Copyright 0 1997 by the Genetics Society of America

Genetics of Hybrid Sterility and Hybrid Breakdown in an Intersubspecific Rice ( O y a sativa L.) Population

Zhikang Li,**+ Shannon R. M. Pinson,: Andrew H. Patemon,” William D. Parks and James W. Stanselt *Plant Genome Mapping Laboratory, Department of Soil and Crop Sciences, ”Department of Biochemistry and Biophysics, Texas A&”

University, College Station, Texas 77843, tTexas A&M University Agricultural System Research and Extension Center, Beaumont, Texas 77713 and xUSDA-ARS, Beaumont, Texas 77713

Manuscript received September 26, 1996 Accepted for publication December 19, 1996

ABSTRACT F, hybrid sterility and “hybrid breakdown” of F2 and later generations in rice (Oryza sativa L.) are

common and genetically complicated. We used a restriction fragment length polymorphism linkage map and F4 progeny testing to investigate hybrid sterility and hybrid breakdown in a cross between “widely compatible” 0. sativa ssp. japonica cultivar Lemont from the Southern U.S. and ssp. indica cultivar Teqing from China. Our results implicate different genetic mechanisms in hybrid sterility and hybrid breakdown, respectively. Hybrid sterility appeared to be due to recombination within a number of putative differentiated “supergenes” in the rice genome, which may reflect cryptic structural rearrange- ments. The cytoplasmic genome had a large effect on fertility of both male and female gametes in the F, hybrids. There appeared to be a pair of complementary genes that behaved like “wide compatibility” genes. This pair of genes and the “gamete eliminator” ( S I ) or “egg killer” (S-5) may influence the phenotypic effects of presumed supergenes in hybrids. Hybrid breakdown appeared to be largely due to incompatibilities between indica and japonica alleles at many unlinked epistatic loci in the genome. These proposed mechanisms may partly account for the complicated nature of postreproductive barriers in rice.

S PECIATION, the process by which an ancestral lin- eage splits into two or more reproductively isolated lineages, is the core of evolution. A fundamental com- ponent of this process is the origin of reproductive isola- tion. In the modern synthesis, DOBZHANSKY (1936) and MULLER (1940) proposed that general incompatibilities between differentiated “complementary” (epistatic) genes often accounted for postzygotic isolation, since these genes affect viability or fertility of hybrid progeny but not the parental species. However, STEBBINS (1958) proposed that formulation of reproductive barriers might differ considerably from one group of organisms to another.

In plants, rates of genetic divergence between popu- lations are affected by breeding systems ( JAIN 1975). Asian cultivated rice (Oryza sativa L.) is a predominantly self-pollinating species. The differentiation of 0. sativa into two subspecies, ssp, indica and ssp. japonica, has been well documented (OKA 1988). Considerable dif- ferences between the two subspecies both in morphol- ogy and at the molecular level [isozymes and restriction fragment length polymorphism (RFLP)] are associated with their adaptability to different environments (GLASZMANN 1987; OKA 1988; WANG and TANKSLEY 1992; LI and RUTGER 1997). Varying degrees of hybrid sterility (sterility in F1 hybrids) and hybrid breakdown

Corresponding author: Zhikang Li, Department of Soil and Crop Sci- ence, Texas A&M University, College Station, TX 77843-2474. E-mail: zli@tam2000.tamu.edu

Genetics 145 1139-1148 (April, 1997)

(sterility and weakness in F2 or later generations) are commonly seen in crosses between indica and japonica varieties ( STEBBINS 1958; OKA 1988).

Hybrid sterility in rice has been a subject of extensive investigation but its basis remains controversial. Al- though hybrid sterility due to chromosomal aberrations has been suggested (HENDERSON et al. 1958; YAO et al. 1958; SHASTRY and MISRA 1961; DELORES et al. 1975), it could not, in the majority of cases, be attributed to cytologically detectable abnormalities (CHU et al. 1969). In addition, several lines of evidence implicate involve- ment of major genes in which alleles at a single locus or two complementary loci could either cause sterility or recover fertility in hybrids (KITAMURA 1962; OKA 1974 and 1988; IJSEHASHI and ARAKI 1986). Certain rice varieties produce fertile F1 hybrids when crossed to ei- ther ssp. indica or ssp. japonica and thus are called “widely compatible varieties” (MORINAGA and KURI- YAMA 1958; JENNINGS 1966; IKEHASHI and ARAKI 1986). Hybrid sterility and hybrid breakdown often coincide in indica/japonica crosses, but hybrid breakdown is sometimes found in advanced-generation progenies from completely fertile F1 hybrids. OKA (1974) found a pair of complementary recessive sterility genes that caused hybrid breakdown in indica/japonica crosses. While OKA (1988) speculated that there may be many sets of such complementary sterility genes responsible for hybrid breakdown in rice, this hypothesis remains untested.

1140 %. Li rt al.

Here, we investigate the genetic basis of postrepro- ductive barriers in rice. Using DNA markers, we sought to determine the number and locations of quantitative trait loci (QTL) and their epistatic interactions, which were associated with hybrid sterility and hybrid break- down in rice.

MATERIALS AND METHODS

Plant materials: Two 0. sativa cultivars, Lemon1 and Teq- ing, were the parentq in the study. Lemont is a semidwarf commercial variety from Southern U.S., belonging to the ja- ponica group based on data of isozymes and RFLP (XIE 1998; LI and RUTGER 1997). It is also widely compatible since it produces fertile Fl’s when crossed with typical indica and , l a - ponicn cultivars (XIE: 1993; Z. L I , unpublished data). Teqing is a semidwarf indica cultivar from China that generally produce highly sterile F,’s when crossed with typical japonica varieties. Reciprocal F, crosses were made between the parents. Two hundred sixty-seven F2 plants from the Lemont/Teqing cross were randomly selected to produce 255 F:I lines (12 F, plants produced X ) from each o f the remaining 255 F-, lines were used to reconstruct the Fy marker genotype and to produce 2418 F., lines (seven to 11 F, lines from each of the F:, lines, which in turn derived from a single F2 plant) that were used for replicated phenotyping.

RFLF’ marker genotyping: The genotypes of each of the 2.55 F? plants for 119 RFLP marker loci were determined as described previously ( L I ut nl. 1995). The genotypes of the 2.55 F2 plants for two morphological markers ( C for purple apiculus and gl-1 for leaf pubescence) were also determined from progeny testing of the F4 lines. The resulting map with 115 markers and an average 19.1 cM between markers covered a11 12 rice chromosomes.

