When a knockout is an Achilles’ heel: Resistance to one potyvirus species triggers hypersusceptibility to another one in Arabidopsis thaliana

Abstract The translation initiation factors 4E are a small family of major susceptibility factors to potyviruses. It has been suggested that knocking out these genes could provide genetic resistance in crops when natural resistance alleles, which encode functional eIF4E proteins, are not available. Here, using the well‐characterized Arabidopsis thaliana–potyvirus pathosystem, we evaluate the resistance spectrum of plants knocked out for eIF4E1, the susceptibility factor to clover yellow vein virus (ClYVV). We show that besides resistance to ClYVV, the eIF4E1 loss of function is associated with hypersusceptibility to turnip mosaic virus (TuMV), a potyvirus known to rely on the paralog host factor eIFiso4E. On TuMV infection, plants knocked out for eIF4E1 display striking developmental defects such as early senescence and primordia development stoppage. This phenotype is coupled with a strong TuMV overaccumulation throughout the plant, while remarkably the levels of the viral target eIFiso4E remain uninfluenced. Our data suggest that this hypersusceptibility cannot be explained by virus evolution leading to a gain of TuMV aggressiveness. Furthermore, we report that a functional eIF4E1 resistance allele engineered by CRISPR/Cas9 base‐editing technology successfully circumvents the increase of TuMV susceptibility conditioned by eIF4E1 disruption. These findings in Arabidopsis add to several previous findings in crops suggesting that resistance based on knocking out eIF4E factors should be avoided in plant breeding, as it could also expose the plant to the severe threat of potyviruses able to recruit alternative eIF4E copies. At the same time, it provides a simple model that can help understanding of the homeostasis among eIF4E proteins in the plant cell and what makes them available to potyviruses.


