Proc. Natl. Acad. Sci. USA Vol. 92, pp. 11331-11338, December 1995 Review Uniparental inheritance of mitochondrial and genes: Mechanisms and evolution (organelles/non-/maternal inheritance) C. William Birky, Jr. Department of Molecular Genetics, Ohio State University, 484 West Twelfth Avenue, Columbus, OH 43210

ABSTRACT In nearly all eukaryotes, Left with the example of maize, for which that die before the colony is examined. at least some individuals inherit mito- more progeny have been analyzed (com- This inheritance pattern is similar to that chondrial and chloroplast genes from piled in ref. 4). Crosses of green by ofPelargonium if one looks at single mark- only one parent. There is no single mech- mutant males (variegated plants with mu- ers in clones, except that the ho- anism of uniparental inheritance: or- tant white germ-line cells supported by moplasmic cells in a clone are individuals ganelle gene inheritance is blocked by a sectors of green tissue) produce only rather than differentiated cells within an variety of mechanisms and at different green progeny; the reciprocal cross pro- individual. stages of reproduction in different spe- duces only mutant . Baur found a Another way of looking at the variation cies. Frequent changes in the pattern of different pattern in the geranium (Pelar- in patterns of uniparental vs. biparental organelle gene inheritance during evolu- gonium zonale; Fig. 1 Center): some off- inheritance is to plot frequency distribu- tion suggest that it is subject to varying spring inherited chloroplast genes from tions of the frequencies of from the selective pressures. Organelle genes often the parent only; others, from both two parents in the progeny of different fail to recombine even when inherited parents; and still others, from the male . Many different frequency distri- biparentally; consequently, their inheri- parent only. The reciprocal cross also gives butions have been observed in different tance is asexual. is a mixture of the three different kinds of species and often in different crosses of apparently less important for genes in progeny but in different proportions. The the same species (e.g., refs. 5 and 6). The organelles than for nuclear genes, prob- plants that inherit from both different patterns can be classified accord- ably becadse there are fewer of them. As a parents are variegated with green and ing to whether they contain uniparental or result organelle can be lost because of white clonal sectors. Although the fertil- biparental zygotes, or both, as in Fig. 2. selection for special reproductive features ized (zygotes) are heteroplasmic, Pattern I (Um) is the same as Fig. 1 Left. such as oogamy or because uniparental containing plastids of both genotypes, Pattern II (UmB) is seen in Oenothera inheritance reduces the spread of cyto- these segregate rapidly during vegetative crosses in which most progeny inherit plasmic parasites and selfish organelle cell divisions. Consequently, the mature chloroplast genes from the female parent DNA. plant consists of clonal sectors of ho- only, while a few inherit them from both moplasmic mutant and wild-type cells. parents. Pattern III (UBU) can be subdi- Genes in chloroplasts were first detected The Mirabilis and Pelargonium inheri- vided according to the relative numbers of by Baur (1) and Correns (2) in 1909 tance patterns are often called maternal genomes contributed by the two parents. because their inheritance departed from and biparental, respectively, but this ter- Different crosses involving chloroplast the Mendelian rules. Much later, mito- minology is not generally applicable. genes in Pelargonium or mitochondrial chondrial genes were identified in the Looking at individual progeny, one sees genes in yeast show patterns Illa-IlIc, same way. The non-Mendelian inheri- that Pelargonium crosses produce a mix- depending on the nuclear and organelle tance of organelle genes became manifest ture of maternal, biparental, and paternal genotypes of the parents. Illd (UU) is in two ways: the rapid segregation of zygotes. Additional terminological prob- seen in the blue mussel (Mytilus edulis), in alleles during vegetative (mitotic) repro- lems appear in microorganisms that do which there are two separate uniparental duction and inheritance from one parent not have differentiated male and female lineages, one transmitted via females and only. Vegetative segregation of chloro- (i.e., are isogamous). Fig. 1 Right the other via males (7, 8). Patterns IV plast genes is a consequence of random- diagrams the inheritance of mitochondrial (BUp) and V (Up) are seen in gymno- ness of replication and partitioning of genes in yeast (3). Haploid cells of mating . organelles and organelle DNA molecules types a and a carry different alleles of a In many cases sample sizes are insuffi- at cell division (3). Uniparental inheri- mitochondrial gene (e.g., conferring resis- cient to distinguish between patterns (e.g., tance proved to be even more complex; tance and sensitivity to an antibiotic). between I and II or III). Moreover, the the variety of molecular and cellular Haploid cells fuse