Wheat leaf rust is one of the most common and destructive diseases of wheat (McCallum et al., 2016). It is a production problem in nearly all areas where wheat is grown around the world (Huerta-Espino et al., 2011). Control strategies include early seeding and fungicides, but for economic, environmental and sustainability reasons, efforts to counteract this disease have focused mainly on genetic resistance. Major resistance genes (R), also referred to as seedling genes or race-specific genes, tend to conform to the gene-for-gene concept of plant disease resistance where the R gene recognizes directly or indirectly the product of its avirulence (Avr) counterpart (Flor, 1955). Wheat leaf rust genes Lr1 (Cloutier et al., 2007), Lr10 (Feuillet et al., 2003), Lr13 (Hewitt et al., 2021), Lr21 (Huang et al., 2003) and Lr22a (Thind et al., 2017) are typical race-specific genes with nucleotide binding site (NBS) and leucine-rich repeat (LRR) motifs known to be important for plant-pathogen recognition while the more recently cloned Lr14a encodes a membrane-bound ankyrin repeat protein (Kolodziej et al., 2021). A large number of leaf rust resistance genes are such typical race-specific R genes but race non-specific genes function differently. The latter constitute a subset of adult plant resistance (APR) genes, and confer partial resistance in a race non-specific manner characterized by a slow rusting phenotype.

APR genes become effective as the plant progresses to the adult stage. Some APR genes are race-specific and function similarly to seedling resistance genes, while others are race non-specific and are quantitative or slow rusting in nature. The slow rusting phenotype includes a longer latency period resulting in slower and less severe growth of the pathogen, hence the term partial resistance. Rust pustules are still present on wheat lines that carry these APR genes, but their size and number are reduced and their development is delayed. In hexaploid wheat, Lr34, Lr42, Lr46, Lr67 and Lr68 have been described as non-race specific APR genes (Dyck, 1987; Singh et al., 1998; Hiebert et al., 2010; Herrera-Foessel et al., 2011; Herrera-Foessel et al., 2012; Lin et al., 2022). Lr34 is one of the best and most durable resistance genes for leaf rust, and it also conditions resistance to stripe rust (Yr18), stem rust (Sr57), powdery mildew (Pm38) and other biotrophic diseases such as spot blotch (Dyck, 1987; McIntosh, 1992; Singh, 1993; Spielmeyer et al., 2005; Lillemo et al., 2008; Lillemo et al., 2013). As such, Lr34/Yr18/Pm38/Sr57, called Lr34 for short, should be classified not only as a race non-specific APR gene but more accurately as a multi-pathogen resistance gene (Krattinger et al., 2013; Ellis et al., 2014). Leaf tip necrosis (LTN) is an interesting phenotype of Lr34 (Singh, 1992) that has been used as a selection tool for Lr34 and other APR genes, although Lr34-transgenic durum wheat Stewart expressed the leaf rust Lr34 but not the leaf tip necrosis phenotype (Rinaldo et al., 2017). Lr46 (Lr46/Yr29/Pm39/Sr58/Ltn2) and Lr67 (Lr67/Yr46/Pm46/Sr55/Ltn3) have also been described as multi-pathogen resistance genes that similarly produce an LTN phenotype (Lagudah et al., 2007; Herrera-Foessel et al., 2011).

Lr34 (Krattinger et al., 2009), Lr42 (Lin et al., 2022) and Lr67 (Moore et al., 2015) have been cloned. Lr34 encodes an ATP-binding cassette (ABC) transporter of the ABCG subfamily with two transmembrane (TMD) and two cytosolic nucleotide-binding domains (Krattinger et al., 2009). Lr67 encodes a hexose transporter with an affinity for glucose; susceptible and resistant isoforms of Lr67 differ by two amino acid residues (Moore et al., 2015). These genes are not only structurally different from known R genes, but they encode proteins of different classes. Interestingly, Lr42 is a nucleotide-binding site leucine-rich repeat (NLR) gene, but it is not strictly a seedling gene because it confers broad effectiveness against all races at both seedling and adult plant stages (Lin et al., 2022).

