Introduction

Cassava is one of the most important starch crops grown in the world (Agricultural production statistics 2000–2022, 2023). In sub-Saharan Africa, cassava is particularly important for food security, but its production is threatened by viral diseases (Rey & Vanderschuren, 2017). In particular, cassava brown streak virus (CBSV) and Ugandan cassava brown streak virus (UCBSV), closely related ipomoviruses belonging to the family Potyviridae, are of major concern (Tomlinson et al., 2018; Robson et al., 2023). Independent or simultaneous infection by these viruses cause cassava brown streak disease (CBSD). Symptoms of CBSD include feathery chlorosis of leaves, brown streaking of stems, and necrosis of storage roots. Cassava varieties with varying levels of resistance have been identified and are being deployed, but none of the varieties currently available to farmers are fully immune (Sheat & Winter, 2023; Robson et al., 2023). A highly effective RNAi-based transgenic approach has been developed but faces regulatory hurdles in many target countries (Odipio et al., 2014; Beyene et al., 2016; Wagaba et al., 2016; Tomlinson et al., 2018).

Viruses of Potyviridae are positive sense, single stranded RNA viruses (Valli et al., 2015). Many of these viruses have been studied extensively as a large number are agronomically important pests (Revers & García, 2015; Yang et al., 2021). In many potyvirus-plant pathosystems, members of the host eIF4E gene family function as susceptibility factors and are necessary for completion of the viral life cycle (Robaglia & Caranta, 2006; Naderpour et al., 2010; Hart & Griffiths, 2013; Bastet et al., 2017; Gomez et al., 2019; Gao et al., 2020). This gene family encodes mRNA cap-binding proteins that facilitate recruitment of translation machinery to transcripts (Browning & Bailey-Serres, 2015). The plant eIF4E-family can be broken into three clades: eIF4E, eIF(iso)4E, and nCBP (Gomez et al., 2019). Previous work in Arabidopsis demonstrated that turnip mosaic virus (TuMV) requires a functional copy of eIF(iso)4E for disease (Duprat et al., 2002). Since then, potyvirus resistance loci in many crops have been mapped to eIF4E-family genes and alleles conferring potyvirus resistance often harbor missense mutations (Bastet et al., 2017). Additional studies have demonstrated that potyvirid viral protein genome-linked (VPg) interacts with eIF4E-family proteins (Wittmann et al., 1997; Schaad et al., 2000) confirming that the interacting eIF4E-family member functions as a susceptibility factor (Yeam et al., 2007; Charron et al., 2008). Conversely, viral isolates that break eIF4E-family based resistance have been found to have accumulated compensating mutations in their VPg (Charron et al., 2008; Gallois et al., 2010; Li et al., 2016; Lebaron et al., 2016; Takakura et al., 2018). VPg is covalently linked to the 5’ end of the viral genome. It has been hypothesized that VPg interacts with eIF4E-family proteins to recruit translational machinery to synthesize the viral polypeptide (Sanfaçon, 2015) and complete the viral life cycle. Solving the structure of VPg-susceptibility factor protein complex could facilitate engineering of susceptibility factor alleles that cannot be co-opted by the pathogen. However, co-crystal structures have yet to be produced and the VPg protein complex with eIF4E-family S factors falls below the current minimum size limit required for cryogenic electron microscopy. To date, engineering synthetic eIF4E resistance alleles has involved transferring resistance-conferring mutations from one pathosystem to another with some degree of success (Kim et al., 2014; Bastet et al., 2018, 2019; Zafirov et al., 2021).

We previously found that simultaneous mutations in two cassava eIF4E-family members, nCBP-1 and nCBP-2, strongly attenuated CBSD storage root necrosis and accumulation of CBSV in storage roots of cultivar 60444 (Gomez et al., 2019). Knocking out nCBP-2 alone resulted in an intermediate phenotype, while the ncbp-1 single mutant exhibited no differences from wild type. Aerial symptoms persisted in the ncbp-1 ncbp-2 double mutant, although with less severe stem brown streaking. This incomplete resistance suggests that VPg may interact broadly with the remaining cassava eIF4E-family proteins. Indeed, CBSV VPg could associate with all five cassava eIF4E-family proteins by coimmunoprecipitation (Gomez et al., 2019). As full CBSD resistance may require engineering all cassava eIF4E-family alleles to evade CBSV VPg, we employed targeted and random mutagenesis of nCBP-2 to identify structural determinants of VPg interaction. Furthermore, to better understand the degree of functional redundancy exhibited by cassava eIF4E-family proteins in CBSD, we here expand our CRISPR/Cas enabled genetic approach to characterize the roles of cassava eIF4E and eIF(iso)4E clades in CBSD.

