South China 124 variety of Manihot esculenta was used. SC124 cassava and tobacco plants were grown in soil with Hoagland’s solution, at 26–28°C, with 12 h light at 120–150 μmol quanta m^-2 s^-1 irradiance and 12 h dark cycles.

Total RNA extraction and cDNA synthesis were performed from plant leaves using RNAprep Pure Plant Kit (TIANGEN, DP441, Beijing, China) and RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific, K1622, Waltham, MA, United States), according to the manufacturer’s instruction. The quantitative real-time PCR was performed using cDNA and FastStart Essential DNA Green Master (Roche, 06924204001, Basel, Switzerland) and analyzed using the comparative ΔΔ^C_T method as Wei et al. (2016) described. NtEF1a and MeEF1a were used as internal references for analysis. The primers used for real-time PCR were listed in Supplementary Table S1.

For the vector construction, the coding regions of MebZIP3 and MebZIP5 were first amplified by PCR from plant leave samples. Thereafter, the PCR products were cloned into SpeI and NcoI/SpeI digested modified pCAMBIA1302 (Liu et al., 2015) by restriction enzyme digestion and T₄ ligase ligation, respectively. The primers responsible for vector constructs were listed in Supplementary Table S2, and the restriction enzymes and their cutting sites were marked. The vector cassettes were illustrated in Supplementary Figure S1. After DNA sequencing for confirmation, the recombinant plasmids as well as P19 were transformed into Agrobacterium tumefaciens strain GV3101. After syringe infiltrating into Nicotiana benthamiana leaves as Sparkes et al. (2006) described for 2 dpi, the green fluorescent and DAPI-stained cell nuclei in the infiltrated leaf areas were examined using a confocal laser-scanning microscope (TCS SP8, Leica, Heidelberg, Germany).

The transgenic MebZIP3 and MebZIP5 tobacco plants were generated through Agrobacterium-mediated transformation of MebZIP3-pCAMBIA1302 and MebZIP5-pCAMBIA1302 as Horsch et al. (1985) described. Briefly, 14-day-old sterilized tobacco leaves were incubated in Agrobacterium cell suspension for 10 min. Subsequently, the treated leaves were dried with sterilized tissue paper and placed on full MS-Agar medium for co-cultivation. After 2 days, the leaves were transferred to shoot initiation medium with cephalosporin (250 mg L^-1) and hygromycin (50 mg L^-1) and the surviving seedlings were grown in a greenhouse to produce seeds for further analysis. The T₂ transgenic seedlings were selected on MS medium with 50 mg L^-1 hygromycin, and the green seedlings with long roots were transferred to soil for further semi-quantitative reverse transcriptase-PCR and seed harvest. The transgenic T₃ seedlings were further selected on MS medium with 50 mg L^-1 hygromycin to obtain homozygous lines with no segregation on hygromycin resistance, and the independent transgenic T₃ lines were used for phenotype analysis.

For the vector construction, the partial coding regions of MebZIP3 and MebZIP5 were first amplified by PCR from plat leave samples. Thereafter, the PCR products of these genes were cloned into EcoRI/BamHI digested pTRV2 vector (Liu et al., 2002) by restriction enzyme digestion and T₄ ligase ligation. The primers that are responsible for vector constructs were listed in Supplementary Table S2, the restriction enzymes and their cutting sites were marked. The vector cassettes were illustrated in Supplementary Figure S1. After DNA sequencing for confirmation, the recombinant plasmid (pTRV2-MebZIP3 and pTRV2-MebZIP5) as well as pTRV1 were transformed into Agrobacterium tumefaciens strain GV3101. The GV3101 strains were first cultured in 10 ml of LB liquid medium at 28°C for 12 h, and shaken in the new LB liquid culture to reached OD₆₀₀ at about 2. After diluted to OD₆₀₀ of 1 by 10 mM MgCl₂, 10 mM MES, and 20 mM acetosyringone, the GV3010 strain with pTRV1 and the strain with pTRV2 or MebZIP3-pTRV2 or MebZIP5-pTRV2 were mixed with ratio of 1:1 and co-infiltrated into cassava leaves as Wei et al. (2017) described. After 14 days, the corresponding gene expression assay and disease resistance assay were performed in plant leaves.

