Arabidopsis thaliana ecotype Col-0 was used as the wild type in this study. The gWUS-GFP3 reporter lines were kindly provided by Thomas Laux (University of Freiburg) (Tucker et al., 2008). The of pARR1:ARR1-GFP reporter lines have been described previously (Meng et al., 2017). The WOX11pro:H2B-eGFPreporter lines were kindly provided by Lin Xu (Chinese Academy of Sciences) (Zhai and Xu, 2021). The drm1 drm2 cmt3-11 triple mutant was kindly provided by Xiaofeng Cao (Chinese Academy of Sciences) (Cao et al., 2003).

Seedlings were grown under sterile condition at 20–22°C, with 16 h of white light and 8 h of dark. Segments containing hypocotyl and root were used as explants, which were firstly germinated in GM medium containing 10 μM auxin transport inhibitors naphthylphthalamic acid (NPA), and then transferred onto the medium containing Gamborg’s B5 medium with 2% glucose, 0.5 g/L MES, 10 μM 1-naphthaleneacetic acid (NAA), and 0.8% agar. After 2 days culture, explants were transferred onto SIM containing Gamborg’s B5 medium with 2% glucose, 0.5 g/L MES, 9 μM 2-isopentenyladenine (2-iP) and 0.8% agar for shoot induction. Explants were cultured under full white light. For calculation of shoot regeneration frequency, regenerated tissues containing a meristem surrounded by three or more leaf primordia with a phyllotactic pattern were considered as a shoot.

Olympus SZX-16 stereoscopic microscope (Olympus) was used to observe explants during regeneration procedures. The expression signals of reporter lines were observed using low melting point agarose embedding section. Confocal microscopy images were taken using a Zeiss LSM 880 NLO confocal microscope with a 20 × lens. Multitracking in line scanmode and a 488/561main dichroic filter were used to image GFP and dsRED together (Heisler et al., 2005). A 561-nm laser line and a 600–640-nm band-pass filter were used for dsRED. A 488-nm laser line and a 505–550-nm band-pass filter were used for GFP. Cell outline was stained with Fluorescent Brightener. A 405-nm laser line and a 425–475-nm band-pass filter was used for observation.

Total RNA was extracted using the TRIzol™ Reagent (catalog no. 15596-026, Invitrogen). The full-length cDNA was generated with the RevertAid First-strand cDNA synthesis kit (Thermo). qRT-PCR was performed on a Chromo4 real-time PCR system (Bio-Rad) using SYBR Master mix (Vazyme) with gene-specific primers (Supplementary Table 1). Transcript levels of the examined genes were normalized to that of the housekeeping gene tubulin2. Values shown are the mean ± standard deviation (SD) of three biological replicates.

DNA was isolated using a cetyltrimethylammonium bromide method. DNA methylation assay was performed using DNA Bisulfite Conversion Kit (Tiangen). PCR products amplified with Methylation specific PCR kit (Tiangen) were cloned into Blunt3 vector (TransGen Biotech) and sequenced. Bisulfite sequencing data were analyzed by the CyMATE software. The results returned by CyMATE were put into GraphPad Prism 9.0 to illustrate DNA methylation frequency at CG, CHG and CHH (where H = A, C or T), respectively. Primers were list in Supplementary Table 1.

In order to study the difference between regulatory mechanisms of direct and indirect shoot regeneration, we first tied to establish a bi-directional regeneration system in which shoots can be generated through the two pathways under the same culture condition. For this purpose, we used segments containing hypocotyl and root as explants, and modified a direct regeneration system reported previously by adjusting the hormone concentrations (Rosspopoff et al., 2017). The results show that when explants were treated with 10 μM NAA for 48 h and then cultured in shoot-inducing medium (SIM) containing 9μM 2-iP, shoots were regenerated from both the hypocotyl and the root (Figure 1). After 2 days incubation on SIM (SIM2), the hypocotyl produced callus while the root gave rise to protuberances. Subsequently, both callus and protuberance grew in size and produced shoot meristems at SIM6. At SIM8, leafy shoots were formed. Therefore, in this system, shoots were generated indirectly from hypocotyls and directly from roots simultaneously (Figure 1).

We next compared the cytological features of these two regeneration procedures by observing their histological structures. Consistent with previous studies, NAA treatment promoted the formation of LRPs (Chatfield et al., 2013; Rosspopoff et al., 2017). After transfer to SIM, the LRP gradually grew into roundish converting organ based on cell divisions at multiple orientations (Figure 2A). Leaf primordia initiated at SIM4 and the structure of shoot meristem was established in the following 1–2 days. In comparison, exogenous NAA induced callus formation in hypocotyls (Figure 2B). After 1 day culture in SIM, the callus grew into a flattened structure. Compared with that of LRP at the same stage, the basal part of callus was much wider, which gave rise to converting organs and shoot meristems with significantly larger size in the subsequent stages.