Results

Generation of cassava eIF4E-family mutants

We previously reported that a cassava ncbp-1 ncbp-2 double mutant in the 60444 background displayed increased, but incomplete, resistance to CBSV challenge (Taylor et al., 2012; Gomez et al., 2019). While VPg interacted most strongly with nCBP-1 and nCBP-2 in yeast two-hybrid assays, all five eIF4E-family proteins interacted with VPg in co-immunoprecipitation assays. These results suggest that there is likely functional redundancy among this family of proteins. To test this idea, we used CRISPR/Cas9-mediated genome editing to generate new cassava eIF4E-family mutants in TME419, a farmer preferred cassava cultivar. gRNAs were designed to target the first exon of each target gene and cassava transformations were performed as previously described (Fig. S2) (Taylor et al., 2012; Gomez et al., 2019). A total of 156 recovered lines from 21 transformations were characterized for possible mutations at the target site(s) using amplicon sequencing. These efforts yielded independent single mutant lines for all members of the cassava eIF4E gene family, along with eif(iso)4e-1 eif(iso)4e-2 and ncbp-1 ncbp-2 double mutants (Table 1, S3). All but one mutant line used in this study are predicted to have frameshifting mutations that result in loss of protein function. The lone exception is one of the ncbp-1 ncbp-2 double mutants where both nCBP-1 alleles and one allele of nCBP-2 have frameshifting mutations and the remaining nCBP-2 allele has a deletion that results in loss of amino acids 45 through 51. This allele is designated nCBP-2^K45_L51deland the mutant line is designated ncbp-1 nCBP-2^K45_L51del (Table 1, S3). No differences in growth and development were observed between any of the mutant lines and wild type (Fig. S3-4).

Table 1.

CBSD phenotypes of TME419 eIF4E-family mutants

To determine the contribution of cassava eIF4E family genes to development of CBSD, mutant and wild type TME419 were challenged with CBSV isolate Naliendele in greenhouse trials where stem brown streaking and leaf feathery chlorosis were tracked for approximately three months. The specific CBSV isolate was chosen as it produces strong aerial symptoms quickly after inoculation and reliably results in strong storage root necrosis (Wagaba et al., 2013). At the end of the viral challenges, storage roots were excavated from soil, sectioned, and scored for severity of necrosis as a measure of spread along storage root length and diameter (Gomez et al., 2019). These same root samples were used to measure viral titer by quantitative real-time PCR (qRT-PCR). For aboveground symptoms, eif4e, eif(iso)4e-1, eif(iso)4e-2, and ncbp-1 single mutants had no reduction in disease severity relative to wild type, while ncbp-2 and both eif(iso)4e and ncbp double mutants exhibited small but significant reductions (Figure 1a). For storage root necrosis symptoms in the single mutants, the ncbp-1 and ncbp-2 mutants displayed a decrease in symptoms compared to wild type with the latter showing the strongest phenotype (Figure 1b). Each of the double mutants showed a decrease in symptom with the ncbp-1 ncbp-2 mutant, where all nCBP alleles have frameshifting mutations, having the greatest reduction in necrosis; virtually no necrosis was observed (Figure 1b, d). We further measured viral accumulation in storage roots of wild type, ncbp-1, ncbp-2, ncbp-1 ncbp-2, and ncbp-1 nCBP-2^K45_L51del lines using qPCR analysis and observed that lines with frameshifting mutations in ncbp-2 had significantly lower virus titer (Figure 1c). Taken together, these data suggest that while nCBP-2 is the primary contributor to root disease severity, other eIF4E-family members may also contribute throughout the plant.

Fig. 1

Purified CBSV VPg and cassava eIF4E-family proteins interact in vitro

To further explore functional redundancy within the eIF4E-family with respect to CBSD and, specifically, interaction with VPg, we used a biochemical approach. Using co-immunoprecipitation, we previously demonstrated that CBSV VPg and cassava eIF4E-family proteins form complexes in whole cell lysates using co-immunoprecipitation (Gomez et al., 2019). However, these experiments did not address whether the proteins are directly interacting or if another complex member is present and facilitating the protein-protein association. Thus, we tested if VPg directly associates with eIF4E-family proteins in the absence of additional factors using a GST pull-down experiment. GST-tagged cassava eIF4E-family proteins and VPg, tagged with both 6xHis and 3xFLAG, were expressed in and purified from E. coli (Figure S5). We found that CBSV VPg was pulled down by GST-tagged eIF4E, nCBP-1, and nCBP-2 in comparison to GST alone (Figure 2a, b). Pulldown of VPg by GST-tagged eIF(iso)4E-1 and eIF(iso)4E-2 was also tested but we observed inconsistent enrichment of VPg by GST-eIF(iso)4E-1 & −2, relative to GST alone, across experimental replicates.

