eIF2Bβ confers resistance to Turnip mosaic virus by recruiting ALKBH9B to modify viral RNA methylation

Summary

Eukaryotic translation initiation factors (eIFs) are the primary targets for overcoming RNA virus resistance in plants. In a previous study, we mapped a BjeIF2Bβ from Brassica juncea representing a new class of plant virus resistance genes associated with resistance to Turnip mosaic virus (TuMV). However, the mechanism underlying eIF2Bβ‐mediated virus resistance remains unclear. In this study, we discovered that the natural variation of BjeIF2Bβ in the allopolyploid B. juncea was inherited from one of its ancestors, B. rapa. By editing of eIF2Bβ, we were able to confer resistance to TuMV in B. juncea and in its sister species of B. napus. Additionally, we identified an N⁶‐methyladenosine (m⁶A) demethylation factor, BjALKBH9B, for interaction with BjeIF2Bβ, where BjALKBH9B co‐localized with both BjeIF2Bβ and TuMV. Furthermore, BjeIF2Bβ recruits BjALKBH9B to modify the m⁶A status of TuMV viral coat protein RNA, which lacks the ALKB homologue in its genomic RNA, thereby affecting viral infection. Our findings have applications for improving virus resistance in the Brassicaceae family through natural variation or genome editing of the eIF2Bβ. Moreover, we uncovered a non‐canonical translational control of viral mRNA in the host plant.

Introduction

Viruses represent a significant global threat to agriculture, causing a 10–15% reduction in agricultural production worldwide (Mahy and van Regenmortel, 2009). Turnip mosaic virus (TuMV) belongs to the genus Potyvirus and has a very wide host range of plant species, including many species of Brassicaceae, legumes, ornamentals and weeds (Edwardson and Christie, 1991). The most effective strategy for combating epidemics and yield losses caused by viruses is to develop crop varieties with durable and broad‐spectrum resistance and high yield (Dangl et al., 2013). The Brassicaceae family is of global economic importance, comprising some 3600 species grouped in more than 300 genera (Bailey et al., 2006), and the well‐studied model plant Arabidopsis. They are highly susceptible to TuMV disease (Li et al., 2019; Palukaitis and Kim, 2021; Walsh and Jenner, 2002).

Viruses hijack plant host factors for successful infection. Consequently, mutations in the genes encoding these host factors can confer genetic resistance, a mechanism known as resistance by loss of susceptibility (Pavan et al., 2010; van Schie and Takken, 2014). Such resistance genes have been isolated by exploring natural variation at different loci, with a particular focus on eukaryotic translation initiation factors (eIFs) (Robaglia and Caranta, 2006; Shopan et al., 2020). Viruses interact with eIFs for their translation in the host (Truniger and Aranda, 2009), so loss‐of‐function in eIFs is expected to confer broad‐spectrum resistance representing a recessive resistance. eIF4E has emerged as a major susceptibility factor for RNA viruses (Bastet et al., 2017). Naturally occurring alleles of the host translation initiation factor, eif4e or eif(iso)4e, have been employed to confer resistance to several agronomically important species of single‐stranded positive sense RNA viruses, including those belonging to the Potyviridae family (Bastet et al., 2017; Qian et al., 2013; Wang and Krishnaswamy, 2012). The use of ethyl methanesulfonate (EMS) mutagenesis and Targeted Induced Local Lesions in Genomes (TILLING) to isolate mutations within the eIF4E1 could be an interesting alternative method to generate resistant alleles in tomato (Piron et al., 2010). In addition, clustered regularly interspaced palindromic repeats/CRISPR‐associated protein 9 (CRISPR/Cas9) gene editing techniques allow the generation of resistant eIFs alleles that are resistant to viruses and can compensate for the lack of natural resistance alleles in some crops (Bastet et al., 2019; Mahas and Mahfouz, 2018). Using this approach, mutations of the eIF4E or its alleles have been used to generate stable resistance to viruses in Arabidopsis (Pyott et al., 2016), cucumber (Chandrasekaran et al., 2016), potato (Zhan et al., 2019), tomato (Yoon et al., 2020) and other plants as well (Mahas and Mahfouz, 2018).

An interaction between eIF4E and VPg (viral protein genome linked, a 5′‐terminal protein covalently linked to the genome) of several Potyviridae is critically associated with the success of viral genome translation (Eskelin et al., 2011; Miras et al., 2017) and viral cell‐to‐cell movement (Contreras‐Paredes et al., 2013; Gao et al., 2004), although the exact role of eIF4E in Potyviridae infection is still unclear (Bastet et al., 2017; Revers and Garcia, 2015; Shopan et al., 2020). A common feature of the knockout approaches for specific eIF4E genes is that the resulting resistance is often overcome by viruses interacting with other host factors (Bastet et al., 2017). Therefore, the search for novel targets and natural resistance to potyviruses for potyvirus control has been widely recognized. Although considerable progress has been made in the last decade in understanding the molecular biology of these viruses and the functions of their various proteins (Bastet et al., 2017; Revers and Garcia, 2015), the molecular mechanism of the eIFs‐dependent translation underlying viral resistance remains to be elucidated.

We have previously mapped a gene annotated as eukaryotic translation initiation factor 2b beta (BjeIF2Bβ), as a new class of plant virus resistance genes, which was used to develop resistance to TuMV in Brassica juncea (Shopan et al., 2017). In this study, we investigated the phylogenetic origin of the SNP and Indel variations of eIF2Bβ and confirmed that eIF2Bβ confers the resistance to TuMV in Brassicaceae. We also found that BjeIF2Bβ mediates resistance to TuMV by interacting with BjALKBH9B and modifying the methylation of the viral coat protein RNA.

