Trans‐species synthetic gene design allows resistance pyramiding and broad‐spectrum engineering of virus resistance in plants

Summary To infect plants, viruses rely heavily on their host's machinery. Plant genetic resistances based on host factor modifications can be found among existing natural variability and are widely used for some but not all crops. While biotechnology can supply for the lack of natural resistance alleles, new strategies need to be developed to increase resistance spectra and durability without impairing plant development. Here, we assess how the targeted allele modification of the Arabidopsis thaliana translation initiation factor eIF4E1 can lead to broad and efficient resistance to the major group of potyviruses. A synthetic Arabidopsis thaliana eIF4E1 allele was designed by introducing multiple amino acid changes associated with resistance to potyvirus in naturally occurring Pisum sativum alleles. This new allele encodes a functional protein while maintaining plant resistance to a potyvirus isolate that usually hijacks eIF4E1. Due to its biological functionality, this synthetic allele allows, at no developmental cost, the pyramiding of resistances to potyviruses that selectively use the two major translation initiation factors, eIF4E1 or its isoform eIFiso4E. Moreover, this combination extends the resistance spectrum to potyvirus isolates for which no efficient resistance has so far been found, including resistance‐breaking isolates and an unrelated virus belonging to the Luteoviridae family. This study is a proof‐of‐concept for the efficiency of gene engineering combined with knowledge of natural variation to generate trans‐species virus resistance at no developmental cost to the plant. This has implications for breeding of crops with broad‐spectrum and high durability resistance using recent genome editing techniques.


Introduction
Synthetic biology has emerged as a promising tool to provide advantageous new functions for crop improvement using plant biotechnology (Liu and Stewart, 2015). One aspect of synthetic biology is the possibility of editing the plant's own genes to induce changes in the way the plant interacts with its environment (Baltes and Voytas, 2015). This approach was successfully used to improve water use and drought stress tolerance in tomato (Solanum lycopersicum) and Arabidopsis thaliana by modifying the abscisic acid (ABA) receptor PYRABACTIN RESISTANCE 1 to make it responsive to an agrochemical ligand (Park et al., 2015). Resistance to pathogens is also an important field in which gene engineering can play a major role by restricting the way pathogens hijack host factors essential for their infectious cycle. For example, targeted modification of the JA hormone receptor in Arabidopsis maintains efficient resistance to the bacterial effector from Pseudomonas syringae without impairing the JA recognition pathway (Zhang et al., 2015). It is of particular interest to understand how specific modifications of plant genes, often resulting from nonsynonymous Amino Acid (AA) changes, can generate new functions. These modifications can then be implemented through high-throughput allele mining (Barabaschi et al., 2016) or new genome editing technologies such as CRISPR/ Cas9 (Ma et al., 2016). Overall, gene editing technologies appear as a promising tool to enhance and accelerate plant breeding (Østerberg et al., 2017;Palmgren et al., 2015).
Viruses are obligate intracellular parasites encoding very few proteins and are thus highly dependent on host factors for successful infection. A widely developed strategy to counter viral infections in crops is to modify these factors or inhibit their expression, therefore causing resistance by loss-of-susceptibility (Pavan et al., 2010; van Schie and Takken, 2014). The eukaryotic initiation factor 4E (eIF4E) and its isoform (eIFiso4E) have been proven to be susceptibility factors to several economically important single-stranded positive sense RNA (ssRNA+) viruses (mainly Potyviridae family members) in many crops Wang and Krishnaswamy, 2012). In cells, translation initiation factors 4E are essential elements of the cellular translation initiation process and are in charge of binding the capped mRNA and recruiting the other translation initiation factors (Browning and Bailey-Serres, 2015).
Natural diversity is a great reservoir of eIF4E resistance alleles in crops and their related wild species. Mostly, these resistance alleles encode modified eIF4E proteins, which still retain a functional role in translation initiation while carrying nonsynonymous AA changes, often located in two conserved regions of the protein . Such resistance alleles have been successfully deployed in many crop species such as tomato, barley (Hordeum vulgare), lettuce (Lactuca sativa), melon (Cucumis melo) and pepper (Capsicum annuum) (Nicaise et al., 2003;Nieto et al., 2006;Ruffel et al., 2002Ruffel et al., , 2005Stein et al., 2005). However, such natural virus resistance alleles are not available in some other important crops such as papaya (Carica papaya), Prunus species or cassava (Manihot esculenta), which are challenged by economically important viruses such as Papaya ringspot virus, Plum pox virus and Cassava brown streak virus, respectively (Bart and Taylor, 2017;Garc ıa et al., 2014;Gonsalves, 1998). Moreover, even if resistance alleles are present in related wild species, their introgression into cultivated accessions can be difficult in the case of reproductive incompatibility or genetic linkage drag bringing unwanted traits (Lin et al., 2014). To compensate for this lack of natural alleles, eIF4E gene disruption can be used to induce virus resistance. Such approaches have been successfully used in some species such as Arabidopsis and cucumber (Cucumis sativus L.) (Chandrasekaran et al., 2016;Duprat et al., 2002;Pyott et al., 2016). Because of redundancy among the eIF4E gene family, plants knocked out (KO) for a single translation initiation factor mostly exhibit normal physiology (Bastet et al., 2017). However, as ssRNA + viruses recruit selectively distinct members of the eIF4E gene family, a single KO is often associated with a limited resistance spectrum while simultaneous KO of several members of the eIF4E gene family often leads to lethality or impaired growth Gauffier et al., 2016). Consequently, pyramiding resistances by knocking out several factors to extend the virus resistance spectrum can be impeded by the induction of developmental defects as shown in Arabidopsis  and tomato (Gauffier et al., 2016). It is also expected that redundancy among 4E (i.e. eIF4E or eIFiso4E) genes could reduce eIF4E-based resistance durability by making other 4E factors available to viruses (Bastet et al., 2017). Finally, a comparison between natural functional eIF4E resistance alleles and engineered loss-of-function KO alleles in tomato unveiled unexpected regulatory processes between the members of the eIF4E family making this approach less effective (Gauffier et al., 2016).
