Two wrongs make a right: heat stress reversion of a male-sterile Brassica napus line

We have identified Tic40 as the restorer gene for the MSL system and hypothesize that heat treatment can also restore fertility via a cell cycle-induced developmental slow down.


Introduction
In the coming century, predictions say that climate change will cause the earth's temperature to increase by 3.7 °C (±1.1 °C) (Pachauri et al., 2014). Since most processes which control plant growth and development are known to be heat sensitive, not surprisingly elevated temperatures are regularly attributed to cause yield reductions in economically important crops (Peng et al., 2004;Zhao et al., 2017;Tigchelaar et al., 2018;Dusenge et al., 2019;Jagadish, 2020). The most heat-sensitive stages of plant growth involve key processes of reproductive organ development, with pollen development considered as the most susceptible (Lohani et al., 2020). Pollen function can be disrupted by heat stress in different ways, including reactive oxygen species (ROS) imbalance, differential hormone regulation (Ozga et al., 2017), premature pollen development (Parish et al., 2012), changes in carbohydrate and lipid metabolism (Jain et al., 2010), tapetal cell abnormalities (Ku et al., 2003), and disruption of meiotic cell division (Endo et al., 2009). While heat stress causes male sterility, there are now a few male-sterile plant lines which, in fact, are male fertile after heat stress or when grown at higher than normal growth temperatures (Zhu et al., 2010;Fernandez-Gomez et al., 2020). Understanding how heat stress can reverse male sterility may offer important clues about ways of engineering plants able to survive and reproduce in the current warming climate.
In general, male sterility is not a positive trait for any organism, but it plays a significant role in plant breeding, especially in crop production, to achieve heterosis. Heterosis (or hybrid vigour) describes the outperformance of the hybrid generation in comparison with its two parental lines. Heterosis can result in grain yield increases ranging from 20% to >50% (Tester and Langridge, 2010). Therefore, since male sterility prevents self-pollination of outcrossing crops, it greatly simplifies crop breeding by ensuring cross-pollination. In rapeseed (Brassica napus), significant heterosis effects have been documented and studied (Sernyk and Stefansson, 1983;Grant and Beversdorf, 1985). Presently, several male sterility systems exist, which are used as the main pollination control systems in rapeseed hybrid breeding, namely cytoplasmic male sterility (CMS), cytoplasmic induced male sterility (CIMS), self-incompatibility (SI), chemical hybridization agent (CHA), and genetic male sterility (GMS). Two recessive GMS systems, the 9012AB (also known as 7365ABC) system and the natural male sterility Lembke (MSL) system, have been extensively used in hybrid breeding in China and Europe, respectively (Frauen et al., 2003(Frauen et al., , 2006Deng et al., 2016). Interestingly, not only have both lines been reported to show fertility reversion upon heat treatment, but they also share potentially similar loci regulating male sterility.
Originally, the male sterility of the 9012AB system was thought to be caused by the interaction between three independent loci (BnMs3, BnMs4, and BnRf). However, more recent work has demonstrated that BnMs4 is most probably allelic to BnRf (Dong et al., 2012). Therefore, the male sterility of 9012AB is now understood to be conferred by BnMs3/Bnms3 and the multiallelic BnRf locus, of which the male-sterile allele is BnRf b (Dong et al., 2012;Dun et al., 2014;Xia et al., 2016). After intense mapping, BnMs3 was identified as a homologue of Tic40 (translocon of the inner envelope membrane of chloroplasts 40), a protein required for protein import into chloroplasts (Chou et al., 2003(Chou et al., , 2006Dun et al., 2011Dun et al., , 2014Li et al., 2012;Lee and Hwang, 2019). More recently, the male-sterile allele (BnRf b ) was demonstrated to be a newly formed chimeric gene, which is located to either the chloroplast or the nucleus (Deng et al., 2016;Xia et al., 2016). The BnRf b chimeric gene encodes a protein consisting of several segments and shows, among others, a sequence similarity to the Arabidopsis mitochondrial Hsp70 (Deng et al., 2016;Xia et al., 2016). The introduction of BnRf b into Arabidopsis plants results in complete male sterility (Z. . A mechanism was proposed, whereby BnRf b interacts with the E3 ligase BRUTUS (BnaBTS) on the outer envelope membrane of chloroplasts, which disrupts the normal ubiquitin-proteasome system causing toxic effects (Z. . These toxic effects are then possibly compensated by an interaction between BnMs3 and Toc33 (translocon of the outer envelope membrane of chloroplasts 33) or by environmental heat shock (Z. . To date, no direct interaction between the restorer BnMs3 protein and the male sterility-causing protein BnRf b protein has been observed (Xia et al., 2016).
This study utilized the natural rapeseed MSL system, which was discovered by the Norddeutsche Pfanzenzucht Hans-Georg Lembke (NPZ) breeding company and which is one of the main male-sterile lines used in hybrid production in Europe (Frauen et al., 2003(Frauen et al., , 2006. Previous work on this system demonstrated that it shares many similarities with the 9012AB system used in China. For example, both lines initiate pollen abortion during the microsporocyte stage, both lines display a premature or retarded degradation of the tapetum, interestingly even though the two lines contain different cytoplasm types they share the same temporary maintainer system, and both lines display a heat shock male fertility restoration (Luo et al., 2018). Intriguingly, the MSL system is also regulated by two genes: the restorer gene was identified as BnaC9-Tic40, which is the same gene that was identified for BnMs3 (Dun et al., 2011(Dun et al., , 2014, while the male sterility gene has not yet been defined. In this work, the male sterility gene of the MSL system was determined to be a gene with high sequence similarity to BnRf b from the 9012AB system. The proposed male sterility gene, now called BnChimera, was cloned and found to encode a chloroplast-targeted protein, which, when transformed into Arabidopsis, resulted in male-sterile flowers. We also demonstrated, in contrast to the work on the 9012AB system, that BnChimera directly interacts with the BnaC9-Tic40 restorer. One major quest of this study was to determine how heat treatment results in the reversion of male fertility in the MSL system. Here, we used RNA sequencing (RNA-seq) to identify target genes contributing to, firstly, the male sterility in the MSL lines and, secondly, to the reversion to male-fertile flowers after heat shock. Combining the RNA-seq experiment with previous published results of the 9012AB system, we propose that during the heat treatment, pollen development is stalled or potentially slowed down during pollen meiosis, which provides time for the affected plastids to produce enough fatty acids for the lipids of the pollen cell wall. This describes a contrasting but similar mechanism to that proposed for how cold treatment slowed development and reverted other temperature-sensitive genic male sterility (TGMS) lines to being male fertile (C. J. Zhu et al., 2020).

Plant material and growth conditions
All seeds for the MSL lines, including C545, SORA 1, A-Line, and B-Line, were provided by the NPZ (Hohenlieth, Germany). The C545 line represents a male-sterile line, which is derived from crossing the male-sterile A-line and the male-fertile maintainer B-line. SORA 1 is a commercially available male-fertile line, which comes from the MSL system. Brassica napus plants were grown under greenhouse conditions [standard: 16 h light (200 μmol m -2 s -1 , 23 °C), 8 h dark (18 °C)] until flowering. For the heat stress treatment, plants were placed in a Percival chamber for 3 d at 37 °C with a day/night rhythm of 20 h light and 4 h night. The humidity level was maintained at ~80-90%. These conditions were chosen as they provided the best laboratory conditions for the restoration of fertility. Neither day length nor humidity alone could restore fertility. Humidity was kept high to prevent plants from drying out. Arabidopsis plants were grown under controlled long-day conditions [16 h light (100 μmol m -2 s -1 , 22 °C), 8 h dark (18 °C), and 50% relative humidity] until flowering and then moved to a greenhouse. Nicotiana benthamiana plants were grown under standard greenhouse conditions.

