Pistillody, homeotic transformation of stamens into pistil-like structures, caused by nuclear–cytoplasm interaction in wheat

Authors


* For correspondence (fax +81 776 61 6015; e-mail murai@fpu.ac.jp).

Summary

Homeotic transformation of stamens into pistil-like structures (pistillody) has been observed in a cytoplasmic substitution (alloplasmic) line of wheat (Triticum aestivum L.) cv. Norin 26, which has the cytoplasm of a wild relative species, Aegilops crassa L. On the other hand, an alloplasmic line of wheat cv. Chinese Spring (CS) with Ae. crassa cytoplasm has normal flowers. This is due to the presence in the CS nucleus of a fertility-restoring gene, Rfd1. Deletion mapping analysis revealed that Rfd1 is located on the middle part of the long arm of chromosome 7B. To investigate the function of the Rfd1 gene by a loss-of-function strategy, we produced alloplasmic lines of CS ditelosomic 7BS [(cr)-CSdt7BS] and CS monotelodisomic 7BS [(cr)-CSmd7BS] with the Ae. crassa cytoplasm, and characterized their phenotypes. The line (cr)-CSdt7BS without Rfd1 exhibited pistillody in all florets, and also female sterility. Scanning electron microscopy of the young spikes revealed that the pistillody was induced at an early stage of stamen development. The pistillate stamens often developed incomplete ovule-like structures with integuments instead of tapetum and pollen grains. It is possible that MADS box genes are associated with the induction of pistillody, because the expression of wheat APETALA3 homologue (WAP3) was reduced in the young spikes of (cr)-CSdt7BS. In addition, a histological study indicated that the female sterility in (cr)-CSdt7BS is due to the abnormality of the ovule, which fails to form an inner epidermis and integuments in the chalaza region. The line (cr)-CSmd7BS, hemizygous for Rfd1, showed partial pistillody (51%) and restored female fertility up to 72%. These results suggest that the induction of both pistillody and ovule deficiency caused by the Ae. crassa cytoplasm is inhibited by the Rfd1 gene in a dose-dependent manner.

Introduction

Extensive genetic and molecular studies in two dicot species, Arabidopsis and Antirrhinum, have provided a general understanding of the determination of floral organ identity in higher plants, explained by the ABC model (Davies and Schwarz-Sommer, 1994; Ma, 1994). In this model, three classes of homeotic genes (A, B and C), each expressed in two adjacent whorls of the floral meristem, define floral organ identity. Class A genes function in the outer (first and second) two whorls; class B genes in the second and third whorls; and class C genes in the inner two (third and fourth) whorls. Sepals are formed by class A gene function in whorl 1. Petals in whorl 2 and stamens in whorl 3 require the functions of class A and B or B and C genes, respectively. Carpels are formed by the class C gene alone in whorl 4. In Arabidopsis, these homeotic genes have been cloned, and most encode members of the MADS box family of transcription factors (Riechmann and Meyerowitz, 1997), except for the class A gene APETALA2. Although a limited number of homeotic genes have been isolated in the monocot grasses, recent studies on class A, B and C MADS box genes suggest that the ABC model of flower development for dicots can essentially be extended to rice (Chung et al., 1995; Kang et al., 1995; Kang et al., 1998; Kyozuka et al., 2000; Moon et al., 1999) and maize (Ambrose et al., 2000; Mena et al., 1995, Mena et al., 1996; Schmidt et al., 1993).

The genetic interactions between the bread wheat (Triticum aestivum L.) nucleus and alien cytoplasms of Triticum and Aegilops species have been investigated using cytoplasmic substitution (alloplasmic) lines produced by recurrent backcrossing (Tsunewaki, 1993; Tsunewaki, 1996; Tsunewaki et al., 1996). In higher plants the most commonly observed effect of alien cytoplasm is abnormal pollen development, namely cytoplasmic male sterility (CMS). In a previous study we reported that the alloplasmic line of wheat cv. Norin 26 (N26) with the cytoplasm of wild relative species Aegilops crassa L. shows homeotic transformation of stamens into pistil-like structures (pistillody) when grown under long-day conditions (>15 h light), and named this phenomenon photoperiod-sensitive CMS (PCMS) (Murai and Tsunewaki, 1993). On the other hand, wheat cv. Chinese Spring (CS) does not show PCMS when the Ae. crassa cytoplasm is introduced. Genetic analysis indicated that fertility restoration in alloplasmic CS is controlled by a single dominant gene (designated Rfd1) located on the long arm of chromosome 7B (Murai and Tsunewaki, 1994). These results indicated that pistillody is caused by the interaction between the N26 nucleus and Ae. crassa cytoplasm under long-day conditions, and that the Rfd1 gene in CS prevents the induction of pistillody. Mitochondrial genes have been found to be associated with CMS traits in most plant species so far examined (Breiman and Galun, 1990; Hanson, 1991; Mackenzie et al., 1994). In these CMS systems, various chimeric mitochondrial genes produced by homologous recombination are believed to disturb mitochondrial function at a critical stage of stamen development, causing male sterility. Consequently, it has been suggested that mitochondrial gene(s) of Ae. crassa cytoplasm are involved in PCMS in alloplasmic wheat (Ogihara et al., 1997; Ogihara et al., 1999).