The bacterial pathogen of Xam was first cultured in 10 ml of LB liquid medium at 28°C for 12 h, and shaken in the new LB liquid culture to reach OD₆₀₀ at about 0.6. After diluted to 10⁸ cfu ml^-1 by 10 mM MgCl₂ and 0.05% silwet L-77, the Xam was syringe infiltrated into abaxial side of plant leaves. Then the plants with pathogen infection were grown in the green house. At indicated time-points, at least 20 leaves were harvested in every biological repeat. Plant leaves were gently washed by sterile distilled water for 1 min, then the bacterial populations in plant leaves were quantified using 10 μl five 10-fold dilutions of homogenate in LB medium.

Callose deposition in plant leaves was visualized by callose staining, using alcoholic lactophenol solution, 0.01% (w/v) aniline blue solution, 50% (v/v) glycerol and fluorescence microscope (DM6000B, Leica, Heidelberg, Germany) as Hauck et al. (2003) described.

The endogenous levels of H₂O₂ in plant leaves were extracted and determined using the peroxide-titanium buffer as Wei et al. (2016) previously described.

For the vector construction, the coding regions of MebZIP3 and MebZIP5 were first amplified by PCR from plant leave samples. Thereafter, the PCR products of these were cloned into NdeI/BamHI and NcoI/BamHI digested pGBKT7 by restriction enzyme digestion and T₄ ligase ligation, respectively. The primers that are responsible for vector constructs were listed in Supplementary Table S2, the restriction enzymes and their cutting sites were marked. After DNA sequencing for confirmation, the recombinant plasmids were transformed into yeast strain AH109, according to the manufacturer’s protocol (Clontech, United States). The transformed clones were screened on the SD/-Trp and SD/-His mediums, respectively. The transcriptional activation was evidenced by the growth of yeast cells on SD/-His medium with 5 mM X-α-gal at 30°C for 3 days.

The accession numbers and CDS length of all genes are shown as following: MebZIP3 (KU160294, 1,788 bp), MebZIP5 (KU160296, 1,488 bp), MePR1 (Me07G050300, 492 bp), MePR2 (Me10G089800, 492 bp), MePR3 (Me07G050700, 486 bp), MePR4 (Me07G050400, 492 bp), MeEF1a (AF041463, 1,035 bp), NtEF1a (AY206004, 661 bp).

All results in this study were obtained from at least three biological repeats, and the average values and SDs of these biological repeats were shown. In the meanwhile, asterisk symbols (^∗) indicting the significant differences at p < 0.05 were also shown after ANOVA analysis.

In the previous study (Hu et al., 2016), 77 MebZIPs have been identified in Manihot esculenta Phytozome database v10.3¹. Herein, a phylogenetic tree between MebIP3/MebZIP5 and their homologs from other plant species were constructed (Supplementary Figure S2A), and the results implied the functional similarities among the bZIP proteins in different plants. Moreover, the conserved bZIP domain of MebZIP3 and MebZIP5 was identified (Supplementary Figure S2B), further indicating that bZIPs are conserved during evolution.

Using quantitative real-time PCR, we found that the transcript levels of MebZIP3 and MebZIP5 were significantly regulated after flg22, Xam, SA and H₂O₂ treatments (Figures 1A,B). After flg22 treatment, the transcript levels of MebZIP3 and MebZIP5 were down-regulated at 3 h, but largely up-regulated at 6 h. After Xam treatment, MebZIP3 and MebZIP5 transcripts were significantly induced at 6 h. MebZIP3 transcript was largely increased after SA treatment for 3 and 6 h, while MebZIP5 expression was decreased after SA treatment for 1 h. Moreover, MebZIP3 and MebZIP5 transcripts were largely induced after H₂O₂ treatment for 3 and 6 h (Figures 1A,B). Generally, the transcripts of MebZIP3 and MebZIP5 displayed common expression patterns in response to these treatments, indicating the possible involvement of them in plant disease response. Moreover, we found that MebZIP3 and MebZIP5 were expressed in all assayed organs, with higher transcript levels in cassava stem and storage root than in leaf (Figure 1C).

