*2.1. Transgenic Process Induces Significant Epigenetic Changes in Rice*

To investigate global methylation caused by the transgenic process in plants, we used WGBS to sequence samples of four types of rice: Wuyunjing7 wild type (WYJ7), T0 plants regenerated from callus (CK), wild type plant derived from segregation of *pOsNAR2.1*-*OsNAR2.1* (SDWT) and *pOsNAR2.1*-*OsNAR2.1* plants. T4 generation SDWT plants served as true experimental controls for the *pOsNAR2.1*-*OsNAR2.1* line to study the transgene insertional effect on the epigenome as it was derived from the T0 *pOsNAR2.1*-*OsNAR2.1* heterozygote line (for the selection process see Supplementary Table S1). T4 generation *pOsNAR2.1*-*OsNAR2.1* plants with high yield and high NUE phenotypes have been described in Chen *et al*. [33]. We also used MeDIP sequencing on four more samples, including the wild type of Nipponbare (NP), the knockdown plant of *OsNAR2.1* by RNA interference (RNAi) (describe as r1 in Yan *et al*. [29]), and overexpression plant of *OsNAR2.1* by the ubiquitin promoter (*pUbi-OsNAR2.1*). The relationships between different samples are described in Table 1 as: CK = WYJ7 + callus inducing media(CIM) + shoot inducing media(SIM); SDWT = WYJ7 + CIM + SIM + selection pressure (Hygromycin, Hyg) + transformation with the *pOsNAR2.1-OsNAR2.1* construct + segregation of *pOsNAR2.1-OsNAR2.1* heterozygote line; *pOsNAR2.1-OsNAR2.1* = WYJ7 + CIM + SIM + selection pressure (Hyg) + transformation with the *pOsNAR2.1-OsNAR2.1* construct + *pOsNAR2.1-OsNAR2.1* insertion; *pUbi-OsNAR2.1* = NP + CIM + SIM + selection pressure (Hyg) + transformation with the *pUbi-OsNAR2.1* construct + *pUbi-OsNAR2.1* insertion; RNAi = NP + CIM + SIM + selection pressure (Hyg) + RNA interference (RNAi) construct. Table 1 also shows the values for raw reads, uniquely mapped reads, and normalized cytosine methylation (mC) number in each replicate of four samples for WGBS and filtered reads, aligned reads, and peaks counts of three samples for MeDIP. We sequenced the flanking DNA of *pOsNAR2.1*-*OsNAR2.1* insertion site of in *pOsNAR2.1*-*OsNAR2.1* plants and found that the insertion of the *pOsNAR2.1*-*OsNAR2.1* transgene was between LOC\_Os02g49950 gene and LOC\_Os02g49960 gene in chromosome 2. We also sequenced the flanking DNA of the *pUbi*-*OsNAR2.1* insertion, which was inserted in chromosome 10, between LOC\_Os10g33874 and LOC\_Os10g33900 in *pUbi*-*OsNAR2.1* line (Supplementary Figure S1d,e). Neither of these sites occur in high methylation areas nor in functional genes.


**Table 1.** Samples description in this study.

Our analysis of the sequencing data showed that the total normalized methylated cytosine (mC) numbers differed significantly among the four samples (Figure 1a), with normalized mC counts of approximately 3.9, 3.5, 3.8, and 3.5 million (M) for WYJ7, CK, SDWT, and *pOsNAR2.1-OsNAR2.1*, respectively. The result suggests that the tissue culture (CK) and the transgenic process (*pOsNAR2.1-OsNAR2.1*) of WYJ7 both lead to global DNA hypomethylation. In contrast to CK and *pOsNAR2.1-OsNAR2.1* plants, the mC level in SDWT plants returned to the WT level (WYJ7) after four selfing generations in the field. Since the CK and SDWT both went through regeneration, the result suggested the loss of the transgene in generation increases the methylation in SDWT. Simultaneously, the *pOsNAR2.1-OsNAR2.1* lines showed a significantly higher percentage of symmetric CG sites and lower percentage of asymmetric CHH sites, and CK lines showed a lower percentage of symmetric CHG sites instead (Figure 1b–d), which indicates that while both *pOsNAR2.1-OsNAR2.1* and CK lines show a decrease in global DNA methylation the sites of mC changes were different. Interestingly, in the other two independent transgenic lines, MeDIP sequencing data showed that peak counts in both *pUbi-NAR2.1* and RNAi lines increased compared with NP (Figure 1e). The results suggest that the tissue culture process leads to a decrease in global methylation, but the transgenic process leads to different global methylation changes in different rice backgrounds.

**Figure 1.** Characteristics of sequencing data (**a**) Normalized DNA cytosine methylation numbers in WYJ7, CK, SDWT and *pOsNAR2.1-OsNAR2.1* samples. (**b**-**d**) mCG, mCHG and mCHH ratio in WYJ7, CK, SDWT and *pOsNAR2.1-OsNAR2.1* samples. (**e**) Peak count of methylation in NP, *pUbi-OsNAR2.1* and RNAi. (**f**) Venn diagram of unique and shared genes in CpG methylation states for WYJ7, CK, SDWT and *pOsNAR2.1-OsNAR2.1* samples. (**g**) Venn diagram of peaks in NP, RNAi and *pUbi-OsNAR2.1* samples.

### *2.2. Both Tissue Culture and Transgenic Processes Can Induce Unique Methylation Changes in Genic Regions*

The Venn diagram of WGBS sequencing shows the unique and shared CpG methylation areas of 42,042 genes in the WYJ7, CK, SDWT and *pOsNAR2.1*-*OsNAR2.1* samples (Figure 1f). The results showed that 88.9% of genes in these four samples shared the same CpG methylation area. CK, SDWT, and *pOsNAR2.1*-*OsNAR2.1* had 347, 216, and 381 genes with unique methylation areas, respectively. If only compared with wild-type WYJ7, CK, SDWT, and *pOsNAR2.1*-*OsNAR2.1* showed 1432, 1181, and 1396 genes with unique methylation areas. The Venn diagram analysis of the methylation peaks of the MeDIP-seq data sets for NP, RNAi, and *pUbi-OsNAR2.1* samples showed that the majority of the peaks, approximately 32,000, are shared by the three samples (Figure 1g). Both RNAi and *pUbi-OsNAR2.1*

samples had 10,223 and 11,168 unique peaks, respectively (Figure 1g). The results indicate that both tissue culture and the transgenic processes could induce unique methylation changes in genic regions although neither knockdown nor overexpression of *OsNAR2.1* in transgenic plants caused the majority of DNA methylation changes.