to form zygotes that are stochastic processes that can cause unipa- mechanisms found in different organisms heterozygous a/a and heteroplasmic antrl rental inheritance are probably always op- is matched only by the variety of hypoth- ants. During subsequent mitotic divisions, erating. Consequently, strictly uniparental eses devised to explain the evolution of the the mitochondrial alleles segregate so that inheritance is probably not as common as phenomenon. virtually all daughter cells are homoplas- is generally believed. The inheritance of mic after about 10 cell divisions. In addi- mtDNA in interspecific crosses of mice Genetics tion, some clones produced by individual was believed to be strictly uniparental (9) zygotes contain only antr or only ants cells. until a more sensitive technique (PCR There Are Many Different Patterns of This is also a consequence of relaxed amplification) was used to detect low lev- Uniparental Inheritance. Correns found replication and partitioning of mtDNA: in els of paternal mtDNA (10). These pat- that chloroplasts are inherited only from some zygotes, only mitochondrial ge- the female parent in the four-o'clock nomes from one parent are replicated and Abbreviations: UPI, uniparental inheritance; (Mirabilis jalapa). Strictly maternal inher- partitioned into buds, while those from BPI, biparental inheritance; mtDNA, mito- itance is shown diagrammatically in Fig. 1 the other parent remain in mother cells chondrial DNA. 11331 Downloaded by guest on September 25, 2021 11332 Review: Birky Proc. Natl. Acad. Sci. USA 92 (1995) fixed by stochastic processes in a fraction -4AP~ ~~ p/(m + p) of cells in the progeny, where p and m are the numbers of organelle genomes in the male and female , 0 0 a erys a eryr respectively. The remaining cells will have (iB) no copies of the paternal . In the \ls/ mouse, p/m is estimated to be 1-4 x 10-5 (10). In extreme cases, organelles may be ± @ NS completely excluded from the gametes of one sex by unequal cytokinesis (Table 1, mechanism b); an example is the of the crayfish (23). Organelles can be de- stroyed in the gametes (Table 1, mecha- nism c), as they are in the isogametes of FIG. 1. Examples of uniparental inheritance in maize (Left), Pelargonium (Middle), and the filamentous green alga Temnogyra. In Saccharomyces cerevisiae (Right). another green alga, Bryopsis, organelle terns are for single markers; if recombi- each biparental zygote whenever possible DNA disappears during the differentia- nation occurs, some zygotes classified as and publish the observed frequency dis- tion of male gametes (Table 1, mechanism uniparental for an allele from one parent tribution(s). d). may contain alleles of another locus from There Are Many Different Mechanisms Other mechanisms eliminate organelles is fre- during fertilization. In some organisms the other parent. Recombination of Uniparental Inheritance. The underly- (e.g., the tunicate Ascidia), sperm or- quent in the chloroplasts of Chlamydomo- ing mechanisms of uniparental inheri- ganelles fail to enter the (Table 1, nas (11) and the mitochondria of the fungi tance are as diverse as the patterns (Table mechanism e). Only nuclei are exchanged Saccharomyces (12, 13), Schizosaccharo- 1). The transmission of organelle genes to during sexual reproduction of other or- myces (14), and Aspergillus (15). Recom- offspring can be blocked at any step in the ganisms (Table 1, mechanism f), as during binant restriction fragment-length poly- reproductive process. conjugation of ciliated protozoa. Mecha- morphism (RFLP) patterns have been Prezygotic mechanisms eliminate or- nisms b-f in Table 1 are called monoga- found in some matings of the slime mold ganelles or organelle genomes during ga- metic transmission (13). Physarum, but it is not clear whether these metogenesis. In some organisms, the mei- Some zygotic mechanisms eliminate or- are due to recombination between the two otic divisions produce gametes that are ganelles from the by unvarying parental mitochondrial genomes, in- morphologically identical (). In (deterministic) processes. A striking ex- tragenolbic rearrangements, or plasmid others, unequal cell divisions or differen- ample is the degradation of the chloro- integration (16). In contrast, there are a tial growth or both result in large maternal plast from the male isogamete in the fil- number of cases in which significant num- gametes and small paternal gametes amentous algae Spirogyra and Zygnema bers of biparental zygotes are produced, ( or oogamy). The common but (Table 1, mechanism g). Selective silenc- but the organelle genomes from the two not invariant consequence is an input bias ing (Table 1, mechanism h; ref. 24) is a parents do not recombine. In plants, of organelle genomes in the fertilized egg process in which organelle genomes from screens of progeny from sexual crosses that favors the female parent (Table 1, one parent are selectively degraded in the detected no progeny that were recombi- mechanism a). Alleles of an organelle zygote. The classic example is the enzy- nant for chloroplast genes (refs. 17 and gene from the paternal parent will