The World of Irises pp. 375-415 (1978)

Chapter 26
Iris Genetics
Kenneth K. Kidd

I. INTRODUCTION

Genetics is the science of heredity; the science of how traits are passed from generation to generation. As such, genetics is concerned with the nature of the genetic material, its transmission from cell to cell and from parent to offspring, and the way in which the genetic material controls the development of traits during the life of an individual. The classical tool of genetics is variation, and in irises there is so much variation that understanding the inheritance of even the simplest trait can be complicated. This plethora of variation derives from the complex background of modern garden irises in which diverse characteristics from many species have been brought together by generations of hybridizing.

In many ways the statement of Sturtevant and Randolph (1945) still holds: "The improvement of the garden iris has progressed very rapidly in recent years but our knowledge of the genetics of the various characters concerned with the diversity of types which have been developed has not increased proportionately." We have much new data, but we are also aware of more possible complications that tend to make simple answers suspect. In what follows most of our present knowledge of iris genetics is summarized. None of the statements or conclusions should be accepted as gospel truths; almost all of them are based on data that are insufficient for strong scientific conclusions. This chapter must be considered an introduction to the basic principles and a collection of working hypotheses that interpret available data in the simplest manner. Only by the formulation of hypotheses and testing them through the careful observations of seedlings from testcrosses will our knowledge of iris genetics advance. While the knowledge we do have can benefit the hybridizer, the best advice for any practical breeder who is uninterested in genetics is a qualified paraphrase: "Like tends to beget like."

II. CYTOGENETICS

Iris enthusiasts generally recognize the need for information about the numerical and morphological relationships of the chromosomes in different species. Different numbers and/or types of chromosomes are characteristic of the basic species from which the important groups of garden varieties originated. The chromosomes are stable units of the nucleus of each cell and bear the genes, the plant's hereditary material. Chromosomes are the major factors determining sterility in hybrids between different species and sometimes even in hybrids of closely related varieties. Any breeding program that involves crossing plants of different species or different garden types should consider the chromosomal constitutions of the plants. Many fruitless crosses and much frustration can be avoided by knowing the results to he expected from crossing plants with different chromosomal characteristics.

In an ordinary diploid plant the chromosomes in every cell (except the pollen and ovules) occur in pairs. A haploid set consists of one member of each pair. Most of the chromosomes in the haploid set are visibly different when observed during cell division. They differ in length, they occur with or without a satellite (a small piece attached to the end by a thinner thread), and have different ratios of arm length. The arm lengths are defined by the position of the kinetochore or centromere—synonomous names for the chromosome region (1) that divides last, (2) that orients the chromosomes at division, and (3) by which chromosomes are pulled apart and moved to the separate ends of a dividing cell. These aspects of chromosome form can he seen in idiograms and photomicrographs of chromosomes. The entire complement of chromosomes in a cell of an individual, as depicted in figure 1, is referred to as the karyotype. Virtually all members of a species have the same karyotype, which is often a unique and distinguishing characteristic of the species.

In addition to the visible differences among chromosomes, genetic studies show that each chromosome of the haploid set has a different set of genes (defined by function) and thus carries a unique array of genetic information. The morphological and genetic individuality of each chromosome is maintained from generation to generation.

THE CHROMOSOME CYCLE

The distribution of chromosomes to the reproductive cells during the reduction division and their rejoining during fertilization determine the manner in which the characters of the plant are inherited. Figure 1 shows essential processes of the chromosome cycle, in which there are two crucial stages: alternately producing cells (gametes) with only half the number of chromosomes found in ordinary plant tissue and then fusing two gametes to restore the original number in the zygote, and start a new plant. It is an endless cycle with no specific beginning; arbitrarily we shall start with the production of a new plant.

Figure 1. Diagram of the cell cycle starting with a zygote. Only one pair of genetic factors is illustrated in the diagram. During the development of the adult, each factor of every pair is copied identically at each cell division so that all cells have both members of every pair. In the production of gametes, however, only one factor of each pair is included in any one meiotic product. These meiotic products then develop into the pollen and ovule, which in turn produce the true gametic nuclei. Fusion of the sperm nucleus and the egg nucleus produces the zygote of the next generation and restores the genetic factors to pairs.

The first stage is fertilization, the uniting of the male and female germ cells (gametes) following pollination. In diploids, each gamete contributes one "set" of chromosomes to the fertilized egg. The two sets are "added" together, and thereafter, during the growth of the new individual, every cell throughout its tissues has two sets of chromosomes, called the 2X or somatic number. Preceding every cell division in the life of an individual each chromosome is precisely replicated and the copies separate during the division. The process of chromosomal replication and separation, known as mitosis, assures that each time a cell divides the two daughter cells are identical to each other and to the parent cell in number and kind of chromosomes. Mitosis occurs at every cell division throughout the growth of the plant, except in certain cells in the developing flower bud.

1 For further clarification of terms not fully defined, see Glossary.

Early in the development of the flower buds, when they are yet deep in the iris fans, the germ cells are being formed in the stamens and ovary through a different kind of chromosome division known as reduction division, or meiosis (figure 2). The chromosomes replicate themselves exactly (except at the kinetochore) and the homologs then pair in a zipperlike way. The paired homologous chromosome can be referred to as a bivalent, emphasizing involvement of two chromosomes, or as a tetrad, emphasizing the existence of four chromatids.1 At this tetrad stage crossing-over (recombination) can occur. Next, there are two nuclear divisions with no intervening chromosome replication. At the first meiotic division, the paired, but as yet undivided, homologous kinetochores are separated with the "arms" attached (the arms are now a combination of material from both original chromosomes because of crossing-over). The second division is like a mitotic division, but with half as many kinetochores; the kinetochores behave independently, each splitting, producing the final haploid nuclei and completing the reduction process. At this stage the number of chromosomes has been halved, reduced back down to the gametic number.

The haploid cells produced by the meiotic divisions give rise to the gametophytes that produce the true gametes by subsequent mitoses. In the anther each of the four (nonidentical) haploid nuclei gives rise to one pollen grain. Subsequent mitotic divisions produce the two identical male sperm nuclei that are the actual gametes. In the ovary one meiosis occurs in each ovule or potential seed. In most flowering plants studied three of the four products of that meiosis disintegrate and only one survives. It subsequently undergoes three cycles of mitosis to produce eight identical haploid nuclei. One of the eight becomes the egg cell, the actual female gamete; two of the eight, called polar nuclei, become the female contribution to the endosperm; the remaining five eventually degenerate. Development of the ovule is assumed to he similar in irises, but there are no known studies specifically of irises.

Figure 2. Diagram of meiosis showing the production of four haploid cells from one diploid cell. This diagram illustrates the transmission of two chromosome pairs and three genetic loci, the L, N, and K loci. The original cell is heterozygous for all three loci. Following duplication and pairing of chromosomes, each chromosome is represented by two chromatids. The chromatids can undergo exchanges, actual physical exchanges of blocks of chromosomal material. The first meiotic division separates the unduplicated kinetochores each with two chromatids still attached. As illustrated, the attached chromatids may no longer be completely identical genetically, but they are still homologous. At the second meiotic division the kinetochores divide without further replication of the chromosomes. The four meiotic products illustrated are only one possible outcome of a meiotic division for the particular diploid cell. In other meioses the small dark chromosome could have segregated at the first division with the predominantly dark large chromosome to give different combinations.

Fertilization occurs after the growing pollen tube penetrates the ovule and releases the two sperm nuclei (see also "Pollination," chap. 22). One sperm nucleus unites with the egg nucleus. Each carries a single set of chromosomes with a complete complement of genes; their union produces the diploid fertilized egg (the zygote) which becomes the embryo of the seed by repeated cell division and growth. The other sperm nucleus unites with the two polar nuclei to form a triploid cell which develops into the endosperm of the seed. Other parts of the seed are formed by purely maternal tissue of the ovule. Only the embryo, with equal genetic contributions from both parents, develops into the mature plant. Thus the chromosome sets of the germ cells transmit to the next generation their unique heritage of genes (figure 2).

RECOMBINATION

Little is understood about what causes chromosomes to pair in meiosis, but the pairing can he observed with a microscope; it is specific, not only visually but genetically. The meiotic pairing of chromosomes, termed synapsis, can even be exact down to the subunits of the genes. Crossing-over occurs while the homologs are paired and can most easily be conceived as occurring h' breakage of two chromatids at exactly corresponding positions and subsequent rejoining of the broken ends to form complete chromatids, each composed of parts of both original chromatids (see figure 2). Genetic information originally linked together in one chromosome inherited from one parent can thereby he separated and transmitted in different gametes. The frequency of recombination between two different genetic sites or loci on the chromosome is a function of the actual physical distance between them. Two genes close together will usually he inherited together, while two genes at opposite ends of a long chromosome will he inherited independently.

KARYOTYPE EVOLUTION

An individual chromosome maintains its basic structure through successive mitoses and remains virtually unchanged even in meiosis. The structure of a chromosome is determined by the structure of its constituent genes. Two different forms of the same gene may differ in as little as one subunit of the one to two thousand subunits that make up that gene, but the functions of those two chemically almost identical forms may he vastly different. Recombination at meiosis cannot significantly change the physical nature of the chromosome since the order of the genetic loci does not change. However, just as genes can undergo rare mutations, chromosomes can change by moving groups of loci into new arrangements. Initially such changes occur in one cell of one individual. If the change happens to he included in a germ cell and has some advantage to the offspring then—given many generations—this new arrangement may come to be the normal order in all individuals of the species

As each successive change is incorporated into the species, the karyotype changes. Since species evolve on a time scale during which several changes could have been incorporated, the difference in the karyotypes of two related species is a crude measure of how long the two species have been separated. Thus, contemporary species with similar karyotypes are more likely to be closely related than species with quite unlike karyotypes. Also, the chromosomes that appear similar in two closely related species will be more likely to contain the same loci in much the same order. This is important since the accurate meiotic pairing of the chromosomes depends upon the two homologs having an almost identical set of loci, in an almost identical linear order.