Phenotypic data: The field experiment to phenotype the 2418 F., lines derived from the 255 F2 plants was conducted in 1990, as described previously (IJ ut al. 1995). The F, lines were drill planted into the field at the Texas A&M University System Agricultural Research and Extension Center in Beau- mont, Texas in 1990. Each F., line was planted in a single row plot 2.4 m long with a spacing of 28 cm between rows. The F, lines were planted in family groups separated by tworow check plots of Gulfmont, a sister line of the female parent Lernont. Gulfmont was used to determine within plot varia- tion and the F, data were adjusted accordingly. Eleven plots of Teqing were also randomly located in the field. Ten pani- cles from different plants in each of the F4 lines were collected and dried at 50” for 72 hr. Panicles from each of the F, lines were hand-threshed and assayed for Several grain yield compo- nent traits including total floret number per panicle (FN), sterile floret (unfilled florets) number per panicle, grain number (filled spikelets) per panicle, 200 kernel weight, and grain weight per panicle. Spikelet sterility (SS) was calculated as the percentage of empty spikelets over total spikelets per panicle for each of the Fa lines. Plant height (pH) and head- ing date (HD) were assayed on each of the F, lines as de- scribed previously ( L I rt al. 1995). PH was used as a trait associated with hybrid weakness since weak and/or slowly growing plants tend to have reduced height (OK\ 1988). The breeding value of each o f the 255 F2 plants for these traits was obtained by averaging its seven to 11 F.+ lines (70- 110 observations) and used in the data analyses. In 1993, the par- ents and reciprocal F, plants were planted in the greenhouse for assay of SS and pollen fertility using I&I staining.

Data analyses: Mapping of QTL used interval mapping (L,ANI)ER and BOTSEIN 1989) and Mapmaker/QTL (L,IN(:oI.N rt rcl. 1992). A threshold of LOD 2.4 was used to claim the

presence of a QTL. Genomic regions with LOD scores be- tween 2.0 and 2.4 were considered as putative QTL. Identified QTL were further evaluated using a nmltiple regression model (ZEN(: 1998, 1994; 1.1 rt (11. 1997). To identify epistasis affecting S S and PH, twz.o-way analyses of variances (ANOVA) were performed between 95 selected markers (noncodomi- nant markers and those with >9.5 missing data were not used) based on the model and the methods by L I rt nl. (1997) using SAS PRO%%) a n d low SS (9.6% for Lemont and 5.8% for Teqing).

In the field experiment, the 2418 F4 lines from the Lemont/Teqing cross had mean SS of 28.3%, signifi- cantly ( P < 0.01) higher than the parents (15.7 2 4.2% for Lemont and 19.1 2 3.5% for Teqing). The F.l distri- bution of SS was slightly skewed, ranging from 8 t o nearly 70% (Figure 1). The breeding values of the 255 F2 plants for sterility had an approximately normal dis- tribution, ranging from 15 to 50%. Sterility was strongly correlated with heading date ( r ( ; = 0.447, P < O . O O O l ) , weakly correlated with plant height ( q , = 0.258, P = O.001), and not correlated with floret number per pani- cle or kernel weight ( , r ( ; = 0.148 and 0.007, respec- tively).

Identification of SS QTL Four QTL affecting SS were mapped to chromosomes 3, 6, 8, and 10 (Table 2 and Figure 2). These QTL had I? ranging from 7.2 to 21.1% a n d collectively explained 60.4% of the geno- typic variance. The presence of two putative QTL ( QL%2 a n d QSs5) on chromosomes 2 a n d 5 WdS suggested by subthreshold LOD scores (LOD = 2.14 and 2.35) . When a full model with the four QTL (LOD > 2.4) fixed was performed on the two putative QTL by Map- maker/QTL and by multiple regression analysis, @.s5 was confirmed (with a LOD of 3.25) and y S s 2 remained suggestive (LOD = 2.1). It was particularly interesting

Reproductive Isolation Barriers in Rice 1141

TABLE 1

Comparison of reciprocal FI hybrids between ‘Lemont’ and ‘Teqing’ for spikelet sterility (SS), pollen sterility (PS), heading date (HD), plant height (PH), panicle length (PL), floret number per panicle (FN), and

kernel weight/1000 grains (KW) in the greenhouse experiment

Lemont 9.6 ? 4.6 4.2 f 3.2 108 79.2 - - - Teqing 5.8 ? 3.2 4.7 f 2.5 117 81.5 - - -

F, (Lemont/Teqing) 16.8 ? 5.6 38.4 f 5.1 98 110.0 24.6 223.0 21.5 R F , (Teqing/Lemont) 50.7 f 6.3 57.0 +- 5.7 98 113.2 25.2 236.2 21.1 F,-RF1“ -33.9*** - 18.6** 0 -3.2 -0.6 -13.2 0.4

“ **, *** represent the significance levels of P 5 0.01 and P 5 0.001, respectively, based on t tests. -, data was not available.

to note that these two putative QTL had zero additive effects but very large underdominance effects of 10.8% (QSs2) and 13.6% ( Q S 5 ) for fertility.

Identification of epistasis affecting SS and PH: Of 4465 two-way ANOVAs performed, the observed num- bers of significant interactions at P < 0.001 were 83 and 74 for SS and PH, of which 18 for SS and 20 for PH were due to linkage of the markers involved in highly significant interactions. Results from multiple regression analyses (with all previously identified QTL fixed in the model) indicated that 38 additional interac- tions for SS and 45 for PH were due to background genetic effects caused by the SS QTL and three plant height QTL (LI et nl. 1995). These were removed from further analyses. The remaining interactions could be divided into two types: (1) interactions between un- linked markers (21 for SS and 11 for PH) that were supposed to be due to epistasis and (2) those between linked markers (six for SS and seven for PH) , which did not represent cases of epistasis but cases where novel phenotypes of SS and PH resulted from rare recombina- tion events within certain genomic regions. The latter cases will be examined separately in a later section.

