-box RNA helicase AtABO6 (He et al. 2012) also affect the splicing of nad4 intron 1. Their rice orthologs were identified by phylogenetic analysis and subjected to Y2H assays to verify interactions between them. The results of our Y2H assays showed that OsMISF68 interacted together with OsABO6, OsDEK35, OsDEK43, OsEMP8, OsPORR1 and itself (Additional file 1: Figure S8), suggesting a possible functional coordination among these proteins. Whether and how these factors synchronously regulate the splicing of nad4 intron 1 have yet to be determined.

Conclusions
Fse5 encodes a PORR domain-containing protein involved in the splicing of nad4 intron 1, which is indispensable for normal mitochondrial structure and function. Abnormal mitochondrial cristae and a reduced respiration rate in the fse5 mutant led to a lower ATP content compared with that in the WT and ultimately affected seed development and viability. These results provide valuable clues for revealing the roles of PORR domain-containing proteins in rice seed development and seedling growth.

Plant Materials and Growth Conditions
The fse5 mutant was isolated from an MNUmutagenized population of W017, a japonica variety whose seeds have excellent taste and texture. An F 2 population for gene mapping was generated by a cross between a heterozygous plant (+/fse5) and indica cv Nanjing 11. Seed germination tests and seedling growth were subsequently carried out in a growth chamber (12 h light/12 h darkness at 30°C). Other materials were grown in a paddy field.

Microscopy
Transverse sections of mature seeds were observed via a Hitachi S-3400 N scanning electron microscope (http:// www.hitachi-hitec.com). Scanning electron microscopy was performed as described previously (Kang et al. 2005). Semithin sections of developing seeds at 12 days after flowering (DAF) were observed via light microscope, and sample fixation and sectioning were performed as described previously (Peng et al. 2014).

Map-Based Cloning of OsPORR1
OsPORR1 was initially mapped using more than 160 polymorphic simple sequence repeat (SSR) and In/Del markers to genotype a small number of mutant F 2 individuals produced from the abovementioned cross. Fine mapping was then performed with tightly linked molecular markers designed from nucleotide polymorphisms between the Nipponbare and 93-11 reference genomes. The molecular markers used for fine mapping are listed in Additional file 1: Table S2.

Genetic Complementation
A vector containing the full-length cDNA sequence of OsPORR1 driven by its native promoter was constructed (the primers used are listed in Additional file 1: Table  S3) and transformed into the fse5 mutant via Agrobacterium tumefaciens. Positive transgenic plants were identified by the hygromycin resistance gene used as the selective marker in the pCUbi1390 vector.

RNA Extraction, Reverse Transcription PCR (RT-PCR) and qRT-PCR
Total RNA was extracted from various tissues using an RNA Prep Pure Plant kit (Tiangen Co., Beijing; http:// www.tiangen.com/en/). First-strand cDNA was synthesized using random hexamer primers for mitochondrion-encoded genes and oligo (dT) for nuclear-encoded genes, and PrimeScript Reverse Transcriptase (TaKaRa; http://www.takara-bio.com) was used for reverse transcription. OsActin1 (Os03g0718150) was used as an internal control (Peng et al. 2014). qRT-PCR was performed and the results were analyzed by an ABI 7500 real-time system. The primers used in RT-PCR for nine mitochondrion-encoded genes with introns and for three introns of nad4 are listed in Additional file 1: Table S4, and the primers used in qRT-PCR for OsPORR1, AOX-and alternative NADH dehydrogenaseencoding genes, introns and corresponding exons of mitochondrion-encoded genes are listed in Additional file 1: Table S5.

Subcellular Localization
The coding region of OsPORR1 was cloned and inserted in a pCAMBIA1305-GFP vector (the primers used are listed in Additional file 1: Table S3), after which the vector was introduced into rice cv Nipponbare via Agrobacterium tumefaciens. Protoplasts of positive seedlings were isolated as described previously (Chen et al. 2006). GFP fluorescent signals were detected by confocal laser scanning microscopy (Leica SP8), and Mito Tracker Orange CMTMRos (Invitrogen, Shanghai) was used to mark the mitochondria.

BN-PAGE and Complex I Activity Assays
Crude mitochondria were extracted from ten-day-old seedlings grown at 30°C in darkness Wu et al. 2019). BN-PAGE and complex I activity assays were then performed as described previously (Wittig et al. 2006;Hao et al. 2019;Wu et al. 2019).

