Gametophytic and zygotic selection leads to segregation distortion through in vivo induction of a maternal haploid in maize.

Production of maternal haploids via a male inducer can greatly accelerate maize breeding and is an interesting biological phenomenon in double fertilization. However, the mechanism behind haploid induction remains elusive. Segregation distortion, which is increasingly recognized as a potentially powerful evolutionary force, has recently been observed during maternal haploid induction in maize. The results present here showed that both male gametophytic and zygotic selection contributed to severe segregation distortion of a locus, named segregation distortion 1 (sed1), during maternal haploid induction in maize. Interestingly, analysis of reciprocal crosses showed that sed1 is expressed in the male gametophyte. A novel mapping strategy based on segregation distortion has been used to fine-map this locus. Strong selection for the presence of the sed1 haplotype from inducers in kernels with haploid formation and defects could be detected in the segregating population. Dual-pollination experiments showed that viable pollen grains from inducers had poor pollen competitive ability against pollen from normal genotypes. Although defective kernels and haploids have different phenotypes, they are most probably caused by the sed1 locus, and possible mechanisms for production of maternal haploids and the associated segregation distortion are discussed. This research also provides new insights into the process of double fertilization.

Introduction

Double fertilization is a complex mechanism that has evolved in flowering plants (angiosperms). Successful double fertilization involves two sperm cells and two female gametes, the egg cell and the central cell, progenitors of the embryo and the endosperm, respectively. Disruption of any part of the process will lead to a block or abnormality in double fertilization.

Haploids can be one of the outcomes after failure of double fertilization. While in most sexually reproducing plant species, the life cycle alternates between a diploid sporophytic phase and a highly reduced gametophytic haploid phase (Walbot ), it is sometimes possible to obtain haploid plants containing the same number of chromosomes in their somatic cells as do normal gametes of the species (Dunwell, 2010). Two methods are generally used to produce haploids. One approach is to produce haploids from male or female gametic cells, but many species and genotypes are recalcitrant to this process (Forster ). Another approach for inducing haploids in some genera is ‘wide crossing’, in which one parental genome is selectively eliminated after fertilization (Kasha and Kao, 1970; Bennett ; Sanei ). Ravi and Chan (2010) found that haploids in Arabidopsis thaliana can be generated through uniparental chromosome elimination by manipulating a single centromere protein, the centromere-specific histone CENH3. In addition to these two approaches, two other methods exist in maize: haploids can be induced either by the ig1 mutant (Kermicle, 1969; Evans, 2007) as the female parent or by Stock6-derived inducers (Coe, 1959; Sarkar and Coe, 1966) as the male parent. The latter method is referred to as in vivo induction of maternal haploids and has become the method of choice in many maize breeding programmes worldwide (Prigge and Melchinger, 2012).

In a recent quantitative trait locus (QTL) mapping study on the in vivo haploid induction rate (HIR) in maize (Prigge ), significant segregation distortion (SD) against the inducer haplotype was observed in bins 1.03–1.05. The SD level reached a maximum value in bin 1.04, where one QTL for haploid induction (qhir1), explaining up to p = 66% of the genetic variance for HIR, was detected. In another study, too, the same region on chromosome 1 was associated with HIR as well as with SD (Barret ). Although considerable progress has been made in the genetic analysis of in vivo haploid induction, the mechanisms behind SD and the genetic characteristics of the qhir1 locus remain unknown.

Segregation distortion is defined as the deviation of genotype frequencies from Mendelian expectations. It can originate from double fertilization failure events that include competition among gametes or from abortion of the gamete or zygote (Faris ; Fishman ). Similar to other QTLs, SD loci are cryptic, but may carry an important evolutionary function because they control the viability of individuals facing natural selection (Xu, 2008). If a cluster of closely linked markers displays SD, one or several of them or tightly linked genes become candidate genes for viability selection and, thus, are no longer selectively neutral. Grini reported that by using five visible recessive markers on chromosome 1, seven gametophytic mutations that were induced by ethyl methanesulphonate (EMS) could be recognized by the SD of the nearby markers. Moreover, gametophytic mutants affecting various aspects of pollen development and function in Arabidopsis were identified through genetic screens for SD (Lalanne , ).

In this study, SD at the qhir1 locus was given top priority. First, the genomic region responsible for SD, subsequently referred to as the segregation distortion 1 (sed1) locus, was narrowed down to a 450kb region. It was found that both gametophytic and zygotic selection are involved in haploid induction and these two mechanisms together lead to strong SD of sed1. Furthermore, the role of gametophytic expression of sed1 in the fate of the zygote was discussed to shed light on the mechanism underlying maternal haploid induction in maize.

Materials and methods

Plant materials

The four inducer lines CAUHOI (Li ), CAU5, CAU079, and UH400 (Prigge ) with HIRs of ~2, 10, 9, and 8%, respectively, all carrying the R1-nj marker gene, and seven elite maize inbred lines with null HIR (1680, B73, Jing 24, Z58, C7-2, BY815, and GY923) were used. Hybrid ZD958, which is one of the most popular maize cultivars grown in China, and shows good purple pigmentation in the endosperm and embryo when crossed with inducers carrying the R1-nj marker gene, was used as a tester for determining the HIR. For pollen competitive ability tests, three female parents were used: hybrid ND5598 with long ears, sweet corn inbred CS1, and waxy corn inbred CW2. For production of crosses, standard practices in genetic field experiments with maize were followed to ensure that no contaminating pollen reached the silks: (i) all tassels of female parents were removed and the ears of seed parents were covered with paper bags prior to silk emergence; and (ii) silks were cut back the day before pollination, so that all exposed silks were of the same length, making it easier to receive the pollen.