Transient expression of fluorescent proteins in Nicotiana benthamiana
The full coding sequence of the BnChimera gene or only the sequence encoding the very N-terminal 545 amino acids were combined as N-terminal green fluorescent protein (GFP) fusions via Goldengate cloning (Binder et al., 2014;Chiasson et al., 2019). For the transient expression of the respective fluorescent proteins, leaves of 4-to 6-week-old N. benthamiana plants were infiltrated with Agrobacterium, and isolation of protoplasts was performed as described (Schweiger and Schwenkert, 2014). Fluorescence signals were detected by confocal laser scanning microscopy (Leica TCS SP5) as previously described (Schweiger and Schwenkert, 2014).

Stable transformation of Arabidopsis
For the assessment of the functionality of BnChimera in Arabidopsis, the full coding sequence of BnChimera and various truncated versions were assembled under the control of the putative native promoter of BnChimera via Goldengate cloning (Binder et al., 2014;Chiasson et al., 2019). The stable transformation was performed using Agrobacterium and the floral dip method (Clough and Bent, 1998). Positively transformed plants were selected by spraying with Basta © and confirmed by genotyping using the primers listed in Supplementary Table S1.

Pollen staining
Brassica napus pollen were stained for viability using acetocarmine solution (1% acetocarmine in 45% acetic acid which was first refluxed for 24 h and then filtered) (Heslop-Harrison, 1992). Pollen were examined under a light microscope (Leica DM1000) after staining.

Yeast two-hybrid assays
Yeast two-hybrid assays were performed using the Matchmaker Gold Yeast two-hybrid system (Clontech). All genes were cloned without predicted chloroplast-targeting peptides and predicted transmembrane domains. Bait plasmids (pGBK) of BnChimera and its truncated versions were cloned as fusions to the Gal4 DNA-binding domain via Goldengate cloning and transformed into the Y2HGold yeast strain (Chiasson et al., 2019). Prey plasmids (pGAD) of the various B. napus Tic40-like and BnaC9-Tic40 point mutations were cloned as fusions to the Gal4 activation domain via Goldengate cloning and transformed into the yeast strain Y187 (Chiasson et al., 2019). Each bait and prey combination was mated and plated first on selection medium lacking the amino acids leucine and tryptophan (-Leu,-Trp). To test for positive interactions, positively mated combinations were inoculated overnight cultures in liquid -Leu,-Trp medium. Overnight cultures were then diluted and spotted onto -Leu,-Trp solid medium or medium lacking leucine, tryptophan, adenine, and histidine (-Leu,-Trp,-Ade,-His). Growth on -Leu,-Trp,-Ade,-His medium indicated a positive interaction. For positive and negative controls, the murine p53 and Lamin (Lam) were mated with the SV40 large T-antigen, respectively (Li and Fields, 1993). For the analysis of protein expression in yeast, immunoblotting was performed on selected colonies. Colonies of mated interaction partners were inoculated in -Leu,-Trp medium and grown overnight at 30 °C. Cells were pelleted by centrifugation (700 g, 5 min), washed in 1 mM EDTA, and resuspended in 2 M NaOH. The same volume of 50% trichloroacetic acid (TCA) was added to NaOH. After centrifugation (14 000 g, 20 min, 4 °C), ice-cold acetone was used to wash the pellet. The centrifugation step was repeated, and 5% SDS and the same volume of SDS loading buffer were added to the pellet. The cells were vortexed and Tris base added if the sample buffer turned yellow. The yeast extract was incubated (37 °C, 15 min, shaking), centrifuged (14 000 g, 15 min), and the supernatant used directly for SDS-PAGE or stored at -20 °C until use.

Isolation of genomic DNA of B. napus
Genomic DNA isolation of B. napus leaves was done with the innuPREP Plant DNA kit (Analytikjena, Jena, Germany) according to the manufacturer's instructions.

RNA isolation from B. napus
RNA was isolated from unopened flower buds of 2 mm from the indicated genotype and treatments using the RNeasy Plant Mini Kit (Qiagen) according to the manufacturer's instructions. Plants were grown in three separate batches with several buds taken from different plants from each batch representing the three replicates used in the RNA-seq for each genotype and temperature treatment. RNA was DNase treated using the Turbo DNase I kit (Ambion) according to the manual. RNA integrity was assayed via agarose gel electrophoresis and quantified using a Nanodrop.

RNA-seq
Preparation of RNA-seq libraries (three replicates per genotype and temperature treatment) and sequencing using 150 bp paired-end mode was performed by Novogene Biotech (Beijing) using standard Illumina protocols. Quality control was performed using the FastQC software. Transcript abundance as transcripts per million and estimated counts were quantified on the gene level against the previously published B. napus reference genome using Salmon (Chalhoub et al., 2014;Patro et al., 2017). Differentially expressed genes (DEGs) were calculated using the 3DRANseq pipeline and were selected based on a log2-fold change ≥1 or ≤ −1 with an adjusted P-value of ≤0.05 (Guo et al., 2020). Gene Ontology (GO) term annotation of the B. napus genome was performed using Blast2GO (Conesa and Gotz, 2008). GO term enrichment analysis was performed using ClusterProfiler and enrichplot with a Holm-Bonferroni test correction (Yu et al., 2012).

Light microscopy
For microscopic analysis, we used 2 mm long unopened flower buds from the indicated genotypes and environmental treatments. Buds were fixed immediately after harvesting with 2.5% (w/v) glutaraldehyde (4 °C, at least 24 h) in 75 mM cacodylate buffer (2 mM MgCl 2 , pH 7.0), rinsed several times with fixative buffer, and subsequently post-fixed with 1% (w/v) osmium tetroxide for at least 2.5 h in fixative buffer at 20 °C. After five washing steps in distilled water, samples were stained with 1% (w/v) uranyl acetate in 20% acetone, dehydrated with a graded acetone series, and embedded in Spurr's low viscosity epoxy resin (Spurr, 1969). For light microscopy, semi-thin sections (1-2 µm) were cut with a glass knife (Pyramitome 11800, LKB).

qRT-PCR
The synthesis of cDNA from 500 ng of purified RNA (three replicates per genotype and temperature treatment) was performed with the iScript cDNA synthesis kit (BIORAD) as instructed by the manufacturer's manual. For quantitative reverse transcription-PCR (qRT-PCR), a 20 μl reaction contained 10 μl of LightCycler FastStart Essential DNA Green Master mix (Roche), 2 μl of diluted cDNA (10-fold dilution in H 2 O), and 1 μM oligonucleotides (Supplementary Table S1). The reactions were performed using the LightCycler96 (Roche) with the following program: one cycle at 95 °C for 60 s, 45 cycles of a three-step amplification (95 °C for 10 s, 60 °C for 10 s, 72 °C for 10 s), and finally a melting curve was performed. The 2 ΔΔ-ct method was used to calculate fold changes (Pfaffl, 2001). Both BnActin (BnaA02g00190D) and Tic40 (BnaA02g03180D) were used as reference genes for normalization using the method outlined in Hellemans et al. (2007), as they were found independently from the RNA-seq data to not change significantly in the conditions used.