Figure 2.

Targeted nCBP-2 mutations fail to disrupt interaction with VPg

If CBSV VPg can coopt all cassava eIF4E-family proteins for completing the viral lifecycle, engineering cassava to be resistant to CBSV may require identifying alleles of eIF4E-family genes encoding mutations that disrupt VPg binding but still allow for translation-initiation activity. In other plant-potyvirid pathosystems, specific mutations within eIF4E-family members have proved sufficient to block interaction with VPg while maintaining normal protein function (Figure 2c) (Ashby et al., 2011; Kim et al., 2014). One well studied example is TuMV VPg and Arabidopsis eIF(iso)4E. In this system, mutation of a highly conserved tryptophan (W92L) and semi-conserved lysine (K149E) in Arabidopsis eIF(iso)4E disrupts binding of TuMV VPg in pull-down and co-immunoprecipitation experiments. Overexpression of this mutant eIF(iso)4E allele in Chinese cabbage confers resistance to TuMV (Miyoshi et al., 2006; Kim et al., 2014). To test the hypothesis that specific eIF4E gene family alleles might avoid VPg binding, we took a yeast two-hybrid approach using different alleles of nCBP-2. We recapitulated the W92L/K149E mutations from Arabidopsis eIF(iso)4E in cassava nCBP-2, resulting in nCBP-2^W118L/N170E (Figure 2d). Whereas the eIF(iso)4E^W92L/K149E no longer associated with TuMV VPg in a co-immunoprecipitation experiment (Figure 2e), we observed that CBSV VPg was similarly present in both wild type nCBP-2 and nCBP-2^W118L/N170E pull downs (Figure 2f). VPg - nCBP-2^W118L/N170E interaction was also observed through yeast two-hybrid assay (Figure 2g, h).

TuMV and CBSV belong to two different genera of Potyviridae, Potyvirus and Ipomovirus. It is possible that viruses from these distinct genera have evolved mechanistic differences in how their VPg proteins interact with host eIF4E-family S factors. To determine VPg protein conservation between potyviruses and ipomoviruses, VPg sequences from five major potyvirus clades were aligned with VPg sequences from seven of the eight known ipomovirus species (TomMMoV, SPMMV, SqVYV, CocMoV, CVYV, UCBSV, CBSV) (Figure S6). Across all VPg sequences analyzed, several highly conserved residues are observed, as are striking genus-specific differences; four stretches of amino acids are present only in one genus or the other (pink boxes). One of these stretches coincides with an α-α loop in PVY VPg thought to directly interact with the ventral side of eIF4E (Coutinho de Oliveira et al., 2019); in ipomoviruses this region is eight amino acids shorter. Additionally, a highly conserved aspartate residue (D77) necessary for TuMV VPg interaction with Arabidopsis eIF(iso)4E (Léonard et al., 2000), is not present in ipomoviruses, indicating this residue may have a less conserved role in interactions between ipomovirus VPgs and eIF4E-family proteins.

A yeast two-hybrid screen identifies nCBP-2 mutants that lose VPg affinity

Since mutations from the known Arabidopsis resistance allele did not block nCBP-2 – VPg interaction, we performed a yeast two-hybrid screen for nCBP-2 mutants that lose affinity for VPg (Figure 3a). Mutant nCBP-2 amplicons were generated by error-prone PCR and cloned into the bait vector by in vivo recombination in EGY48 yeast already containing both lacZ reporter and VPg prey vectors (Figure 3a). From roughly 1000 yeast transformants, 51 nCBP-2 mutants were isolated that lost interaction with VPg (Table 2, Table S4). Forty-eight unique amino acid substitutions were observed across forty-six unique nCBP-2 mutants. Of the unique mutants, 27 had mutations affecting a single residue or the introduction of a frameshifting deletion while 19 contained two or three mutated residues. Nine residues were isolated more than once, with up to three unique amino acid substitutions identified. Mutants were then filtered to identify residues with varying degrees of conservation (Figure S7) across cassava eIF4E-family proteins resulting in 15 being re-cloned and validated for loss of VPg interaction in yeast using a quantitative β-galactosidase activity assay. All but two mutants, nCBP-2^L147H and nCBP-2^R196S, were confirmed to confer total loss of β-galactosidase activity in yeast and suggest commensurate loss of VPg affinity (Figure 3b, 5b, S8). The R196S mutant yeast did not have any reduction of β-galactosidase activity relative to yeast expressing wild-type nCBP-2, while L147H mutant yeast exhibited an intermediate reduction in β-galactosidase activity. We further validated by western blot analysis that the yeast used for two-hybrid experiments expressed nCBP-2 variants and CBSV VPg (Figure S8).