CHROMOSOME HOMOLOGIES AND FERTILITY

A hybrid between two species with quite different karyotypes has one complete haploid set of chromosomes from each parent. It therefore has two representatives (alleles) of every gene (locus) necessary for the plant to live and grow. The loci, however, are not arranged in the same way in the two different genomes (the haploid complements of genes in a diploid species; the smallest complete set of loci). In ordinary cell division (mitosis) this presents no problem since each individual chromosome functions independently; the plant can live and grow and may even be very vigorous. But in meiosis the chromosomes must function in pairs and such pairs do not exist in the hybrid. In the extreme case, each chromosome from one species finds that its set of loci are scattered among several of the chromosomes of the other species. Pairing with any of these chromosomes is impossible. At the first division of meiosis normally paired kinetochores separate, but in this hybrid with no paired kinetochores the process breaks down and no functional gametes are produced. Such is the usual cause of hybrid sterility.

Diploid interspecies hybrids in irises show virtually the entire range from essentially complete fertility to complete sterility. Intermediate levels of fertility occur when most chromosomes have enough homology that they pair at least sometimes. In spite of the differences in the karyotypes of various species with the same chromosome number, it appears that the most important consideration for the hybridizer is chromosome number. Two diploid species, or even two named varieties, will be likely to yield at least partially fertile hybrids if they have the same chromosome number and are in the same subsection of the genus. However, exceptions occur and cannot easily be predicted in advance. Other genetic causes of sterility, such as specific genes for sterility, are independent of chromosome behavior, and offspring have been obtained from hybrids between species with different chromosome numbers. Some plants from wide crosses are unexpectedly fertile. Moreover, although a plant that is nearly sterile only rarely sets and yields viable seeds, such less fertile hybrids can be important in the development of new varieties. Unfortunately, progress along such lines is generally very slow.

III. INHERITANCE IN DIPLOIDS

MENDEL'S FIRST LAW

Gregor Mendel laid the foundation of diploid genetics in 1865. Over a hundred years later the laws and concepts that he described are still valid. Mendel's first law is that the hereditary characteristics observed in an organism are determined by pairs of particles (factors). When the gametes are formed the factors segregate so that only one of each pair is included in each gamete. Half of the gametes have one factor of the pair, half of the gametes have the other factor. When the male and female gametes fuse to form a zygote, the pair of factors is restored. The factors, the hereditary units or genes, can exist in different forms associated with the different states of the hereditary characteristic, but a given factor does not change when passed from one generation to the next. The observed characteristics of the organisms in different generations may show considerable variation because these characteristics are produced anew in each individual and are determined by the particular combination of hereditary factors present in the first cell of that individual.

RELATIONSHIPS OF GENOTYPE TO PHENOTYPE

Mendel also drew the distinction between the genotype (number and kind of hereditary units present) and the phenotype (appearance of the organism), and described several possible relationships between the two. To make the examples more relevant to the subject of this book, Mendel's observations will he translated from peas to diploid irises. He observed that in plants that bred true for a trait such as self color or plicata pattern, both factors of the pair were identical (homozygous), either two "self" factors or two "plicata" factors. But in the cross between these two types, the hybrids (F1 or first filial generation) showed only one character, self color, even though one factor of each type was present (heterozygous). Mendel called the characteristic expressed in the F1 the dominant characteristic; the one not expressed he called the recessive characteristic. By extension and as a convenient shorthand the factors have been commonly referred to as dominant and recessive respectively, e.g., "a dominant gene." This terminology has never been technically correct and it is frequently contusing. Experiments with many different species have shown that dominance does not always occur and is only a relative phenomenon, not an intrinsic property of the factors themselves. For example, the plicata pattern in irises is recessive to self color but dominant to one type of white. Thus, dominance is a statement about the relationship between a particular heterozygous genotype and the phenotype it demonstrates.

When one hereditary factor controls two or more phenotypic characteristics, the situation is even more complex: one may be dominant and the other recessive. A certain yellow coat color in mice is dominant to wild-type color, and the heterozygote is yellow. However, the homozygote for that same hereditary factor dies as an embryo; lethality is recessive to viability. One hereditary factor here determines both a dominant trait, yellow coat color, and a recessive trait, lethality. It is obvious that this factor could not be called either dominant or recessive.

The example just given also illustrates pleiotropy, the general phenomenon of one hereditary factor affecting more than one phenotypic characteristic. Pleiotropy is common and undoubtedly occurs in irises. However, no examples are known as yet.

LOCI AND ALLELES

The hereditary units are actually parts of the chromosomes and exist in a specific linear order that is unique for each chromosome pair. In addition to the word gene, two other terms are useful in referring to these hereditary units. Locus, based on the position on a particular chromosome, refers to the factor pair and the general function or characteristic affected; allele refers to the particular information about that function to be found at that locus on one chromosome. In a population of several individuals, more than two different alleles (forms) of a given locus may exist, as, for example, at the plicata locus. Only two alleles (representatives) of any locus occur in the cells of any diploid individual—one on each chromosome of the pair—but those two may be different (forms) from the two alleles (representatives) occurring in another individual. The plicata locus contains information about anthocyanin distribution in the flower; at least four different forms of that information may exist; Pl is the allele which says, "uniform distribution"; pl is the allele which says, "plicata pattern"; pllu says, "luminata pattern"; and pla says, "no anthocyanin produced." Thus, the four alleles at the plicata locus can be paired to form 10 different types of individuals: Pl/Pl, Pl/pl, pllu/pla, etc. Many loci, even in the very hybrid modern irises, probably have only one kind of allele and all plants will be identical and homozygous at such loci. However, allelic variation is so common in plants and animals that even in a species with little or no apparent variation, no two clones will be genetically identical or homozygous at all loci.

PREDICTING PROGENY FOR MONOHYBRID CROSSES

To predict the outcome of any cross three steps are necessary. First, determine the types and frequencies of gametes of each parent. Second, multiply these gamete arrays together to produce the zygote array. Third, add together the genotypes with identical phenotypes to produce a phenotype array. For a cross in diploids involving only one locus—a monohybrid cross-this procedure is relatively simple. The first step is easy since only two types of diploids exist: homozygotes produce only one type of gamete; heterozygotes produce two types in equal frequencies of 50 percent each. At the second step there can be three types of matings—both plants homozygous, one homozygous and one heterozygous, and both heterozygous. Each of these will be considered as a separate case.

In the first case all offspring will be identical, all homozygous if both parents have the same allele or all heterozygous if the parents have different alleles. In the second case, one plant produces two types of gametes in equal numbers and the other plant produces only one type. The two types of zygotes have equal probabilities, a 1:1 ratio of the two genotypes, heterozygous like the heterozygous parent and homozygous like the homozygous parent. If both parents had the same phenotype—the dominant trait—all seedlings will have that trait. If the homozygous parent was the recessive phenotype, there will be a 1:I ratio of phenotypes corresponding to the two genotypes. This is a backcross, the cross of a heterozygote to either homozygous parental type. If the cross involves the homozygote for the recessive trait it is called a testcross, because it can be used to detect heterozygotes.

The third case is the most complex of the three because both parents are producing two kinds of gametes. The three steps are illustrated in figure 3. The zygotes occur with probabilities of 1/4 for the homozygote for the allele for the dominant trait, 1/2 for the heterozvgote, and 1/4 for the homozygote for the allele for the recessive trait. The phenotypic probabilities for the seedlings are 3/4 for the dominant trait, 1/4 for the recessive trait; this gives the familiar 3:1 Mendelian ratio. Progeny from two heterozygous parents are called the F2 (second filial) generation because they represent the intercross of F1 plants.

MENDEL'S SECOND LAW

If two loci, say those for plicata and purple leaf base, are segregating (i.e. both are heterozygous) in a plant and these two loci are located on nonhomologous (of different types; not constituting a pair) chromosomes, then in the formation of gametes the inclusion of a given allele at the plicata locus will not affect which allele at the leaf base locus is also included in that gamete, i.e., segregation for the two loci is independent (random). This principle of independent assortment is Mendel's second law.

Figure 3. Diagram of segregation of plicata pattern vs full coloration in diploid irises. The two homozygous plants represent the parental (P1) generation. All of the offspring from these two plants will be identical heterozygotes at this locus, these are the F1 or first filial generation. Self-pollinating or intercrossing two of these heterozygous F1 plants produces the F2 or second filial generation.

Among the seedlings of a cross segregating for two loci—a dihybrid cross—each trait alone will show one of the monohybrid Mendelian ratios just discussed. The ratios for the two traits jointly will reflect the independent assortment.

PREDICTING PROGENY FOR DIHYBRID CROSSES

The procedure for predicting the outcome of such a dihybrid cross is the same as for a monohybrid cross, but four types of gametes are possible when the plant is heterozygous at two loci. The principle of independent assortment explains the equal frequencies of the types of gametes. For a plant heterozygous for both plicata and purple leaf bases, Pl/pl; Pb/pb, the four types of gametes occur with equal probabilities: 1/4 Pl/Pb, 1/4 Pl pb, 1/4 pl Pb, and 1/4 pl pb. Figure 4 shows how the Punnett Square is used to determine the genotypes of seedlings from two such plants, their probabilities, and the expected phenotype frequencies, a ratio of 9:3:3:1 of the four phenotypes.

If one trait gives a 1:1 testcross ratio and the other gives a 3:1 ratio, the joint testcross ratio will be 3:3:1:1. If the two loci segregating both affect the same characteristic and are complementary, then a 9:7 ratio can occur. For example, the diploid Daystar is homozygous for y1, a recessive absence of yellow and the diploid I. reginae is homozygous for y2, a recessive absence of yellow controlled by a different locus, and also heterozygous for y1. A cross of the two gives 50 percent yellow, 50 percent nonyellow. The yellow seedlings will all be Y1y1; Y2y2 and the cross between any two of them should give 9/16 Y1-Y2-, 3/16 Y1-y2y2, 3/16 y1y1Y2- and 1/16 y1y1y2y2. Since all of the last three types lack the yellow pigment, the phenotype ratio for the cross is 9 yellows to 7 with little or no yellow pigment.

LINKAGE

Independent segregation is not a general law, however, since the alleles at loci close together on the same chromosome are more likely to segregate together than independently. This is the phenomenon of genetic linkage. Linkage is usually not absolute and two alleles originally on the same homolog will be found separated in some of the gametes. This results from recombination between homologous chromosomes during meiosis as discussed earlier. While all types will occur among the seedlings from two double heterozygotes, they will not be in a 9:3:3:1 ratio. The deviation from that ratio will vary with different pairs of loci depending on how close together they are on the chromosome.

