Results

Epistatic Interactions Modulate Plant Growth in Response to Temperature.

Different RILs derived from crosses between the Arabidopsis accessions Landsberg erecta (Ler) and Cvi, Sha, An-1, Eri-1, Kas-2, or Kond (12) were grown at moderate (20 °C) and low (14 °C) temperature. Reduction of rosette area in response to low temperature was most obvious among Ler × Kas-2 and Ler × Kond RILs. A subset of RILs (15% from Ler × Kas-2 and 24% from Ler × Kond; supporting information (SI) Fig. S1) were dwarf at 14 °C but resembled parental lines at 20 °C. We therefore selected these 2 populations for QTL detection and further study. Common QTL for rosette area were detected among these populations on the bottom of chromosome 3 (QTL 3) and the top of chromosome 4 (QTL 4) (Fig. 1A). In both populations, Ler alleles on QTL 3 contributed to a reduction of rosette area, whereas Ler alleles on QTL 4 produced the opposite effect. Because the dwarf phenotype was observed in RILs but not in the parental lines, we reasoned that the dwarfism likely results from an epistatic interaction between parental alleles. In both populations a common epistatic interaction between 2 loci was found (Fig. S2). These loci are located on QTL 3 and QTL 4. An additional epistatic interaction was revealed in Ler × Kas-2 involving loci located on QTL 4 and at the top of chromosome 5 (QTL 5) (Fig. 1A and Fig. S2). Together, the data indicate that a 3-way interaction between QTL 3, QTL 4, and QTL 5 in Ler × Kas-2 and a 2-way interaction between QTL 3 and QTL 4 in Ler × Kond condition variation of rosette area. To test this hypothesis, an ANOVA test was performed. The 3-way interaction in Ler × Kas-2 explained 35.5% (P_value < 0.001) and the 2-way interaction in Ler × Kond 36.3% (P_value < 0.001) of the variation in rosette area, and only a specific allele combination lead to dwarf plants (Fig. 1B). We concluded that the main-effect QTL are interacting. We refer to lines carrying such allele combinations as incompatible.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Detection of QTL of rosette area and cell death at low temperature. (A) LOD trace for the QTL detection involved in variation of rosette area and cell death at low temperature in Ler × Kas-2 and Ler × Kond RIL. (B) Rosette area and cell death score of Ler × Kas-2 (Top) and Ler × Kond (Bottom) RILs sorted according to the allelic values at interacting loci QTL 3, QTL 4, and QTL 5. Values with different letters are significantly different at level P < 0.001 in a Student-Newman Keuls test (SNK). n, number of RIL from each class; bars, SD.

The allelic forms that result in dwarfism on QTL 3 and 4 behaved recessively in both populations, whereas the incompatible Kas-2 allele on QTL 5 was dominant. All crosses derived from incompatible Ler × Kas-2 and Ler × Kond RILs resulted in dwarf F₁ plants at low temperature. This phenotype and the recessive nature of QTL 4 alleles suggest that the dwarf phenotype is controlled by the same loci in both populations.

Additional Modifiers in the Genetic Background Influence Growth of Incompatible Lines.

Even though all of the incompatible lines exhibited dwarfism at low temperature, some variation was observed among them, reflecting incomplete penetrance of the dwarf phenotype in certain lines (Ler × Kas-2 RIL 4 in Fig. 2A). We concluded that additional modifiers influence growth traits. We developed a near isogenic line (NIL) harboring a Ler introgression on QTL 3 in a homogeneous Kas-2 genetic background (Fig. S3). The NIL grown at low temperature exhibited severe stunting that was only partially suppressed by moderate temperature (Fig. 2A). Hence, the Kas-2 background in incompatible Ler × Kas-2 lines exacerbates growth defects in response to temperature. The flowering time of the NIL was delayed by weeks compared with parental lines when grown on soil at 14 °C (Fig. 2B) but resembled the Kas-2 parent when grown at 20 °C (Fig. 2C). However, the NIL was ultimately able to flower even under nonpermissive conditions (14 °C) and produce viable seeds. Hence, NIL plants harboring the incompatible allele interaction can still produce progeny and thus overcome complete reproductive isolation. Dwarfism was also suppressed by growing incompatible lines in vitro under high humidity (Fig. 2D).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Developmental phenotypes of incompatible lines. (A) Five-week-old Ler × Kas-2 incompatible lines (14, 4, and NIL) and parental accessions Ler and Kas-2 at 14 °C (Left) and 20 °C (Right). (B) Eight-week-old NIL at 14 °C and at 20 °C (C) compared with Ler and Kas-2 parents. (D) Two-week-old NIL grown at 14 °C under high humidity conditions.

