**3. Discussion**

RNA editing is a common biological process in the plastids and mitochondria of land plants [30]. It is believed to be a post-transcriptional correction mechanism to circumvent mutations in the organellar genomes [11]. Since RNA editing is an energy-consuming and complex process, it must serve important biological functions, or otherwise it should not be maintained in higher plants during evolution. One possible advantage of RNA editing is to provide an additional regulatory mechanism to organelle physiology. RNA editing was shown to play important roles in organellar tRNA maturation [31] and intron splicing [18,19]. Several hundreds of RNA editing sites on mitochondrial transcripts and differential RNA editing were observed in our previous study on *A. thaliana* [15]. We showed that overexpression of AtPAP2, a phosphatase dually targets the outer membranes of chloroplasts and mitochondria [32,33] and plays a role in the import of pMORF3 into mitochondria [5], might influence cyt *c* biogenesis by modulating RNA editing of *ccm* transcripts through its interaction with the MORF proteins [15].

Here, the objective is to examine how nodule formation affects the degrees of RNA editing in soybean mitochondrial transcripts and investigate its possible biological purpose. We identified 631 RNA editing sites with at least 15% of edited transcripts in all three biological replicates of any one of the samples (Supplementary Table S2). Only 12% of these sites were differentially edited between any two of the three samples. One of the mitochondrial transcripts that underwent extensive RNA editing was the *matR* transcript (Table 3), which encodes an intron maturase that mediates group-II intron splicing. In a previous study, 510 RNA editing sites were identified in Arabidopsis mitochondrial transcripts and 124 sites were differentially edited between the wild-type and a high-energy, fast-growing, transgenic line [15]. There are 12 RNA editing sites in Arabidopsis *matR* transcripts, but none of them were differentially edited between the two lines [15]. Hence, the differential editing of *matR* transcripts is specific in our soybean samples and is likely to play a role in nodule function (Table 2). While plant mitochondria have lost most of their intron-encoded ORFs, a single intron maturase gene (*matR*) is retained in the fourth intron of *nad1* in seed plants [21] (Figure 2). In addition to the mitochondrial *matR* gene, phylogenetic analysis shows that the nuclear maturase genes are also conserved between *A. thaliana* and soybean, since the homologous sequences of all four Arabidopsis *nMat* homologs could

be found in the soybean genome (Figure 2). Among the six soybean nuclear maturases, only a gene homologous to Arabidopsis *nMat4* exhibited a 2X higher expression in N than in UR (Supplementary Table S4).

By suppressing the expression of *matR* in Arabidopsis, the splicing efficiencies of *nad1* introns 3 and 4, *nad4* intron 3, *cox2* intron 1, and *ccmFc* intron 1 were downregulated, indicating that AtMatR carries out the splicing of these introns [27]. In soybean, the splicing efficiencies of these introns were higher in N and SR than in UR (Figure 3a). It should be noted that the abundance of *matR* transcripts were 27% and 38% higher in the SR and the N samples than in the UR sample (*p* < 0.005), respectively (Supplementary Table S1). Hence, the changes in the splicing efficiencies of these introns could be due to an increase in *matR* transcript abundance and/or due to the changes in RNA editing. The *nad4* intron 1 was not identified as a *matR*-mediated splicing site in Arabidopsis by *matR* knockdown experiments. However, it was identified in the MatR-ribonucleoprotein complex in an RNA co-immunoprecipitation experiment [27]. Hence, it is possible that RNA editing of *MatR* and the subsequent changes in its amino acid sequence might affect the composition of its associated ribonucleoprotein complex, thus affecting the splicing efficiency of *nad4* intron 1 by the other maturases. Alternatively, MatR might have an additional splicing activity toward *nad4* intron 1 in soybean. A P-type PPR protein, MISF68, was recently shown to be essential for the splicing of *nad4* intron 1 in *A. thaliana*, but the maturase responsible for its splicing has not yet been identified [34]. Nevertheless, several nuclear-encoded splicing factors can independently regulate the intron splicing in plant mitochondria, such as *nMat1-4*, mitochondrial transcription termination factors (*mTERF*s), and some PPR proteins. They are responsible for the splicing regulation of multiple introns in the mitochondrial genomes [35,36]. However, the most significant change between SR and N came from *nad4* intron 1 in the qRT-PCR results, suggesting that the splicing factor which targets the *nad4* intron 1 may play a role in nodule function.

