Cotton Fiber Development Requires the Pentatricopeptide Repeat Protein GhIm for Splicing of Mitochondrial nad7 mRNA

Abstract

Pentatricopeptide repeat (PPR) proteins encoded by nuclear genomes can bind to organellar RNA and are involved in the regulation of RNA metabolism. However, the functions of many PPR proteins remain unknown in plants, especially in polyploidy crops. Here, through a map-based cloning strategy and Clustered regularly interspaced short palindromic repeats/cas9 (CRISPR/cas9) gene editing technology, we cloned and verified an allotetraploid cotton immature fiber (im) mutant gene (GhImA) encoding a PPR protein in chromosome A03, that is associated with the non-fluffy fiber phenotype. GhImA protein targeted mitochondrion and could bind to mitochondrial nad7 mRNA, which encodes the NAD7 subunit of Complex I. GhImA and its homolog GhImD had the same function and were dosage-dependent. GhImA in the im mutant was a null allele with a 22 bp deletion in the coding region. Null GhImA resulted in the insufficient GhIm dosage, affected mitochondrial nad7 pre-mRNA splicing, produced less mature nad7 transcripts, and eventually reduced Complex I activities, up-regulated alternative oxidase metabolism, caused reactive oxygen species (ROS) burst and activation of stress or hormone response processes. This study indicates that the GhIm protein participates in mitochondrial nad7 splicing, affects respiratory metabolism, and further regulates cotton fiber development via ATP supply and ROS balance.

Pentatricopeptide repeat (PPR) proteins are RNA-binding proteins that regulate RNA processing events in plant organelles (Andrés et al. 2007; Yagi et al. 2013). The PPR family consists of two types denoted P and PLS. The P-type PPRs are composed of typical P motifs, and each P motif contains 35 amino acids, while the PLS-type PPRs are composed of P-type motifs and longer L-type and shorter S-type motifs (Small and Peeters 2000). Arrays of tandem PPR motifs form a super-helical ribbon-like sheet and are responsible for special RNA binding. PLS-type PPRs can be further classified into PLS, E, E⁺ and DYW subgroups based on their C-terminal domains (Lurin et al. 2004; Filipovska and Rackham 2013).

PPR proteins always contain signal peptides in the N-terminal, generally leading PPR proteins to plant mitochondria and/or chloroplasts to regulate organellar RNA metabolism. The PLS-type PPR proteins are mainly involved in site-specific RNA editing and P-type PPR proteins in translation activation/repression, stabilization and cleavage (Saha et al. 2007; Schmitz-Linneweber and Small 2008; Barkan and Small 2014). PPR proteins are essential for plant growth and development by regulating photosynthesis, respiratory metabolism and restoration of cytoplasmic male sterility (Saha et al. 2007). Mitochondrial Editing Factor9 (MEF9) is required for RNA editing of the site nad7-200 in the mitochondrial nad7 mRNA in Arabidopsis thaliana (Takenaka 2010). Maize Chloroplast RNA Processing 1 (CRP1) is necessary for the translation of petB and petD, transcription activation of petA and further affected the cytochrome f/b6 complex and Photosystem I (Barkan et al. 1994). Both P-type and PLS-type PPR proteins can promote intron splicing in chloroplasts and mitochondria. The P-type PPR protein Defective Kernel 2 (Dek2) is involved in nad1 splicing and the dek2 mutant has reduced levels of the exon 1-2 fragment and mature nad1 transcripts (Qi et al. 2017a). PLS-type PPR protein OTP70 is involved in rpoC1 splicing, and otp70 mutant shows a near-total lack of splicing of rpoC1 transcripts (Chateigner-Boutin et al. 2011). Arabidopsis Slow Growth 4 (SLO4) can regulate nad4 RNA editing and splicing of nad2 intron1 (Weißenberger et al. 2017). PPR4 and PPR78 are special cases, as they promote the trans-splicing of organellar RNA (Schmitz-Linneweber et al. 2006; Zhang et al. 2017). It was hypothesized that intron splicing is closely related to the intron fragments (Barkan and Small 2014), but few mechanisms for this have been verified by structural proteomics.

Mitochondria are the center of cellular energy homeostasis and redox regulation (Sweetlove et al. 2007). Via the electron transport chain, electrons transfer from NADPH via Complex I (NADH dehydrogenase) or Succinate via Complex II (Succinate dehydrogenase) to Coenzyme Q (CoQ), and then to Complex III (Cytochrome c reductase), Cytochrome c, Complex IV (Cytochrome c oxidase) and Complex V (ATP synthase), and finally generate ATP ${\mathrm{O}}_2^{-}$ for cellular activities (Siedow and Day 2000; Dudkina et al. 2006; Matsuzaki et al. 2009). Complex I is responsible for the first step catalysis of respiratory chain and its dysfunction can cause reorganization of cellular respiration and influence metabolic processes in many organelles such as plastids, mitochondria, peroxisomes, leading to destroyed consequences for plant and animal growth and development (Fromm et al. 2016). Based on the rapid development of high-throughput sequencing technology, the subunits of nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7, and nad9 have been identified in many higher plant mitochondrial genomes (Chomyn et al. 1986; Sugiyama et al. 2005). In recent decade, some PPR proteins have been shown to regulate nad genes. For example, several Arabidopsis PPR proteins have been reported to be responsible for RNA splicing, such as OTP43 (organelle transcript processing 43) for nad1 intron1 trans-splicing (de Longevialle et al. 2007), ABO5 (abscisic acid overly sensitive 5) for nad2 intron3 cis-splicing (Liu et al. 2010), BIR6 (buthionine sulfomixine-insensitive root 6) for nad7 intron splicing (Koprivova et al. 2010), TANG2 and OTP439 for nad5 intron2 and 3 splicing (Colas des Francs-Small et al. 2014), SLO3 (growth3) and MTL1 (mitochondrial translation factor1) for nad7 intron2 splicing (Hsieh et al. 2015; Haïli et al. 2016), and SLO4 (growth4) for nad2 intron1 splicing (Weißenberger et al. 2017). The roles of several PPR proteins in maize have also been verified, EMP10 causes loss of nad2 intron1 splicing and severely affects Complex I activity, and ultimately producing non-viable maize kernels (Cai et al. 2017). Dek37 affects cis-splicing of mitochondrial nad2 intron1 and seed development (Dai et al. 2018), and PPR20 is required for nad2 intron3 splicing (Yang et al. 2020). In addition, maize EMP12 simultaneously affects the splicing of nad2 introns1, 2 and 4 to regulate seed development (Sun et al. 2019). Based on these, identification of PPR proteins and further unraveling of their specific roles in respiratory pathways will provide insights into organelle metabolism in plants.