Broad spectrum resistance, i.e., little or no race-specificity, is often observed when a gene is brought into a gene pool from a distant relative (Ellis et al., 2014). This is not the case for Lr34+ (the resistant allele), whose origin post-dates the formation of hexaploid wheat (Krattinger et al., 2011; Krattinger et al., 2013; Dakouri et al., 2014). Lr34 is located on Triticum aestivum chromosome 7D. Independent haplotyping of several hundred accessions of Aegilops tauschii (D genome ancestor) revealed that only the Lr34- haplotype H2 was present in this gene pool, indicating that the susceptible Lr34- allele was the ancestral gene and that the H1 haplotype (Lr34+) arose after the interspecific crosses between the T. turgidum and Ae. tauschii progenitors that produced T. aestivum (Krattinger et al., 2011; Krattinger et al., 2013; Dakouri et al., 2014). Lr34 has orthologs in rice and sorghum, both of which correspond to the susceptible allelic form, and is not present in maize, Brachypodium, barley and rye, where it is suggested to have been deleted (Krattinger et al., 2013). In bread wheat, homoeologs are present on chromosomes 7A and 4A but only the 4A homoeolog is putatively functional (Kang et al., 2011).

A total of five haplotypes (H1-H5) have so far been described in spring wheat (Krattinger et al., 2009; Lagudah et al., 2009; Dakouri et al., 2010; Dakouri et al., 2014) and an additional two in winter wheat (Lagudah et al., 2009; Cao et al., 2010). In spring wheat, the resistant (Lr34+) haplotype H1 is null/C for the mutation sites in exons 11/12 and the known susceptible (Lr34-) haplotypes H2-H4 are all TTC/T. The susceptible haplotypes H2, H3 and H4 differed by the presence of single nucleotide variants in intron 4 and exon 10 (Dakouri et al., 2010; Dakouri et al., 2014). In a germplasm collection of 700 accessions, 698 belonged to the main H1-H4 haplotypes but accessions Odesskaja 13 and Koktunkulskaja 332 had a rare null/T haplotype at exons 11/12, corresponding to haplotype H5 (Dakouri et al., 2010; Dakouri et al., 2014). These two accessions had an intermediate leaf rust phenotype in the field and it was not possible to categorically determine whether either or both mutations in exons 11 or 12 imparted resistance to leaf rust.

It was originally believed that only the mutation in exon 11 was needed to recover full LR34 functionality (Chauhan et al., 2015). However, it is still unclear whether both mutations contribute to the functionality. Here, recombinant inbred line (RIL) populations from Odesskaja 13/Thatcher, Odesskaja 13/Thatcher-Lr34 near-isogenic line (NIL) RL6058, Koktunkulskaja 332/Thatcher and Koktunkulskaja 332/RL6058 were developed. The severity of leaf rust infection on these lines in the field was measured in six disease nurseries over three years to determine the functionality of the LR34 proteins as encoded by the three segregating haplotypes present in these populations: null/C, null/T and TTC/T. Here we provide phenotyping data to quantify the contribution of both mutations and use protein structure modeling of LR34 to support our results.

Lr34 is a multi-pathogen resistance gene that confers slow rusting resistance in wheat. Its cloning led to several studies about its nature, origin, evolution, distribution, structural organization and functionality (Krattinger et al., 2009). The five Lr34 haplotypes (H1-H5) reported in spring wheat are defined based on mutations in intron 4 and exons 10, 11 and 12 (Dakouri et al., 2014). Resistant genotypes (Lr34+) all have the H1 haplotype, while H2, H3 and H4 haplotype germplasm is susceptible (Lr34-) (Dakouri et al., 2010; Dakouri et al., 2014). The common differences between these two phenotypic classes are the mutations in exons 11 and 12. The resistant H1 haplotype is null/C while the susceptible H2, H3 and H4 haplotypes are all TTC/T for exons 11/12 ( Table S1 ). Because of the proximity of these two mutations, the vast majority of the accessions (698/700) previously tested were either null/C or TTC/T; hence it was not possible to partition the causal role of these two mutations using this extensive germplasm. However, the H5 haplotype is null/T, a haplotype that could be useful for addressing this question. This could not be clearly addressed by Dakouri et al. (2014) because only two accessions, both with intermediate severity ratings, possessed this haplotype. To confirm each mutation’s contribution to leaf rust resistance, populations that segregated for these different haplotypes were developed. To do so, the null/T haplotype accessions Odesskaja 13 and Koktunkulskaja 332 were crossed to Thatcher and its Lr34 NIL RL6058, and RIL populations segregating for haplotypes H1, H2 and H5 were developed. The severity of leaf rust infection in the field at two locations per year over three years was then measured. To date, accessions with the TTC/C haplotype have not been identified, and only three of the possible four classes segregated in these crosses.