| INTRODUC TI ON
Plant pathogens are responsible for significant agricultural and economic losses worldwide . Genetic resistance is an important trait looked for in crop improvement programmes as it provides an effective and environmentally safe strategy to control pathogens in fields. One important challenge of breeding is to be able to translate known resistance mechanisms from species where they have been characterized to species where they have not yet been. This is particularly important for orphan crops to speed up breeding and make up for the lack of investment (Jacob et al., 2018).
Dominant genetic resistances based on R genes, relying on response to the detection of pathogen avirulence factors, are very well developed and can sometimes be transferred to other species (Lacombe et al., 2010; for review Rodriguez-Moreno et al., 2017), but often with limited success (van Wersch et al., 2020). On the contrary, it is expected that resistances by loss of susceptibility, based on the absence or incompatibility of host susceptibility factors required for pathogen proliferation, are possibly easier to translate among crops as they are supposed to target conserved mechanisms. For example, resistance to powdery mildew associated with Mildew resistance locus o (Mlo) genes characterized in barley (Hordeum vulgare) can be set up in tomato (Solanum lycopersicum) or apple (Malus domestica) (Kusch & Panstruga, 2017). Resistance to Xanthomonas species obtained by regulating the expression of the glucose transporter SWEET genes reported in rice (Oryza sativa) (Oliva et al., 2019) is conserved in cassava (Manihot esculenta) (Cohn et al., 2014) and cotton (Gossypium spp.) (Cox et al., 2017). Finally, the role of translation initiation factors eIF4E in susceptibility to many single-stranded positive-sense RNA (ssRNA+) viruses in crops has been shown to extend to caliciviruses affecting animals and humans (Hosmillo et al., 2014;Robaglia & Caranta, 2006). However, translating resistance mechanisms by loss of susceptibility can be hindered by gene redundancy among susceptibility factors belonging to small multigene families. Susceptibility factors having overlapping functions could offer diversified baits to the pathogen to hijack, and thus provide new routes of susceptibility or resistance-breaking opportunities. This is exemplified by resistance associated with translation initiation factors eIF4E to potyviruses in crops (Gallois et al., 2018).
Translation initiation factors eIF4E and their paralogous counterparts eIFiso4E are a major source of resistance to members of the Potyviridae family as well as related ssRNA+ viruses and constitute traits of great agronomical importance (Robaglia & Caranta, 2006;Wang & Krishnaswamy, 2012). Since their discovery in pepper (Capsicum annuum) 20 years ago (Ruffel et al., 2002), they have become a model for the implementation of resistance through both classical breeding and biotechnology, using allele mining, and diverse ways of gene inactivation (RNAi, TILLING, CRISPR) (Schmitt-Keichinger, 2019). Natural eIF4E resistance alleles are characterized in many crops by the presence of a small set of nonsynonymous mutations associated with amino acid changes in specific regions of the eIF4E protein. As a result, the eIF4E resistance alleles encode proteins that are still functional in their translation initiation function, but can no longer be recruited by the virus, leading to resistance.
However, such resistance alleles are not present in the natural variability of some crops. For instance, only dominant resistance genes have been characterized in potato (Solanum tuberosum) in which infection by potato virus Y (PVY) induces a strong tuber necrosis phenotype (Glais et al., 2015). A similar situation occurs in other economically important crops such as papaya (Carica papaya) and cassava challenged by papaya ringspot virus (PRSV) and cassava brown streak virus (CBSV), respectively (Bart & Taylor, 2017;Gonsalves, 1998). When functional eIF4E resistance alleles are not available, biotechnological approaches to transfer resistance can be carried out by knocking down or out the susceptibility factors. Although associated in some cases with efficient resistance, the resulting plants could be affected in their development because of the important physiological role of eIF4E, and be prone to limited resistance spectrum or to resistance breaking because of gene redundancy (Bastet et al., 2017). Overall, we posit that, taking opportunity of the genome editing tools associated with CRISPR-Cas9, a good strategy would be to copy in eIF4E mutations that have been selected by natural variation rather than knocking out the gene. Indeed, using Arabidopsis thaliana as a model, we showed that such mutations could be copied across species, namely from pea (Pisum sativum) eIF4E resistance alleles to the Arabidopsis eIF4E1, effectively transferring resistance (Bastet et al., 2018). We further showed that the more relevant mutations could be inserted in transgene-free plants by genome base editing, allowing resistance at no yield cost (Bastet et al., 2019).
Because potyviruses are known to selectively require different eIF4E paralogs to establish infection (eIF4E, eIFiso4E, or the atypical nCBP factor) (Duprat et al., 2002;Gomez et al., 2019;Nicaise et al., 2007;Sato et al., 2005), we reasoned that reshuffling this selectivity by inactivating one paralog could affect global plant susceptibility. In the present study, we took advantage of the A. thaliana-potyvirus pathosystem where clover yellow vein virus (ClYVV) recruits eIF4E1 while turnip mosaic virus (TuMV) uses the counterpart eIFiso4E to accomplish their infection cycles. We show that knocking out eIF4E1, to generate resistance to ClYVV, is at the same time responsible for a remarkable increase in susceptibility towards TuMV. This aggravated phenotype to TuMV infection is associated with an increase in virus load throughout the plant and can be averted by the deployment of a functional eIF4E1 resistance allele. We propose that because potyviruses specifically hijack different eIF4E factors, the strategy of developing resistance by eIF4E loss of function could be jeopardized by existing or emerging viruses able to recruit the remaining paralogs in the plant.