Fig. 3

Table 2.

Translation initiation activity of isolated nCBP-2 mutants

Previous work with Arabidopsis eif4e or eif(iso)4e mutants suggested that both genes were functionally redundant for translation initiation. However, eif4e mutants exhibit subtle reductions in fertility and biomass in addition to delayed flowering time (Bastet et al., 2017, 2018). Similarly, simultaneous knockout of tomato eIF4E1 and eIF4E2 incurs a growth penalty (Gauffier et al., 2016). T93C is a yeast eif4e mutant with chromosomal eIF4E replaced with LEU2 (Altmann et al., 1989). This yeast strain can survive on galactose media due to a plasmid-based eIF4E gene that is galactose inducible but is repressed in the presence of glucose (Figure 4a). Coding sequences of nCBP-2 mutants confirmed to lose affinity for CBSV VPg by quantitative yeast two-hybrid were cloned into pG1.1 under control of the constitutive yeast GPD promoter and transformed into T93C. Transfer of these T93C transformants from galactose- to glucose-containing media revealed that only three nCBP-2 mutants rescue loss of eif4e: nCBP-2^{A20T/H94Q/L200P}, nCBP-2^L51F, and nCBP-2^S138P (Figure 4bc, 5cd).

Fig. 4

Mutating the conserved HPL motif of nCBP-2 can affect VPg binding and disease

Among the three nCBP-2 mutants that retained translation initiation activity, nCBP-2^L51F stood out as having a mutation in the highly conserved HPL motif. In yeast, this motif participates in docking with eIF4G (Grüner et al., 2016, 2018). In addition, this amino acid is deleted in the nCBP-2^K45_L51del cassava mutant line that displayed increased resistance to CBSD (Figure 5a). The HPL motif is found in many plant eIF4E-family proteins and as such, a mutation of the leucine may be suitable for engineering resistance to other potyvirids. We tested the nCBP-2^K45_L51del mutant for interaction with VPg by quantitative yeast two-hybrid. Yeast expressing this mutant exhibited no β-galactosidase activity, like yeast expressing nCBP-2^L51F (Figure 5b, S9a). We confirmed this implied loss of VPg interaction by observing strongly reduced VPg binding in GST pulldowns with both mutant proteins (Figure 6). However, comparing wild-type nCBP-2, nCBP^L51F and nCBP-2^K45_L51del for translation initiation activity in T93C yeast (τιeif4e) showed that while nCBP-2^L51F was able to rescue loss of eIF4E in T93C it was to a lesser degree than wild type nCBP-2, and nCBP-2^K45_L51del was not able to complement T93C (Figure 4d, e). Leucine 51 of nCBP-2 corresponds to leucine 28 within Arabidopsis eIF(iso)4E. We hypothesized that recapitulating the L51F substitution in Arabidopsis eIF(iso)4E would have the similar effects on TuMV VPg binding. Using the same quantitative yeast two-hybrid approach as above we observed that the L28F mutation strongly disrupted eIF(iso)4E interaction with TuMV VPg (Figure 7a, S9b). However, it is not able to rescue T93C eif4e mutant yeast (Figure 7b, c).

Fig. 5

Fig. 6

Fig. 7

Discussion

Potyviridae is the largest plant-infecting family of RNA viruses comprised of 249 viruses across 12 genera (Walker et al., 2022). The genus Potyvirus is perhaps the best studied as it contains 206 of the 249 known viruses belonging to the family Potyviridae. In comparison, only eight species of viruses belonging to the genus Ipomovirus have been identified and much less is known about molecular ipomovirus-host interactions (Dombrovsky et al., 2014). While the genome structure and genes encoded by potyviruses and ipomoviruses are similar, and thus it is reasonable to expect a degree of conservation in virulence strategies, it is also likely that differences have arisen given the evolutionary divergence of the two genera. For CBSV in particular, we note that while eIF4E-family proteins are clear susceptibility factors, the breadth of family members that contribute to disease appears broader than typically found for potyviruses (Bastet et al., 2017). Furthermore, while resistance-conferring amino acid substitutions in eIF4E-family proteins have been effectively transferred between potyvirus-pathosystems, we find in this study that Arabidopsis eIF(iso)4E mutations that break TuMV VPg interaction fail to do so for the cassava nCBP-2 interaction with CBSV VPg. These observations highlight the need for further characterization of ipomovirus-host molecular interactions.