Mutants are powerful tools for studying the molecular mechanisms of genes. The immature fiber (im) mutant in cotton (Gossypium hirsutum L.) was discovered by Kohel et al. (1974) and has a non-fluffy fiber phenotype, where mature fibers have a low dry weight and fiber fineness (Kohel et al. 1974; Kohel and McMichael 1990; Kohel et al. 2002; Rong et al. 2007; Wang et al. 2012a; Kim et al. 2013a). By re-sequencing DNA pools from F₂ populations of im mutant and its near-isogenic line, G. hirsutum acc. TM-1, a PPR encoding gene, Gh_A03G0489, was tentatively identified (Thyssen et al. 2016). In addition, the transcriptome of fibers between im mutant and TM-1 showed that the differentially expressed genes (DEGs) were mainly involved in stress and intracellular repair (Kim et al. 2013b). The biological processes related to cellulose synthesis, secondary wall biogenesis, cell wall thickening and sucrose metabolism were also significantly down-regulated in im mutant (Wang et al. 2014). Nevertheless, there are some challenges in studying gene functions in cotton fiber, and the molecular mechanism of the candidate im gene was still unknown.

In this study, integrated map-based cloning, Clustered Regularly Interspaced Short Palindromic Repeat (CRISPR)/Cas9 and Virus induced gene silencing (VIGS) assays, we confirmed that Gh_A03G0489, encoding a P-type PPR protein and designated as GhImA, is responsible for the im mutant. Functional analysis indicated that GhIm targeted mitochondria and displayed a dosage-dependent involvement in nad7 precursor mRNA splicing. In the im mutant, null GhImA led to the dosage-insufficiency of GhIm proteins in cotton, the shortage of mature nad7 transcripts, the reduced activity in Complex I, and the increase in reactive oxygen species (ROS) levels. Insufficient ATP and toxic ROS accumulation led to non-fluffy fiber phenotype. Our studies indicate that GhIm plays a dosage-dependent role in mitochondrial nad7 splicing, which provided energy for cotton fiber development.

Materials and Methods

Plant materials

G. hirsutum acc. Texas Marker-1 (TM-1), with normal fluffy fibers, was the genetic standard line of upland cotton. The immature fiber mutant (im), G. hirsutum acc. im, with non-fluffy fiber phenotype, was controlled by a recessive gene on the Chr.A03 in tetraploid cotton (Wang et al. 2012a). TM-1 and the im mutant were considered as the near isogenic lines (NILs). Another two upland cotton accessions, with normal fluffy fibers, I4005 and CSIL028, were used to produce genetic populations for mapping the GhImA gene. I4005 displayed an excellent fiber quality and CSIL028 carried a 44.88% homozygous Chr.A03 segment from G. barbadense cv. Hai7124 in the background of TM-1 (Wang et al. 2012b). These accessions were self-pollinated and planted for years in the experimental field of Nanjing Agricultural University under normal field growth conditions.

Fresh leaf tissues were collected for DNA extraction and genetic analysis. TM-1 vegetative tissues from two-week-old seedlings (root, stem and leaf), fibers from the developing bolls of 10 days post anthesis (DPA) and 20 DPA were sampled for expression pattern analysis. RNA-seq was performed using fibers from two NILs, TM-1 and the im mutant, at 13 DPA, 16 DPA, 19 DPA, 22 DPA and 25 DPA, respectively. Samples were quick-frozen in liquid nitrogen and stored at -70°C before use.

Map-based cloning of the GhImA locus

Three mapping populations were developed by crossing the im mutant with three cotton accessions with fluffy fiber phenotype. A total of 342 individuals/lines were utilized from (CSIL028 × im) F₂/F_2:3 populations (Wang et al. 2014). Another two F₂/F_2:3 populations from the cross of TM-1×im and I4005 × im, with 737 and 1837 individuals/lines respectively, were developed. The fiber characteristics were investigated for each segregating population.

Simple sequence repeat (SSR) and single nucleotide polymorphism (SNP) markers on Chr.A03 were developed to target the candidate im gene. Linkage analyses and construction of a physical map were then conducted by Join-Map 4.0 (Ooijen and Voorrips 2001) and MapChart 2.3 (Voorrips 2002).

Phylogenetic tree construction

Phylogenetic tree was constructed using MEGA-X software (Kumar et al. 2018). The PPR protein sequences from other species, including Solanum tuberosum, Solanum lycopersicum, Durio zibethinus, Theobroma cacao, Herrania umbratica, Corchorus capsularis, Quercus suber, Malus domestica, Hevea brasiliensis, Manihot esculenta, Jatropha curcas, Ricinus communis, Populus trichocarpa, Citrus unshiu, Vitis vinifera, Arabidopsis lyrata, Arabidopsis thaliana and Zea mays, were downloaded from NCBI (https://www.ncbi.nlm.nih.gov/).

Creation of the CRISPR/Cas9 knockout mutant

For sgRNA design, target sites were identified on the website of CRISPR-P (http://crispr.hzau.edu.cn/CRISPR2/). As im mutant was controlled by a single nuclear recessive gene and mapped on Chr.A03, the sgRNA was chosen to specifically target GhImA and the primer sequences were listed in Table S1 in File S1 (Supplemental Material). sgRNA fragments were generated by annealing two complementary 23 nt oligos and were ligated to BsaI (R3535L, New England Biolabs) digested pKSE401 plasmid by Golden Gate reactions as previously described (Xing et al. 2014). The resulting constructs were transformed into E. coli competent cells and positive clones were identified by colony PCR and verified by Sanger sequencing. Then, the assembled CRISPR/Cas9 plasmid was transformed into Agrobacterium tumefaciens strain LBA4404, and the CRISPR/Cas9 vector was introduced into cotton G. hirsutum acc. W0 by Agrobacterium tumefaciens-mediated procedure as described previously (Li et al. 2009).

Gene cloning

The cDNA or genomic DNA of the candidate genes were amplified from TM-1 and the im mutant, and the PCR products were confirmed by sequencing. Total DNA and RNA were isolated as described previously (Paterson et al. 1993; Jiang and Zhang, 2003). First-strand cDNA was synthesized based on reverse transcription of 2 μg RNA using the reverse transcription polymerase reaction system (Promega, USA) and the final concentration was adjusted to ∼100 ng/μL with “one drop spectrophotometer OD-1000+” (OneDrop, Nanjing, China).

High-Fidelity ExTaq DNA Polymerase (TaKaRa Biotechnology Co. Ltd. China) was employed to amplify the genomic and cDNA sequences. pMD19-T vector (TaKaRa Biotechnology Co. Ltd. China) and E. coli strain DH5α were used for the transformation of target fragments. At least six clones for each gene were picked randomly and sequenced. Primers for PCR amplification were designed with Primer Premier 5 (http://www.premierbiosoft.com) and listed in Table S1 in File S1.

Expression analysis

Gene-specific primers for qRT-PCR were designed with Beacon Designer 7.91 (http://www.softpedia.com/get/Science-CAD/Beacon-Designer.shtml) based on the CDS sequences close to the 3’-UTRs. Subgenomic-specific primers were developed by analyzing the SNP sites between the GhImA and GhImD genes and were designed with WebSNAPER-SNAP (http://pga.mgh.harvard.edu/cgi-bin/snap3/websnaper3.cgi). The nad7 gene primer was designed specifically for the exon segments to prevent interference from abnormal transcripts. qRT-PCR was performed on a BIO-RAD CFX96^TM real-time PCR system (BIO-RAD, USA). The relative expression level was calculated using the 2^-△CT method (Livak and Schmittgen 2001) with three biological replicates. The expression levels of nuclear-encoded GhEF1α (GenBank ID: DQ174251) and mitochondrial 18S rRNA (GenBank ID: 24679539) were used as the endogenous controls.

