Suppression among alleles encoding nucleotide-binding–leucine-rich repeat resistance proteins interferes with resistance in F1 hybrid and allele-pyramided wheat plants



The development of high-yielding varieties with broad-spectrum durable disease resistance is the ultimate goal of crop breeding. In plants, immune receptors of the nucleotide-binding–leucine-rich repeat (NB-LRR) class mediate race-specific resistance against pathogen attack. When employed in agriculture this type of resistance is often rapidly overcome by newly adapted pathogen races. The stacking of different resistance genes or alleles in F1 hybrids or in pyramided lines is a promising strategy for achieving more durable resistance. Here, we identify a molecular mechanism which can negatively interfere with the allele-pyramiding approach. We show that pairwise combinations of different alleles of the powdery mildew resistance gene Pm3 in F1 hybrids and stacked transgenic wheat lines can result in suppression of Pm3-based resistance. This effect is independent of the genetic background and solely dependent on the Pm3 alleles. Suppression occurs at the post-translational level, as levels of RNA and protein in the suppressed alleles are unaffected. Using a transient expression system in Nicotiana benthamiana, the LRR domain was identified as the domain conferring suppression. The results of this study suggest that the expression of closely related NB-LRR resistance genes or alleles in the same genotype can lead to dominant-negative interactions. These findings provide a molecular explanation for the frequently observed ineffectiveness of resistance genes introduced from the secondary gene pool into polyploid crop species and mark an important step in overcoming this limitation.


In crop production, breeding for resistant plant genotypes is widely considered to be the most sustainable strategy for preventing yield losses due to pathogen infestation. Plant breeders constantly renew the set of cultivars offered to farmers by incorporating resistance loci from the primary, secondary and tertiary gene pools into breeding germplasm. Molecular cloning of the underlying genetic constituents for pathogen defense has shown that many of the genes with a major resistance effect encode intracellular resistance (R) proteins with an N-terminal coiled-coil (CC) or TOLL/interleukin-1 receptor (TIR), a central nucleotide-binding (NB) and a C-terminal leucine-rich-repeat (LRR) domain (Marone et al., 2013). These proteins, designated NLR proteins, usually provide strong resistance that is frequently associated with a hypersensitive response (HR), a form of programmed cell death that prevents the spread of biotrophic pathogens in particular. The NLR proteins are specifically activated by the direct or indirect recognition of avirulence (Avr) molecules that are delivered from the pathogen into the host cell. These are mainly effector proteins that usually support the virulence of the pathogen. Co-evolution of host and pathogen populations leads to diversification or a fast turnover at the genomic R and Avr loci, resulting in a large allelic diversity and race specificity of the R-gene-mediated effector-triggered immunity (ETI) (Dodds and Rathjen, 2010).

Thus, the use of R-gene-based resistance in crop plants has the drawback of rapid loss of effectiveness. This is especially true for genetically uniform agricultural ecosystems that create a high selection pressure on pathogen populations. Stacking of multiple highly effective, redundantly acting R genes each covering a broad race spectrum is considered to be a promising strategy for a more sustainable use of race-specific resistance in agriculture (Dangl et al., 2013). Due to the redundancy in recognition the pathogen will have to evolve multiple Avr genes simultaneously to gain virulence on such R-gene-pyramided plants – this is an unlikely event, and hence pyramiding is expected to extend the durability of R-gene crop resistance (McDonald and Linde, 2002).

As an alternative to the stacking of different R genes, different allelic variants of the same R gene can also be combined. For dominant R genes this is possible in a heterozygous form in F1 hybrids. A genetically stable combination of various alleles can be achieved by the use of transgenic approaches, for example via the cross of transgenic lines having different alleles inserted at random sites. Using this approach Bieri et al. (2004) selected lines expressing both the Mla1 and Mla6 powdery mildew resistance specificities from the Mla locus in barley. Additive resistance was also obtained when L6 was combined with the L2 or L10 alleles of a flax rust resistance gene in one genotype (Chen et al., 2007). These are promising examples of how, next to genetic diversity at different loci, the allelic diversity can also be exploited for improving resistance.

A challenge for the combination of different R genes or alleles is their potential functional incompatibility with the genetic background or among the stacked genes/alleles themselves. Incompatibility between resistance genes may result in autoimmunity and this hybrid necrosis sets a barrier for hybridization (Bomblies and Weigel, 2007). In contrast to this immunity-activating effect, the genetic background may also lead to loss of resistance activity. This resistance-suppression phenomenon is frequently observed and results in a significant limitation for resistance breeding, especially in polyploid crop species (e.g. Hanusová et al., 1996; Nelson et al., 1997; Knott, 2000; McIntosh et al., 2011; Chen et al., 2013; Liu et al., 2013 and references therein). Although this is a widespread problem, the underlying molecular determinants and mechanisms have remained elusive.