Structure-function analysis of VPg interaction with eIF4E-family susceptibility factors would strongly benefit from crystal or cryo-EM structures of the complex; these have yet to be successfully generated (Coutinho de Oliveira et al., 2019). Most recently, an NMR approach solved the structure of PVY VPg (Coutinho de Oliveira et al., 2019). The PVY VPg and human eIF4E interaction interface was also mapped through monitoring chemical shift perturbations of either protein during NMR spectroscopy. It is unclear how well these findings will translate to diverse potyvirid VPg interactions with cognate plant eIF4E-family proteins, especially for the plant specific eIF(iso)4E clade and considering significant divergence of ipomovirus VPg sequences from those of potyviruses. Thus far, attempts to design eIF4E-family resistance alleles have relied on inference-based approaches (German-Retana et al., 2008; Ashby et al., 2011; Kim et al., 2014; Bastet et al., 2018, 2019; Zafirov et al., 2021). In contrast, here we utilize a loss-of-affinity screen of randomly mutagenized nCBP-2 variants to identify key residues necessary for interaction with CBSV VPg. Only three nCBP-2 amino acid residues that we found to be necessary for interaction with VPg had previously been described in potyvirus-plant pathosystems. These amino acid residues, D106, V164, and R168 in nCBP-2, are highly conserved across eIF4E-family proteins. In pepper, the pvr1²allele of eIF4E is responsible for resistance to isolates of PVY and TEV, and the amino acid corresponding to nCBP-2 D106 is one of three causal amino acid substitutions (Kang et al., 2005). That aspartate in pvr1² is substituted with asparagine (D to N). For cassava nCBP-2, two separate D106 substitutions were observed; D106V and D106H, the latter of which co-occurred with an A41T substitution. Only the nCBP-2^D106V mutant was tested for translation initiation activity and was found deficient in rescuing T93C eif4e yeast. It is unclear if a D106N mutation in nCBP-2, as found in pvr1², would similarly break VPg interaction while preserving its translation initiation activity. nCBP-2 V164, where mutation to leucine disrupts CBSV VPg interaction, corresponds to V167 in eIF4E of pea. The pea eIF4E^V167A variant confers resistance to PSbMV and can rescue eif4e yeast. In contrast, cassava nCBP-2^V164L does not rescue T93C eif4e yeast. Whether the V164A mutation in nCBP-2 would have the same effects as in pea eIF4E is an open question. Lastly, for nCBP-2 R168, the previously described NMR structural study demonstrated that corresponding residues in human eIF4E and wheat eIF(iso)4E are involved in the interaction with Potato virus Y VPg (Coutinho de Oliveira et al., 2019). Like nCBP-2^D016V and nCBP-2^V164L, nCBP-2^R168C also failed to rescue T93C.

Of the fourteen nCBP-2 loss-of-VPg-interaction mutants that we tested for translation initiation activity, eleven were nonfunctional. The three nCBP-2 variants that retained function were nCBP-2^{A20T/H94Q/A148V}, nCBP-2^L51F, and nCBP-2^S138P. For nCBP-2 ^{A20T/H94Q/A148V}, A20 resides outside of the conserved core of eIF4E family proteins while H94 and A148 are present in alpha helices found on the dorsal side of nCBP-2; further work is required to identify which individual or combination of these amino acids are required for CBSV VPg interaction. S138 is similarly located in a dorsal alpha helix while L51 is closer to the dorsal side and located N-terminal of a beta strand. Evidence from a study on PVY VPg interaction with human eIF4E indicates that VPg interfaces with the ventral side of eIF4E (Coutinho de Oliveira et al., 2019). If this also applies to CBSV VPg interaction with nCBP-2, the four aforementioned amino acid residues likely affect VPg binding in an indirect manner. Of particular importance for this study, the L51F substitution impacts the HPL motif, which was also affected in our cassava ncbp-1 nCBP-2^K45_L51del line. This enabled assessment of the importance of that motif for CBSD symptom severity. Unlike nCBP-2^L51F, nCBP-2^K45_L51del did not rescue T93C eif4e deficient yeast. Surprisingly, despite inability to strongly interact with VPg and complement eif4e activity, the ncbp-1 nCBP-2^K45_L51del root disease phenotype was intermediate to wild type and ncbp-1 ncbp-2, indicating that this mutant version of nCBP-2 is sufficient for the development of some disease symptoms. A similar result has also been observed in pea eIF4E mutants where of eight mutants unable to rescue eif4e mutant yeast, two were still able to confer susceptibility to PsbMV (Ashby et al., 2011). In such mutants, it may be that low levels of residual translation initiation activity are sufficient for completion of the viral life cycle.