Semi-quantitative RT-PCR was conducted for mitochondrion-encoded genes analysis. The gene sequences were downloaded from https://www.ncbi.nlm.nih.gov/nuccore/KR736345.1/(Chen et al. 2017b). The primers were designed with Primer Premier (www.PremierBiosoft.com) and amplicon products included the open reading frame (ORF) of target genes (Table S1 in File S1). The expression levels of the nuclear-encoded GhEF1α (GenBank ID: DQ174251) and mitochondrial 18S rRNA (GenBank ID: 24679539) genes were used as the endogenous controls.

Subcellular localization

The full-length ORFs of the GhImA and GhImD genes without stop codons from TM-1 were inserted into the pBINPLUS.GFP4 vector by the Gateway LR recombination reaction (Invitrogen, CA, USA) to generate the 35S: GhImA-GFP and 35S: GhImD-GFP expression vectors, which were transformed into Agrobacterium strain GV3101. The primers were listed in File S1. The Agrobacterium cells were collected and suspended in the infiltration medium (10 mM MgCl₂, 10 mM MES and 100 µM acetosyringone with pH = 5.6). After incubation for 3 h at 28 °C, the suspensions were infiltrated into fully expanded leaves of N. benthamiana. Fluorescence signals were detected 2-3 days after infiltration using laser confocal fluorescence LSM710 microscopy (Zeiss, Germany). The empty vector (GFP) and mitochondria tracker (Mito-Tracker) were used as the negative and positive controls, respectively.

Prediction of GhIm-binding sequences

PPR repeat profile of GhImA protein was predicted using ScanProsite online software (https://prosite.expasy.org/). Each PPR motif always connects with one RNA base and the 6^th and 1’^st amino acids are known as the PPR code, the combinations of the 6^th and 1’^st amino acids within each motif confer RNA-specific binding characteristics (Barkan et al. 2012; Yin et al. 2013; Shen et al. 2015; Huang et al. 2018).

BN-PAGE and Complex I in-gel assays

The mitochondria of 22 DPA fibers from TM-1 and im mutant were isolated separately by a plant mitochondria isolation kit (Biohao, Wuhan, China). The enriched mitochondria were re-suspended in sample buffer (25 mM Bis-Tris, 20% glycerin, pH 7.0) adding n-dodecyl-β-D-maltoside to a final concentration of 1% and incubated on ice for 60 min. The samples were centrifuged at 20000×g for 60 min at 4 °C and added to the loading buffer for BN-PAGE. The concentration of the separation gel was from 5 to 13.5%. At first electrophoresis was run at 50 V, and 25 V was added every 20 min to the final 150 V until the loading dye migrated to the edge of the gel. The gel was stained by coomassie brilliant blue. To analyze Complex I in-gel activity, the gel was immersed in the assay buffer (0.05 M MOPS/KOH, pH 7.6, 0.2 mM NADH, and 1 mM nitrotetrazolium blue). The reaction was stopped with 40% methanol and 10% acetic acid.

RNA pull-down assay

To better express GhIm protein in prokaryote, the sequences of GhImA and GhImD were performed codon optimization using Codon OptimWiz instrument (https://www.genewiz.com.cn/Public/Services/Gene-Synthesis/mmz/). The resulting sequence was introduced into pGEX-4T-1 construct to express GhImA and GhImD recombinant proteins, then the GST-labelled GhImA and GhImD proteins were purified for subsequent RNA-Pull down analysis. Sense and antisense nad7 mRNA sequence were amplified (primers listed in Table S1 in File S1), labeled with Biotin and purified with RNeasy Mini Kit (Qiagen, Germany). The RNA-Pull down was carried out according to the instruction of Pierce™ Magnetic RNA-Protein Pull-Down Kit (Thermo Scientific™, USA). Western blot was performed using GST antibody (http://www.agrisera.com/en/index.html) followed above mentioned method.

Measurement of Complex activity and histochemical detection of ROS

The activities of Complex I for 22 DPA fibers were examined with a Complex detection kit (CominBio, Suzhou, China). Complex I activity was calculated with three biological replicates as the degradation rate of Cytochrome C per minute. The kinetics of Cytochrome C were detected at 550 nm (ε = 1.91 × 10⁴/mol/cm).

ROS levels were examined with H₂DCFDA staining. Cotton leaves were incubated in 10 µM H₂DCFDA (pH = 3.8) in the dark for 30 min, then fluorescence signals were detected. H₂O₂ content was detected using a 3,3 N-Diaminobenzidine Tertrahydrochloride (DAB) peroxidase color development kit (Solarbio, Beijing, China) and H₂O₂ detection kit (Jiancheng Bio, Nanjing, China), respectively. Cotton leaves were incubated in 1.0 mg/mL DAB solution (pH = 3.8) in the dark for 10 h and then decolorized in decolorization solution (Ethanol: Acetic acid: Glycerol = 8:1:1). All samples were washed three times with ddH₂O before further investigation.

VIGS assay

A 503 bp homologous fragment from both GhImA and GhImD in TM-1 was inserted into the pTRV2 VIGS vector (TRV: GhIm) via the Gateway LR recombination reaction (Invitrogen, CA, USA), which was then transformed into Agrobacterium strain GV3101. The primers were listed in Table S1 in File S1. The Agrobacterium cells were collected and suspended in the infiltration medium (10 mM MgCl₂, 10 mM MES and 200 µM acetosyringone with pH = 5.6), After incubation for 3 h at 28 °C, the suspensions were mixed with the Agrobacterium strains containing the pTRV1 vector at a ratio of 1:1 (TRV: GhIm) and infiltrated into the mature cotyledons of TM-1 plants. Untreated plants (CK) and empty vectors (TRV : 00) were used as negative controls and the CLA1 (KJ123647, TRV: CLA1), 1-deoxyxylulose 5-phosphatesyn-thase gene was used as the positive control.

RNA-seq analysis

Fibers from TM-1 and im mutant plants at 13 DPA, 16 DPA, 19 DPA, 22 DPA and 25 DPA were sampled for RNA-seq analysis. Library DNA was checked for concentration and size distribution in an Agilent2100 bioanalyzer before sequencing with an Illumina HiSeq 2500 system according to the manufacturer’s instructions (HiSeq 2500 User Guide). A total of 422,509,057 raw paired-end sequenced reads with a length of 101 bp were generated, resulting in 61 GB of raw data (PRJNA436644). After trimming off of adapters and filtering out low quality bases and short reads, 346,318,805 clean reads were obtained, leading to 42.4 GB of clean data. Subsequently, RNA-seq data from TM-1 and im mutant plants were mapped to the TM-1 transcript database (Zhang et al. 2015).