Our results indicated that exons 11 and 12 mutations independently and additively contribute to leaf rust resistance. H1 haplotype (null/C) RILs were significantly more resistant than H5 haplotype (null/T) RILs which, in turn, were significantly more resistant than H2 haplotype (TTC/T) RILs ( Figures 2 , 3A, B ). The same effect was observed in crosses with Odesskaja 13 and Koktunkulskaja 332 ( Figures 3A, B ), indicating that this is not specific to one genetic background. Overall, the RILs in the Odesskaja 13 crosses were slightly more resistant than their counterparts in the Koktunkulskaja 332 crosses harboring the same haplotypes ( Figures 2 , 3A, B ); differences in their genetic backgrounds being the most likely cause, although this could not be attributed to any specific leaf rust seedling genes based on our screening with ten races. Additionally, when the effects of each mutation were analyzed separately, RILs with the “null” exon 11 allele were significantly more resistant than RILS with the “TTC” allele in both crosses ( Figures 3C, D ). For exon 12, RILs with the “C” allele were significantly more resistant than RILs with the “T” allele ( Figures 3E, F ). Marker csLV46, linked to Lr46, segregated in the Koktunkulskaja 332 crosses. Though it was not as effective as Lr34, the resistance conditioned by the Koktunkulskaja 332 putative Lr46 gene was additive to the resistance conditioned by both Lr34 exon mutations ( Figure 4 ).

The TTC to null exon 11 mutation corresponds to the deletion of a phenylalanine residue at position 546 in the first TMD of LR34, and the T to C exon 12 mutation converts a tyrosine into a histidine at position 633, also in the first TMD (Krattinger et al., 2009). Chauhan et al. (2015) transformed barley with Lr34 constructs called M1 (null/T; equivalent to H5) and M2 (TTC/C; no known haplotype). The M1 transgenic lines displayed leaf rust resistant and LTN phenotypes. In contrast, the M2 transgenic lines were susceptible to rust and showed no LTN, suggesting that, in barley, deletion of the phenylalanine residue was sufficient to provide Lr34-based disease resistance, and that the exon 12 encoded tyrosine residue alone was insufficient for leaf rust resistance functionality. Transformation of wheat with the M1 construct also conferred partial resistance in seedling tests. This agrees with our results that support the functional association with the phenylalanine deletion. However, our results demonstrate that both mutations independently contribute to leaf rust resistance in wheat and are additive in the genetic backgrounds of Odesskaja 13 and Koktunkulskaja 332. Because neither variety possesses effective seedling resistance genes, the field severity ratings measured reflect only APR genes.

Gene stacking or pyramiding has long been an effective strategy against wheat leaf rust. Wheat germplasm with three or more seedling resistance genes had significantly lower severity than accessions with fewer genes (Dakouri et al., 2013). APR genes such as Lr34 were also reported to enhance the effectiveness of other R genes (German and Kolmer, 1992; Vanegas et al., 2008). Combining multiple APR genes is also beneficial as lines with multiple APR genes were reported to have near immunity against rusts (Singh et al., 2000). Silva et al. (2015) reported interactions between Lr68, Lr34 and Sr2 in Uruguay, where they stated that, in their environments, the positive effect of Lr68 exceeded that of Lr34. They hypothesized that the enhanced expression of Lr34 at colder temperatures might be the reason for their observations. Lillemo et al. (2008) found a non-additive interaction between Lr34 and Lr46, while Herrera-Foessel et al. (2009) reported only a marginal improvement of the gene combination. However, a recent report indicated the additive effect between the two genes (Bokore et al., 2022); the latter being consistent with our findings in the Koktunkulskaja 332 populations, where Lr34 and an APR gene postulated to be Lr46 were both effective and additive. Indeed, assuming that lines with csLV46 carry the APR gene Lr46, RILs carrying both APR genes displayed significantly lower rust infection severity than RILs with a single APR gene which, in turn, had lower infection ratings than lines with neither gene. The molecular basis for Lr46 was hypothesized to differ from that of Lr34, i.e., the former is not an ABC transporter, an observation that prompted Lagudah (2011) to recommend using the gene combination as a preventative strategy should any of the two mechanisms be overcome. While there is no evidence of either gene being ineffective, our results suggest a certain degree of additivity between the two genes, which could be explained by different modes of action. However, this is not direct evidence because many more alternative scenarios could explain the additivity.