Haploid identification

Two methods were employed for determination of the HIR. The first relied on self-pollinated ears and was called S-HIR. Here, all normal kernels obtained by selfing were planted in the field to infer their ploidy status. Haploid plants can be distinguished from diploid plants by their shorter stature, slender weak stems, erect and narrow leaves, and reduced growth rate. The S-HIR was calculated by the following formula: S-HIR=(number of haploid plants/total number of plants)×100%. The second method relied on test crosses with hybrid ZD958 to evaluate the HIR of inducers and was called T-HIR. Here, testcross seeds with purple endosperm (indicating a regular triploid endosperm resulting from successful fertilization with inducer pollen) and a colourless embryo (indicating a haploid embryo of ZD958 origin) were designated as putative haploid seeds following the guidelines published by Li . To verify the effectiveness of the R1-nj colour marker classification and to account for misclassification of seeds (Prigge ), all putative haploids were grown in Beijing and visually scored for plant characteristics as described above. The following formula was applied to calculate T-HIR: T-HIR=(number of putative haploids/total number of purple endosperm kernels)×K×100%, where K is a correction factor given by the ratio K=(number of true haploids/number of putative haploids).

Kernel classification

Kernels obtained from self-pollinated and cross-pollinated ears could be readily classified into three categories according to their morphological features: (i) normal kernels, with normal embryo and endosperm, including normal diploids and haploids (Fig. 1B 1, B2); (ii) endosperm abortion (EnA) kernels, with shrunken endosperm and kernel size smaller than that of normal haploids (Fig. 1B 4–B8); or (iii) embryo abortion (EmA) kernels, with normal endosperm but without an embryo (Fig. 1B3). Moreover, some embryo sacs did not show any development, which might be caused by early seed abortion or lack of fertilization, leading to a poor seed set rate (Fig. 1A2, A3).

Fig. 1.

Morphological differences in maize ears and seeds from various progeny. (A) Selfing ears of inbreds B73 and UH400 and ears from their reciprocal cross. (A1) Selfing ear of B73 showing good seed set; (A2) ear of the cross B73×UH400 showing poor seed set with many EnA kernels; (A3) selfing ear of UH400 with poor seed set and many EnA kernels; (A4) ear of the cross UH400×B73 showing good seed set. (B) Different kernel types after maternal haploid induction. (B1) diploids; (B2) haploids; (B3) EmA kernels; (B4–B8) different types of EnA kernels. From B4 to B8, the kernel size and volume progressively decreased, reflecting an impairment of kernel development at different developmental stages.

Evaluation of ear traits

In Experiment 1, 11 maize inbred lines listed in Table 1, comprising four inducers and seven normal maize lines, were used for morphological analysis of the progeny produced by self-pollination. In addition, the progeny of five populations from UH400 and 1680 were also analysed (Table 1). Based on kernel classification, besides S-HIR, the following traits were determined for all ears: normal kernel number (NKN), normal seed set rate (NSSR), endosperm abortion rate (EnAR), and embryo abortion rate (EmAR) using the formulae described in the footnote of Table 1. All traits were evaluated for at least three ears for each entry.

Table 1.

Kernel traits on selfing and cross-pollinated ears of different maize germplasm with and without haploid induction ability of the pollinator.

Parent		Ear and kernel traitsa				Haploid induction
Female	Maleb	NKN (per ear)	NSSR (%)	EnAR (%)	EmAR (%)	S-HIR (%)	T-HIR (%)
Male parents with haploid induction ability
CAU5	ⓧ	73	24.33	65.71	9.59	11.89	10.9
CAUHOI	ⓧ	125	60.19	8.76	1.54	1.31	1.94
CAU079	ⓧ	142	45.08	38.53	8.45	8.80	9.84
UH400	ⓧ	95	29.32	44.77	6.32	10.43	9.26
1680×UH400	ⓧ	373	71.45	24.40	1.87	1.93	2.53
1680	UH400	290	43.11	48.78	8.98	8.50	NDc
1680×UH400	UH400	210	39.91	51.86	10.47	8.86	NDc
Average		186.86	44.77	40.40	6.75	7.39	6.81
Male parents without haploid induction ability
1680	ⓧ	320	93.14	1.84	0	0	0
B73	ⓧ	357	91.16	1.11	0	0	0
Jing 24	ⓧ	287	91.77	0	0	0	0
Z58	ⓧ	320	91.65	0	0	0	0
C7-2	ⓧ	386	96.44	0	0	0	0
BY815	ⓧ	350	95.45	1.02	0	0	0
GY923	ⓧ	392	93.32	0	0	0	0
UH400	1680	287	91.40	0	0	0	NDc
1680×UH400	1680	467	93.40	0	0	0	NDc
Average		351.78	93.08	0.44	0	0	0

NKN, normal kernel number; NSSR, (nunber of normal kernels/number of embryo sacs)×100%; EnAR, (number of EnA kernels/number of embryo sacs)×100%; EmAR, (number of EmA kernels/number of normal kernels)×100%;

ⓧ, self pollination

ND, not determined.

