Copyright 0 1994 by the Genetics Society of America

X A Detailed RFLP Map~ of Cotton, Gossypium hirsutum Gossypium barbadense: Chromosome Organization and Evolution in a Disomic Polyploid Genome

Alesia J. Reinisch," Jian-mjn Dong,* Curt L. Brubaker,t David M. Stelly," Jonathan F. Wendelt and Andrew H. Paterson" *Department of Soil and Crop Sciences, Texas A&M University, College Station Texas 77843-2474, and +Department of Botany, Iowa State University, Ames, Iowa 50011 Manuscript received January6, 1994 Accepted for publication July12, 1994

ABSTRACT We employ a detailed restriction fragment length polymorphism(RFLP) map to investigate chromo- some organization and evolution in cotton, a disomic polyploid. About 46.2% of nuclear DNA probes detect RFLPs distinguishing Gossypium hirsutum and Gossypiumbarbadense; and 705 RFLP loci are assembled into 41 linkage groups and 4675 cM. The subgenomic origin (A us. D) of most, and chro- mosomal identityof 14 (of 26), linkage groups is shown. TheA and D subgenomes show similar recom- binational length, suggesting that repetitive DNA in the physically larger A subgenome is recombina- tionally inert. RFLPs are somewhat moreabundant in the D subgenome. Linkage among duplicatedRFLPs reveals 11 pairs of homoeologous chromosomal regions-two appear homosequential,most differ by in- versions, and at least one differs by a translocation. Most homoeologies involve chromosomes from different subgenomes, putatively reflecting the n = 13 to n = 26 polyploidization event of 1.1-1.9 mil- lionyears ago. Several observations suggest that another,earlier, polyploidization event spawned n = 13 cottons, at least 25 million years ago. The cotton genome contains about 400-kb DNA per cM, hence mapbased gene cloning is feasible. The cotton map affords new opportunities to study chro- mosome evolution, andto exploit Gossypium genetic resources for improvementof the world's leading natural fiber.

HE genus Gossypium L. has long been a focus of raploids contain two distinct genomes, which resemble T genetic, systematic andbreeding research. Gos- the extant A genome of G. herbaceum (n= 13) and D sypium comprises about 50 diploid and tetraploid spe- genome of G. raimondii Ulbrich (n= 13), respectively. cies indigenous to Africa, Central and South America, The A and D genome species diverged from a common Asia, Australia, the Galapagos, and Hawaii(FRYXELL ancestor about 6-1 1 million years ago (WENDEL1989). 1979, 1992). Cultivated types derived from fourspecies, The putative A X D polyploidization event occurred namely G. hirsutum L. (n= 2x = 26), G. barbadense L. in the New World, about 1.1-1.9 million years ago, (n= 2x = 26), G. arboreum L. (n = x = 13), and G. and required transoceanic migration of the maternalA herbaceum L. (n= x = 13), provide the world's leading genome ancestor (WENDEL1989, WENDELand ALBERT natural fiber, cotton,and are also a major oilseed crop. 1992), which is indigenous to the Old World (FRYXELL Cotton was among the first species to which the Men- 1979). Polyploidization was followed by radiation and delian principles were applied (BALLS1906), and has a divergence, with distinct n = 26 AD genome species now long history of improvement through breeding,with sus- indigenous to Central America (G. hirsutum), South tained long-term yield gains of 7-10 kg lint/ha/yr America ( G. barbadense, G. mustelinum Miers ex Watt), (MEREDITHand BRIDGE1984). The annual world cotton the Hawaiian Islands (G. tomentosum Nuttall ex See- crop of ca. 65 million bales (of 218 kg/bale) , has a value mann), and the Galapagos Islands ( G. danuinii Watt) of ca. US$15-20 billion/yr. (FREXELL1979). Diploid species of the genusGossypium are all n = 13, Variation in ploidy among Gossypium spp., together and fall into 7 different "genome types," designated A-G with tolerance of aneuploidy in tetraploid species of Gos- based on chromosome pairing relationships (BEASLEY sypium, has facilitated use of cytogenetic techniques to 1942; ENDRIZZIet al. 1984). A total of 5 tetraploid explore cottongenetics and evolution. Among 198 mor- (n = 2x = 26) species are recognized. All tetraploid phological mutants described in cotton, 61 mutant loci species exhibit disomic chromosome pairing (KIMBER have been assembled into 16 linkage groups, through 1961). Chromosome pairing in interspecific crosses be- the collective results of many investigators. Using nul- tween diploid and tetraploid cottons suggests that tet- lisomic, monosomic, and monotelodisomic stocks, l lof

Genetics 138 829-847 (November, 1994) 830 A. J. Reinisch et al. these linkage groups have been associated with chro- DNA probes: Mapped DNA probes included PstI-genomic mosomes (ENDRIZZIet al. 1985). fragments from G. raimondii (prefix G: 75 probes), G. her- Our objectives wereto use low copy DNA markers to baceum subsp. africanum Watt (Mauer) (accession A,-73;pre- fix A 192 probes), and G. hirsutum accession “TM-1” (pre- investigate cotton genome and chromosome organiza- fured M: 21 probes; and P 78 probes), lowcopy genomic tion at themolecular level, in a cross of G. hinuturn X restriction fragments selected from a libraly of G. barbadense G. barbadense. We have established a detailed restriction cv. “Pima S6” DNA cut with a mixture of five different blunt- fragment length polymorphism (RFLP) map of cotton, cutting fourcutters (prefmed LXP: 1 probe and PXP:40 identifymg genomic origins of, and homoeologies probes; to be described elsewhere; X. Zhao and AHP,in prepa- ration), genomic probes derived from G. hirsutum cv. “Tam- = among, most ofthe linkage groups in tetraploid (n 2% cot GNCH” and containing Not1 sites (prefixed pVNC), and = 26) cotton, and characterizing the nature and fre- cDNAs (143 probes) from a library prepared from drought- quency of rearrangements which distinguish homoe- stressed tissue ofG. hirsutum accession ”T25.”All probes were ologs. Identification of homoeologous chromosomal re- prepared by PCR amplification of plasmid or phage DNA, gions reveals the approximate locus of DNA probes using M13 primers (SP010and SP030, Operon, Alameda, Cali- monomorphic in one subgenome but polymorphic in fornia), followed by chromatography through homemade Sephadex G50 (Sigma) spuncolumns (SAMBROOKet al. 1989). the other, and representsa means to greatly increase the DNA extraction, electrophoresis, blotting, and hybridiza- informativeness of genetic maps in polyploids such as tion: DNA extractions followed PATEMONet al. (1993).Replica cotton. Characterization of the comparative organiza- blots wereprepared using each of the restriction enzymes used tion of homoeologous chromosomes promises to in- for polymorphism screening (below). DNA electrophoresis, crease the markerdensity of molecular maps in disomic blotting, probe labeling, Southern hybridization, and autora- diography were as described by CHITTENDENet al. (1994). ex- polyploids, facilitating both genetic and physical cept that after liquid hybridization, filterswere washed at 2X, mapping applications. lX, and 0.5X SSC (instead of three washes at 0.1X). Individual DNA probes were screened for polymorphisms MATERIALSAND METHODS distinguishing the grandparents (G. hirsutum race “palmeri” Genetic stocks: The genetic mapping population com- and G. barbadense acc. “K101”) using 4-6 restriction enzymes prised 57F2 individuals froma cross between single individuals (EcoRI,EcoRV, HindIII, XbaI in all cases,and BamHI, and CfoI of G. hirsutum race “palmeri” (see BRUBAKERand WENDEL in a subset of cases), and filters werewashed for 20 min 1993) and G. barbadense acc. “K101.” These accessions were each in 2X, lX, and 0.5X SSC, 0.1%SDS at 65” prior to selected because they are largely homozygous, and relatively autoradiography. primitive, thus free from the interspecific introgression that Data analysis Linkage maps and related statistics were de- characterizes somepopulations of these two species from sym- termined using MapMaker 1.0 (LANDER et al. 1987; provided patric portions of their indigenous ranges (PERCYand WENDEL by L. PROCTOR,DuPont Co.) running on a Macintosh Quadra 1990; WENDELand ALBERT 1992; BRUBAKERet al. 1993), as well 700 under System ’7.01. A LOD (base 10-log of the ratio be- as cultivated types (PERCYand WENDEL 1990;G. WANCand A. tween odds of linkage and odds of non-linkage) score of 4.0 H. PATERSON,unpublished data). Self-pollinated progeny from was used toinfer linkage in two-point analyses: sincethe cotton the “palmeri“ and “K101” parents were used for preliminary RFLP map spans nearly 5000 cM, the standard LOD 3.0 used surveys of DNA polymorphism, prior to mapping. for smaller genetic maps is not sufiiciently stringent. lnitial Genomic identityof RFLP alleles (see Figures2 and 3) were frameworks of markers were evaluated using the “compare” inferred based upon comparison of G. barbadense and G. hir- function, and only orders preferred by a LOD of at least 2.0 sutum parents to accessionsrepresenting the diploid progeni- (e.g.,a 100-fold difference in likelihood) over alternate orders tor genomes of tetraploid Gossypium, i.e., the A genome dip were accepted. New markers were added to the frameworks loids G. arboreum (accession 447) and G. herbaceum using the “try” function, and the maximum likelihood orders (accession 4 A,97), and the D genome diploids G. triEobum of each linkage group ultimately verified by the “ripp1e”func- (Mocino & Sese ex DeCandolle) Skovsted (un-named acces- tion. Segregation ratios were calculated SASusing (SM Institute, sion) and G. raimondii (un-named accession). 1989). Deviation of segregation ratios from the Mendelian ex- Genetic stocks monosomicfor G. barbadense chromosomes pectation was assessed by calculating chiquared values. 1, 2, 4, 6, 9, 10,17, 20 and 25, and stocks mono-tele Since the mapping population is tetraploid, determination disomic for G. barbadense chromosome arms>Lo, 14L0,15Lo, of allelism between genomic restriction fragments unique to l8L0, 22Sh and 26Sh were used to identlfy linkage groups the respective grandparents is more complicated than in dip- corresponding to these chromosomes. A monosomic substi- loids, or largely diploidized paieopolyploids.Basically, allDNA tution stock hasa single chromosome from a donor genotype polymorphisms fit one of the following three scenarios: (1) (in our case, G. barbadense acc. “Pima 3-79”) substituted for Only one of the two homozygous grandparents had a unique the corresponding pair of chromosomesof the recipient gene fragment. This was commonly observed, as many fragments type (in our case, G. hirsutum acc. “TM-1”). A mono-telo- werepresumably masked by co-migrating duplicated frag- disomic stockis similar, except that a single chromosome arm ments. Consequently, mapping was based on presence or ab- from the donor is substituted for a chromosome pair in the sence of an individual fragment, because heterozygotes could recipient. In the series of monosomic and mono-telodisomic not be reliably(e.g., qualitatively) distinguished from homozy- substitution stocks, an individualgenetic stock exhibitsthe G. gotes for the variant fragment. (2) Each of the two homozy- barbadense allele only for loci fallingon the substituted chro- gous grandparents had one unique fragment. In principle, the mosome (or arm). hsignment of chromosome numbers in two fragments could be allelic, or could represent polymor- tetraploid cotton is based on pairing relationships in diploid phisms at two different loci. These alternatives were readily X tetraploid crosses, with chromosomes 1-1 3 corresponding distinguished in most cases by the following logic: Ifthe two to the A subgenome, and chromosomes 14-26 corresponding fragments are allelic, all individualsin the mapping population to the D subgenome (UMBER1961). must have at least one of the two. On the other hand, if the Organization of Cotton Chromosomes 831