In the light of these observations, the best strategy to develop eIF4E-based resistance would be to design functional alleles by introducing point mutations in the gene, mimicking naturally occurring resistances, instead of knocking them out (Bastet et al., 2017). In this regard, several studies have shown that eIF4E or eIFiso4E could be engineered as resistance alleles, in tomato, potato (Solanum tuberosum) and Chinese cabbage (Brassica rapa), as attested by the transgenic ectopic expression of resistance alleles under the control of a strong constitutive promoter in a susceptible genetic background (Cavatorta et al., 2011;Kang et al., 2007;Kim et al., 2014).
In the present paper, we aimed at extending this approach by mimicking allele replacement and addressing the functional role of the modified eIF4E protein using the well-characterized Arabidopsis thaliana/potyvirus pathosystem (Ouibrahim and Caranta, 2013).
We report for the first time the construction of a synthetic resistance allele of eIF4E1 by introducing six nonsynonymous mutations deduced from pea eIF4E allelic variability. We show that a phenotypic defect associated with eif4e1 loss of function, a bolting delay, can be complemented by the synthetic allele, while maintaining the resistance to the potyvirus Clover yellow vein virus (ClYVV). Significantly, by maintaining function in planta, this synthetic resistance allele allows resistance pyramiding with an eifiso4e loss-of-function allele and confers new resistances to another important potyvirus species as well as to previously described eIFiso4E-resistance-breaking (RB) isolates. In addition, these double-mutated plants displayed resistance to a polerovirus species. This study therefore establishes a proof-of-concept for editing susceptibility genes to create new genetic resistances and expand resistance spectrum without affecting the plant development and agronomic traits.

Design of a synthetic eIF4E1 R resistance allele in Arabidopsis
The wild-type Arabidopsis eIF4E1 gene from the Columbia (Col) accession was chosen as a target to design a synthetic resistance allele. eIF4E1 encodes a susceptibility factor to the potyvirus Clover yellow vein virus (ClYVV) and, consequently, inactivation of eIF4E1 (eif4e1 KO ) is associated with resistance to ClYVV (Sato et al., 2005). No natural eIF4E1 resistance allele is known in Arabidopsis, so we looked for data on genetic resistance determinants in other species. We focused on eIF4E1 natural variation in pea (Pisum sativum), because the natural eIF4E resistance allele (known as sbm-1, cyv-2 or wlv) from pea accession JI1405 confers resistance to several potyviruses, notably ClYVV (Andrade et al., 2009;Bruun-Rasmussen et al., 2007;Gao et al., 2004). The eIF4E1 protein in Arabidopsis is highly similar to its counterpart in pea (81% homology/89% similarity) (Gao et al., 2004) (Figure 1a). Another resistance allele isolated from pea accession PI269818 and conferring resistance to one pathotype of Pea seed-borne mosaic virus (PsbMV) was also considered in this study. These two pea resistance alleles are associated with the presence of five AA substitutions and one deletion within the protein sequence: W62L (i.e. the tryptophan at position 62 is substituted by a leucine in the protein encoded by the resistance allele), A73D, A74D, G107R and N169K for JI1405 line, as well as the deleted serine at position 77 (S77) in the PI269818 accession. These AA changes are mainly located in two regions (regions I and II) close to the cap-binding pocket according to the eIF4E 3D structure (Marcotrigiano et al., 1997;Robaglia and Caranta, 2006) ( Figure S1). In Arabidopsis, the AA located at similar positions (AA 69,80,81,114 and 176) are, in most cases, identical to the AA in the pea susceptible allele (Figure 1a). The synthetic Arabidopsis eIF4E1 allele was obtained by substituting the AA present in the five locations by the corresponding AA from the resistant pea accession JI1405. The position of the deletion in the resistant pea accession PI269818 was substituted by an alanine in the Arabidopsis eIF4E1 sequence. The 3D predicted structure of the Arabidopsis eIF4E1 protein carrying the six modifications W69L, T80D, S81D, S84A, G114R, and N176K mimics the structure of the pea resistant alleles. No drastic changes in the protein structure were predicted, hinting that the function of this mutated form of eIF4E1 in translation initiation should not be strongly affected (Figure 1b,c and Figure S1).