| Loss of function in eIF4E1 is associated with an increased severity of symptoms on infection by TuMV
To assess the resistance generated by disruption of genes encoding translation initiation factors 4E, we compared the phenotypic responses on infection with TuMV-GFP UK1 of three Arabidopsis genotypes: a wild-type susceptible accession Columbia-0, a line knocked out for eIF4E1 (eif4e1 KO ), resistant to ClYVV, and a line knocked out for eIFiso4E (eifiso4e KO ), resistant to TuMV. The virusinduced symptoms were recorded at 14 and 21 days postinoculation (dpi) (Figure 1a). Infection of wild-type plants resulted in typical TuMV symptoms including leaf distortion and stunted growth. On the contrary, eifiso4e KO plants did not exhibit any disease symptoms at both postinoculation stages, consistent with their previously described full resistance towards TuMV (Duprat et al., 2002).
Interestingly, TuMV-induced disease symptoms in the eif4e1 KO line were clearly more pronounced in comparison to wild-type plants.
Two main phenotypic patterns of this enhanced susceptibility were visible at the leaves formed prior to and after the viral inoculation step in the eif4e1 KO mutant (Figure 1a, leaves formed prior to inoculation are labelled by white arrowheads while leaves formed after the inoculation are labelled by black arrowheads).
The leaves already formed before the inoculation step did not show any visible size or morphology alterations between wild-type and eif4e1 KO plants. However, as early as 14 dpi, a precocious yellowing was observed in eif4e1 KO plants (Figure 1a, white arrowheads).
Imaging total chlorophyll fluorescence suggested a reduction of chlorophyll levels, a trait characteristic of leaf senescence and typically observed at later developmental stages (Pružinská et al., 2005).
The tissue senescence in eif4e1 KO plants was even more striking at Altogether, we conclude that the Arabidopsis eif4e1 KO mutant displays strong disease symptoms and increased developmental malformations on TuMV infection.

| TuMV accumulation is favoured in the eif4e1 KO mutant
The increased severity of the symptoms induced by TuMV in eif-4e1 KO plants prompted us to explore whether this phenotype could be associated with a higher viral load. To compare the accumulation of TuMV, we took advantage of the GFP coding sequence inserted in the TuMV UK1 clone (Beauchemin et al., 2005). Because the signal of GFP fused to TuMV can be used as a proxy for viral accumulation (Abdelkefi et al., 2018), fluorometric analysis with a GFP-camera was applied as a nondestructive assessment of the viral infection at 14 and 21 dpi. Phenotypic response to TuMV infection is known to be isolatedependent as different TuMV isolates induce contrasting disease symptoms (Sanchez et al., 2015). To test if the developmental defects and the higher virus load observed in the eif4e1 KO mutant occur specifically in response to infection with the TuMV-GFP UK1 isolate, we performed an inoculation test with the CDN1 isolate of TuMV (Duprat et al., 2002). Once again, the eif4e1 KO mutant was more strongly affected than wild-type plants ( Figure 2a). As observed with TuMV-GFP UK1 inoculation, eif4e1 KO plants infected by TuMV CDN1 displayed an approximately 33% decrease in weight (mean fresh rosette weight = 0.43) compared to wild-type plants (mean fresh rosette weight = 0.65; Figure 2b). Importantly, the accumulation of TuMV CDN1 was about three times higher in eif4e1 KO than in wild-type F I G U R E 3 Complementation of eIF4E1 loss of function suppresses the enhanced susceptibility towards turnip mosaic virus (TuMV). (a) Phenotypic comparison of representative plants upon TuMV-GFP UK1 infection at 17 days postinoculation (dpi). Photographs were taken under natural light conditions (left panel) and under wavelengths specific for chlorophyll excitation (middle panel) or green fluorescent protein (GFP) excitation (right panel) by using GFP Camera (PSI) fluorescence imaging. Fluorescence is represented by false colours ranging from blue (low intensity) to red (high intensity). Two independently obtained eif4e1 KO ;eIF4E1 complemented lines were used in the analyses. (b) Rosette fresh weight analysis of plants inoculated with TuMV-GFP UK1 at 17 dpi. The aerial part was weighed for five wild-type mockinoculated plants and at least seven TuMV-GFP UK1-inoculated plants of each genotype. (c) Accumulation analysis of TuMV-GFP UK1 at 17 dpi. Viral accumulation was detected by double antibody sandwich-ELISA for TuMV coat protein (CP) on five wild-type mock-inoculated and at least seven TuMV-GFP UK1-inoculated plants of each genotype. Different letters depict significantly different groups identified by Kruskal-Wallis statistical tests at p < .05 inoculated plants ( Figure 2c). Therefore, the enhancement of viral accumulation extends to the TuMV CDN1 isolate, suggesting that the over-accumulation of TuMV in eif4e1 KO is consistent and not isolatedependent, although it would be of interest to test more TuMV isolates.
Taken together, these results suggest that the accumulation of TuMV is favoured in the absence of eIF4E1.