TABLE 1 Hybridization of homologous and homoeologous cotton DNA probes to EcoRIdigested genomic DNA from A, D and AD genome cottons

A. Number of restrictionfragments B. Fraction of restrictionfragments hybridizing to genomic DNA strongly hybridizingto genomic from different genomes. DNA from different genomes. Source of genomic DNA Source of genomic DNA

probe’ A D Source ofA DNA probe’ D AD A Total AD Average

A genome PstI fragments 0.65 157 0.69 1550.62 1870.64 499 D genome PstI fragments 0.56 178 0.60 1610.52 2320.53 571 AD genome PstI fragments 149 126 0.59226 0.60 500.63 1 0.55 AD genome cDNAs 167 154 262 0.57583 0.51 0.53 0.54 Total 651 596 907 2154 Average per DNA probe 2.71 0.58 2.96 0.56 4.12 0.58 0.60 ‘A total of 55 DNA probes from each of the four classes were used. Only 55 D genome probes were available which provided the necessary data; probes from the other three classes were chosen randomly from candidates for which equivalent datawas available.

two fragments represent polymorphisms twoat unlinked loci, nomes of both diploid and tetraploid cottons (Table l, an average of 1/16 of the individuals in the mapping popu- Figure la). Based on a sample of 220cotton DNA probes lation shouldnot show either fragment (assuming Mendelian for which equivalent data were available[55 G. hirsutum segregation ratios), and in a population of 57 individuals, there is approximately 95% confidence of seeing such an in- cDNAs, and 55 PstI-digested genomic sequencesderived dividual. In cases where one or more such individuals were from each of three species, G. herbaceum (A genome), observed,the fragments weremapped as independent G. raimondii (Dgenome) and G. hirsutum (AD ge- dominant/recessive loci,as in case #1, above. If no such in- nome)], the average “low copy” cotton genomic probe dividual was observed, the fragments were tentatively mapped as alleles. This algorithm should be reliableall casesfor except hybridizes to 2.71 genomic EcoRI fragmentsin those rare cases wherethe two fragments represent linked du- the (n= 13) D genome, 2.97 genomic EcoRI fragments plications in “trans”configuration. This finalpossibility is dif- in the ( n = 13) A genome and 4.12 genomic EcoRI ficult to exclude with certainty in any genetic mapping ex- fragmentsin the (n = 26) AD genome. Since DNA periment, although the numberof cases fitting this scenario probe lengthsaveraged ca. kb (ranging fromca. 0.5 to is small. Should the alternative fragments represent different 1 loci separatedby an appreciable recombinational distance, at- 2.0 kb), while EcoRI genomic fragments averaged ca. tempting to map them as alleles would result in conflicting 4 kb, only about one-fourth of the genomic probes information, and either imprecise mapping or an inordinatelyshould contain internal EcoRI sites, giving a mean for large expansion of the most likely map interval (relative to its true “single copy” probes of 1.25 fragments per probe. recombinational length without this “locus”).A few such cases were found, and the fragments were re-mapped as indepen- Even in contemporary diploid (n = 13) cottons, most dent loci. (3) Eachof the two homozygous grandparents DNA probes detect more thantwice this number of frag- showed multiple unique fragments. Thisis similar to scenario ments, with an even larger number of fragments found (2), except that the possibility of allelism must be evaluated in tetraploid (n = 2x = 26) cottons. between all possible pairs of loci derived fromalternate grand- parents. In most cases, allelism could beinferred notonly by Most probes didnot hybridize equallywell to different complementarity betweenone pairof genomic fragments, but genomic restriction fragments. EcoRI genomic frag- also by lack of complementarity toother genomic fragments. ments denoted as “weaklyhybridizing” (less than 10%of Nomenclature: Genetic loci detectedby DNA markers have signal in thelane, based on visual assessment), ac- been designatedby the name of the DNA probe detecting the counted for44% of fragments inthe A genome diploids, locus. If a probe detectedRFLPs at multiple loci, letters(a, b, c, . . . ) were arbitrarily assignedto these loci. If a locus could 42% of fragments in the D genome diploids, and 40% be assigned to a particular chromosome (by analysis of aneu- of fragments inthe tetraploids (Table1). In eachof the ploid stocks as describedabove), thelocus name was followed three genomes, DNA probes hybridizing to multiple by an indication of chromosome number (e.g., pAR042.05, fragments detected similarly increasing proportions of for chromosome 5). Linkage groups not yet assigned to chromosomes are des “weakly hybridizing” fragments (Figure 1, b-d). While ignated by the subgenome from whence they appear to be we acknowledge that ourvisual classification of“strong” derived (A, D or U if unknown)based upon alloallelic us. “weak”bands may be somewhat confounded with the evidence(see below), and numerically by decreasing number of bands per lane (e.g., multicopy probes are recombinational length within a subgenome. more likely to include copies which explain less than 10% of signal), it remains clear that most low copy RESULTS probes detect both “strong” and “weak” bands. Sequence duplication in diploid and tetraploid Gos Low copy DNA sequences in the A, D and AD ge- sypium genomes: Genomic hybridization patterns of nomes have not diverged extensively, as PstIdigested cDNAs and hypomethylated (PstIdigested) genomic se- genomic probes from each source showed only small quences revealed considerable duplication in the ge- differences in detection of genomic fragments across 832 A. J. Reinisch et al.

70

60

50

40

30

20 .A a OD FIGURE1.-Copy number and hy- 10 AD bridization affinity (signal intensity) of DNA probes in A, D, and AD ge- 0 nome cottons. (a) Histogram of the 123 4 5 6 7 8 9 1011121314 number of DNA probes detecting 1,

1.0 -I 2, 3, or up to 14 (the maximum ob- - Q served)genomic EcoRI fragments, 0.8 - - - for the A, D and AD genomes. The - - average “low copy” cotton genomic Q probe hybridizedto 2.71 genomic EcoRI fragments in the (n = 13) D d - b genome (represented by G. raimon- dii),2.97 genomic EcoRI fragments in y = 0.12540 + 0.56371’LOG(x) 8 = 0.771 the (n= 13) A genome(G. fmhxmm), ,.~.~.~.l.l.~.~and 4.12 genomicEd fragments in the 0 123 4 5 6 7 8 9 1011121314 ( n = 26) AD genome ( G. himLcum). The most highly duplicated probes hybrid- 1.0 1 izedtoll (D),12(AD),and14(A)fhg- 0.8 * ments,respectively. (b-d)Fraction of 1 -.. c e ‘’weaklyhybridizing” restriction frag- - / Q 0.6 - / L ments detected, us number of total re- striction fragments detected, in theA, D 0.4 - - and AD genomes,respectively. Most ” C probes did not hybridize equaUy well to 0.2 - y = 0.090493 + 0.61139’LOG(x) = 0.781 Merent genomic restriction fragments, 0.0 - i~i~i~i~l~l~i~l~l~l~l~~but detecting an increasing proportion Q 12 3 4 5 6 7 8 91011121314 of “weakly hybridizing” hgments with increasing total number of fragments. 1.o 1

0.8 El T P - 0.6

0.4

0.2 L nn y =- 0.036219 + 0.65373’LOG(~) F? = 0.780 thedifferent genomes (Table 1). Cloned genes suggest that we should detect a lower level of duplica- (cDNAs), and PstI sequences derived from the smallest tion, all other factors being equal. One difference which (D) genome (c = 0.68 pg; GEEVERet al. 1989), detected may have afforded us greater power to detect weakly appreciably moregenomic fragments than PstI se- homologous genomic DNA fragments is the fact that we quences from the A ( c = 1.05 pg; GEEVERet al. 1989) or used -2 X as much DNA on genomic Southern blots as AD genomes (c = 1.8-2.5 pg; ARUMUNGANATHAN and did GALAUet al. (1988); specifically, 2-5 pg us. 1-3 pg (stoi- EAIUE 1991; MICHAELSONet al. 1991). chiometrically adjustedin proportion to the cvalue ofthe The present results regarding sequence duplication respective genomes, in both experiments). The effect of differ somewhat from previously published results which differences in blottingprotocol, and other conditions,can- suggested that most Lea genes are represented in only not be evaluated based upon present information. one copy in diploid Gossypium, and two copies in tet- Frequency of DNA polymorphisms between G. hir- raploid Gossypium (GALAUet al. 1988). The higher strin- sutum and G. barbadense: Averaged across all classes of gency we employed in our final wash (0.5 X SSC at 65” probes, 46.2% of DNA probes detected one or more herein, us. 1 X SSC at 68”; GALAUet al. 1988) would DNA polymorphisms between G. hirsutum “palmeri” OrganizationChromosomes of Cotton 833

TABLE 2

Observed and expected frequencies of RFLPs between G. hirsutum race ‘‘palmm”’ and G. barbadense acc. “KlOl with different numbers of restriction enzymes

B. No. of enzymes per probe detecting RFLPs A. Average frequency of Enzyme probes detecting RFLPS Obs. Exp. 7‘ (1 d.f.)’ BamHI 0.10 0 170 594 350 CfOI 161 0.19 447 1 179 EcoFU 0.22 2 33 147 235 EcoRV 0.20 3 103 65 22 HzndIII 0.17 10 4 58 234 XbaI 0.16 5 15 1 257 6 12 0 5681 Total 1108 1108 6559 Only probes for which data from all six restriction enzymes was available were used in the analysis. Expected values were calculated from the average frequencies of RFLF’s for each restriction enzyme or combination of restriction enzymes, using the binomial expansion, and summed across the Dossible combinations of restriction enzvmes. ExDected values are rounded to the nearest integer in this table, however exact values were used to calculate chi-square statistics. All values had a deviation significant at 0.001 or less. and G. barbadense “K101,” based on evaluation of 782 not attempted. Our results underestimate the true fre- PstI-genomic probes and 326 anonymous cDNAs quency of such RFLPs, as we preferentially mapped screened with the same six restriction enzymes. The fre- DNA probe X restriction enzyme combinations show- quency of RFLPs for genomic clones (47.0%) was not ing the maximum numberof bands differing between significantly different from that for cDNAs (44.2%). parents. [Since determination of allelism in a tetra- DNA probes showing no polymorphism, or showing ploid requires genetic analysis, simple observation of polymorphism with each of several enzymes are found apparent differences between parents also underes- far more frequently than expected, while probes show- timates the frequencyof “dominant” polymorphisms- ing RFLPs with only one ortwo enzymes werefound far from our initial screen of grandparental genotypes, less frequently than expected (Table 2). The tendency only 27.5% of RFLPs exhibited a single unique band, for restriction enzymes with different target sites to re- while genetic mappingshowed that 36% of RFLPs seg- veal RFLPs with common DNA probes suggests that an regated as “dominant-recessive” markers.] appreciable numberof RFLPs between G. hirsutum and Deviations from Mendelian segregation ratios were G. barbadense may be due to genomic rearrangements widespread, as is commonplace in interspecific crosses. (e.g., insertions/deletions or localized duplications and Averagedacross theentire genome, segregation fit slippage at repeat units) which occur betweenthe resmc- closely with the Mendelian expectation for an F, popu- tion sites of severalWerent enzymes. GALAUet al. (1988) lation, with 25.5% of loci homozygous for the G. hir- previously suggested this, basedon similar observations. sutum allele, 22.8% homozygous for the G. barbadense Segregation: Among more than1200 DNA probes ex- allele, and 51.7% heterozygous. However, specific re- amined with four to six restriction enzymes, 563 DNA gions of the genome showed marked deviations from probes revealed RFLPs at 705 loci, including 455 loci the average. A total of 42 distinct regions, on 26 of the (average 1.25 per probe) detected by PstI-genomic se- 41 linkage groups, showed significant deviation from the quences, 187 loci (average 1.31 per probe) by cDNAs, Mendelian expectation (at the nominal 0.05 level: re- and 63 loci (average 1.17 per probe) by other cotton gions depicted in Figure 2). Heterozygote excess was genomic sequences. Among the 705 mapped RFLP found in 16 regions, while a deficiency of heterozygotes loci, 452 (64.1 %) were scored as co-dominant alleles was found in one region. Nine regions showed a defi- at single loci, 122 (17.3%) were scored for presence ciency of G. hirsutum homozygotes, including three re- or absence of the G. hirsutum fragment,and 131 gions showing an overall deficiency of the G. hirsutum (18.6%) were scored for presence or absence of the allele. Sixteen regions showed a deficiency of the G. bar- G. barbadense fragment.This high proportion of badense homozygote, including six regions showing an “dominant-recessive”markers presumably results overall deficiency of the G. barbadense allele. from polyploidy, whereby thealternate allele is There appears to be no strong bias in gross nuclear masked by monomorphic co-migratingduplicated genome composition, despite the fact that thecytoplasm DNA fragments. In many cases, this was evident from was derived from G. hirsutum. Further, no expected stoichiometric diminution of signal from the “mono- single-locus genotype was completely absent from the morphic” restriction fragment (filter zone) in the par- population. However, the knowledge that segregation ent exhibiting the unique fragment; however, quan- distortion occurs in crosses between these species titative discrimination of genotypes in thesecases was (which segregate for a number of agriculturally impor- 834 A. J. Reinisch et al.