All six nonsynonymous substitutions were introduced by PCRbased directed mutagenesis on a 3.4-kb construct covering the whole Arabidopsis eIF4E1 gene including its endogenous promoter. This allele (called hereafter eIF4E1 R ) was introduced into Arabidopsis plants knocked out for eIF4E1 (eif4e1 KO lines), and three independent transgenic lines were selected for further analyses. The correct expression of the eIF4E1 R transgene was checked in all three lines ( Figure S2), as well as the sequence of Figure 1 Design of a synthetic Arabidopsis resistance allele eIF4E1 R . (a) Alignment of eIF4E protein sequences from pea susceptible (JI2009) and resistant (JI1405 and PI269818) accessions with the Arabidopsis thaliana eIF4E1 (AteIF4E1). The position of the amino acids that differ between susceptible and resistance pea accessions is highlighted in grey. Black triangles above the alignment indicate the mutations introduced in Arabidopsis eIF4E1 to create the synthetic eIF4E1 R . (b and c) Three-dimensional predicted structure of AteIF4E1 (b) and of eIF4E1 R (c) based on homology modelling using the pea eIF4E 3D structure as a template (PDB ID: 2WMC-C). The modified amino acids between AteIF4E1 and eIF4E1 R are coloured in red with side chains shown. the expressed eIF4E1 mRNA in all lines. As controls, a construct harbouring a wild-type eIF4E1 was introduced into the same eif4e1 KO background, as well as an unrelated transgene, expressing the GUS reporter gene under control of the CaMV 35S promoter. eIF4E1 R encodes a functional protein that complements eif4e1 KO bolting delay Although 4E translation initiation factors are largely redundant in their functions (Browning and Bailey-Serres, 2015;Combe et al., 2005;Duprat et al., 2002;Nicaise et al., 2007), the eif4e1 KO Arabidopsis plants were found to display a consistent 7-day bolting delay when compared with the wild-type Columbia accession (Col WT) (Figure 2a, b, Bastet et al., 2017). To assess whether eIF4E1 R encodes a functional translation initiation factor, the ability of eIF4E1 R to complement this bolting delay was tested. Indeed, eIF4E1 R restored the wild-type timing of bolting (Figure 2a,b). Furthermore, total protein extracts were subjected to cap analog-affinity assay and, on all independent transgenic lines expressing either eIF4E1 or eIF4E1 R , eIF4E1 proteins were shown to bind the cap analog, confirming that eIF4E1 R encodes a functional protein ( Figure 2c).

eIF4E1 R is a resistance allele to ClYVV
We next addressed whether the six mutations introduced in the synthetic allele were sufficient to confer resistance to ClYVV, a potyvirus that requires an eIF4E1 factor for its viral cycle. All plants were mechanically inoculated with ClYVV and the viral accumulation was measured 30 days postinoculation (dpi) by double antibody sandwich enzyme-linked immunosorbent assay (DAS-ELISA) ( Figure 3). As expected, ClYVV accumulated in both Col WT and eIF4E1-complemented plants while no virus accumulation was observed in eif4e1 KO plants. Interestingly, no virus accumulation was detected in the eif4e1 KO plants complemented by the eIF4E1 R construct suggesting that this mutated factor cannot be recruited by ClYVV. Altogether, these results show that the six AA changes introduced in eIF4E1 R are sufficient to convert eIF4E1 into a resistance allele to ClYVV, while maintaining its physiological functionality. This is similar to the resistance genes that have been previously identified in crops or in their wild relatives . eIF4E1 R rescues the lethality associated with the double mutation eif4e1 KO eifiso4e KO and does not impact plant yield It was previously shown that the resistance spectra associated with either eif4e1 KO or eifiso4e KO , to ClYVV and Turnip mosaic virus (TuMV), respectively (Duprat et al., 2002;Lellis et al., 2002), could not be combined in the same plant because of plant lethality Patrick et al., 2014). We speculated that since eIF4E1 R encodes a functional cap-binding factor, it may successfully be combined with loss-of-function alleles affecting both eIF4E1 and eIFiso4E and could therefore rescue the lethal phenotype. To verify this hypothesis, two previously described independent T2 transgenic lines [eif4e1 KO / eif4e1 KO eIF4E1 R ] (homozygous for eif4e1 KO allele but complemented by eIF4E1 R ) were crossed with [eif4e1 KO /eif4e1 KO eIFiso4E/eifiso4e KO ] (homozygous for eif4e1 KO but heterozygous for eifiso4e KO ) plants. The resulting F3 plants knocked out for eIF4E1 and eIFiso4E but complemented by eIF4E1 R were viable (Figures 4a and S3). The rescue of the lethal phenotype by eIF4E1 R further confirmed that this allele was fully functional in planta. Consistently, the eif4e1 KO /eif4e1 KO eifiso4e KO /eifiso4e KO plants complemented by eIF4E1 R (named hereafter eif4e1 KO eifiso4e KO eIF4E1 R ) did not show any delayed bolting phenotype compared with wild-type plants (Figures 4a and S4). In addition, no decrease in dry weight or in seed yield was observed (Figure 4b,c and S4). In conclusion, the synthetic eIF4E1 R allele can fully complement the lack of both eIF4E1 and eIFiso4E isoforms in planta.