| Complementation of eIF4E1 loss of function restores wild-type susceptibility to TuMV
To confirm that the eif4e1 KO severe developmental phenotype is caused by the lack of eIF4E1, we checked the effect of TuMV infection on eif4e1 KO plants complemented by a full-length genomic eIF4E1 under the control of its native promoter and 3′ untranslated region (UTR) (Bastet et al., 2018). Two independently obtained Hence, the complementation of eif4e1 KO function is sufficient to revert to a wild-type susceptibility to TuMV, showing that disruption of eIF4E1 is the sole cause for the increased susceptibility and higher virus load in the eif4e1 KO mutant.

| TuMV overaccumulation is not associated with a change in eIFiso4E protein accumulation in the eif4e1 KO mutant
Previously, it was demonstrated that an eIFiso4E knockout allele causes an enhanced accumulation of eIF4E1 protein in Arabidopsis, suggesting the existence of a regulatory mechanism among eIF4E family members (Duprat et al., 2002). We hypothesized that the difference in viral accumulation in eif4e1 KO could be due to a difference in the accumulation of eIF4E factors. Hence, a greater amount of eIFiso4E in eif4e1 KO lines could provide a more abundant host factor for TuMV and possibly favour viral accumulation.
To address this question, we compared the accumulation of eIF4E1 and eIFiso4E in the different Arabidopsis genotypes in the presence of TuMV infection. Western blot analysis revealed higher eIF4E1 protein levels in eifiso4e KO (Figure 4a) corresponding to an approximately 50% increase as determined by band intensity quantification (see Figure S3a,b for short-exposure images and band intensity quantification), confirming the results demonstrated by previous studies (Duprat et al., 2002). However, no changes in eI-Fiso4E accumulation could be detected between wild-type and eif-4e1 KO (Figures 4a and S3c).
To examine whether a subtler variation in eIFiso4E levels occurs in eif4e1 KO , we performed an m 7 GTP pull-down on protein extracts obtained from wild-type and eif4e1 KO mutant plants. This allows selective purification of the proteins associated with the cap-binding complexes in the total plant extract, including eIF4E and eIFiso4E.
Western blot analysis on m 7 GTP-bound fractions (OUTPUT) confirmed the presence of eIFiso4E proteins but no change in its accumulation pattern was observed in the eif4e1 KO  These results rule out that a significant increase in eIFiso4E protein level could directly explain the overaccumulation of TuMV in eif4e1 KO .

| Virus evolution is not responsible for the enhanced susceptibility of the eif4e1 KO mutant towards TuMV
The viral genome-linked protein (VPg) of potyviruses has been identified as a key virulence determinant of potyvirus infection.
The amino acid substitutions E116Q and N163Y in the central and C-terminal part of TuMV VPg are known to expand the target range of TuMV in Arabidopsis by allowing it to recruit the eIF4E1 factor and consequently to overcome resistance mediated by eIFiso4E loss of function (Bastet et al., 2018;Gallois et al., 2010). In addition to this  (Table S1). Furthermore, identical VPg polymorphisms resulting in the nonsynonymous amino acid substitutions N120D and S114L were present in the viral progeny obtained from both wild-type or eif4e1 KO inoculated plants, suggesting no preferential selection caused by the plant genotype (Figure 5a, green and blue vertical lines). Importantly, none of the nucleotide changes in the VPg sequenced from wild-type and eif4e1 KO infected plants corresponded to the E116Q and N163Y substitutions associated with resistance breaking (Bastet et al., 2018;Gallois et al., 2010). Altogether, these results suggest that TuMV multiplied in the eif4e1 KO mutant did not acquire distinct VPg mutations compared to TuMV multiplied in wild-type plants.
To determine whether propagation of TuMV in eif4e1 KO results in an increase of aggressiveness in a VPg-independent manner, extracts from either wild-type or eif4e1 KO infected plants were back-inoculated to the same two genotypes and plant susceptibility was scored at 21 dpi. Our results show that whether the TuMV back-inoculum originated from wild-type or from eif4e1 KO plants, severe disease symptoms were observed only in eif4e1 KO plants ( Figure 5b). Accordingly, the strong weight reduction and increased viral loads, characteristic of the TuMV hypersusceptibility, were found only in eif4e1 KO back-inoculated plants (Figure 5b,c). Therefore, the susceptibility outcome towards TuMV depends on the genotype of the back-inoculated plants, regardless of the origin of the inoculum.
These results demonstrate that no selection pressure is imposed by the absence of eIF4E1, ruling out the hypothesis that the hypersusceptibility of eif4e1 KO is caused by direct evolution of the virus affecting its aggressiveness. Moreover, back-inoculation further confirms the high TuMV accumulation and increased disease symptoms associated with eIF4E1 loss of function.