be matic degradation of chloroplast DNA 18). Two recombinant plastid genomes difficult to detect because they will be contributed by the mt- from were recovered by stringent selection in plants reared from a very large number of fused somatic cells (19, 20), showing that chloroplast genes can recombine; the ex- 4 treme scarcity of detectable recombina- 100 t x ?IL tion in crosses is probably due to a very low I I frequency of chloroplast fusion. No re- 10 , I _l combinants of the male and female mito- chondrial lineages in blue mussel have 8 been found, even though both genomes 6ioj IVIIIII have been present in the fertilized egg and 4110o |U UmB UBU BUp | Up 0 2 germ line cells of embryos in every gen- be eration (7, 8) for over five million years 0 15 35 5575 95 0 15 35 55 75 95 0 15 35 55 75 95 0 15 35 55 75 95 0 15 35 55 75 100 (21, 22). The absence of recombination in a) 100 many organisms means that the inheri- I.. 80- tance of organelle genes is effectively 60- Ila Ilib IIIc IIId asexual in those cases, even when it is 40- UmBU UBU UBUp l UU biparental. The division of patterns of inheritance 20 013555 S 9 015 35 5575 95I0i 15 35 5575 95 0 15 5S5 10 into these classes is somewhat artificial Percent transmission because different species, or even differ- ent crosses in some species, produce from FIG. 2. Frequency distributions. The organisms and organelles are, left to right: maize to 100% zygotes. Conse- chloroplasts or mitochondria (I), Chlamydomonas reinhardtii chloroplasts (II), geranium chlo- 0% uniparental roplasts (III), Chlamydomonas mitochondria (IV), and sequoia chloroplasts or mitochondria (V). quently it would be more appropriate for The lefthand parent is the female sex (maize, geranium, sequoia) or mt+ mating type many purposes to treat uniparental inher- (Chlamydomonas). Patterns I-V contain various combinations of uniparental (U), uniparental itance as a quantitative trait. If this is to be maternal (Um or U+), uniparental paternal (Up or U-), and biparental (B) zygotes; thus, UBU done, geneticists will have to estimate the is roughly equal numbers of the two types of uniparentals, plus some biparentals. The diagrams frequency of markers from each parent in are idealized and do not precisely represent observed distributions. Downloaded by guest on September 25, 2021 Review: Birky Proc. Natl. Acad. Sci. USA 92 (1995) 11333

Table 1. Mechanisms of uniparental inheritance (30). Moreover, mitochondria and chloro- plasts can be preferentially transmitted Mechanisms Examples from different parents: male and female, Prezygotic (anisogamy/oogamy) respectively, in some gymnosperms (31, a. Unequal cell divisions and differential growth 32); and mating types minus and plus, produce large egg and small sperm Mouse Mus M (10) respectively, in C. reinhardtii (33). b. Exclusion of organelles from gametes during partitioning Crayfish M (23) Evolutionary History c. Degradation of organelles in gamete Green alga Temnogyra C (93) d. Degradation of organelle DNA in gamete Bryopsis M,C (94, 95) Most Organisms Show Some Degree of Fertilization Uniparental Inheritance of Mitochondria e. Exclusion of organelles of one parent from and Chloroplasts. The inheritance of or- zygote Tunicate Ascidia M (70) ganelle genes has been studied in a large f. No organelles exchanged Ciliate Paramecium M (56) number of flowering plants, one green Zygotic/deterministic alga, a few and in a few animals, g. Selective silencing/degradation of organelle Green alga Spirogyra C (96) fungi, oomycetes, mycetozoa, ciliated pro- h. Selective silencing/degradation of organelle tozoa, and plasmodia. Nearly every spe- DNA Chlamydomonas C,M (33) cies produces a substantial proportion of i. Partitioning of parental organelles into uniparental zygotes. Exceptions are found separate cells Cylindrocystis C (96) in the yeasts S. cerevisiae and Schizosac- j. Exclusion of organelles from embryonic tissue Gymnosperm Larix C,M (31) charomyces pombe in which some crosses Zygotic/stochastic produce less than 10% biparental zygotes k. Exclusion of organelles from embryonic Angiosperm Pelargonium C (5). Only in some ascomycete fungi do the tissue (25) progeny of all zygotes receive organelle 1. Random replication only Yeast Saccharomyces M (3) genes from both parents. Heterokaryons M, mitochondria; C, chloroplast. formed by hyphal fusions in these fungi are analogous to zygotes in that some of Chlamydomonas zygotes. In another first and second cell divisions, leaving the the parental nuclei can fuse and then green alga, Cylindrocystis, the zygote con- embryo with chloroplasts from only one undergo meiosis to produce haploid asco- tains two chloroplasts from each parent. parent or the other (6). Random replica- spores. In Aspergillus nidulans, mitochon- However, these chloroplasts do not fuse or tion of organelle genomes may also play a drial genomes from both parents persist in divide, aad the four products of meiosis role in uniparental inheritance. Organelle all heterokaryons and recombine before each receive one chloroplast and, hence, genomes within a cell or organelle are they segregate into homokaryotic