Protein production
Recombinantly produced 15 N-labelled Bna-Tic40 variants A10 and C9, as well as 13 C, 15 N-labelled A10 were purchased from CRELUX (Martinsried, Germany) and used for all spectroscopic experiments at the indicated concentrations and buffer conditions.

Structure prediction
A structural model of Arabidopsis thaliana Tic40 (residues 298-447) was obtained from the AlphaFold Databank (https://alphafold.ebi.ac.uk) (Jumper et al., 2021;Tunyasuvunakool et al., 2021) and validated with experimental NMR data as well as a previously determined NMR structure of the C-terminal NP domain of Tic40 (Kao et al., 2012).

Heat stress restoration of fertility of the Brassica napus male-sterile C545 line
Under standard growth conditions, the C545 line displays normal vegetative growth, but typically presents a male-sterile phenotype (Fig. 1A). The male-sterile phenotype was characterized as flowers with shorter filaments and severely degraded anthers, completely devoid of pollen (Fig. 1A). However, upon short heat treatment at 37 °C, male fertility could be restored (Fig. 1A). The flowers now exhibit normal anthers, which contained viable pollen, visualized by a red staining with acetocarmine (Fig. 1B). Previous work on similar B. napus male-sterile lines had demonstrated that male sterility was caused via exaggerated vacuolation and abnormal expansion of the tapetal cells, which crushed the developing tetrads (Wan et al., 2010;Zhu et al., 2010;Dun et al., 2011). This led to an inability to distinguish middle layer cells from the others at approximately stage 9 of anther development. In stage 9 of anther development, microspores generate the exine wall and become vacuolated, which is followed by stage 10 with the initiation of tapetum degeneration (Sanders et al., 1999). To observe what occurs inside the anthers, 2 mm long buds were selected and analysed under light microscopy. Flower buds with a length of 2 mm should represent anthers at approximately developmental stages 9-10. Accordingly, in the SORA 1 control line, free microspores are readily visible. However, in the untreated C545 line, microspores cannot be distinguished from the surrounding tapetal cells ( Fig. 1C; Supplementary Fig.  S1). This is consistent with previous results demonstrating that the tetrads were typically surrounded with thick callose, which is degraded in male-fertile plants (Wan et al., 2010;Zhu et al., 2010;Dun et al., 2011). Anthers were also analysed after heat treatment at 37 °C for 3, 7, and 14 d, respectively ( Fig. 1C; Supplementary Fig. S1). Even though in the C545 line the tapetal cell layer may still be enlarged compared with the fertile SORA 1 line, fully developed microspores can be observed ( Fig. 1C; Supplementary Fig. S1). These results demonstrate that male fertility can be (partially) restored in the C545 line via a short 3 d heat treatment at 37 °C and that the physiological causes of this male sterility are consistent with previous reports of similar breeding systems (Wan et al., 2010;Zhu et al., 2010;Dun et al., 2011).

The restorer gene of the MSL system is BnaC9-Tic40
To determine the restorer gene of the MSL system, crosses between a male-sterile MSL A-line (rr) and fertile restorer line (RR) and subsequent selfing of the fertile F 1 plants resulted in a large F 2 progeny (n=2000) that were used for phenotyping of fertility/sterility and fine mapping of the Restorer (Rf) gene. Furthermore, a BAC (bacterial artificial chromosome) library generated from the above-mentioned MSL A-Line (rr) was screened to this end with simple sequence repeat-(SSR) flanking Rf markers (HMR0875, HMR1750, and HMR1882) to identify BAC clones harbouring the rf gene. At each step, screening of the BAC library was accompanied by phenotypic evaluation of the F 2 progeny to identify recombinant genotypes and validate the markers mentioned above.
Two BAC clones (BAClone27 and BAClone3) fished with the marker HMR0875 were subjected to Roche 454 Next Generation sequencing (Rx Biosciences, Ltd). After assembly, 236 contigs with lengths varying between 220 bp and 11 328 bp were recovered from the BAClone3 and subjected to physical mapping with the Rf markers HMR0875, HMR1750, and HMR1882 ( Supplementary Fig. S2) to identify linkage and recombination events (Supplementary Table S2).
Contig_8 exhibited a tight linkage with the marker HMR1750 (no recombinants were found with this marker) and was then subjected to an ORF search to identify genes underlying the linked region, and to BLAST searches against The Arabidopsis Information Resource (TAIR; Huala et al., 2001). The position of the marker HMR1750 was identified to be on the ORF of a Tic40-like protein (At5g16620, BnaC9-Tic40) representing a small 2 bp deletion leading to a frameshift and truncated protein ( Supplementary Fig. S2). Single nucleotide polymorphism (SNP) markers differentiating between the mutant (short Tic40, rr) and restorer gene (native Tic40, RR) were designed and validated on a large set of rr, Rr, and RR B. napus genotypes. The 2 bp deletion in BnaC9-Tic40 was confirmed via PCR and Sanger sequencing using the RCP170 primer pair ( Supplementary Fig. S2).

The male sterility of C545 is caused by a chloroplasttargeted chimeric protein
Due to the phenotypic and genetic similarities in the observed male sterility of C545 and knowing that BnaC9-Tic40 could act as a restorer gene for male fertility, we reasoned that the MS gene is most probably also the same or similar to that of recently characterized systems Deng et al., 2016;Xia et al., 2016). The putative MS gene of the MSL system was also mapped to a region on chromosome A07, which is very close to the regions recently described by Deng and Xia and their co-authors (Deng et al., 2016;Xia et al., 2016). In both cases, a similar chimeric gene was reported: amino acids 41-214 display a homology to the Arabidopsis At4g37510, which encodes an RNase III-like protein; amino acids 214-518 display a homology to At1g80070, which encodes ABNORMAL SUSPENSOR 2 (SUS2) or PRE-RNA PROCESSING 8 (PRP8) and which is a conserved member of the spliceosome (Grainger and Beggs, 2005); amino acids 762-1375 display a homology to At4g37910, which encodes a mitochondrial Hsp70 (mtHsc70-1); and amino acids 519-761 display no homology to any known protein (Deng et al., 2016;Xia et al., 2016) (Fig. 2A). The only major difference between the two previous published works is that Xia et al. (2016) localized the protein to chloroplasts whereas Deng et al. (2016) detected it in the nucleus. Therefore, we first determined if this same chimeric gene could also be found in the MSL system. For this, the three primer sets HY1, HY2, and HY3, respectively, described by Xia et al. (2016) were used for PCR amplifications on DNA isolated from the following lines from the MSL system: the male-sterile A-line and C545 line, the male-fertile maintainer B-line, and SORA 1, which is a male-fertile restorer line. Only for the A-line and C545 line did we obtain positive PCR fragments for all three primer pairs ( Supplementary Fig.  S3). These fragments were confirmed via sequencing and contained fragments identical to those identified previously (Deng et al., 2016;Xia et al., 2016). This indicates that the MS gene in the MSL system is most probably the same chimeric gene as described before (Deng et al., 2016;Xia et al., 2016).
Since the subcellular localization of this chimeric protein (from now on for simplicity referred to as BnChimera) is currently in dispute, we first investigated if it is a chloroplast protein like the restorer BnaC9-Tic40. To determine the subcellular location of BnChimera, two different GFP constructs were generated. The first construct contained the full coding sequence of BnChimera fused to the N-terminus of GFP. The second contained the first 545 N-terminal amino acids (BnChimera N-terminus) fused to the N-terminus of GFP. A transient expression of both constructs in tobacco epidermal cells showed a clear chloroplast localization, as evidenced by the overlap with chlorophyll autofluorescence (Fig.  2B). This clearly demonstrates that both the restorer gene BnaC9-Tic40 and BnChimera code for proteins targeted to the chloroplast. Here, it must be mentioned that the GFP signals of both BnChimeria constructs were consistent with either a stromal or a thylakoid localization within chloroplasts and not in the envelope membranes. This observation is important when interpreting the possible mode of action of BnChimera.
If BnChimera is in fact the MS gene, its protein should be able to confer male sterility on Arabidopsis plants. A plasmid was constructed that contained the BnChimera coding sequence under the control of its putative native promoter and transformed into Arabidopsis by floral dip. After selection of primary transformants with Basta®, positively transformed plants were allowed to grow normally. All the plants transformed with the BnChimera construct were male sterile and like the Brassica C545 flowers, as mature anthers failed to develop (Fig. 2C). Since Arabidopsis is self-pollinating, this also led to the development of extremely short empty siliques (Fig. 2C). This demonstrates that BnChimera can cause male sterility, indicating that it is the MS gene from the MSL system. We were then interested in which parts of the protein cause the male-sterile phenotype. To analyse this, five further constructs of the BnChimera were made: the N-terminus, C-terminus, middle, N-terminus+middle, and middle+C-terminus. None of these constructs resulted in male sterility in Arabidopsis (Fig. 2C), demonstrating that only the full-length protein of BnChimera is responsible for causing the male-sterile phenotype. A similar result was also observed recently by others (Z. .