To explore the im candidate genes, GSNAP software (Wu and Nacu 2010) was used for mapping the reads to the TM-1 genome database (Zhang et al. 2015). SNPs and InDels between TM-1 and the im mutant were obtained with vcftools software (Danecek et al. 2011).

For DEG analysis, Tophat software was used for mapping the reads to the TM-1 genome database (Trapnell et al. 2009). The expression level of each gene obtained from Cuffdiff software (Trapnell et al. 2012) was represented as fragments per kilobase of transcript sequence per millions base pairs sequenced (FPKM). DEGs were identified by Cuffdiff and were required to have a 2-fold change (q value ≤ 0.05) and expression level of FPKM ≥3.0.

Correlations between RNA-seq data and qRT-PCR were analyzed using 20 representative genes. The R² value was obtained by the ratio of $\hspace{0.17em}{\text{log}}_2^{\left(\text{FPKM}\right)}$ values from RNA-seq against the $\hspace{0.17em}{\text{log}}_2^{(\text{Relative}\hspace{0.17em}\hspace{0.17em}\text{expression}\hspace{0.17em}\hspace{0.17em}\text{level})}$ from qRT-PCR value.

Statistical analysis

The experimental data was presented as means with standard derivations. For variance analysis, the SPSS (Statistical Package for the Social Sciences) was used to assess statistically significant differences based on student’s t-test or Tukey’s test. * and ** represent significant differences at P < 0.05 and P < 0.01 levels respectively. Lowercase a, b, c represent the significance at P < 0.05 level respectively.

Data Availability Statement

Strains and plasmids are available upon request. Supplemental files available at FigShare. The primers used in this study, predicted genomic polymorphic loci between TM-1 and the im mutant, detection on polymorphic loci during target region in Chr.A03 in im mutant and three different cotton accessions with fluffy fiber genotype, and GO enrichment analysis of differentially expressed genes between TM-1 and the im mutant are listed as Table S1-S4 in File S1, respectively. Supplemental Figures and legends are listed in File S2. The sequences files of G. hirsutum acc. TM-1 are available at http://www.cottongen.org/. Transcriptome data have been deposited in GenBank with the accession number: PRJNA436644. Supplemental Material available at figshare DOI: https://doi.org/10.25386/genetics.13265153.

Results

Map-based cloning of the GhImA locus

Mature cotton fibers exhibited a fluffy state after the capsule dehisced and fiber cells dehydrated, while the fibers of immature fiber (im) mutant displayed non-fluffy phenotype (Figure 1A). By analyzing the F₂ genetic population of G. hirsutum CSIL028×im, we found that the im mutant was controlled by a single nuclear recessive gene localized at A03 chromosome in tetraploid cotton (Wang et al. 2012a). To fine-map the Im gene, two materials of G. hirsutum acc. TM-1 and G. hirsutum acc. I4005, both with normal fluffy fibers, were used to individually cross with im mutant to construct F₂ mapping populations. As a result, 737 and 1,837 individuals were generated in TM-1×im and I4005×im F₂ populations respectively. Using the three segregating populations, the target gene was delimited to a 1.93-Mb DNA interval flanked by markers of NAU8194 and S3059 on Chr. A03, and we named the target gene GhImA. According to the reference genome sequence of the tetraploid cotton G. hirsutum acc. TM-1 (Zhang et al. 2015), the region contains 43 open reading frames (ORFs) (Figure 1B).

Figure 1.

Map-based cloning and identification of the Im gene. A. TM-1 and im mutant were near isogenic lines and the im mutant exhibited a non-fluffy fiber phenotype in contrast to TM-1. Scale bar, 10 mm. B. The Im locus was mapped to an interval of approximately 1.93 Mb on chromosome A03 on the physical map of tetraploid cotton TM-1 with 43 ORFs annotated (Zhang et al. 2015). The schematic structure of the candidate Im gene ORF29 displayed 22 bp deletions in the im mutant, resulting in a premature stop codon TAG (in red font). SP, P and S indicate the N-terminal signal peptide, PPR motifs in PPR proteins, respectively.

To further identify the candidate gene, transcriptome sequencing of TM-1 and im mutant was performed using fibers at different developmental stages. In total, 36 SNP/InDel polymorphic loci (18 ORFs) between TM-1 and the im mutant were predicted in the interval flanking the target locus (Table S2 in File S1), of which 14 ORFs (25 polymorphic loci) with FPKM ≥1.0 were amplified and sequenced using TM-1 and im mutant genomes. As shown in Table S3 in File S1, three polymorphic loci on Gh_A03G0461, Gh_A03G0487 and Gh_A03G0489 were confirmed between TM-1 and the im mutant. Further, polymorphisms of these loci were also detected in the other two mapping parents, CSIL028 and I4005. Finally, only the InDel locus on Gh_A03G0489 displayed a consistent polymorphism between fluffy fiber parents of TM-1/CSIL028/I4005 and non-fluffy fiber im mutants. Thus, Gh_A03G0489 was preliminary considered as the candidate gene for GhImA. Then we used the InDel polymorphic marker of the candidate gene to detect correlations with fiber phenotype in these three genetic populations. The resulting genotypes were consistent with the fiber fluffy or non-fluffy phenotypes, and no cross-over was detected in the 2,916 individuals (Figure S1 in File S2).

To determine whether loss of function of GhImA locus is responsible for the non-fluffy fiber phenotype, we used CRISPR/Cas9 gene editing technology to create the GhImA-specific knock-out mutants (Figure 2A). Three homozygous editing GhImA gene lines with different edited types, named respectively as m1, m2 and m3, were identified in T₂ generation (Figure 2, B-D). We cloned and sequenced the GhImD (Gh_D03G1048) from wild type W0 and GhImA-edited lines, no off-target effect for GhImD was found in all mutant plants. Fiber phenotype in the three GhImA locus-edited mutants showed non-fluffy fibers, just same as that in im mutant (Figure 2E).

Figure 2.

Identification of genotype and fiber phenotype from CRISPR/Cas9-mediated GhImA-editing lines in T₂ generation. A. sgRNA target site analysis for gene editing using CRISPR/Cas9. The sgRNA target site and the PAM regions are highlighted in red and gray background, respectively. sgRNA was designed to specifically target GhImA rather than GhImD. The white letters are the Xmn I restriction enzyme site, which located in sgRNA. B. Restriction enzyme assay of the mutants at the target sites. None of digestion fragment detected in mutants. M, Marker. The m1, m2, m3 represent the three independent mutant lines, respectively. C. Sanger sequencing of genome DNA from edited plants at GhImA target sites. Nucleotide deletions and insertions are shown in red letters, details are labelled at right. The gaps between the omitted nucleotides are in dotted line. D. The detailed chromatograms at the target sites are illustrated. PAM sequences are underlined with red. Insertion is underlined with blue. E. Mature cotton bolls from T₂ plants showed non-fluffy fiber phenotype as im mutant. Scale bar, 10 mm.

Taken together, Gh_A03G0489 was responsible for the GhImA gene and disruption of this gene caused the non-fluffy fiber phenotype in im mutant.