Linkage must exist in irises since there are only a limited number of chromosomes, but the association of two characteristics, e.g., yellow and pink as discussed in Garden Irises, can have many possible explanations other than linkage. That particular association is now believed to be due to the biosynthetic pathway common to both pigments. The only way to demonstrate linkage is through the analysis of segregation ratios. This is difficult to do in irises and requires very large numbers of seedlings. It will probably be some time before we will be able to discuss examples of linkage in irises, except hypothetically.

PARENTAL GENOTYPES: PlplPbpb

Figure 4. Punnet square for a dihybrid diploid cross. The Punnet square serves as a simple way to combine gametes from 2 parents into all possible combinations. The procedure involves placing gametes from one parent across the top, those from the other parent across the side. Then, in each square the gametes from that row and that column are added together and their probabilities multiplied. In this example both parents are doubly heterozygous and each produces four gamete types in equal frequencies; hence, there are 16 zygote types also in equal frequencies. Some zygote types are genetically identical since it doesn't matter which allele came from which parent. Others are phenotypically identical. To determine phenotype ratios, those zygote types with identical phenotypes are simply added together. In the illustration the different phenotypes are represented by the different shadings.

IV. TETRAPLOID CYTOGENETICS

Modern tall bearded irises contain four sets of chromosomes and are, hence, tetraploid. Triploid, pentaploid, octaploid (three, five, and eight sets respectively) and the other terms in this series are self-explanatory given the Greek roots for the numerals. Anything with more than two sets of chromosomes is called polyploid.

The ratios Mendel found and the laws he formulated for diploid plants will not apply to the genetics of tall beardeds. In these irises chromosomes do not exist in pairs, but in quartets, and not one but two from each kind are included in the gametes. Consequently, each locus is present four times, and each gamete contains two alleles of the four. The different possible ways that chromosomes may pair lead to different segregations of the alleles on those chromosomes. The meiotic behavior of tetraploids is the major determiner of the possible gametetypes and their probabilities.

Tetraploids are generally divided into two types depending upon the relationships among the four basic chromosome sets. If all four sets are homologous, i.e., there are four representatives of each chromosome, the tetraploid is called an autotetraploid. If one or more of the chromosome sets is not homologous to the others, the tetraploid is called an allotetraploid. The chromosome numbers in autopolyploids are given in reference to the genomic chromosome number represented by x. Thus I. trojana is 2n=4x=48. The 2n means that the haploid or gametic number of chromosomes is 24. The 4x indicates that the species is tetraploid with a genomic chromosome number of 12. Allopolyploids, unfortunately, cannot be so represented if there are genomes present with different chromosome numbers. The notation introduced by Werckmeister (1960) whereby the various polyploids are represented not by the total chromosome but by the genomic numbers is a very convenient system and immediately indicates many major fertility problems. In this system the autotetraploid I. trojana is 12 12/12 12, the allotetraploid standard dwarfs are 12 8/ 12 8, and the arilmedians (arilbred x standard dwarf) are 12 10/12 8. (See also Kidd 1969, and Wilkes 1975.)

AUTOTETRAPLOIDS

Because each chromosome type exists in four homologous copies, meiotic pairing in autotetraploids is not the same as in diploids. One possibility is that the homologs form bivalents, in which case two bivalents are formed for each type, and the two homologs forming a bivalent in any one meiosis can be a random choice from among the four. Thus, in one meiotic cell of an AAaa variety, pairing may be A/A, a/a, while in another it could be A/a, A/a. Random bivalent formation results in chromosome segregation.

In contrast, as a different possibility, all four chromosomes can be involved in one pairing complex, called a quadrivalent. This can happen because a given chromosome can pair with one homolog in one region and another homolog in a different region. Quadrivalents give rise to chromatid segregation.

Chromatid and chromosome segregation are also functions of the position of a locus relative to the kinetochore. If a locus is far enough from the kinetochore, the amount of exchange will mean that alleles at the locus will segregate largely independently of their particular kinetochore. When paired in meiosis, the chromosomes will have doubled, except for their kinetochores, so that each locus is actually represented eight times. In an ideal autotetraploid the homologs pair randomly to form quadrivalents and each chromatid behaves independently with respect to exchanges with homologous chromatids. Thus, segregation for a locus far from the kinetochore essentially amounts to choosing two alleles for a gamete from a population of eight alleles, one on each chromatid (chromatid segregation), instead of two alleles from a population of only the original four alleles (chromosome segregation). For such a locus one would obtain from an AAaa plant 3/14 AA, 8/14 Aa, and 3/14 aa gametes. The segregation at a locus more tightly linked to its kinetochore will yield gamete arrays with probabilities intermediate between those for the unlinked or distal loci and those for the tightly linked or proximal loci. Loci at the kinetochore would give 1/6 AA, 4/6 Aa, and 1/6 aa gametes for the same AAaa genotype (chromosome segregation). The exact segregation probabilities for a single locus in these tetraploids depend upon the position of the locus; such positions are not known for any loci in irises.

This difference in segregation frequencies may not seem large but can be dramatic and significant. Consider an AAAa autotetraploid. If the locus is completely linked to the kinetochore, it is impossible to obtain an aa gamete. However, if the locus is far enough from the kinetochore to behave independently, one finds 15/26 AA, 12/28 Aa, and 1/28 aa gametes! Because some of the probabilities are so small, determining ratios in tetraploids is virtually impossible and requires extremely large numbers of seedlings from a single cross. For example, studies of segregation at the plicata locus have shown that an allele present only once in a plant will segregate as two copies in one gamete about once in a hundred gametes, only a quarter of the maximum possible (Kidd 1976): several hundred seedlings were required.

AMPHIDIPLOIDS

The most regular of the allotetraploids with nonhomologous genomes are the amphidiploids with two chromosome sets from one species and two chromosome sets from a different species. As the name indicates, such a plant behaves as a diploid for both chromosome sets. At meiosis the two chromosome sets from one diploid species pair with each other as do the chromosome sets from the other species: each chromosome has one and only one homolog and pairs only with that one. Each gamete therefore contains one complete set of chromosomes from each species. Such tetraploids generally are fertile because meiosis is regular.

2. Simonet preferred the term "syndesis" rather than the more usual
"synapsis" for the association of homologous chromosomes.

Simonet (1934) described an excellent example of a new amphidiploid, I. autosyndetica Simonet2 (2n=46 or 11 12/11 12) originating from I. hoogiana x I. macrantha. This new species is typical of true amphidiploids in its very regular meiotic behavior, normally forming exactly 23 bivalents. This regularity is the result of strict autosynapsis (pairing of homologous chromosomes both of which are descended from the same species). While its meiotic behavior is usually completely regular, an amphidiploid's genetic behavior is quite different from that of both the diploids and the autotetraploids: there are four alleles at most loci, but segregation occurs only between pairs of alleles as in diploids. Each chromosome has only one homolog, and for this pair of chromosomes segregation occurs exactly as in diploids. However, the bd on this pair of chromosomes are also represented on other chromosomes: each locus is present on two separate chromosome pairs. If an allele at a given locus is included in a gamete, it is impossible for the allele on the homologous chromosome to be included too. If one considers the amphidiploid of genotype EEee, the segregation will be quite different if it is E/E; e/e from the segregation if it is E/e; E/e. In the first case it is homozygous (producing only one kind of gamete) and hence truebreeding. In the second case it is doubly heterozygous and will yield 1/4 EE, 1/4 Ee, 1/4 eE and 1/4 ee gametes; it behaves in this case exactly as though it were a diploid with duplicate loci controlling one characteristic, and indeed it could be considered as such.

The behavior of amphidiploids is particularly important in iris hybridizing since two major groups of hybrids behave essentially as amphidiploids. The standard dwarf bearded irises derived from tetraploid tall bearded x I. pumila have been shown by Randolph and Heinig (1951) to be regular in forming 20 bivalents at meiosis—12 bivalents formed by the two tall bearded genomes and 8 bivalents formed by the two I. pumila genomes. The 44 chromosome arilbreds introduced by Clarence G. White seem to be similar with 12 pairs of tall bearded chromosomes and 10 pairs of aril chromosomes. While meiotic studies are needed for confirmation, breeding data and chromosome counts provide strong support for this belief. Randolph (1964) feels the term amphidiploid cannot strictly be applied to the majority of arilbreds because several species have contributed to each "genomic" set. Similar arguments would prevent its use for many standard dwarfs as well. Nonetheless, all available evidence suggests that these two groups behave genetically the same as true amphidiploids. Calling them amphidiploids is reasonable.

In these amphidiploid hybrids most loci are duplicated in the two genomes, but behave as separate, independent diploid loci. A recessive allele at a locus in one genome cannot have its phenotype expressed in the plant unless the same or a similar allele occurs and becomes homozygous at the duplicate locus in the other genome. Among amphidiploid arilbreds there are no plicatas, no tangerine pinks, and essentially no whites. All three of these will be difficult, if not impossible, to obtain since the alleles necessary seem not to exist in the aril genome and new mutations may be required. In crosses involving the standard dwarfs these patterns have occurred; the plicatas and recessive whites are due to the presence of a pla or equivalent allele in the I. pumila genome (chap. 8).

The assumption so far in this discussion has been that if a locus (defined primarily by function) exists in one iris species, the same locus, though possibly with distinct types of alleles, exists in all other iris species. This simplifies genetic discussions and may be valid generally, but there are exceptions. In the course of evolution, a nonessential locus may be lost entirely in some species. In other species, mutations at a locus may produce an entirely new function usually after a duplication of the old locus so that no loss of the original function is involved, but sometimes at the expense of the old function. Irrespective of mechanism, some loci may exist in only one or a few species, and other loci may be absent from one or a few species. Flower color and pattern usually show more variation because they are relatively nonessential characteristics of the plant; these irregularities are very important to the hybridizer.