In this study we tested the allele-pyramiding approach for chromosome 1AS-localized Pm3, a CC-NLR-coding gene that mediates race-specific resistance against powdery mildew (Blumeria graminis f. sp. triticiBgt) in wheat and of which 17 functional alleles have so far been described (Yahiaoui et al., 2004, 2006; Srichumpa et al., 2005; Bhullar et al., 2009, 2010). We investigated the pairwise combination of five different alleles of Pm3 in F1 hybrids and stacked transgenic lines. We show that a quantitative suppression among the Pm3 alleles themselves frequently limits the efficiency of the resistance combination. We demonstrate that the suppression does not take place at the transcriptional or the translational level and that the suppression activity is delimited to the LRR domain. Our results provide molecular evidence that non-activated alleles of NLR resistance genes can block the activity of their resistant counterparts, which has major consequences for their use in crop resistance breeding.


Different Pm3 alleles cannot be stably combined in one genotype by classical genetics. Therefore we wanted to explore pyramidization of transgenic Pm3 alleles to improve resistance to powdery mildew. First, the additive or non-additive action of different Pm3 alleles was tested in the F1 progeny of crosses between lines/cultivars (cv.) that carry different alleles of Pm3. Wheat cv. ‘Kolibri’ carrying Pm3d or cv. ‘Michigan Amber’ (M. Amber) carrying Pm3f were crossed with the landrace ‘Chul’ carrying Pm3b. The presence of the two different Pm3 alleles in the F1 plants was confirmed by PCR amplification of allele-specific Pm3 markers (Tommasini et al., 2006). For leaf segment infection tests of the F1 hybrid plantlets we selected powdery mildew isolates that differentiate the resistance specificity of the Pm3 alleles (Figure 1a): isolates Bgt 97011 and Bgt 98229 are avirulent on wheat differential lines for Pm3d and Pm3f (AvrPm3d, AvrPm3f) but virulent on Pm3b differential lines (avrPm3b). In contrast, the isolates Bgt 07298 and Bgt 07201 are avirulent on Pm3b (AvrPm3b) but virulent on Pm3d (avrPm3d) or Pm3f (avrPm3f), respectively. Previous studies have shown that Pm3a-f are dominant alleles of the Pm3 resistance gene (Briggle, 1966; Zeller et al., 1993). Assuming additive gene action, we expected the F1 plants which carry both Pm3 alleles in a heterozygous state to be completely resistant to all of the tested Bgt isolates. However, we observed low (10–37% infected leaf area for F1 Chul × Kolibri) to high (60–96% infected leaf area for F1 Chul × M. Amber) levels of infection at 7 days post-infection with the isolates Bgt 97011 and Bgt 98229 (both AvrPm3d/f) (Figure 1b,c). The F1 hybrids remained fully resistant towards the Pm3b-avirulent isolates Bgt 07298 or Bgt 07201. The parental cultivars of the crosses, ‘Kolibri’, ‘M. Amber’ and ‘Chul’, displayed complete resistance towards all matching avirulent isolates (Figure 1b,c). In addition, F1 hybrids between the non-Pm3-carrying line ‘Bobwhite SH 98 26’ and Pm3f line ‘M. Amber’ remained resistant towards the AvrPm3f isolate Bgt 97011 (Figure 1c). To investigate whether the genetic background contributes to the observed incomplete resistance in F1 hybrids we crossed Pm3b- (Chul/8*CC) and Pm3f-near isogenic lines (M. Amber/8*CC). Both lines have cv. ‘Chancellor’ (CC) as the recurrent parent. F1 plants from these crosses were analyzed by infection tests. As for F1 Chul × M. Amber, we observed full resistance of the F1 Chul/8*CC × M. Amber/8*CC plants towards the Pm3b-avirulent isolate Bgt 07201 (4% infected leaf area), but also high levels of susceptibility with the Pm3f-avirulent isolates Bgt 97011 and Bgt 98229 (77–97% infected leaf area) (Figure 1b). Overall, these results reveal normal gene function of Pm3b (Briggle, 1966; Yahiaoui et al., 2004) but show incomplete resistance in the investigated F1 hybrids containing two Pm3 alleles for Pm3d and Pm3f in one or two genetic backgrounds, respectively. These data suggest that Pm3d- and Pm3f-mediated resistance can be weakened by the quantitatively acting negative activity of a suppressor present both in ‘Chul’ and in the ‘Chul’-derived chromosomal regions of Chul/8*CC.

Figure 1.

Suppression among Pm3 alleles leads to incomplete resistance in F1 hybrids.

Hybrids originating from parental lines expressing the Pm3b, Pm3d or Pm3f allele are incompletely resistant to powdery mildew isolates that are avirulent on the Pm3d or Pm3f parental lines.

(a) Virulence profiles of Blumeria graminis f. sp. tritici (Bgt) isolates used in this study.