In cassava the eIF4E-family has five members and we previously demonstrate that all five proteins interact with CBSV VPg by co-immunoprecipitation (Gomez et al., 2019). We now find, through pulldown experiments with purified proteins, that nCBP-1, nCBP-2, and eIF4E can directly interact with CBSV VPg. This validates our approach for identifying nCBP-2 residues crucial for VPg interaction. In contrast, our pulldown experiments with VPg and eIF(iso)4E-1 and −2 were uninformative. It may be that the cassava eIF(iso)4E proteins have lower affinity for CBSV VPg than the nCBPs and eIF4E, precluding detection of interaction via GST pulldown. This is supported by our earlier work using yeast two-hybrid to dissect CBSV VPg and cassava eIF4E-family interactions; there, the combination of CBSV VPg with eIF(iso)4Es produced weaker reporter activity than when VPg was combined with nCBPs (Gomez et al., 2019). A more granular understanding of differences in cassava eIF4E-family protein affinity for CBSV VPg may be achieved using techniques like surface plasmon resonance or isothermal titration calorimetry (Upadhyay et al., 2024).

To determine the contribution of each eIF4E-family gene to CBSV symptoms, and therefore the biological relevance of interaction with VPg, we have used CRISPR/Cas9 to generate single mutants of all five genes, as well as some double mutants, in cassava variety TME419. Notably, the ncbp double mutant has near total reduction of storage root necrosis, commensurate with strong reductions in viral titer. Of the two nCBP genes, nCBP-2 appears to be the main driver for the improved disease response as evidenced by the results of the single nCBP mutants, and that nCBP-2 is more highly expressed than nCBP-1 (Gomez et al., 2019). As noted in our previous work, the expression difference between nCBP-2 and nCBP-1 is even more pronounced in storage roots and may be the reason for differential disease response in roots and aerial tissues of ncbp-2 and ncbp-1 ncbp-2 mutants (Gomez et al., 2019). The disease data are consistent between mutants in both 60444 (Gomez et al., 2019) and TME419. Beyond the nCBP clade, the eIF(iso)4E clade appears to have a small contribution to disease as aerial symptom severity and storage root necrosis is slightly reduced when both eIF(iso)4E-1 and eIF(iso)4E-2 are simultaneously knocked out. This finding also argues against the ncbp-1 ncbp-2 root disease phenotype being simply due to additively knocking out family members. It remains to be seen if higher order mutant combinations would have an even greater effect on reducing CBSD disease severity. It will also be important to test the utility of ncbp-1 ncbp-2 mediated resistance against the natural diversity of cassava brown streak viruses in field trials.

A key consideration going forward is that any benefits conferred by mutations in eIF4E-family proteins with respect to viral resistance may also be accompanied by growth and developmental penalties (Gauffier et al., 2016, Bastet et al., 2017). Of the four residues identified in nCBP-2 mutants that appear to avoid this penalty, three (L51, H94, and A148) are conserved within all cassava eIF4E-family proteins and the fourth (S138) is conserved within three members. Targeted base editing of some or all of these residues across the entire eIF4E-family may allow for improved CBSV response while maintaining essential protein translation functions. Indeed, we demonstrate here that transferring the L51F mutation from nCBP-2 into Arabidopsis eIF(iso)4E, as an L28F mutation, results in loss of affinity for TuMV VPg. As base editing approaches mature alongside improvements with gene targeting, combining these approaches with loss-of-VPg interaction screens may facilitate isolation and implementation of novel eIF4E-family resistance alleles. This may allow for countering the edge viruses hold in the host-pathogen evolutionary arms race.

Materials and Methods

Plasmid vector construction

gRNAs for CRISPR/Cas9-based mutagenesis of cassava eIF4E-family genes were selected using the CRISPOR webtool (Concordet & Haeussler, 2018). Cas-OFFinder was then used to screen gRNAs for off-target activity against a TME419 reference based genome assembly produced using data (SRX1393295) from Bredeson et al. and the AM560-2 v6.0 reference genome (Table S1) (Bae et al., 2014; Bredeson et al., 2016). gRNAs with off-targets containing less than 2-mismatches were excluded from consideration. Finally, gRNAs were cloned into the pDIRECT 21C CRISPR binary vector as previously described by (Čermák et al., 2017).

Plant transformation, genotyping, and growth parameters

Cassava cultivar TME419 was transformed as previously described (Veley et al., 2023). DNA was extracted from recovered transformants using the CTAB procedure. PCR was then used to amplify 240 to 428 bp regions encompassing the gRNA binding sites. Purified amplicons were sent to GENEWIZ for amplicon sequencing. Amplicon sequencing data was analyzed using the CRIS.py python program (Connelly & Pruett-Miller, 2019). Plantlets of transgenic lines were then maintained in tissue culture (Taylor et al., 2012).