One clear example of a trait in some species which appears to have no corresponding genetic locus in other species is the signal spot of the dwarf and aril irises, and the genetic basis for this pattern may be different in the two groups. There is tentative evidence in some arilbreds (Kidd 1964) that the signal spot shows diploid inheritance, consistent with an amphidiploid behavior of the C. G. White arilbreds and the existence of the signal-spot locus only in the aril genome. Similar situations may exist in the tall bearded hybrids, but since the absence of a locus is generally indistinguishable from the presence of a nonfunctioning allele at that locus, such occurrences will be difficult to demonstrate even if they do occur.

OTHER ALLOTETRAPLOIDS

Many types of allotetraploids exist in irises. These show a spectrum of fertility from as high as the amphidiploids to complete sterility. The degree of fertility can often be correlated roughly with the fraction of chromosomes that have homologs. In the amphidiploid arilbreds and standard dwarfs essentially every chromosome has a homolog. Backcrossing of these to a tall bearded variety is common, producing the Mohr-type arilbreds (chap. 11) and the intermediates (chap. 10). Each of these types has three sets of tall chromosomes and one set of nonhomologous chromosomes (either of aril origin or from I. pumila). Unmatched chromosomes are the cause of the reduced fertility present in both of these groups.

ARILMEDIAN HYBRIDS

An intercross of an amphidiploid arilbred with an amphidiploid standard dwarf would give a hybrid with two sets of tall bearded chromosomes, one set of aril chromosomes, and one set of I. pumila chromosomes. With two completely unmatched sets of chromosomes these hybrids will generally have low fertility. The extreme example of an allotetraploid, which probably does not exist in irises, would be one with four different sets of chromosomes, which were not able to pair. Such a plant would be sterile because meiosis, which depends on homologs pairing, could not occur. Only unreduced gametes (see below) would allow such a hybrid to reproduce.

ALLOTETRAPLOID FERTILITY

Fertility and genetic behavior of allotetraploids is even more complex in hybrids containing both homologous and homeologous ("almost identical") chromosomes. During meiosis of such hybrids, some cells will have the normal pairing of two chromosomes, others the association of three chromosomes of a set of four, and still others the association of all four similar chromosomes as in an autotetraploid. Because of the random occurrence of each type of pairing, it would be impossible to predict what segregation ratios might result. Conversely, studying observed ratios might be largely uninformative. This pessimistic conclusion specifically applies to the modern tall bearded hybrids.

Heinig and Randolph (1963) studied the meiotic behavior of tetraploid iris species and tall bearded varieties. Their observations indicate that for many and possibly most chromosomes tetrasomic pairing can occur, even among the technically allotetraploid cultivars. However, in any one variety not all chromosomes showed such pairing and the number that formed quadrivalents varied among the cultivars studied. Nevertheless, enough homology exists among the n=12 genomes to allow occasional allosynapsis (the pairing of chromosomes from different species). Thus, in tall bearded hybridizing it is possible that any allele can eventually be recovered as a tetraploid homozygote with four doses.

THE DIPLOID x TETRAPLOID CROSS

Triploids usually arise from a cross of a diploid by a tetraploid, the 2x x 4x cross. It would be expected that such crosses would yield full seed pods since both parents are balanced fertile types and that are balanced fertile types and that most if not all of the seeds would produce triploid seedlings. Such is not the case. Diploid x tetraploid crosses are notoriously difficult to effect--many pollinations may yield few pods, each with but a few misshapen seeds. The greatest surprise is that when the resulting seedlings are examined, a high percentage are tetraploid (Simonet, in Report of the Florence Symposium, 1963); the diploid parent has contributed a gamete with the full somatic (diploid) number of chromosomes.

As explained in section II of this chapter the number of chromosomes in a normal gamete is reduced to half the somatic number. Unreduced gametes instead contain the full somatic number of chromosomes. Although they rarely occur in irises, they are important. While no iris studies have been done, the types of errors have been studied in other plants. Two situations can be distinguished in irises. Normally fertile plants can produce unreduced gametes through partial failure of a meiotic division or of one of the postmeiotic divisions. Such gametes are meiotic products and are not genetically identical to the parent because of recombination and segregation. However, plants normally sterile because of nonhomologous chromosome sets cannot undergo a normal meiotic reduction division. Any diploid gamete they produce is likely to be genetically identical to the somatic cells of the plant.

Highly effective screens against triploid embryos exist in some plants, e.g. potato (see Kidd 1969 for references), for which results are more spectacular than in irises. Simonet (1963) presents convincing evidence that a similar phenomenon exists in irises and accounts for the high number of tetraploids from diploid x tetraploid crosses, much higher than the frequency of unreduced gametes. Many tetraploids are known to have arisen from such crosses, and in Simonet's controlled study 40 percent of the offspring were tetraploids.

Studies in irises document more examples of unreduced female gametes than of unreduced male gametes, but this probably does not reflect a greater frequency of unreduced gametes occurring in the female. A single unreduced ovule in an otherwise sterile plant will very likely produce a good seed if viable pollen is used in sufficient quantity. However, germination of a pollen grain on the stigma is stimulated by the presence of other viable, germinating pollen grains in close proximity. Thus, a few good grains among many aborted ones probably are not able to function. A diploid, sterile because of nonhomologous genomes, e.g., William Mohr, will be "fertile" only through the occasional unreduced gamete and then only as a female parent.

When normally fertile diploids are used, unreduced pollen grains can function because of the surrounding viable, reduced, normal pollen. Simonet (1963) documented several instances and found no significant difference in frequency of unreduced ovules and pollen. Both can occur, but in some cases unreduced female gametes would be easier to recover.

Modern tall bearded irises are based on the spectacular results obtained from unreduced diploid gametes in diploid x tetraploid crosses (Randolph, Garden Irises, chap. 23). In arilbreds interesting, if not quite as spectacular, results from unreduced gametes dramatically illustrate the advantages of such crosses. However, many pollinations yielding few seeds make this type of hybridizing difficult, slow, and often discouraging. The hybridizer should only pursue such a course if endowed with considerable time and perseverance.

V. TETRAPLOID GENETICS

Tetraploids, in general, are bigger. Since the tetraploid nucleus is twice as large as the diploid nucleus, a cell which contains a tetraploid nucleus is almost always somewhat larger than one with a diploid nucleus: the jump from diploid to tetraploid increases plant and flower size by slightly increasing the size of each cell. More important than increased size, however, is the greatly increased variability in tetraploids. Diploids have only three possible genotypes at a locus with two types of alleles, six at a locus with three types of alleles. Tetraploids have , respectively, five and fifteen possible genotypes. These different genotypes may have slightly different phenotypic expression, frequently referred to as a dosage effect, but the different phenotypes may not always show a strict dosage relationship. Tetraploid plicatas are often considered an example of the greater variation possible in tetraploids, but this may be due only to the greater variety of modifying genes in modern tetraploids since diploid plicatas show almost as much variation as tetraploids.

TETRAPLOID SEGREGATION FREQUENCIES

The iris hybridizer has been able to capitalize upon the variability of tetraploids to create the spectacular tall bearded garden varieties. This variation, however, occurs at the expense of predictable genetic behavior. Most of the complicated genetic behavior of tetraploids has been discussed in conjunction with the cytogenetics of tetraploids. In summary, the segregation ratios, or more precisely the gamete types and associated probabilities, are determined by three parameters in tetraploids: the genotype, the position of the locus on the chromosome, and the pairing relationships of the chromosomes. In diploids only the first, genotype, is involved. While the genotype of a plant may be predicted from its phenotype and parentage, in tetraploids the second two parameters cannot be estimated before offspring are raised. Even working backwards from the observed segregation ratios, the contributions of these two parameters will be confounded. Table I has been prepared as an aid in prediction for the hybridizer (Kidd 1969). This table differs in many respects from a similar table in Garden Irises; the differences are explained by Kidd (1969).

Table 1. The exact and approximate (in parentheses) probabilities of recovering the recessive phenotype, i.e., the genotype aaaa, from various types of tetraploid crosses involving segregation for two alleles. For random chromosome pairing the minimum value is observed when only bivalents are formed or the locus is tightly linked to the kinetochore. The maximum is observed when tetrasomic pairing regularly occurs and the locus is distant from the kinetochore. For more explanation of the derivation of this table see Kidd (1969).

Values in Table 1 are given as probabilities rather than ratios. The use of probabilities is preferable to ratios in that probabilities can be easily combined and used to determine the number of seedlings that should be raised from a cross (see Kidd 1968). The values given under "random" chromosome pairing are those for loci near the kinetochore (minimum) and for loci far from the kinetochore (maximum) assuming quadrivalent formation (tetrasomic pairing). The maximum probability is not obtained under strictly random allelic (or chromatid) segregation but results from more complicated processes. Under "selective" pairing the maximum values correspond to the ratios given in Garden Irises. Other selective pairings will not give these ratios, and the values for this alternative behavior are given in the "minimum" column. Ignoring selective pairing for the moment, almost all actual values will be somewhere between the maximum and minimum. Because most tall bearded irises are complex allotetraploids, chromosome pairing may differ between even full sibs and an exact value, even if known for one variety, would not necessarily he valid for any other clone.

PREDICTING PROGENY FOR TETRAPLOID CROSSES

Table I can be used directly to determine the expected progeny frequencies from crosses involving only two alleles, if the genotypes of the plants being crossed are known. For example, both Emma Louisa (Gypsy Lullaby x Memphis Lass) and Diplomacy (Rococo x Whole Cloth) should be Pl Pl pl pl. A cross of the two would be type 5 in the table. Reading across for that line for the frequency of plicatas among the seedlings shows a minimum of 1/36 and a maximum of about 1/20. Among about 50 seedlings there should be a few plicatas, but not many, and depending on the vagaries of chance there might he no plicatas. Another trait in this same cross, dominant inhibition of anthocyanin in the standards, IS, illustrates a cross of type 8 since both plants are probably IS iS iS iS. The frequency of "no inhibition" (the recessive trait) will be at least 1/4 (25 percent) and possibly as high as 10/35 (about 29 percent). Conversely, some inhibition in the standards will occur in 71 to 75 percent of the seedlings.