| An edited eIF4E1 N176K resistance allele, encoding a functional protein, does not trigger an increased susceptibility towards TuMV
While generating KO in eIF4E1 is efficient to drive resistance to ClYVV (Bastet et al., 2018;Sato et al., 2005) that relies on this translation initiation factor in Arabidopsis, we show here that such an approach comes at the price of an increased susceptibility to a virus targeting the eIFiso4E paralog. As previously suggested (Bastet et al., 2017), we reasoned that an alternative option would be to mimic functional resistance eIF4E1 alleles rather than knocking them out.
In our previous work, we used the CRISPR-nCas9-cytidine deaminase method to convert the Arabidopsis eIF4E1 susceptibility allele to ClYVV into a functional resistance allele in a transgene-free manner (Bastet et al., 2019). The protein encoded by this allele contains the single amino acid substitution N176K, imitating a mutation that occurs naturally in a pea eIF4E allele associated with resistance to potyviruses. When modified in the wild-type Arabidopsis eIF4E1 gene, it confers complete resistance to two distinct ClYVV isolates F I G U R E 4 The eIFiso4E protein levels are not increased in the eif4e1 KO mutant. (a) Western blot analysis on total protein extracts from mock-inoculated wild-type and TuMV-GFP UK1-inoculated plants at 21 days postinoculation (dpi). eIF4E1 and eIFiso4E protein accumulation was analysed by western blot on total plant protein extracts using specific antibodies. Equal loading was checked by western blot using anti-actin antibodies. (b) Western blot analysis on m 7 GTP-purified protein extracts from mock-or TuMV-GFP UK1-inoculated wild-type and eif4e1 KO plants at 21 dpi. Total protein extracts (INPUT) were pulled down with γ-aminophenyl-m 7 GTP (C 10 -spacer)-agarose beads (OUTPUT). After purification, the output fraction was analysed by western blot using anti-eIFiso4E antibody while equal loading control was checked on total protein extracts (INPUT) by western blot using anti-actin antibodies. Each lane represents immunoblotted protein extracts obtained from a single independent plant F I G U R E 5 The increased susceptibility of the eif4e1 KO mutant towards turnip mosaic virus (TuMV) is not caused by virus evolution. (a) A graphic representation of nonfixed amino acid substitutions at TuMV VPg following TuMV-GFP UK1 multiplication on wild-type and eif4e1 KO plants. Total RNA extracted from six wild-type and 10 eif4e1 KO TuMV-GFP UK1-inoculated plants was reverse transcribed and the VPg coding region was sequenced. Each box represents a VPg sequence obtained from an independent wild-type or eif4e1 KO TuMV-GFP UK1inoculated plant. Vertical lines indicate the positions of the detected nonfixed amino acid substitutions. The VPg polymorphisms N120D and S114L observed in both wild-type and eif4e1 KO are represented in green and blue, respectively. The positions of amino acid substitutions associated with eifiso4e KO resistance breaking (Gallois et al., 2010)  TuMV-GFP UK1, plants harbouring the eIF4E1 N176K allele showed a level of GFP intensity similar to wild-type rather than the enhanced signal imaged in eif4e1 KO plants (Figure 6a). In accordance, the accumulation of TuMV CP was lower in the eIF4E1 N176K lines than in the eif4e1 KO plants (Figure 6c). This suggests that the eIF-4E1 N176K resistance allele does not trigger TuMV overaccumulation as the knockout of eIF4E1 does. Similar results were observed at a later stage of infection such as 21 dpi ( Figure S4).
These results demonstrate that a functional resistance allele tailored with CRISPR/Cas9 base-editing technology is able to avoid the increased susceptibility to TuMV imposed by the knockout in eIF4E1, while providing resistance to ClYVV. Hence, it appears that tailored functional eIF4E alleles allow the generation of resistance while circumventing drawbacks associated with knocking out these genes.