Mapping strategy for sed1

In Experiment 2, three F2 populations derived from the cross 1680×UH400 were used to narrow down the sed1 region responsible for SD in bins 1.03–1.05. In all cases, SD of individual markers in the target region on bin 1.04 was evaluated by a χ2 test based on the three genotypes within each F2:3 family derived from recombinant F2 plants. First, nine recombinants in the QTL confidence interval of qhir1 described in a previous study (Prigge ) were identified in a sample of 186 plants. By analysing the genotype of the offspring from three recombinant F2 plants (R1, R2, and R3), the SD level for each F2:3 family could be assessed, which allowed delimitation of the sed1 locus to the region between markers umc1917 and bnlg1811.

In an attempt to narrow down the region further, a second and third sample of 2000 and 6000 F2 plants, respectively, were generated to obtain more recombinants. This led to the identification of 48 and 125 recombinants, respectively, by using the markers umc1917 and bnlg1811. Moreover, 12 new markers (primer sequences for each of these markers are given in Supplementary Table S1 available at JXB online) were used to genotype the recombinants in order to confirm their exchange boundary. These markers represented the subset of polymorphic markers between both parents from a set of 250 newly developed markers with primers designed on the basis of the B73 genome sequence, which spans a physical distance of 7.85Mb between markers umc1917 and bnlg1811. Each confirmed recombinant F2 plant was evaluated for T-HIR; moreover, each recombinant with a heterozygous genotype in this region was self-pollinated to obtain a large number of seeds for evaluating the SD level in each resulting F2:3 family. The genotypes of at least 50 plants in each F2:3 family were determined to ascertain that SD could be detected with sufficient power. PCR-based marker development and genotyping were conducted following the protocol of Yang .

F2:3 families originating from 10 independent F2 recombinants were tested for their SD level at the sed1 locus. Pollen of the three most important recombinant F2 plants (R1, R4, and R5) in F2 was used to pollinate the female parent 1680. The segregation ratios for each haplotype in the resulting backcross progeny were calculated by using either of the markers umc1917 or bnlg1811, depending on which one segregated in the particular cross.

SD detection in different cross combinations

In Experiment 3, the 15 populations listed in Table 2, each with at least 50 plants, were analysed for SD in the sed1 region. Again, SD from expected Mendelian segregation ratios was assessed by a χ2 test using either marker X18 or umc1917.

Table 2.

Analysis of segregation distortion in selfing or crossing progeny of different combinations of crosses between female and male parents with high (H), low (L), and no (N) endosperm abortion rated. Parents with induction ability are underlined.

Female	Malea	Genotypeb			χ2 test
		A	H	B	A:H:B=1:2:1	(A+B):H=1:1	A:H=1:1	B:H=1:1
Cross combination N×H; marker X18
1680×UH400	ⓧ	67	102	17	28.62***	1.55
GY923×UH400	ⓧ	84	111	32	23.93***	0.07
1680×UH400	1680	19	28	0			1.36
1680×UH400	UH400	11c	41	46				0.18
1680	1680×UH400	32	4	0			20.25***
Cross combination N×L; marker X18
1680×CAUHOI	ⓧ	31	24	9	19.13***	3.52
1680×CAUHOI	1680	44	49	0			0.17
1680×CAUHOI	CAUHOI	3c	33	28				0.09
1680	1680×CAUHOI	38	0	0			36.03***
Cross combination L×H; marker umc1917
CAUHOI×UH400	ⓧ	68	111	42	6.12*	0
CAUHOI×UH400	CAUHOI	71	53	1c			2.33
CAUHOI×UH400	UH400	4c	51	53				0.01
1680	CAUHOI×UH400	85	84	0			0
Cross combination H×H; marker umc1917
CAU5×CAU079	ⓧ	11	19	14	1.23	0.57
CAU5×CAU079	CAU079	7c	24	35				4.34

ⓧ, self pollination.

A refers to the homozygous genotype of the first parental line, B to the homozygous genotype of the second parental line, and H to the heterozygous genotype.

Haploid.

*, **, ***Significant at probability levels 0.05, 0.01, and 0.001, respectively.

Endosperm DNA extraction

In Experiment 4, endosperm DNA from different types of kernels on self-pollinated ears harvested from F1 plants of cross 1680×UH400 was extracted according to the following protocol: small pieces of endosperm sampled from single seeds without destruction of the embryo germination capacity were placed directly in a 96-well PCR plate. Subsequently, 100 µl of sodium hydroxide (0.1M) was added to each well and the plate was incubated at 99.9 °C for 10min. After 1min of centrifugation with a centrifugal force of 168 g, 100 µl of 1× TE buffer (pH=2.0) was added to each well and mixed, and the plate was centrifuged again with a centrifugal force of 671 g for 2min. The clear DNA-containing solution was then used for further PCR analysis.

Pollen competitive ability tests

In Experiment 5, three maize genotypes (CS1, CW2, and ND5598) were chosen as female parents. The four inducers (CAUHOI, CAU5, CAU079, and UH400) and the three inbred lines 1680, CS1, and CW2 were used as pollen sources to evaluate their pollen competitive ability (PCA).