Since probabilities for independent traits can be multiplied, the probability (or expected frequency) of amoena (or neglecta) plicatas from this cross would he between a low of 71/100 x 1/36 = 71/3600 = .0197 (nearly 2 percent) and a high of 3/4 x 1/20 = 3/80 = .0375 (nearly 4 percent). Both are low frequencies and many seedlings would have to be raised to he sure of obtaining even one amoena (or neglecta) plicata. Keppel (1971) reported on a cross of type 9 for IS and type 7 for plicata. He did obtain the expected segregation for Pl and pl, but did not obtain the expected frequencies of IS among the plicata seedlings. He attributed the discrepancy to the difficulty of classifying plicata seedlings for slight suppression of anthocyanin in the standards since there may be only a very narrow band of pigment to begin with. That seems the most likely explanation.

Table I is more difficult to use if three alleles are present in one or both parents of a cross. For such cases, the procedures used for diploids are applicable: determine gamete types and frequencies and then multiply directly or use a Punnett square. The important difference from diploids is that different gamete types will not be in equal frequency. Because there is a range of expected frequencies for each gamete type, either a single rough approximation can he made or separate estimates made for the minimum and the maximum.

Another initial approach is to consider complete linkage to the kinetochore and each allele identified separately. Then only six types of gametes are possible, all in equal frequency. Figure 5 diagrams how, predictions for such a cross could he made. This does not consider any gamete containing two copies of a single chromatid. Specifically, the pod parent produces no pla pla gametes. Table I shows that pla homozygotes could occur from this cross (type 2) as frequently as 1/108. This will not change the frequencies of the other types of seedlings very much.

The method outlined in figure 5 could be used for loci far from the kinetochore, but instead of six gamete types there would be 28—all possible combinations of two picked from eight copies of the locus, a pair of duplicates for each of the four alleles. While logically simple it is an arduous task. Since only rough estimates are necessary anyway, the six gamete types in figure 5 should suffice with the qualification that gametes with two copies of any of the four alleles could occur rarely and therefore homozygotes for such alleles are possible.

VI. GENETICS OF SPECIFIC TRAITS: FLOWER COLOR AND PATTERN

INTRODUCTION

Of the several characteristics in irises showing single locus inheritance, the best understood are flower characters such as color and color patterns. The color and pattern loci can generally he divided into those affecting anthocyanins (blues and lavenders) and those affecting carotenoid pigments (yellows and pinks). These two pigment families are biochemically distinct and seem to be genetically independent. (See chapter 25 for more information on the pigments in irises.) Patterns involving combinations of both pigment types, such as blends, pink plicatas and red-bearded blues, are mostly easily studied by considering each pigment separately. In this way simple genetic patterns are more likely to be observed. Identifying and understanding the individual effects of single loci provide the necessary foundation for understanding the combined effects of many loci.

Figure 5. Modified Punnett square for predicting results of a tetraploid cross. The genotypes of the two kinds are given at the top of the figure. Each allele is considered as a separately identifiable allele, symbolized A1, etc. Assuming random chromatid segregation and complete linkage to the kinetochore, the six gamete types possible are easily determined. Each of the types listed across the top and side of the diagram occurs with equal frequency, 1/6. Since in the pod parent A1 and A2, are actually identical, the gametes have been grouped accordingly and the frequencies added to give the values on the left side of the diagram. In the pollen parent A3 and A4 are both pla alleles; a different grouping for the pollen is therefore required, and the appropriate groupings and frequencies are given across the top of the figure. The widths of the columns and rows are proportional to the gamete frequencies, so the various boxes are drawn to scale in terms of frequency among the total progeny. All of those genotypes in the cross hatched area will be selfs. Only one-twelfth of the seedlings have no Pl allele; these will be plicatas. As can be seen in the lower right hand corner of the diagram, this one-twelfth of the seedlings will be, in turn, 2/3 with 2 pla alleles and 1/3 with 3 pla alleles.

While a nonanthocyanin flower may be white, it may also be yellow, pink, orange, etc., because of carotenoid pigments. Therefore, to emphasize that each pigment type is being considered separately, any nonanthocyanin or reduced anthocyanin phenotype or a plant with such a phenotype will be symbolized RA and any absent or reduced carotenoid phenotype or plant will be symbolized RC. The word reduced is emphasized because small amounts of pigment are found in several phenotypes which appear on casual inspection to lack that pigment. Also, a restriction of pigment to only a small part of the flower may have the same visual effect as complete absence of pigment. Thus, a clear distinction between presence/absence of pigment and pigment distribution or pattern is sometimes quite difficult.

For most loci known to affect pigments, reduction of pigment is recessive to its presence. It is therefore convenient to extend the symbolism to RRA for a recessively inherited reduction in anthocyanin (or a plant with that phenotype) and similarly to RRC for carotenoids. There are also at least two dominantly inherited inhibitors of anthocyanin. These will be discussed in the following sections.

ANTHOCYANIN VARIATION AND PATTERNS

THE PLICATA LOCUS

This locus affects the presence and distribution of anthocyanin pigments. The dominant pattern is actually no pattern-full pigmentation with no pattern; the allele responsible is symbolized Pl. The plicata pattern, for which the locus is named, is recessive to full pigmentation and is due to a second allele, pl. A third allele, pla, with the superscript an abbreviation for "all-white," produces a third phenotype: an RRA with complete absence of anthocyanin in the homozygote (see Kidd 1976). At least one other allele, pllu is believed to exist and produce the luminata pattern (Sturtevant 1951; MIS Study Group 1972). Like the plicata pattern, luminata is recessive to full pigmentation and dominant to the pla RRA.

The two patterns, plicata and luminata, are almost the reverse of each other. In plicatas the pigment is at the margins of the petals with the center of the petal largely unpigmented, the hafts and style arms are pigmented, and if there is stitching it lies over the veins in the petal. In luminatas the pigment is in the center of the petal with the edges often unpigmented; the hafts and style arms are unpigmented or essentially so. The pigmentation of the petal is usually not uniform, but marbled, being lighter or absent over the veins, as exemplified by Moonlit Sea. Plicata and luminata show no dominance with respect to each other—if both pl and pllu are present in the absence of Pl, both patterns are expressed. This luminata-plicata pattern has in the past been called the "fancy" or "fancy plicata" pattern, a pattern that is hard to define.

Results from many crosses are consistent with the luminata pattern being determined by another allele at the plicata locus. However, because of their complex ancestry, none of these crosses completely excludes the possibility of luminata being determined by an allele at a completely separate locus, as suggested by Werckmeister (1971) who uses havelberg as the name for the pattern. The available data are illustrated by the following crosses. Blushing Blonde crossed with a pla homozygote gave one seedling, but this was a much more intense luminata than Blushing Blonde. This one seedling is typical of several crosses showing that luminata and pla RRA do not complement in hybrids. It also suggests that the paleness of the luminata pattern in such varieties as Blushing Blonde is not a dosage effect, but due to modifiers. Matterhorn x Moonlit Sea produced six luminatas and three RRAs, demonstrating that Moonlit Sea carries the pla allele but no Pl or pl alleles. May Sky also produces gametes lacking the luminata allele but always carrying pl or pla alleles. In standard dwarfs the luminata pattern segregates only in those lines carrying the pla allele homozygous on the I. pumila chromosomes. The segregation ratios, though not well documented, appear to be diploid. For all these crosses the most parsimonious explanation puts luminata at the plicata locus.

Many "modern" plicatas are known to carry the pla allele. April. Melody has been observed in many crosses to give 50 percent plicatas, 50 percent RRA when crossed to a pla homozygote (Kidd 1976); thus, it has only one pl allele and three pla alleles. Such good evidence is not available for other varieties, but the following probably have two pla alleles: Tea Apron, Blueberry Trim, Pink Ember. Other varieties are known to carry only one pla allele (Kidd 1976). Any immediate descendants of April Melody have a high probability of carrying the pla allele.

The plicata series may contain more than four alleles. An irregular wash of color on the backs of the falls, most intense on those parts exposed on the outside of the bud, is another pattern found in many plicatas, luminatas, and pla RRAs. This celestar pattern (so named for an early variety in which it was very pronounced) is definitely inherited and may be determined by a plc allele. However, it has not been well studied, partly because it is difficult to detect in strongly colored plicatas and luminata-plicatas and partly because it is not considered a desirable pattern. Variations of the plicata pattern, such as pigment largely confined to the haft areas and bases of the standards, are probably determined by other loci acting as modifiers.

PLICATA IN THE STANDARD DWARFS

Though the plicata pattern and its allele derive from garden diploids (Schreiner 1971), the locus exists in at least two different species. It exists in I. pallida (Megson and Megson 1968; Tearington and Tearington 1969) and in Cretica and other I. pumila varieties. In both species it is the pla allele (or one indistinguishable from it) that marks the locus. The pla allele was not previously known to exist in diploids or in tetraploid dwarfs. The existence of the pla allele in some I. pumila clones has allowed the plicata and luminata patterns to be introduced into the standard dwarfs; the most notable example is the first standard dwarf plicata, Dale Dennis, in the first generation from Cretica. Plicatas have segregated out in advanced generations of standard dwarfs from other I. pumila clones. The Pl, pl, pla and pllu alleles on the tall bearded chromosomes are expressed just as in diploid tall bearded varieties whenever the dwarf half of the amphidiploid is homozygous for the pla allele from I. pumila. Until the pla allele was introduced in a homozygous state from Cretica, the plicata alleles in the tall bearded chromosomes had never been expressed in standard dwarfs. This is a good example of the complexities introduced by the duplicate nature of amphidiploids.

SEGREGATION FREQUENCY OF LOCUS Using data from crosses involving several different plicata varieties, Kidd (1976) calculated that in tall bearded varieties tetrasomic segregation did occur at this locus. The frequency of pl pl gametes from Pl Pl Pl pl plants was slightly less than 1/100. While this segregation frequency confounds the distance from the kinetochore with the frequency of selective pairing, the fact that it is greater than zero is proof that the locus is not completely linked to the kinetochore and that tetrasomic pairing does occur at least occasionally.