BnChimera can directly interact with the restorer BnaC9-Tic40
Due to the observations that both the BnChimera and the restorer gene BnaC9-Tic40 encode chloroplast-targeted proteins, we explored if their proteins can functionally interact. For this, we performed yeast two-hybrid assays, with the bait plasmids containing constructs from BnChimera and either of its three individual sections (N-terminus, middle, or C-terminus) and The first 42 amino acids contain a predicted chloroplast-targeting peptide. Amino acids 41-214 display a homology to At4g37510, encoding an RNase III-like protein (coloured orange). Amino acids 214-518 display homology to At1g80070, encoding SUS2, which is required for pre-mRNA splicing (coloured red). Amino acids 519-761 display no homology to any known protein (coloured white). Amino acids 762-1375 display homology to At4g37910, which encodes a mitochondrial Hsp70 (mtHsc70-1, coloured blue). (B) Protoplasts of transiently transformed tobacco leaves expressing either the full coding sequence or the N-terminus of BnChimera tagged to the N-terminus of GFP, monitored by laser scanning microscopy. Chlorophyll autofluorescence (red channel) indicates the location of chloroplasts. Scale bar=5 µm. (C) On the left are schematic diagrams of the various constructs of BnChimera under the control of the putative BnChimera promoter transformed into Arabidopsis. On the right are representative images of the phenotypes observed for the flowers and siliques after transformation with either construct. A typical wild type (Col-0) is shown as a control. Scale bars=2 mm except for the BnChimera siliques photo, where it represents 3 mm.
prey plasmids containing either of the four Tic40-like proteins identified in the B. napus genome (Fig. 3). To ensure proper expression, all chloroplast-targeting signals and transmembrane domains were removed in the bait and prey plasmid constructs. Apart from the positive control, the only combination of bait and prey plasmids which resulted in a positive interaction was with the BnChimera middle section and BnaC9-Tic40 (Fig.  3B). While the full-length BnChimera bait construct should also code for the middle section, no interaction was observed. Further analysis revealed that the full-length BnChimera is not expressed in yeast, possibly due to its large size ( Supplementary  Fig. S4). So, the only Tic40 protein which positively interacts Fig. 3. The middle non-homologous section of BnChimera directly interacts with BnaC9-Tic40. (A) Schematic diagram of the domains of BnChimera used in yeast two-hybrid assays. (B) Four versions of BnChimera (full sequence, N-terminus, middle, and C-terminus) were tested for interactions with the four Tic40 proteins (BnaA2, BnaA10, BnaC2, and BnaC9) from Brassica napus. Bait and prey plasmids were individually transformed into the correct yeast strains and combinations were mated. Mated yeast were then plated on either growth medium (SD medium -Leu,-Trp) to assay correct mating or on selection plates (SD medium -Leu,-Trp,-Ade,-His) to assay for positive interactions. Positive interactions appear as white colonies on the selection plates. For positive and negative controls, murine p53 and Lamin (Lam) were each mated with the SV40 large T-antigen. (C) The middle section, which positively interacted with BnaC9-Tic40, was divided into four overlapping fragments (1, 2, 3, and 4) and re-assayed for the interaction with BnaC9-Tic40. Yeast two-hybrid assays were performed as in (B). (D) Fragment 1 from the middle section of the BnChimera was again divided into three overlapping fragments (A, B, and C) and tested for interaction with BnaC9-Tic40 in the same manner as in (B).
in yeast with BnChimera is BnaC9-Tic40, counter-intuitively interacting to the non-homologous middle section. This is in contrast with previous work where no interaction between BnaC9-Tic40 and BnChimera was observed (Xia et al., 2016;Z. Zhang et al., 2020).
To further determine the amino acids enabling the interaction of BnaC9-Tic40 with the BnChimera middle section, the middle section was divided into four overlapping parts (Fig. 3C). The yeast two-hybrid assay was then repeated solely against BnaC9-Tic40 (Fig. 3C). In this instance, section 1 was the only section interacting with BnaC9-Tic40 (Fig. 3C). In addition, when this section 1 was further divided into three more overlapping sections, part A showed the strongest interaction (Fig. 3D), indicating that these approximate 50 amino acids are the most important for the interaction between BnChimera and BnaC9-Tic40.
Since the B. napus genome encodes for four different Tic40-like proteins and only BnaC9-Tic40 interacts in yeast with BnChimera, and since it is also responsible for the restoration of fertility, we next sought to identify the difference between BnaC9-Tic40 and the other three Tic40-like proteins. A multiple sequence alignment demonstrates that the C-termini of the four Tic40-like proteins are highly similar (Fig. 4A). Interestingly, only five previously described amino acids are unique to BnaC9-Tic40 and are present in all plant species but not in the other three B. napus Tic40-like proteins (Fig. 4A) (Dun et al., 2014). To test the functionality of these five amino acids in BnaC9-Tic40, a site-directed mutagenesis was performed followed by a repetition of the yeast two-hybrid assays (Fig. 4B). In all cases, the amino acids were mutated to a corresponding amino acid present in BnaA10-Tic40, which is the closest to BnaC9-Tic40. The interaction between BnaC9-Tic40 and the BnChimera middle section was abolished only if amino acids 321 (F321V) and 343 (R343L) were mutated (Fig. 4B). Interestingly, Z.  also demonstrated that a similar mutation of amino acid 321 abolished the restorer activity of the BnaC9-Tic40 protein.