Viral challenge and disease phenotyping

Cassava disease challenges with CBSV isolate Naliendele, including inoculation, aerial and below ground disease severity assessments, and qPCR-based quantification of storage root viral titer, were carried out as described previously (Gomez et al., 2019). For data analysis, wild type disease severity was compared across all challenges and no significant differences were found (Figure S1). Data for all challenges were then aggregated to increase sample sizes for statistical analysis.

Co-immunoprecipitation and GST-pulldown experiments

Co-immunoprecipitation was performed as described previously (Gomez et al., 2019). For GST-pulldown experiments, cassava eIF4E-family genes in pENTR/D-TOPO, from Gomez et al., were cloned into pDEST15 (Invitrogen) by Gateway recombination (Invitrogen). pDEST15 encoding cassava eIF4E-family genes and pGEX-4T-1 (encoding unfused GST, Cytiva) were transformed into BL21-AI E. coli (Invitrogen) and transformants were used to start overnight seed cultures. Seed cultures were then used to inoculate 250 mL of LB broth. Once inoculated cultures were grown to an OD600 of 0.4, IPTG (pGEX-4T-1) or arabinose (pDEST15) were added to final concentrations of 0.5 mM or 0.2%, respectively, to induce recombinant protein production. Induced cells were incubated for four hours at 28°C. Cells were then pelleted and resuspended in 4 mL of lysis buffer (125 mM Tris, 150 mM NaCl, 1 mM DTT, 5 mM MgCl₂, 10 mM CaCl₂, pH 7.4) supplemented with 1X cOmplete protease inhibitor (EDTA-free, Roche). Lysozyme (GoldBio) was added to a concentration of 1 mg/mL and the mixture was allowed to incubate at room temperature for 15 minutes. 50 uL DNAse I (Thermo) was then added to the mixture, followed by sonication on ice to ensure complete lysis and elimination of DNA induced viscosity. The resulting cell lysate was centrifuged at 4°C, maximum speed, for 20 minutes. 3.5 mL of clarified lysate was transferred to a new tube and EDTA was added to a final concentration of 1 mM. GST and GST-eIF4E proteins were then purified from clarified lysate using glutathione magnetic agarose beads (Pierce) according to manufacturer’s protocol. Glutathione was removed from GST and GST-fusion protein eluates by buffer exchange with Amicon Ultra centrifugal filters (Millipore). CBSV 6xHis-VPg-3xFLAG-6xHIS was produced and purified for pulldown experiments as described in (Gomez et al., 2019). GST pulldown was performed by mixing 3 ug of GST or GST-eIF4E proteins with 3 ug of VPg and 10 uL of resuspended glutathione magnetic agarose beads (Pierce) in buffer A (125 mM Tris, 150 mM NaCl, 10 mM DTT, 1 mM EDTA, pH 7.4) supplemented with 1X cOmplete protease inhibitor (EDTA-free, Roche). Samples were incubated at 4°C for one hour. Beads were then washed four times with buffer A and proteins were eluted with Laemmli sample buffer for analysis by western blotting. Antibodies used in this study are described in Table S2.

For comparison of VPg pulldown by nCBP-2 variants across experiments, nCBP-2 and VPg eluate band intensities were first measured using Image Lab (Bio-Rad). Within an experiment, eluate VPg band intensities were first adjusted to correct for differing pulldown efficiencies of each nCBP-2 variant using the following equation where VPg_X is VPg band intensity for sample X and GST_X is full length GST fusion protein band intensity for sample X:

Adjusted VPg intensities were then expressed relative to the VPg intensity of the wild-type nCBP-2 sample for that experiment:

Yeast two-hybrid assays

All yeast two-hybrid experiments were performed using the EGY48 strain harboring the pSH18-34 lacZ reporter plasmid along with pJG4-5 and pEG202 bait and prey plasmids (Golemis & Khazak, 1997). pJG4-5 CBSV VPg bait plasmid and pEG202 nCBP-2 prey plasmids were created by using In-Fusion cloning (Takara) to insert CBSV VPg or nCBP-2 coding sequences into pJG4-5 or pEG202 linearized with EcoRI and XhoI. For qualitative yeast two-hybrid experiments, EGY48 pSH18-34 yeast transformed with bait and prey plasmids were first selected for on SD –ura –trp –his media supplemented with 2% glucose (Gietz & Schiestl, 2007). Transformants were then plated onto SD –ura –trp –his media supplemented with 2% galactose, 1% raffinose, 1X BU salts, and X-gal. Blue/white screening was performed three to four days later. Quantitative yeast two-hybrid was performed using a spectrophotometric β-galactosidase activity assay (Amberg et al., 2006).