sectors chloroplast genes from only one parent or chosen randomly, with respect to geno- (15, 34). This avenue of sexual reproduc- the other (Table 1, mechanism i). Multi- type or origin, for replication. Within a tion might be viewed as the only exception cellular animals and plants have addi- zygote (or zygote clone), genomes from to the rule that all organisms produce tional options because early divisions of one parent may be replicated by chance some uniparental zygotes. But the mito- the zygote separate embryonic and ex- more often than those from the other chondrial genes are inherited maternally, traembryonic cells, so organelles from one parent, or some genomes may be de- through the cleistothecial parent, in sexual parent can be eliminated by being parti- graded (turnover). These processes prob- crosses between heterokaryon-incompat- tioned into the extraembryonic cells (Ta- ably play a major role in uniparental in- ible strains in which nuclei migrate only a ble 1, mechanism j). For example, in the heritance in yeast. They are all conse- short distance from each mycelium into fertilized egg of the gymnosperm Larix, quences of the fact that organelle genomes the other (35). future embryonic cytoplasm is segregated are relaxed, with no reinitiation block that Although some degree of uniparental into a special region that includes only would prevent genomes from replicating inheritance appears to be a nearly univer- paternal plastids, while a majority of mi- more than once per cell division. In con- sal phenomenon, it must be remembered tochondria are maternal. The localization trast, eukaryotic nuclear genomes are that only a few species have been studied of maternal plastids in alfalfa is evidently stringent, having cis-acting blocks to reini- in most major groups of organisms, and under genetic control: high-transmitting tiation and consequently lacking the sto- only chloroplast inheritance in flowering females localize all plastids in the apical chastic processes that can cause uniparen- plants has been studied in a reasonably part of the unfertilized egg so that all of tal inheritance (3). large and diverse number of species. More- them enter the embryo after the first cell Some organisms reduce the contribu- over, there are major groups in which or- division, whereas some maternal plastids tion of one parent at more than one stage. ganelle gene inheritance has never been are localized in the basal part of the egg in In animals, for example, the sperm con- studied. These include the nonflowering low-transmitting females and are parti- tribute very few mtDNA molecules to the plants; golden-brown, brown, and yellow- tioned into the extraembryonic suspensor zygote, and random replication probably ; dinoflagellates; and most in- (30). reduces this contribution to zero in most vertebrate animals. In some of these Random or stochastic processes can individuals. Moreover, sperm mitochon- groups there is cytological evidence for also eliminate organelles from the zygote dria are degraded in the fertilized eggs of uniparental inheritance (36). However, or embryo (3). Organelles from different rodents (26-28). In the honeybee, pater- cytological data are usually not as strong parents may segregate into embryonic and nal mtDNA constitutes about one-fourth as genetic data for determining the mode extraembryonic cells by chance (Table 1, of the total mtDNA in newly laid eggs of inheritance of organelle genes: light mechanism k). This is seen in green x because of polyspermy, but the paternal microscopy may lack the resolution to white crosses in Pelargonium, in which mtDNA is degraded or replicates slowly or identify proplastids and mitochondria in some very early embryos contain only not at all and is undetectable in larvae gametes or distinguish between those white plastids, while the extraembryonic (29). from different parents in the egg, the loss suspensor cells contain green plastids The diversity of mechanisms of unipa- of fluorescent staining with DAPI or an- (25). This suggests that sometimes all the rental inheritance is further demonstrated tibodies may not distinguish between loss plastids from one parent are partitioned, by the fact that most of these mechanisms of DNA and its dispersal, and electron by chance, into extraembryonic cells at the can be found in one taxon, the seed plants microscopy is limited to very small sam- Downloaded by guest on September 25, 2021 11334 Review: Birky Proc. Natl. Acad. Sci. USA 92 (1995) ples of gametes or zygotes. Moreover, parsimonious reconstructions of the mi- ther way, the organelle genomes are ef- cytological evidence is one-sided: the ab- tochondrial data all have strictly unipa- fectively asexual genetic systems in sexu- sence of organelles from gametes or their rental (maternal) inheritance at the root ally reproducing organisms. Why do the destruction in zygotes is evidence for uni- of that tree. However, the chloroplast and advantages of sexual reproduction out- parental inheritance, but the presence of mitochondrial trees both show a great deal weigh those of asexual reproduction for organelles from both