To investigate the cause of the different meristem size that regenerated from the two regeneration types, we examined the transcriptional levels of genes involved in shoot regeneration. The selected genes encode transcriptional factors regulating auxin/cytokinin signaling (ARF5 and ARR1) or stem cell identity (LBD16, PLT1, SCR, WOX5, WOX11, and WUS). qRT-PCR revealed that transcriptional levels of ARF5, ARR1, LBD16, PLT1, and SCR exhibited similar dynamic patterns between direct and indirect regeneration procedures, suggesting their conserved roles in the two different regeneration pathways (Figure 3). Transcripts of WUS was not detectable during NAA-treatment stage. However, SIM-incubation caused obvious increase of WUS expression, which was more significant in the hypocotyl explants. Transcription of WOX5 and WOX11 was induced by exogenous NAA but decreased during SIM culture. The NAA-mediated alteration of WOX11 and WOX5 was more pronounced in hypocotyl explants than that in root. The results suggest that conversion of cell identity might be differently regulated between direct and indirect regeneration.

To get more insights into the cell fate transition process, we visualized the spatio-temporal expression signals of WOX5, WUS, and WOX11, respectively (Figures 4, 5). The pWOX5:RFP; gWUS-3GFP double reporter lines revealed that WOX5 and WUS were expressed in similar patterns in hypocotyl and root explants (Figure 4). After 48 h NAA-treatment, WOX5 was expressed in the middle cell layers in both callus and LRP. At SIM1, the expression signal of WOX5 vanished while that of WUS was initiated in a few cells. As the callus and LRP grow in size, WUS expression expanded into larger domains. When shoot meristem was formed, WUS expression was confined to the organizing center. The most obvious difference between the two types of explants is that the expression domain of WOX5 at the end of NAA-treatment in callus was larger than that in LRP.

In the pWOX11:H2B-eGFP lines, GFP signals were first detected in pericycle cells at 6 h of NAA-treatment (Figure 5). Different from that of LRP, where WOX11 was induced in about 8 cells before periclinal division at 12 h, expression signals were visible in more than 15 continuous cells in the initiating callus. During the primary cell divisions, the signals were also observed in newly proliferated cells. When organized cell files were established, WOX11 was expressed in the founder cells at basal part of the callus and LRP. Therefore, number of founder cells in the incipient stage was tightly associated with the regeneration types and the size of regenerated meristems. Relatively more WOX11-expressing founder cells gave rise to callus which produce larger meristem, whereas less founder cells led to the formation of LRP that regenerate smaller meristem.

We next intend to explore the factors regulate WOX11 expression. Previous studies showed that non-CG methylation is involved in acquisition of pluripotency (Shemer et al., 2015). It has been shown that non-CG DNA methylation is almost completely lost in the triple mutant of DOMAINS REARRANGED METHYLTRANSFERASE1/2 CHROMOMETHYLASE3 (drm1 drm2 cmt3) (Cokus et al., 2008; Stroud et al., 2014). We thus examined the shoot regeneration capacity of the drm1 drm2 cmt3 (ddc) triple mutant using the bi-directional system described above. As a result, both the frequency and the number of regenerated shoots per explant were significantly increased in root of the ddc triple mutant compared with those of wild type (Figure 6). However, the regeneration ability of hypocotyl did not show obvious changes between the mutant and the wild type. The results indicate that non-CG DNA methylation negatively regulates direct shoot regeneration but did not affect indirect regeneration. No obvious phenotype was observed in drm1 and cmt3 single mutants, as well as drm1 drm2 double mutant, suggesting the functional redundancy among DRM1, DRM2 and CMT3.

To determine whether the expression of WOX11 is mediated by non-CG DNA methylation, bisulfite sequencing was performed to compare DNA methylation between the ddc triple mutant and the wild type. The results illustrate that after NAA-treatment for 1 day, 13 sites of the genomic fragments 1,036–1,529 bp and 2,635–3,044 bp upstream of the coding sequence were hypermethylated in wild type. However, the level of methylation in the same sites were substantially decreased in the ddc mutant (Figure 7). Correspondingly, compared with that of wild type, transcriptional level of WOX11 was significantly higher in the ddc root during NAA-treatment (Figure 8A). On the contrary, in the hypocotyl explants at the same stages, both the methylation and the expression of WOX11 did not show obvious difference between ddc and wild type (Figure 8B and Supplementary Figure 1). The results demonstrate that non-CG DNA methylation negatively regulate WOX11 transcription during LRP formation but had no influence on callus.

De novo shoot regeneration comprises three steps, including the activation of initial cells, acquisition of regenerative competency and establishment of shoot meristem (Sang et al., 2018b). WOX11 controls the former two steps by promoting the first cell fate transition and activating WOX5 expression (Liu et al., 2014; Hu and Xu, 2016; Zhai and Xu, 2021). Subsequently, genes responsible for shoot meristem maintenance such as WUS regulate the third step. Therefore, if non-CG methylation regulate shoot regeneration via modulating WOX11 expression, the ddc triple mutant would produce more LRP. To test this hypothesis, we examined lateral root number. The results show that the ddc triple mutant give rise to significantly more lateral roots than that of wild type, indicating an increase in LRP formation (Figure 8C).

Previous study showed that after transfer to hormone-free medium, auxin-induced callus which resembles LRP can be converted to roots (Atta et al., 2009). To analyze the difference of NAA-induced callus/LRP formation between the ddc triple mutant and wild type, we transferred explants after 48 h NAA-treatment to hormone-free medium. As a result, the ddc triple mutant generated significantly more lateral roots than wild type (Figure 8D). However, the number of adventitious roots derived from hypocotyls did not demonstrate obvious changes, suggesting that callus formation capacity was similar between the ddc triple mutant and wild type (Figure 8E). These results suggest that non-CG methylation is implicated in shoot regeneration through mediating WOX11 expression.