CD and NMR studies of BnaC9 and BnaC10-Tic40 proteins
The structures of the two Bna-Tic40 variants C9 and A10 were first analysed by far-UV CD spectroscopy (Fig. 5A). Both variants showed a typical spectrum indicative of an α-helical secondary structure. Since the initial NMR spectral quality of the C9 variant was not sufficient for a more detailed NMR analysis ( Supplementary Fig. S5), we chose to use the A10 variant for further experiments. In order to improve the stability of the A10 variant, we added the α-helixstabilizing solvent TFE and monitored the gain in secondary structure by CD spectroscopy. As shown in Fig. 5B, the addition of up to 10% TFE led to a gain in the α-helical secondary structure content, while higher TFE concentrations did not have any additional effect. Thus, we added 10% TFE to 15 N-labelled A10 and C9 samples (Fig. 5C) and compared the NMR spectral signatures using 2D-[ 15 N, 1 H]-HSQC experiments. The presence of TFE markedly improved the spectral qualities of both Tic40 variants. However, A10 still displays a much better spectral quality than the C9 variant. In these NMR spectra, the positions and the intensities of the signals are highly sensitive to changes in structure and dynamics. The analysis of the spectra shows a large number of changes in the signal positions as well as intensities originating from only a relatively low number of sequence variations between these two variants, suggesting that the structure and dynamics are altered.
In order to probe the structural state of the A10 variant that appeared to be most promising for a more detailed NMR analysis, we recorded a set of three-dimensional NMR experiments to obtain sequence-specific resonance assignments (Fig. 5D). The obtained 13 C-NMR chemical shift information was further used to determine the location of the secondary structure elements in the protein (Fig. 5E). Only α-helical secondary structure could be determined with this methodology, in line with the CD data. Furthermore, the variant residue positions that lead to abolished binding to BnChimera (F321V, R343L) in A10 are both located in α-helical secondary structure elements.
Next, we utilized the recently developed software AlphaFold (Jumper et al., 2021) to derive a structural model of the investigated Tic40 fragment (residues 305-457). The deposited model of Tic40 from A. thaliana, showing high sequence identity with Bna-Tic40, was in very good agreement with the location of the α-helical secondary structural elements obtained with NMR (Fig. 5F). Furthermore, the available NMR solution structure of the C-terminal NP domain of Tic40 (Kao et al., 2012) from A. thaliana overlaid very well with the predicted structural model as well as the herein experimentally determined α-helical secondary structure elements. The elongated shape of the model and the lack of larger hydrophobic clusters, which would be expected for a compactly folded protein, are consistent with the relatively low NMR spectral quality. The two mutations that exhibited the most pronounced effect on the ability to bind to BnChimera are located in the N-terminal segment with position 321 oriented toward the interior, whereas position 343 appears to be partially surface exposed. Thus, mutation at these positions most probably has an impact on the folding state of the N-terminal segment. This is supported by the observed reduced NMR spectral quality for the C9 variant indicating that this protein exhibits a higher degree of flexibility and a weakening of the folding state.
These observations allow for the conclusion that the binding between BnaC9-Tic40 and the BnChimera to restore fertility depends either on interactions mediated by the specific amino acids F321 and R343 or, more globally, by the less compact folding state of C9 that might facilitate the exposure of hydrophobic surfaces.

Transcriptomic analysis of C545 before and after heat treatment
Three different RNA samples were isolated for this study and used in two different comparisons (Fig. 6A). For comparison, one RNA was isolated from 2 mm long unopened buds from both C545 and SORA 1 flowers (C545 versus SORA 1, Fig.  6A), representing a comparison between sterile and fertile buds. In the second comparison, a third RNA sample was isolated from C545 buds after a 3 d heat treatment at 37 °C, which represents a comparison of heat stress fertility-restored buds and male-sterile buds from the same genotype (C545 heat versus C545, Fig. 6A). Isolated RNA samples were sequenced using the Illumina platform and subjected to RNA-seq analysis. For the two comparisons, a total of 20 743 DEGs (log2-fold change ≥1 or ≤ −1 with an adjusted P-value of ≤0.05) were identified with 4029 common DEGs (Fig. 6B). The DEGs for Comparison one, C545 versus SORA 1, contained 3677 down-regulated and 4398 up-regulated genes (Supplementary Table S3). In Comparison two, C545 heat versus C545, we identified 8558 down-regulated and 8140 up-regulated genes (Supplementary Table S3), indicating that the heat treatment resulted in more changes to the transcriptome than between untreated male-sterile and fertile buds.
For an understanding of the biological significance of the gene expression changes, a GO analysis was performed on the down-and up-regulated DEGs from both comparisons ( Fig. 6C; Supplementary Table S4). The GO analysis provides an insight into the important roles of enriched groups of genes in regulating male sterility and in restoring fertility in the C545 line. For the comparison of C545 versus SORA 1, the GO term enrichment analysis of the down-regulated Red highlighted amino acids represent essential amino acids for the interaction between BnaC9-Tic40 and the middle fragment of BnChimera. Orange highlighted amino acids represent amino acids which are different only in the BnaC9-Tic40 but which are not essential for the interaction with BnChimera. (B) Yeast two-hybrid assays to determine essential amino acids required for the interaction of BnaC9-Tic40 with the BnChimera middle fragment. Yeast two-hybrid assays were performed as in Fig. 3.
DEGs displayed an enrichment of GO terms related to 'fatty acid biosynthetic process' including 'l-phenyalanine catabolic process', 'S-adenosylmethioninamine biosynthetic process', 'phosphatidylcholine biosynthetic process', 'glucosinolate catabolic process', 'cinnamic acid biosynthetic process', and 'purine ribonucleoside salvage' (Fig. 6C; Supplementary Table S4). This indicates that the chloroplast localization of BnChimera possibly leads to a disruption of fatty acid synthesis in chloroplasts, ultimately resulting in male sterility. Interestingly, other chloroplast GO terms including 'photosystem I' and 'photosystem II' also appear, which could indicate broader chloroplast metabolic impediments besides simply fatty acid synthesis ( Fig. 6C; Supplementary Table S4). Also, polyamine biosynthesis appears to be down-regulated in C545 in comparison with SORA 1 (Fig. 6C; Supplementary Table S4). GO enrichment of the up-regulated DEGs in the C545 versus SORA 1 comparison demonstrated that buds in the C545 line are obviously stressed. Several stress-related GO terms are significantly enriched, including 'response to chitin', 'systemic acquired resistance', 'response to oxidative stress', 'response to wounding', 'response to oomycetes', 'defence response to fungus', 'hydrogen peroxide catabolic process', and 'cellular oxidant detoxification' (Fig. 6C; Supplementary Table S4). The finding of 'anaerobic respiration' in the up-regulated DEGs is possibly related to the fact that the microspores are constantly surrounded by a layer of thick callose, which may interfere with oxygen diffusion, possibly ending in the suffocation of the microspores leading to male sterility (Figs 1C, 6C; Supplementary Table S4).
GO enrichment analysis of the up-and down-regulated DEGs after heat treatment of the C545 line revealed some interesting results. For down-regulated DEGs, many GO terms were associated with the cell cycle, including 'mitotic cell cycle phase transition', 'male meiosis II', 'mitotic cell cycle', 'cytokinesis by cell plate formation', 'resolution of meiotic recombination intermediates', 'preprophase band assembly', 'meiotic sister chromatid cohesion, centromeric', and 'homologous chromosome pairing at meiosis' (Fig. 6C; Supplementary  Table S4). This possibly indicates that cell cycle progression is    (Jumper et al., 2021). The model for the C-terminal NP domain is in perfect agreement with the experimental NMR structure (Kao et al., 2012) (2LNM). Between the two domains, an additional α-helical segment (helix 6) was detected by NMR and is also visible in the structural model. inhibited or slowed down upon heat treatment, which could explain why pollen development is readily aborted during heat treatment in most plant species. A cell cycle arrest could also be linked to groups of GO terms enriched in 'DNA replication initiation', including 'mitotic chromosome condensation', 'nucleosome positioning', 'DNA replication, synthesis of RNA primer', and 'mitotic DNA replication imitation' (Fig. 6C; Supplementary Table S4). Interestingly, several GO terms associated with 'regulation of cyclin-dependent protein serine/threonine kinase activity' such as 'regulation of cytokinesis', 'regulation of mitotic spindle organization', and 'mitotic spindle assembly checkpoint' are also found within the down-regulated DEGs ( Fig. 6C; Supplementary Table S4). The above, when also accompanied by a down-regulation of GO terms encompassing 'sporopollenin biosynthetic process', 'pollen development', 'pollen exine formation', 'anther wall tapetum development', and 'anther development' would all suggest that after heat treatment, pollen development should be inhibited or blocked, but is clearly not in C545 (Figs 1A, 6C; Supplementary Table S4). Strikingly, fatty acid biosynthesis appears to be down-regulated in the C545 line after heat treatment ( Fig. 6C; Supplementary Table S4). Surprisingly, within the after heat treatment up-regulated DEGs, many of the same GO terms which were enriched in the C545 versus SORA 1 comparison are also found in the up-regulated DEGs of the C545 heat versus C545 comparison ( Fig. 6C; Supplementary Table S4). Many of these enriched GO terms represent stress responses. These include heat stress-related GO terms 'response to heat' and 'heat acclimation'. Other GO terms relate to 'response to chitin' or 'defence response to oomycetes' (Fig. 6C; Supplementary Table S4). This could indicate that, since the C545 line is already showing indications of stress, the buds are already primed to respond to the heat stress treatment.