Pistillody has also been observed in an alloplasmic line of T. durum L. with A. caudata L. cytoplasm (Kihara and Tsunewaki, 1961). In this case, however, a suppresser gene(s) against the induction of pistillody has not been identified. The induction and suppression of pistillody we report here is a unique system, in which it is demonstrated that the stamen malformation is controlled by the interaction between nuclear and cytoplasmic (mitochondrial) genes. Our ultimate goal is to understand the genetic and molecular mechanisms underlying the induction of pistillody caused by the nuclear–cytoplasm interaction in the alloplasmic wheat. The deficiency of class B MADS box genes causes homeotic transformation of stamens into carpels in Arabidopsis (Goto and Meyerowitz, 1994; Jack et al., 1992) and Antirrhinum (Sommer et al., 1990; Tro¨bner et al., 1992). Thus it is possible that pistillody in the alloplasmic wheats was generated by modifying the expression of MADS box genes belonging to class B. How could mitochondrial gene(s) affect MADS box gene expression? The characterization of this phenomenon is a fascinating subject concerned with homeotic changes of floral organs caused by nuclear–cytoplasm interaction.

Here we initially describe the induction and suppression of pistillody caused by the interaction between the nuclear gene Rfd1 and Ae. crassa cytoplasm. To delineate the role of the Rfd1 gene by a loss-of-function strategy, we produced by chromosome engineering an alloplasmic line of CS wheat lacking the long arm of chromosome 7B, and examined the phenotype by morphological and histological analyses in comparison with the euplasmic (with wheat cytoplasm) line. Telocentric chromosomes (telosomes) cannot be used in diploid organisms because of their lethal effects. However, in polyploid species such as hexaploid wheat, the deleterious effects of the telosomes can be avoided by the duplication of genetic materials in them. A series of ditelosomic lines of CS wheat, which is deficient for a pair of long arms or short arms of each homologous chromosome, has been developed (Sears and Sears, 1978). Using the ditelosomic 7BS line of CS with a pair of telosome 7BS, we produced alloplasmic lines of ditelosomic 7BS [(cr)-CSdt7BS] and monotelodisomic 7BS [(cr)-CSmd7BS] with Ae. crassa cytoplasm. The line (cr)-CSmd7BS has a combination of single chromosome 7B with a telosome 7BS as the counterpart. Therefore the line (cr)-CSmd7BS has one dose of the Rfd1 gene, whereas the line (cr)-CSdt7BS carries no Rfd1. The results indicate that the Ae. crassa cytoplasm induces abnormal development of ovule in pistils as well as pistillody in stamens, and the Rfd1 gene prevents the induction of both pistillody and ovule deficiency in a dose-dependent manner.

Results

Wheat spikelet and floret structures

The inflorescence of a wheat plant (spike, ear or head), developed at the tip of the stem, is composed of spikelets. The spikelets are arranged to form two opposite rows along the main axis or rachis (Figure 1a). The number of spikelets per spike is determined by the timing of terminal spikelet initiation, which depends on the genotype and the environmental conditions. In this respect, the rachis meristem in wheat is determinate. The spikelet is composed of florets joined at the axis (rachilla) alternately on opposite sides, and encompassed by two small bract leaves called glumes (Figure 1b). There are multiple florets (usually six to eight) in each spikelet, of which a few in the apical positions may be sterile due to hypoplasia. In contrast to barley, rice and maize, the rachilla meristem in wheat is classified as indeterminate. In each floret, the reproductive organs are enveloped by two leaf-like structures, a lemma and a palea (Figure 1c). The lemma bears an awn at its tip in cv. Norin 26 (Figure 1b), but there are also awnless cultivars such as Chinese Spring. It was recently suggested that the lemma and palea are homologous to dicot sepals in rice and maize (Ambrose et al., 2000; Kang et al., 1998). An individual wheat flower contains one pistil, three stamens and two lodicules (Figure 1d). The pistil, which has a unilocular carpel, is the female part of the flower and consists of the ovary containing the ovule and two filamentous styles, each terminating in a feathery stigma. The stamen, the male part of the flower, is composed of a filament and an anther containing pollen grains. The lodicule is considered to be a modified petal in maize and rice (Ambrose et al., 2000; Kang et al., 1998; Kyozuka et al., 2000), which swells during anthesis, forcing the lemma and palea apart to facilitate pollination of the stigma from the dehisced anther.

Figure 1.

Wheat inflorescence structure.

(a) The wheat inflorescence (spike, ear or head) is composed of spikelets (Sp) attached at the nodes of zig-zag rachis (Rs). The rachis meristem is determinate because of the formation of a terminal spikelet (TS). Scale bar, 2 cm.