GhIm encoded a P-type pentatricopeptide repeat protein

Genomic DNA and cDNA sequences of GhImA and GhImD were obtained from TM-1 and the im mutant by PCR and RT-PCR analysis, respectively. In TM-1, sequence alignment analysis showed that GhIm genes (GhImA and GhImD) had no intron with an open read frame of 3,072 bp, encoding a total of 1,023 amino acids. There were 95 SNP loci between GhImA and GhImD in TM-1, of which 49 were non-synonymous sites. Subgenome-specific primers for GhImA and GhImD (Figure S2 in File S2) showed that both GhImA and GhImD were equally expressed in vegetative organs (root, stem and leaf) and fibers (10 DPA and 20 DPA) (Figure 3A). The GhIm had 40.93% sequence similarity with Arabidopsis At5g61990.1, which encoded a pentatricopeptide repeat protein. Motif analysis showed that GhIm had 24 P-motifs and one S motif, and therefore was a P-type PPR protein (Figure 1B and Figure S3 in File S2). Phylogenetic relationships suggested that GhIm had higher identities with PPR proteins from Durio zibethinus and Theobroma cacao, whereas very low identities with PPR proteins from Arabidopsis and Zea mays whose functions are involved in the intron splicing of nad1, nad2 and/or nad4, inferring that GhIm belongs to a novel PPR protein with different functions from PPRs reported (Figure S4 in File S2).

Figure 3.

Expression pattern of GhIm and subcellular localization of the GhIm protein. A. Expression pattern of GhIm in different tissues of TM-1, including root, stem, leaf, fibers at 10 DPA and 20 DPA, respectively. B. Comparison of GhIm expression in TM-1 and im mutant plants at different fiber developmental stages of 13 DPA, 16 DPA, 19 DPA, 22 DPA and 25 DPA. The expression levels of the GhEF1α gene (DQ174251) were used as the internal control. The relative expression level was calculated using the 2^-△CT method with three biological replicates. Significant differences in expression between TM-1 and the im mutant at the same stage of fiber development are shown (*P < 0.05 and **P < 0.01, Student’s t-test). C. The full-length CDSs of GhImA and GhImD from TM-1 were fused to green fluorescence protein (GFP), and then transiently expressed in tobacco leaf epidermal cells. Mito-Tracker fused to RFP (Mito) was used as a control. Scale bar, 20 μm.

Compared to TM-1, the GhImA in the im mutant had a 22-bp deletion in the ORF region, resulting in a premature stop codon (TAG) after 184 amino acid residues (Figure 1B). There were no polymorphic loci in GhImD between TM-1 and the im mutant. We also examined GhIm expression patterns in TM-1 and im mutant fibers and found a significant difference in the 13, 16, 19, 22 and 25 DPA fibers, in which the transcripts of GhImA in the im mutant were significantly lower than that in TM-1. However, in 25 DPA fibers, the transcripts of GhImD were higher in the im mutant than in TM-1 (Figure 3B), resulting in higher expression in im mutant for the total expression levels of GhIm in 25 DPA fibers.

Bioinformatics analysis indicated that the GhIm was localized to mitochondria under the guidance of an N-terminal signal peptide. To confirm this, 35S: GhImA-GFP and 35S: GhImD-GFP constructs were injected into tobacco leaf respectively. Confocal microscopy showed that signals from 35S: GFP construct distributed throughout the cell, while signals from both 35S: GhImA-GFP and 35S: GhImD-GFP constructs located along the cell membrane, and co-located with the red fluorescence signals from mitochondria tracker (Mito-Tracker). The results indicated that both GhImA and GhImD proteins localized in mitochondria (Figure 3C).

The GhIm protein potentially participates in mitochondrial nad7 processing

To gain insight into an overview changes for mitochondrial transcripts in im fiber cells, 36 mitochondrial protein-encoding genes were amplified by RT-PCR in 22 DPA fibers of TM-1 and im mutant using both nuclear-encoded GhEF1α and mitochondrial 18S rRNA genes as the endogenous controls. The results showed that mature transcripts of nad7 in im mutant exhibited an obvious decreased abundance compared with that of TM-1, and other mitochondrial genes had no indistinguishable expression between TM-1 and im mutant (Figure 4A and Figure S5A in File S2), implying that GhIm might participate in the expression regulation of nad7 in developing fiber cells.

Figure 4.

Investigation on mature transcripts and the splicing efficiency of mitochondrion-encoded genes in fibers of TM-1 and im mutant. A. RT-PCR analysis of 36 mitochondrion-encoded transcripts in fibers of TM-1 and im mutant. The primers were designed to amplify the full length of coding region for each transcript. Mitochondrion genes were referred to Gossypium raimondii mitochondrial genome database (GenBank: KR736345.1). RNA was from 22 DPA fibers of TM-1 and the im mutant. GhEF1α (GenBank: DQ174251) was used as a reference. B. Schematic representation of cotton mitochondrial nad7 gene and RT-PCR analysis of nad7 pre-mRNA in TM-1 and im mutant. C. The splicing analysis of introns in nad7 pre-mRNA. The solid arrow showed the unspliced intron2. W: TM-1; M: im mutant. D. Splicing efficiency of mitochondrial introns. The ratio of mature transcripts to unspliced fragments was used for measuring differences in splicing efficiency. The splicing efficiency of 23 introns for nad1(4), nad2(4), nad4(3), nad5(4), nad7(4), rps3(1), rps10(1), cox2(1) and ccmFC(1) were detected, showing that the intron2 splicing efficiency of nad7 was potentially affected in im mutant.

Based on released cotton mitochondria complete genome sequence (GenBank: KR736345.1), the nad7 gene contains five exons and four introns, and four cis-splicing events are needed for the complete maturation of nad7 transcript (Figure 4B). To further examine the splicing of nad7 precursor transcript, we designed four specific primers to amplify fragments containing each of the four introns of nad7 transcript in TM-1 and im mutant. In im, the cis-spliced fragment of nad7 mRNA joining exon2/exon3 was clearly attenuated and the unspliced fragment signal containing nad7 intron2 was stronger than that in the TM-1 (Figure 4C, Figure S5B and Figure S6 in File S2). In addition, nine mitochondria genes containing intron(s) were also detected for their splicing efficiency analysis. Totally, 23 introns for nad1(4), nad2(4), nad4(3), nad5(4), nad7(4), rps3(1), rps10(1), cox2(1) and ccmFC(1) were investigated. We confirmed that the intron2 splicing efficiency of nad7 was significantly affected in im than that in TM-1 (Figure 4D). Furthermore, using ScanProsite online software (https://prosite.expasy.org/), 25 PPR motif sequences from GhIm were used to mine the probable recognition sites, and potential candidate binding sequences of XXUXUCAUXCGC(A/C)GGACXXCAU(G/U)XX were predicted, where X represents any nucleotide (A/T/C/G). Further, the candidate binding sequence (CACTTTGAATTTTGGACCTCAACAT) was detected in nad7 exon1.

Taken together, GhIm protein might participate in mitochondrial nad7 processing and affect respiratory pathways.