DEGs encoding chloroplast-targeted proteins
Since it has now been established that both BnChimera and BnaC9-Tic40 interact and are localized to chloroplasts, a more detailed analysis of DEGs coding for chloroplast-targeted proteins was performed. To identify chloroplast-targeted proteins in B. napus, firstly a list of known chloroplast proteins from A. thaliana was prepared using data from the SUBA4 database (Hooper et al., 2017). A B. napus protein was then considered as chloroplast localized if its top BLAST hit was from the Arabidopsis chloroplast list. This analysis identified a list of 4716 genes which potentially encode chloroplast-localized proteins in B. napus. Of these 4716 genes, 1382 genes were identified as DEGs in our dataset and 203 genes overlap between the two different comparisons ( Fig. 7A Table S5). This again demonstrates a stronger influence of the heat treatment on gene expression than any difference between untreated male-sterile or fertile flowers. GO enrichment analysis of DEGs coding for chloroplasttargeted proteins indicated that many of the same processes in chloroplasts are down-regulated in both comparisons ( Fig. 7B; Supplementary Table S5). For example, in both the C545 versus SORA 1 and C545 heat versus C545 comparisons, GO terms such as 'fatty acid biosynthetic process', 'acetyl-CoA biosynthetic process from pyruvate', 'photosynthesis, light harvesting in photosystem I', and 'photosystem I' were over-represented ( Fig. 7B; Supplementary Table S5). This indicates that in both the sterile untreated and heat-treated fertile flowers, although fatty acid synthesis is potentially the main site of inhibition, photosynthesis may also be perturbed, which is most probably a secondary effect (Fig. 7B, Supplementary Table S6). In reference to GO terms of the up-regulated DEGs, in the C545 versus SORA 1 comparison, only two enriched GO terms were identified while for the heat-treated comparison again several stress-related GO terms were found, including 'response to water deprivation', 'leaf senescence', 'response to red light', and 'response to cold' (Fig. 7B; Supplementary Table S6). When combining all these transcript data, what is lacking is a definitive chloroplast function in the down-regulated population of C545 versus SORA 1 and in the up-regulated population after the heat treatment, or vice versa. Not observing such a group makes it difficult to pinpoint the exact inhibition, which leads to either the male sterility in the C545 line or to the restoration of fertility after the heat treatment.

Analysis of several affected functional groups
Since no clear pathway reversion caused by the heat treatment was obvious, we looked closer at some more consistent changes in the transcriptome data. In both comparisons, fatty acid biosynthesis was down-regulated. Indeed, we identified all genes encoding proteins involved in chloroplast lipid biosynthesis to be differentially expressed in at least one of the comparisons. These chloroplast lipid biosynthesis genes were identified by taking the homologues of the previously published proteins (Holzl and Dormann, 2019). Out of 230 genes involved in chloroplast lipid biosynthesis, 99 genes were DEGs within the two comparisons ( Fig. 8A; Supplementary Table S7). These 99 chloroplast lipid biosynthesis-associated DEGs were mostly down-regulated in both comparisons but more so after the heat treatment of C545 flowers (Fig. 8A). Therefore, lipid synthesis impairment could potentially be the reason for the male sterility as lipids are required for pollen cell wall formation, essential for pollen development and male fertility (Hater et al., 2020). Because BnaC9-Tic40 is a chloroplast protein import component, we also analysed DEGs known to be involved in chloroplast protein import ( Fig. 8B; Supplementary Table S7). However, only 13 DEGs were identified out of a possible 130 genes coding for chloroplast import components in the Brassica genome. This indicates that there are no major changes in the abundance of protein import components and that protein import into the chloroplast is not affected. As expected, in the C545 versus SORA 1 comparison, the transcript coding for BnaC9-Tic40 is also down-regulated, which is not surprising as it contains a premature stop codon and would produce a truncated non-functional protein ( Fig. 8B; Supplementary Table S7).
One of the interesting findings from the GO term enrichment analysis was that many cell cycle-related GO terms were significantly enriched within the down-regulated DEGs in the C545 line after heat treatment (Fig. 6C; Supplementary Table  S4). This may hint at a delay in pollen development within the flowers, slowed down or blocked after the heat treatment. Indeed, pollen development is considered as the most heatsensitive stage in plant development (Lohani et al., 2020). So, the DEGs from both comparisons were searched for genes related to the following selected GO terms: 'pollen development', 'anther wall tapetum development', 'pollen exine formation', 'sporopollenin biosynthetic process', 'anther development', and 'meiosis'. Overwhelmingly, the majority of DEGs from these GO terms are down-regulated after the heat treatment of the C545 line ( Fig. 8B-H; Supplementary  Table S7). This is a little surprising considering that in some cases these genes are not down-regulated between male-sterile C545 and fertile SORA 1 plants. Within these down-regulated cell cycle genes, developmental DEGs are homologues of some well-known genes related to anther and pollen development including: ABORTED MICROSPORES (AMS) (Sorensen et al., 2003), MYB80 transcription factor (MYB80) (Phan et al., 2011), MYB35 (Zhu et al., 2008), cysteine protease CEP1 (Zhang et al., 2014), basic helix-loop-helix proteins 89 and 91 (bHLH89 and bHLH91) (Zhu et al., 2015),  Table S7). This illustrates that in the C545 line after heat treatment, counterintuitively anther and pollen developmental genes are down-regulated, which, when combined with a potential lipid synthesis bottleneck, may lead to fertility restoration.