The magnitudes and characteristics of additive di- genic epistatic estimates: Table 3 shows the interaction effects between homozygotes in the 21 and 11 interac-

Frequency

“ c 4O0 t n n II n I mean of F4 lines 300

0 F2 breeding values

200

100

0 5 11 17 23 29 35 41 47 53 59 65 68

Spikelet sterility (SS, in %)

FIGURE 1.-Frequency distributions of the mean spikelet sterility of 2418 F4 lines and the breeding values of the 255 F2 plants (10 times the observed frequecies) of the Lemont X Teqing cross.

tions affecting SS and PH, and the markers involved. For SS, the mean I? explained by individual interac- tions was 7.62 % 1.28% and ranged from 6.18 to 10.84%, significantly higher than the mean l? (6.05 ? 1.85%, obtained by ANOVA) of the four SS QTL. The average value of 45 significant interaction effects (based on t-tests) was 5.3 ? 2.9% (ranging from 2.7 to 21.2%), which was slightly larger than the doubled mean addi- tive effect (4.9 +- 0.4%) of the SS QTL. No interactions were detected between the SS QTL. However, three of the 32 different markers involved in the interactions flanked three SS QTL (QSs3, QSs6, and QSsX), which were involved in five interactions (Tables 4 and 5). The remaining 27 (84.4%) interactions occurred between 29 unlinked non-QTL markers. On average, each of the 32 markers was involved in 1.31 interactions, suggesting the presence of higher-order interactions.

Evidence of sterility arising from “incompatibilities” between indica and japonica alleles at epistatic loci: When the digenic interaction effects between homozy- gotes in Table 3 (except for the one between CD0348b and RZ660 that will be discussed later) were classified into the parental types and the recombinant type (inter- actions between indica alleles and jnponicn alleles), we found that sterility was usually associated with the re- combinant type interaction (Table 4). The sum of 11

TABLE 2

QTL and their estimated effects on percentage SS detected by MapMaker/QTL in the Lemont X Teqing rice cross

Flanking R“ (; Q T L markers II d (%) LOD

QL%3 RG348-RG944 3.9 -2.0 11.7 5.40 QSsh GRG424 3.1 -1.2 9.6 2.88 QSS8 RG20-RG1034 -4.9 1.2 21.1 4.23 QSSlO RG752-E786 2.9 2.4 7.2 3.31 QSS5 RG556-gl-1 -0.2 13.6 6.4 3.25 QSs2 RG598bRG139 -0.1 10.8 4.1 2.14 Multi-QTL model 60.4 17.64

” a is the additive effect due to substitution of the Lemont allele by the corresponding Teqing allele at QTL, and d is the dominance effect associated with the heterozygote. LOD scores are based on a “free genetics” model (PA’rERSON et al. 1991), including both additive and dominance effects.

I142 %. 1.i 111.

C M I 2 3 4 5 6 7 9 IO I 1 12

---RGXl I I KG173 RG532 R7.2XRx RG472 R(iJ47 DO3XXr

RG4XZa

CDO177

CDOX7

RGYlOn

RG41X

RG403

RGIX?

RG13

RG716

RG34h CDOS44

RZ7hX

RG433

RGh53 - SSQTL 0 Purative“supergenc”regions affecting PH

Pur;lrive“super~ene”regions causing S S

(5.06 to 10.295% (Table 4), which was lower than the ~nean I?’ (8.00 f 6.12%) of the four PH QTI. itlcntified previously (LA NI. 199.5). N o interaction was tletectetl between the PH QTI. ( h I>/ d . 199.5), but two markers flanking PH QTI, (Ql’h? antl Ql’lt9) were involved i n tligenic interactions. Thc remaining nine (81.8%) i n - teractions occurred between 18 ranclom markers. On average, each of the 20 markers was involved i n 1 . I interactions.The mean o f the 19 significant interxtion effects hetween homozygotes was 7.0 5 2.0 cm (ranging from 4.1 to 12.1 cm) , which w a s smaller than the dou- I~led lnean additive effect (7.8 f 2.9 cm) of the four PH QTI, (LI d d . 199.5). The sum of the parental inter- action cffccts resulted i n 32.1 cm increased height. I n contrast, the sum of the recombinant interaction effects resulted i n decreased PH by 4.9 cm.

Putative “supergenes” affecting SS and PH in rice: Table .5 shows three (SS) antl t w o (PH) genomic regions within which significant interactions were de- tected between linked markers. I n these cases, large phenotypic emects \vert almost exclusively associated with the recombinant genotypes. I t was realized that the recomhinant genotypes homoz)pus a t the flanking markers may actually be heterozygous somewhere i n the region since the prohahility that crossing-over events occurred at the same point of this region in hoth n1ale and female gametes is extremely l o w .

SS: Six pairs of linked markers represented three ge- nomic regions i n chromosolnes 5, i, and 11, within which recombination resulted i n tlrasticAly reduced fertility. Thc first case was two interactions representing a 14c,M genomic region betwren RG207 and $1 on

Reproductive Isolation Barriers in Rice 1143

TABLE 3

Interaction effects ( T ~ ) on spikelet sterility (in %) and plant height (in cm) between homozygous alleles at unlinked marker pairs in the Lemont X Teqing rice cross

R' Digenic genotypes"

Chromosome Marker 1 Marker 2 Chromosome (%) 1L/2L 1 L/ 2T 1T/2L 1T/2T

Spikelet sterility 1 1 1 1 1 1 1 2 2 3 ? ? ? ? 3 3 6 6 6 7 9 Plant height 1 1 1 1 1 1 2 ? 5 6 7

RG236 RZ390a RZ288b RG532 RZ390a RG447 RZ390a RG520 RG171 RG418 RZ676 RG9 1 Oa CD087 RG482a RG944 RG104 CDO109b CD0348b CDO109b RG30 RG757

RG381 RG532 RZ776 RZ390a CD0388a RZ776 RG437 RZ761 RG13 RZ2 RG4

RG9 1 Oa CDOlO9a RG 143 RG143 RG424 RZ276Q RG90 1 RG1034 RG103 RG143 RG182 c RG1034 RG1034 CD098 RZ397 RG4 RZ660 CD098 RG463 RZ660