(b) A spikelet is removed from the rachis. It consists of multiple (usually six to eight) florets attached at the rachilla (Ra). Although the rachilla meristem is indeterminate, a few florets in apical positions are sterile due to hypoplasia. The first and second florets (1F and 2F) are indicated. Two small bract leaves (glumes; Gl) enclose the spikelet. Scale bar, 1 cm.

(c) Close-up of an opened floret. In each floret the reproductive organs [pistil (Pi) and stamens (St)] are enveloped by two leaf-like structures, a lemma (Le) and a palea (Pa). The lemma and palea are detached from each other to make the reproductive organs visible. Scale bar, 2 mm.

(d) An individual flower contains one pistil (Pi), three stamens (St) and two lodicules (Lo). Palea (Pa) is also indicated. In this figure, the pistil, stamens, lodicules and palea are removed from the rachilla. Scale bar, 2 mm.

Aegilops crassa cytoplasm induces pistillody in alloplasmic wheat

To investigate the interaction between Ae. crassa cytoplasm and Rfd1 located on the long arm of chromosome 7B, we produced a ditelosomic 7BS line of wheat cv. Chinese Spring (CS) with the Ae. crassa cytoplasm [(cr)-CSdt7BS]. This cytoplasmic substitution (alloplasmic) line does not contain the Rfd1 gene as it lacks the long arm of chromosome 7B. Under natural daylength conditions (<15 h) in the field, the line (cr)-CSdt7BS showed pistillody in all florets (Figure 2; Table 1). In contrast, a euplasmic ditelosomic line with the wheat cytoplasm (CSdt7BS) developed normal stamens. On the other hand, almost all stamens were normal in alloplasmic CS [(cr)-CS] under natural daylength conditions (<15 h) in the field. The induction of pistillody caused by the Ae. crassa cytoplasm was originally observed in an alloplasmic line of wheat cv. Norin 26 grown only under long-day conditions (>15 h) (Murai and Tsunewaki, 1993). The present results indicate that Ae. crassa cytoplasm essentially induces pistillody independent of long photoperiod, and the Rfd1 gene located on chromosome 7B prevents the induction of pistillody. Figure 2 shows the floral organs in the euplasmic (CSdt7BS) and alloplasmic [(cr)-CSdt7BS] lines. Despite the pistillody of the stamens, the lodicules remained unchanged in (cr)-CSdt7BS, indicating that homeotic alteration was restricted to the staminate portion. The monotelodisomic 7BS line of CS with the Ae. crassa cytoplasm [(cr)-CSmd7BS], which is hemizygous for Rfd1, showed a reduced frequency of pistillody (51%; Table 1), indicating that the Rfd1 gene functions in a dosage-dependent manner. In the floret of (cr)-CSmd7BS, the frequency of normal stamen appearance was significantly higher in the central position than in the lateral positions (Table 2), indicating that the stamens at the lateral positions tend to be more influenced by the Ae. crassa cytoplasm. Normal stamens found in the florets of (cr)-CSmd7BS produce normal pollen in normal quantities (Table 1). This indicates that the influence of the Ae. crassa cytoplasm is restricted to the homeotic transformation of stamens, and there is no effect on pollen development when the stamens develop normally.

Figure 2.

Floral organs (pistil, stamens and lodicules) of the euplasmic and alloplasmic lines with Aegilops crassa cytoplasm grown under the natural daylength conditions (<15 h) in the field.

(a) Pistil (Pi), stamens (St) and lodicules (Lo) in the euplasmic ditelosomic 7BS line of Chinese Spring (CSdt7BS). The pistil, stamens and lodicules are removed from the rachilla. Scale bar, 2 mm.

(b) Pistil (Pi), transformed stamens (TSt) and lodicules (Lo) in alloplasmic ditelosomic 7BS line of Chinese Spring [(cr)-CSdt7BS]. The pistil, stamens and lodicules are removed from the rachilla. Note that the lodicules remain unchanged. Scale bar, 2 mm.

Table 1.  Fertility of reproductive organs and selfed seed fertility in the euplasmic and alloplasmic lines grown under the natural daylength conditions (<15 h) in the field
LineMale fertilityFemale fertility
Pistillody
frequency
(% ± SE)
Pollen
fertility
(%)
Crossed seed
fertility
(% ± SE)
Selfed seed
fertility
(% ± SE)
  1. Values with the same letter do not differ significantly at the 5% level from each other when Duncan's new multiple range test is applied.