Discussion
The MSL system from Europe and the 9012AB/7365ABC system from China are two of the world's main pollination control systems for hybrid rapeseed production (Chen et al., 1993;Frauen et al., 2003Frauen et al., , 2006. Even though rapeseed is the second most economically important edible oil seed crop, the identity of the genes for both male sterility (MS) and restorer of fertility (Rf) in either system remained unknown until recently. In fact, both systems seemed to have the same temporary maintainer system and showed very similar morphological observations during pollen abortion (Luo et al., 2018). In this study, we present data that show that both the MSL system and 9012AB share the same MS gene called BnChimera (also known as Bnrf b ), which encodes a chloroplast-targeted chimeric protein (Dun et al., 2011(Dun et al., , 2014Li et al., 2012;Deng et al., 2016;Xia et al., 2016;Z. Zhang et al., 2020). Using Arabidopsis, it could be demonstrated that BnChimera requires all its domains to cause male sterility. The restorer of fertility gene in both systems was shown to be BnaC9-Tic40. However, in contrast to previous work (Xia et al., 2016), we found a direct and specific interaction between the BnChimera and BnaC9-Tic40 proteins. Interestingly, both systems display a heat stress-reversible male fertility, which we attribute to a slowdown of pollen development. Our results also led us to offer an alternative hypothesis regarding how BnChimera causes male sterility, about how BnaC9-Tic40 rescues male fertility, and about how heat stress may also revert the male-sterile phenotype.

How does the male sterility system work in the MSL and 9012AB lines?
In previous work on the 9012AB male sterility system, an extremely complicated explanation about how the system functions was proposed (Dun et al., 2011(Dun et al., , 2014Li et al., 2012;Deng et al., 2016;Xia et al., 2016;Z. Zhang et al., 2020). The explanation was that the chloroplast-targeted BnChimera interacts with the nuclear-localized E3 ligase BTS and thereby redirects BTS to the chloroplast outer envelope membrane. There, it is thought to affect chloroplast protein translocation by disrupting the normal ubiquitin-proteasome system causing toxic effects within the chloroplast, ultimately resulting in male sterility (Z. . These defects can be overcome via either heat treatment, which suppresses the interaction between BnChimera and BTS as demonstrated in yeast two-hybrid experiments (Z. , or via an interaction between BnaC9-Tic40 and Toc33. This interaction then sends a signal to the nucleus to inhibit BnChimera expression or to remove the BnChimera protein through an unknown mechanism, thus eliminating its toxic effects (Z. . Here, we would like to offer a more straightforward hypothesis about how these and the MSL system function based on our own and previously published results. Firstly, there is no evidence in any previously published work or in this study that the BnChimera is located at the outer envelope of chloroplasts. In fact, the GFP results presented here and in previous work would support a stromal location for BnChimera. In most cases, outer envelope proteins generally display a halo-like pattern around the chloroplast (Breuers et al., 2012;Teresinski et al., 2019). However, in our analysis and in those of others (Xia et al., 2016), the BnChimera GFP pattern is rather diffuse and distributed within chloroplasts more like a stromal protein. In fact, BnChimera does not contain any predicted transmembrane domains, which makes it unlikely that BnChimera could interact in vivo with anything on the outer envelope of chloroplasts or with a nuclear-localized protein. More plausible is that BnChimera is targeted to the stroma where it interacts with currently unidentified protein(s) that results in the disruption of, for example, fatty acid synthesis which would lead to male sterility (Fig. 9A). Disruption of processes affecting plastid fatty acid synthesis and export have been previously demonstrated to cause male sterility in plants (Li et al., 2015;T. Zhu et al., 2020;Zhang et al., 2021). A possible involvement of BnChimera in lipid biosynthesis was previously proposed after conducting cutin and wax measurements (Xia et al., 2016). A disruption of fatty acid biosynthesis caused by BnChimera is also supported by our RNA-seq experiments, since down-regulated DEGs were enriched in processes of fatty acid and lipid biosynthesis. It is also possible that chloroplast membrane lipid trafficking is affected, for example FAX1, which is responsible for the export of fatty acids from plastids, is heavily up-regulated in C545 (Li et al., 2015). Ultimately, this disruption of fatty acid synthesis or export from plastids leads to male sterility due to the inability to synthesize any lipids required for the pollen cell wall (Fig. 8A).
Secondly, we could demonstrate that the restorer gene product BnaC9-Tic40 directly interacts with BnChimera. Since this involves the C-terminus of BnaC9-Tic40, which is known to face the stroma, this also strongly indicates a stromal location of BnChimera. A direct interaction between the MS and Rf gene products allows a more clear-cut explanation for the restoration of male fertility. Since in the male-sterile lines bnaC9-tic40 contains a premature stop codon, the protein is missing its C-terminus, which is responsible for the interaction with BnChimera (Dun et al., 2011). Our hypothesis here is that, when full-length BnaC9-Tic40 is present, it interacts with BnChimera within the stroma of chloroplasts and prevents BnChimera from interfering with fatty acid synthesis (Fig. 9C). This may probably be just by way of competition. Since out of the four Tic40-like proteins encoded within the B. napus genome only BnaC9-Tic40 can interact with BnChimera, it is the only one which can confer the restorer function. Interestingly, the amino acid phenylalanine at position 321, which was demonstrated to be essential for the restorer function of BnaC9-Tic40, is also essential for the interaction with BnChimera ( Fig. 4B) (Xia et al., 2016;Z. Zhang et al., 2020). It must also be pointed out that any interaction between Toc33 and Tic40 in the outer envelope membrane is not under scrutiny here. There is no evidence for any Tic40 being located within the outer envelope, and it is considered as an inner envelope membrane marker protein with its C-terminus facing the stroma (Stahl et al., 1999;Bedard et al., 2007;Chou et al., 2003). Therefore, it is extremely unlikely that the C-terminus of BnaC9-Tic40 interacts in vivo with Toc33.