Cumulative resistance to several potyviruses in eIF4E1 Rcomplemented plants As eIF4E1 and eIFiso4E are required for ClYVV and TuMV multiplication, respectively, eif4e1 KO eifiso4e KO eIF4E1 R plants, together with control lines including single mutants affecting eIF4E1 or eIFiso4E alone, were separately challenged with these two viruses, and the accumulation of viruses was measured by ELISA ( Figure 5). As expected, Col WT plants were susceptible to both viruses while eif4e1 KO and eifiso4e KO plants were resistant to ClYVV and TuMV, respectively. Remarkably, ClYVV or TuMV did not accumulate in the eif4e1 KO eifiso4e KO eIF4E1 R lines. Lettuce mosaic virus (LMV) and Plum pox virus (PPV) that both rely on the isoform eIFiso4E (Duprat et al., 2002;Nicaise et al., 2003) were further inoculated onto the similar mutated lines, and the results showed that the eif4e1 KO eifiso4e KO eIF4E1 R lines were fully resistant to these potyviruses as well ( Figure S5). Therefore, by complementing the loss-of-function for both eIF4E1 and eIFiso4E, the synthetic eIF4E1 R allele allows combination of the resistance spectrum associated with each individual mutation, leading to an overall broad resistance spectrum to several potyviruses. eIF4E1 R confers resistance to a potyvirus relying on both eIF4E1 and eIFiso4E Watermelon mosaic virus (WMV) is a potyvirus that mainly affects cucurbit plants, but also Arabidopsis. Previously, the Arabidopsis Cap Verde island accession (Cvi) was identified as a source of partial resistance to WMV relying on the chloroplastic PHOS-PHOGLYCERATE KINASE 2 gene (cPGK2) (Ouibrahim et al., 2014). Unlike most potyviruses, no resistance to WMV associated with loss-of-function alleles of either eIF4E1 or eIFiso4E has been shown so far ( Figure S6). A hypothesis is that WMV could recruit either eIF4E1 or eIFiso4E as do Pepper veinal mottle virus (PepVMV) or Chilli veinal mottle virus (ChiVMV) in Capsicum species (Hwang et al., 2009;Ruffel et al., 2006). In agreement with this assumption, the eif4e1 KO eifiso4e KO eIF4E1 R lines were also fully resistant to WMV since no virus accumulated in the inoculated plants 3 weeks postinoculation ( Figure 6). This result shows that, beyond allowing the pyramiding of the resistance spectra conferred by both mutated eIF4E and eIFiso4E, the strategy outlined here can extend the resistance to potyviruses using both factors.

eIF4E1 R restrains TuMV resistance-breaking isolates
The durability of resistance is a major issue for plant breeding and it is crucial to develop and release resistances that are efficient towards resistance-breaking (RB) emerging isolates. Significantly, eIF4E-mediated resistance can be overcome by RB potyviruses (Sanfac ßon, 2015; Wang and Krishnaswamy, 2012). Previously, two RB isolates of TuMV able to overcome eifiso4e KO -mediated resistance were isolated in Arabidopsis (Gallois et al., 2010). Each isolate, TuMV-E116Q and TuMV-N163Y, possesses a single nonsynonymous mutation (mutation E116Q or N163Y, respectively) in the avirulence factor VPg (viral genome-linked protein), that is responsible for the resistance overcoming. It had been hypothesized that the overcoming of eifiso4e KO -mediated resistance could be achieved by the ability of both isolates to recruit another translation initiation factor, namely eIF4E1 (Gallois et al., 2010).
As often with resistance-breaking strains (Moury et al., 2004), TuMV-E116Q and TuMV-N163Y accumulate weakly in eifi-so4e KO lines ( Figure S7). Therefore, to monitor plant infection, we inoculated the same plants as before with a viral isolate expressing the GFP reporter gene. The mutations E116Q and N163Y were each introduced into a TuMV-GFP infectious cDNA clone (Beauchemin et al., 2005), generating TuMV-E116Q-GFP and TuMV-N163Y-GFP, respectively. Plants were agro-inoculated, and the GFP accumulation in plants was detected 21 dpi using a GFP-imaging camera (Figure 7a). TuMV-E116Q-GFP and TuMV-N163Y-GFP, but not TuMV-GFP accumulated in eifi-so4e KO plants, confirming their ability to overcome the eifiso4e KO -mediated resistance. However, the two eif4e1 KO eifiso4e KO eIF4E1 R lines were fully resistant to both resistancebreaking TuMV isolates. The absence of virus accumulation in the eif4e1 KO eifiso4e KO eIF4E1 R lines was moreover confirmed by RT-PCR on total RNA extracted from the inoculated plants ( Figure 7b). Altogether, our results show that one nonsynonymous mutation either at position 116 or 163 in the VPg allows RB-TuMV viruses to use a new translation initiation factor, eIF4E1, in addition to eIFiso4E, and that the engineered eif4e1 KO eifiso4e KO eIF4E1 R genotype could significantly improve virus resistance durability. Translation initiation factors 4E are mostly associated with resistance to potyviruses but resistances to other ssRNA+ viruses, harbouring different genomic features, have also been described . For example, loss-of-function in eIF4E1 was shown to be implicated in a delay of susceptibility to the carmovirus Turnip crinkle virus (TCV) (Yoshii et al., 1998). We found that the engineered eif4e1 KO eifiso4e KO eIF4E1 R plants were fully susceptible to TCV, showing that the eIF4E1 R allele does not constitute a resistance allele to TCV ( Figure S8). Likewise, the engineered eif4e1 KO eifiso4e KO eIF4E1 R plants were fully susceptible to Grapevine fanleaf virus (GFLV) and Arabis mosaic virus (ArMV), two ssRNA+ nepoviruses ( Figure S9).