(b) Results of leaf segment infection tests with parental lines and F1 hybrids originating from crosses between cultivars expressing Pm3b, Pm3d and Pm3f. Three independent crosses per hybrid combination (1–3) were analyzed. Infections were done with Bgt isolates that differentiate the parental resistance specificities. Data for at least eight replicates are presented with box and whisker plots and the median values are shown with black horizontal lines.

(c) Pictures of leaf segment infection tests taken at 7 days post-infection showing suppression of Pm3f in F1 Chul × M. Amber. Bgt isolates 97011 and 07201 differentiate the Pm3b- and Pm3f-mediated resistance as shown by the resistance specificity of wheat differential lines Chul and M. Amber. Infections of cultivar Bobwhite SH 98 26 (BWS) and F1 BWS × M. Amber are shown as controls.

We used previously developed transgenic Pm3aHA, Pm3b, Pm3cHA, Pm3dHA and Pm3fHA lines to further study if the genetic background might account for the suppressed resistance. They all have the susceptible genetic background of the spring wheat line Bobwhite SH 98 26 where no Pm3 allele is present (Brunner et al., 2011, 2012). In addition, we generated transgenic lines Pm3bHA and Pm3bmyc in the same genetic background expressing the Pm3b allele with a C-terminally fused single hemagglutinin (HA) or c-myc (myc) epitope tag, respectively. All the lines exhibited race-specific powdery mildew resistance over multiple generations.

We used the transgenic Pm3 lines that carry a single Pm3 allele inserted at random sites in the genome to pyramid the Pm3 alleles in pairs in Pm3 double-homozygous lines (Pm3x/y) that stably inherit two Pm3 alleles. One to three independent crosses between Pm3 lines were made and the F1 progeny were allowed to self-pollinate for three more generations (F4). The segregation of the individual Pm3 alleles was analyzed in the F3 or F4 generations with allele-specific Pm3 markers, and double-homozygous lines were selected. For the combination of Pm3bmyc with Pm3fHA we additionally selected the corresponding sister lines in the F3 generation, i.e. null segregants for Pm3fHA [Pm3bmyc/(ΔfHA)] or Pm3bmyc [Pm3(Δbmyc)/fHA] that are homozygous for either Pm3bmyc or Pm3fHA, respectively.

The resistance of all double-homozygous lines to powdery mildew was examined along with their parental lines (and sister lines for Pm3bmyc/fHA) in infection tests with three Bgt isolates that differentiate the parental resistance specificities. All the parental lines and sister lines exhibited the expected resistance specificities and were either completely resistant (<3% average infected leaf area observed) or completely susceptible (>66% average infected leaf area observed) towards the tested Bgt isolates (Figure 2a and Figure S1 and S2 in Supporting Information). In total, seven allele combinations were analyzed in which Pm3b combinations were redundantly investigated using the untagged, HA-tagged, and myc-tagged Pm3b-fusion variants. Remarkably, Pm3cHA/dHA was highly susceptible to Bgt 07298 (avrPm3d and AvrPm3c); Pm3aHA/b, Pm3aHA/bmyc, Pm3b/fHA, Pm3bHA/fHA, Pm3bmyc/fHA and Pm3cHA/fHA showed intermediate resistance or high susceptibility to Bgt 97011 and also in some cases lower resistance to Bgt 98229 (both avrPm3b/c and AvrPm3a/f) (Figures 1a, 2 and S1). Here, inoculations with the isolate Bgt 97011 always led to a higher level of infection compared with the isolate Bgt 98229. This shows that the degree of susceptibility in the affected lines depends on the Bgt isolate used for infection. This indicates that the diverse virulence potential of different Bgt isolates quantitatively influences the fungal infestation on suppressed Pm3 lines. Summarizing these observations, we found incomplete resistance with one or two isolates in infection tests with Pm3 double-homozygous lines of four allele combinations. In all these cases only the resistance mediated by one of the two combined Pm3 alleles was compromised. This is consistent with the observations in the F1 hybrids of the crosses of Pm3 cultivars (Figure 1b). From these results we infer that the incomplete dominance observed with the F1 hybrids of some Pm3 cultivar crosses was not due to influences of the genetic background. Instead the incomplete resistance in the F1 plants and the double-homozygous lines is based on negative epistatic effects between the Pm3 alleles themselves. Interestingly, a reduction of Pm3fHA-mediated resistance was observed in all three independent combinations of Pm3b with Pm3f, but the degree of susceptibility varied depending on the parental Pm3b line (e.g. Pm3b/fHA had 22% infected leaf area for Bgt 97011, Pm3bHA/fHA had 46% and Pm3bmyc/fHA had 77%). Similar observations were made for the two combinations of Pm3a with Pm3b (Pm3aHA/b and Pm3aHA/bmyc). Collectively, this dependence on the particular transformed construct or transgenic event shows the quantitative nature of suppression.

Figure 2.