How does heat stress revert male sterility?
Since BnChimera is unlikely to be localized to the outer envelope of chloroplasts, heat stress inhibition of the interaction with BTS is also unlikely. In our observations, transcript abundance of BnChimera after heat treatment does not significantly change ( Supplementary Fig. S6). It is assumed that even after heat treatment, the BnChimera protein is still found in chloroplasts; however, it cannot be completely ruled out that the BnChimera is unstable after heat treatment or that its interaction with other proteins is disrupted. Therefore, we offer an alternative hypothesis about how heat treatment reverses male sterility.
The developmental process of pollen, including microgametogenesis, depends on a strict and timely coordination of meiosis, mitosis, cell growth, and expansion (Sanders et al., 1999). One of the reasons why BnChimera causes male sterility is that during the relatively fast development of pollen, chloroplasts cannot synthesize sufficient material for pollen wall formation, which ultimately leads to defective microspores and male sterility. These effects are then reversed via heat treatment. From the RNA-seq analysis of the C545 malesterile line, the largest group of down-regulated DEGs after heat treatment is involved in cell cycle progression including meiosis and mitosis, as well as in pollen development. It has been demonstrated before that heat stress affects pollen meiosis (Sakata et al., 2000). Therefore, we hypothesize that during heat stress pollen development is slowed down, which allows chloroplasts more time to synthesize sufficient material for pollen cell wall development. This combination of slower development and slower synthesis of cell wall material results in fertile pollen (Fig. 9B). A similar phenomenon was recently described in Arabidopsis TGMS lines, in which lower temperatures restored fertility by slowing pollen development (C. J. Zhu et al., 2020). Interestingly, in one of these studies, the authors concluded that, at least in Arabidopsis, heat treatment did not restore fertility but reduced it (C. . These experiments, however, were performed on Arabidopsis T-DNA insertional mutants, whereas in rapeseed the mutation is not a knockout. The development of the male reproduction system in plants has been demonstrated many times to be extremely sensitive to adverse environmental conditions (De Storme and Geelen, 2014). This includes both heat and cold stress whereby heat stress is typically observed to show the premature disappearance of the tapetal cell layer in conjunction with microspore development issues (Ku et al., 2003;Abiko et al., 2005;Oshino et al., 2007). Cold stress, on the other hand, displays an almost opposite effect in that the tapetum persists right up until the mature pollen stage (Mamun et al., 2006;Oda et al., 2010). In both cases, the environmental temperature results in male The C545 line under normal conditions is male sterile because bnac9-tic40 is truncated at the C-terminus and therefore unable to interact with BnChimera. This allows BnChimera to disrupt, block, or inhibit fatty acid synthesis within chloroplasts. Male sterility results from chloroplasts' inability to produce enough fatty acids in time for the rapid pollen development. (B) At higher temperatures, pollen development is severely slowed down, which in turn provides the chloroplasts with time to produce enough lipids, restoring male fertility. (C) This is equivalent to the restorer line situation, in which BnaC9-Tic40 contains its complete C-terminus and can interact with BnChimera. In this situation, the flowers are male fertile as the interaction with BnaC9-Tic40 blocks the BnChimera-mediated inhibition of fatty acid synthesis. Created with BioRender.com. sterility; however, both heat and cold stress have also been demonstrated to cause male fertility restoration (Fernandez-Gomez et al., 2020;C. Zhang et al., 2020;J. Zhu et al., 2020). Heat stress fertility restoration was recently shown to be important at an early stage of pollen development, more specifically prior to pollen mitosis I (Fernandez-Gomez et al., 2020). This may be relevant to the C545 line analysed here as the DEGs were enriched for cell division machinery in flowers heat treated early on in flower development. This could be the developmental slowdown we have hypothesized. Cold stress, on the other hand, leads to an overall slowdown of development not just on any specific pathways (C. J. Zhu et al., 2020). It also leads to tapetum layer retention, which is already evident in the C545 sterile flowers, so cold treatment may only compound the tapetum developmental problems at normal growth temperatures. Heat stress was also demonstrated to lead to chloroplast overdevelopment (Oshino et al., 2007). Therefore, it could also be hypothesized that in the C545 line the BnChimera could block chloroplast development at normal growth temperatures, but chloroplast development can then be induced during the heat stress; something to consider since the BnChimera is localized to chloroplasts. However, since the observed changes in gene expression after heat treatment support a reduction in transcripts related to cell cycle and cellular division, it is possible that the short (3 d) heat treatment leads to a temporary halt or slowing down of pollen development most probably related to cellular division, which allows the perturbed chloroplasts more time to synthesize sufficient material for pollen cell wall development.

Potential for using BnChimera to create thermotolerant pollen
Due to the adverse effects of climate change, pollen thermotolerance is becoming an increasingly important economic trait for breeding. Therefore, the ability of the C545 line to produce seeds at high temperatures potentially could be used to help create thermotolerant pollen. This means that understanding how the MSL system can produce viable pollen after heat treatment becomes more relevant. The ability of the C545 line to survive heat could be related to the genes we identified as up-regulated DEGs in the C545 line in comparison with SORA 1. The vast majority of these DEGs were related to various stress responses. Due to this large number of stress-related up-regulated DEGs, the C545 line is already primed for responding to stress. Similar observations have been made before whereby the adverse effects of heat stress can be circumvented to a certain extent when plants experience a 'pre-conditioning treatment' (Chaturvedi et al., 2021). This is normally achieved by exposing plants to a mild stress treatment, which is followed by a brief recovery period whereby an acquired thermotolerance is induced, allowing the plants to survive a normally lethal heat stress (Chaturvedi et al., 2021). The capacity for plants to acquire thermotolerance has been attributed to the ability of cells to produce and store certain proteins, which enhance their resistance to higher temperatures (Larkindale and Vierling, 2008). This was recently demonstrated for tomato plants, in which heat treatment of 50 °C results in reduced pollen germination rates. However, if plants are pretreated for a brief period at 32 °C followed by a short recovery phase at 25 °C, this enhanced the tolerance to 50 ° C (Firon et al., 2012;Jegadeesan et al., 2018). Of course, thermotolerance is not limited to reproductive tissues, as a gradual increase in temperature versus a sudden 40 °C heat treatment of seeds also showed that the slower acclimation increased thermotolerance (Stone and Nicolas, 1995). In the MSL system, the expression of the full-length BnChimera not only produces sterile pollen, but also induces enough changes within the transcriptome, which equate to a stress pre-treatment. Thus, if the expression of the BnChimera could be used to induce this pre-treatment effect without causing male sterility, this would be invaluable. Potentially using the fragments of BnChimera, as we have here, could produce a similar transcriptome stress response but no male sterility, which warrants further experimental work. Alternatively, BnChimera could be transformed into other crop plants for inducing stress responses and then the BnaC9-Tic40 could be transiently expressed via the newly developed spray-on viral transfection technology to produce viable pollen (Massel et al., 2021;Torti et al., 2021). This would allow a tuneable expression and control of pollen development.
In conclusion, this study has identified the MS gene of the MSL system and demonstrated that the regulation of the MSL system occurs within chloroplasts. Furthermore, using RNAseq provided new insights into the thermosensitive male sterility system, displaying a high temperature fertility restoration. Overall, the results may prove useful for future plant breeding strategies as the MSL sterile lines display signs of a conditional pre-treatment to stress, potentially allowing the plants to survive heat. When combined with the need to produce crop plants, which are resistant to a future warming climate, these findings could be used to design plants which display thermotolerance during pollen development.

Supplementary data
The following supplementary data are available at JXB online. Fig. S1. Light microscopy of developing anthers from SORA 1 and C545 before and after heat treatment. Fig. S2. Identification of BnaC9-Tic40 as the restorer gene of the MSL system. Fig. S3. Determination of the existence of the MS gene in different lines of the MSL system. Fig. S4. BnChimera expression analysis from yeast. Fig. S5. 2D-NMR evaluation of the Tic40 variants A10 and C09. Fig. S6. qRT-PCR of the BnChimera transcript before and after heat treatment. Table S1. Primers used in the study. Table S2. Assembled contigs of clone BAC3 and physical positions of the Rf-linked markers. Table S3. List of differentially expressed genes. Table S4. GO enrichment data. Table S5. List of differentially expressed genes encoding chloroplast-targeted proteins. Table S6. GO enrichment data for DEGs encoding chloroplast-targeted proteins. Table S7. List of DEGs used in the heatmaps in Fig. 7.

Author contributions
JS, BB, and CC: conceptualization and design; PS: performing and coordinating most experiments; CE: performing the yeast two-hybrid assays; IG: performing microscopy; AA, FD, and GL: providing the mapping data for the Rf gene; KK and FH: performing NMR and analysing the structural data. All authors reviewed the manuscript.

Conflict of interest
AA, FD, and GL are employed by the company NPZ Innovation GmbH. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Data availability