| D ISCUSS I ON
Knocking out genes encoding factors required for the infection cycle of potyviruses appears to be a straightforward approach to achieve resistance (Mäkinen, 2020;Pyott et al., 2020). Here, our results demonstrate that such a strategy can be double edged as besides providing resistance to ClYVV, knocking out eIF4E1 makes plants especially vulnerable to infection by TuMV. As it stands, this is the first time such an Achilles' heel has been experimentally shown for eIF4E-based resistance to viruses in plants. Our report reinforces the idea that resistance based on disrupting eIF4E alleles should be avoided (Bastet et al., 2017)  counterparts) (Browning & Bailey-Serres, 2015). The disruption of a single gene coding for the cap-binding 4E or the scaffolding 4G subunits is known to impact the accumulation of their counterparts, suggesting the existence of a feedback mechanism that regulates the homeostasis of translation initiation factors in the plant cell (Combe et al., 2005;Duprat et al., 2002;Lellis et al., 2010). In the light of these observations, a disruption of eIF4E homeostasis due to the absence of eIF4E1 could lead to a compensatory increase of eIFiso4E accumulation. Higher eIFiso4E levels could supply a more abundant host factor for TuMV accumulation, and therefore explain the enhanced susceptibility of the eif4e1 KO mutant. We ruled out this hypothesis by demonstrating that there was no significant difference of eIFiso4E protein accumulation between eif4e1 KO and wild-type plants (Figures 4 and S3).
However, even if the absence of eIF4E1 does not directly affect the overall accumulation of eIFiso4E, it could still be responsible for a more available eIFiso4E target for TuMV accumulation. It is noteworthy that both the eIF4E cap-binding and the eIF4G scaffolding components of the cap-binding complexes are required for the establishment of potyvirus infection, suggesting that the entire eIF4F/ eIFiso4F complex is recruited during the potyvirus infection cycle (Nicaise et al., 2007). Several studies have put forward the idea that rearrangements in the composition of these complexes could modulate the availability of eIF4E factors for potyviruses. Kang et al. (2007) first proposed that the ectopic expression of the pepper eIF4E resistance allele pvr1 in a susceptible tomato accession could confer potyvirus resistance by saturating the cellular pool of eIF4G proteins with an eIF4E factor that cannot be recruited by the virus, and thus render endogenous eIF4E susceptibility factors inaccessible for viral multiplication. In a similar way, Gauffier et al. (2016) showed that in the wild tomato species Solanum habrochaites, the Our results reveal that besides supplying resistance to viruses, the withdrawal of eIF4E susceptibility factors comes at the expense of an unanticipated increase in susceptibility. Interestingly, a similar trade-off between broad resistance spectrum and resistance efficiency occurs in pepper. In the 'Perennial' cultivar of pepper, a deletion of 82 nucleotides in the pvr6 locus leading to a disruption of an eIFiso4E gene has been shown to provide resistance to chilli veinal mottle virus (ChiVMV) and pepper veinal mottle virus (PVMV) when combined with pvr2 alleles encoding functional eIF4E factors (Caranta, 1997;Moury et al., 2005;Rubio et al., 2008). An unexpected drawback of this gain of resistance was illustrated by Quenouille et al. (2014) who demonstrated that the natural knockout allele pvr6 was also linked to an increased resistance breaking of the pvr2 3 resistance allele by PVY. Moreover, it was found that this higher resistance breakdown frequency correlated strongly with higher PVY loads in pvr6-carrying pepper accessions (Quenouille et al., 2016). Although the direct involvement of eIFiso4E in these events remains to be functionally validated, it appears that in the pepper-PVY pathosystem, eIFiso4E absence could favour viral accumulation in a way reminiscent of how eIF4E1 loss of function triggers TuMV overaccumulation in the current model. These results mirror our findings by suggesting that although supplying resistance, using knockout alleles in breeding programmes could decrease the resistance efficiency towards viruses for which multiplication does not rely on the mutated factor.
Another important setback of resistance mediated by knocking out eIF4E has been exemplified in tomato in which the potyviruses PVY and TEV require the eIF4E1 and eIF4E2 factors for multiplication. While a TILLING-induced disruption in eIF4E1 generated resistance to a single PVY isolate, a naturally occurring functional eIF4E1 counterpart Sh-eIF4E1 PI24 -pot1 enlarged the resistance spectrum to several isolates of PVY and TEV (Gauffier et al., 2016). Indeed, a wide resistance spectrum to these isolates could be achieved by disrupting both eIF4E1 and eIF4E2 factors but had a severe effect on plant development, whereas the functional eIF4E1 PI24 allele did not impose any fitness cost. Similarly, we show that the increased susceptibility towards TuMV governed by eIF4E1 knockout can be avoided by using a functional eIF4E1 N176K allele that retains a wide spectrum resistance to several potyviruses (Bastet et al., 2019) ( Figure 6). These direct comparisons of two strategies illustrate the superiority of functional resistance alleles over knockout eIF4E alleles in the generation of efficient resistance to potyviruses.
In addition to suggesting arguments for the long-term inefficiency of resistance mediated by knocking out eIF4E factors, the cases in tomato and pepper described above raise the question of how regulation between eIF4E could play a role in resistance to potyviruses.
Knocking out eIF4E family members appears to disrupt the tight homeostasis between these proteins and to have direct repercussions on their availability for viruses (Michel et al., 2019). Indeed, the large majority of resistance alleles identified by studying natural variation encode functional proteins and only a very few alleles coding for nonfunctional proteins have been identified, and only for eIFiso4E (for review, see Bastet et al., 2017;Wang & Krishnaswamy, 2012).
It has been suggested that natural eIF4E loss of function alleles are not selected as their inactivation may be associated with adverse developmental defects but also because they could be of limited use for providing resistance to viruses: this is illustrated by the analysis of engineered eIF4E KO alleles especially in tomato and Arabidopsis (Bastet et al., 2017;Gallois et al., 2010;Gauffier et al., 2016). The advent of CRISPR-based gene inactivation will allow such studies to be extended in new species. It will be of major interest to assess the long-term effects on the plant development, resistance spectrum, and resistance durability for recently developed resistances based on eIF4E knockout in cucumber (Cucumis sativa) and cassava, for example (Chandrasekaran et al., 2016;Gomez et al., 2019). An alternative to developing durable and broad-spectrum resistances to potyviruses may, however, reside in the use of naturally occurring functional eIF4E resistance alleles. Novel gene-editing technologies such as CRISPR/Cas9 coupled with Cas9 cytidine and adenine deaminase fusions represent a promising approach for the engineering of resistance by the precise editing of susceptibility alleles in crops in which natural variability is lacking (Mushtaq et al., 2019;Veillet et al., 2020).
In conclusion, our findings reinforce the idea, initially built on observations made in the major crops tomato and pepper, that resistance based on knocking out eIF4E factors should be avoided in plant breeding. Here, results gathered in Arabidopsis suggest that such a strategy could expose the plant to the severe threat of potyviruses able to recruit alternative eIF4E copies. At the same time, it provides a simple model to explore the mechanism behind the availability of eIF4E factors to viruses; understanding this could help to make resistance more efficient in crops. It is important to question whether this observation is restricted to the small multigene families of translation initiation factors or if it can be generalized to other susceptibility factors.