Pm3-mediated resistance is suppressed in several Pm3-double-homozygous lines and the level of fungal infestation on the suppressed plants depends on the Blumeria graminis f. sp. tritici (Bgt) isolates.

(a) Leaf segment infection tests with Bgt isolates 97011, 98229 and 07201 are shown. These isolates differentiate the Pm3bmyc- and Pm3fHA-mediated resistance as shown by the resistance specificity of the Pm3bmyc/(ΔfHA) and Pm3(Δbmyc)/fHA sister lines. Pictures were taken at 7 days post-infection.

(b) Summary of results of leaf segment infection tests with transgenic lines homozygous for two different Pm3 alleles and differentiating Bgt isolates. The arrow colors indicate mean infected leaf areas of the double-homozygous lines for the tested Bgt isolates and the corresponding suppression levels of the double-homozygous line compared with the respective resistant parental line (light gray boxes). The direction of the arrows indicates the suppression activity of the Pm3 allele on the base of the arrow on the resistance activity originating from the Pm3 allele on the tip of the arrow. Detailed results of infection tests are presented in Figures S1 and S2.

We did not detect suppression for the combinations of Pm3a with Pm3c, Pm3a with Pm3d, and Pm3b with Pm3d: separate infections of the respective double-homozygous lines with three Bgt isolates resulted in no or very little infestation (<6% average infected leaf area) (Figures 2b and S2). Thus, while we detected incomplete Pm3d resistance with the F1 Chul × Kolibri hybrids (Figure 1b) we did not detect suppression in the Pm3b/dHA, and Pm3bmyc/dHA double-transgenic lines. We explain this discrepancy by the quantitative character of suppression where the suppression may be too weak to cause an obvious loss of resistance in these combinations of Pm3-overexpressing transgenic lines, but may be sufficient to compromise the resistance of F1 hybrids expressing the Pm3 alleles under native conditions. However, it is also possible that suppression is only occurring in a subset of Pm3-allele combinations and, therefore, does not affect the allele combinations where we observed additive resistance.

To characterize the molecular basis of the suppression effects we selected the Pm3bmyc/Pm3fHA combination for which we found strong Pm3f suppression and for which the different epitope tags enable allele-specific protein analyses. To test whether the Pm3 activity is affected by transcriptional silencing we performed Pm3fHA and Pm3bmyc allele-specific reverse transcription, quantitative real-time PCRs (RT-qPCR) with the double-homozygous lines Pm3bmyc/fHA and their sister lines from all three crosses. The Pm3 alleles are expressed from a strong maize ubiquitin promoter in all these lines. From six experiments, two independent ones on each of the three crosses, we detected a minimal (2.1-fold) but significant difference of Pm3fHA-expression levels in Pm3bmyc/fHA lines compared with the Pm3fHA-expressing sister lines Pm3(Δbmyc)/fHA in only one case (Figure 3a). The only significant reduction of Pm3bmyc expression in Pm3bmyc/fHA lines compared with the respective Pm3bmyc/(ΔfHA) sister lines was measured in the same experiment and for the same cross (two-fold reduction) (Figure S3). Given that Pm3fHA-mediated resistance is suppressed in all Pm3bmyc/fHA lines of all three crosses and that Pm3bmyc-mediated resistance is not affected in the Pm3bmyc/fHA line where we detected a reduced expression, there is no correlation between the resistance phenotypes and differences in Pm3 expression. Therefore, we conclude that suppression of Pm3fHA in the Pm3bmyc/fHA lines is not based on transcriptional silencing.

Figure 3.

Suppression of Pm3fHA activity in double-homozygous Pm3bmyc/fHA lines occurs neither at the transcriptional nor at the translational level.

(a) Reverse transcription, quantitative real-time PCR results for the incomplete resistant Pm3bmyc/fHA lines and the resistant Pm3(Δbmyc)/fHA sister lines show only minimal differences of Pm3fHA-transcript accumulation. Mean values (represented by bars) and single data points (crosses) of two independent experiments each with three replicates are shown. Values are normalized to the lowest value (set to 1) within each experiment. Significant differences are indicated by asterisks (Student's t-test < 0.05; n.s., non-significant).

(b) Immunoblot (IB) analysis of PM3Bmyc (upper panel) and PM3FHA protein (lower panel) in leaves of Pm3bmyc/fHA lines and the respective sister lines from three independent crosses. Ponceau S membrane staining of ribulose-1,5-bis-phosphate carboxylase/oxygenase is shown as control for equal loading of total protein.

Next, we wanted to test whether reduced protein abundance might be the reason for the suppression of Pm3fHA-mediated resistance. With immunoblots using anti-myc or anti-HA antibodies, respectively, we separately analyzed the PM3Bmyc or PM3FHA proteins in the double-homozygous Pm3bmyc/fHA lines in comparison with the sister lines. Similar band intensities indicate that similar levels of PM3Bmyc and PM3FHA protein are produced in the leaves of Pm3bmyc/fHA lines and the corresponding sister lines (Figure 3b). This suggests that the incomplete PM3FHA-mediated resistance phenotype in Pm3bmyc/fHA is not correlated with a reduced amount of the PM3FHA resistance protein and indicates a suppression mechanism at the post-translational level.