An Interacting Locus on Chromosome 3 Maps to a RPP1-Like Cluster.

A total of 768 F₂ plants generated from the cross between Ler × Kas-2 incompatible lines and Kas-2 and segregating for QTL 3 were used to fine map the locus to a 71.4-kb interval between genes At3g44600 and At3g44700 (based on Col sequence; www.arabidopsis.org). In Col, this region contains 11 genes, including 2 RPP1-like TIR-NBS-LRR genes (13) and 3 transposable elements. A set of 270 F₂ plants derived from a Col and Kas-2 cross were grown at low temperature and genotyped. Lines providing Col alleles on QTL 3 and Kas-2 on QTL 4 and 5 did not show stunting, indicating that Col alleles are compatible with Kas-2 and that polymorphism(s) between Ler and Col lead to incompatibility with Kas-2. The incompatible phenotype was also evident in Landsberg ERECTA × Kas-2 F₂ lines (Table S1), thus the Ler allele on QTL 3 represents a natural allele that was present in the parental Landsberg accession. A BAC spanning QTL 3 in Ler was isolated from the Ler BIBAC library (14) and sequenced. Compared with the compatible Col haplotype, the Ler sequence contains a genomic insertion of 65 kb between At3g44620 and At3g44630 that has 5 genes encoding TIR-NB-LRR proteins (R1–R5 in Fig. 3) with sequence similarities to RPP1-like genes, and 1 gene with a predicted TIR- domain (R6; Fig. 3). Together with the RPP1-like At3g44630 and At3g44670 homologs in Ler (630 and 670; Fig. 3) a polymorphic cluster of 7 TIR-NB-LRR genes exists in Ler compared with 2 TIR-NB-LRR genes in Col (Fig. 3). The sequences of other genes on QTL 3 At3g44610, At3g44620, At3g44660, At3g44680, At3g44690, and At3g44700 were mostly conserved (Fig. 3). We concluded that one or more Ler RPP1-like gene homologs are the likely determinants of interacting QTL 3. Bomblies et al. (7) mapped an interacting QTL for temperature-sensitive hybrid necrosis to the same locus in a cross between Arabidopsis accessions Uk-1 and Uk-3. Hybrids from the cross between Ler and Kas-2 to Uk-1 and Uk-3 accessions were fully compatible in the F₁ and F₂ generations (Table S1). Thus, the Uk-1 × Uk-3 and Ler × Kas-2 allele incompatibilities map to an overlapping locus on chromosome 3 but differ in their allelic effects and epistatic interactions (Fig. S2) (7).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Comparison of the genetic architecture in QTL 3 between Ler and Col. Marker positions are indicated by arrows. Insertions are indicated in triangles (yellow in Ler, orange in Col). Duplications are shown in dotted blue lines. Genes and their orientations are represented. The 3 digits next to the genes are identifiers for the last corresponding AGI numbers (At3g44XXX). The Col At3g44630 and At3g44670 RPP1-like genes and homologs in Ler are shown in light blue. Inserted genes in Ler encoding TIR-NB-LRR proteins (R1-R5) or truncated forms (only TIR domain in R6) are indicated in dark blue. Transposable elements are shown in black boxes. C, copia; L, LINE; I, transposase IS4.

Low Temperature Deregulates Cell Death Programs in Incompatible Lines.

Temperature and humidity influence the activation of plant defenses against pathogens (11). Dependence of the RIL dwarf phenotypes on these environmental factors and our positioning of QTL 3 to a RPP1-like TIR-NB-LRR cluster implied that a deregulated TIR-NB-LRR disease-resistance pathway contributes to the Ler × Kond and Ler × Kas-2 incompatibilities. We examined dwarf Ler × Kas-2 RIL and NIL for the occurrence of cell death at low temperature by staining leaves with trypan blue (TB) (15). In all cases, growth at the higher temperature (20 °C) suppressed or attenuated cell death (Fig. 4A), consistent with a correlation between cell death and stunting of plant growth. Dispersed areas of necrosis were also seen in the parental Kas-2 line at 14 °C, although Ler plants did not initiate cell death at either temperature (Fig. 4A). Therefore, Ler and Kas-2 appear to have different thresholds for induction of cell death programs that are revealed at low temperature. Cell death also segregated in incompatible Ler × Kond RILs although neither parent displayed lesions at low temperature (Fig. S4).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Cell death and disease-resistance phenotypes of incompatible lines. (A) Spontaneous cell death revealed by TB staining of 3-week-old leaves of incompatible lines (4, NIL) and parental accessions grown at 14 °C (Left) or 20 °C (Right). (B) Infection phenotypes of the same lines as in (A). Two-week-old plants grown at 14 °C (Left) or 20 °C (Right) were inoculated with virulent H. parasitica isolate Cala2 at 18 °C. Microscopic examination of TB-stained leaves to reveal dead plant cells and pathogen mycelium growth was performed 4 days postinoculation. (Scale bars, 500 μm.)