RG437 RG944 RG104 RG424 RG716 Rz777 RG9 1 RG190a CD098 RG241b RG1022

3 3 4 4 6 9

12 8

11 4 5 6 8 8

10 12

7 9

10 9 9

2 3 ? 6 6 9 4 4

10 10 11

6.24 7.26 6.32 8.46 7.53 7.26 7.93 7.66 6.18 7.25 6.34 6.63 8.59 7.22 7.47 8.73

10.84 8.08 7.88 7.72 8.51

6.64 6.42 6.06 8.92 7.46 7.66 6.08 7.74

10.26 6.47

10.29

-6.4**** 3.3*

-0.2 1.3 4.7***

-1.2 3.1*

-7.7"""" 0.1 5.1***

-2.8" 2.7*

-3.8"" -2.5 0.8 2.6 7.7****

-0.0 1 .0

-6.7**** -2.4

-8.0*** 0.0 4.1" 8.6*** 6.3**

12.1**** 0.1

-0.4 3.0

-3.3 -1.0

2.1 -3.9** -2.6

6.8**** 0.5 1.6

-2.1 0.3 3.5*

-1.4 2.8*

-2.2 -1.5

3.1* 1.6 3.7**

-4.7** -0.9

6.7*** 6.4**** 9,2****

1.1 6.0**

-4.2" -2 .3 -2.8 -0.4

1.6 -4.7* 10.4****

2.8 -7.7***

3.1" 0.9 5.5****

-2.7* -4.7***

4.1** -0.2 -1.7

3.0** -1.6 -0.7

5.4*** 3.8** 5.7*** 6.1**** 1 .0

-7,5**** -5.9*** -2.9

1 .o -1.0

2.9 4.5" 1.7

-7.4*** -8.9***

0.9 7.0**

-6.5** -3.5

6.4** 3.1

-3.9"" 1.7 1.8 0.6 1 .0 8.0**** 6.7****

-0.3 -3.9"" 3.9** -3.5**

2.3 6.2****

-3.7** -0.0 -(jJ****

1.8 21.2**** - 1.6

1.4 4.6**

-5.1* 1.7 5.9** 2.6 3.6* 1.2 3.1 8.3***

- 1.4 2.2

- 1.4

" *, **, ***, and **** represent the significance levels of P 5 0.05, 0.01, 0.001, and 0.0001 based on t tests. 1L/2L and IT/ 2T are the parental genotypes where L and T represent homozygotes of Lemont and Teqing, and 1L/2T and 1T/2L are two homozygous recombinant genotypes.

chromosome 5 where an underdominant QTL (QSs5) was identified. This QTL had a zero additive effect (0.1 %) and a very large dominance effect for increased SS by 13.6%. Again, significantly increased SS was asso- ciated with the recombinants within this region. The second case was a 15-cM-long region flanked by RG711 and RG678b on chromosome 7, which was supported by the data from three pairs of linked markers flanking this region. While we did not detect any SS QTL at this region, one class of the two homozygous recombinants generated by crossing over within this region signifi- cantly ( P < 0.0001) increased SS by 16.1%. This was unlikely to be caused by mistyping of the marker geno- types because the probability that the same mistyping error occurred in three markers detected by different enzyme digestions and scored in different hybridization blots was very low. The third was a 26cM-long genomic

region flanked by RZ781 and RG1022 on chromosome 11. One of the classes of homozygous recombinants (the genotype 1L/2T) had 6.2% increased SS (and 49.3 reduced florets per panicle). The other homozygous recombinant (the genotype 1T/2L) had a reduced SS by 5.1%, which could be attributed to its effect on re- duced floret number (data not shown). Two heterozy- gous recombinant classes also resulted in significant effects for increased SS. In addition, the heterozygotes at this region increased SS by 4.8% ( P = 0.03).

PH: The seven interactions between linked markers affecting PH involved two genomic regions. Similar to the cases for SS, there were two genomic regions on chromosomes 2 (between RG256 and RG139) and 6 (between RG179 and CD0544) where crossing over produced recombinants with significantly reduced height. It was also noted that the putative "underdomi-

1144 Z. Li et al.

TABLE 4

Characterization of significant additive interaction effects (T+) on spikelet sterility and plant height between

homozygous alleles at unlinked marker pairs in the Lemont/Teqing rice cross

Spikelet sterility Plant height

No. of Zr, No. of C T , ~ T'/ (%) rt/ (cm)

Parental types Lemont +" 6 -0.8 4 22.1 Teqing + 4 8.3 2 9.1 Total + 10 7.5 15 31.2

- 5 1

6 1

- 11 7

-

Recombinants + 16 52.5 5 -4.9 - 6 6

" + and - represent the parameters for increased SS or PH and decreased SS or PH, respectively.

nance" QTL for fertility was detected in the putative supergene region on chromosome 2.

DISCUSSION

Genetic analysis of postreproductive isolation barri- ers between 0. sativa ssp. indica and ssp. japonica may shed light on the process of speciation in plants, and provide important information for rice improvement. Our results from QTL mapping and interaction analy-

ses suggest several genetic mechanisms that might be responsible for hybrid sterility and hybrid breakdown in rice.

Genetic mechanisms causing hybrid sterility in rice: Three genetic mechanisms, cytoplasmic gene(s), putative differentiated supergenes, and putative com- plementary R m genes, have been implicated in hybrid sterility by our results.

Cytoplasmic gene(s): The impact of cytoplasmic gene(s) on hybrid sterility has been reported in rice (CHANG et al. 1990; PHAM 1990). In the present study, sterility of the reciprocal Lemont/Teqing F, hybrids differed as much as 33.9%. Since 43-57% normal pol- len of the reciprocal F, plants was sufficient to fertilize the female gametes, such a big difference in SS was best accounted for by the cytoplasmic gene(s) on the female gametes. The effect of cytoplasmic gene(s) on hybrid sterility may provide an explanation for the observed varietal specificity of some wide compatibility varieties (KUMAR and VIRMANI 1992; LIN et al. 1992).