OTHER RRAS

THE WHITE-1 AND WHITE-2 LOCI In addition to the "all-white" due to the pla allele, there are at least two other loci in tall bearded irises with alleles producing a recessive reduction in anthocyanin. These have been designated by Randolph and Sturtevant (Garden Irises, 1959) as the white-1 and white-2 loci. Megson and Megson (1970) have confirmed Sturtevant's (1951) finding that in diploids there is an RRA that is not related to the plicata locus. Pluie d'Or, Daystar, and Gold Imperial are examples of this RRA and define the white-1 locus. These w1 w1 homozygotes have virtually no visible anthocyanin in the flower. Sturtevant (1951) emphasized that some anthocyanin is always present at least as veining at the bases of the petals, but Gold Imperial has no visible anthocyanin. This might represent another allele at the locus or modifying effects at other loci. Dayspring is the only registered tetrapioid variety known to carry the allele (Sturtevant 1951), but the allele is believed to be present in several other varieties. While no other tetraploids are documented to be homozygous w1, Impala is an RRA with prominent veining at the petal bases, suggesting it may be homozygous w1.

White-2 was the designation given to the locus in tetraploids segregating whites from blue parents. Senorita Ilsa, from Helen McGregor x Sylvia Murray, was the type homozygote for w2, although several other varieties are likely to carry the same allele (Randolph and Sturtevant in Garden Irises 1959; Watkins 1955). In the w2 homozygotes absence of anthocyanin is usually more complete than in w1 homozygotes like Dayspring.

The w2 RRA is largely equivalent to the "b-white" of Vallette (1961), although she considered "b-white" equivalent to w1 homozygotes, e.g. Dayspring. It is certain that not all varieties she called "b-white" were homozygous for the same allele. Sturtevant, too, did not have evidence that all the varieties he considered as possible w2 homozygotes were genotypically the same. However, crosses made by Sturtevant, as well as others made by Vallette show that there is probably at least one RRA genetically distinct from w1, and pla. While Sturtevant's crosses (Randolph and Sturtevant in Garden Irises) do not conclusively show that w2 is different from pla or w1, they do show that the RRA from Gloriole is different from w1. Vallette (1961) crossed White Satin (from Blue Champagne x Sylvia Murray and probably genetically the same RRA as Senorita Ilsa) with Matterhorn; there were nine light-to-medium bitones, two medium blue selfs, one pastel amoena, five plicatas, three cool whites, two greenish whites, and two pure whites. Clearly White Satin is not a pla RRA. These results suggest that White Satin is w2 w2 w2 w2 Pl pl pla pla and Matterhorn is W2 W2 w2 w2 pla pla pla pla, and taken together with the usually different phenotype of the w1 RRA, support the existence of a white-2 locus. A complete understanding will require much more data.

A diploid RRA that complements with both pla homozygotes and w homozygotes has been found by Megson and Megson (1975). The type plant for this new locus is La Neige, a white with a tiny blue flush below the heard. Only one additional locus seems to he involved, but since it is not known whether it corresponds to the tetraploid w2 locus, the locus has not been assigned a name. The crosses by which Megson and Megson (1975) demonstrated this new locus also showed Mendelian segregation for a pattern not previously described in iris literature: random irregular violet spots on a light background. The crosses suggest that this pattern, termed "maculosa," may be produced by an allele of the white-1 locus. Thus, the white-1 locus may have three alleles related much as the three common alleles at the plicata locus: one for full color, one for pattern, and one for no color.

DOMINANT REDUCTION OF ANTHOCYANIN

Most white, yellow, and pink tetraploid tall bearded varieties have the common dominant inhibition of anthocyanin. The allele responsible, designated I, is virtually absent or very uncommon among diploid tall bearded hybrids. This is probably an example of a gene that has an inhibitory effect on pigment in modern hybrids but manifested some other function in its original genetic context. Sturtevant and Randolph (1945) and Randolph and Sturtevant (1959) have documented its spread among tetraploids from Kashmir White through such varieties as Snow Flurry. They speculated it may be ultimately derived from dwarf iris species. At the diploid level dominant inhibition has been shown in crosses with I. imbricata. For example, from the cross I. imbricata x Pocahontas (a medium violet plicata) Megson obtained 74 seedlings ranging from white to cream white, none with more than a trace of anthocyanin.

Whatever the origin of the common inhibitor in tetraploids, its inheritance pattern in modern hybrids is clear. A white variety (or RA) with only one "dose" of inhibitor crossed to a blue or lavender variety (necessarily homozygous for absence of the inhibitor) will give half white (or RA) and half colored seedlings. Varieties with more than one I allele will give predominantly RA seedlings (types 4 and 7 in Table 1). Crosses of two RA varieties will only rarely produce seedlings with full anthocyanin pigmentation (types 1-3, 5, 6, and 8 in Table 1). Thus, this is a useful characteristic for breeding whites, yellows, and pinks because most seedlings, even in outcrosses, will have the anthocyanin suppressed.

The way this allele acts on pigment production is not understood. Inhibition is often not complete and a slight amount of pigment is present, as in the blue whites such as Snow Flurry. This incomplete inhibition can be influenced by modifiers and is not solely a dosage effect since some varieties with only one I allele show apparently complete inhibition of visible anthocyanin. However, regardless of the visual appearance or the number of I alleles, not all anthocyanin is inhibited. In the dominant RA varieties tested anthocyanin is present as leucoanthocyanin, the colorless "pseudobase" that can be converted in vitro to a colored form by strong acid (Werckmeister 1955, 1960).

The failure of the dominant inhibition to be observed in hybrids between tall bearded irises and aril irises is well documented. Since the anthocyanin produced by the arils and the hybrids is the same as in the tall bearded varieties, this is most likely an inability of the inhibitor to affect the pigment-synthesizing apparatus of the arils.

DOMINANT INHIBITION OF ANTHOCYANIN IN THE STANDARDS

A different allele, IS, presumably at a different locus than I, produces the dominant amoena pattern. This allele has a known origin from dwarf species and was only introduced into tetraploid tall bearded irises around 1950. In 1944 Paul Cook crossed a yellow form of I. reichenbachii with Shining Waters to produce Progenitor (registered in 1951). By a series of backcrosses to tall bearded varieties he transferred the IS allele into essentially tall bearded varieties. In this way he eventually derived Whole Cloth and several other blue "amoenas" (Galyon and Warburton 1975a, 1975b). Cook obtained sufficient data on segregating progeny to demonstrate that the amoena pattern was inherited as a simple single-gene dominant.

Among Cook's introductions the range of patterns produced by the Is allele was evident. The dominant amoena pattern of Whole Cloth is but one manifestation. Often the inhibition of anthocyanin in the standards is incomplete, as in in Melodrama, producing a bitone or near amoena. In other varieties, anthocyanin is present only as a band of color around the edges of the falls, as in Emma Cook. Though a plicata-like distribution, this is a distinctly different pattern from plicata and does not show the dotting and/or stitching common in plicatas. In some extreme cases, the band of pigment may be only a narrow line or may be absent altogether. Thus the patterns range from incomplete suppression of anthocyanin in the standards, to complete suppression in the standards but none in the falls, to complete suppression in standards and partial suppression in the falls, to complete suppression in both standards and falls.

The various patterns may be largely the result of a dosage effect whereby varieties with one Is allele have the least suppression and those with three or four Is alleles show nearly complete suppression. However, modifiers must also be involved, particularly in the difference between bitones (near amoenas) and clean amoenas. Varieties of both types have been shown to segregate approximately 50 percent for Is, indicating only one copy of Is was present (Galyon and Warburton 1975a, 1975b). Paul Cook considered that the bordered-falls pattern might be a recessive trait produced by a separate locus segregating in the blue amoena line. However, the pattern did not breed true in all cases, suggesting the modification could not be strictly a recessive trait. Despite the numerous crosses made since the 1950s and the hundreds of varieties carrying the Is allele, the question of dosage effects and modifiers is not resolved.

A similar pattern has more recently been introduced into tall bearded irises from I. balkana. Balkan Glacier and Prophecy are two [405] of the few irises introduced from this origin. I. balkana produces less clear amoenas, usually with more irregular pigmentation on the falls. Since it is closely related to the dwarf parent of Progenitor, the differences in the patterns produced are probably due to modifiers rather than being caused by a second Is allele.

VARIEGATA AND RECESSIVE AMOENA

Since the introduction of the dominant amoena pattern, little hybridizing has been done with the "recessive" amoenas and variegatas. The same pattern, originating from I. variegata is responsible for both, the only difference being the absence of yellow in the amoenas and its presence in variegatas. There are many unanswered questions concerning the genetics of this pattern. While generally considered recessive, the complete absence of anthocyanin in the standards does not always breed true as some neglectas often result from a cross of two amoenas (Randolph and Sturtevant 1959). Crosses of an amoena with a self-colored variety often produce mostly neglectas (Randolph 1950), suggesting that neglecta may even be a dominant trait and amoena a modification due to other loci. A dosage effect could also explain the results. No conclusions are possible without additional data.

SIGNAL SPOTS IN ARILS AND ARILBREDS

A simple definition of this spot of anthocyanin, or additional anthocyain, in the centers of the fall petals at the ends of the beards. That definition is too simple since characteristics other than pigment are also involved. In the pure oncocyclus and in some hybrids the signal spot is not solely a spot of pigment. The area also has a velvety texture, usually due to each cell having a small papilla. In addition, carotene pigment may also be concentrated in this area; some pure oncocyclus have yellow signals. Since the two types of pigment concentration and the papillate cells are presumably under different genetic control, the term "signal spot" may be used to refer to different genetic characteristics.

Though no generic studies have been done in pure arils, the signal spot considered solely as anthocyanin pigment appears basically dominant since diploid hybrids, such as the regeliocyclus varieties and William Mohr, have signals. However, that does not indicate whether one or more loci are involved. Kidd (1964), based on crosses in arilbreds, suggested that one major locus was involved but that the heterozygous phenotype was affected by modifiers. While that explanation may be too simplified, no data are available to suggest an alternative hypothesis. The problem is complicated by the virtual absence of arilbred varieties that completely lack all traces of additional anthocyanin at the end of the beard. The anthocyanin [406] signal spot is inherited independently of anthocyanin in the rest of the flower. Indeed, one RRA arilbred seedling had no visible anthocyanin, even at the hafts or bases of the petals, except for a large signal spot.