To investigate the subcellular location of MebZIP3 and MebZIP5, the coding regions of these genes were fused with GFP and transiently expressed in Nicotiana benthamiana leaves. The control vector (35S::GFP)-transformed leaves displayed GFP in both cell nuclei and membrane, consistent with many previous studies (Wei et al., 2017). The GFP signals of MebZIP3-GFP and MebZIP5-GFP were co-localized with DAPI-stained cell nuclei in the infiltrated leaf areas, as marked by the arrow, suggesting that MebZIP3 and MebZIP5 are specifically located in cell nucleus (Figure 2).

Since bZIPs belongs to transcription factor family, transcription activation assays of MebZIP3 and MebZIP5 were performed in yeast cells. The coding regions of MebZIP3 and MebZIP5 were fused to the GAL4 DNA binding domain in pGBKT7, and the constructs were transformed into yeast strain AH109. As evidenced by the growth of yeast cells and LacZ staining on SD/-His medium with 5 mM X-α-gal, the yeast cells transformed with MebZIP3-pGBKT7 and MebZIP5-pGBKT7 had transcriptional activity (Supplementary Figure S3), suggesting the transcriptional activities of MebZIP3 and MebZIP5 in yeast cells.

To further reveal the in vivo roles of MebZIP3 and MebZIP5, the transgenic plants overexpressing MebZIP3 or MebZIP5 were generated in tobacco. After selection on MS medium with hygromycin, the resistant T₁ transgenic seedlings were transferred to soil, and the gene expressions in the overexpressing lines were confirmed by semi-quantitative reverse transcriptase-PCR (Figures 3A,B and Supplementary Figure S4). The corresponding MebZIP3 or MebZIP5 could be examined in the transgenic MebZIP3 or MebZIP5 overexpressing tobacco lines, but could not be amplified in the WT tobacco leaves (Figures 3A,B and Supplementary Figure S4). The transgenic T₂ and T₃ seedlings were further selected on MS medium with 50 mg L^-1 hygromycin to obtain homozygous lines with no segregation on hygromycin resistance. Because no PCR product was detected in the WT sample by semi-quantitative reverse transcriptase-PCR (Figures 3A,B), quantitative real-time PCR was performed to show the relative transcript level in different transgenic T₃ lines (Figures 3C,D). Based on the gene transcript level, three independent transgenic T₃ lines were used for the phenotype analysis of MebZIP3 (OE1, OE2, and OE3) and MebZIP5 (OE2, OE5, and OE7).

Although Nicotiana benthamiana is non-host of Xam, its leaves can be infected by Xam with disease symptom and pathogen growth. To investigate the function of MebZIP3 and MebZIP5 in plant disease resistance, the leaf surfaces of WT, MebZIP3, and MebZIP5 transgenic lines were infected with 10⁸ cfu ml^-1 of Xam. At 2, 4, and 6 dpi, three MebZIP3 (OE1, OE2, and OE3) and three MebZIP5 (OE2, OE5, and OE7) overexpressing lines exhibited significant less bacterial number in the leaves in comparison to that of WT (Figures 4A–C). Moreover, when Xam was infected, the H₂O₂ and callose depositions were substantially higher in the overexpressing plant leaves than those in WT (Figures 4D,E). These results suggested that overexpression of MebZIP3 and MebZIP5 conferred improved disease resistance.

To further confirm the in vivo roles of MebZIP3 and MebZIP5 in cassava defense resistance, we construct the MebZIP3- and MebZIP5-silenced plants through VIGS. As evidenced by the lower transcript of MebZIP3 or MebZIP5, the VIGS plants (pTRV-MebZIP3 and pTRV-MebZIP5) were successfully acquired (Figure 5A). In comparison to mock plants, the VIGS plant (pTRV-MebZIP3 and pTRV-MebZIP5) leaves showed more bacterial number (Figure 5B), lower transcripts of defense-related genes (PR1, PR2, PR3, and PR4) (Figure 6), less callose depositions and lower levels of H₂O₂ in plant leaves upon Xam infection (Figures 7A,B). Thus, MebZIP3 and MebZIP5 are essential for plant disease resistance in cassava.