Cbr. 4(A) L. G. DO1 L G. A05 89.75 cM 173.9 cM 117.3 cM

G 1045(A 1575)c, 7.7 4.9 PXP4-58(A) \ 7.8 19.3 &I p AR 738a.OqA) 10.6 19.4 4.0 1.6 pAR219b 1.1 0.0 M 16- 125a 0.0 1.7 A 1543.04 2.3 1.7 A 1638 6.1 4.5 PI 1-38(A) 2.7 3.3 pAR197b~ 6.7 6.8 G1033a.O4@W , 4.6 11.6 VNC58 p ' 21.9 8.3 R230 pA 7B.q 3.9 pAR049b 0.0 A 1172(A) 8.5 16.2 17.4 10.9 17.2 12.3 A7 'A 1620(D)

Cbr. @A) Cbr. 25( D) L G. DO4 Cbr. 7qA) 195.77 cM 153.90 cM 143.63 cM 214.62 cM

7.5 2.0 1.5 6.0 p VNC24 2.9 9.4 2.9 39.1 17.8 10.0 10.4 2.0 8.0 7.8 1.5 0.0 4.6 5.8 10.8 8.6 2.8 1.8 12.3 0.0 0.1 12.2 7.0 4.4 3.7 3.6 12.9 1.2 5.0 19.5 P12-13 7.1 2.9 5.6 16*8 11.0 9.0

P5-3Zb (A)

FIGURE2.-RFLP map of the cotton genome. The cotton map presently includes41 linkage groups,and spans 4675 cM, using the KOSAMBI(1944) cM function. A total of 683 loci link to this map; anadditional 22 loci (not shown) were mappedbut do not show statistically significant linkage yet. The identity, and genomic affinity of 15 linkage groups assigned to chromosomes is indicated. Nomenclature for loci, and for linkage groups not yet assigned to chromosomes,is described inMATEW AND METHODS. Genomic affinities of linkage groups represent a consensus of data from several markers (inferred as described in RESULTS, and Figure 3). Evidence for genomic affinity of individual markers (denoted as shown in Figure 3) is presented in bold. Solid lines connecting probes on different linkage groups indicate putative evidenceof homoeology, basedon duplication of three or more loci on common linkage groups. Dotted lines connecting LG DO1 and A05, and LG A02 and DO3 reflect probes which show two loci on one homoeolog, and a single locus on another, thus the orthologous loci are unknown. Dotted lines connecting Organization of Cotton Chromosomes 835

Chr. 20 (D) 270.2 cM y:z 23.6 Chr. 9(A) 186.58 cM P13-6 7.9 @) P2-5 D A R068 Chr. (D) 5.9 23 Chr. 5(A) A 1567 DAR 144g 122.3 cM 3.0 244.27 cM A 1259 G1112d n 2.9 n pAR16S I A 17371," 5.8 A 1650(A) P8-36 12.8 - 11.3 P1-336 pAR117.09 pAR257b 2.3 0.9 A 1690 M 16-45 A1270b.05( ) pVNC164a 0.0 0.9 pA R085b P12-12a 15.0 0.0 P5-6 1b pA R262 pA R328 pAR127 1.3 0.0 A 1751c pA R278a pA ROE5 pAR083 0.0 9.9 pAR053 PXP1-48 pA R279 1.9 7.8 G1386 (D ) P2-9 A 1737a A 1270a 1.3 12.9 A 1246 ~1112e A 1471b (D)P3-23 1.1 8.8 G1112a G1180 A 1744.0qD) A 1471a 3.8 3.3 xM 16-125b P6-58 (D)P3-4 PXP3-30 19.6 pAR157 2.6 PXP G 12671, G1267a 10.5 1-9a 10.7 P6- 12 G 1050 -A 1606 3.6 10.4 P1-33a PXP4-23 A1517 pA R 12am) 26.0 0.0 P6-256 pA R240 @)pARZo9 0.9 612286 7.4 P10-62.0SfAA) pARoo8 1.3 7.0 p A R022 17.9 5.3 A 1733 6.1 A 18386 G1228a A1174 0.2 M 16-02 DO2 0.1 P16-114 20.9 L G. A 11aAA) Al318a 68'78 cM 3.9 A1778- pA R099b PXP3-3 9 9.4 0.9 pA ROO3 M16-185 Ml6-150- - RA 0.9 pARl25- A 1296(D) p A R200 - 0.0 2.7 PXP3-2qA) A 17396 G1219- 3.5 1.2 A 1660 pAR137&d$)AR137a.20- 8.3 5.1 pA RO38 L A03 G 1004- 14.0 G. 1.9 M16-117 A 1378- 1.4 2.0 A1341- G1054 1.8 8.0 pA R 7 78 G 1045(A 1575)a A1701a -A1701b- 5.6 9.2 - " G1025a G1025b 2.4 7.6 - M16-118 - PXP2-4 1 - 20.2 17.5 - A 1532 P5-2 - 3.7 4.5 - PXP4-49 - 0.0 0.0 - 0.0 0.0 - 8.9 1.8 - 0.0 1.1 - 3.2 2.9 - 0.0 1.9 - 1.8 0.5 - 4.5 5.4 3.8 - 5.1 9.5 13.8 11.2

~ 14.4 5.6 - 0.0 1.8 3.4 - 4.1 2.1 30.1 0.0 14.3 - 0.0 L 4.0 1.9 3.7 10.1 21.0 /A A1214p 1700 0.0 4.3 1;j-z 9.8 0.9 1.8 3.9 A 12146 P6-57a 4.0 1.o (D)pAR138bJ P657bd(A)G1199 FIGURE2.Pontinued chromosome 4 to LG DO1 and A05 reveal putative evidence of "paleo-homoeology." The criteria for inferring this, as well as a description of three additional cases which could not be shown directly in the figure, are described in RESULTS. Regions of seg- regation distortion are indicated by filled rectangleswhich span the marker loci showing significant deviations fromthe Mendelian expectation, and end halfway between the most distal locus showing distortionand the nearest locus for which segregation does not deviate significantly (0.05 level) fromthe Mendelian expectation. Shading of the rectangles indicatesthe nature of segregation distortion, as shown in legend. 836 A. J. Reinisch et al.

Chr. 77(D) A1608 suppl. A 1194'$A 17076 ili 4'8 cM A 1482.17 Chr. 22(D) 305.3 cM (D)G1071

G1074 Chr. 77(D) A1583 121.38 cM A 1310 15.1 L G. DO5 132.1 cM 11.8 8.8 13.1 L G. A06 9.5~~ 98.2 cM 20.0 pA ROO6 26.8 n A 1252 12.5 4.1 6.2 0.0 7.0 4.3 0.9 4.3 2.7 4.6 8.9 9.4 7.9 14.3 0.0 3.5 5.6 3.6 2.0 13.6 0.0 7.8 3.6 14.5 0.0 12.1 3.1 pA R051 12.4 A 1524 n 11.4 A 1264 4.9 A 1740 0.0 0.0 A 1719bMA 171ga 3.5 2.2 P9-3a PXP2-78 1.5 9.0 1557A pA R 144h 9.1 16.8 A 1780 pA R218a 14.4 3.6G1005 1807 A 0.0 0.0 A 1685a- A 16856 0.4 ~1776 G1045(A1575)d 13.5 4.9 DAR177a- 9.3 M16-41 22.4 L. G. U03 pA R 188 7.3 50.1 cM pA R057 1.3 A 1625 21.2 Pll-72 14.8 pA R078a 10.6 A 1826a

FIGURE 2.Pontinued tant traits) suggests that subsequent QTL mapping stud- Recombination: Among the 705 loci at which segre- ies, or phenotype-based DNA pooling experiments gation has been determined, 683 (96.8%) have been to isolate diagnostic markers (MICHELMOREet al. 1991; assembled into 41 linkage groups of two or more loci, GIOVANNONIet al. 1991) should compensate for segre- with the remaining 22 not yet linked to the map(Figure gation distortion with increased population size (WANC 2). The map presently spans 4675 centimorgans (cM), and PATEMON1994). with DNA marker loci distributed at average intervals of Organization of Cotton Chromosomes 837

L.G. A02 LG. DO3 Chr. 7(A) Chr. 75(D)

205.2 cM 155.7 cM 169.3- cM p2-58$1128.6~~~ cM ~ 5.0 3.8 G1114 P10-56a-)&4.4A 1590e 12.1 -E 11.9 P1-8 (D)pAR019.15 6.4 A 1698 10.7 20.1 0.0 A 1216 8.2 - pAR099a - I I 1-38.2 4 A 1549.07 11.9 12.8 A 1706 2.9 A 1204.0 1 M 16-78 6.8 2.0 A 1731 11.9 1197 pAR297.OqA) (D)P5-39 5.8 6.4 A 1679 6.9 2.7 1658 A 1686a.OqAA) 2.7 14.8 M 16-88 8.6 4.9 1010 pA R226\(DIDY\ 16866 0.0 3.4 A 1412fAA) 6.5 16.9 A 1643 (DDID) P5-32 A 1348 0.0 6.0 A 1691c 0.1 10.1 pA ROE8 G1171 3.2 10.4 A 1 108 (D/D pAR309 4.2 3.4 61097 A 1720 8.1 10.3 4.6 0.1 VNC1646 A 1204.0 1 pA Roll 10.3 5.3 1.5 3.6 A 1553 P5-24 pA R132 7.5 4.6 1.0 5.0 A 1168 pA R07 A 1225 3.6 0.9 A 1590a -( 31.2 0.0 0.0 0.0 M 16-200 A 1667a 0.0 5.9 61013 ApAR121 1588 0.0 0.0 A 140 1 0.0 16.2 A 1109 A 15906 A 1155 14.4 0.0 pAR118 6.4 0.0 A 1257 A 1340 3.6 0.0 P 1-26 pAR118a 3.9 7.9 P5-3 1 P9-54 2.1 3.6 P5-9 A 1632 22.0 1.3 A A 1643 5.4 4.3 P5- 18 A 1562 13.0 0.0 pA R077a A 1097 5.3 16.4 61018 A1107 26.0 7.4 pA R245 12.8 P1-46a P5-377.0 P1-46a e3.1 A 17g4Aw pxp3-42 A 1658 AA) P5-4 17.9 'J 12.3 4.1 GllOl pA R248 5.6 G1021 G 1078(A) pA R 123 LG. A01 LG. U06 J 205.75 cM 19.14 cM 20.8 G1276a(D/D) 4 1214d 0.0 A 1552a -A 15526 20.6 A 1676 A 1417qD\ A 14176 1.5 A 1520A \ ' A 15208 12.1 p A R238 A 1835 P10-7 3.4 12.5 pA R265c P1-19 4.4 A 1208a 14.9 P11-16