Results

Natural variation of eIF2Bβ is associated with TuMV resistance in Brassicaceae

Given that the TuMV‐associated resistance gene, eIF2Bβ, has been mapped to the A subgenome of B. juncea, we selected one of its ancestors, B. rapa, and the model plant Arabidopsis, to study their natural variation and trace the source of resistance. Microsynteny pattern analysis revealed that eIF2Bβ has one copy in Arabidopsis thaliana, and two copies in the B. rapa and B. juncea A subgenomes. It is noteworthy that the resistance copy was located on the A01 chromosome in both B. rapa and B. juncea (Figure S1). Through association analysis of BjeIF2Bβ genotypes and phenotypes after TuMV inoculation using selected accessions from the B. juncea resequencing panel, we identified two SNPs (SNP_1: C/G and SNP_2: G/A) and one Indel (90 bp) associated with TuMV resistance resulting in 7 haplotypes (Figure 1a). Among these variations, SNP_1 was identified as a significant variant associated with TuMV resistance in B. juncea. Analogously, we investigated polymorphisms of eIF2Bβ in B. rapa resequencing panel, which possessed the same two SNPs, and observed that the types of Indel were more complicated than those observed in B. juncea. A summary of four Indel types of BreIF2Bβ was provided, with type_1 and type_2 being identical to the Indel observed in susceptible and resistant lines of B. juncea, respectively (Figure S2). By sequence alignment, it was found that the Indel type_3 (106 bp) and type_4 (105 bp) only shares rare parts that are similar to the Indel in B. juncea (Figure S3). The phenotypes and TuMV isolate accumulation of all cultivars were found to correspond to the eIF2Bβ haplotypes, thereby confirming the correlation of SNPs of eIF2Bβ with resistance to TuMV (Figure 1b and Figure S4). The results demonstrated the association between eIF2Bβ haplotypes and phenotypes following TuMV inoculation in the A. thaliana Col‐0, as well as in the selected accessions of B. rapa and B. jucnea. The evidence indicated that the natural SNPs of eIF2Bβ are phylogenetically derived from its ancestor, B. rapa. Furthermore, no such SNPs and Indel variations were identified in the A. thaliana eIF2Bβ based on the Arabidopsis 1001 Genomes (http://1001genomes.org/).

Figure 1.

Natural variation of eIF2Bβ is associated with resistance to TuMV in Brassicaceae. (a) Haplotypes of BjeIF2Bβ using SNPs and Indel variation with resistant and susceptible phenotypes in B. juncea. (b) Associative analysis of phenotypes and haplotypes of eIF2Bβ following inoculation with TuMV isolate in selected accessions of B. juncea, B. rapa and A. thaliana. (c) Phylogenetic tree of the A subgenome and haplotypes of eIF2Bβ with SNPs using resequencing panels of B. rapa and B. juncea. The presence or absence of the indel is indicated by the symbols ‘+’ and ‘−’, respectively. The letters ‘R' and ‘S’ indicate the resistance and susceptibility, respectively.

In order to investigate the phylogenetic relationship and eIF2Bβ haplotypes of B. rapa and B. juncea, a phylogenetic tree of the A subgenome was constructed using resequencing panels of 198 B. rapa accessions and 208 B. juncea accessions. The eIF2Bβ haplotypes of all accessions were displayed in the vicinity of the phylogenetic tree (Figure 1c, Tables S1 and S2). The light green and light red ring sectors represented B. rapa and B. juncea species, respectively. The results demonstrated that the eIF2Bβ exhibits greater polymorphic variation in B. rapa than that in B. juncea. This observation may be due to the fact that B. rapa is one of the progenitors of B. juncea, having originated earlier and undergone selection and domestication for a longer period, thus accumulating more variation. In the meantime, we discovered that the earlier diverged subgroup of B. juncea, such as root type and oilseed type (Oilseed_ACE), possessed a greater number of resistant genotypes than other subgroups of B. juncea. This indicated that earlier diverged varieties retain a diverse array of resistant genotypes that could be employed in TuMV resistance breeding. These results indicated that the eIF2Bβ is a promising genetic locus and could be widely used for TuMV resistance breeding in Brassicaceae.

Editing of eIF2Bβ confers resistance to TuMV in B. juncea and B. napus

In the Brassica genus crops, the allopolyploid B. juncea and its sister species, B. napus, share a common A subgenome derived from an independent variety of the ancestral B. rapa. A comparison of the eIF2Bβ gene sequences of B. juncea var. XLH and B. napus var. Westar revealed that they are highly conserved in terms of the amino acid sequence (Figure S5). Both two accessions possessed the SNP that predicts susceptibility to TuMV. In order to ascertain the functionality of eIF2Bβ, we employed the CRISPR‐Cas9 technique to generate eIF2Bβ edited lines in B. juncea and B. napus. Following the validation of the mutations on the eIF2Bβ, three and two independent homozygous eIF2Bβ‐edited lines of B. juncea and B. napus, respectively, were generated in the T2 generation (Figure 2a,e and Figures S6, S7). In B. juncea, the BjeIF2Bβ‐edited lines (bjeif2bβs) exhibited resistance to TuMV compared with the wild type (WT), as evidenced by the observed phenotypes and the accumulation of TuMV isolates. The systemically infected leaves of WT exhibited typical mosaic and shrunken phenotypes, whereas the bjeif2bβs displayed no obvious symptoms (Figure 2b). The transcriptional and translational expression of TuMV‐CP in systemically infected leaves of WT and bjeif2bβs was confirmed by qPCR and Western blotting. These results indicated that almost no TuMV isolate accumulated in bjeif2bβs (Figure 2c,d).

Figure 2.