Several *nad* genes are retained in the mitochondrial genome but are not transferred to the nucleus during evolution, probably because they encode the hydrophobic membrane subunits (NAD1-6) of the complex I [37]. Some *nad* introns are subjected to RNA editing (Supplementary Table S3). The latest research reveals that some RNA editing sites on the intronic sequences of maize mitochondrial *nad7* transcript can affect its intron splicing [19]. For example, in maize, it was previously shown that splicing would be abolished without the C-to-U editing at a specific position in two *nad7* introns [19]. In this study, 12 intronic RNA editing sites were identified, 11 of which were situated in *nad1*/*2*/*4*/*5*/*7* introns, including a site located in *nad4* intron 1 (base 102195, Supplementary Table S3). RNA editing of these intronic sequences could potentially affect intron splicing in soybean.

We observed the highest splicing efficiency of the *nad4* intron 1 in N compared to UR and SR (Figure 3), which might, in turn, affect the NAD4 protein abundance and supercomplex formation (Figure 4). In mammalian cells, spliced mRNA yielded more proteins than identical mRNA not made by splicing, possibly due to an enhanced association of spliced mRNA with polyribosomes [38]. This phenomenon is also observed in viruses [39]. Likewise, a higher splicing efficiency of the first intron of *nad4* in nodule (Figure 3a) might enhance the efficiency of translation. Evidentially, we detected a higher NAD4 protein abundance in nodule comparing to roots (Figure 4c) despite a similar *nad4* transcript abundance in all samples (Supplementary Table S1). NAD4 is a membrane component of complex I. In a maize *nad4* mutant (NCS2), while the missing NAD4 results in the destabilization of the NDH complex, a smaller complex can still exhibit complex I activity in the in-gel enzyme assay [40]. In BN-gel, complex I has two forms, either alone or forming a supercomplex (I + III2) with dimeric complex III, and their ratios are 40%/60% in plants, mammals, and fungi [41]. Our result showed that the mitochondria in nodules contain less I + III2 supercomplex, but contain a similar amount of free complex I to that of the root samples (Figure 4). Factors such as differential intron splicing, differential RNA editing of various *nad* transcripts and a change in NAD4 abundance could collectively affect the formation of the I + III2 supercomplex during nodulation. It was suggested that the I + III2 supercomplex may enable a more efficient electron transfer from complex I to complex III, as the close proximity of the two complexes enables efficient channeling of reduced quinol [42]. The lower

abundance of supercomplex I + III2 in nodule mitochondria might affect electron transport rate in the mETC [41].

In nodules, nitrogen fixation by *S. fredii* is an extensive energy-consuming process. To fix one molecule of N2, 16 ATP molecules and eight electrons (and eight H<sup>+</sup>) are required [43]. C4-dicarboxylates (succinate, fumarate, and malate) were suggested to be the primary carbon source provided to the rhizobia by the root cells [1], and a recent study showed that malate, but not succinate or fumarate, is the essential dicarboxylate for bacteroid growth and symbiosis [44]. Normal root mitochondria do not exhibit substantial glycine decarboxylase (GDC) as there is no photorespiration in roots [45]. However, a proteomic study showed that nodule mitochondria contain a significantly higher amount of GDC, but a lower amount of ATP synthase α and β chains than uninoculated roots [46]. A high glycine flux is present in nodule mitochondria due to nitrogen fixation and ureide biosynthesis [47]. The presence of GDC in nodule mitochondria will generate a large amount of NADH. If the electron transport in the mETC is tuned down by the downregulation of the formation of supercomplex I + III2 [42], more NADH generated from GDC can be exported from the nodule mitochondria in the form of malate, which can then be consumed by the bacterioid for energy production. In this study, we reported the changes in RNA editing and intron splicing of the mitochondrial genome in soybean nodules. The relationship among RNA editing, intron splicing, and supercomplex I+III2 formation is complicated and requires further investigation.