~ 10.6 LEGEND FOR pA R274a SEGREGAllON DlSTORllON 22.2 Excess of heterozygotes PXP2-25 Deficiency of hetenzygotes 26.5 bficiency of G. hilsutum homozygotes P5-lla Chr. 78( D) =Deficiency of G. hilsutum allele 24.6 cM 15.7 Deficiency of G. babadense homozygotes 7.6 A 1713a(AA)-A 17136 bficiency of G. babadense allele 0 0.0 G11256(Ak(DD)G1125a PXP 1-69 5.5 pA R046 PXP3-25 19.4 A 1364 14.5 1258a(A) (A)A 1647.18 A 1428a 17.7 - pA R282

FIGURE2.4ontinued

7.1 cM along the chromosomes. Estimates of recom- 26 chromosomes of cotton are presently representedby bination in cotton based upon chiasmata counts indi- 41 linkage groups, and that we can detect linkage at up cate a minimum overall map length of 4660 cM (STELLY to 30cM, filling thegaps between existing linkage 1993), suggesting that most regions of the genome are groups should add at least450 cM (=30 cM X 15 gaps) already covered by the present map, although gaps in to the map, for a minimum overall length of 5125 cM. some linkage groups remain to be filled. Given that the Once the map reaches a densityof about one marker per 838 A. J. Reinisch et al.

L. G. UO1 L. G. A04 L. G. U02 Chr: 2(A) 200.34 cM 120.22 cM 63.82 cM 51.9 cM

A 1826b A 1719c G 1148 10.6 23.7 34.7 - 1.7 A 1325 p BA M325b 6.8 P2-35c pAR151.OqD) 3.3 G I185b P9-6 1 4.3 0.0 A I 146.02 7.0 pA R027 P2-356 2.1 29.2 pA R318.04AA) 18.1 A 16914A) - 14.0 pA R316 pA R257a 12.4 pA R260c A 1436 23.5 L G. Dl1 L G. DO9 7.05 cM 33.7 -1 r pAR278b 3.3 cM

L G. U04 L. G. A07 L G. A08 28.4 cM 27.65 cM 21.77 cM

0.1 M 16-85 M16-45c

L G. Dl0 L. G. U05 L G. U07 L G. UO9 6.86 cM 26.1 0 cM 16.21 cM 6.27 cM

A 1619a 6.8 PI-246 P12-1qD)

FIGURE2.-Continued

5 cM, or a total of about 1025 markers, there should be somes 5, 14, 15, 18 and 20 is suggested by single loci, fewer than 1% of intervals between markers measuring which are neither corroborated nor contradictedby any >25 cM (TANKSLEYet al. 1988), and the map should“link other locus on the linkage group. These determinations up” into 26 linkage groups corresponding to the 26 were based upon single plants derived from isoline de- gametic chromosomes of cotton. velopment projects for each monosome and telosome. Assignment of linkage groups to chromosomes: A Abnormal meioses occasionally generate new forms of subset of the mapped DNA probes were hybridized to aneuploidy from existing aneuploids, thus independent genomic digests of a series of monosomic and monote- corroboration of these results with newly isolated stocks is lodisomic substitution stocks, with a single G. bu&xknse underway. Further, additional aneuploid stocks are being chromosome substituted forone pair of G. himturn chre characterized to identtfy chromosomesand chromosome mosomes. This method is conceptually similar to use of arms corresponding to additional linkage groups. chromosomedeficient lines to determine chromosomal Three pairs of (tentatively) identified chromosomes location of DNA probes (HELENTJARISet al. 1986; GALAU showed homoeology in our map, chromosomes 1 and et al. 1988), except that assignment of probes to chro- 15,chromosomes 5 and 20, and chromosomes 6and 25. mosomes is based on detection of an RFLP, rather than Homoeology between chromosomes 1 and 15 is sup- on dosage. Based on multiple genetically linked loci ported by the observation that duplicated mutations in which correspond to common aneuploid substitution three morphological traits map to this pair of chromo- stocks, we havetentatively determinedthe linkage somes (ENDRIZZIet al. 1984),with one inversion in order. groups which correspond to chromosomes 1, 2, 4, 6, 9, Homoeology between chromosomes 6 and 25 has been 10, 1 7, 22 and 25 (Figure 2). The identity of chromo- suggested previously (ENDRIZZIand RAMSAY 19’79),based Organization of Cotton Chromosomes 839 upon similar phenotypes of plants monosomic for each of these chromosomes. Although several mutant phenotypes have been assigned to chromosome 5, no evidence is available regarding chromosome 20. Classi- cal evidence for homoeology between chromosomes 7 and 16, and chromosomes 12 and 26 (ENDRIZZIet al. 1984), cannot yet be corroborated by molecular evi- dence as the linkage groups corresponding to these FIGURE3.-Inference of genomic affinity for individual chromosomes have not yet been determined. DNA marker loci. Genomic restriction fragments which re- Several tentatively identified chromosomes show ho- vealed RFLPs between G. hirsutum and G. barbadense were moeology to as yet unassigned linkage groups. For ex- evaluated for genomic affinity by comparison to correspond- ample, chromosome9 appears homoeologous toa link- ing genomic digests from G. raimondii and G. trilobum (D genome) andG. arboreum and G. herbaceum (A genome). Six age group(LG) initially named D06. Based on basic scenarios were found. (1) A and D genomes have dif- independent evidence of homoeology between chromo- ferent restriction fragments, one polymorphic restriction frag- somes 9 and 23 (CRANEet al. 1994),we have tentatively ment in AD genome co-migrates with one (or both, in this changed the name of LG DO6 to chromosome 23. Un- case) fragments in a diploid genome (D, in this case). In the assigned linkage groups showing partial or complete ho- case shown, the locus would be denoted“DD,” because it co- migrated with fragments unique to Dthe genome but present moeology to chromosomes 10, 14, 17 and 22 have been in both D genome types. This was deemed to represent two identified; however, the identitiesof the homoeologs to “units”of information in supportof a D genome affinity of the these chromosomes remain unknown. chromosomal region near this locus.(2) A case similar to # 1 Deducing the genomic origin of linkage groups in wouldbe denoted “D,” forco-migration with a fragment allotetraploid cotton: Some DNA probesdetected unique tojust oneof the two D genome types studied. This was deemed to represent only “unit”one of information in support genomic fragments in tetraploid cottons which were of D genome affinity. (3) If each of two allelic polymorphic shared with either A or D genome ancestors, but not restriction fragments in ADthe genome cottons correspondto both. In a subset of these cases, polymorphism between a polymorphic restriction fragment in one of the two D ge- G. hirsutum and G. barbadense permitted mapping of nome types, the locus is denoted “D/D,” and two “units”of one or bothof the homoeologous genomic fragments. information in supportof D genome finitywere inferred. (4) This case exemplifies the risk of artifacts due to convergent Such “alloalleles” (e.g., orthologousdiploidgenome- gain/loss of restriction sites. The available information war- specific DNA fragments) have been suggested to be a rants the same designation as case #2 (‘ID”). The most par- means of deducing the diploidorigin of particular link- simonious explanation of this scenario is that oneD genome age groups (or chromosome segments) in tetraploid ancestor and oneof the tetraploid subgenomes have indepen- cotton (GALAUet al. 1988). dently incurred mutations resultingin new co-migrating frag- ments. However,there is no a priori way of determiningwhich The degreeof support for inferencesof subgenomic of the AD subgenomes was involved in the mutation. An al- origin of alloalleles varied among RFLP loci and linkage ternative explanation would be polyphyletic evolution of the groups (logic summarized in Figure 3). Among 276 two tetraploid species, involving eachof the two diploid spe- mapped loci for which the A, D, and AD genomes could cies. No information is inferred. (5) In this case, one AD ge- be compared, 115 yielded information suggesting the nome fragment corresponds to a fragment unique to one A genome type. AnotherAD genome fragment corresponds to genomic origin of particular restriction fragments. a fragment unique to a D genome type. Consequently, there These 115 loci yielded a total of 159 comparisons (in is conflicting information of equal weight. Such cases were some cases, a locus yielded more than onepiece of evi- considered ambiguous, and no informationis inferred. (6) In dence of genomic origin; see Figure 3). Among these, this case, the polymorphic band(s) is/are unique to the tet- 138 (86.8%) fit with the most parsimonious genomic raploid, and no information is inferred. origin of their respective linkage group. Based on al- of alloallelic data fora chromosome agreedwith classical loallelic loci, the genomic origin of 33 of the present41 subgenomic assignment of chromosomes in all 14 linkage groups, including all of the known chromo- (100%) cases. Individual pieces of information from al- somes, has been determined. One linkage group (LG loallelic probes on known chromosomes were correct in UO1) showed conflicting information (2 A, 2 D), and 63 (86.3%) of 73 cases. We conclude that alloallelic in- seven small linkage groups (totaling 224.8 cM and 34 formation is a reliable, although notinfallible, indicator loci) harbored no alloallelic loci. of genomic origin of a chromosome or linkage group. To evaluate the reliability of our subgenomic infer- For cotton, agreement among three independentpieces ences, we calculated the frequency at which alloallelic of alloallelic information affords ca. 99% confidence information coincided with the classical assignment of (1 - [ 1 - 0.8631’) of correct subgenomic assignment of cotton chromosomes to subgenomes based upon pair- a chromosomal region. ing relationships in diploid X tetraploid hybrids (A = Six of the 10 loci where alloallelic information dis- chromosomes 1-13 D = chromosomes 14-26). Based agreed with classicalsubgenomic assignment of chromo- on the 14 chromosomes we have identified, the majority somes occurred as linked pairs on threelinkage groups. 840 A. J. Reinisch et al.