GhIm protein binds to nad7 mRNA and affects Complex I activity

In fiber developmental process, nad7 expression was continuously up-regulated from 13 to 25 DPA up to a peak at 22 DPA in TM-1, however, nad7 kept a relative low expression in im mutant, and significantly lower at 16 DPA, 19 DPA, 22 DPA and 25 DPA fibers compared with that in TM-1 (Figure 5A).

Figure 5.

The GhIm protein was essential for splicing of mitochondrial nad7 precursor mRNA. A. Comparison of nad7 expression in fibers of TM-1 and im mutant. The expression levels of the GhEF1α (GenBank: DQ174251) were used as the internal control. The relative expression level was calculated using the 2^-△CT method with three biological replicates. Significant differences in expression between TM-1 and the im mutant at the same stage of fiber development are shown (*P < 0.05 and **P < 0.01, Student’s t-test). B. Detection of respiratory Complex I activity in fibers of TM-1 and im mutant. Complex I activity was calculated as the degradation rate of NADH per minute with three biological replicates. Significant differences in activity between TM-1 and the im mutant at 19 and 22 DPA of fiber development are shown (*P < 0.05 and **P < 0.01, Student’s t-test). C. RNA Pull-Down assay showed that the GhImA and GhImD proteins could bind to nad7 mRNA. GhImA and GhImD proteins were purified in E. coli and nad7 mRNA was transcribed with synthetic nad7 DNA as a template. The pull-down assay was carried out using the Pierce^TM Magnetic RNA-Protein Pull-Down Kit (Thermo). Compared with the Input and antisense nad7 RNA, both GhImA and GhImD can directly bind to sense nad7 RNA.

RNA-Protein Pull-Down analysis revealed interactions between the GhIm protein and nad7 mRNA. Compared with the input, GST empty vector and antisense mRNA, GST-GhImA protein-nad7 mRNA and GST-GhImD protein-nad7 mRNA complexes were detected using GST antibody, demonstrating both GhImA and GhImD could directly bind to nad7 transcript (Figure 5C). These results indicate that GhIm protein could bind to nad7 mRNA, participating in the cis-splicing of mitochondria nad7 intron2.

The nad7 protein is a component of Complex I in the mitochondria. Using fiber cells at 22 DPA, blue native polyacrylamide gel electrophoresis (BN-PAGE) and In-gel NADH dehydrogenase activity assays were performed to investigate the potential influence on mitochondrial complexes in TM-1 and im mutant. Due to fiber tissues specificity, it is difficult to extract enough amounts of mitochondria proteins in fiber secondary cell wall developmental process. Nevertheless, we found that there was a bit weak difference of the assembly of mitochondrial Complex I and NADH dehydrogenase activities between TM-1 and im mutant (Figure S7 in File S2). Further, we measured the activity of Complex I in fibers, the data showed that the activities of Complex I in TM-1 increased from 16 DPA to 25 DPA fibers. In the im mutant, Complex I activity generally showed a similar trend to that of TM-1, but was significantly lower in 19 DPA and 22 DPA fibers (Figure 5B).

GhImA affected respiratory metabolism and cellulose synthesis in fibers

To clarify the roles of the GhImA protein in cotton fiber development, we analyzed RNA-seq data from TM-1 and im fibers at 13 DPA, 16 DPA, 19 DPA, 22 DPA and 25 DPA (Table S4 in File S1). In total, 6,084 DEGs were identified with 3,101 genes down-regulated and 3,543 genes up-regulated in the im mutant compared with TM-1 (q ≤ 0.05, fold change ≥2.0 and FPKM ≥3.0). Gene Ontology (GO; http://bioinfo.cau.edu.cn/agriGO/) analysis indicated that DEGs were mostly related to three biological processes, respiratory metabolism, cell wall/cellulose synthesis and stress responses.

The GhIm protein was found to target mitochondria, regulate Complex I activity, and affect ATP generation. DEGs related to respiratory metabolism included “copper ion export” (GO:0060003), “establishment or maintenance of transmembrane electrochemical gradient” (GO:0010248), “detoxification of copper ion” (GO:0010273), “respiratory burst” (GO:0045730), “respiratory burst involved in defense response” (GO:0002679) and “response to oxidative stress” (GO:0006979) (Figure 6A). In plants, electrons transfer from complexes to ${\mathrm{O}}_2^{-}$ and generate ATP via ATP synthase (Sweetlove et al. 2007; Matsuzaki et al. 2009). Alternative oxidases (AOXs) can be activated to maintain the tricarboxylic acid cycle and electron transport when Complexes I, III and IV are unable to function properly (Vanlerberghe and Ordog 2002; Izabela and Anna 2003; Dai et al. 2018). qRT-PCR analysis showed that the transcription of ATP synthase subunits genes, ATPase subunit 1, ATPase subunit 4 and ATPase subunit A was greatly inhibited in the im mutant at 22 and 25DPA, while the transcription of AOXs: AOX2 and AOX1A was activated at 22 and 25DPA (Figure 6B).

Figure 6.

Analysis of DEGs between TM-1 and the im mutant. A and C. Gene ontology (GO) analysis of DEGs between TM-1 and im mutant related to respiratory metabolism (A) and secondary cell wall synthesis (C). DEGs were required to have a 2-fold change (q value ≤ 0.05) and expression level of FPKM ≥3.0 at any of the tested fiber developmental stages (13 DPA, 16 DPA, 19 DPA, 22 DPA and 25 DPA) between TM-1 and the im mutant. B and D. qRT-PCR analysis of genes related to respiratory metabolism (B) and secondary cell wall synthesis (D) in TM-1 and the im mutant. The relative expression level was calculated using the 2^-△CT method with three biological replicates. Significant differences in expression between TM-1 and the im mutant at the same stage of fiber development are shown. Statistical significance was determined by Student’s t-test. *P < 0.05 and **P < 0.01.

Respiratory metabolism provides energy for cellular activities (Sweetlove et al., 2007; Matsuzaki et al. 2009). With ATP shortage, cell wall development is inhibited. During the fiber development stages (from 13 DPA and 25 DPA), cellulose is necessary for secondary cell wall synthesis (Basara and Malik 1984; Lee et al. 2007), therefore, several DEGs involved in cell wall/cellulose synthesis were also enriched, including “carbohydrate metabolic process” (GO:0005975), “cell wall organization or biogenesis” (GO:0071554) and “carbohydrate biosynthetic process” (GO:0016051) (Figure 6C). Genes involved in secondary cell wall synthesis like cellulose synthase (CESA) genes and COBRA Like (COBL) genes were also detected in TM-1 and im mutant fibers (Figure 6D). CESA4B and CESA7B had similar expression patterns in TM-1 and im mutant fibers, but had lower expression levels in the im mutant at some stages. In TM-1 fibers, the transcripts of CESA8B, COBL9 and COBL13 were continuously up-regulated from 16 DPA to 19 DPA, and then declined on 22 DPA and 25 DPA; while in im mutant fibers, the transcription of CESA8B, COBL9 and COBL13 peaked at 22 DPA; displaying delayed expression patterns, as reported before (Wang et al. 2014). Moreover, at 22 DPA and 25 DPA, the expression levels of CESA8B, COBL9 and COBL13 were higher in the im mutant than in TM-1. When the electron transport chain was inhibited in mitochondria, the production of oxygen free radicals was increased and ROS release was induced (Matsuzaki et al. 2009). DEGs involved in biological processes such as “Response to stimulus” (GO:0050896), “Response to stress” (GO:0006950) and “Cell death” (GO:0008219) were also enriched (Table S4 in File S1). In addition, there was a higher correlation between RNA-seq data and qRT-PCR validation (R² =0.87) (Figure S8 in File S2), indicating that RNA-seq data were accurate and robust.