Finally, the role of translation initiation factors in resistance to polerovirus, such as Turnip yellows virus (TuYV), Beet mild yellowing virus (BMYV) and Beet western yellows virus-USA (BWYV-USA), had previously been investigated in Arabidopsis (Reinbold et al., 2013). TuYV was shown to require the Arabidopsis eIFiso4G1 to infect the plant, whereas BMYV and BWYV infection was compromised in plants devoid of functional eIF4E1, although the latter resistance was only partial. To assess whether eIF4E1 R could confer resistance to viruses other than potyviruses, the eif4e1 KO eifiso4e KO eIF4E1 R lines were challenged with these three polerovirus species.
Aphid-mediated inoculation was performed on controls and eif4e1 KO eifiso4e KO eIF4E1 R plants. Viral accumulation was detected by DAS-ELISA 21 dpi (Figure 8). All genotypes were fully susceptible to TuYV, consistent with the virus requirement for eIFiso4G1 and not eIF4E1 or eIFiso4E. Intermediate resistance to BMYV and BWYV-USA associated with eIF4E1 loss-of-function was confirmed as described before (Reinbold et al., 2013) but interestingly while susceptibility to BMYV was restored in eif4e1 KO eifiso4e KO eIF4E1 R lines, this genotype was fully resistant to BWYV-USA. In conclusion, the presence of the synthetic resistance allele eIF4E1 R in an eif4e1 KO eifiso4e KO genetic background is associated with a full resistance to a polerovirus isolate.

Discussion
Natural eIF4E-based resistance alleles are largely used to introduce virus resistance in crops by breeding. Genetic engineering could help provide new sources of resistance in crops that are devoid of such natural resistance. In this study, we developed a synthetic allele, eIF4E1 R that can make up for the lack of natural resistance alleles, and we gave evidence of its effectiveness in an Arabidopsis-based pathosystem. We showed that eIF4E1 R mimics natural resistance alleles when replacing endogenous eIF4E while maintaining biochemical (binding to cap analog) and physiological (seed yield, biomass) functions. A full assessment of this genotype's resistance to biotic or abiotic stresses as well as of any changes in translatome could allow to analyse more subtle changes in the plant physiology that could be associated with this new allele.
By being fully functional, eIF4E1 R allows combination of the resistance spectra associated with both loss-of-function alleles in eIF4E1 and eIFiso4E (i.e. the resistance to ClYVV associated with eif4e1 KO and the resistance to TuMV, PPV and LMV associated with eifiso4e KO ). Such association had been infeasible so far because of the lethality of the double mutant Patrick et al., 2014). Moreover, the same genetic combination involving eIF4E1 R allowed us to extend the Arabidopsis resistance to WMV, a potyvirus for which no 4E-mediated resistance has been identified so far. Interestingly, this resistance is more efficient than the one previously described based on the plant host factor chloroplastic PHOSPHOGLYCERATE KINASE 2 (Ouibrahim et al., 2014). This shows that WMV is able to use the two main translation initiation factors in Arabidopsis indifferently, namely eIF4E1 and eIFiso4E isoforms. This is a very similar situation to the one that has been described for susceptibility to PepVMV and ChiVMV in pepper, where resistance relies on both a Ca-eIF4E1-pvr2 functional resistance allele and a natural Ca-eIFiso4E-pvr6 KO allele (Hwang et al., 2009;Ruffel et al., 2006). We therefore suspect that the ability of certain viral species to recruit several isoforms could be more frequent than expected, and by generalization, the genericity of 4E factors in susceptibility to ssRNA+ viruses could be more universal than previously thought (Figure 9).