In each of these three cases (chromosome 14, LG A03, 6 and 141.9 cM (92.2%) of chromosome 25. The du- LG UOl), thetwo deviant loci are consecutive along the plicated loci are homosequential. Three additional DNA linkage group (with referenceto alloallelic probes). probes mappingto these chromosomes show duplicated This suggests the possibility that exchanges of chromatin loci which fall on different linkage groups. Alloallelic between homoeologous chromosomes mayhave oc- evidence indicates that chromosome 6 is A genome- curred during theevolution of tetraploid cottons. How- derived, while chromosome 25 is D genomederived. ever, by the criteria we suggest in the preceding para- Linkagegroups DO1 and A05r These are deemedho- graph, three pieces of informationare necessary to moeologous based upon seven duplicated loci, span- make this inference, thus more data is needed to sub ning 152.7cM (87.8%) ofLGD01 and108.7 cM (92.3%) stantiate the possibility of homoeologous exchanges. of LG A05. At least two inversions are necessary to ac- Determination of homoeology from map positions of count fordifferences in locus order. Four additional du- duplicated lock A total of 106 (18.8%)DNA probes de- plicated DNA probes mapping to these chromosomes tected RFLPs at two loci, 13 (2.3%) probes detected show duplicated loci. Two of these are on different link- RFLPs at threeloci, 2 (0.36%) at four loci and 1 (0.18%) age groups, but the othertwo, PAR1 97 (on LG Dol) and at six loci. We preferentially mapped probe X enzyme G1045 (on LG A05), each show a duplicated locus on combinations which showedmultiple polymorphic frag- chromosome 4 (an A genome chromosome), possibly ments; hence, these data overestimate the frequency at indicating “paleo-homoeology” (see below). which DNA clones detect multiple RFLPs. Chromosome 10 (Agenome) and LG 004: These are In many cases, loci linked to one another had dupli- deemed homoeologous based upon 4 DNA probes, cated sites also linkedto one another. This could occur by spanning 75.2 cM (35.0%) of chromosome 10 and 66 at least three mechanisms: (1) chance associations between cM (45.9%) of LG D04. A single inversion can account loci generated by retrotransposon-like or other duplica- for the difference in order between loci. Eleven other tionmechanisms, (2) recent polyploidy and (3) pa- duplicated loci are found on these two groups. Two of leopolyploidy. If duplicated loci are randomly distributed these (A118?a, b on chromosome 10) represent a and linkage groups are of equal size, synteny between n proximal duplication, while three show duplicated sites duplicated DNA markers in a map of 41 linkage groups widely distributed throughoutthe genome. Several would occur with a likelihood of (1/41)(”-’). [We recog- additional loci show evidence consistent with nize that each of these assumptions probablyare violated paleopolyploidy (see below). to some degree, and that this calculationrepresents only Chromosome 9 (A genome)and chromosome 23: a crude estimate of the likelihood of the underlying event.] These are deemed homoeologous based upon 5 DNA Thus, synteny of twoduplicated loci by chance is somewhat probes, spanning 76.3 cM (40.9%) of chromosome 9 improbable ( ca. 2.4%), butsynteny betweenthree or more and 78.1 cM (63.9%) of chromosome 2?. A single ter- loci by chance is highlyimprobable (ut.0.06%), andclearly minal inversion can accountfor the difference in order suggests that duplication of a chromosome (or segment), between markers on these linkage groups. Six additional rather than independent local duplications of short se- probes show unlinked duplications. We note that chro- quences, is the most likely explanation. We have consid- mosome 23 has been identified based on prior evidence ered synteny of three or more duplicated loci as evidence of homoeology to chromosome 9 (CRANEet al. 1994), as of homoeology. no aneuploid genetic stocks were available. The inver- By evaluating the distribution of linked duplicated sion we detected is consistent with the physical linkage loci acrossthe map,we find evidence for duplication of of 5s and 18-28s rDNA sites in the same arm of chro- chromosomes or segments of23 of our 41 linkage mosome 23, and opposite arms of chromosome 9 (18- groups, covering 1668 cM (36% of the genome). These 28s on short arm; CRANEet al. 1994). homoeologous relationships are based upon association Linkage groups A03 and 002: These are deemed ho- among 62 (51%) of the 122 duplicated probes we moeologous basedupon 4 DNA probes, spanning 26 cM mapped. In all cases, anyone region of a linkage group (15.4%) of LG DO2 and 54 cM (31.8%) of LG A03. The was associated with only one region of another linkage duplicated loci map in the same order on the two chro- group, when associations based upon three or more mosomes. Nine loci falling outsidethe region of homoe- duplicated loci are considered. ology map to unlinked sites, severalon very small linkage Homoeologous relationships between linkage groups groups. Only one otherduplicated locus occurs inthe re- are described below (and shown in Figure 2). Inall cases gion of homoeology, withfive sites on other chromosomes. where structural mutations were inferred, homosequen- Chromosome 5 (A genome)/chromosome 20 (Dgenome) tial order was rejected at LOD scores of 2.0 or greater, + LG 007: These are deemed homoeologous based unless otherwise specified. upon 13 DNA probes (9 on chromosome 20,4 on LG Chromosome 6 (A genome) and chromosome 25 (0pome): D07). One region of 122.7 cM (50.2%) of chromosome These are deemed homoeologous based on six dupli- 5 corresponds toa region of 127.6 CM (48.5%) of chro- cated loci, spanning 150.4 cM (76.8%) of chromosome mosome 20, while another region of 31.4 cM (12.8%) OrganizationChromosomes of Cotton 841 of chromosome 5 corresponds to19.3 cM (28.6%) of LG (22.4%) on LG U03, a linkage group of unknown sub D07. A single terminal inversion can explain the differ- genomic origin. The loci occur in homosequential order ence in marker orders between chromosome 5 and on the two linkage groups. The regions of homoeologyto chromosome 20, while another terminal inversion dis the different linkage groups do not overlap. tinguishes chromosome 5 and LG D07. This association Homoeology of chromosome 22 to two other linkage of chromosome 5, an A genome chromosome, toparts groups may be the result of a translocation-or, the two of two different D genome linkage groups could indicate other groups (DO5 + U03) could be different parts of that the relevant A and D genome chromosomesdiffer the same chromosome. New markerswill be necessary to by a translocation, or that the two D genome linkage distinguish between these hypotheses. groups areparts of the same chromosome. We found no Finally, we note thatthis is the only instance where two evidence of linkage between chromosome 20 and LG linkage groups inferred to be of the same subgenome D07; however, LGDO7 is relatively small(67.4 cM) to be (D) show evidence of homoeology. However, the sub- an entire chromosome. genomic identity of LG DO5 is based on sharing of only Linkage groupsA02 and003: These are deemedho- a single restriction fragment between the D and AD ge- moeologous based upon three DNA probes, spanning nomes (cf. Figure 3, case 2), an observation which could 32 cM (16%) of LG A02 and 26 cM (17%) of LG D03. easily have explanations other than homoeology. At least one inversion is necessary to explain the differ- Linkage groupA06 andchromosome I4 (Dgenome) + ences in order among the duplicatedloci. The location chromosome 17 (D genome): In the lower region of LG of the breakpoint is unclear, as one probe (A1590)de- A06, three loci spanning 55.8 cM (56.8%) correspond to tects two loci on LG DO3 at similar hybridization inten- 26 cM (17.4%) of chromosome 14. The difference in sity, and it is unclear which ofthese is orthologous to the order of loci on the two groups can be explained by a locus on LG A02. Seven other loci on the two chromo- single inversion. The upper region of L.G. A06 shows somes map to unlinked sites. association to chromosome 17, based upon only two Chromosomes 1 and 15: These are deemed homoe loci. Although this is not sufficient (by our criteria de- ologous basedupon two duplicated DNA probes, spanning scribed above) to claim homoeology, we note that the 45 cM (26%) of chromosome 1 and 58 cM (45%) of chro- corresponding regions of LG A06 and chromosome 1 7 mosome 15. As described above,we have not inferred ho- do not show homoeology to any other genomic regions. moeology inother cases unlessthree duplicated loci were The apparenthomoeology of LGA06 to two different D found- however, this case supported is by classicalevidence genome chromosomes strongly suggestsa translocation that mutations in three morphological traitsmap to these or Robertsonian fusion distinguishing the A and D ge- two chromosomes (ENDRIZZIet al. 1984),with one inversion nome chromosomes. Two other markers on chromo- in order. The duplicated loci we have mapped do notde- some 17 detect unlinked duplicated loci. tect the inversion, but span less than half of each chro- Possible evidence of segmental paleo-homoeology: mosome. Three otherduplicated loci on the two chromo- In at least five cases, we found “nested”associations, be- somes map to unlinked sites. tween (segments of) different homoeologous pairs of Linkage groups A01 andU06: These are deemed ho- linkage groups. While any one such association of two moeologousbased on 3 markers, spanning 20.6 cM duplicated loci might well be seen by chance in a ge- (10.0%) of LG A01 and 15.6 cM (81.5%)of LG U06. The nome with as manychromosomes and duplicated loci as duplicated loci are homosequential.At the opposite end of cotton (see likelihood calculation, above), the occur- LG AO1, over 120 cM away, an association of two markers rence offive such events suggestspaleopolyploidy. is found between LG A01 and chromosome 18. Four other These associations are described below, andone duplicated loci on LG A01 map to unlinked sites. One, example is illustrated (Figure 2; chromosome 4 with pAR274b, has not yet linkedto the map, and the (putatively LG DO1 and A05): homoeologous) pAR274a locus corresponds to the center Chromosome 4//LGDOI-LGAOS: Probes GI 045and of the gap between LG U06 and chromosome 18. PAR197 map to loci 55.9 cM apart on chromosome 4, Chromosome 22 (D genome) and LG DO5 + LG U03: and duplicated sites map to loci on LG DO1 and A05, In the upperregion of chromosome 22,s loci spanning respectively, which are an estimated 40 cM apart (esti- 93.5 cM (30.6%) correspondto a region of 86 cM matedfrom distances separating nearby duplicated (66.7%) on LG D05. The homoeologous regions are markers). largely homosequential, with the most likely order of Chromosome IO-LG D04//LG A03-LG 002: Probes markers suggesting a small terminal inversion, although AI 759and P6-57 map to loci near the upper ends of an alternate markerorder with no inversion is only LOD chromosome IO-LG D04, and duplicated sites map 1.59 less likely(thus cannot be ruled out at the standardabout 40 cM apart in a region homosequential between LOD 2.0 threshold). LG A03 and LG D02. In the lower regionof chromosome 22, three loci span- Chromosome IO-LG D04//chromosome 5-chromosome ning 41.8 cM (13.7%) correspond to a region of 45.1 cM 20: Probes AI 751 and P5-61 map to loci 1.3 cM apart 842 A. J. Reinisch et al. on chromosome 5, in a region which is homoeologous tivity of other sequenceduplication mechanisms in cot- to chromosome 20. Duplicated loci map to homeolo- ton, andfurther support the need for determining gous regions of chromosome 10 and LG D04, separated homoeology based on several locialong a linkage group by about 36 cM (estimated from distances separating (as we have done; see above). nearby duplicated markers). Most of the homoeologous chromosomes in n = 26 Chromosome 5-chromosome 20//LG UO1: Probes cottons aredistinguished by one ormore inversions. In A1 432and PPI-9 map to loci 51.7 cM apart on LG only two cases are putative homoeologs homosequential UO1, and are estimated to be more than 100 cM apart over the entire region of the chromosome for which on chromosome khromsome 20 (estimated from dis- homoeology can be inferred, although large blocks of tances separating nearby duplicated markers). sequence conservation are evident in most cases. Four Chromosome 20 + LG D07//chromosome 4: Probes pairs clearly differ by at least one inversion, and two PAR138 and PAR21 9 map about21 cM apart on chro- additional pairs differ by at least two inversions. At least mosome 4, and are unlinked on chromosome 20 and LG one pair shows a translocation, while several additional D07. Chromosome 20and LG DO7 each show homoeology pairs may represent eithertranslocations or simply small to Merent parts of chromosome 5, and it is unclear linkage groups which have not yet linked up. whether they are part of the same linkagegroup or not. Detection of rearrangements among thehomoeologs Several other “nested”associations weredetected, but of cotton is not surprising, in light of the 6-11 million all involved one DNA probe that mapped to six loci, years of divergence of n = 13 cottons from a common suggesting that these are due tomechanisms other than ancestor (WENDEL1989). The observation that aneu- polyploidy. ploid D X AD genome hybrid cottons form >12 bivalents per cell suggests that few if any translocations differen- tiate D genome diploid and tetraploid chromosomes DISCUSSION (KIMBER 1961). However, A X AD hybrids showevidence Molecular mapping of disomic polyploids can becon- of at least two translocations distinguishing n = 13 A ducted with similar efficacy to molecular mapping of genome chromosomes from their n = 26 A genome diploids, although polyploidy necessitates a more rigor- counterparts (GERSTEL1953), suggesting that the A and ous examination of allelismbetween polymorphic DNA D subgenomes also differ by translocations, which is con- fragments (see above). Further, polyploidy tends to in- sistent with our observations. A companion study (C.L. crease the frequency of polymorphic bands masked by BRUBAKER,A. H. PATERSONand J. F. WENDEL,unpublished monomorphic co-migratingduplicated fragments, data) will describe mapping of a subset of the DNA precluding qualitative identification of heterozygotes probes used here inF, populations of G. arboreum X G. at such loci. Fortuitously, however, one occasionally herbaceum (Agenome) and G. raimondii X G. trilobum finds polymorphism at each of two homoeologous (D genome). This should permitdiscrimination of struc- loci, affording unique opportunities tostudy chromo- tural mutations that occurred duringdivergence of the some evolution, and to economize in genetic map- n = 13A and D genomes, from mutations that occurred ping. after formation of the n = 26 AD genome. Divergence of homoeologous chromosomesin n = 26 Despite structural rearrangement and reliable biva- cotton: Mapping of multiple RFLPs derived from indi- lent pairing in tetraploid cotton (KIMBER 1961),low copy vidual cotton DNA probes reveals the consequences of DNA sequences on the homoeologous chromosomes of genome-wide duplication in “tetraploid” (n= 2x = 26) both A and D diploid (n = 13) genomes, and the cor- cotton. We tacitly assume that most such associations responding AD tetraploid (n= 26) genome retain con- between linkage groups are a result of the New World siderable similarity. Low copy DNA probes derived from A X D hybridization event responsible forthe either theA or D genomes areof similar effectiveness at occurrence of “tetraploid” (n = 2x = 26) cottons. detecting homologous and heterologous copies at mod- Based on criteria described above, we have tentatively erate stringency (0.5 X SSC, 65”). A high frequency of identified (at least parts of) 11 of the 13 expected ho- “dominant/recessive” RFLPs suggests that many homoe- moeologous pairs in n = 26 cottons based upon 62 du- ologous DNA sequences retain flanking 6-base restric- plicated DNAprobes, spanning 1668 cM or 35.6% ofthe tion sites from common ancestors of A and D genome genome. Among the 265 mapped duplicated loci, 124 cottons-thus one polymorphic restriction fragment is (46.8%)are accounted for by homoeology, and 16 masked by monomorphic co-migrating duplicated (6.0%) by “paleo-homoeology.”Additional loci may be fragments. accounted for by these mechanisms as the remaining Extent of structural rearrangementof cotton chromo- small linkage groups coalesce. The occurrence of proxi- somes relative to chromosomesof other taxa: The na- mal duplications, and duplicated loci inconsistent with ture and extent of chromosomal rearrangements dis- homoeologous relationships, together with indepen- tinguishing taxa provides one means by which the dent evidence (VANDERWIELet al. 1993) reflect the ac- relative “genetic distance” between disparate taxa can be Organization of Cotton Chromosomes 843 assessed. For example, the two homoeologous chromo- netic observations and the relatively high gametic chro- some sets (A and D) of n = 26 Gossypium are much mosome number (DAVE 1933; SKOVSTED 1933;ABRAHAM more extensively rearranged than chromosomes of the 1940), and more recently (SUITER1988; WENDELand genera Lycopersicon and Solanum, which differ by only PERCIVAL1990) stemming from the observation that five inversions and no translocations (TANKSLEYet al. gene numbers for allozymeencodingloci exceed expec- 1992). Cotton homoeologs also exhibit more extensive tations for “true” diploids (cf. GOTTLIEB 1982). In our rearrangementthan the chromosomes of Brassica results, hybridization patterns of low copy genomic se- oleracea (n = 9)and Brassicacampestris (n = lo), quences reveal an appreciable level of DNA sequence which differ primarily by structural rearrangements ac- duplication even in present-day “diploid” (n= 13) cot- counting for the difference in chromosome number tons, although we note thatour results show a somewhat ( SLOCUM1989). Extension of this comparison to other higher level of DNA sequence duplication than prior Gossypium genomes (B, C, E, F, G) awaits future results. results (GALAUet al. 1988; discussedabove). Further,in In contrast, homoeologous cotton chromosomes are at least four cases, associations of two pairs of linked more similar than the chromosomes of Zea and Oryza duplicated loci were observed within individual subge- (AHN and TANKSLEY1993) which are distinguished by nomes of tetraploid Gossypium, possiblyreflecting con- several translocations. Finally, homoeologous cotton served linkage blocks from an earlier polyploidization. chromosomes are far more similar than the chromo- At least one homoeologous pair (chromosome IO-LG somes of Lycopersicon and Capsicum (TANKSLEYet al. D04) shows the possibilityof two different putative 1988) or Arabidopsis and Brassica (KOWALSKIet al. “paleo-homoeologs” in only a small part of its length. 1994), which are extensively rearranged and exhibit These different putative regions of paleo-homoeology only small islands of conserved order. are non-overlapping. The fact that we could only docu- The extentof rearrangement between homoeologous ment such associations with two pairs of in loci each region cotton chromosomes most closely parallels that found suggests that n = 13 cottons may have undergone much between Zea mays and Sorghum bicolor (HULBERTet al. structural rearrangement and/or sequence evolution sub- 1990; WHITKUSet al. 1992), where long homosequential sequent to the presumed paleopolyploidization event. stretches of chromatin are occasionally interrupted by In suggesting the possibility that n = 13 cottons are inversions, but few if any translocations are evident. paleopolyploid, we need to account for the observation Evidence repding paleopolyploidy of n = 13 cot- thatmore than 30%of the DNA probesexamined tons: It is apparent that themajority of duplicated loci hybridized to only one (size of) genomic restriction mapped in n = 26 cotton became duplicated as a con- fragment in the A and D genomes (Table 1; Figure 1). sequence of the relatively recent hybridization event Some of these apparent “single” fragments may repre- (1.1-1.9 million years ago) that led to the evolution of sent two co-migrating duplicated sequences, which origi- polyploid Gossypium. This is supported by the identifi- nated via polyploidization, and retain the restriction cation of homoeologous linkage groups covering 35.6% sites present in the putative n = 6-7 ancestor. Alter- of the genome, as shown above. In addition to recent nately, duplicated sequences may have diverged to the polyploidy, however, evidence implicates other mecha- extent that genomic probes no longer detect “paleo- nisms in the generation of duplicated sequences. For homoeologs.” Weview this latter alternative as more example, the observation that DNA probes hybridize to likely, i.e.,that the n = 6-7 to n = 13 transition is an- a larger number of genomic fragments in the n = 13 A cient, primarily because paleopolyploidization must genome than in the n = 13 D genome cannot be ex- have antedated the evolution of the genus Gossypium plained by recent polyploidy (Table 1). Similarly, the (which is estimated to be at least 25 million years old; fact that individual DNA probes revealed up to six RFLP WENDELand ALBERT 1992), andindeed, must have loci demonstrates sequence duplication that is not ex- antedated the entire tribe Gossypieae and the closely plicable by recent polyploidy alone. Additional dupli- related tribe Hibisceae, where all genera have high ga- cations may have occurred by a variety of mechanisms, metic chromosome numbers (FRWELL 1979).Accordingly, including replicative transposition (VANDERWIELet al. it would not be surprising if many genomic probes canno 1993), chromosome segment duplication, and, of par- longer detect paleehomoeologous DNA sequences. ticular relevance here, more ancient polyploidization Relationship between physical size and recombina- events involving progenitors of present-day “diploid” tional length of subgenomes: Identification of linkage (n= 13) Gossypium species. groups corresponding to the A and D subgenomes of New evidence from genomic hybridization patterns n = 26 cotton (Figure 2) permits comparison of prop and linkage relationships among duplicated RFLP loci erties of the two subgenomes. Despite considerable dif- supports the hypothesis that “diploid” (n= 13) cottons ferences in DNA content (GEEVERet al. 1989), and chro- may be paleopolyploids, derived from anevent or events mosome size (KIMBER 1961), theA and D subgenomes of involving ancestors with fewer chromosomes. This hy- n = 26 cottons exhibit virtually identical recombina- pothesis was suggested decades ago, based on cytoge- tional length (2119 and 2140 cM, respectively;Table 3). 844 A. J. Reinisch et al.