GhImA and GhImD homologs affected respiratory metabolism and plant growth and development in a dosage-dependent manner

The im mutant was an allotetraploid cotton, and mutation of GhImA caused non-fluffy fiber phenotypes. To explore the roles of the GhImD gene, we conducted virus-induced gene silencing to interfere with the transcription of both GhImA and GhImD in TM-1 (Figure 7). When the second leaves of positive control TRV: CLA1 plants displayed an obvious photobleaching phenotype after two weeks of agroinfiltration (Figure 7A), the expression levels of GhImA and GhImD in GhIm-silenced plants (TRV: GhIm) were significantly lower than those in the untreated plants (CK) and TRV: 00 plants (Figure 7B). After four weeks, the TRV: GhIm plants was significantly lower compared with CK and TRV: 00 plants (Figure 7C), and nearly stopped growing at five weeks (Figure 7D). We also suppressed the transcription of GhIm in im mutant and obtained similar results (Figure S9 in File S2).

Figure 7.

Plant phenotypes after silencing GhIm in TM-1. A. Phenotypes of control (CK), empty vector (TRV : 00), positive control (TRV: CLA1) and GhIm-silenced (TRV: GhIm) plants at two weeks post-agroinfiltration. Scale bar, 35 mm. B. Transcript levels of GhImA and GhImD in leaves of CK, TRV : 00 and TRV: GhIm plants. The relative expression level was calculated using the 2^-△CT method with three biological replicates. Statistical significance was determined by t-test. Statistical significance: *P < 0.05 and **P < 0.01. C. Plant height comparison of CK, TRV : 00 and TRV: GhIm plants after GhIm was silenced. **P < 0.01, Student’s t-test with 20 biological replicates for each treatment. D. Phenotypes of CK, TRV : 00, TRV: CLA1 and TRV: GhIm plants at five weeks post-agroinfiltration. Scale bar, 35 mm.

TM-1, im mutant and TRV: GhIm (in im mutant) plants contained four, two and zero copies of the GhIm gene, respectively. Correspondingly, the expression levels of ATP synthase (ATPase subunit 1, ATPase subunit 4 and ATPase subunit A) reduced significantly with decreasing GhIm transcripts, while the expression levels of AOX1A increased significantly, and AOX2 showed higher expression in TRV: GhIm than that in im and TM-1 (Figure 8A). Quantitative and qualitative analysis revealed that the ROS or H₂O₂ content increased in TM-1, im mutant and TRV: GhIm plants (Figure 8, B and C). Taken together, these results suggest that respiratory metabolism and ROS burst correlated positively with the copy numbers of GhIm genes in cotton.

Figure 8.

Respiratory metabolism and reactive oxygen balance in different GhIm copies. A. Expression levels of genes related to respiratory metabolism. Statistical significance was determined by Tukey-test. Lowercase a, b, c represent the significance at P < 0.05 level respectively. B. Detection of ROS and H₂O₂ content. Cotton leaves were incubated in 10 µM H₂DCFDA (pH=3.8) for 30 min and 3,3 N-Diaminobenzidine Tertrahydrochloride (DAB) stain for 12 h. Scale bar, 10 mm. C. Measurement of H₂O₂ content. TM-1, im and TRV: GhIm (in im mutant) plants have transcripts from four copies (AADD), two copies (aaDD) and zero copies (aadd) of GhIm, respectively. The second leaves were sampled about five weeks after the TRV strain was injected. All measurements were performed with three biological replicates. Statistical significance was determined by Tukey-test. Lowercase a, b, c represent the significance at P < 0.05 level, respectively.

Discussion

Molecular mechanism of the GhImA gene in cotton

Mitochondria are the center of cellular energy and redox homeostasis via the respiratory chain (Siedow and Day 2000; Dudkina et al. 2006; Sweetlove et al. 2007; Matsuzaki et al. 2009). There are five inter-coordinated complexes in mitochondria respiratory metabolism for generating ATP and ${\mathrm{O}}_2^{-}$ supplying cellular activities. When Complexes I, III and IV are unable to function properly, alternative oxidases (AOXs) can bypass other Complexes, catalyze the oxidation of ubiquinol and transport of O₂ to H₂O but without ATP generation (Vanlerberghe and Ordog 2002; Finnegan et al. 2004). PPR proteins always regulate RNA processing metabolism in mitochondria and chloroplasts (Yagi et al. 2013; Barkan and Small 2014). Several reports have described the critical roles of PPR proteins in mitochondrial respiratory pathways (Andrés et al. 2007; Liu et al. 2010; Takenaka, 2010; Manavski et al. 2012; Haïli et al. 2013; Qi et al. 2017a and 2017 b; Lee et al. 2017). The mutation of PPR proteins leads to the splicing defects of mitochondrial genes and affects the assembly and/or stability of Complexes (Cohen et al. 2014; Chen et al. 2017a; Qi et al. 2017a; Sun et al. 2019).

Cotton (Gossypium hirsutum L.) is one of the important industrial crops, providing an important source of renewable textile fibers and oilseeds. However, few studies have reported the functions of PPR proteins in post-transcriptional processing in cotton. In this present study, we cloned GhImA gene, which is a P-type PPR protein encoding gene, and its encoded protein targets mitochondria. We confirmed that both GhImA and GhImD could bind to nad7 mRNA in mitochondria and participate in nad7 precursor mRNA splicing (Figure 4 and Figure 5). Compared with TM-1, mature transcripts of nad7 decreased and unspliced forms of intron2 of nad7 increased in the im mutant. As a result, Complex I pathway was damaged in the mutant. Koprivova et al. (2010) identified a pentatricopeptide repeat (PPR) protein which implicated in splicing of intron1 of nad7 transcripts in Arabidopsis, affecting assembly of Complex I and resulting in moderate growth retardation. Here, our reported PPR protein, GhIm participated in splicing of intron2 of nad7 transcripts, and its mutation ultimately leading to non-fluffy fiber phenotype in cotton.

In summary, absence of the GhImA protein affected nad7 precursor mRNA splicing, further disturbed Complex I, along with down-regulation of ATP synthase related genes and up-regulation of AOX. This is the first report on PPR protein participating in mitochondrial gene processing in cotton. Further, how GhIm participates in nad7 transcriptional modification is interesting and requires further exploration.