| Plant and virus material
A. thaliana ecotype Columbia-0 (Col-0) was used as the wild-type accession for all experiments. Mutations in the eif4e1 KO and eifiso4e KO homozygous lines were caused by a T-DNA insertion in eIF4E1 (At1g18040; SALK_145583) and a dSpm transposon insertion in eIFi-so4E (At5g35620; Duprat et al., 2002), respectively. Both mutations are in the Columbia-0 (Col-0) background. The complemented eif4e1 KO ; eIF4E1 lines carries a genomic At4g18040 eIF4E1 fragment (spanning 1,500 bp of the promoter region and 150 bp of the 3′ UTR) (Bastet et al., 2018). Two independently obtained T 3 eif4e1 KO ;eIF4E1 lines were used in the analyses. The CRISPR-nCas9-cytidine deaminase-induced eIF4E1 N176K lines were obtained by agrotransformation of Col-0 plants with the pDICAID_nCas9-PmCDA_NptII_eIF4E1 construct as previously reported (Bastet et al., 2019). Two independently obtained transgene-free T 4 CRISPR-induced eIF4E1 N176K lines were used in the analyses. The TuMV CDN1 isolate was used as previously reported (Duprat et al., 2002). For TuMV-GFP UK1 infection, a binary vector was kindly provided by Jean-François Laliberté (Beauchemin et al., 2005).