This suggests that the muchof the DNA comprising the diversity to be equivocal with respect to the issue of (physicallylarger) A genome may be recombinationally polyploid monophyly us. polyphyly. The magnitude of inactive. Much of the difference in DNA content be- observed differences in RFLP diversity between subge- tween the A and D diploid genomes is attributable to nomes is small, the RFLPs remaining unassigned to a moderately and highly repetitive elements (GEEVERet al. subgenome are of sufficient number to eliminate this 1989; X. ZHAO, R. WINGand A. H. PATERSON,manuscript difference, we have little evidence regarding therelative submitted for publication), suggesting that the repeti- activity in different subgenomes of processes which have tive fraction may be recombinationally inert. This con- given rise to new RFLP diversity since polyploid forma- cept is supported in a general sense by the observation tion, and evidence from alloallelic polymorphisms may that recombinational length of the genomevaries much be biased by convergence between D-genome and less than DNA content, across a wide range of plant and tetraploid restriction fragments. animal species. However, in no prior case have recom- Application of DNA markers to cotton improvement: bination rates in two genomes differing inDNA content The cotton RFLP map is a starting point in use of DNA been comparedso directly, e.g.,within the gametes pro- markers to manipulate genetic determinants of agricul- duced by a common F, plant. turally important traits, and characterize the basis of ag- Subgenomic distribution of RFLPs, and its implica- ricultural productivity of cotton. Numerous introgres- tions for theorigin of polyploid cotton:A long-standing sion events may be detectable using existing DNA question in Gossypium evolution is whether poly- markers, including transfer of numerous genomic re- ploidization occurred only once (“monophyletic ori- gions from G. hirsuturn into cultivated G. barbadense, as gin”)or two or more times (“polyphyletic origin”). well asintrogression of specifictraits such as Verticillium Whilemost evidence supports a single hybridization wilt resistance (STATEN1971), bacterial blight resistance event leading to the evolution of tetraploid Gossypium (STATEN1971), nectariless leaves (TYLER1908), restora- species (ENDRIZZIet al. 1984), polyphyly has been sug- tion of cytoplasmic male-sterity (MEYER 1975; WEAVER gested based on cytogenetic information (DAVIE 1933, and WEAVER1977), and improved fiber quality (CULP 1935; KA”ACHER 1959,1960),flavonoid profiles (PARKS and HARRELL1974: CULPet al. 1979). As in many largely et al. 1975), and seed protein electrophoretic patterns self-pollinated crops, the gene pools of each of the culti- (JOHNSON 1975). Chloroplast DNA restriction site data vated cotton species show onlymodest levels of DNA poly- clearly demonstrate that only a single parental A ge- morphism. Routine applicationof DNA markers tocotton nome was involved in polyploid formation (WENDEL breeding may benefit fromtechnologies such as 1989), but few data critically address whether this ma- microsatellite-based markers, which are being superim- ternal A genome parent hybridized with one, or more posed on the existing RFLP map (ZHAOet aL 1994). than one, D genome species. Prospects for high density mapping and mapbased Herein, we note thatapproximately 10% more RFLPs cloning in cotton and other polyploids: Several poten- were detected in the D subgenome than the A subge- tial complications associated withmapbased cloning in nome (329 us. 295; Table 3). If only a single hybridiza- disomic polyploids are partly compensated for by some tion event led to polyploid formation, then both sub- unique advantages of polyploid genomes. Utilization of genomes would be equivalently aged and might be genetic information from homoeologous relationships expected to reveal similar levels of RFLP diversity (as- will accelerate development of a high density map of suming equal rates of sequence divergence). contrast,In DNA markers in polyploid genomes with many chro- if different D genome species were involved in the par- mosomes. About 21.7% of DNA probes segregated for entage of G. barbadense and G. hirsutum, one might RFLPs at two or more loci, accounting for 265 (37.5%) expect greater D subgenome than A subgenome RFLP of the mappedloci. These duplicated loci provide a skel- diversity in modern tetraploids. eton for inferring theapproximate homoeologous locus Evidence from alloallelic polymorphisms could, in of DNA probes monomorphic in one subgenome but principle, support or contradict a polyphyletic origin of polymorphic in the other,increasing the markerdensity n = 26 cotton. Specifically, scenario #3 in Figure 3 of themap. As additional markers are mapped itshould could be consistent with a polyphyletic origin of n = 26 be possible to catalog homoeologous relationships over cotton. Nine D subgenome loci fit scenario #3, while most of the genome, and identify breakpoints of rear- only three A subgenome loci do so. However, an un- rangementswith higherresolution. Once theremaining usually high rate of polymorphism between the D ge- homoeologous relationships are identified, and the den- nome parentsraises the possibility that a larger portion of sityof duplicated loci becomes greater, formal algo- the D genome matches are due to convergence, rather rithms for genetic map integration( cf. BEAVI~and GRANT than conservation. 1991; STAM1993) can be employed to infer “unified While one might be tempted to speculate that our maps” of particular homoeologous chromosomal re- data supporta polyphyletic origin of n = 26 Gossypium, gions, greatly increasing the density of DNA markers in we view the apparent differences in subgenome RFLP maps of disomic polyploids such as cotton. Ultimately, OrganizationChromosomes of Cotton 845