Some CC-NLR resistance proteins are known to form multimeric complexes before pathogen perception (Ade et al., 2007; Maekawa et al., 2011). Based on these findings, we presumed that protein interactions might be important for the suppression mechanism, and for this reason we analyzed whether protein complexes containing different PM3 proteins can be found in plant cells. We performed co-immunoprecipitation experiments with Pm3bmyc- and Pm3fHA-co-infiltrated leaf material from Nicotiana benthamiana in which we had previously shown that PM3 is functional (Stirnweis et al., 2014). Here, myc-tagged PM3Bmyc co-precipitated with HA-tagged PM3FHA protein demonstrating that these two proteins interact (Figure 4a). No similar interaction was detected with the same analysis using primary leaves of the stable transgenic Pm3bmyc/fHA wheat lines. There, levels of PM3 protein are very low and the immunoprecipitation and detection efficiency for PM3FHA or PM3Bmyc, each fused with only a single epitope, may be insufficient. Using N. benthamiana, we also tested for the interaction of PM3FHA with myc-tagged hPM3-1Bmyc. This protein is encoded by a homolog of Pm3 originating from wheat homoeologous chromosome 1B and has 78% similarity to the amino acid sequence of PM3B (Hurni et al., 2013). The hPm3-1B gene is present in cv. ‘Chancellor’ that was used as recurrent parent in many near-isogenic Pm3 differential lines, and is therefore expected not to suppress Pm3-mediated resistance. The detection of hPM3-1Bmyc protein in the PM3FHA precipitate showed that PM3FHA and hPM3-1Bmyc are also present in a common protein complex (Figure 4a). This indicates that protein interaction with a PM3-like protein per se is not sufficient for the suppression of PM3FHA.

Figure 4.

Agroinfiltration experiments in Nicotiana benthamiana reveal that PM3Bmyc and its homolog hPM3-1Bmyc physically interact with PM3FHA, but only PM3Bmyc suppresses PM3FHA-mediated hypersensitive response (HR) via its leucine-rich repeat (LRR) domain.

(a) Co-immunoprecipitation of c-myc-tagged PM3B or hPM3-1B with hemagglutinin (HA)-tagged PM3F from co-infiltrated N. benthamiana leaves. Images of immunoblots (IB) before (two upper panels) and after (two lower panels) immunoprecipitation (IP) with anti-HA agarose beads are shown.

(b) Overlapping infiltrations of autoactive, HR-inducing Pm3f_D501VHA (upper circles) with Pm3bmyc, the negative control GUS, and hPM3-1Bmyc (lower circles) for the detection of phenotypic interactions in the overlap. Pictures were taken at 5 days post-infiltration (dpi).

(c) Deletion analysis for PM3Bmyc suppression activity. Constructs for PM3Bmyc fragments and Pm3f_D501VHA were co-infiltrated and suppression of the HR was scored in comparison with a co-infiltration with the negative control (GUS) at 5 dpi in at least three independent experiments with at least four replicates per experiment (0, no HR suppression; ++++, consistently complete HR suppression). Exemplary pictures are shown. Due to variable phenotypes these pictures cannot be representative for all co-infiltrations. The top row of the drawing shows the PM3 domain structure with the CC (blue), NB (red), Spacer (Sp, gray) and LRR (green) domains. Black bars indicate the portion of PM3B encoded by the respective construct. Numbers give the respective positions in PM3B of the outermost amino acids of the fragments.