PCA was assayed by applying two different kinds of pollen to the same ear at several time intervals. The exact process was as follows (Supplementary Fig. S1 at JXB online): first, for hybrid ND5598, pollen from the four inducers and non-inducer 1680 was applied separately to the different ears. Then 0, 0.5, 1, 1.5, 2, 2.5, 3.5, and 6h later, the ears pollinated first with pollen from the inducer lines were pollinated again with pollen from line 1680. Vice versa, the ears initially pollinated with pollen from line 1680 were pollinated with pollen from the four inducer lines. In order to ensure that an equal amount of pollen was allocated to each ear, newly collected pollen was divided into equal parts by volume by using 1.5ml tubes as a measuring cup. On the basis of the purple pigmentation of the endosperm, it was possible to distinguish between the two types of pollen applied by the dual pollination method. For the female parents CS1 or CW2, the same protocol was employed except that pollen from line 1680 was replaced by pollen from lines CS1 or CW2. At least five ears were used as replicates for each treatment.

In Experiment 6, pollen germination speed and pollen tube growth were compared between UH400 and 1680. First, fresh pollen of inbreds UH400 and 1680 was collected separately at 09:00h on the same day. Then, the pollen was divided into a number of equal parts by volume and pollinated separately on the silks of the tester hybrid ZD958. After intervals of 1, 2, 4, and 8h, silk tissue from ears pollinated with excess pollen of UH400 or 1680 were fixed in 9:1 v/v ethanol:acetic acid at 4 °C. The aniline blue staining protocol described by Lausser was adopted to observe pollen tube germination. Specimens were mounted with fresh staining solution on a slide with a coverslip and squashed by gentle squeezing; subsequently, they were examined on a fluorescence microscope (OLYMPUS BX51) with near-UV excitation, using wavelengths ranging from 315nm to 380nm. In order to guarantee high accuracy of the experiment, each treatment (three ears) and each ear (five silks) were analysed using aniline blue staining.

Genetic analysis of haplotype transmission through male and female gametes

To gain a better understanding of the gametophytic transmission of sed1, the process from the parental F1 to the F2 generation was taken as an example (Supplementary Fig. S2 at JXB online). Several possibilities affecting SD were taken into consideration. Let p and q represent the frequency of wild-type haplotypes (defined as ‘+’) in the male and female gametes involved in fertilization, respectively, then 1–p and 1–q represent the frequency of ‘sed1’ haplotypes in the male and female gametes, respectively. The survival rate of zygotes with genotype ‘+/+’, ‘+/sed1’, ‘sed1/+’, and ‘sed1/sed1’ is denoted by r1, r2, r3, and r4, respectively. Moreover, the common notation A, B, and H was used to describe the genotypes ‘+/+’, ‘sed1/sed1’, and ‘+/sed1’ or ‘sed1/+’, respectively. Taken together, the final frequency of each genotype in the F2 population can be described by the following formulae:

where T=q×p×r1+q×(1–p)×r2+(1–q)×p×r3+(1–q)×(1–p)×r4 is the survival rate of the zygotes summed over all three genotypes.

To obtain estimates for each parameter, the populations derived from the two parents 1680 and UH400 were used as a model. First, the mode of sed1 transmission in the backcross population (1680×UH400)×1680 from Experiment 3 was analysed for estimating q. Secondly, the ear and kernel traits of crosses from Experiment 1, where inducers served as female parents (sed1/sed1) and non-inducers (+/+) as male parents, were analysed to estimate r1 and r3. Thirdly, EnAR of kernels from the backcross (1680×UH400)×UH400 from Experiment 1 were analysed to obtain estimates for r2 and r4. Finally, a maximum likelihood estimator for p was calculated based on the above formula using a numerical algorithm implemented in software R 13.0 (http://www.r-project.org).

Results

The haploid induction process entails both haploid and defective kernels

In Experiment 1, estimates for S-HIR ranged from 1.31% to 11.89% and for T-HIR from 1.94% to 10.9% when the pollen parent was an inducer line, whereas S-HIR and T-HIR were null for all combinations in which the pollen parent was a non-inducer line (Table 1). The first group averaged only half the number of normal kernels (NKN) per ear of the second group, but had a high rate of defective kernels, as reflected by means for EnAR (40.4%) and EmAR (6.75%). Moreover, S-HIR for individual male parents in this group was tightly associated with T-HIR and defects in kernel traits of their progeny (Fig. 2). In comparison, materials without induction ability had a mean of EmAR and EnAR of exactly or close to zero. Ears of the cross 1680×UH400 and the backcross (1680×UH400)×UH400 had high values for both S-HIR (8.50% and 8.86%, respectively) and EnAR (48.78% and 51.86%, respectively), similar to those of the self-pollinated ear of UH400. In contrast, the reciprocal crosses UH400×1680 and (1680×UH400)×1680 showed null S-HIR and almost no defective kernels. In further crosses with various inducers as the female parent and normal lines as the male parent, no haploids and few defective kernels were found (Supplementary Fig. S3A at JXB online) and the ear phenotype resembled that of the F1 of cross UH400×1680.

Fig. 2.