Finally, we described earlier that TuMV can overcome eIFiso4E loss-of-function through either one of two nonsynonymous mutations in the viral VPg coding sequence (Gallois et al., 2010). We hypothesized that because eIFiso4E was no longer available, the viral mutations associated with resistance breaking may extend the VPg's ability to recruit new host factors, possibly other translation initiation factors. Indeed, we show here that eif4e1 KO eifiso4e KO eIF4E1 R lines were fully resistant to both RB-TuMV isolates. This means that E116Q and N163Y mutations in the VPg are two independent paths to the same RB mechanism, allowing the virus to recruit eIF4E1 as well as eIFiso4E in the plant. eIF4E-based resistance breaking has often been associated with the recruitment of the same eIF4E factor, as shown by the pepper/potyvirus co-evolution process (Charron et al., 2008). This work presents for the first time RB events associated with the recruitment of another 4E isoform. This allows us to anticipate that resistance based on eIF4E loss-of-function could be easily overcome, leading to serious consequences for the design of durable genetic resistance (Bastet et al., 2017). In conclusion, the genetic combination using the synthetic allele designed in this study is associated with a full resistance to all seven potyviruses tested. Overall, our results show that combining mutations in several 4E translation initiation factors is highly likely to expand the resistance spectrum to other potyviruses, but might also extend the resistance durability.
More generally, translation initiation factors were also found in crops and in Arabidopsis to be involved in resistance to ssRNA+ viruses with different genome organizations to potyviruses . This includes the resistance to viruses from genera such as Polerovirus (family Luteoviridae) and Sobemovirus (family Tombusviridae), harbouring a much shorter VPg (around 90 AA in length) displaying little homology with the Potyvirus VPg (around 190 AA). This also extends to Carmovirus (family Tombusviridae) members, including Melon necrotic mosaic virus, whose genomes are uncapped (Jiang and Lalibert e, 2011). To demonstrate whether the efficiency of the eIF4E1 R allele, specifically designed as a resistance allele to potyviruses, could be extended to other ssRNA+ viruses, we challenged the eIF4E1 R -complemented plants with polerovirus species (TuYV, BMYV, BWYV-USA), nepoviruses (GFLV-F13, GFLV-GHu, ArMV) and carmovirus (TCV). We did not observe any resistance to TCV, nor to the three nepoviruses assayed. However, we found a complete resistance to one polerovirus species (BWYV-USA) out of the three tested. It is worth noting that the same polerovirus species have been previously used to challenge loss-of-function mutants affecting translation initiation factors in Arabidopsis, with the eif4e1 KO mutant showing partial resistance to BMYV and BWYV-USA (Reinbold et al., 2013). Significantly, in our study, eIF4E1 R complementation restores full susceptibility to BMYV while it triggers full resistance to BWYV-USA, which could be of interest for sugarbeet breeding programmes. Recently, potyvirus VPg has been suggested to interact with host eIF4E through a conserved noncanonical 4E-binding domain (Miras et al., 2017) which interestingly does not seem to be conserved in VPg from other groups of ssRNA+ viruses. Other viruses are likely to recruit eIF4E through different structural binding domains and would therefore only be marginally affected by potyvirus-specific nonsynonymous AA changes in eIF4E regions I and II as the ones introduced in the synthetic eIF4E1 R allele . For example, the Arabidopsis eIFiso4E binds the viral protease rather than the VPg of the nepovirus Tomato ringspot virus (L eonard et al., 2002) and the carmovirus Melon necrotic spot virus (MNSV) binds eIF4E directly through an RNA motif at the 3 0 end of the viral genome (Truniger et al., 2008). Consistent with this, the melon eIF4E natural allele nsv associated with resistance to MNSV displays a very idiosyncratic AA change at the C-terminal end of the eIF4E protein sequence (Nieto et al., 2006). This suggests that, to adjust to the different interaction patterns developed by these viruses, specific additional mutations-possibly outside regions I and II-should be characterized and introduced into eIF4E or eIFiso4E protein sequences, to tailor specific resistances. Besides, ssRNA+ viruses can also use other translation initiation factors. For example, sobemovirus Rice yellow mottle virus in rice and polerovirus TuYV in Arabidopsis both use the large scaffolding translation initiation factors eIFiso4G through the binding of their VPgs (H ebrard et al., 2010;Reinbold et al., 2013), suggesting that these factors may also be designed to trigger specific resistance.
The innovative approaches developed here can be fully translated to crops in order to develop amenable resistances to control losses associated with virus infection using environment friendly strategies. Mostly, by targeting simultaneously several translation initiation factors 4E and by combining lossof-function alleles with modified functional alleles, our approach is likely to trigger resistance without impairing crop agronomical performance, a major requisite for plant breeding. By introducing six independent mutations in the eIF4E1 coding sequence, similar to those naturally present in pea, the eIF4E1 R synthetic allele confers a very robust resistance to potyviruses. We show that mutations associated with resistance can be easily transferred from one species to another to design alleles based on the endogenous relevant genes, as previously shown in Solanaceae (Cavatorta et al., 2011;Kang et al., 2007). Furthermore, by expressing the modified eIF4E1 R under the control of its own regulatory sequences, the complementation of a knockout allele can be easily performed  without the use of foreign DNA sequences such as strong constitutive promoters and terminators. This approach, also called cis-genesis, constitutes a major issue for public acceptance (Ilardi and Tavazza, 2015;Schouten et al., 2006) and a strategy similar to the one developed in this study in Arabidopsis could be easily transposed to any crops. This can also be carried out by new breeding techniques (NBT) such as CRISPR/Cas9 technology particularly with the use of Cas9-cytidine deaminase fusion, allowing precise base substitution in situ in the genome of plants of agronomical interest (Komor et al., 2016;Murovec et al., 2017;Shimatani et al., 2017). Genetic diversity could then be optimally exploited in a trans-species manner to provide new engineered sources of resistance.