ATP insufficiency and ROS burst caused immature fibers

The im mutant was first discovered by Kohel et al. (1974) and its candidate gene was provisionally identified as Gh_A03G0489 by Thyssen et al. (2016), but no further molecular mechanism was clarified. In previous studies, differential expression analyses suggested that the majority of DEGs between TM-1 and the im mutant are related to stress responses, cellular respiration (Kim et al. 2013a) and carbohydrate metabolism (Wang et al. 2014). However, the exact mechanism remains to be explored.

Here we confirmed that Gh_A03G0489 (GhImA) is responsible for the im phenotype based on map-based cloning, CRISPR/Cas9 and VIGS analysis. Molecular analysis showed that the GhImA protein targets mitochondria and regulates respiration pathway. Through the electron transport chain, mitochondria provide energy for cellular activities and produce superoxide anion to balance ROS homeostasis (Sweetlove et al. 2007). GhImA showed a constitutive expression in different vegetative and reproductive tissues, while GhImA mutation in the im mutant brought about abnormal phenotypes in fiber cell wall synthesis. There are At- and Dt-subgenome homologs in allotetraploid cotton. We speculated that even though loss of function of GhImA in im mutant, the energy supply might be sufficient to maintain its growth during the vegetative growth stages, but more energy was needed for cellulose synthesis and fiber development. The energy deficiency due to GhImA mutation brought about non-fluffy fiber phenotype in the im mutant.

Mitochondrial electron transport can produce ${\mathrm{O}}_2^{-}$ ROS. As an intracellular signal molecule, ROS is necessary for plant development, regulating adaptation to stress and defense, as well as other processes (Venditti et al. 2001; Matsuzaki et al. 2009). During the fiber development stages, H₂O₂ is considered to be a signal for secondary wall synthesis in cotton, with peak production at 16 DPA - 20 DPA (Potikha et al. 1999; Kurek et al. 2002), however, excessive ROS can cause damage, affect cellular metabolism and even stimulate programmed cell death (Chen and Lesnefsky 2006; Matsuzaki et al. 2009; O-Uchi et al. 2014). In the im mutant, inhibition of the electron transport chain of Complex I in respiratory pathway increased H₂O₂ levels or led to ROS burst (Figure 6 and Figure S10 in File S2). Gene Ontology analysis indicated that DEGs between TM-1 and im mutant fibers are related to stress or hormone response, and biological processes related to cell death (GO:0008219) were also enriched (Table S4 in File S1), that might be toxic to plant growth and development, and further exacerbate the immature fibers in the im mutant.

Subgenomic-dependent dosage effect of GhImA and GhImD in regulating plant growth and development

In recent years, mutations in PPR genes have been reported in Arabidopsis, maize, rice and tomato (van Rooijen et al. 2017; Manavski et al. 2012; Tang et al. 2017; Yang et al. 2017a and 2017 b), but none has been reported in polyploidy plants. In contrast to diploid plants, polyploidy plants were derived from fusions of different diploid progenitors, and contain gene redundancy and homologous genes, which bring about various genetic effects. Thus, mutations in PPR proteins in diploid plants might be difficult to survive or show significant phenotype difference. For example, mutations in PPR4 in maize, ABA overly-sensitive 5 (ABO5) in Arabidopsis and PPR6 in rice were seedling lethal after germination due to insufficient respiration or photosynthesis (Schmitz-Linneweber et al. 2006; Liu et al. 2010; Tang et al. 2017), while polyploidy plants could have more tolerance due to the functional complementation of homologous genes.

The allotetraploid cotton genome contains the A and D subgenomes, which originated from a single hybridization event between the A- and D- diploid species (Wendel 1989). Wild type allotetraploid cotton has four copies of the GhIm gene, named AADD. Although polymorphic loci exist between the two sub-genomic genes, both GhImA and GhImD expressed equally in different tissues and their encoding proteins targeted into mitochondria to bind nad7 mRNA in mitochondrial with same functions. In im mutant, absence of the GhImA protein leads to a dramatic reduction in the mature nad7 transcripts, however GhImD can partially compensate the absence of GhImA during fiber development stages from 13 DPA to 25 DPA. Especially in 25 DPA fibers, the transcripts of GhImD were higher in the im mutant than in TM-1 and effectively complement the deficiency of GhImA for fiber development, implying that GhImA and GhImD proteins have similar roles with dosage-dependent effect.

By silencing transcripts, we observed that plants with different copies of the GhIm gene had different growth and development phenotypes. TM-1 plants with four copies of the GhIm gene (AADD), or three copies (AaDD; from CSI028×im, TM-1×im and I4005×im F₁ plants) had normal seed germination, seedling growth, flowering and ripening. The im mutant with two copies of the GhIm gene (aaDD) showed no significant difference in vegetative growth compared to TM-1 plants, but displayed the immature fiber phenotype. When the GhIm gene was completely silenced (TRV: GhIm; aadd), the plants appeared to stop growing and even died. At the same time, the transcription of ATP synthase also reduced with decreasing copy numbers of GhIm, and the alternative oxidases were activated. Correspondingly, respiration-induced ROS levels increased when GhIm copy number decreased. We also found that AOX1A might play more important role than AOX2 in coding for alternative oxidase in the mitochondrial electron transfer chain, which is consistent with the previous report (Ng et al. 2013). These results further confirmed that the GhImA and GhImD genes have dosage-dependent effect, and synergistically regulate growth and development in tetraploid cotton.

Though several PPR proteins have been reported in regulating plant growth and development, the underlying molecular mechanisms are still unclear. In this present study, we identified a nuclear-encoded PPR protein from cotton, which can bind to mitochondria nad7 mRNA for assisting its pre-mRNA splicing. Here, we postulated a probable working model in which GhIm protein plays an essential role in facilitating the organelle mRNA splicing. In wild type, PPR genes from At- (GhImA) and Dt- subgenome (GhImD) produce sufficient proteins to function for nad7 pre-mRNA splicing, lead to enough mature nad7 transcripts, and generate normal Complex I activities to allow the plant to grow healthy and produce fluffy fibers. However, mutation of GhImA in the im mutant leads to insufficient GhIm proteins to work for nad7 pre-mRNA splicing, resulting in less nad7 mature transcripts and decreased Complex I activities. Meanwhile, along with the disturbed respiratory chain pathways, alternative oxidase metabolism is up-regulated, ROS burst and stress or hormone response processes are activated. As a result, non-fluffy fiber phenotype is produced. In addition, GhImA and GhImD act in a dosage-dependent manner. In im mutant, the defect of insufficient GhImA could be compensated by GhImD at some developmental stages, for example, GhImD was highly expressed at late fiber developmental stages to produce more PPR proteins for compensating respiratory metabolism, leading to the non-fluffy fiber phenotype but not development arrest (Figure 9).

Figure 9.

Postulated working model for GhIm. In TM-1, PPR genes from At- (GhImA) and Dt- subgenome (GhImD) produce sufficient proteins to function for nad7 pre-mRNA splicing, lead to enough mature nad7 transcripts, and eventually show the fluffy fiber phenotype. In im mutant, mutation of GhImA in the im mutant lead to insufficient GhIm proteins to work for nad7 pre-mRNA splicing, resulting in less nad7 mature transcripts and non-fluffy fiber phenotype.