Putative supergenes: Supergenes are defined as groups of tightly linked genes, within which recombination will cause reduced fitness (DARLJNGTON and MATHER 1949). The five genomic regions on chromosomes 2, 5, 6, 7, and 11, revealed both by QTL mapping (underdomi- nant SS QTL) and by interaction analyses in the present study, appear to behave as such supergenes. A unique characteristic of these putative supergene regions was that reduced fitness (sterility, reduced height, and grain number per panicle) was exclusively associated with het- erozygotes. It is conceivable that a limited number of differentiated supergenes may cause hybrid sterility in

TABLE 5

Large phenotypic effects of crossing over within certain genomic regions on percentage spikelet sterility and on plant height revealed putative supergene regions in the rice genome

~~~~~~ ~

Genomic regions Interval R' Flanking markers ( r ) (%) 1L/2T 1L/2T 1T/2L 1T/2T 1L/2H 1T/2H 1H/2L 1H/2T

Digenic genotypes"

Spikelet sterility

5 RG207 gl-1 0.14 4.20 -0.9 3.8** 14.6**** 1 .o 2.3 0.3 1.5 2.1 5 RG556 gLJ 0.13 4.43 -1.1 3.4** 1.3 2.3 1.7 3.2"" 4.0** 7 RG4 RG678b 0.19 4.40 0.0 11.7**** - -1.2 - 1 . 1 1.5 0.1 0.1

7 RG711 RG30 0.26 5.64 -2.2 15.1**** -0.3 -1.0 1.3 0.6 2.7* 0.8

-

7 RG711 RG678b 0.15 4.62 -0.7 16.1**** -3.9"" -0.7 0.2 1.3 1.9 1.8

11 RZ781 RG1022 0.18 8.04 -2.6 6.2**** -5.1*** 2.0 3.0* 0.4 3.2* -0.4

Plant height

2 RG256 RG139 0.19 5.52 4.1* 4.4* -9.4*** -1.1 -4.0* 1.6 0.3 -1.0 2 RG256 RG598b 0.17 4.51 2.1 3.1 -10.9**** -0.8 -1.3 1.9 2.7 -0.8

6 RG424 E 7 6 8 0.29 3.34 -2.9 -0.4 -16.3**** -0.9 1.9 1.1 2.5 1.7

6 RG179 E 7 6 8 0.29 5.43 -2.9 -3.8* -16.5**** -1.0 2.9 1.4 3.1 1.9 6 RG716 RZ768 0.24 6.71 -1.2 -4.3* -18.2**** -0.6 3.3 0.3 2.3 3.2

-~ 6 RG424 CD0544 0.26 3.50 -2.6 -2.5 -17.1**** -1.0 1.4 1.2 2.1 2.3

6 RG179 CD0544 0.24 5.34 -3.8 -6.1** -17.4**** 0.1 2.3 0.9 3.8 0.9

" L, T, and H represent the Lemont alleles, the Teqing alleles, and the heterozygotes at the flanking markers; *, **, ***, and **** indicate that the interaction effects are different from zero at P 5 0.05, 0.01, 0.001, and 0,0001, respectively. -, cases of a missing genotype. Underlined markers flanked two regions where underdominance SS QTL were identified.

RcprodtIcIivc Isolatic

cM 2 2 5 5

0 RG.520 P-RGS20 I ',:E: 1 RZ3YOh HC.556 RGSS6.A

Ij :: ~ -: RGZW 20 R Z 4 4 6 X '-RZM6x "256 RZ273 40 . -RG256 gr- I 1 u 2 ~ 1 60 RG139 .LRZ260 + ~-CD0718,RZ386 RG139 1 RG403 RG4Q3 RG182 RZ213 'S'"' (RZ476) RG I82 :I$?$$ I20 RG83 I40 RG13 RGSSS RG13 m386 160 I :

(Caussc cf ol. 1994) W 7 1 8

... RG470

I80 ... RG470

RG437 Rz476

U S 9 9 200

220

2 4 1 L 1 RG346 I RG346 260 ( Y u e f d . 199s: (Li er ol. unpublished) Caussc cf ol. 1994) (Li " a'' 199s)

F K ; ~ . I W : :~.-(:oml';lrison o l ' ricr Iinktgr m;qx sr~ggcbsrs t w o illvcrsions in rhr pt l c t t ivc . supwgvw regions o n ricr chromo- S ~ I I I C S 2 ; t n d 5 .

specific crosscs. Each of these clifferentiatetl supergenes indiviclwlly m;y only cause partial sterility, since only thc g;\metcs that arc recombinant within any specific stlpcqcnc rcgion will be affectctd. Further, different supergenes m;~!' havc different cffccts, and their phentr typic cf'fccts on sterility and/or viability i n hybrids are rxprctctl t o l ,c crlmdativc. In agreement with these expectations. the ovcrall heterozygosity (cstimatcd from a l l 11.5 m;llk-s throrlghout the genome) o f intli- vitltlal F2 plants was not rc+ltctl t o SS o f their derived F I lines ( r = 0 . 0 1 3 ) , b u t the heterozygosity at the putative supcrgcnc regions (flanking markers) was significantly associatcttl w i t h SS ( r = 0.43.5, 1'< 0 . 0 0 0 1 ) . These results suggest that these tlilrercntiatctl supergenes may be an important C ; I I I S ~ of' hybrid sterility i n rice.

Although the nature o f ' these putative suprrgenes is u n k n o w n , some ma!' he c t y p i c sl,rrclrr,nlrc~~rl-angc~nents (cytologically undctcctahlc minor chromosomal ahcrra- tions) (S-n-lrslss 19.50, 19.58). A direct comparison he- t~vren o u r maps (constructcd from the F2 population and a set of219 tlcrivctl RILs from the Lcmont/Tcqing cross) and the ricc RFLP maps pd~lishctl by C A L W I ; PI d . (19!)4) suggests the presence of inversion polymor- phism i n thc putative supergene regions on chromo- somes 2 ;und 5 (Figure 3 ) . Rcducetl fitness in heteroz\l- gous plants :wising from crossing over within the invcrtctl regions is expected if gcnes w i t h i n these re-

) I I kuricrs in l i i w I 1 - 1 3

gions arc related to fitness traits. Segregation distortion against the 0 . s d i r w s.s/l. itlclietr allcle w i t h i n chromo- some 5 supergene region was also rrportetl i n several cases invol\ing , j~rpot l ic /r / ir t t l ie~r crosses (/;/I L.ls I>/ c d . 1992). M'c further ohset~ctl a 1:s ratio against the ittdicct allele at gl-I i n I28 R C I F l [(I,cmont X Tcqing) X 1.c- mont] plants, which suggested that the distorted scg~.ca- gation within this rcxgion WIS clue t o gametic sclcction. O u r rcsdts indicated that the phcnot!.pic eff'ecr o n k r - tility o f this sllpe1-genc was rlcpcntlent 0 1 1 the cyto- plasmic genotype. This is much like the case of meiotic drive i n nlouse i n which sllppresion o f rccorn1)ill;~tion t l w t o inversions within thc I complex appeared t o occur only i n fcmalrs (FKIS(:I I A ~ T 198.5).

Supergenes ( o r the presumed cryptic Iyhritlit!~) arc cxpcctrtl t o inflrlcncc hybrid l,re;ddo\w primarily i n c * d y segregating gencwtions, and their c f fc~ t s will tli- minish ;IS the progeny ;Ipproach complete llomoz!y)s- ity by sclfing.