Using the RoseTTAFold and Alphafold2, the 3D structural fold of LR34 was predicted. As suggested, the predicted 3D models of LR34 were found to contain 12 transmembrane helixes, with six in each TMD (Krattinger et al., 2011). The structural locations of the two non-synonymous mutations in LR34+ and LR34- were further analyzed to show that they are located in the transmembrane helix TMH-2 and TMH-4, respectively, in close proximity to the putative substrate binding sites/channel. The aromatic side chain of the phenylalanine (PHE546) residue in LR34- located in TMH-2 appears to create a structural obstacle in the substrate translocation channel ( Figure 5 ). The functional significance of phenylalanine residue within transmembrane helixes of membrane transporters is well established, and its role in substrate binding, channel gating and transport activities have been reported (Kuo et al., 2003; Rojas et al., 2007; Madej and Kaback, 2013; Alegre et al., 2016; Ganz et al., 2019; Selvam et al., 2019). The second mutation, histidine (HIS634) in LR34-, replaced by tyrosine (TYR633) in LR34+, in TMH-4 of TMD1, is also located near the substrate binding sites/channel. The tyrosine residues are known to stabilize the positive charges within the membrane electric field and contribute to substrate transport function (Dougherty, 1996; Chen et al., 2000; Wu et al., 2008). Likewise, the histidine residues also serve critical roles in the TMDs of transmembrane proteins (Wiebe et al., 2001; Stewart et al., 2007; Wang et al., 2013). In particular, histidine residues are known to undergo protonation state changes in a pH-dependent manner, resulting in a conformation change in transmembrane helixes, change in pore sizes, and to influence the substrate transport activity (Todt and McGroarty, 1992; Chen et al., 2000; Harrison, 2008; Martfeld et al., 2016). Thus it is reasonable to speculate that the substitution of the tyrosine by histidine in LR34-, may result in improved stability, flexibility, and additional structure conformation changes altering substrate translocation activities. This may explain the functional differences between LR34 of H1, H2 and H5 haplotypes. Additionally, at the sequence level, both the histidine/tyrosine residues at position 633/634 are conserved among all plant LR34 orthologues (Krattinger et al., 2011), implying the functional significance of both tyrosine and histidine residues in TMH-4 of LR34.

Lr34 encodes an ABCG transporter that confers race non-specific resistance to wheat leaf rust (Krattinger et al., 2009). Arabidopsis PEN3/PDR8, another plant ABC transporter, confers non-host resistance to the barley powdery mildew fungal pathogen Blumeria graminis f. sp hordei, but fungal hyphae growth and infection were observed in pen3 mutants (Stein et al., 2006). However, these ABC transporter’s exact mode of action in the disease resistance mechanism remains largely unexplored. One of the functions of ABC transporters is the translocation of one or more substrates across membranes. For instance, in Nicotiana plumbaginifolia, NpPDR1, an ABCG protein responsible for resistance to the fungal pathogen Botrytis cinera, was shown to transport sclareol and to participate in basal plant defense (Kang et al., 2011). Similarly, in Arabidopsis, ABCG25, ABCG36 and ABCG40 are known to transport multiple substrates (Campbell et al., 2003; Kim et al., 2007; Strader and Bartel, 2009; Kuromori et al., 2010; Lu et al., 2015). The ABCG transporter encoded by Lr34 was originally hypothesized to either export metabolites that affect fungal growth or to provide resistance through a senescence-like process (Krattinger et al., 2009). More recently, abscisic acid (ABA) has been shown to be a substrate of the Lr34 encoded ABCG transporter in wheat (Krattinger et al., 2019) and also when it is expressed in barley (Bräunlich et al., 2021). Assuming that this is its only mode of action, our results suggest that both the phenylalanine deletion and the tyrosine to histidine substitution independently modify the protein sufficiently to alter the translocation of ABA. The conformation changes caused by one or the other substitution is sufficient to disrupt its function to a certain degree, and this function is disrupted to a greater extent when both mutations are present. LR34 has also been reported to be involved in phospholipid transport (Deppe et al., 2018) and phenylpropanoid accumulation (Rajagopalan et al., 2020), but these may be indirect effects of LR34-dependent ABA allocation (Banasiak and Jasinski, 2022).