Genome editing of eIF2Bβ confers resistance to TuMV in B. juncea and B. napus. (a) gRNA target site and the resulting missense mutation in the BjeIF2Bβ gene in B. juncea. (b) Symptoms of leaves after 2 weeks of TuMV inoculation in WT and BjeIF2Bβ‐edited lines of B. juncea. (c) Transcriptional expression of TuMV‐CP after 2 weeks of TuMV inoculation determined by RT‐qPCR in WT and BjeIF2Bβ‐edited lines of B. juncea. (d) Translational expression of TuMV‐CP after 2 weeks of TuMV inoculation determined by Western blotting in WT and BjeIF2Bβ‐edited lines of B. juncea. (e) gRNA target site and the resulting missense mutation in the BneIF2Bβ gene in B. napus. (f) Symptoms of leaves after 2 weeks of TuMV inoculation in WT and BneIF2Bβ‐edited lines of B. napus. (g) Transcriptional expression of TuMV‐CP after 2 weeks of TuMV inoculation determined by RT‐qPCR in WT and BneIF2Bβ‐edited lines of B. napus. (h)c Translational expression of TuMV‐CP after 2 weeks of TuMV inoculation determined by Western blotting in WT and BneIF2Bβ‐edited lines of B. napus. Values are means ± SD in qPCR results. Actin protein was employed as a loading control in Western blotting.

In B. napus, no obvious mosaic symptoms were observed in either the WT or BneIF2Bβ‐edited lines (bneif2bβs) following TuMV inoculation (Figure 2f). We attempted to extend the observation period after inoculation to 30 days and reinoculated the leaves in case TuMV was absent. Necrosis was observed in inoculated leaves, and no systemic infection symptoms were apparent. Previous research indicated that the symptom observed in B. napus was seemingly dependent on the interaction of different resistance genes with distinct TuMV isolates (Jenner and Walsh, 1996). The operation of friction inoculation also produced necrosis in leaves, which made it challenging to assess the resistance based on the observed symptoms. Subsequently, composite samples were collected from inoculated and uninoculated leaves on each plant to detect virus accumulation at the molecular level. The results of RT‐qPCR and Western blotting demonstrated that the accumulation of virus in bneif2bβs was significantly reduced in comparison to WT at both the mRNA and protein levels (Figure 2g,h). Although the accession Westar possessed some resistance to TuMV_ZJ, the results indicated that the mutants exhibited increased resistance to TuMV. These results confirmed that editing of eIF2Bβ confers resistance to TuMV in B. juncea and B. napus, and suggest that the eIF2Bβ is a compelling target for potyvirus control using genome editing techniques in Brassicaceae.

Subcellular localization of eIF2Bβ

In order to gain further insight into the potential functions of BjeIF2Bβ, we transiently expressed BjeIF2Bβ‐GFP to visualize its subcellular localization in A. thaliana protoplasts. The fluorescence signals were observed to be distributed in several cell organelles. Consequently, a number of genes localized to different A. thaliana cell organelles were selected as co‐localization markers by fusing them to the mCherry protein. The fluorescence signals indicated that BjeIF2Bβ was distributed throughout the cytosol, plasma membrane, nucleus and endoplasmic reticulum (Figure 3). In addition to cytoplasmic and nuclear localization, our results demonstrated that BjeIF2Bβ accumulation was also detected in the membrane system. In particular, the endoplasmic reticulum localization may be associated with TuMV vesicles, thereby influencing plant resistance to the virus. These results indicated that BjeIF2Bβ exhibits a diverse subcellular localization, suggesting that it may perform multiple biological functions beyond eukaryotic translation initiation.

Figure 3.

Subcellular localization of BjeIF2Bβ. The green fluorescence signals indicate fused protein of BjeIF2Bβ and green fluorescent protein (GFP). The red fluorescence signals represent the marker genes, which are located in different organelles. The pink fluorescent signals indicate chloroplast autofluorescence. OASA1, cytosol localized marker; PIP2;1, plasma membrane localized marker; HY5, nucleus localized marker and PDI6, endoplasmic reticulum localized marker. Bars, 25 μm.

Interaction between eIF2Bβ and ALKBH9B

To investigate the potential molecular mechanism by which eIF2Bβ mediates TuMV resistance. A cDNA library from B. juncea var. XLH was employed to screen for candidate proteins that interact with BjeIF2Bβ via a yeast two‐hybrid (Y2H) library screening. A protein with a conserved domain 2OG‐Fell_Oxy_2 was annotated as Arabidopsis ALKBH9B, which was named BjALKBH9B after (Figure 4a). Subsequently, a Y2H assay utilizing full‐length BjALKBH9B demonstrated that BjeIF2Bβ can interact with BjALKBH9B in yeast cells (Figure 4b). This interaction was further verified by a bimolecular fluorescence complementation assay (BiFC), in which P2YN‐BjALKBH9B and P2YC‐BjeIF2Bβ were co‐expressed in a N. benthamiana leaf cell (Figure 4c). In addition to cytoplasmic fluorescence, we also observed some small cytoplasmic granule fluorescence signals in the BiFC assay. Furthermore, a co‐immunoprecipitation (co‐IP) assay was employed to confirm this interaction in vivo (Figure 4d). Another protein from the 2OGD gene family, BjuVB05G39250, which possesses a 2OG‐FeII_Oxy_3 domain, was selected as a negative control (Figure 4a). BjuVB05G39250 was unable to interact with BjeIF2Bβ in all the three assays, thereby demonstrating the unique interaction between BjeIF2Bβ and BjALKBH9B. These results indicated that BjeIF2Bβ can specifically interact with BjALKBH9B. Given that AtALKBH9B has been demonstrated to regulate virus resistance (Martinez‐Perez et al., 2017), it can be postulated that BjeIF2Bβ may mediate TuMV resistance through its interaction with BjALKBH9B.

Figure 4.