CS0.0 ± 0.0c95.394.2 ± 2.4a94.2 ± 1.4a
(cr)-CS8.8 ± 2.2c97.790.4 ± 4.3a,b96.8 ± 0.7a
CSmd7BS0.0 ± 0.0c95.885.1 ± 5.3a,b,c93.0 ± 2.8a
(cr)-CSmd7BS50.8 ± 7.3b93.172.0 ± 6.0c42.8 ± 3.1b
CSdt7BS0.0 ± 0.0c96.179.7 ± 5.4b,c90.9 ± 3.1a
(cr)-CSdt7BS100.0 ± 0.0a0.0 ± 0.0d0.0 ± 0.0c
Table 2.  Frequency of normal stamen formed in the floret of (cr)-CSmd7BS
Position of stamenFrequency (%)
Right20.5 ± 4.9
Centre93.1 ± 3.7
Left11.0 ± 5.6

Surprisingly, the line (cr)-CSdt7BS showed complete female sterility in addition to complete pistillody, while the euplasmic counterpart and alloplasmic CS were female fertile (Table 1). This means that the Ae. crassa cytoplasm causes female sterility together with pistillody, and that the Rfd1 gene also restored female fertility. The line (cr)-CSmd7BS tended to have reduced female fertility (72%) compared with (cr)-CS (90%), indicating a dosage effect of the Rfd1 gene on female fertility restoration. The selfed-seed fertility reflected pistillody frequency and female fertility (Table 1). The line (cr)-CSdt7BS showed no seed set due to complete pistillody and female sterility, whereas the line (cr)-CSmd7BS showed 43% restored seed fertility because of incomplete pistillody and restored female fertility.

Chromosomal mapping of Rfd1, a suppressor gene against the induction of pistillody by Ae. crassa cytoplasm

Deletion mapping was performed to locate the Rfd1 gene in the long arm of chromosome 7B. Because the line (cr)-CSdt7BS is completely female sterile (Table 1), the line (cr)-CSmd7BS was crossed with a series of chromosome 7BL deletion lines. Distribution of pistillody frequency in the F1 progenies of the cross (cr)-CSmd7BS × CS 7BL deletion lines was observed with nine deletion lines (Figure 3a). In the F1 progenies of the deletion lines with a fractional length (FL) of distance from the centromere of 0.24–0.40, complete pistillody plants appeared in about half the progenies. On the other hand, normal plants were obtained in the progenies of the cross with FL 0.63–0.86 deletion lines. These results clearly indicated that the Rfd1 gene is located within the region FL 0.40–0.63 (Figure 3b).

Figure 3.

Deletion mapping of Rfd1, a suppresser gene against the induction of pistillody caused by the Aegilops crassa cytoplasm.

(a) Distribution of pistillody frequency in F1 progenies of the cross (cr)-CSmd7BS × a series of CS 7BL deletion lines. The break-point position for each deleted chromosome was calculated as the fractional length (FL) of the distance from the centromere.

(b) Physical map of the long arm of chromosome 7B with C-banding. Fractional length (FL) measurements are shown on the left, and location of Rfd1 on the right.

Pistillody occurs at an early stage of stamen development

To determine at what point in development the stamens transformed into pistil-like structures in the (cr)-CSdt7BS florets, we analysed the early developmental stage of young spikes in CSdt7BS (normal) and (cr)-CSdt7BS (pistillody) by SEM (Figure 4a,b). The development of florets begins in the basal positions of each spikelet and progresses toward the distal positions. Therefore we can observe floret initiation by growth stage order within a spikelet. In the normal spikelet of CSdt7BS, the stamen primordia were clearly distinguishable from the pistil primordia in the first to third florets, due to their tetralocular forms (Figure 4c). In contrast to CSdt7BS, the stamen primordia of (cr)-CSdt7BS never attained a tetralocular form in any floret position, but instead resembled the pistil primordia (Figure 4d). This indicates that the homeotic transformation of stamens into pistil-like structures occurs during an early stage of stamen development.

Figure 4.

SEM of young spikes at the floret differentiation stage in the lines CSdt7BS (normal) and (cr)-CSdt7BS (pistillody).

(a) A young spike with developing spikelet (Sp) in CSdt7BS. In the spikelet, eight primordia of florets are detectable. The first and second florets (1F and 2F) are indicated. Scale bar, 500 µm.

(b) A young spike with developing spikelet (Sp) in (cr)-CSdt7BS. Eight primordia of florets are detectable in the spikelet, and the first and second florets (1F and 2F) are indicated. Same magnification as (a).

(c) Close-up of the spikelet with developing florets in CSdt7BS. Pistil (Pi) primordia initiate at the centre of the floret, which is surrounded by stamen (St) primordia. Within a spikelet the development of florets begins in the basal positions and progresses toward the distal positions. Note that the developing stamen primordia begin assuming a tetralocular shape, which are distinguishable from the pistil primordia at the first to third floret. Lemma (Le) and glume (Gl) primordia are also indicated. Scale bar, 200 µm.

(d) Close-up of the spikelet with developing florets in (cr)-CSdt7BS. Note that the stamen (St) and pistil (Pi) primordia are undistinguishable from each other in shape. Lemma (Le) and glume (Gl) primordia are also indicated. Same magnification as (c).