7Kr p l r l n l i r w K t n p t w / r t d ils rnotl~jiw: M'hile the pre- sumed supergenes may cause vaning tlcgt-ecs of l l ! h k l sterility, they do not adeqrlatcly explain the indcpcn- tlence of' hybrid sterility and hybrid bre;lktlo\\*n i n t!lc I,emont/Tcqing cross, o r i n many other crosscs whcrc cytopl;tsmic cffc~ts \\we not ;I k~ctor. Thc presence of witlc compatibility gcne(s) that c m p r c ~ ~ 1 t the abor- tion of F, hybrid gametes has been suggcstecl i n ricc (KIT~\\IL.IL\ 1962; I K I . I I . \ S I I I ;lnd AIG\KI 1986; K ~ . \ I . \ I < and \'IK\I:\SI 1992), tomato ( KK:K 1 S(i(i, 197 I ) , wheat (LOIX;I

1146 Z. Li et al.

Unlike many reported mutations that cause nuclear male and/or female sterility in rice (Hu and RUTCER 1992), the recessive alleles at the putative complemen- tary loci do not kill gametes in their original parent (Teqing) but only in the indica/japonica hybrid back- ground. Then, a pertinent question is, what kind of genes are they and how can they cause SS or recover fertility in hybrids? Our speculation is that if the gene or gene pair influences recombination frequencies be- tween homologous chromosomes, like the RMl gene in Petunia hybrida (CORNU et al. 1989), they may behave like gamete killers or wide compatibility genes. As dis- cussed above, a primary mechanism causing hybrid ste- rility is inferred to be recombination within differenti- ated supergenes located in several regions of the rice genome. Then, any gene(s) that can influence homolo- gous recombination may behave like wide-compatibility (dominant) or gamete killers (recessive) by affecting recombination frequencies in many supergene regions. Thus, the major gamete killer near the wx and Cloci on chromosome 6 reported in previous studies (KITAMURA 1962; IKEHASHI and ARAKI 1986; SANO 1986, 1990; KU- MAR and VIRMANI 1992), including one of the (putative) complementary genes identified in the present study, may be recombination-modulating genes. Large effects of S-5 and S, leading to two- to fivefold variation in recombination frequencies along their surrounding ge- nomic regions were observed in rice (IKEHASHI and AKAKI 1986; SANO 1990), which strongly suggested the true nature (recombination modulating) of the wide compatibility gene or the gamete eliminator gene in rice.

Control of recombination may also be influenced by a polygenic system, as described by BROOKS and MARKS (1986). PFEIFFER and VOCT (1990), SALL (1990) and FATMI et al. (1993) reported polygenes that affect local recombination frequency primarily in a region-specific manner in maize, barley, and soybeans. Such a poly- genic system may also influence postreproductive isola- tion of rice and may explain some of our epistatic inter- actions that resulted in reduced SS but were not attributable to the related traits.

Genetic mechanisms causing hybrid breakdown in rice: SS and hybrid weakness in F2 and later genera- tions from indica/japonica crosses, or hybrid breakdown as defined by STEBBINS (1958), are much more compli- cated than F1 sterility since these properties are influ- enced by large numbers of genes functioning in both gametophytic and sporophytic stages, as well as by envi- ronment. Our results indicated that the primary mecha- nism causing hybrid breakdown in rice is the uncou- pling of coadapted indica and/or japonica gene complexes by recombination.

The presence of such coadapted gene complexes as a result of natural and/or artificial selection in related plant populations has long been suggested &LARD et al. 1972; WEIR et al. 1974; CLEGG 1978; &LARD 1988; OKA 1988). In rice, intersubspecific progeny tend to

quickly revert back to their parental types as genera- tions advance (Om 1988; SATO 1990), strongly sug- gesting the presence of indica and japonica gene com- plexes. Our results indicated that sterility arises primarily from the incompatible interactions between indica (Teqing) and japonica (Lemont) alleles at many unlinked loci in the genome. This implies that epistasis is an important factor in maintaining the integrity of these gene complexes, in the indica and japonica gene

Our results provide direct evidence for the hypothesis that hybrid breakdown results primarily from dishar- monic interactions between unlinked loci, which refers to the complementary gene system (DOBZHANSKY 1936; MULLER 1940; STEBBINS 1958; OKA 1988). Genes in this complementary system appear to directly (QTL) or in- directly (epistasis) affect fitness correlates such as head- ing date and floret number per panicle. The presumed coadapted gene complexes appear to have large effects on hybrid breakdown but little impact on hybrid steril- ity since the F, plants were largely fertile and vigorous. This suggests that dominance relationships at many of these epistatic loci afford fertility to F1 plants, as pre- dicted by recent theoretical work (ORR 1995; TURELLI and ORR 1995).

Recombination is a key genetic element for origin of reproductive isolation: Both hybrid sterility and hybrid breakdown appear related to recombination. In self- pollinated plants like rice, chromosomal mutations may evolve quickly since they impose minimal genetic load on a population of homozygous individuals. Reproduc- tive isolation mechanisms based on cryptic chromo- somal rearrangements may be more effective. Genes that suppress homologous recombination at meiosis may enhance the formation of reproductive isolation (hybrid sterility).

Sterility and weakness can also arise in interspecific hybrid progenies from random assortment of nonho- mologous chromosomes (DOBZHANSKY 1936; MULLER 1940; STEBBINS 1958). Our results indicated that hybrid breakdown may involve large numbers of genes and complex high order gene interactions, as reported in Drosophila (CABOT et al. 1994; PALOPOLI and WU 1994) and predicted by recent theoretical work (ORR 1995; TURELLI and O m 1995).