Interaction of BjeIF2Bβ with BjALKBH9B. (a) Comparison of the conserved domain of the 2OG‐FeII_Oxy superfamily between BjALKBH9B, AtALKBH9B and another BjALKBH9B homologue (BjuVB05G39250). (b) Interaction between BjeIF2Bβ and BjALKBH9B through Y2H. (c) Interaction between BjeIF2Bβ and BjALKBH9B by BiFC. (d) Interaction between BjeIF2Bβ and BjALKBH9B by Co‐IP. BjALKBH9B homologue (BjuVB05G39250) was employed as a negative control in the interaction assay. Bars, 25 μm.

BjeIF2Bβ recruits BjALKBH9B to modify methylation of CP RNAs

The results presented above indicate a need for further investigation into the specific molecular mechanism by which BjeIF2Bβ and BjALKBH9B co‐regulate TuMV resistance. First, a co‐localization assay was performed among BjeIF2Bβ, BjALKBH9B and 6K‐NIa of TuMV, representing the cytoplasmic vesicles derived from the endoplasmic reticulum. The fluorescence signals demonstrated that both BjeIF2Bβ and BjALKBH9B were present in vesicles, which provided a shelter for virus replication and protein translation (Figure 5a). This result provided an opportunity to approach BjeIF2Bβ and BjALKBH9B with TuMV mRNA or protein at the cellular spatial level.

Figure 5.

eIF2Bβ recruits ALKBH9B to modify the methylation of viral CP RNAs. (a) Co‐localization of BjeIF2Bβ, 6K‐NIa of TuMV and BjALKBH9B. (b) Identification of the m⁶A methylation peak by RIP‐seq in the TuMV genome after TuMV inoculation in WT and BjeIF2Bβ‐edited lines of B. juncea. (c) BjeIF2Bβ and BjALKBH9B target CP mRNA by RIP‐qPCR. Values are means ± SD (two‐sided t test, n = 3). (d) Translation efficiency of TuMV‐CP in WT and BjeIF2Bβ‐edited lines of B. juncea. (e) Translation efficiency of TuMV‐CP in WT and lines of overexpressed BjALKBH9B of B. juncea. (f) Binding affinity of BjALKBH9B and BjeIF2Bβ dimer complex with TuMV‐CP RNA detected by MST assays. BjALKBH9B and GUS complex mixed with CP RNA were employed as a negative control. Each binding experiment was conducted three times independently. (g) A proposed working model in which eIF2Bβ recruits ALKBH9B to modify the methylation of viral RNAs to influence TuMV epidemic.

Subsequently, considering the potential demethylation function of BjALKBH9B, a meRIP‐seq analysis was performed using TuMV‐infected WT and BjeIF2Bβ‐edited lines (bjeif2bβs) of B. juncea to determine the level of m⁶A modification in TuMV mRNA. Following the alignment of sequencing reads to the TuMV reference genome, it became evident that pronounced peaks were present in the RNA region encoded by CP (Figure 5b). Furthermore, we observed that the m⁶A modification status of CP mRNA in bjeif2bβs was increased compared to that of WT (Figure 5b). In particular, another m⁶A peak was observed in bjeif2bβs mRNA, distinct from the first observed in CP mRNA (Figure 5b). The results confirmed that the absence of BjeIF2Bβ is associated with an increased m⁶A modification status in CP mRNA. RIP‐qPCR assays were then conducted utilizing B. juncea protoplasts overexpressing BjeIF2Bβ‐FLAG or BjALKBH9B‐FLAG in conjunction with TuMV virions (Figure S8c). Two pairs of qPCR primers were designed to amplify fragment 1 (F1) and fragment 2 (F2), which completely covered the m⁶A peak region of CP mRNA (Figure S8a). The qPCR results of FLAG‐IP demonstrated that both BjeIF2Bβ and BjALKBH9B bind to the m⁶A peak region of CP mRNA in vivo, in comparison to the negative control IgG (Figure 5c and Figure S8b). To demonstrate that BjeIF2Bβ is a critical factor in the function of BjALKBH9B, we designed a series of microscale thermophoresis (MST) assays. The results proved that when BjALKBH9B‐GFP or BjeIF2Bβ‐GFP was mixed with CP RNA for MST experiments, the binding affinity between them was weak (Figure S9a–c,e). However, when BjALKBH9B‐GFP and BjeIF2Bβ were mixed first and CP RNA was added afterwards, the binding affinity of this sample became evident. The Kd value was determined to be 4.05 μM, and the signal‐to‐noise ratio was found to be 13.1. When BjeIF2Bβ was replaced by GUS, no binding was detected (Figure 5f and Figure S9d,f). These results indicated that BjeIF2Bβ is an important enabler in the process of BjALKBH9B binding to CP RNA. Furthermore, no ALKB, 2OG‐FeII_Oxy or related domains were identified in the TuMV_ZJ encoding proteins following a domain scanning using the Pfam database (Figure S10), which indicated that TuMV_ZJ isolate does not possess the capacity to remove its own m⁶A methylation. These results demonstrated that BjeIF2Bβ can promote the binding of BjuALKBH9B to CP mRNA, which ultimately contributes to the observed variation of m⁶A methylation level of TuMV.

Finally, in consideration of the possibility that BjeIF2Bβ may regulate translation initiation, we conducted experiments to determine whether BjeIF2Bβ or BjALKBH9B would alter viral translation efficiency. The accumulation of TuMV protein at the single‐cell level was quantified using protoplasts via Western blotting. The results confirmed that the viral translation efficiency was significantly reduced in the absence of BjeIF2Bβ (Figure 5d). Similarly, TuMV protein accumulation was significantly increased in OE‐BjALKBH9B_FLAG groups compared to OE‐FLAG groups (Figure 5e). The results demonstrated that the absence of BjeIF2Bβ and BjALKBH9B can result in a reduction in viral translation efficiency, and that the two proteins function in a synergistic manner.