Pistillate stamens have an incomplete ovule-like structure instead of tapetum and pollen grains

The pistillate stamens of (cr)-CSdt7BS were transversely sectioned through the middle, and examined for histological modifications. The pistillate stamens contained ovule-like structures instead of tapetum and pollen grains, indicating variable patterns in formation of the ovule-like structures. Some pistillate stamens showed abnormal development of the inner epidermis (Figure 5a), and in other stamens there was an ovule-like structure either with an inner integument (Figure 5b), or with both inner and outer integuments (Figure 5c). However, the vascular bundle system connecting the ovule and ovary was not developed. In this case, no seed set was observed in the pistillate stamens due to their abnormality (data not shown). Although the homeotic changes of stamens occur at an early stage of stamen development (Figure 4), the pistillate stamens have incomplete ovule-like structures instead of complete ovules.

Figure 5.

Transverse sections of pistillate stamens in the line (cr)-CSdt7BS. All scale bars, 0.5 mm.

(a) A pistillate stamen developing an abnormal inner epidermis (IE).

(b) A pistillate stamen developing an ovule-like structure (Ol) with an abnormal inner epidermis (IE) and inner integument (Ii). The vascular bundle system, which connects the ovule and ovary, was not developed.

(c) A pistillate stamen developing an ovule-like structure (Ol) with an abnormal inner epidermis (IE), and an inner and outer integument (Ii and Io). The vascular bundle system was not developed, as in (b).

Ae. crassa cytoplasm also affects ovule formation in genuine pistils

Transverse and longitudinal sections of the pistils were also analysed by microscopy to investigate the cause of female sterility in the (cr)-CSdt7BS pistils. In normal pistils, the ovule is surrounded by an inner and outer integument, then enveloped by an inner epidermis (Figure 6a). A longitudinal section of normal pistils in CSdt7BS revealed the vascular bundle system connected with the ovule at the chalaza region (Figure 6b). Unlike normal pistils, sterile pistils in (cr)-CSdt7BS failed to form an inner epidermis and integuments in the chalaza region, and the vascular bundle system was malformed (Figure 6c–e). Furthermore, the pistils in (cr)-CSdt7BS were generally smaller than in CSdt7BS. These observations suggest that female sterility of (cr)-CSdt7BS pistils is due to the disruption of ovule development.

Figure 6.

Transverse and longitudinal sections of pistils in lines CSdt7BS and (cr)-CSdt7BS. All scale bars, 0.5 mm.

(a) Transverse section of the middle part of normal pistil in CSdt7BS. The ovule (Ov) is surrounded by an inner and outer integuments (Ii and Io), then enveloped by an inner epidermis (IE). The chalaza (Ch) region, a connection of the ovule and vascular bundle system, is also indicated.

(b) Longitudinal section of the middle part of normal pistil in CSdt7BS. The ovule (Ov) is connected with the vascular bundle (VB) system at the chalaza (Ch) region. Inner epidermis (IE), outer and inner integuments (Io and Ii), and micropyle (Mi) are also indicted.

(c) Transverse section of the middle part of a sterile pistil in (cr)-CSdt7BS. Contrary to the normal pistils in CSdt7BS, the sterile pistil failed to form an inner epidermis and integuments in the chalaza (Ch) region.

(d) A transverse section of the middle part of a sterile pistilin (cr)-CSdt7BS, which has an abnormal cell aggregate (CA) at the chalaza (Ch) region.

(e) A longitudinal section of the middle part of a sterile pistil, same type as (c), in (cr)-CSdt7BS. The sterile pistil failed to form an inner epidermis and integuments in the chalaza (Ch) region, leading to malformation of the vascular bundle system.

Expression of wheat APETALA3 homologue WAP3 is reduced in young spikes of (cr)-CSdt7BS

The homologues of Arabidopsis floral organ identity genes APETALA1 and APETALA3, WAP1 (wheat APETALA1; formerly TaMADS#11) and WAP3 (wheat APETALA3; formerly TaMADS#51), were isolated by screening a cDNA library from young wheat spikes at the floret differentiation stage (Murai et al., 1998). Phylogenetic analysis using the deduced amino acid sequences revealed that WAP1 belongs to the class A MADS box gene family, and is closely related to maize APETALA1 homologue ZAP1 (Mena et al., l995), and that WAP3 clusters with maize and rice B function genes, Silky1 (Ambrose et al., 2000) and OsMADS16 (Moon et al., 1999). The accumulation of WAP1 and WAP3 transcripts was examined by quantitative RT–PCR analysis with cDNAs from young spikes at the floret differentiation stage, when stamen primordia are differentiated and developed (Figure 7). The same level of the WAP1 gene expression was observed in both lines CSdt7BS and (cr)-CSdt7BS. However, line (cr)-CSdt7BS showed a large reduction in WAP3 gene expression in comparison with CSdt7BS. This suggests that reduced WAP3 expression could be associated with the induction of pistillody in (cr)-CSdt7BS.