Impacts of genetic mechanisms affecting SS on segre- gation distortion and gene mapping: Segregation dis- tortion has been commonly observed in wide crosses between distantly related taxa. In our study, distorted segregation at 18 marker loci (LI et ai. 1995) could be attributed to selection at either gametophytic and/or sporophytic stages as well as to random sampling varia- tion. In the Lemont/Teqing population, 14 of the 18 markers showing segregation distortion were located on four genomic regions on chromosomes 3, 5, 6, and 9, three of which were associated with SS QTL. Severely distorted segregation due to the differentiation of co- adapted gene complexes may occur in segregating pop-

pools.

Reproductive Isolation Barriers in Rice 1147

ulations from divergent parents, as was noted in an indica/juponicu recombinant inbred population ( W ~ G et al. 1994). Selection for differentiated coadapted epi- static genes could explain “pseudolinkage” between some unlinked markers (WANG et al. 1994).

Differentiated supergenes or cryptic chromosomal variation may have an impact on genomic studies. First, breakdown of supergenes by crossing over would result in recombinants with reduced fertility, and thus re- duced genetic distances between markers in affected genomic regions. Second, presence of cryptic structural variation between parents of a mapping population may cause problems in determining marker order within a supergene region since the parents may have different orders of the affected markers. Thus, in populations generated from wide crosses including interspecific or subspecific crosses, information about the number and effects of such supergene regions is valuable in design- ing gene mapping and map-based cloning experiments.

We are grateful to Dr. TRUDY F. C. MACKAY and two anonymous reviewers for many critical comments and suggestions in the early draft of this manuscript. We also thank S. D. TANKSLEY and S. R. MCCOUCH for providing us with the DNA probes. This research was supported by USDA-ARS, Southern Plain Area, The Texas A&M Uni- versity System Agricultural Research and Extension Center, The Texas Rice Research Foundation, and the Texas Advanced Technol- ogy Program grants to W.D.P., Z.L. and J.W.S.

LITERATURE CITED

ABBO, S., and G. LADIZINSKY, 1994 General aspects of hybrid embryo abortion in the genus Lens L. Heredity 7 2 193-200.

ALw, R. W., 1988 Future directions in plant population genetics, evolution and breeding, pp. 1-18 in Plant Population Genetics and Germplasm Resources, edited by A. H. D. BROWN, M. T. CLEGG, A. L. KAHLER and B. S. WEIR. Sinauer Assoc. Inc., Sunderland, MA.

ALLARD, R. W., G. R. BABBEI., M. T. CLECG and A. L. KAHLER, 1972 Evidence for coadaptation in Avena barbata. Proc. Natl. Acad. Sci. USA 69 1820-1824.

BROOKS, L. D., and R. W. MARKS, 1986 The organization of genetic variation for recombination in Drosophila melanogaster. Genetics 114 525-547.

UOT, E. L., A. W. DAVIS, N. A. JOHNSON and C. I. Wu, 1994 Genet- ics of reproductive isolation in the Drosophila simulans clade: com- plex epistasis underlying hybrid male sterility. Genetics 137: 175- 189.

CAMERON, A. G., and R. MOAV, 1957 Inheritance in Nicotiana tabu- cum. XXVII. Pollen killer, an alien genetic locus inducing abor- tion of microspores not carrying it. Genetics 42: 326-335.

CAUSSE, M. A,, T. M. FULTON, Y. G. CHO, S. N. AHN, J. CHUNWONGSE et al., 1994 Saturated molecular map of the rice genome based on an interspecific backcross population. Genetics 138 1251- 1274.

CHANC, T. T., Y. PAN, Q. CHU, R. PEIRIS and G. C. LORESTO, 1990 Cytogenetic, electrophoretic, and root studies of javanica rices. Proceedings of the Second International Rice Genetics Sympo- sium, International Rice Research Institute, Manila, Philippines, pp. 21-31.

CHU, Y. E., H. MORISHIMA and H. I. Om, 1969 Reproductive barri- ers distributed in cultivated rice species and their wild relatives. Jpn. J. Genet. 44: 207-233.

CLEGC, M. T., and R. W. ALLARD, 1972 Patterns of genetic differenti- ation in the slender wild oat species Avena barbata. Proc. Natl. Acad. Sci. USA 69: 3043-3048.

CORNU, A,, E. F m m and C. MOUSSET, 1989 A genetic basis for the variation in meiotic recombination in Petunia hybrida. Genome 32: 46-53.

DARLINGTON, C. D., and K. MATHER, 1949 The E k t s of Genetics. Allen and Unwin, London.

DAVIS, A. W., E. G. NOONBERG and C. I. WU, 1994 Evidence for com- plex gene interactions between conspecific chromosomes under- lying hybrid female sterility in the Drosophila simulans clade. Ge- netics 137: 191-199.

DELORES, R. C., T. T. CHANC and D. A. RAMIREZ, 1975 Cytogenetics of sterility in F, hybrids in indica X japonica varieties of rice. Cytologia 40: 639-647.

DOBZHANSKY, T., 1936 Studies on hybrid sterility. 11. Localization of sterility factors in Drosophila pseudoobscura hybrids. Genetics 21: 113-135.

FATMI, A,, C. G. PONELEIT and T. W. PFEIFFER, 1993 Variability of recombination frequencies in the Iowa Stiff Stalk Synthetic (&a mays L.). Theor. Appl. Genet. 86: 859-866.

FRISCHAUF, A.M., 1985 The T/t complex of the mouse. Trends Genet. 1: 100-103.

GLASLMANN, J. C., 1987 Isozymes and classification ofAsian rice vari- eties. Theor. Appl. Genet. 74: 21-30.

HENDERSON, M. F., B. P. YEH and B. EXNER, 1958 Further evidence of structural differentiation in the chromosomes as a cause of sterility in inter-varietal hybrids in rice. Cytologia 2 4 415-422.

Hu, J., and J, N. RUTGER, 1992 Pollen characteristics and genesis of induced and spontaneous genetic male-sterile mutants in rice. Plant Breeding 109: 97-107.