| Plant culture and virus inoculation
Plant seeds were directly sowed on soil or on Murashige and Skoog

| Fluorometric camera analysis
Monitoring of TuMV-GFP infection was carried out using a closed fluorometric camera FluorCam FC 800-C/1010-GFP (Photon System Instruments) equipped with chlorophyll and GFP filters.  TuMV was added (Agdia). Following 2 hr incubation at 37 °C, the plate was washed as described above and diluted alkaline phosphatase-conjugated secondary antibody (1:200) was added. Plates were incubated at room temperature for 2 hr and after a wash step, p-nitrophenylphosphate substrate was added. Plate reads were carried out by spectrophotometry at 405 nm periodically in the 30 min following substrate addition.

| Total protein extraction and m 7 GTP pulldown assay
For total protein analysis, extracts were prepared by grinding equal amounts of TuMV-GFP UK1 systemically infected leaves in Laemmli buffer and boiling samples for 5 min. Proteins contained in the supernatant were recuperated following centrifugation for 10 min at 15,000 × g. For the m 7 GTP pull-down assay, total protein extracts were prepared by grinding whole mock or TuMV-GFP UK1-inoculated plants in liquid nitrogen and suspending 100 mg homogenized tissue in a binding buffer (40 mM HEPES/KOH pH 7.6, 100 mM KCl, 1 mM dithiothreitol, 10% glycerol, 1% phenylmethanesulphonyl fluoride and 1 × protease inhibitor cocktail [Roche]). After centrifugation for 10 min at 15,000 × g at 4 °C, the protein-containing supernatant fraction (INPUT fraction) was recovered and incubated at 4 °C overnight with 50 µl of prewashed immobilized γ-aminophenyl-m 7 GTP (C 10 -spacer)-agarose beads (Jena Bioscience). To remove any unbound proteins, the m 7 GTP-agarose beads were centrifuged at for 1 min at 15,000 × g and washed four times with the binding buffer at 4 °C. Finally, the proteins linked to the m 7 GTP-agarose beads were eluted by boiling in 50 µl Laemmli buffer (OUTPUT fraction).

| VPg cDNA sequence analysis
Total RNA was extracted from TuMV-GFP UK1 systemically infected leaves using TRI-reagent (Sigma-Aldrich). Reverse transcription was performed on 500 ng RNA using a PrimeScript RT reagent kit (Takara) with oligo-dT and random hexamer primers.

| Statistical analysis
Kruskal-Wallis statistical tests were performed using the pgirmess package on R software (http://www.r-proje ct.org/).

ACK N OWLED G EM ENTS
We thank Christophe Robaglia for inspiring discussions and keen support. We are grateful to Sylvie German-Retana for critical reading and helpful comments on the manuscript. We thank Petyo Zafirov for technical advice on the preparation of the manuscript figures. Funding was provided by ANR-POTYMOVE (ANR-16-CE20-000803).

DATA AVA I L A B I L I T Y S TAT E M E N T
Data available on request from the authors.