Plots of S-HIR against different ear and kernel traits. (A–D) Plots of S-HIR against NSSR, EnAR, EmAR, and T-HIR in 11 maize inbred lines (four inducers and seven normal maize lines) and five progeny derived from crosses between line 1680, inducer UH400, or their single cross (see Table 1). Open circles with a cross, pollen contains sed1 male gametes; filled traingles, pollen that did not contain sed1 male gametes.

SD and its features in different genetic populations is controlled by sed1

By screening for SD in F2:3 progeny of 10 (heterozygous) recombinant F2 plants in Experiment 2, the region of sed1 between the two markers X22 and X93 could be narrowed down (Fig. 3) with a physical distance of 455kb, based on the B73 reference sequence. Moreover, estimates of T-HIR for each of these 10 recombinants were closely associated with the level of SD (Fig. 4). Recombinants with significant SD had a mean T-HIR of 1.64±0.19%, whereas those without SD had a significantly (P < 0.001) smaller mean T-HIR of 0.26±0.09%.

Fig. 3.

Fine-mapping of sed1 and segregation distortion (SD) in F3 progeny or backcross progeny to 1680 of F2 plants recombinant for the sed1 region. Left side: the physical position of markers mapping to bin 1.04 is shown on top of the bars. The bars reflect the genotype of the F2 plants (open bars, homozygous for the 1680 haplotype; shaded bars, heterozygous for the UH400 haplotype). Right side: number of F3 progeny or backcross progeny in each genotype class and χ2 test for deviation from expected Mendelian segregation ratios. The S-HIR of the parental recombinant F2 plants is also given. *, **, *** Significant at probability levels 0.05, 0.01, and 0.001, respectively.

Fig. 4.

Plot of T-HIR of 10 recombinants in the F2 generation against χ2 test statistic for SD. Stars, recombinants with significant SD; filled circles, recombinants without SD.

Based on the EnAR determined in Experiment 1 (Table 1), the four inducers and two common maize lines were classified into three groups: 1680 and GY923 were non-abortion (N) lines; CAUHOI was a low (L) abortion line (8.76%); and UH400, CAU5, and CAU079 were high (H) abortion lines with an EnAR of 44.77, 65.71, and 38.53%, respectively. In Experiment 3, different crosses among these groups (N×L, N×H, L×H, H×H) were used to investigate the universality and characteristics of SD in sed1. In general, F2 populations derived from crosses of type N×L and N×H showed significant (P < 0.01) SD. In the backcross populations, if the sed1 haplotype was present in the female gametes of crosses of type (sed1/+)×(+/+), the segregation ratios at markers X18 or umc1917 were in agreement with normal Mendelian expectations. However, if the male gametes carried the sed1 haplotype, for example in the cross (+/+)×(sed1/+), high significant (P < 0.01) SD was observed. In F2 populations derived from crosses of type L×H, significant (P < 0.05) SD against the sed1 haplotype derived from the H parent was found, but, when the F1 hybrid CAUHOI×UH400 of type L×H was used as male parents to pollinate the female parent 1680, no SD was found in the progeny (n=169). Moreover, no significant SD was detected in this region for the backcross, when L×H hybrid CAUHOI×UH400 was used as female parent. In the H×H crosses, no significant SD was detected for both the F2 populations and CAU079 backcross populations.

Genotypic analysis of the different kernel types in segregating population

High HIR was always associated with a high rate of defective kernels and SD (Figs 2, 4). For investigating the association between the latter two traits, 7544 kernels from 35 ears of the F2 population 1680×UH400, segregating for the sed1 locus, were classified into different kernel classes (Table 3). Randomly sampled kernels in each class were assayed for their endosperm genotype. In addition, the ploidy level for all putative haploid kernels was determined by planting and genotyping each plant for 20 markers polymorphic between UH400 and 1680. In the 74 putative haploids, 86.5% of the kernels were finally determined to be haploids. After correcting for misclassification in determination of putative haploids, strong selection was found for the presence of the sed1 haplotype from the inducer in the endosperm of haploids or defective kernels (Table 3). Notably, all haploids (n=64) and EmA kernels (n=66) harboured at the UH400 haplotype marker locus X18 in their endosperm.

Table 3.

Genotypes at marker X18 of different kernel types in the F population from 1680×UH400.

Type	Selected kernel no.	Genotyped kernel no.	Genotypea (endosperm)			Genotypea (seedling)
			A	H	B	A	H	B
Normal kernel	5892	5499	2031	2847	621	NDc	NDc	NDc
Haploidb	64	64	0	33	31	33	0	31
EnA kernel	1438	183	2	96	85	NDc	NDc	NDc
EmA kernel	150	66	0	38	28	NDc	NDc	NDc

Genotype of different classes: A=Sed1 homozygous, H= heterozygous, B= sed1 homozygous.

Confirmed as haploids by planting and genotyping of the seedlings.

ND, not determined.

Pollen competitive ability (PCA) of sed1 pollen

A potential reason for SD could be differences in PCA (Xu ; Lu ). The dual-pollination method used in Experiment 5 allows the PCA of haploid inducers and common (non-inducer) lines to be distinguished (Table 4). When female parents were first crossed with a non-inducer line, followed by application of pollen from an inducer at different time intervals (0–2h), up to 98.86% of the kernels showed colourless endosperm, indicating successful pollination by the non-inducer lines after an interval of 0.5h. If inducer pollen was first applied, followed by pollination with a non-inducer at different time intervals (0–6h), up to 88.2% of the resulting kernels had colourless endosperm, as was also the case after simultaneous pollination with a 1:1 mixture of pollen from both sources.