To examine whether the suppression of Pm3 is independent of factors from the powdery-mildew pathogen we established an assay in Nicotiana that allows us to investigate the phenotypic aspects of interactions of PM3 in the absence of mildew. We performed overlapping infiltrations with Agrobacterium tumefaciens strains transferring either the construct under investigation or Pm3f_D501VHA, a version of Pm3f coding for an autoactive form of the protein due to a mutation in the MHD motif. This aspartate-to-valine substitution renders many resistance proteins, including PM3, autoactive which leads to the induction of a hypersensitive response (HR) after agroinfiltration (Stirnweis et al., 2014). Hence, the programmed cell death that can be observed after agroinfiltration of Pm3f_D501VHA resembles a resistance response activated by perception of an avirulent powdery mildew isolate. The PM3F_D501VHA-induced cell death was completely suppressed by PM3Bmyc in the infiltration overlap at 5 days post-infiltration (dpi) while the negative control GUS did not reduce the PM3F_D501VHA-mediated HR in the overlapping infiltration zone (Figure 4b). This indicates that the suppression of PM3F by PM3B is independent of components from the powdery mildew fungus. We also found that PM3F_D501VHA-mediated HR was not markedly influenced in overlapping infiltrations with hPm3-1Bmyc (Figure 4b). This is in accordance with the observation that hPm3-1B does not appear to interfere with Pm3-mediated resistance in wheat. An immunoblot analysis of the proteins in leaf material co-infiltrated with Pm3f_D501VHA and Pm3bmyc or hPm3-1Bmyc and harvested at 43 h post-infiltration (hpi) shortly before the onset of the HR shows that the PM3F_D501VHA levels did not significantly differ between infiltrations with and without an active suppressor gene (Figure S4a). Overall, these results demonstrate that the Nicotiana system recapitulates the suppression effects observed in wheat and reveal that intrinsic protein properties make the difference between the suppressing PM3B and the non-suppressing hPM3-1B.To investigate which part of the PM3B protein causes suppression we co-infiltrated constructs for fragments of PM3Bmyc and for PM3F_D501VHA in N. benthamiana and examined the HR at 5 dpi. Co-infiltration of Pm3f_D501VHA with the Pm3b_CC-NBSmyc construct comprising amino acids (aa) 1–602 of PM3B led to a HR that was at least as intense as with the GUS negative control, indicating that the CC-NBS domains are not responsible for the suppression but rather enhance the PM3F_D501VHA-induced HR (Figure 4c). In contrast, the Pm3b_LRRmyc (aa 525–1415) construct comprising the complete LRR domain very efficiently suppressed PM3F_D501VHA-mediated HR. We split the LRR into an N-terminal (Pm3b_Sp-LRR15myc, aa 525–983) and C-terminal part (Pm3b_LRR15-ENDmyc, aa 949–1415) and observed that the HR-suppression property is encoded in the N-terminal fragment. When we shortened the N-terminal fragment, co-infiltration with the construct Pm3b_Sp-LRR12myc (aa 525–879) still showed a strong reduction of PM3F_D501VHA-induced HR, whereas with shorter fragments HR suppression was only rarely observed (Pm3b_Sp-LRR10myc, aa 525–826) or not detected (Pm3b_Sp-LRR8myc, aa 525–774) (Figure 4c). This gradual loss of suppression activity from Pm3b_Sp-LRR15myc to Pm3b_Sp-LRR8myc may have its origin in the ever shorter size of the fragment or may reflect the importance of the PM3B aa 826–983 for the suppression. An observation that supports the latter hypothesis was that the Pm3b_LRR10-ENDmyc construct (aa 826–1415) frequently displayed suppression in contrast to the Pm3b_LRR15-ENDmyc construct. The construct Pm3b_LRR10-15myc (aa 826–983) was still not sufficient to suppress the PM3F_D501VHA-induced HR. Immunoblot analysis of the protein levels in co-infiltrated Nicotiana leaves shortly before the onset of HR (about 30 hpi for Pm3b_CC-NBSmyc) showed that the abundance of PM3F_D501VHA is not significantly altered by suppressing and non-suppressing fragments and all PM3B fragments formed stable proteins (Figure S4b). In addition, infiltrations without Pm3f_D501VHA showed that none of the Pm3b constructs induced HR by themselves (Figure S5). In summary, we infer from the deletion analysis that the suppression activity of PM3B towards PM3F is situated in the Spacer-LRR domain and here the N-terminal half plays the major role.


The pyramiding of resistance genes or their alleles in a single genotype by the generation of F1 hybrids or genetically stable non-segregating lines is a promising concept for the combination of gene specificities and effectiveness, and for an extension of their durability (McDonald and Linde, 2002; Dangl et al., 2013). Combinations of resistance loci leading to additive gene action have been reported in a number of plant species (e.g. Liu et al., 2000; Hu et al., 2012; Zhu et al., 2012) and the successful stacking of alleles was also reported for two NLR resistance genes (Bieri et al., 2004; Chen et al., 2007). However, there have been a number of reports describing observations of weakened or lost resistance when the source of resistance is introgressed into another genetic background. This is well described for resistance breeding in polyploid crop plants where resistance loci derived from lower-ploidy species of the secondary or tertiary gene pool are often suppressed in the polyploid species or in synthetic polyploids (e.g. Hanusová et al., 1996; Nelson et al., 1997; Knott, 2000; McIntosh et al., 2011; Chen et al., 2013; Liu et al., 2013 and references therein). Incomplete resistance in F1 hybrids, as seen in this study for the Pm3 F1 hybrids, is also known (e.g. Islam et al., 1992; Wilson and McMullen, 1997; Kim et al., 2012) but the causes were often attributed to a gene-dosage dependency. Only few studies so far have given hints that the suppression of a resistance locus might originate from its combination with the corresponding, dominantly acting, susceptible allelic locus: For example, it was shown in Arabidopsis thaliana that the TIR-NLR-WRKY-resistance gene Rrs1, originally classified as recessive by classical genetics, behaved as a dominant gene when introduced as a transgene (Deslandes et al., 2002). A rust resistance gene of soybean showed dominance in some crosses but recessiveness in others. There it was found that the genetic determinant of the suppression co-segregated with the allelic, susceptible resistance locus (Garcia et al., 2011). The results of our study now demonstrate at the molecular level that incompatibility among alleles of an NLR resistance gene can cause suppression of resistance.