HUTTER, P., and M. ASHBURNER, 1987 Genetic rescue of inviable hybrids between Drosophila melanogaster and its sibling species. Nature 327: 331-333.

IKEHASHI, H., and H. ARAIU, 1986 Genetics of F, sterility in the remote crosses of rice. Proceedings of the International Rice Genetics Symposium, International Rice Research Institute, Ma- nila Philippines, pp. 119-130.

JAIN, S. K., 1975 Population structure and the effects of breeding system, pp. 15-36 in Crop GeneticResources for Today and Tommow, edited by 0. H. FRANKEL and J. G. H A W S . Cambridge University Press, Cambridge.

JENNINGS, P. R., 1966 Evolution of partial sterility in indica X japonica rice hybrids. Int. Rice Inst. Tech. Bull. 5: 63.

KITAMURA E., 1962 Genetic studies on sterility observed in hybrids between distantly related varieties of rice, w z a sativa L. Bull. Chugoku Agr. Exp. Sta. Series A 141-205.

KUMAR, R. V., and S. S. VIRMANI, 1992 Wide compatibility in rice ( W z a sativa L.). Euphytica 6 4 71-80.

LI, Z. K., and J. N. RUTGER, 1997 Geographic distribution and multi- locus structure of isozyme variation in rice (Ofyza saliva L.). Theor. Appl. Genet. 145: 453-465.

LI, 2. K., S. R. M. PINSON, J. W. STANSEI. and W. D. PARK, 1995 Identi- fication of QTL for heading date and plant height in rice using RFLP markers. Theor. Appl. Genet. 91: 374-381.

LI, Z. K., S. R. M. PINSON, W. D. PARK, A. H. PATEWON and J. W. STANSEL, 1997 Epistasis for three grain yield components in rice (Olyza sativa L.). Genetics 145: 453-465.

LIN, s. Y., H. IKEHASHI, S.YANAGIHARAandA. mWASHIMA, 1992 Segre- gation distortion via male gametes in hybrids between Indica and Japonica or wide-compatibility varieties of rice (Olyzu sativa L.). Theor. Appl. Genet. 8 4 812-818.

LOEGERING, W. Q., and E. R. SEARS, 1963 Distorted inheritance of stem-rust resistance of timstein wheat caused by a pollen-killing gene. Canad. J. Genet. Cytol. 5: 65-72.

MORINAGA, T., and H. KURIYAMA, 1958 Intermediate type of rice in the subcontinent of India and Java. Jpn. J. Genet. 7: 253-259.

MULLER, H. J., and G. PONTECORVO, 1940 Recombinants between Drosophila species, the F1 hybrids of which are sterile. Nature 146: 199.

OW, H. I., 1974 Analysis of genes controlling F1 sterility in rice by the use of isogenic lines. Genetics 77: 531-534.

Om, H. I., 1988 Functions and genetic bases of reproductive barri- ers, pp. 181-205 in Chigin of Cultivated Rice edited by. Elsevier, New York.

ORR, H. A., 1995 The population genetics of speciation: the evolu- tion of hybrid incompatibilities. Genetics 139 1805-1813.

PALOPOLI, M. F., and C. I. WU, 1994 Genetics of hybrid male sterility between Drosophila sibling species: a complex web of epistasis is related in interspecific studies. Genetics 138: 329-341.

PFEIFFER, T. W., and S. VOGT, 1990 Variability for recombination frequencies in the MI2 soybean population. Crop Sci. 30: 545- 549.

1148 Z. Li et al.

PHAM, J. L., 1990 Genetic diversity and intervarietal relationships in rice (Oryza sativa L.) in Africa. Proceedings of the Second International Rice Genetics Symposium, International Rice Re- search Institute, Manila, Philippines, pp. 55-65.

RICK, C. M., 1966 Abortion of male and female gametes in the to- mato determined by allelic interaction. Genetics 53: 85-96.

RICK, C. M., 1971 The tomato Ge locus: linkage relations and geo- graphic distribution of alleles. Genetics 67: 75-85.

SAIJ., T., 1990 Genetic control of recombination in barley. I. Varia- tion in linkage between marker genes. Hereditas 112: 171-178.

SANO, Y., 1986 Sterility barriers between Oryza sativa and 0. glaber- rima. Proceedings of the International Rice Genetics Sympo- sium, International Rice Research Institute, Manila, Philippines,

SANO, Y., 1990 The genetic mature of a gametic eliminator in rice. Genetics 125: 183-191.

SATO, Y. I . , 1990 How was rice differentiated into indica and japon- ica. Proceedings of the International Rice Genetics Symposium, International Rice Research Institute, Manila, Philippines, pp. 45-53.

SHAS’I‘KY, S. V. S., and R. N. MISRA, 1961 Pachytene analysis o n olyza sativa L. Sterility in japoniceindica hybrids. Curr. Sci. 30: 70-71.

STEBBINS, G. L., 1950 Variation and Evolution in Plants. Columbia University Press, New York.

STEBBINS, G . L., 1958 The inviability, weakness, and sterility of inter- specific hybrids. Adv. Genet. 9: 147-215.

pp. 109-118.

T U ~ I L I , M., and H. A. O m , 1995 The dominance theory of HAL DANE’S rule. Genetics 140: 389-402.

XIE, F. M., 1993 Hybrid performance and genetic diversity as re- vealed by RFLPs in rice ( W z a sativaL.). Ph.D. dissertation, Texas A&M University.

WANC, G. I.., 1992 RFLP mapping of major and minor genes for blast resistance in a durable resistant rice cultivar. Ph.D. disserta- tion UP at Los Banos Philippines.

WAN(;, Z. Y., and S. D. TANKSLEY, 1992 Restriction fragment Icngth polymorphism in Oqza sativa L. Genome 32: 11 13-1 118.

WATANABE, T. K., 1979 A gene that rescues the lethal hybrids be- tween Drosophila melanogaster and D. simulans. Jpn. J. Genet. 5 4 325-331.

WEIR, B. S., R. W. AI.IARD and A. L. KAHLEK, 1974 Further analysis of complex allozyme polymorphisms in a barley population. Ge- netics 78: 911-919.

Wu, C. I., and A. W. DAVIS, 1993 Evolution of postmating rcproduc- tive isolation: the composite nature of Haldane’s rule and its genetic bases. Am. Nat. 142: 187-212.

YAO, S.Y., M. T. HENI)ERSC)N and H. E. JOUON, 1958 Cryptic struc- tural hybridity as a probable cause of sterility in inter-varictal hybrids in cultivated rice. Cytologia 23: 46-55.

Yu, Z. H., S. R. MCCOUCH, T. KINOSHITA, S. SATO and S. D. TANI(SI.EY, 1995 Association of morphological and RFLP markers in rice ( O q u n tati7m L.) . Genome 38: 566-574.

Communicating editor: T. F. C. MACKAY