Table 4.

Pollen competitive ability test for normal and haploid inducing lines after pollination with pollen from Male 1 at 09:00h and sequent pollination with pollen from Male 2 at different time intervals from 0 (=equal parts by volume) to 6 h.

Parent			Time interval (h)	Traitsa
Female	Male 1	Male 2		Kernel no.		PEKR (%)		P-value
				Up	Down	Up	Down
ND5598	1680	UH400	0.5	248	278	1.44	1.31	NDb
ND5598	1680	UH400	0	225	255	16.89	7.05	0.0004**
ND5598	UH400	1680	0.5	193	189	17.1	5.82	0.0002**
ND5598	UH400	1680	1	201	152	36.81	28.95	0.024*
ND5598	UH400	1680	1.5	215	160	88.37	37.5	0**
ND5598	UH400	1680	2.5	189	194	97.88	45.36	0**
ND5598	UH400	1680	3.5	191	118	95.81	94.07	0.2446
ND5598	UH400	1680	6	122	104	95.08	97.12	0.7828

Kernel no. number of normal kernels in different parts of the ear; PEKR, purple endosperm kernel rate=(number of purple endosperm kernels/total number of normal kernels)×100%

ND, not determined.

*, **Significant at probability levels 0.05 and 0.01, respectively.

From ears first pollinated with pollen of UH400 and afterwards pollinated with pollen from 1680 at different time intervals (0–2.5h), most of the kernels with purple endosperm (indicating fertilization with inducer pollen) were found in the upper part of the ear (Fig. 5D) and the number of kernels with purple endosperm significantly (P < 0.05) differed between the upper and lower parts of the ear (Table 4). The same results were observed if pollen from CS1 or CW1 was used instead of pollen from 1680 (data not shown).

Fig. 5.

Pollen competitive ability (PCA) in haploid inducers and non-inducer maize inbreds. (A1) and (A2) Germination of 1680 pollen on ZD958 silks; (B1) and (B2) germination of UH400 pollen on ZD958 silks (scale bars are 100 µm). (A1 and B1) Pollen tubes were stained with aniline blue 1h after pollination. (A2 and B2) Pollen tubes were stained with aniline blue 8h after pollination; the pollen tube is shown in blue colour. (C) PCA tested by dual pollination, from left to right: (C1) ear of ND5598 first pollinated with 1680 pollen, then pollinated with UH400 pollen after 1h; (C2) ear of ND5598 pollinated with a mixture of equal amounts of UH400 and 1680 pollen at the same time; (C3) ear of ND5598 first pollinated with UH400, then pollinated with 1680 after 1h; (C4) ear of ND5598 first pollinated with UH400, then pollinated with 1680 after 2h; (D) ears of ND5598 first pollinated with UH400, then pollinated with 1680 after 1h, showing clusters of purple endosperm kernels in the distal part of the ear, and colourless endosperm kernels concentrated mainly at the base of the ear.

Results from in vivo pollen germination in Experiment 6 showed that most pollen grains from line 1680 germinated within 1h (Fig. 5A1), whereas most of the UH400 pollen grains did not germinate within this time frame (Fig. 5B1). However, the latter began to germinate after 1h, and most of them eventually germinated like the pollen of 1680 (Fig. 5B2).

Genetic characterization of sed1 pollen mutants

The crosses (1680×UH400)×1680 and (1680×CAUHOI)×1680 belong to the (sed1/+)×(+/+) type of cross. For both, the frequencies of ‘sed1/+’ and ‘+/+’ genotypes were according to the χ2 test in agreement with the expected Mendelian segregation ratio of 1:1 (Table 2). Consequently, q=1–q=0.5 and the sed1 haplotype from the female parent did not contribute to SD.

For the reciprocal crosses 1680×(1680×UH400) and 1680×(1680×CAUHOI), which belong to the (+/+)×(sed1/+) type of cross, severe SD from the expected 1:1 genotypic frequencies of sed1/+ and +/+ was found (Table 2). Thus, the sed1 haplotype from the male parent was the reason for the SD. Considering that defective kernels and haploids could be detected only rarely when inducers served as female parents (sed1/sed1) and non-inducers (+/+) served as male parents, it follows that r1=r3=1. The χ2 tests for a ratio of 1:1 between homozygous (A+B) versus heterozygous (H) genotypes at the sed1 locus in the (sed1/+)×(sed1/+) type of F2 population indicated no significant deviation from the expected Mendelian segregation ratios (Table 2), leading to the conclusion that r2=r4=r. In the cross (1680×UH400)×UH400, EnAR was 51.8%, from which it is concluded that r was ~50%. Hence, the maximum likelihood estimator for p obtained from Equations 1–3 was 0.6 when inserting q=0.5, r1=r3=1, and r2=r4=r=0.5 (Fig. 6). Based on this value, for normal kernels (n=5499), the genotype frequencies in the F2 population of cross 1680×UH400 were A:H:B=0.375:0.500:0.125, which were in close agreement with the observed frequencies of 36.9, 51.8, and 11.3%, respectively (Table 3). In the self-pollinated ears of F1 plants from cross 1680×UH400, the EnAR amounted to ~24.40% (Table 1), which was again in excellent accordance with the predicted value of 25% based on the value of the parameters given above (Fig. 6).