Plant materials and growth conditions
Arabidopsis thaliana Columbia-0 (Col) plants were used in this study as a wild-type control for all tests. A. thaliana accessions Figure 7 eif4e1 KO eifiso4e KO eIF4E1 R plant resistance spectrum extends to resistance-breaking TuMV isolates. Controls and eif4e1 KO eifiso4e KO eIF4E1 R lines were challenged with either GFP-tagged TuMV, GFP-TuMV-E116Q (RB) or GFP-TuMV-N163Y (RB). Viral accumulation was assessed using GFPcam (PSI) camera-imaging (a). Fluorescence intensity is shown in false colour from blue (low) to red (high intensity) with the plant pot reflecting light, allowing visualization of the outline of the plant position. (b) Detection of TuMV mRNA by RT-PCR amplification of the VPg coding sequence for both TuMV-GFP and TuMV-N163Y-GFP 21 dpi in the noninoculated leaves. ADENINE PHOSPHORIBOSYL TRANSFERASE 1 (APT1) gene was used as a control for RNA extraction. Landsberg erecta (Ler) and Cap Verde island (Cvi) were also used as susceptible and resistant controls, respectively, for WMV infection according to (Ouibrahim et al., 2014). eif4e1 KO and eifiso4e KO lines are, respectively, homozygous for the T-DNA insertion in eIF4E1 (At1g18040; SALK_145583) and the transposon dSpm insertion in eIFiso4E (At5g35620; Duprat et al., 2002) in a Col background. For genetic crosses, immature flowers were emasculated and cross-pollinated manually.
For growth on plates, seeds were surface-sterilized for 10 min in 95% ethanol with 0.1% Tween 20 and plated on Murashige and Skoog (MS) medium, supplemented if necessary with 10 mg/L hygromycin B. Seedlings were transferred to soil after 2 weeks. For nepovirus, polerovirus, TCV, PPV and LMV resistance assays, seeds were directly sowed on soil and individually transferred 2 weeks later.
Plants were grown in a culture chamber at 20-24°C with a cycle of 16-h light (fluorescent light at 100 lmol photon/m 2 /s) and 8-h dark for the flowering assay and nepovirus resistance assay, with a cycle of 8-h light and 16-h dark for ClYVV, TuMV, WMV, RB-TuMV, LMV, PPV, and TCV and 10-h light and 14-h dark for TuYV, BMYV-2itb and BWYV-USA resistance assays. For the flowering and fresh/dry weight assays, the different genotypes of plants were randomized in the culture chamber.

Plasmid construction and plant transformation
A 3533-bp genomic -At4g18040-eIF4E1 fragment (spanning 1500 bp of the promoter region and 150 bp of the 3 0 UTR) was amplified on Columbia genomic DNA using primers Z4148-F and Z4148-R (Table S1) and subcloned into pDONR207 using BP gateway recombination (Invitrogen), resulting in plasmid pJL631. Site-directed mutagenesis associated with the following AA changes-W69L T80D S81D S84A G114R and N176K-were iteratively introduced with the QuikChange II Site-Directed Mutagenesis Kit (Stratagene) using specific primers. All primers are listed in Table S1. All constructs were sequence-checked. The resulting eIF4E1 R genomic construct was then cloned into the binary vector pMDC099 (Curtis and Grossniklaus, 2003) using gateway LR recombination and subsequently introduced in Arabidopsis genome by Floral Dip agrotransformation (Clough and Bent, 1998). Transformants were selected on MS plates supplemented with 10 mg/L hygromycin B.
The LMV-AFVAR1 isolate, differing by a single amino acid change in the CP N-terminal region from its LMV-AF199 progenitor (Decroocq et al., 2009), was selected for its capacity to systemically infect Columbia-0. This virus was inoculated mechanically onto Arabidopsis rosette leaves 6 weeks after sowing. Virus accumulation in apical noninoculated leaves was analysed by DAS-ELISA assay with polyclonal antibodies anti-LMV (German-Retana et al., 2008) 21 days postinoculation (dpi).
RB-TuMV plasmids were constructed from pCambia TuMV-GFP, a gift from Jean-Franc ßois Lalibert e (Beauchemin et al., 2005), by introducing into the VPg coding sequence the E116Q and N163Y mutations, respectively, by PCR-based mutagenesis using the QuikChange â II XL Site-Directed Mutagenesis kit (Agilent, Santa Clara, CA) with primers Z4653-F/Z4653-R for E116Q and Z4654-F/Z4654-R for N163Y. Presence of the mutations was checked by sequencing (Genoscreen, Lille, France), and the resistance-breaking properties of these variants were assessed on eifiso4e KO plants. The plasmids were transformed into the C58S pMP90 agrobacterium strain and agro-inoculated in leaves. GFP accumulation in plants was imaged 3 weeks after inoculation using a closed fluorometric camera FluorCam FC 800-C/1010-GFP (Photon System Instruments, Drasov, Czech Republic) equipped with a GFP filter. Fluorescence is represented in false colours.