Figure 7.

Comparison of WAP1 and WAP3 transcript levels in lines CSdt7BS (normal) and (cr)-CSdt7BS (pistillody) by quantitative RT-PCR.

PCR amplifications were performed with the first-strand cDNAs synthesized from DNase-digested total RNA of young spikes at the floret differentiation stage. PCR patterns are shown with 26 and 30 cycles for the Ubi-1 and WAP1 genes, and with 30 and 34 cycles for the WAP3 gene, to indicate that the amplification at fewer cycles for each gene is in the exponential range of amplification. The Ubi-1 gene is used as control, showing that the same amount of cDNA was used in the PCR reactions of the two lines.

Discussion

Pistillody triggered by a long photoperiod was initially reported in an alloplasmic line of wheat cv. Norin 26 (N26) with Ae. crassa cytoplasm (Murai and Tsunewaki, 1993). This alloplasmic line showed pistillody under long-day conditions of 15 h or longer, whereas normal stamen formation occurred under short-day conditions of 14.5 h or less. This phenomenon was called photoperiod-sensitive cytoplasmic male sterility (PCMS). Murai and Tsunewaki (1994) identified a single dominant fertility-restoring (Rf) gene against the PCMS located on the long arm of chromosome 7B in wheat cv. Chinese Spring (CS), which they named Rfd1. The presence of the Rfd1 allele results in normal formation of stamens in the alloplasmic CS line, suppressing the influence of Ae. crassa cytoplasm. In the present study, we observed that line (cr)-CSdt7BS, which lacks the Rfd1 gene, exhibited pistillody in all florets under short-day conditions (<15 h) in the field (Figure 2; Table 1). This demonstrates that the Ae. crassa cytoplasm essentially induces pistillody in wheat independent of long photoperiod, and the Rfd1 gene prevents the induction of pistillody. A suppressor gene of Rfd1, which functions under long-day conditions, could be responsible for the induction of pistillody (PCMS) under long-day conditions in the alloplasmic N26. Alternatively, N26 might have gene(s) suppressing the influence of Ae. crassa cytoplasm, which act under short-day conditions. Furthermore, there is a possibility that the PCMS in the alloplasmic N26 is connected with the gene(s) for photoperiodic response in this cultivar.

Cytoplasmic male sterility (CMS) is known to be caused by functional incompatibility between the nuclear and mitochondrial genomes. It has been reported that mitochondrial mutations, such as multiple intragenic recombination and insertion of sequences of unknown origin, are associated with CMS (Breiman and Galun, 1990; Hanson, 1991; Mackenzie et al., 1994). In the PCMS system, the mitochondrial gene orf25 is the most likely candidate because, among 12 mitochondrial genes investigated, only orf25 showed alterations in both gene structure and transcription patterns between the euplasmic and alloplasmic lines of Ae. crassa cytoplasm (Ogihara et al., 1997). The restoration of fertility by nuclear Rf gene(s) has been shown to be involved with RNA maturation and/or editing of the transcripts from CMS-associated mitochondrial genes in several plant species such as rice (Iwabuchi et al., 1993), Brassica (Singh et al., 1996) and sorghum (Tang et al., 1996). In the present PCMS system in wheat, differential processing of the mitochondrial orf25 gene was observed among the euplasmic and alloplasmic lines (Ogihara et al., 1999). However, no restoration in the transcription pattern of orf25 was detected in alloplasmic CS having Rfd1, based on Northern blot analysis with mRNA extracted from etiolated seedling. This suggests that the Rfd1 gene could function on a specific developmental stage and/or tissue, as in the cases of sunflower (Monéger et al., 1994) and common bean (Abad et al., 1995). Alternatively, the Rfd1 gene might act post-translationally, as in radish (Krishnasamy and Makaroff, 1994) and rapeseed (Bellaoui et al., 1999). Based on these analogies, it is likely that the suppressive effect of the Rfd1 gene against the induction of pistillody by Ae. crassa cytoplasm affects the function of PCMS-associated mitochondrial gene(s) (possibly orf25) rather than that of genes for floral organ formation directly. Cloning and characterization of the Rfd1 gene would help us to understand this point.