Fig. 6.

Genetic transmission of sed1 in the F2 population 1680-UH400. For the +/sed1 F1 plants, there is no distortion in the female gametes, and transmission of both haplotypes occurs at a frequency of 50%. Because of the poor PCA of sed1 gametes, male gametes face competition and selection in the gametophytic stage, and the transmission of sed1 gametes is reduced to 40%. The male and female gametes then fuse together, and one sperm of the pollen fuses with the egg cell to produce the embryo while the second fuses with the central cell to form the endosperm. Because sed1 is a male gametophyte-expressed gene and will influence the fate of the zygote, the kernel survival rate will be 50% for both ‘+/–‘ and ‘–/–‘ zygotes.

Discussion

Gametophytic and zygotic selection

Several studies reported that gametophytic selection (Mangelsdorf and Jones, 1926; Lavery and James, 1987; Lin ) or zygotic selection (Rick, 1963; Gadish and Zamir, 1986) might be the mechanism underlying SD. However, no report indicated that both selections together influence haplotype transmission. In the present study, the endosperm genotype of kernels with haploid formation or defects in the self-pollinated ears of F1 plants from cross 1680×UH400 showed strong selection for the presence of the sed1 haplotype from the inducer. In all haploid and EmA kernels, the endosperm consistently showed the UH400 haplotype (Table 3). In EnA kernels, too, the endosperm genotype of all (n=183) kernels except two indicated the presence of the sed1 haplotype. These two exceptions are most probably due to nutritional competition among sporophytes (Sheridan ), because a low frequency of EnA kernels (0.44%) was also found in the common (non-inducer) maize lines. In Experiment 5, PCA of inducers was lower than that of common materials, suggesting that the sed1 haplotype is selected for not only in male gametophytes but also in the sporophytes after fertilization. One explanation for the pleiotropic effect of the sed1 locus on both gametophytic and sporophytic development might be that it plays a central role in cell development.

The estimates for the survival rates r1–r4 and parameter p were based on F2 and backcross populations of cross 1680×UH400. Obviously, the strength of gametophytic and zygotic selection may depend on the parental lines, as indicated by the variation in EnAR and EmAR values for progeny produced by self- and cross-pollination (Table 1).

Embryo–endosperm interactions

Defective kernels caused by defects in endosperm development, embryogenesis, or both were found in dek mutants (Neuffer and Sheridan, 1980; Scanlon ) and interploidy-intraspecific crosses (Johnston ; Birchler, 1993). It was observed that defective kernels occur concomitantly with haploids in the haploid induction process. Compared with haploid kernels (Fig. 1B2), defective kernels differed in their appearance and showed a continuous distribution in size (Fig. 1B3–B8). In particular, the embryo aborted in EmA kernels (Fig. 1B3), but their endosperm developed quite well and resulted in ‘germless’ kernels. This corroborates that the embryo phenotype is not dependent on the endosperm genotype (Scanlon, 1997). Since most EnA kernels lacked an embryo, it is speculated that defective kernels and haploids are caused by the same genetic mechanism. Thus, defective kernels might also be haploids that suffered from an early disruption of embryogenesis, leading to a disruption of endosperm development at different stages. An additional mechanism leading to zygote suspension during seed development might also be involved, but this warrants further research.

Mechanism for induction of maternal haploids and defective kernels

There are two main hypotheses to explain the maternal haploid induction in maize. One is that the haploid embryo can be formed from an unfertilized egg cell (Sarkar and Coe, 1966; Barret ) and the other is chromosome elimination. A third hypothesis, named semigamy (Turcotte and Feaster, 1967), suggests failure of karyogamy in the fertilized egg cell, which results in chimeras of haploid maternal and paternal tissue that develop into the proper embryo tissue and the suspensor cells, respectively.

Many male gametophytic mutations have been studied well and several of them such as cdka-1, fbl17, Chromatin Assembly Factor-1 (CAF-1), and meiosis pathway mutants (fas1, fas2, and msi1) were described, which cause the formation of bicellular pollen harbouring a single germ cell due to failure of germ cell division. The single sperm of mutants cdka-1, fbl17, and msi1 can still fertilize the egg cell or central cells, leading to single fertilization events (Iwakawa ; Nowack ; Chen ; Kim ; Aw ). In contradiction to the hypotheses of single sperm-like fertilization and semigamy, Sarkar and Coe (1966) reported that all pollen grains produced by inducer lines are tricellular and, in previous studies (Fischer, 2004; Li ), genome fragments from the inducer were found in a few haploid plants. Furthermore, except for some extreme examples such as B chromosomes, it is known that each of the two sperm cells can fertilize either the egg cell or the central cell (Chen ; Frank and Johnson, 2009; Aw ). Consequently, if the single sperm that carried the sed1 haplotype fused with either the egg cell or the central cell and eventually resulted in a haploid kernel or a defective kernel, respectively, then the endosperm of aborted kernels should not always carry the sed1 haplotype. However, in contradiction to the single fertilization hypothesis, all but two defective kernels (including haploids, EnA kernels, and EmA kernels) in the F2 population of cross 1680×UH400 carried the sed1 haplotype. Therefore, the present results corroborate that double fertilization also happens in maternal haploid induction, and chromosome elimination in the embryo is the underlying cause of maternal haploid induction in maize. Because the endosperm of all defective kernels contains the paternal haplotype of sed1 in the mature stage, chromosome elimination seems not to occur in the endosperm.