Discussion

Viruses are thought to cause about half of all plant diseases, leading to massive losses in global agricultural production. The challenges that lie ahead in genetic engineering for resistance offer opportunities for further advances in sustainable crop protection. The eukaryotic translation initiation factors represent priority targets for target for RNA virus resistance (Bastet et al., 2017; Gal‐On et al., 2017). The use of CRISPR/Cas9 has successfully conferred resistance to viruses by editing the eIF4E and eIF4G to confer viral immunity in several crops without affecting plant physiology (Bastet et al., 2017, 2019; Chandrasekaran et al., 2016; Gomez et al., 2019; Macovei et al., 2018). Nevertheless, a common feature of the knockout strategies for specific eIFs is that the resulting resistance is often overcome by viruses interacting with other host factors from the multigene family resulting from ancestral gene duplications (Bastet et al., 2017). It is worth considering at the host‐virus co‐evolution, which can result in the host losing resistance (Charron et al., 2008). The engineering of CRISPR‐Cas9‐based resistance against viruses necessitates the identification of additional host factors required by viruses as new targets. Recessive resistance to viruses in the allopolyploid Brassica crops is very rare, despite extensive research in another allopolyploid Brassica species, B. napus, where only dominant resistance has been found (Walsh and Jenner, 2002). Here, we have demonstrated a new promising target for resistance to TuMV, not only by natural variation and molecular‐assisted selection, but also by editing of the eIF2Bβ in Brassicaceae.

In the Brassicaceae family, the prevalence and recurrence of polyploidization means that these genomes have massively duplicated genes, which could reinforce their adaptations due to relaxed purifying selection (Cheng et al., 2018). Although redundancy among eIFs probably buffers adverse phenotypes in polyploids caused by loss‐of‐function of individual eIFs, it is an Achilles' heel for resistance by providing multiple viral susceptibility factors (Bastet et al., 2017). Given the pivotal role of translation initiation in cellular processes, cumulative loss‐of‐function mutations could have strong deleterious effects even if individual eIFs mutants show little or no phenotype. Precise structural analyses of eIFs have already been carried out, providing new insights into the structural basis of viral resistance by identifying the specific amino acid substitutions within eIFs that are associated with resistance without affecting its translation initiation function (Ashby et al., 2011; Bastet et al., 2017; German‐Retana et al., 2008). It is possible to design eIFs resistance alleles using point mutations or base editing that preserve translational functions, and maximize the spectrum of target viruses, while reducing the possibility of being overcome. This approach could be used in crops to develop efficient new resistance to viruses without compromising yield.

The m⁶A modification has emerged as an essential regulator of numerous fundamental aspects of plant RNA biology, including RNA processing, maturation and decay, nuclear export and translation (Gilbert et al., 2016). In mammals, the m⁶A demethylation has been implicated in the viral life cycle of and cellular response to viral infection (Williams et al., 2019). Furthermore, the m⁶A demethylation activity is also a regulatory strategy in plants that controls the cytoplasmic replication of AMV (alfalfa mosaic virus), but not CMV (cucumber mosaic virus) (Martinez‐Perez et al., 2017). In addition, m⁶A methyltransferase B has also been reported to be associated with the stability of viral genomic RNA and has been identified as a positive regulator of wheat yellow mosaic virus infection (Zhang et al., 2022). The dynamics of m⁶A modification during the interaction of rice and viruses may represent an important plant‐virus regulatory strategy in gene expression (Zhang et al., 2021). The m⁶A demethylase ALKBH9B has been identified to function in several RNA virus infections, such as AMV (Martinez‐Perez et al., 2017) and PVY (potato virus Y) (Yue et al., 2022). To date, it is known that an ALKB domain encoded by the viral genomic RNA has been identified in the plant virus families Closteroviridae, Alphaflexiviridae, Flexiviridae, Betaflexviridae and Secoviridae (Bratlie and Drablos, 2005; Halgren et al., 2007; Moore and Meng, 2019; van den Born et al., 2008). In the Potyvirus family, one of the largest genus of plant RNA viruses, the ALKB domain within the P1 protein has been identified in Endive necrotic mosaic virus and another virus of a putative new species within this family, but not either in TuMV (Yue et al., 2022), or in our study. It has been reported that ALKBH9B interacts with the AMV, which has a profound effect on the viral infection cycle (Martinez‐Perez et al., 2017). This evidence allowed us to further investigate the role and mechanism of ALKBH9B in viral RNA infection. Here, we found that BjeIF2Bβ confers resistance to TuMV by interacting with BjALKBH9B to modify the m⁶A of the CP of TuMV, thereby facilitating viral infection. In addition, ALKBH9A and ALKBH9C were found to be dispensable for the AMV cycle, suggesting that ALKBH9B activity is highly specific to viral RNA infection (Martinez‐Perez et al., 2021). These results indicate that the virus may exploit the host ALKB protein to modify the methylation of the viral genomic RNA, thereby facilitating its infection when it lacks ALKB protein within its genomic RNA. It is therefore also worth investigating whether the natural allele of ALKBH9B is associated with resistance to viral infection, which may be a useful target for the molecular breeding for TuMV resistance and beyond.