To screen for nCBP-2 mutants that lose affinity for CBSV VPg, pEG202 nCBP-2 was used as template for mutagenic PCR (Genemorph II, Agilent). Primers immediately flanking the second and penultimate codon of nCBP-2, preserving the start and stop codons during mutagenic PCR, were used with the low mutation rate protocol from the Genmorph II kit (Agilent). pEG202 empty vector was then digested with EcoRI and SacI and the fragment containing the yeast HIS marker was gel purified (Qiagen). A third pEG202 fragment containing the yeast 2μ origin of replication was produced by PCR with ends that overlap the nCBP-2 mutant amplicons and pEG202 HIS marker fragment by 30 base pairs; co-transformation of all three fragments into yeast result in assembly of an intact pEG202 nCBP-2 plasmid through yeast recombinational cloning. Mutant nCBP-2 amplicons, pEG202 HIS fragment, and pEG202 2μ fragment were mixed in a 4:1:1 molar ratio and co-transformed into EGY48 yeast harboring the pSH18-34 and pJG4-5 CBSV VPg plasmids. Transformed yeast were directly plated onto SD –ura –trp –his media supplemented with 2% galactose, 1% raffinose, 1X BU salts, and X-gal. Four days post transformation, white or light blue colonies were restreaked onto a new plate for further visual assessment of reduced β-galactosidase activity. The yeast were then checked by PCR for the presence of full length nCBP-2 insert in pEG202 using vector backbone specific primers. nCBP-2 mutant alleles were then amplified from insert positive yeast using Q5 Hot Start polymerase (NEB) and analyzed by Sanger sequencing.

Yeast eif4e complementation assay

3xFLAG and nCBP-2 or mutant nCBP-2 coding sequences were cloned into pG1.1 linearized with NcoI and BamHI using In-Fusion (Takara) to produce pG1.1 3xFLAG-nCBP-2. 3xFLAG was amplified from pB7-HFC (Huang et al., 2016) while nCBP-2 variants were amplified from pEG202 yeast two-hybrid plasmids. pG1.1 plasmids were then transformed into the conditional yeast eif4e mutant strain T93C (eif4e::LEU2 ura3 trp1 leu2 [pGAL1-eIF4E URA3]) (Altmann et al., 1989). Transformants were plated on SD –ura –trp dropout media containing 2% galactose and 1% raffinose and allowed to grow for five to seven days. Colonies of transformed yeast were then resuspended in sterile water and a ten-fold dilution series ranging from OD600 = 10^-1 through 10^-4 was plated on SD –ura –trp dropout media containing 2% glucose to examine for rescue by pG1.1 encoded eIF4E-family gene variants. Plates were imaged after 5-7 days.

Data Availability

The source data for all results presented in this study are provided in accompanying supplementary files.

Acknowledgements

We would like to thank the DDPSC PGF team for their hard work in caring for our plants. We thank Claire Albin for her assistance with CBSV challenges and Nigel Taylor for allowing us to use Taylor Lab greenhouse space. Yeast strain T93C, pG1.1 At_eIF4E1, and pG1.1 empty vector were generously provided by Dr. Karen S. Browning. Z.D.L. would like to thank Susan Lin, Wei-Yang Lin, Chun-Hwa Lin, and Jocelyn Lin for their love and support. Z.D.L. was funded by USDA NIFA award 2019-67012-29626 and the Bill & Melinda Gates Foudnation (OPP1125410, R.B.S., S.E.J., J.C.C.). E.J.D., Z.V.B., H.T., and E.H. were funded by the DDPSC NSF-REU program, DBI-1659812, −2050394, −1659812, and −2244275, respectively.

Additional information

Author Contributions

Conceptualization: Z.D.L.; Methodology: Z.D.L., J.C.C., R.S.B.; Investigation: Z.D.L., M.K.S., G.L.H., E.D.M., Z.V.B., K.B., H.T., E.H., G.J., K.B.G.; Visualization: Z.D.L.; Writing – original draft: Z.D.L.; Writing – reviewing & editing: Z.D.L., K.B.G., J.C.C., R.S.B.

Additional files

Supplemental Figures S1-9

Table S1. TME419 gRNA off-target sites

Table S2. Antibodies used in this study

Table S3. Gene editing outcomes for TME419 eIF4E-family mutants

Table S4. All nCBP-2 loss-of-VPg-affinity mutants identified by yeast two-hybrid