Similarly to rice and maize, an individual wheat flower consists of pistil, stamens and lodicules, and is enclosed with two leaf-like structures known as the lemma and palea (Schmidt and Ambrose, 1998) (Figure 1). The most distinctive feature of grass flowers is the lack of obvious sepals and petals. In rice, recent studies on class A, B and C homeotic genes revealed that the lemma and palea are equivalent to sepals, and lodicules to petals (Kang et al., 1998; Kyozuka et al., 2000). Similarly, in maize the lemma and palea have been recognized as modified sepals and the lodicules as modified petals (Ambrose et al., 2000). These findings strongly indicate that the ABC model of floral organ development in dicots can be extended to rice and maize, and probably to wheat. According to the ABC model, the homeotic change of stamens is caused by loss of function of class B genes, and is associated with alteration of the petals. In Arabidopsis and Antirrhinum, loss-of-function mutants in class B genes display homeotic alterations in both stamens and petals (Bowman et al., 1991; Sommer et al., 1990). Furthermore, transgenic rice with loss-of-function of class B genes exhibited homeotic alterations of the lodicules to palea- or lemma-like structures, and the stamens to carpel-like organs (Kang et al., 1998). In this study we observed that in the young spike of (cr)-CSdt7BS exhibiting pistillody, the expression level of a wheat class B MADS box gene, WAP3, was greatly reduced in comparison with CSdt7BS with normal stamens (Figure 7). This suggests that the induction of pistillody caused by the Ae. crassa cytoplasm in alloplasmic wheat is associated with change of MADS box gene expression. In the pistillate floret of (cr)-CSdt7BS, however, the lodicules developed normally (Figure 2), indicating that floral organ identity in whorls 2 and 3 could be determined separately by homeotic genes in wheat. In contrast to Arabidopsis and Antirrhinum, a class B gene, GREEN PETAL (GP; identical to pMADS1) is required to specify only the petal identity of whorl 2, and is not involved in the determination of stamen identity in Petunia (van der Krol et al., 1993). A double-mutant analysis with GP and class A homeotic mutants suggested that a postulated class B gene, ‘PhBX’, works in whorl 3 but not in whorl 2 for the formation of stamens (Tsuchimoto et al., 2000). Based on these findings, we suggest that a whorl 3-specific class B gene is involved in the phenotype of the alloplasmic line (cr)-CSdt7BS. The WAP3 gene must be a candidate for such a gene. More precise molecular studies on wheat homeotic genes, including the MADS box genes (Murai et al., 1997; Murai et al., 1998), could elucidate the mechanism of induction of pistillody.

The induction and suppression of pistillody presented here is unique in that the incompatibility between nuclear and mitochondrial genes induces a homeotic transformation of floral organs. How does a mitochondrial gene induce homeotic transformation of stamens into pistil-like structures in alloplasmic wheat? Mitochondrial DNA analyses of somatic hybrids revealed that alterations in mitochondrial DNA are associated with homeotic transformation of stamens in Nicotiana (Kofer et al., 1991) and in Brassica (Wang et al., 1995). These observations strongly suggest that homeotic gene function could be influenced by mitochondrial gene(s), resulting in homeotic alterations of floral organs. The expression patterns of homeotic genes should be examined in alloplasmic [(cr)-CSdt7BS] and euplasmic (CSdt7BS) wheat lines to clarify the relationship between the mitochondrial orf25 gene and nuclear homeotic genes.

In this study, we observed ovule-like structures in the pistillate stamens of (cr)-CSdt7BS (Figure 5), as reported in carpeloid stamens of Arabidopsis (Bowman et al., 1999) and Antirrhinum (Sommer et al., 1990). It is noteworthy that the pistillate stamens in (c)-CSdt7BS often develop ovule-like structures with an inner epidermis and integuments. Together with the induction of pistillody, the pistils of (cr)-CSdt7BS failed to form an inner epidermis and integuments in the chalaza region of the ovule, resulting in female sterility. These findings suggest that the expression of genes for ovule development, such as AINTEGUMENTA (Elliott et al., 1996; Klucher et al., 1996) in Arabidopsis, could also be influenced by the nuclear–cytoplasm interaction.

In conclusion, the present study demonstrates that pistillody, the homeotic transformation of stamens into pistil-like structures, is controlled by the nuclear–cytoplasm interaction in alloplasmic wheat, i.e. mitochondrial gene(s) (possibly orf25) in Ae. crassa cytoplasm versus the nuclear Rfd1 gene. Moreover, these results suggest that the nuclear–cytoplasmic interaction could affect ovule development. Molecular dissection of this phenomenon provides a fascinating link between the nuclear–cytoplasm interaction and homeotic conversion of floral organ identity.

Experimental procedures

Plant materials

Cytoplasm of Aegilops crassa L. was transferred to Triticum aestivum L. cv. Chinese Spring (CS) by repeated backcrosses (Tsunewaki, 1993; Tsunewaki, 1996; Tsunewaki et al., 1996). The alloplasmic line obtained, Ae. crassa/11*CS [(cr)-CS], was crossed as the female parent with the ditelosomic 7BS line of CS (CSdt7BS) as the male parent, which is deficient for a pair of long arms of chromosome 7B (Sears and Sears, 1978). The F1 hybrid (monotelodisomic 7BS; (cr)-CSmd7BS) was then backcrossed to the CSdt7BS line as the pollen parent to produce the alloplasmic ditelosomic 7BS line with Ae. crassa cytoplasm [(cr)-CSdt7BS]. Because the telosome 7BS is normally transmitted through the female gametes, the progenies segregated the monotelodisomics [(cr)-CSmd7BS] and ditelosomics [(cr)-CSdt7BS] in the ratio 1 : 1 (data not shown). The (cr)-CSdt7BS plants obtained were verified by somatic chromosomal staining with acetocarmine during root-tip mitosis. For deletion mapping of Rfd1, (cr)-CSmd7BS was crossed with a series of chromosome 7BL deletion lines (Endo, 1988; Endo and Gill, 1993).