Polerovirus inoculations were carried out as previously described (Reinbold et al., 2013). Briefly, BMYV (isolate 2ITB), BWYV-USA and TuYV were inoculated using Myzus persicae membrane-fed on purified virus for 24 h. Ten viruliferous aphids were then transferred onto each test plant for an inoculation period of 4 days. Detection of viral infection was performed by DAS-ELISA (Loewe, Kronach, Germany) 21 dpi.
TCV isolate M (Oh et al., 1995) inoculation with infectious clone contained in pBIN61 was carried out by agro-infiltration on 5-week-old plants. Infection was detected by symptom development 23 dpi.
All results presented are mean values from at least 15 independent plants per genotype, unless indicated otherwise, and error bars represent standard errors. The threshold for susceptibility is represented by a line on each graph and refers to absorbance value at 405 nm in ELISA equal to three times the mean value for healthy controls.
A 575-bp fragment covering the central region of the TuMV VPg coding sequence was similarly amplified by RT-PCR from inoculated plants with primers Z4735-F/Z4735-R.

Antibodies and Western blot analysis
Total protein extracts in Laemmli buffer from 4-week-old plantlets were used for Western blot analysis. Electrophoresis by SDS-PAGE was performed on equal amount of protein extracts, and proteins were transferred to Amersham TM Protran â Premium nitrocellulose membranes (GE Healthcare, Chicago, Illinois, United States). To ensure the correct transfer of proteins and equal loading, membranes were stained with Ponceau S solution (Sigma-Aldrich, St Louis, MO), before being incubated with anti-actin or anti-eIF4E1 antibodies at 1/5000 and 1/2000 dilution, respectively. Membranes were then incubated with goat horseradish peroxidase-labelled anti-mouse serum for detection of actin (1/5000 dilution) (Sigma-Aldrich, St Louis, MO) and goat horseradish peroxidase-labelled anti-rabbit serum for detection of eIF4E1 (1/2000 dilution) (Sigma-Aldrich, St Louis, MO). Horseradish peroxidase activity detection was carried out using a LumiGLO Reserve chemiluminescent substrate kit (KPL, www.kpl.com) and X-OMAT LS films (Kodak, New York).
The polyclonal serum targeted against the full Arabidopsis thaliana eIF4E1 protein was produced by Proteogenix (Oberhausbergen, France) using the method previously described .

Cap-binding assay
Total protein extracts obtained by grinding 4-week-old plants were suspended in a 40 mM HEPES/KOH pH 7.6, 100 mM KCl, 1 mM dithiothreitol, 10% glycerol, 1% phenylmethanesulphonyl fluoride and protease inhibitor cocktail (Roche, Basel, Switzerland) resuspension buffer. Samples were centrifuged for 10 min at 15000 g, and the supernatant fractions (corresponding to INPUT fraction) were added to 50 lL of 7 methyl-GTP sepharose beads (GE Healthcare, Chicago, Illinois, United States) and incubated overnight at 4°C. After four washes with the resuspension buffer detailed above, samples were centrifuged 1 min at 15 000 g at 4°C. Pellets were resuspended in 30 lL of Laemmli buffer before being boiled for 5 min to elute proteins bound to the cap analog (corresponding to OUTPUT fraction). Protein detection was carried out by Western blot.

Dry/fresh weight and fertility rate measurements
Three-week-old plants (20 per genotype) were cut, and aerial parts were weighed to calculate fresh weights. Each plant was then dried for 24 h in a 100°C heating chamber before evaluating the dry weight.
Total seed production of 10 plants per genotype was assayed by collecting and weighing all the seeds of each plant. A pool of 100 seeds per genotype was also weighed to ensure that there was no difference in seed average weight between genotypes.

Statistical analyses
Kruskal-Wallis statistical tests were performed using the pgirmess package on the free software R (http://www.r-project.org/).

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Figure S1 Three-dimensional predicted structure of Pisum sativum eIF4E proteins. Figure S2 eIF4E1 expression analysis in controls and complemented lines. Figure S3 Genotyping of eif4e1 KO eifiso4e KO eIF4E1 R plants. Figure S4 Phenotype of the eif4e1 KO eifiso4e KO eIF4E1 R plants. Figure S5 eif4e1 KO eifiso4e KO eIF4E1 R plants resistance spectrum extends to LMV and PPV. Figure S6 The Cvi accession is partly resistant to WMV, but eif4e1 KO and eifiso4e KO single mutants are both susceptible. Figure S7 Low accumulation of resistance-breaking TuMV isolates in eifiso4e KO plants as detected by DAS-ELISA Figure S8 eif4e1 KO eifiso4e KO eIF4E1 R plants resistance spectrum does not extend to carmovirus TCV. Figure S9 eif4e1 KO eifiso4e KO eIF4E1 R plants resistance spectrum does not extend to nepoviruses.