8 .G .- I 2 n n nn 00000000 W tl -8 I -33 el 22 e, $ c, 3 3 mm 00000000 W2 a gJ3 3 e ** mM 3 h 24 e I co .^ $ K W W mQ, CI g mmm*mm*m* * zg z 23 el z ii gg 3 3 33 E 00 2wmWWmWW? B 2 3 mm owmm3- m m2 g: m n * B 5 .o nEl e a VI 2 Od 0 0 zk h 00 mooooooa m 2 2 .-U e W g h - v1 i;O 00 3 *oooooo* * s I &I 0 0 3 -0I & M 1 1 9 82 2 e 0 a W n 3m*mr-o3m E 0000000 v1 a - 8 .C kg Y) s s s 3333333 tl 29 8 fg el a$ sj s M s SssSSss 3 - .-E CII$ 8 mQ,*wm*nam n*r.- 8 z O2 Ei 'g mmn 000 0 w m 3 2 a5 $4 g 1 "3 e c 0 rg .-v) g 8 m & *g 5 '49 P-" m - 5 a?' 8 000 0 '3 g E, gGzqgclgul2 mmzzWwm- %f%Z r-x 8"$ z 33 z 3 : CQ 8 y mz 0 P- 4- .- g v B WW c mr-mOWaWQIm 3mnm nna 00 Q, .9 g g m*3a3mm- mnaw e %2% 2 m *g g E3 el m c, @ n P v E Ag 8 M 2g ,+gr-z+z,+zm v?zQ, ci ??-"I c mmr- 2 r-r-n 0 Pn 4- r- 3 2 a onm 3 n z ZGWzZhnzam hnnsz g n3* 3 3 K m 44 i% nE 2 E E 21 1 s v) Q,z-mm**mm Vmr-W z" gi r-im .C 3" n W %2 1 m W c m '2 -$ $? 3 e$? 0$ 2E El- '5 2, 24 Q,=-Wm*"mn *mP-m r-em "3 co $8 1 1 w n m .^E r- -@ 1 .CU *$ g f - 0 Of M a ~Or-v\R\*~.p\coma3 8 cym a03 s 1cyocyrv0111 ooo cy0 c, 3 g .E niin . . . nnn g CI 2 ZZE 3 -2 0 220Sf0&&& 000 c $0 % a.5; e VVelU~el"U" elelel M Vcl h 2% Ef 1 co '3 a @$ 3 T .g s. P - am*wm*nmmnnw~* 0 a 00000 0 s& &2 10 5 & *le s2 B$ - s. g 8 8 5 3 82 "i E A 3 0 .- 00000 0 0 3 g 5 L&lggggC3ggm3m*mo3 mnm2zP-3 * 2 z 5 el L- 4 A F .- d & cl E omr-- 0 L-*mr-m m i g$ mmmm 2 8ZZS cd .-. G '3 s".e 04 8 U& 8 a 2: 2 .8 ab &3 22 C??Zm 2 @!?Zb 0 2ommm 2g 4 5 2s 32;Q, g gssz mamma ; 2 3 3 3 n 2.3 m 882z

$?& m= mnr-w w WWW3 0 $ 3wa3m 3 W 2 r- g.." g 2,s I &I $8 p 2 3% 1 qm2 mmr-w m mmw-P- 0 3wn3a 2&na :.9u 2 - 0 .i 71 9: OWW33mnm~ * r--00 ~~~~~-400000003 ov.cyo0 z 08 i i & i i isoybean, wheat, oat, morphic loci be mappedgenetically, however this might canola, tobacco, peanut and many others). be accomplished either by screening more restriction A detailed map of DNA markers, in conjunction with enzymes for DNA polymorphisms, by using special tech- well developed cytogenetic tools, offers the opportunity niques such as denaturing gradientgels, by sequencing to utilizethe polymorphic genus Gosy@um and its diverse alternate alleles to identify base substitutions, or by relatives such as Hibiscus (kenaf, roselle) and Aklnwschus other techniques. (okra), to investigate plant chromosome evolution in ex- The physical size ofa centimorgan in cotton is not pro- quisite detail,and to have a major impacton improvement hibitive to mapbased cloning, but the lengthy genetic map of one of the world’s oldest and most important crops. will require a large number of markers inorder to be suf- ficiently close to most genes for “chromosome walking.” We appreciate a cDNA library from N. TROLINDER,genomic probes With a genetic map length of ca 5000 cM (Figure 2), and from I. KWON,J. PRICE, T. MCKNIGHT,V. RASTOGI, and X. Zmo, technical physical size of 2246 Mb (ARUMUNGANATHANand EARLE support from L. CHITTENDEN,D. CZESCHIN,J. GARZA, D.RASKA, D. 1991), the average physical size ofa centiMorgan in cot- SARMIENTO,R. VAYLAY, and C.VASQUEZ-ROSADO, and helpful com- ton is about 400 kb, only slightlylarger than thatof Ara- ments from D. ALTMAN,A. H. D. BROWN,C. CRANE,R EL-ZIK,J. GANNAWAY,R. KOHEL, T. MCKNIGHT,E. PERCIVAL,J. PRICE, D. RAY, W. bidopsis (ca. 290 kb), and smaller than that of tomato SMITH,P. THAXTON,N. TROLINDER,T. WIWNS, R. WING,and three (ca. 600 kb), both species in which map-based gene anonymous reviewers. Aspects of this work were supported by the cloning has been accomplished (GIRAUDATet al. 1992; U.S. Department of Agriculture (91-37300-6570 (to A.H.P.), ARONDEL et al. 1992; MARTIN et al. 1993). However, even with and 92-37300-7655 (to D.M.S.), Texas Agricultural Experiment Sta- tion (to A.H.P. and D.M.S.), Texas Higher Education Coordinating the advantage affordedby homoeologousinformation, the Board (999902-148 to A.H.P.), and National Science Foundation genetic map of 5000 cM will require ca 3000 DNA markers (8918041 to J.F.W.). Subsets of the mapped DNA probes are avail- to map at average 1 cM density, and the physical genome able for academic research from A.H.P. by written request. of 2246 Mb will require ca 75,000 YACs/BACs of average size 150 kb for 5X coverage. The suggestion above that repetitive DNA is recombinationally inert will further im- LITERATURE CITED prove prospectsfor mapbased cloning in cotton, as a siz- ABRAHAM,P., 1940 Cytological studies in Gossypium. I. Chromosome able fraction of the genome may contribute little or noth- behavior in the interspecific hybrid G. arboreum X G. stocksii. ing to the 5000 cM genetic length. Finally,we note that the Indian J. Agr~c.Sci. 10 285-298. h,S., and S. D.TANKSLEY, 1993 Comparative linkage maps of the estimate of physicalgenome size we haveemployed (2246 maize and rice genomes.Proc. Natl. Acad. Sci. USA90: 7980-7984. Mb: ARUMUNGANATHAN and EARLE 1991) falls at the median ARONDEL,V., B. LEMIEUX,I. HWANG, S. GIBSON,H. M.GOODMAN et al., of several estimates et al. 1989; et al. 1992 Mapbased cloning of a gene controlling omega-3 fatty (GEEVER MICHAEL~~N acid desaturation in Arabidopsis. Science 258: 1353-1354. 1991; GOMEZ et al. 1993), which ”uy by *30%, introducing ARUMUNGANATHAN,K, AND E. D. EARLE,1991 Nuclear DNA content of a corresponding range into our figures. some important plant species. Plant Mol. Biol.Reptr. 9 208-218. BALLS,W. L, 1906 Studies in Egyptian cotton, pp. 29-89 in Yearbook Mapbased cloning in polyploids such as cotton in- Khediv. Ap’c. Soc. for 1906. Cairo, Egypt. troduces a new technical challenge not encountered in BEASLEY,J. O., 1942 Meiotic chromosome behavior in species hybrids, diploid (or highly diploidized) organisms, e.g., that vir- haploids, and induced polyploids of Gossrplzlm Genetics n: 25-54. BEAVIS,W. D., and D.GRANT, 1991 A linkage map based on infor- tually all “single-copy”DNA probes occurat two or more mation from four F2 populations of maize (Zea mays L.). Theor. unlinked loci. This makes it difficult to assign Y mega- Appl. Genet. 82: 636-644. base DVA clones to their site of origin. One possible BRUBAKER,C. L., and J. F. WENDEL,1993 Molecular evidence bearing on the specific status of Gossypium lanceolatum Todaro. Genet. approadh to this problem is the utilization of diploids in Res. Crop Evol. 40 165-170. physical mapping and map-based cloning; for example, BRUBAKER,C. L., J. A. KOONTZand J. F. WENDEL,1993 Bidirectional exploiting the relatively smallgenome of D genome dip cytoplasmic and nuclear introgression in the New World cottons, Gossypiumbarbadense and G. hirsutum. Am. J. Bot. 80 loid cottons (approximately 65% of the size of the A 222-227. genome, and 39% of the tetraploid; GEEVERet al. 1989). CHIITENDEN,L. M., K F. SCHERTZ,Y.-R. LIN, R. A.WING and A. H. However, since D genome cottons produce minimal fi- PATERSON,1993 A detailed RFLP map of Sorghum bicolor X S. propinquum suitable for highdensity mapping suggests ancestral ber, and are notcultivated, they are of limited utility for duplication of chromosomes or chromosomal segments. Theor. dissecting the fundamentalbasis of cotton productivity. Appl. Genet. 87: 925-933. The relatively larger size ofthe A genome (GEEVERet al. CRANE,C. F., H. J. Pma, D. M. SWY, D. G. CZESCHm, JR. and T. D. Mcx- NIGHT, 1994 Identificationof a homoeologous chromosome pairby 1989), despite low copy sequence complexity similar to in situ hybridization to ribosomalRNA loci in meioticchromosomes that of the D genome (Table 1; also see GEEVERet al. of cotton ( Cossypiurn hinuturn L.) . Genome (in press). 1989), suggests that the repetitive fraction may provide Cuw, T. W., and D. C. mu,1974 Breeding quality cotton at the Pee Dee Experiment Station, Florence S. C. USDAF’ubl.ARS5-30. a means of characterizing the genomic identity of indi- U.S. Government Printing Office, Washington, D.C. vidual YACs from tetraploid cottons (X. ZHAO, R. WING Cum, T. W., D. C. HARRELL and T. KEm, 1979 Some genetic impli- cations in the transfer of high-fiber strength genes to upland cot- and A. H. PATERSON, manuscript submitted for publica- ton. Crop Sci. 19: 481-484. tion). Such an approachmay prove generally applicable DAVIE,J. H., 1933 Cytological studies in the Malvaceae and certain to mapbased cloning in other major crops, manyof related families. J. Genet. 28: 33-67. OrganizationChromosomes of Cotton 847