THE PUMILA SPOT PATTERN

This spot of anthocyanin pigment in the center of the fall petal is believed to be inherited as a single-locus dominant trait. The size of the spot can be greatly affected by modifiers and possibly also by a dosage effect — in some cases virtually the entire fall can be colored. Though much discussed, the genetics of this trait has little solid documentation from specific protected crosses. Genetic studies of the spot in the related I. pseudopumila by Pray (pers. com.) have not yet been able to show simple inheritance patterns. Thus, the inheritance of spots in I. pumila may be more complex than commonly believed.

ANTHOCYANIN VARIATION IN BEARDED IRISES

Only one different anthocyanin has been studied genetically. The diploid tall bearded variety Floridor has been shown by Werckmeister (1963; Werckmeister et al. 1966) to have a different glycoside of delphinidin which they call floridorin. Werckmeister backcrossed a plicata seedling from Floridor x a red-violet seedling to Floridor. Of 88 seedlings, 45 had the unusual dove-blue color of Floridor and 43 had an ordinary violet color. The difference thus segregates as a single locus which could be designated the "floridorin locus." Presence of floridorin is recessive; the common anthocyanin delphanin results from genotypes FF and Ff; floridorin is produced instead by the genotype ff.

ANTHOCYANIN VARIATION IN BEARDLESS IRISES

Two genetically distinct whites (both RRA's) exist among the Siberian irises: one occurs in I. sibirica and the other occurs in I. sanguinea. lntercrosses of the two whites produce purples (Tiffney 1971). Vaughn (1974) has summarized data on crosses involving anthocyanin pigmentation in the Siberians. In the sibirica varieties, both red and pink coloring segregate from blue/purple coloration, thereby defining one locus responsible for the distinction between the two groups of colors. lntercrosses among reds and pinks produce more reds and pinks; reds and pinks crossed with sibirica-type whites produce pale reds and pale pinks. When reds or pinks are crossed with the sanguinea-type whites, blues result. Thus, the red and pink colors seem to be controlled by the locus that produces the sibirica-type white.

Vaughn (unpublished) has also studied the pigments of one representative of each color type: blue, red, pink, doll (sibirica) white, and clear (sanguinea) white. The blue flower had delphanin, the red had both delphanin and ensatin (a glycoside of malvidin), the pink had only ensatin, the sibirica white had small amounts of ensatin, and the clear (sanguinea) white had neither. The clear white did have large amounts of a related flavone, tentatively identified as swertisin, suggesting a blockage in the biosynthesis of anthocyanins that results in an increased concentration of flavone. Thus, the biochemical evidence is consistent with the genetic in suggesting that at least two complementing loci are involved.

One of the loci is postulated to be a color locus which controls, in order of dominance, the four phenotypes, blue, red, pink, and sibirica-type white, by a series of four alleles: C for blue, cr for red, cp for pink, and c for white. The clear (sanguinea) whites are homozygous are at a separate locus, the white locus.

The inheritance of flower color in I. versicolor has been studied by Tiffney (chap. 19) and is similar to that in the Siberians.

GENETICS OF CAROTENOID PIGMENTS
INTRODUCTION

There are several, apparently independent, loci controlling carotenoid pigmentation of flowers. These loci determine whether carotenoids are present, which specific carotenoid pigments are present, and in what pattern the pigments are distributed. At least three separate loci, all complementary, are necessary for carotenoid pigments to be present in the flowers of the n=12 diploid bearded irises. The pigment in the tangerine pinks, lycopene, is primarily controlled by another single locus. Color patterns for the carotenoid pigments are known to be inherited, but no data are yet available on the mode of inheritance and the number of loci involved in determining the various patterns.

The synthetic pathway for the carotenoid pigments is completely independent of that for the anthocyanin pigments and under separate genetic control, Witt (1971) questions the complete independence of patterns for the two types of pigment. The exceptional examples she cites suggest that some developmental processes do affect both pigment types. However, the patterns for the two types of pigments seem to be independent in the vast majority of crosses.

RECESSIVE REDUCTION OF CAROTENOIDS

In the tetraploid tall bearded irises yellow pigment is usually considered dominant to absence of yellow pigment. However, clear studies of its inheritance have only been done in diploids (Megson and Megson 1975). The inheritance of these pigments may be controlled by as complex a genetic structure as the numerous loci involved in anthocyanin production and patterns. The existence of three separate complementation groups, and hence three separate loci, is clearly documented. The three loci have been tentatively labeled Yellow-I, Yellow-2, and Yellow-3. The diploid variety Daystar is the type example for the homozygote recessive at the Yellow-I (Y1) locus: y1y1. The species Iris reginae is the type example for a homozygote recessive at the Yellow-2 (Y2) locus: y2y2. The variety Chartier is the type for a Yellow-3 (Y3) homozygous recessive: y3y3. These three type cultivars have little apparent yellow pigment (they are RRC "whites"), but when intercrossed in any combination can produce seedlings with full yellow pigmentation.

Understanding the observed ratios from diploid crosses is complicated by the fact that complete complementation among the three classes of nonyellows has not yet been observed. While any cross between a y1 nonyellow, a y2 nonyellow and a y3 nonyellow will give some yellow offspring, only 1:1 ratios of yellow: nonyellow have been observed. Thus, though Daystar is homozygous y1y1 it must also be heterozygous at one of the other loci, probably the Y3 locus. The yellow variety Gold Imperial when crossed with either Daystar or Zero gives 50 percent yellow, 50 percent nonyellow and hence is heterozygous at the Y1 locus. Gold Imperial, when crossed with I. reginae, a y2 homozygote, also gives approximately 50 percent yellow and 50 percent nonyellow seedlings. The same has been observed for a cross of Gold Imperial with Chartier, the y3 homozygote, though in this case only three yellow and two nonyellow seedlings were raised.

These three complementary loci are probably not loci in the biochemical pathway producing carotenoid pigments.. Indeed, a homozygous recessive for the Y1 locus has visible amounts of yellow pigment; it simply has a very restricted distribution of that pigment; both Daystar and Zero (a Megson seedling that is a y1y1 homozygote) have yellow at the haft but are otherwise white. Moreover, many carotenoid pigments are essential copigments in the photosynthetic process in chloroplasts and must be present in the leaves of any viable plant. Only if completely separate genes control synthesis of the pigment in the flower, might these loci be parts of the biosynthetic pathway. They are more probably mutants in a developmental or differentiation sequence. The sequence would control the distribution of the pigment and whether or not it is synthesized in flowers, but these would not be alleles for the actual enzymes involved in the biosynthesis.

DOMINANT REDUCTION OF CAROTENOIDS?

A dominant inhibitor of carotenoids may exist. Nicholls (1933) first suggested this possibility to explain his difficulty in getting any yellow seedlings from Kashmir White and related varieties. Later. Randolph and Sturtevant (1959) reported a cross of the white Spanish Peaks by the yellow amoena, Mystic Melody; all of the seedlings were whites, with only some yellow in the center of the flower. Since Mystic Melody is believed to carry all the necessary alleles for full yellow color, this result suggests that Spanish Peaks carries a dominant inhibitor of carotenoid pigments. This hypothesis should be investigated by crossing Spanish Peaks with yellows, pinks, or blends which are not related to the yellow amoenas.

TANGERINE PINK

The tangerine pink irises represent the first occurrence of a new pigment identified in irises. These irises have a different carotenoid pigment, lycopene. In most of the plants with this trait, the major pigment is lycopene and ordinary beta-carotene is in very reduced amounts. The production of lycopene instead of carotene appears to be controlled by a single locus, called logically the tangerine locus. The t allele, when homozygous, appears to disert pigment synthesis from carotene into lycopene. Since the biosynthetic pathways for carotene are not well understood, the exact mechanism of this diversion of pigment synthesis is unknown (chap. 25). The t allele is not sufficient for the production of lycopene, all of the remaining genetic and biosynthetic apparatus for carotene synthesis must also be present. Thus, the tangerine pink irises are basically yellow irises, but with a different pigment produced. Realizing this, it is easy to understand that any pattern available for yellow irises could also occur with lycopene instead of carotene, i.e. as a pattern in pink instead of yellow.

Since pink (lycopene) is recessive to yellow but production of carotenoids is dominant to "no carotenoids produced," it is easily seen that in a cross of a pink iris with one lacking carotenoid pigments, say a blue amoena, most seedlings will have yellow pigment. The pink parent provides the necessary genes for producing the yellow pigments and the other parent provides the T allele preventing the "diversion" to lycopene.

Virtually all tall bearded pinks carry the I allele, which inhibits anthocyanin. Unless two pinks are also homozygous for I, it is possible for them to produce seedlings showing blue as well as pink pigments. Datebook arose directly from a cross of two pinks, but other mauve to raspberry varieties, such as Raspberry Ripples and Pretty Karen, are genetically the same: tttt iiii. lntercrossed they produce seedlings with both pigments; crossed to pinks they can produce pinks because of the introduction of the I allele. The only pink which definitely does not carry the I allele is Summer Silk; it is homozygous for pla.

Pinks have never occurred among I. pumila seedlings, but Berlin (1969) has reported that a pink seedling resulted directly from a cross of a pink tall bearded by a white I. pumila clone collected in Hungary. Warburton (pers. com.) also reports getting a pink SDB directly from a pink tall bearded crossed with an I. pumila clone. These data, together with the reported origins of other SDB pinks (Brown 1970; Roberts 1970). suggest that I. pumila does not carry the genes for lycopene synthesis, but does carry alleles which block carotenoid synthesis, as determined in I. pumila, in such a way that complementation with tall bearded loci does not occur and lycopene is synthesized by the tall bearded genes if no T allele is present. I. pumila may carry the t allele or may lack the tangerine locus altogether.

CAROTENOID AMOENAS

The genetics of the yellow amoena pattern in tall bearded irises is not well understood. In crosses of amoenas with unrelated yellow selfs, only yellow selfs occur among the progeny (Randolph and Sturtevant 1959). In intercrosses of amoenas, bitones as well as amoenas occur, but there are no selfs. Reduction of carotenoid pigments in the standards thus appears recessive, but cannot be ascribed to only one locus. The degree of reduction, either to bitone or to amoena, seems to involve genetically independent factors, but the data suggest that at most only a few loci are involved. Theoretically, the more recent pink amoenas differ only in being homozygous for the t allele.