In eukaryotic cells, protein synthesis starts at the 5′ untranslated region, either via a cap‐dependent or cap‐independent mechanism (Hinnebusch et al., 2016). m⁶A in the 5′UTR can be translated in a cap‐independent manner, with 5′UTR m⁶A bypassing 5′ cap‐binding proteins to facilitate translation under stress in the Hela cell (Meyer et al., 2015). Furthermore, m⁶A drives extensive mRNA circularization by METTL3‐eIF3h, which enhances translation (Choe et al., 2018). These findings indicated a link between eukaryotic translation initiation factors and the epitranscriptome. In mammalian cells, the integrated stress response (ISR) results in a reduction in protein synthesis by inhibiting the eukaryotic translation initiation factor 2B (eIF2B) (Bogorad et al., 2018). The activity of eIF2B is inhibited by phosphorylation of its substrate eIF2 by several stress‐induced kinases, which in turn triggers the ISR. A structural analysis revealed that eIF2B is a heterodecameric complex comprising two copies of each of the α‐, β‐, γ‐, δ‐ and ε‐subunits. Of these, the α‐, β‐ and δ‐subunits constitute the regulatory subcomplex (Kashiwagi et al., 2016, 2019). Moreover, the viral factors involved in the interplay between m⁶A methylation and viral infection should be addressed. Indeed, the viral CP is the preferential protein that encounters the plant immune response after viral invasion, where the process of CP recognition with the m⁶A methylation is a particularly intriguing question. The mRNA of viral CP has been reported to be modified by methyltransferase B in wheat (Zhang et al., 2022) and by ALKBH9B in our study, which affects viral accumulation in the host, suggesting the potential application of m⁶A methylation in plant virus control strategies.

Taken together, we propose a model that illustrates how eIF2Bβ directly interacts with ALKBH9B to mediate TuMV resistance. In the WT, eIF2Bβ recruits ALKBH9B to bind to the TuMV mRNA and modify its m⁶A status. This maintains the normal translation of viral proteins and viral infection. In eif2bβ mutants, the absence of eIF2Bβ results in a reduction in ALKBH9B binding to viral mRNA, leading to an increased status of m⁶A methylation, which in turn restricts the level of viral protein translation and thus affects the virus epidemic (Figure 5g). Meanwhile, the natural variation of eIF2Bβ simultaneously offers interesting perspectives for further molecular design of viral resistance in Brassicaceae crops and beyond.

Experimental procedures

Plant materials and TuMV isolate inoculation

A resequencing panel comprising 208 accessions of B. juncea (Kang et al., 2021; Yang et al., 2021) and 198 accessions of B. rapa (Cheng et al., 2016) from previous publications was employed to verify the SNP and Indel genotypes of eIF2Bβ. Accessions of B. rapa (ZLC, SHAYC, FS212, ZYMJ, SC and YQ49), B. juncea (T84‐66 and 03D102) and A. thaliana Col‐0 were selected for phenotypic and genotypic analyses following inoculation with the TuMV_ZJ isolate. The susceptible accessions B. juncea var. XLH and B. napus var. Westar were employed for eIF2Bβ editing via the CRISPR‐Cas9 technique.

Leaves infected with the TuMV_ZJ isolate were harvested and ground in 20 mL of 1 X PBS buffer to obtain the inoculum. When the plants had reached the three‐to‐four true‐leaf stage, the leaves were abraded with silicon carbide to create wounds. 100 μL of inoculum was dripped onto the wounds, and the leaves were subsequently rinsed with running water after a 20‐min waiting period. Following inoculation for 9–15 days, the plants exhibited virus symptoms.

Genotyping and phylogenetic analysis of eIF2Bβ in B. rapa and B. juncea

Microsynteny pattern analysis was performed using MCScanX (Wang et al., 2012), and eIF2Bβ sequences of A. thaliana, B. rapa and B. juncea were obtained separately from TAIR10 (https://www.arabidopsis.org/) and BRAD (http://brassicadb.cn).

For phylogenetic analysis, the resequencing data of B. juncea and B. rapa were aligned to the reference genome of B. juncea, T84‐66, using Sentieon pipeline (https://github.com/Sentieon). Subsequently, variants retrieved from the A subgenome of B. rapa and B. juncea were extracted for the construction of phylogenetic tree. The low‐quality variants were filtered using vcftools (version 0.1.16) (Danecek et al., 2011), with the following parameters: ‐‐minDP 4, ‐‐maxDP 100, ‐‐minGQ 10, ‐‐minQ 30, ‐‐min‐meanDP 3, ‐‐maf 0.05, ‐‐max‐missing 0.8. The variation loci were then annotated using snpEff (Cingolani et al., 2012). Fourfold degenerate synonymous sites (4dTVs) were extracted from the annotated vcf format file and then transferred to the phylip format (Felsenstein, 1989) using a Python script. A maximum‐likelihood (ML) phylogenetic tree based on 4dTVs was constructed using IQ‐TREE (v2.1.4‐beta) with best model (UNREST + FO + I + G4) and 1000 replicates for ultrafast bootstrap (Minh et al., 2020). The phylogenetic tree and the eIF2Bβ haplotype of B. rapa and B. juncea were visualized using iTOL (https://itol.embl.de/) (Letunic and Bork, 2021).

Construction of eIF2Bβ edited lines

Two sgRNAs of each eIF2Bβ in B. juncea and B. napus were designed using the CRISPR‐P 2.0 tool (http://crispr.hzau.edu.cn/CRISPR2/) (Lei et al., 2014). The corresponding oligonucleotides were synthesized and inserted into the vector pKSE401 using the Golden Gate technology (Xing et al., 2014). The confirmed recombinant vectors were then introduced into Agrobacterium tumefaciens (strain GV3101). Regeneration plants were achieved based on the Agrobacterium‐mediated transformation in Brassicaceae family established in our laboratory. In order to identify the mutational message, genomic DNA was extracted from transgenic plants using the CTAB method to amplify the region containing the target sites. The transgenic plants that had been successfully edited were self‐pollinated until they were homozygous for the mutation, which was then used in subsequent assays.