DAVIE,J. H., 1935 Chromosome studies in the Malvaceae and certain MICHELMORE,R W., I. PARANand R V. KESSELI, 1991 Identification of related families. 11. Genetica 17: 487-498. markers linked to disease resistance genes by bulked segregant ENDRIZZI,J. E., and G. RAMSAY, 1979 Monosomes and telosomes for 18 analysis: a rapid method to detect markers in specific genomic of the 26 chromosomes of Gossyjnum himtum Can. J. Genet. Cytol. regions using segregating populations. Proc. Natl. Acad. Sci.USA 21: 531-536. 88 9828-9832. ENDRIZZI,J. E., E. L. TURCO-ITEand R. J. KOHEL,1984 Qualitative ge- PARKS,C. R., W. L. EZELL,D. E. WIWS and D. L. DREER, 1975 VII. netics, cytology, and cytogenetics, pp. 81-129 in Cotton, edited The application of flavonoiddistribution to taxonomic problems byR. J. KOHELand C.F. LEWIS. ASA/CSSA/SSSA Publishers, in the genus Gossypium. Bull. Torrey Bot. Club 102 350-361. Madison, Wisc. PATERSON, A. H.,and R. A. WING, 1993 Genome mapping in plants. FRYXELL,P. A., 1979 The Natural History of the Cotton Tribe. Texas Curr. Opin. Biotechnol. 4: 142-147. A&M University Press, College Station, Tex. PATERSON,A. H., S. D. TmKsLEYand M. E. Somu,1991 DNA mark- FRYXELL, P. 1992 A., Arevised taxonomic interpretation of Gossypium ers in plant improvement. Adv. Agron. 46: 39-90. L. (Malvaceae). Rheedea 2 108-165. PATERSON,A. H., C. BRUBAKERand J. F. WENDEL,1993 A rapid method Gmu, G. A., H. W.BASS and D.W. HUGHES,1988 Restriction frag- for extraction of cotton (Gossypium spp.) genomic DNA suitable ment length polymorphisms in diploid and allotetraploid for RFLP or PCR analysis. Plant Mol. Biol. Rptr. 11: 122-127. Gossypium: assigning the lateembryogenesisabundant (Lea) al- PERCY,R. G., andJ. F. WENDEL,1990 Allozyme evidence for the origin loalleles in G. hirsutum. Mol. Gen. Genet. 211: 305-314. and diversificationof Gossypiumbarbadense L. Theor. Appl. GEEVER,R F., F. KAmw and J. E. ENDRIZZI,1989 DNA hybridiza- Genet. 79 529-542. tion analyses ofa Gossypium allotetraploid and two closely related SAMBROOK,J., E. F. FRITSCHand T. MANmns, 1989 Molecular Cloning: diploid species. Theor. Appl. Genet. 77: 553-559. A Laboratory Manual, Ed. 2, pp. E.37-E.38. Cold Spring Harbor GERSTEL,D. U., 1953 Chromosomal translocations in interspecific Laboratory, Cold Spring Harbor, N.Y. hybrids of the genus Gossypium. Evolution 7: 234-244. SAS Institute, 1989 SAS/STAT User’sGuide, Version 6, Ed. 4. SAS GIOVANNONI,J. J., R.A. WING,M. W. GANALand S. D.TANKSLEY, Institute, Inc., Cary, N.C. 1991 Isolation of molecular markers from specific chromo- SKOVSTED,A.,1933 Cytologicalstudiesin cotton.I. The mitosisandmeie somal intervals using DNA pools from existing mapping popu- sis in diploid and mploid Asiatic cotton. Ann. Bot. 47: 227-258. lations. Nucleic Acids Res. 19 6553-6558. SKOVSTED,A., 1937 Cytological studies in cotton. lV. Chromosome GIRAUDAT,J., B. M. HAUGE,C. VALON,J. SMALLE,F. Pmcx et al., conjugation in interspecific hybrids. J. Genet. 34: 97-134. 1992 Isolation of the ArabidopsisABI3 gene by positional clon- SLOCUM,M. IC, 1989 Analyzing the genomic structure of Brassica ing. Plant Cell 4: 1251-1261. species and subspecies using RFLP analysis, pp. 73-80 in Devel- GOMEZ, M.,J. S.JOHNSTON, J. R. ELLISONand H. J. PRICE,1993 Nuclear opment and Application of MolecularMarkers to Problems in 2c DNA content of Gossypium hirsutumL. accessions determined Plant Genetics,edited by T. HELENTJARISand B. BURR.Cold Spring by flow cytometry. Biol. Zentrbl. 112: 351-357. Harbor Laboratory, Cold Spring Harbor, N.Y. GOTI-LIEB,L. D., 1982 Conservation and duplication of isozymes in STAM,P., 1993 Construction of integrated geneticlinkage maps by plants. Science 216: 373-380. means of a new computer package: JOINMAP.Plant J. 3: 739-744. HELENTJARIS,T, D. F. WEBERand S. WRIGHT,1988 Use of monosornics STATEN,G., 1971 Breeding Acala 1517 cottons, 1926-1970. New to map cloned DNA fragments in maize. Proc. Natl. Acad. Sci. Mexico State Univ. College of Agric. and Home Econ. Memoir USA 83 6035-6039. Series No. 4. HULBERT,S. H., T. E. RICHTER,J. D. AXTELL and J. L. BENNETZEN, STELLY,D. M., 1993 Interfacing cytogenetics with the cotton genome 1990 Genetic mapping and characterization of sorghum and mapping effort. Proc. Beltwide Cotton Impvt.Conf., pp. related crops by means of maize DNA probes. Proc. Natl. Acad. 1545-1550. Sci. USA 87: 4251-4255. SUITER,K. A., 1988 Genetics ofallozyme variation in Gossypium JOHNSON,B. L., 1975 Gossypium palmeri and a polyphyletic origin of arboreum L. and Gossypium herbaceum L. (MALVACEAE).Theor. the New World cottons. Bull. Torrey Bot. Club 102: 340-349. Appl. Genet. 75: 259-271. KAMMACHER,P., 1959 Relations cytologiques entre l’espece sauvage Gos- TANKSLEY,S. D., J. C. MILLER,A. H. PATERSONand R. BERNATZKY,1988 syjn’um raidii Ulb. et lesespeces cultivees de cotton- Molecular mapping of plant chromosomes. Proc. 18th Stadler niers G. himtum L. er G. badma!ense L. Cotton Fib. Trop. 13 1-22. Genetics Symposium, pp. 157-173. Wcmm, P., 1960 Observations cytologiques sur deux hybrides F, TANKSLEY,S. D., M. W. CANAL, P.J. PRINCE,M. C. DE VICENTE,M. W. entre especes cultivees tetraploidsde cottoniers et I’espece diploid BONIERBALEet al., 1992 Highdensity molecular linkage maps of sauvage Gosypum raidiiUlb. Rev. Cytol. Bil. Veg. 22 1-32. the tomato and potato genomes. Genetics 139: 1141-1160. KIMBER,G., 1961 Basisof the diploid-like meiotic behavior of TYLER,F. J., 1908 The nectaries of cotton, Part V, pp. 45-54. U.S. polyploid cotton. Nature 191: 98-99. Dept. Agr. Plant Ind. Bull. No. 131. KOSAMBI,D. D., 1944 The estimation of map distances from recom- VANDERWIEL,S., P. D. F. VOWAS and J. F. WENDEL,1993 Copia-like bination values. Ann. Eugen. 12 172-175. retrotransposable element evolution in diploid and polyploid cot- KOWALSKI,S. P., T.N. LAN, K. A. FELDMANNand A. H. PATERSON, ton (Gossypium L.) J. Mol. Evol. 36: 429-447. 1994 Comparative mapping of Arabidopsis thaliana and Bras- WANG,G., and A. H. PATERSON,1994 Prospects for using DNA pooling sica oleracea chromosomes reveals islands of conserved organi- strategies to tag QTLs withDNA markers. Theor. Appl. Genet. (in zation. Genetics 138: 499-510. press). LANDER,E. S., P. GREEN,J. ABRAHAMSON,A. BARLOW, M. J. DALYet al., WEAVER,D. B. and J. B. WEAVER.,1977 . Inheritance of pollen fertility 1987 MAPMAKER an interactive computer package for con- restoration in cytoplasmic male-sterile upland cotton. Crop Sci. structing primary genetic linkage maps ofexperimental and natu- 17: 497-499. ral populations. Genomics 1: 174-181. WENDEL,J. F., 1989 New World cottons contain Old World cytoplasm. MARTIN, G.B., S. H. BROMMONSCHENKEL,J. CHUNWONGSE, A. FRARY, M. W. Proc. Nat. Acad. Sci. USA 86 4132-4136. GANALet al., 1993 Mapbased cloning of a protein kinase gene WENDEL,J. F., and V. A. ALBERT,1992 Phylogenetics of the Cotton conferring disease resistance in tomato. Science 262 genus ( Gossypium): Character-stateweighted parsimony analysis 1432-1436. of chloroplast-DNA restriction site data and its systematic and MEREDITH,W.R., and R R.BRIDGE, 1984 Genetic contributions to biogeographic implications. Syst. Bot. 17: 115-143. yield changes in upland cotton, pp. 75-86 in Genetic Contribu- WENDEL,J. F., and A. E. ~WAL1990 Molecular divergence in the Ga- tions to Yield Gains of Five Major Crop Plants, edited by W. R. lapagos Island-Baja California species pair, Gossypiurn klotschianum FEHR.Crop Science Society of America, Madison, Wisc. Anderss., and G. davidsaii Kell. Plant Syst Evol. 171: 99-115. MEYER,V. 1975 Male sterility from Gossypium harknessii. J. Hered. WHmUs, R.,J. DOEBLEY,M. LEE, 1992 Comparative genome mapping 66 23-27. of sorghum and maize. Genetics 132: 1119-1130. MICHAELSON,M. J., H. J. PRICE,J. R. ELLISONand J. S. JOHNSTON, ZHAO, X, Y LIN,AH PATERSON,1994 Genetic mapping of DNA mic- 1991 Comparison of plant DNA contents determined by Feul- rosatellites from cotton. Plant Genome I1 Proceedings, p. 74. gen microspectrophotometry and laser flow cytometry. Am. J. Bot. 78: 183-188. Communicating editor: A. H. D. BROWN