The pattern is not limited to tall bearded irises. It has been observed in seedlings from Sylvia (a yellow self) selfed by Witt (pers. com.). In a repeat of that cross one amoena was observed among 25 yellow selfs and bitones (Kidd, unpublished). This low frequency in diploids suggests a complex inheritance. Other examples of the pattern exist in the SDBs and in other groups, for example, in the lB Sea Foam

In the arilbreds, yellow amoena is a common pattern that occurs in progeny from two nonamoena parents. It thus appears to be determined by one to only a few loci. It may have been introduced on the aril chromosomes, since some oncocyclus species have a yellow amoena pattern in combination with a self or bitone anthocyanin pattern. Unfortunately, the modern arilbred hybrids are so heterozygous and their parentage is so complex that no meaningful genetic studies of this trait have yet been possible.

BEARD COLORS

Beard color is not a simple trait, but has all the complexities of flower color. Beard color seems to be partially, but not completely, independent of the color of the rest of the flower. Both anthocyanins and carotenoid pigments may be present in the beard, separately or together, and, in many caws, irrespective of whether the same pigments are present in the rest of the flower. Patterns sometimes occur within the beard itself, either as pigment distribution within an individual beard hair or as different colored hairs in different parts of the beard. Thus, beard color is a collection of many separate traits, the inheritance patterns of which are not well understood. Only a few simple statements can be made with certainty.

The relationship of pink to yellow carotene applies to the beard as well as to the rest of the petal. Any color pattern which has yellow carotene pigments in the beards can be represented by pink pigment instead of the yellow. Since this also applies b the concentration of pigment, it is easy to understand why tangerine pink beards are often more intensely colored than the petal of the same flower: in most yellow flowers the beard has a higher concentration of pigment and hence a darker, more intense color than the petal. Similarly, pink ("red") beards on while flowers or on blue flowers are the genetic equivalent of the yellow bearded white and the yellow bearded blue with the substitution of pink lycopene for the yellow carotene.

No conclusions can be reached on the inheritance of presence/absence of carotenoid pigments in the beard. In some cases there appear to be all-or-none differences; in others there seem to be only concentration differences with all intermediates present.

Blue beards are a more difficult subject still. They exist in many different types of irises, and their genetics may be simple. Certainly in the standard dwarf bearded hybrids and in some arilbreds in which the blue beard has been introduced through a genome other than that of the tall bearded, there is a major element of dominance of the pattern over the colors in the "pore" tall beardeds. That does not necessarily imply a single locus, however. In the tall bearded hybrids in which the blue beard cannot be directly attributed to recent dwarf ancestry, the trait has resulted from strong selection and planned breeding programs. It seems likely that several major loci, as well as modifiers, may be involved.

VII. GENETICS OF OTHER SPECIFIC TRAITS

PURPLE LEAF BASES

Purple color at the base of the leaves occurs in many varieties of tall bearded irises and the amphidiploid arilbreds and standard dwarfs. It appears to be inherited as a dominant trait, as illustrated by a cross of Moonlit Sea (green bases) x Matterhorn (purple bases). Eight seedlings had green bases and eight had purple leaf bases, indicating that Matterhorn has the genotype Pb pb pb pb. Variation in the intensity of pigment among varieties, and even seasonally in the same variety, makes this difficult to study in some crosses.

SELF-STERILITY ALLELES

Many types of plants have ways to present self-fertilization. Some irises, and particularly the bearded irises, appear to have a system of self-sterility (S) alleles: a pollen tube's growth is inhibited if it contains an S allele also present in the tissue of the style. This mechanism normally prevents self-fertilization and will also prevent cross-fertilization between closely related clones if they have the same S alleles. In all cases, the plants are normally fertile, as male and female, with clones having different S alleles. In a random sample of varieties many different S alleles would be found and unrelated clones having identical S genotypes would be rare.

In the complex garden hybrids, these mechanisms can break down. However, the examples of successful self-fertilizations are noteworthy because self-pollinations only occasionally set seed.

HORNS, SPOONS AND FLOUNCES

Most irises showing any of these controversial projections are descended from seedlings raised by Lloyd Austin. Austin (1961) explained the origin of his line as coming from two seedlings raised by Sydney Mitchell; one was later introduced as Advance Guard (Midwest x San Francisco or sib). At least two other times horns have arisen independently (chap. 20), also descended from Mitchell seedlings and Sass varieties. Henry Sass has stated that horns had been noticed in the Sass seedling fields in the 1930s. The Sass line may thus be the origin of the genes for horns.

Austin described the inheritance of the projections as "partially dominant" because in crosses of a horned clone by a nonhorned variety some of the seedlings were horned, but only between 10 percent and a maximum of 30 percent, less than expected for a completely dominant trait. However, other evidence suggests a recessive pattern of inheritance; the trail originated from inbreeding at least three separate times and homed seedlings have been documented as having arisen from intercrossing of nonhorned seedlings descended from homed varieties though themselves nonhorned.

The genetics of the different manifestations of the projection—as a horn, a spoon, or a flounce—is also unclear. Intercrossing varieties with only horns can produce some seedlings with no projections, as well as other seedlings with the larger projections, including flounces. A simple explanation for these different manifestations would be a dosage effect. Thus, homed varieties would have al most two copies of the allele and, when intercrossed, could segregate out some plants with one or no copies of the homed allele. The intercross would also segregate plants with three and four copies of the allele. These would be the spooned and flounced seedlings. This dosage hypothesis predicts that intercrossing two flounced varieties would give almost all flounced seedlings. Adequate data of this type do not seem to be available. An alternative explanation for the different projections would rely on other loci modifying the projection. The modifier hypothesis predicts that one locus controls presence or absence of the projection and a second locus controls the form of the projections, if present.

Both of these hypotheses seem compatible with available data. For instance, Austin (1961) obtained two flounced varieties from a cross of Mulberry Snow (homed) with Happy Birthday (nonhorned). On the dosage hypothesis, Happy Birthday would have to carry at least one copy of the homed allele which would be unexpressed because of only "partial dominance" with a single dose present. According to the modifier hypothesis, Happy Birthday would be required to carry modifiers producing flounces, not expressed in happy Birthday because no "homed" allele is present. Which of these hypotheses is correct remains to be tested by further crosses. It is clear, however, that at most a very few loci have the major effect in the expression of this trait.

BLOOM SEASON AND REBLOOMING

The time at which an iris blooms is determined by an unknown physiological regulation which responds to environmental cues such as temperature and photoperiod Within the normal spring season variation among varieties for bloom time is under genetic control. Most seedlings from two very early blooming parents will also bloom early in the season; most seedlings from late blooming parents will also be late blooming.

While different types of irises have distinct bloom periods—the dwarfs and arilbreds before the tall bearded, for example—in general, the variation in timing appears to be continuous with no discrete differences segregating in crosses. That continuum implies that many loci, each of small effect, are involved, but does not prove it. Whatever its nature, the genetic variation exists and can be used in breeding programs.

Genetic variation, of an equally obscure nature, also exists for reblooming. Among tait bearded irises, reblooming takes many forms—summer, fall, or winter rebloom, and even everblooming. The different types appear to be genetically related, probably as variations due to modifiers of the basic reblooming tendency. This is important for breeders in cold climates because varieties that only rebloom in milder climates may still produce good cold climate rebloomers. The inheritance patterns of reblooming are very similar to those of the horns, spoons, and flounces: they are neither dominant nor recessive, and are not obviously due to any simple genetic system.

If a reblooming tendency is merely a disruption of the normal physiological regulation of bloom time, several genotvpically unrelated types of reblooming might exist and show complementation. However, there is evidence that reblooming tendencies in different groups of bearded irises may be genetically similar. One example is the case of the Sass intermediates (chap. 10). The "Pumila hybrids" (amphidiploid SDB hybrids in today's terminology) raised by the Sasses sometimes bloomed on and off during the summer. They marked those that did rebloom and crossed them with Autumn King and other reblooming tall beardeds. From these crosses they obtained their reblooming intermediates.

OTHER CHARACTERISTICS

Fragrance is not generally considered a positive attribute of irises, though a few varieties do have pleasant fragrances. Some irises have extremely unpleasant smells; still others have no apparent fragrance at all. Because fragrances are specific biochemical substances, their inheritance should be easy to study. However, virtually nothing in the was' of genetics of these substances has been done. It is not even known whether scent is inherited in an essentially dominant or essentially recessive pattern, though biochemical considerations suggest scent would be a dominant with respect to no scent.

Branching shows great variation among tall bearded irises because of the quite different types of stalks in the ancestral species. There are no known single genes having a major effect on branching. Rather, many loci appear involved. Rough general rules for the genetics of a trait that is polygenic are (1) offspring are roughly intermediate between the parents, (2) the better the quality of the parents, the better the offspring are likely to be, (3) parents far from the norm will have offspring less extreme. That third point is especially important since good branching is an extreme expression and hence will be difficult to maintain even using two good parents. Of course, in all cases, variations will occur among the sister seedlings, so some will be better than others and a few may even be better than the parents.

Bud count similarly shows great variation. Besides being influenced by branching, it is a function of the number of buds on each branch. It is probably controlled by many kid independent of branching.

Substance in tall bearded irises has been improved greatly n the last few generations of breeding and selection. It is a complex trait probably related to strength of each cell wall, number and arrangement of the cells in the petal, turgor in each cell, etc., and is best treated as a polygenic trait. Other types of irises probably also have a multiple gene control of substance.

"Lace" at the petal margins has not been well studied genetically. It is neither a dominant trait (since two parents with no lace can produce a laced seedling) nor a recessive (since two laced parents can produce a nonlaced seedling). A dosage effect and or modifiers are possible explanations.

Beard width in most irises is reasonably narrow with multicellular hairs arising primarily along the midrib of the fall petal. Oncocyclus irises and many arilbreds have wide diffuse beards with hairs arising from the petal surfaces instead of or as well as from the midribs. In arilbred hybrids this trait does segregate and is probably due to very few loci. However, not enough crosses have been studied to define its inheritance. Because the wide beard is determined by genes solely from the oncocyclus ancestors and it occurs in arilbreds, "wide beards" is dominant to the "narrow beards" characteristic of tall bearded irises.