Measurement of ATP Content and Respiration Rate
The ATP content of twelve-day-old seedlings grown at 30°C in darkness was measured with an ATP assay kit (Beyotime, Shanghai), and the respiration rate of nineday-old seedlings grown at 30°C was measured by a liquid-phase oxygen electrode (Hansatech, UK).
Additional file 1: Figure S1. Homozygous fse5 seeds were produced by a heterozygous plant (+/fse5). Segregation of grains with normal vitreous and floury (black, red arrows) kernels occurred among all seeds harvested from a single heterozygous plant (+/fse5) and was viewed by an X-ray viewer (PD-HA). Bar, 1 cm. Figure S2. Phenotypes of seeds and seedlings of the WT and fse5 mutant lines. A-D, 1000-grain weight (A), and length (B), width (C) and thickness (D) of mature WT and fse5 seeds. E-H, Total starch (E), amylose (F), protein (G) and lipid (H) contents of mature WT and fse5 seeds. I, Germination percentages for WT and fse5 seeds at 7 DAS in culture dishes. J, Percentages of seedlings grown from WT and fse5 seeds at 9 DAS in soil. K, Heights of 9-day-old seedlings grown from WT and fse5 seeds. The values are the means ± SDs. **, P < 0.01, Student's t test. Figure S3. OsPORR1 KO lines generated via CRISPR/Cas9. A, KO target site in the genomic sequence of OsPORR1. B, Target sequences of the OsPORR1 allele in four independent positive lines of Nipponbare. Single-nucleotide insertions (red letters, A, T and G) occurred in KO-140, KO-155 and KO-172, and a 32-nucleotide deletion (red dotted line) was present in KO-174. The black box indicates the protospacer-adjacent motif (PAM) sequence. C, Phenotypes (upper panel) and transverse sections (lower panel) of seeds from Nipponbare and KO lines. Bars, 1 mm. Figure S4. GUS staining of various tissues from a ProOsPORR1:GUS transgenic plant. Left-to-right, young seedling, root, stem, leaf, leaf sheath, panicle and developing seed. The promoter of OsPORR1 was inserted into a pCAMBIA1381Z vector, which was then introduced into Nipponbare via Agrobacterium tumefaciens transformation. GUS staining was performed as described previously (Zhang and Muench, 2015). Bars, 2 mm. Figure  S5. qRT-PCR analysis of 23 mitochondrial introns. Primers spanning adjacent exons were used to amplify fragments of mature mitochondrial transcripts (upper panel), and primers spanning adjacent exons and introns were used to amplify fragments of mitochondrial precursor mRNAs (lower panel) Chen et al. 2017;Lee et al., 2017). The results of nad4 exon 1-exon 2 (spliced fragment) and its precursor fragment are indicated by a black box. ND, not detected. Three biological replicates were assessed via qRT-PCR, and OsActin1 was used as an internal control for data normalization. The values are the means ± SDs. Figure S6. Immunoblot analysis of major starch synthases. Total proteins were extracted from developing WT and fse5 seeds at 15 DAF and separated by SDS-PAGE (Takemoto et al., 2002;Wang et al., 2010). Translation elongation factor 1α (EF-1α), encoded by Os03g0177400, served as the loading control. GBSSI, Granule-bound starch synthase I. SSIIa, Starch synthase IIa. SBE, Starch branching enzyme. AGPS2b, ADP-glucose pyrophosphorylase 2b. AGPL2, ADP-glucose pyrophosphorylase large subunit 2. Figure S7. Multiple sequence alignment of PORR domains in OsPORR1, ZmWTF1 and AtWTF9. ZmWTF1, GRMZM2G403797; AtWTF9, At2g39120. Figure S8. Y2H assays showing that MISF68 interacts together with splicing factors of nad4 intron 1. Empty-AD and Empty-BK were used as controls. Fulllength cDNAs of OsABO6 (LOC_Os01g02884), OsDEK35 (LOC_Os03g50500), OsDEK43 (LOC_Os05g11700), OsMISF68 (LOC_Os02g16650), OsEMP8 (LOC_Os08g41380) and OsPORR1 were cloned into pGADT7 or pGBKT7 vectors. The yeast transformation and screening procedures were performed according to the manufacturer's instructions (TaKaRa Bio, Kusatsu, Japan). DDO, Double dropout media (SD/−Trp-Leu). QDO, Quadruple dropout media (SD/−Trp-Leu-His-Ade). Table S1. Segregation of vitreous and floury grains from seven heterozygous plants (+/fse5). Table S2. Primers used for mapping. Table S3. Primers used for vector construction. Table S4. Primers used for splicing analysis. Table S5. Primers used for qRT-PCR analysis.