QTL governing necrosis at 14 °C were mapped in the Ler × Kas-2 and Ler × Kond RIL populations. Four main-effect QTL were identified in Ler × Kas-2 and Ler × Kond RILs explaining, respectively, 36% and 42% of the phenotypic variation of cell death (Fig. 1A). Three cell-death QTL colocated at the same loci in both populations (Fig. 1A) and 2 colocated with previously identified interacting QTL involved in growth variation (QTL 3 and 4; Fig. 1A). The average cell death of the different genotypic classes determined by 3-way and 2-way epistatic interactions in Ler × Kas-2 and Ler × Kond was measured (Fig. 1B) and revealed that combinations of incompatible alleles giving dwarf phenotypes in both populations are also responsible for increased cell death at low temperature (Fig. 1B).

Incompatible Lines Exhibit Enhanced Pathogen Resistance.

We measured whether the low temperature-conditioned epistatic interactions correlate with increased disease resistance by inoculating Ler × Kas-2 RIL and NIL with an isolate of the biotrophic oomycete pathogen Hyaloperonospora parasitica (Cala2) (16) that infects both parents (Fig. 4B). Seedlings were cultivated at 14 °C or 20 °C for 2 weeks and then moved to 18 °C for inoculation, as this is optimal for the pathogen. H. parasitica Cala2 was able to colonize Ler and Kas-2 after growth of plants at low temperature, although pathogen development in Kas-2 was slower than in Ler, consistent with the occurrence of sporadic cell death in Kas-2 leaves at low temperature (Fig. 4 A and B). Pathogen colonization was strongly inhibited in all incompatible Ler × Kas-2 lines grown at 14 °C or at 20 °C, as shown for the moderately incompatible RIL 4 and strongly incompatible NIL (Fig. 4B). Notably, RIL 4 that exhibited no cell death at 20 °C (Fig. 4A) responded to pathogen infection by initiating cell death and resistance in a manner reminiscent of a hypersensitive response (17) (Fig. 4B). Lesioning in RIL 4 was dependent on the pathogen because mock-inoculated plants did not initiate cell death (data not shown). We concluded that pathogen infection probably reduces the threshold for activation of defense processes that also lead to hybrid incompatibility at low temperature.

Disproportionate SA Pathway Activation in Incompatible Lines.

We selected the Ler × Kas-2 RIL population for further characterization of plant defense pathway activation. Transcripts of genes regulated by oxidative stress (GST-1) (18), SA (EDS1, PR-1) (19), or jasmonic acid (JA; PDF1.2) (19) were quantified by real-time PCR in a set of lines grown at 14 °C and 20 °C. Expression of GST1, EDS1, and PR-1 was enhanced in all dwarf lines and in Kas-2 grown at 14 °C compared with 20 °C (Fig. S5). The Ler parental and semidwarf lines (RILs 4 and 6) exhibited a much less pronounced or no effect of temperature on accumulation of these transcripts. Thus, low temperature-induced expression of GST1, EDS1, and PR-1 in dwarf lines broadly correlates with the occurrence of cell death (Fig. 4A). Notably, JA pathway activation (as measured by PDF1.2 expression) was high in Kas-2 but not in the compatible or incompatible lines grown at low temperature (Fig. S5), pointing to a degree of SA and JA pathway deregulation in Kas-2. SA synthesis and/or signaling prevail in the incompatible lines at low temperature because EDS1 and PR-1 expression remained high but PDF1.2 expression was dampened (Fig. S5).

We measured whether the extent of incompatibility in the Ler × Kas-2 lines correlated with SA accumulation and found that low temperature increased levels of free and total SA in Kas-2 plants and the incompatible lines but not in Ler or the compatible lines (Fig. S6). Strikingly, the NIL accumulated 2- to 3-fold more free SA than Kas-2 at low and moderate temperature (Fig. S6) and displayed a suppression of JA signaling (as measured by PDF1.2 expression) that was strongest at low temperature (Fig. S5). The extreme phenotype of the NIL points to the importance of Ler alleles at QTL 3 for strong activation of the SA pathway in a Kas-2 genetic background.