RT‐qPCR and Western blotting

Total RNA was extracted using TRIzol reagent (Thermo Fisher Scientific, Carlsbad, CA, USA), and 1000 ng of total RNA was reverse transcribed to cDNA using ReverTra Ace qPCR RT Master Mix with gDNA Remover (TOYOBO, Osaka, Japan). A quantitative real‐time PCR was conducted utilizing TOROGreen qPCR Master Mix (TOROIVD, Shanghai, China) on a StepOne Plus system (Applied Biosystems, Foster City, CA, USA). Each analysis was calculated with a minimum of three biological replicates and three technical replicates. The relative transcription levels of the target genes were determined using the 2^−ΔΔCt method (Livak and Schmittgen, 2001). Primers utilized in this study are presented in Table S3.

A total of 0.3 g of TuMV‐infected plant leaves were harvested and lysed in 600 μL plant RIPA buffer (Beyotime, Shanghai, China) to obtain total protein. The supernatant was subjected to protein concentration determination using the Enhanced BCA Protein Assay Kit (Beyotime, Shanghai, China). For the immunoblot, each sample was separated in a 10% SDS‐polyacrylamide gel by electrophoresis, and the protein bands were then transferred to a PVDF membrane (BIO‐RAD, Hercules, CA, USA) by wet electroblotting. Following the blocking with 5% skim milk, the blots were probed with TuMV monoclonal antibody (Green Agriculture Safeguard, Beijing, China), plant actin polyclonal antibody (Abbkine, Wuhan, China) or FLAG‐tag antibody (ZENBIO, Chengdu, China). Subsequently, HRP‐conjugated goat anti‐mouse (ZENBIO, Chengdu, China) or goat anti‐rabbit (Agrisera, Vännäs, Sweden) secondary antibodies were incubated with the blots. The detection signals were developed using ECL reagent and subsequently visualized using the ChemiDoc MP imaging system (BIO‐RAD, Hercules, CA, USA).

Subcellular localization of BjeIF2Bβ and its co‐localization with viral protein and BjALKBH9B

Subcellular localization of eIF2Bβ was performed using Arabidopsis protoplasts according to previously published methods (Yoo et al., 2007). Arabidopsis protoplast extraction and transformation were conducted in accordance with the instructions of Arabidopsis Protoplast Preparation and Transformation Kit (Coolaber, Beijing, China). The OASA1 (Alvarez et al., 2010), PIP2;1 (Li et al., 2011), HY5 (Li et al., 2022) and PDI6 (Yuen et al., 2013) were employed as cell organelle markers for the cytosol, plasma membrane, nucleus and endoplasmic reticulum, respectively. The recombinant vectors of BjeIF2Bβ fused with GFP, the viral protein (6K2_VPg_Pro from TuMV_ZJ isolate) fused with CFP and BjALKBH9B fused with mcherry were transformed into Agrobacterium strain GV3101. Following centrifugation at 4000  g for 10 min, the solution was resuspended in buffer (10 mM MES pH = 5.7, 10 mM MgCl₂, 150 μM acetosyringone) with OD₆₀₀ diluted to 1.0. Following a 3‐h incubation at 28°C, injection was prepared in a 1:1:1 ratio (eIF2Bβ: 6K2_VPg_Pro: ALKBH9B) and co‐infiltrated into N. benthamiana leaves. After infiltration for 48 h, the detection of fluorescent signals was identified and documented using a confocal laser scanning microscope (Nikon, Tokyo, Japan).

Protein–protein interaction assay

For yeast two‐hybrid (Y2H) assays, recombinant AD and BD vectors were co‐transformed into Y2HGold competent cells utilizing a PEG/LiAc‐based method. The transformed yeast cells were placed on two different media: SD/‐Trp‐Leu and SD/‐Trp‐Leu‐His‐Ade/X‐α‐gal. Transformations carrying empty pGADT7 or pGBKT7 vectors were performed to test for autoactivation and toxicity. The pGADT7‐T and pGBKT7‐53 combinations and pGADT7‐T and pGBKT7‐lam combinations were used as positive and negative controls, respectively.

For bimolecular fluorescence complementation (BiFC) assays, the CDS of BjeIF2Bβ, BjALKBH9B and BjuVB05G39250 were cloned into P2YN and P2YC vectors, respectively. The recombinant vectors were transformed into Agrobacterium strain GV3101 and the subsequent experiments were similar to the co‐localization assay.

For co‐immunoprecipitation (Co‐IP) assays, Agrobacterium strain GV3101 carrying eIF2Bβ‐GFP and ALKBH9B‐3Flag was co‐infiltrated into N. benthamiana leaves. Following infiltration for 48 h, 0.5 g of leaves were harvested and ground in 1 mL IP buffer (50 mM Tris–HCl pH = 7.5, 150 mM NaCl, 10 mM MgCl₂, 5 mM DTT, 0.1% Triton X‐100 and protease inhibitor cocktail) to extract total protein. After centrifugation at 8000  g for 20 min at 4°C, the supernatant was incubated with 40 μL of anti‐Flag‐M2 magnetic beads (Merck, Darmstadt, Germany) for 4 h at 4°C with rotation. Following 5 washes with IP buffer at 4°C, the bead‐bound protein complex was eluted in 200 μL elution buffer (5 mM EDTA and 200 mM NH₄OH). After individual mixing with SDS loading buffer and boiling for 10 min, the samples were processed for Western blotting in accordance with the previously described methodology.

