*2.2. Di*ff*erential RNA Editing in Root Nodules*

RNA-seq reads were mapped to the mitochondrial genome. A total of 631 RNA editing sites, with at least 15% edited reads in all three biological replicates of any one of the three samples, were identified (Supplementary Table S2). The average editing degrees of the three groups (N, SR, and UR) were compared, and 74 sites with ≥ 15% differences in the editing degrees between any two of the three tissues were identified (Supplementary Table S3). Out of these 74 editing sites, 12 sites were intronic, 23 sites were synonymous, and 39 sites were non-synonymous. There were a few observations: (i) UR showed higher editing degrees in *atp1-1*, *atp1-2*, *atp1-3*, and *atp1-4* than the other two samples, whereas all these sites are synonymous; (ii) There were 29 differentially edited sites in *nad* transcripts (*nad1*, *nad2*, *nad4*, *nad4L*, *nad5* and *nad7*). Most of these sites were intronic (11 sites) or synonymous (nine sites). Only six and three sites in *nad1* and *nad4*, respectively, could lead to amino acid substitutions; (iii) There were 18 differentially edited sites in ribosomal protein transcripts (*rpl5*, *rps1*, *rps4*, *rps10* and *rps12*). Only one site was intronic (*rps10*), but all the other sites could lead to amino acid substitutions. Comparing N to UR, 16 sites in *rps 1*, *4*, *10*, and *12* had higher editing degrees, while two sites in *rpl5* had lower editing degrees.

#### *2.3. matR Transcripts Underwent Extensive RNA Editing in Root Nodules*

We identified 17 editing sites on *matR*, of which eight sites had been identified in *A. thaliana* [15] (Table 3) and 11 sites had been reported previously in uninoculated soybean by comparing genomic DNA and cDNA sequences [28]. Out of these 17 editing sites, the degrees of RNA editing of eight sites were higher in the N than UR and SR, of which seven could lead to amino acid substitutions. To confirm that these sites were actually edited, RT-PCR and Sanger sequencing were performed on three biological replicates of each sample (Supplementary Figure S1). In general, the Sanger sequencing data confirmed the next-generation sequencing (NGS) data, and both showed that N had a higher degree of editing than UR and SR in multiple editing sites (Figure 1 and Table 2).



**Table**

of editing between the samples were labelled by different letters.