Estimation of male and female fertilities

Male and female fertilities in the three alloplasmic lines, (cr)-CS; (cr)-CSmd7BS; (cr)-CSdt7BS, and their normal (wheat) cytoplasm counterparts, were examined using the materials grown under natural daylength conditions (<15 h) in the field. To investigate the fertility of the male reproductive organs, two characters, pistillody frequency and pollen fertility, were measured. Pistillody frequency (%) was determined as the percentage of pistillate stamens in the first and second florets of all spikelets using five ears per line. Pollen fertility (%) was estimated as the percentage of normal pollen observed in the anther of the first florets. Female fertility was estimated from crossed seed fertility (%) of five ears per line as follows. Ears of the maternal parents were emasculated and bagged to prevent cross-pollination. After 2–3 days the stigmas of emasculated florets were pollinated with fresh pollen of CS. The crosses were made using the first and second florets of the middle spikelets in each spike. About 30 days after pollination, the number of florets with and without seeds was recorded for each spike. Crossed seed fertility was the percentage of successful crosses over the total number of florets pollinated. Selfed seed fertility (%) was estimated by the seed setting rate of the first and second florets of all spikelets in two to three bagged ears per line.

Deletion mapping

To determine the chromosomal position of the Rfd1 gene, the line (cr)-CSmd7BS was crossed as a female parent with nine deletion stocks having various deficiencies in the long arm of chromosome 7B (Endo, 1988; Endo and Gill, 1993). The break-point position for each deleted chromosome was calculated as a fractional length (FL) of the distance from the centromere. Pistillody frequency (%) was evaluated for 17–20 F1 plants by the same method as described above.

Scanning electron microscopy

Young spikes at the floret differentiation stage were vacuum infiltrated with 2% glutaraldehyde and 2% paraformaldehyde in cacodylate buffer, and prefixed overnight at 4°C. The samples were rinsed three times in water, and fixed in OsO4 in the same buffer. After fixation, the samples were dehydrated in a graded ethanol series, then critical point-dried in liquid carbon dioxide and mounted on stubs. The stubs with prepared material were coated with platinum by an ion spatter (Hitachi E1010), and examined with an SEM (Hitachi S-4700) at an accelerating voltage of 10 kV.

Light microscopy

Stamens and pistils of pistillate florets in (cr)-CSdt7BS and normal pistils in CSdt7BS were embedded in 5% agarose without fixation. Sections (50 µm thick) were made by Micro-Slicer (Dohan EM, Kyoto, Japan), and light microscopy was performed with a Nikon Optiphoto-2.

Quantitative RT–PCR

Total RNA was extracted from young spikes at the floret differentiation stage in lines CSdt7BS and (cr)-CSdt7BS, using ISOGEN (Nippon-gene, Tokyo, Japan). DNase-digested total RNA (4.5 µg) was reverse-transcribed with oligo-dT primer, and the first strand cDNA was obtained using the First-strand synthesis kit for RT–PCR (Amersham, Little Chalfont, Buckinghamshire, England). PCR primers for RT–PCR were designed from the sequences of WAP1 (formerly TaMADS#11; DDBJ Accession Number AB007504) and WAP3 (TaMADS#51; AB007506) cDNA clones (Murai et al., 1998) as follows: WAP1-L (ATCAGACTCAGCCTCAAACA), WAP1-R (TAGAGACGGGTATCATGGAA), WAP3-L (TACAAGAAGAAGGTGAAGCA) and WAP1-R (ACAACCAAAGTCATAGCACA). As a control, a fragment from the wheat ubiquitin gene (Ubi-1) was amplified using the primers Ubi-1 L (GCATGCAGATATTTGTGAA) and Ubi-1R (GGAGCTTACTGGCCAC). PCR reactions were performed with 26 and 30 cycles for Ubi-1 and WAP1 genes and with 30 and 34 cycles for the WAP3 gene. The amplification at fewer cycles for each gene was in the exponential range of amplification. PCR fragments were separated on 1.5% agarose gel, stained with ethidium bromide, and photographed.

Acknowledgements

We are grateful to Dr T. R. Endo for providing us with the seed stocks of 7BL deletion lines. Thanks are also due to Dr M. Fukutomi for technical advice in SEM, and Dr A. P. Aryan for PCR primer information about the Ubi-1 gene and for critical reading of the manuscript. This work was supported in part by a Grant-in-Aid for Scientific Research (B) from the Ministry of Education, Science and Culture of Japan (No. 12460006, to K.M.) and from Fukui Prefectural Government (to K.M.).

Ancillary