SA Pathway Activation Is Necessary for Hybrid Incompatibility.

We investigated the contribution of SA accumulation to hybrid incompatibility by transforming the moderately incompatible Ler × Kas-2 RIL 4 with a constitutively expressed bacterial Salicylate Hydroxylase gene (NahG) that converts SA to catechol (19). Two independent homozygous NahG transgenic lines were selected that had detectable NahG expression by qRT-PCR (not shown). Incompatible RIL 4 plants have reduced leaf size compared with the parents at low temperature (Fig. 5A). In contrast, the rosette area of RIL 4-NahG plants was not significantly different to the parental lines (Fig. 5A). Both RIL 4-NahG lines were susceptible to H. parasitica infection at low temperature and did not exhibit cell death in response to the pathogen in contrast to the nontransformed RIL 4 (Fig. 5B). Therefore, SA accumulation is necessary for growth retardation and the enhanced resistance response of RIL 4.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Suppression of incompatibility by SA depletion. (A) Percent rosette area relative to Ler of 3-week-old plants: incompatible line RIL4 at 14 °C (RIL4), F₁ progeny from a cross between RIL4 and NIL (RIL4 × NIL), RIL4 line transformed with Salicylate hydroxylase (RIL4-NahG), and F₁ progeny from a cross between RIL4-NahG and NIL (RIL4-NahG × NIL). (B) Infection phenotypes of RIL4, RIL4-NahG, RIL4 × NIL, and RIL4-NahG × NIL lines (see [A] for details of lines). Two-week-old plants grown at 14 °C were inoculated with virulent H. parasitica isolate Cala2 at 18 °C. Plant cell death and pathogen colonization was monitored as in Fig. 4. (Scale bars, 500 μm.)

RIL 4 and RIL 4-NahG were crossed with the extreme dwarf Ler × Kas-2 NIL. F₁ progeny from each cross differed only in the presence or absence of NahG transgene and were homozygous at the incompatible interacting loci (QTL 3, 4, and 5). Growth and resistance phenotypes of the F₁ hybrids were analyzed at low temperature. RIL 4 × NIL F₁ hybrids were dwarf (Fig. 5A) and resistant to H. parasitica infection (Fig. 5B), whereas RIL 4-NahG × NIL F₁ hybrids were similar in size to the parental lines (Fig. 5A), did not exhibit cell death, and became susceptible to pathogen infection (Fig. 5B).

Incompatible phenotypes were also suppressed by depletion of SA through introduction of the sid2–1 mutant defective in the isochorismate synthase 2 gene responsible for pathogen-induced SA biosynthesis (20). Quadruple homozygous lines harboring the incompatible allelic interaction on QTL 3, 4, and 5 in a sid2 genetic background were isolated. These lines were not dwarf at low temperature and did not show spontaneous cell death (Fig. S7 B and D), whereas a wild-type SID2 allele conferred stunted growth and spontaneous cell death in the same incompatible genetic background (Fig. S7 C and E). We concluded that the extreme dwarfism, enhanced pathogen resistance, and necrosis exhibited by incompatible lines at low temperature depend on SA accumulation.

We then examined the effect of a mutation in the enhanced disease susceptibility (EDS1) gene because this is an essential regulator of SA production and defense signaling triggered by TIR-NBS-LRR resistance proteins (11, 21). A null eds1 mutant in Ler (eds1–2) was crossed to Kas-2 and 192 F₂ plants genotyped with markers spanning four loci: QTL 3, QTL 4, QTL 5, and EDS1. Quadruple homozygous (eds1–2/eds1–2, QTL 3: Ler/Ler, QTL 4: Kas-2/Kas-2, and QTL 5: Kas-2/Kas-2) were isolated. No segregation of dwarfism was observed in progeny from F₂ single-locus segregating lines or eds1–2 homozygous-incompatible lines grown at 14 °C (Fig. S8). Thus, dwarfism driven by the incompatible allele interaction requires EDS1. The eds1 mutation also restored susceptibility to H. parasitica isolate Cala2 in incompatible Ler × Kas-2 lines grown at low temperature (Fig. S8). These data show that environmentally conditioned activation of immune and cell death responses in the incompatible hybrids requires SA signaling through the EDS1 pathway. They also support our conclusion that a Ler TIR-NB-LRR gene(s) is the determinant of interacting QTL 3.