Combining these findings with those of the companion publication (Hurni et al., 2014), showing that rye-derived Pm8-mediated resistance in wheat can be suppressed by its ortholog Pm3, also suggests that other NLR resistance activities might be compromised by closely related NLR proteins (e.g. encoded by alleles, orthologs, homoeologs or paralogs) by the same suppression mechanism. Indeed, there are several indications that the identified mechanism (how a non-functional resistance protein suppresses a resistant counterpart) is of wider significance.

For instance, Nelson et al. (1997) genetically mapped the suppressor of chromosome 2B-localized leaf-rust resistance gene Lr23 in hexaploid wheat to the homoeologous locus on chromosome 2D, suggesting a susceptible homoeolog of Lr23 as suppressor. Furthermore, it was shown that expression of a version of the bacterial NLR-resistance protein RPS2 that is inactivated by mutations in the CC domain has a dominant-negative effect on the wild-type RPS2-mediated resistance in Arabidopsis thaliana (Tao et al., 2000). Moreover, the viral TIR-NLR-resistance protein N of Nicotiana is suppressed by co-expression of N variants inactivated by P-loop mutations, or by deletion of, or mutations in, the TIR domain (Dinesh-Kumar et al., 2000). Similar to our observation for PM3, it was also reported that cell death induced by an autoactive version of the NLR Prf can be suppressed by co-expression of its LRR in a Nicotiana infiltration system (Du et al., 2012). Finally, the described suppression mechanism might also be relevant for resistance proteins with an extracellular LRR (eLRR) domain. This hypothesis is based on results of Barker et al. (2006) showing that inactive Cf-9 variants with C-terminal deletions in the eLRR have a dominant-negative effect on the wild-type Cf-9 activity in tomato. These examples are all consistent with our findings from the deletion analysis showing that the N-terminal part of the LRR domain is the major determinant of suppression.

Recently, Williams et al. (2014) showed for the RPS4/RRS1-NLR pair that heteromeric, TIR-domain independent complexes between these NLR proteins are formed and that co-expression of the RRS1-TIR domain is sufficient to suppress the HR induced by expression and homodimerization of the RPS4-TIR domain. Our results and data from Du et al. (2012) both show that co-expression of the LRR domain is sufficient to suppress the HR induced by an autoactivated full-length NLR protein. These data suggest that multiple domains are potentially involved in the co-suppression and co-activation of NLR-protein complexes (Williams et al., 2014).

The results of this study present molecular evidence that suppression of Pm3 is based on dominant-negative, post-translational effects among the proteins involved. These effects are of a quantitative nature, as indicated by suppression differences between different Pm3b constructs. They are most likely independent of fungal components, as suggested by the HR suppression in the Nicotiana infiltration system, and could involve interactions of PM3 proteins. Thus, we propose the following model for the suppression by alleles: PM3 proteins form complexes, exclusively PM3-homomeric ones when only one allele is present and PM3-homomeric as well as PM3-heteromeric ones when multiple alleles are present. In contrast to homomeric complexes, heteromeric complexes might be incompatible for signaling, or even block it, thereby sequestering the active protein pool of each combined PM3 variant. This also provides an explanation for the quantitative nature of suppression where the ultimate phenotypic outcome depends on the virulence potential of the pathogen race, the individual PM3 protein level and the recognition and activation efficiency of the PM3 proteins encoded by different alleles (Stirnweis et al., 2014). This scenario also implies that the successful pyramiding of the Pm3a and Pm3c, Pm3a and Pm3d, and Pm3b and Pm3d alleles is based on limited quantitative suppression due to an optimal combination of PM3-protein levels. The observation that hPM3-1B does not suppress PM3, even though the proteins interact as shown by co-immunoprecipitation, indicates that the inactivation of PM3-protein complexes is a complicated process that possibly also depends on particular protein features which differ between PM3B and hPM3-1B.

Overall, the suppression mechanism found in this study is possibly widespread, especially in polyploid species where resistance genes might not only be suppressed by other alleles but also by the homoeologous genes located on the different subgenomes. It can be a limiting factor for gene- or allele-pyramiding approaches as well as for the transfer of resistance genes into species where an ortholog is present. The results of this study suggest that in such cases the mutagenesis, silencing, gene editing or replacement of the orthologous suppressor gene is a possibility to bypass the unwanted suppression of resistance. The introduced N. benthamiana infiltration system offers an opportunity to easily verify suppression activities between cloned genes.