Furthermore, it should be noted that the ATP transporters utilize the energy of ATP to trigger conformational changes in the TMD that consequently permits the transport of molecules across membranes (Hollenstein et al., 2007). This is referred to as the ATP-switch model (Higgins and Linton, 2004). In many cases, the substrate binding channel or sites resides between the transmembrane domains TMD1 and TMD2 (Locher, 2016; Lefevre and Boutry, 2018) but the interactions between domains, the order of the activation steps and their consequences on ABC transporter functionality remains unclear (Rees et al., 2009). Of note, a gain-of-function mutation caused by a single amino acid change was also reported for the Arabidopsis PDR9 gene, which encodes the ABC transporter ABCG37 that imparts resistance to auxinic herbicides (Ito and Gray, 2006). Despite these similarities, further structural, biochemical and functional studies are necessary to better understand the specific roles of the variant amino acid residues in ABC transporters such as LR34. Using the germplasm developed herein and knowing that ABA is a substrate of the LR34 ABCG transporter enable us to design future experiments that could shed further light on the mechanisms of resistance and susceptibility and on this phenomenon of intragenic additivity. Heterologous expression of Lr34 in barley demonstrated the conservation of the mechanism between wheat and barley while establishing that Lr34 in barley imparted resistance against leaf and stem rusts as well as powdery mildew (Risk et al., 2013). While the exact mechanism remains unknown, Risk et al. (2012) suggested that Lr34 may induce pathogenesis-related gene transcripts but it does not directly regulate them, a hypothesis that seems to have been confirmed by the transcriptomics study of Lr34 barley transgenic lines (Chauhan et al., 2015).

Here we refer to the phenylalanine deletion (null) in exon 11 and the tyrosine (T) to histidine (C) substitution in exon 12 as two gain-of-function mutations from a phenotypic perspective because they both independently contribute additively to APR resistance to leaf rust. However, knowledge of the substrate(s) of LR34- is lacking. Therefore, it remains unclear whether the LR34- function(s) is altered by the two substitutions in LR34+. In other words, is this an added function or a case of neofunctionalization? This is an interesting question that remains to be elucidated.

In conclusion, the data clearly demonstrated that the two non-synonymous mutations located in exons 11 and 12 of Lr34 were independently additive to adult plant resistance against leaf rust in wheat. This was observed in independent sets of crosses with both Odesskaja 13 and Koktunkulskaja 332 and did not appear specific to the genetic background. The germplasm developed in this study consists of H1 (null/C), H2 (TTC/T) and H5 (null/T) haplotypes which show degrees of resistance to leaf rust in the following order: H1>H5>H2. Krattinger et al. (2019) and Bräunlich et al. (2021) demonstrated that ABA was the substrate of LR34. The germplasm developed herein could be used to clarify the role of each of the two read mutations in ABA transport. If the premise is that both mutations alter the transport channel and that these structural changes affect the transport of ABA, then this can be evaluated using the germplasm developed herein. For instance, the total and relative concentration of ABA in leaves and changes in ABA fluxes in the seedlings can be measured in each haplotype. These results should also have important implications for wheat breeding because Lr34 is one of the most widely deployed and durable resistance genes, and it confers resistance against several wheat pathogens.

The original contributions and datasets presented in the study are all included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding authors.

SC and BM designed the experiment, performed the phenotyping and the data analyses and co-wrote the manuscript. ER made the crosses, performed the molecular analysis and phenotyping. BK did the protein modeling and wrote the relevant sections. All authors read and approved the final version of the manuscript.