Several hypotheses have been proposed to explain uniparental chromosome elimination in interspecific crosses (Gernand ). For example, when CENH3 null mutants in Arabidopsis expressing altered CENH3 proteins are crossed with the wild type, chromosomes from the mutant are eliminated, and haploid progeny are produced (Ravi and Chan, 2010). In maize, maternal haploid induction is most probably not caused by a mutation in the CENH3 gene, because the maize CENH3 gene is located on chromosome 6 (http://www.maizesequence.org) and no CENH3 homologue exists in maize bin 1.04.

The sed1 haplotype in male gametophytes affects seed development

There is evidence for an extensive overlapping of sporophytic and gametophytic gene expression in higher plants. For example, Tanksley examined isozyme profiles of nine enzyme systems in tomato (Solanum lycopersicum) and found that 60% of the structural genes coding for them in the sporophyte were also expressed in the gametophyte. Studies on synthetics in maize also showed that PCA due to pollen tube growth rate and kernel development are largely controlled by genes expressed in both tissues (Ottaviano ). In contrast, comparative analyses of the Arabidopsis pollen transcriptome revealed that nearly 40% of the genes were exclusively expressed in the pollen (Honys ). Based on the results from both ear trait evaluation and SD detection of the plants from the cross combinations (+/+)×(sed1/+ or sed1/sed1), both defective kernels and significant SD were always detected; however, in the reciprocal crosses, almost no defective kernels and SD were found. This clearly demonstrates that the gene(s) underlying the sed1 locus are expressed only in the pollen.

When inducer lines were used for pollinating the female parents, some ovaries develop normally, resulting in normal diploid kernels, some ovaries stop developing at different stages, leading to defective kernels, while a small proportion of kernels develop as haploids (Fig. 1B). There might be two explanations: (i) the sed1 locus was heterozygous in the inducer and the lethality of homozygous embryos maintains heterozygosity over generations; or (ii) the gene(s) underlying in vivo haploid induction display gametophytic expression with incomplete penetrance. The progeny-test method was employed to examine the HIR of the selfing progeny from inducer line UH400 (Supplementary Table S2 at JXB online). Since no bimodal distribution for T-HIR was observed in the offspring, it is concluded that the sed1 locus was homozygous for the UH400 haplotype, which rules out the first explanation.

Epigenetic regulation can be one cause of incomplete penetrance (Petronis, 2001). Epigenetic modification of the DNA of centromeric chromatin influences the precise deposition of the CENH3 protein in centromeres (Zhang ). Since binding of the CENH3 protein to centromeric chromatin is crucial for normal functioning of the centromere in proper segregation of the sister chromatids (Birchler ), it is hypothesized that the sed1 locus exerts an epigenetic, dosage-dependent modification of the chromosomes (Fig. 7). This model assumes that the expression of sed1 differs between the pollen of plants homozygous for the sed1 haplotype. Therefore, the epigenetic modification of the chromosomes caused by sed1 can differ between pollen grains. A strong modification of the sperm cell chromosomes will lead to kernel abortion or haploid formation. If the pollen is less epigenetically modified, this will not affect the double fertilization process and this finally leads to normal kernel formation. Therefore, a small proportion of the kernels become haploid but the majority develop into normal diploid kernels. Further research is needed to test this hypothesis. In particular, cloning of the gene(s) underlying the sed1 and qhir1 locus may help to uncover the genetic and functional basis of in vivo haploid induction in maize, especially whether only a single gene or a cluster of genes is involved, and this should also reveal its relationship to SD.

Fig. 7.

Model for sed1 as a male gametophyte-expressed gene(s) that influence the fate of the zygote. (A) sed1 is expressed only in the male gametophyte, and not in the female gametophyte or in the sporophyte. The gene product can epigenetically modify chromosomes of the sperm cell, because the dosage of sed1 can differ between pollen grains; in some grains, the chromosomes are highly modified (in black), whereas, in others, the chromosomes are less modified (in grey). The sed1 gene product could reduce pollen competitive ability but not affect pollen viability. (B) The high degree of modification of the paternal chromosomes in the sperm cell leads to kernel abortion or chromosome elimination. The remaining, less modified pollen will not affect the double fertilization process, thus leading to normal kernel formation. (This figure is available in colour at JXB online.)

Supplementary data

Supplementary data are available at JXB online

Figure S1. Schematic diagram for the pollen competitive ability (PCA) test procedure for lines 1680 and UH400.

Figure S2. Punnett square for genetic transmission of the sed1 haplotype through male and female gametes in F2 populations with different survival rates ri for the different genotypes.

Figure S3. Ear phenotype in reciprocal crossing and selfing ears.

Table S1. Oligonucleotide primers for newly developed molecular markers for fine-mapping sed1.

Table S2. Progeny testing of T-HIR for UH400 plants by two normal inbred lines.