Experimental procedures

Transgenic Pm3 lines

Five transgenic lines were previously described by Brunner et al. (2011, 2012) and were renamed in this study: Pm3aHA corresponds to Pm3a#1, Pm3b to Pm3b#1, Pm3cHA to Pm3c#1, Pm3dHA to Pm3d#1, and Pm3fHA to Pm3f#1. The cloning and transformation procedures for the Pm3bmyc and Pm3bHA transgenic lines are described in Methods S1.

Selection of double-homozygous and sister lines

For the generation of Pm3 double-homozygous lines up to three crosses between individual plants from two different Pm3 transgenic lines were made. The resulting F1 progeny was allowed to self-pollinate for three more generations (F4). The segregation of the individual Pm3 alleles was analyzed in the F3 or F4 generations with allele-specific Pm3 markers and double-homozygous lines were selected based on segregation analysis using at least 20 individual plants per family of each line. The Pm3bmyc/fHA lines as well as the sister lines Pm3bmyc/(ΔfHA) and Pm3(Δbmyc)/fHA, null segregants for Pm3fHA or Pm3bmyc that are homozygous for either Pm3bmyc or Pm3fHA, respectively, were selected based on marker analysis with at least 30 individuals of one family in the F3 generation.

Pm3 marker

Conditions and primers for allele-specific Pm3 markers were used as described (Tommasini et al., 2006). For the detection of the Pm3b and Pm3d alleles in the transgenic context the marker primers had to be modified: For Pm3b the primer sbi144 (5′-TTTAGCCCTGCCTTCATACG-3′) was combined with the primer Pm3b/R (Tommasini et al., 2006); for Pm3d the primers dst003 (5′-AGATGGCAAGCAAGAGGTGT-3′) and dst004 (5′-CAAGCTTAATGCACCCACGA-3′) were used.

Infection tests

Powdery mildew infection tests using leaf segments were performed as previously described by Brunner et al. (2011). Box and whisker plots of the obtained data in Figures 1b,  S1 and S2 were created with the software package R (R Core Team, 2013).

Reverse transcription, quantitative real-time PCR analysis for detection of Pm3fHA and Pm3bmyc expression

Expression of Pm3fHA and Pm3bmyc was separately quantified using a reverse transcription, quantitative real-time polymerase chain reaction assay. For each line, technical triplicates of three biological replicates were analyzed using a CFX96 Real-Time System C1000® Thermal cycler (Bio-Rad, Each biological replicate consisted of three pooled first leaves of 10-day-old plants. GAPDH (UniGene Ta.5104) was included as reference gene. For a more detailed description see Methods S1.

Protein extraction and immunoblot analysis

Protein from primary leaves of wheat was detected as essentially described by Brunner et al. (2012) but using the Chemidoc XRS system (Bio-Rad) for blot development instead of x-ray film. Protein detections from N. benthamiana leaves at the indicated time post-infiltration were performed as described by Stirnweis et al. (2014). Anti c-myc antibodies (rat monoclonal, clone JAC6, sc-56633; Santa Cruz Biotechnology, were used in 1:4000 dilution for the detection of c-myc tagged proteins.

Construction of plasmid vectors for agroinfiltrations

Complementary DNA (cDNA) was synthesized on total RNA of cv. Chul or M. Amber/8*CC with the SuperScript III RT (Life Technologies, enzyme according to the manufacturer's protocol, Pm3b or Pm3f were amplified by PCR with primers TJ065 (5′-TTGGCGCGCCGCGGATGGCAGAGCGGGTGGTCA-3′) and TJ066 (5′-CCCCCCGGGCGGCCGCTCAGCTCCGGCAGGCC-3′) and were cloned with the StrataClone Blunt PCR Cloning Kit (Agilent Technologies, From these and from existing plasmids all genes were cloned into Gateway-compatible entry vectors via Gateway BP Clonase II reactions (Life Technologies). Introduction of modifications and cloning of fragments were achieved by the QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies). By Gateway LR reactions (Life Technologies) all resulting pENTR plasmids were recombined to the binary vector pIPKb004 (Himmelbach et al., 2007) carrying the double-enhanced cauliflower mosaic virus 35S promoter. Detailed primer and cloning information is given in Tables S1–S3.

Agroinfiltrations and co-immunoprecipitation

Transient expressions of vector constructs in N. benthamiana leaves via A. tumefaciens infiltrations and co-immunoprecipitation experiments were performed according to the protocols of Stirnweis et al. (2014). For overlapping infiltrations, A. tumefaciens clones containing the Pm3bmyc, hPm3-1Bmyc, or GUS constructs were infiltrated first, and the infiltrations of Pm3f_D501VHA were done 1–2 h later.


We thank Serverine Hurni for her collaboration on this project. Jochen Kumlehn is acknowledged for providing the pIPKb004 vector. This work was supported by an Advanced Investigator grant from the European Research Council (ERC-2009-AdG 249996, Durable resistance) and by Swiss National Science Foundation grant 310030B_144081/1.