Analysis of viral translation efficiency

The extraction and purification of TuMV virions were performed, and the resulting purified virions were equally transfected into protoplasts extracted from WT and bjeif2b (Kikkert et al., 1997). Similarly, 20 μg BjALKBH9B‐FLAG/FLAG recombinant plasmids and 20 μL purified TuMV virions were co‐transfected into protoplasts of WT. Following a 24‐h incubation period, 2.5 × 10⁶ protoplasts were collected by centrifugation and used for Western blotting in accordance with the previously described methodology.

m⁶A RIP seq assay

mRNAs were enriched from 50 μg of total RNA extracted from mustard leaves 9 days after inoculation with TuMV using VAHTS mRNA capture beads (Vazyme, Nanjing, China). Following the addition of 20 mM ZnCl₂, mRNA fragments of 100–200 nt length were cleaved. A specific m⁶A antibody (Synaptic Systems, Goettingen, Germany) was employed for m⁶A immunoprecipitation. Sequencing libraries were prepared using the Stranded mRNA Library Prep Kit (KC‐Digital, Wuhan, China). The 200–500 bp library products were sequenced on a DNBSEQ‐T7 model PE150 sequencer (MGI Tech, Shenzhen, China). After filtering low‐quality reads and removing duplication bias, the processed sequences were aligned to the TuMV reference genome using the STAR software (version 2.5.3a) with default parameters. Peak calling and annotation were performed using exomePeak (version 3.8) and bedtools (version 2.25.0). The distribution of peaks was analysed using DeepTools (version 2.4.1), and a Python script was employed to identify different m⁶A peaks using the Fisher test.

RNA immunoprecipitation (RIP)‐qPCR assay

A 20 μg BjeIF2Bβ‐FLAG/BjALKBH9B‐FLAG plasmid and 20 μL purified virions of TuMV were co‐transfected into protoplasts extracted from WT. Following a 24‐h incubation, the protoplasts were cross‐linked with 1% formaldehyde and terminated with 125 mM glycine. The collected protoplasts were lysed in 1 mL RIP buffer (150 mM NaCl, 25 mM Tris–HCl pH = 7.4, 5 mM EDTA, 0.5 mM DTT, 0.5% Tergitol NP40, protease inhibitors and RNase inhibitors). After centrifugation, each 100 μL of the supernatant was separated and used as input for RNA extraction and protein detection. A further 30 μL anti‐Flag‐M2 magnetic beads (Merck, Darmstadt, Germany) and magnetic bead‐conjugated mouse IgG (ABclonal, Wuhan, China) were added to the remaining supernatant for immunoprecipitation. The mixture was incubated at 4°C with rotation overnight. Following three washes with RIP buffer at 4°C, the RNA–protein complex was treated with DNase I (Takara, Osaka, Japan) and Proteinase K (Takara). The IP and input RNA were extracted and then reverse transcribed into cDNA for qPCR, as described above. The proteins present in the input and IP samples were collected for subsequent analysis via Western blotting.

Cell‐free protein synthesis and purification and microscale thermophoresis assay

The expression vector pD2P, comprising BjALKBH9B, BjeIF2Bβ and GUS gene, or an empty plasmid, was combined with the RXN solution (Kangma, Shanghai, China) to initiate protein synthesis. Subsequently, the reaction mixture was incubated at room temperature for 6 h, after which Ni‐NTA magnetic beads (Yeasen, Shanghai, China) were added to purify the target protein. A wash buffer with a range of imidazole concentrations was employed to remove impure proteins. The target proteins were then eluted using elution buffer, and the purified proteins were desalted using a SpinDesalt column (Smart‐Lifesciences, Changzhou, China). To verify the purified proteins, SDS‐PAGE and Coomassie blue staining analyses were performed.

In MST assay, BjALKBH9B and BjeIF2Bβ fused with eGFP were employed as Target, and CP RNA product was employed as Ligand. The assay buffer was PBST. Samples containing different kinds of Target and Ligand with 16 concentration gradients were loaded into standard capillaries and fluorescence was detected using the Monolith X instrument (NanoTemper, Munich, Germany). The results were analysed using NanoTemper analysis software in order to estimate equilibrium dissociation constant (Kd) values and signal‐to‐noise ratio.

Conflict of interest

The authors declared no conflict of interest.

Author contributions

J.Y. and M.Z. conceived and designed the research. T.S., Z.L. and S.X. performed the main research. Other authors contributed experiments and discussions. J.Y. and T.S. drafted and edited the manuscript. All authors have read and approved the manuscript.

Supporting information

Figure S1 Microsynteny of the targeted TuMV resistance associated eIF2Bβ among B. juncea, B. rapa and A. thaliana.

Figure S2 Schematic diagram of four Indel genotypes of eIF2Bβ in B. rapa.

Figure S3 Alignment of Indel of eIF2Bβ among B. juncea and B. rapa.

Figure S4 TuMV‐CP expression and TuMV isolate accumulation after 2 weeks of TuMV inoculation in Brassicaceae.

Figure S5 Alignment of eIF2Bβ (BjuA01G04814) from lines susceptible to TuMV with its orthologues from B. napus (BnaC01T0437600WE) in terms of amino acid.

Figure S6 The resulting missense mutation in the eIF2Bβ gene as determined by Sanger sequencing in B. juncea.

Figure S7 The resulting missense mutation in the eIF2Bβ gene as determined by Sanger sequencing in B. napus.

Figure S8 The binding of BjeIF2Bβ or BjALKBH9B to the m⁶A peak region of CP mRNA.

Figure S9 Binding detected of BjALKBH9B, BjeIF2Bβ with CP RNA using microscale thermophoresis (MST) and SDS‐PAGE analysis of purified proteins used in MST assays.

Figure S10 The domain scan results of 10 mature proteins encoded by TuMV_ZJ.

Table S1 Genotyping analysis based on SNPs and Indel of eIF2Bβ in B. juncea varieties.

Table S2 Genotyping analysis based on SNPs and Indel of eIF2Bβ in B. rapa varieties.

Table S3 Primers used in this study.

Data availability statement

All data are available in the main text or the supplementary information. The raw data and materials that support the study are available from the corresponding author upon reasonable request.