parents at one stage of homoplasy, consisting of numerous par- nuclear genes in so many eukaryotes but does not demonstrate biparental inheri- allel evolutionary changes in the terminal not for the organelle genes? To approach tance because transmission may be branches. Unless these frequent changes this question, it is convenient and biolog- blocked at a later stage. For these reasons, in pattern began only recently, they will ically reasonable to consider an organism cytological data on the mode of inheri- confound the parsimony analysis. Thus, with at least occasional sexual reproduc- tance cannot be combined with genetic the available data do not strongly support tion during which organelle genes are data. any ancestral pattern. The evolution of inherited biparentally and recombine in The Evolutionary History of Uniparen- different cytological mechanisms of uni- most zygotes (random replication will al- tal Inheritance Shows Frequent Reversals parental inheritance has not yet been sub- ways produce some uniparental zygotes). and Parallel Changes. How were the ear- jected to phylogenetic analysis. However, One can then compare the consequences liest mitochondria and chloroplasts inher- most of the mechanisms described above of losing sexual reproduction entirely, ited? One approach to this question is to have been observed in studies of flowering which affects both nuclear and organelle deduce the answer from what we know or plant chloroplasts (39), and at least five genes, to those of losing biparental inher- suspect about the organisms and their different mechanisms are operating in an- itance or recombination of organelle endosymbionts. Oogamy probably did not imals (7, 8, 23, 28, 29, 36, 70, 71). It is very genes alone. The literature on the evolu- arise until well after the a-purple bacterial likely that mechanisms, as well as patterns, tionary consequences of asexual repro- ancestor of mitochondria was ingested by of uniparental inheritance have changed duction is extensive; the following treat- a eukaryotic cell with no cell wall, and frequently during evolution. This obser- ment relies heavily on the reviews by sexual reproduction of the host was prob- vation is highly significant for evaluating Kondrashov (72), Brooks (73), and others ably initiated by fusion of whole undiffer- evolutionary explanations of uniparental in the same volumes. entiated cells. The resulting zygotes con- inheritance. The Production of Some Uniparental tained all of the endosymbionts from both Evolutionary Explanations. Most ani- Zygotes by Random Replication Is Un- cells. Initially the endosymbionts from the mals and plants reproduce sexually at least avoidable. We saw in the preceding sec- two gametes were indistinguishable, so part of the time, and this is also true of tion that the symbiotic ancestors of or- that there could be no mechanism to many other eukaryotes. Natural selection ganelles were probably inherited biparen- preferentially eliminate one or the other. presumably favors biparental inheritance tally, with few or no uniparental zygotes. The host was probably unicellular, and so and recombination of nuclear genes over As they became integrated into the cell as there could be no uniparental inheritance strictly asexual reproduction in these or- symbionts, their genomes lost stringent due to random partitioning of symbionts ganisms. In contrast, most of these same replication and partitioning. The produc- between embryonic and extraembryonic organisms have reduced the proportion of tion of some uniparental zygotes by ran- tissue. Uniparental zygotes would only be biparental zygotes further, while many dom replication is therefore probably the produced if symbionts from one parent have eliminated them altogether, and primitive state. Moreover, it probably per- replicated much more often than those those with biparental zygotes often have sists today in the absence of other mech- from the other parent or were destroyed Ei- anisms of inheritance in by chance; most zygotes were probably no recombination of organelle genes. uniparental (as biparental. Consequently, the symbionts were probably inherited biparentally. However, it is unlikely that the endosym- biont genomes recombined because any mechanisms that the endosymbionts might have used to exchange genes while they were free-living would not work in the very different environment inside the host cytoplasm. This scenario may also apply to the cells that ingested the ances- tors of chloroplasts. However, it is also possible that those cells were oogamous or anisogamous, with differentiated cells fus- ing to initiate sex, in which case the sym- bionts may have been inherited uniparen- tally from the beginning. Another approach is to use phyloge- netic analysis to infer the ancestral mode of inheritance from the inheritance pat- terns of organelles today. Figs. 3 and 4 show the most parsimonious reconstruc- tions of the evolution of patterns of uni- parental inheritance in chloroplasts and mitochondria. The pattern of chloroplast inheritance in the ancestor of all of FIG. 3. Phylogenetic tree of inheritance patterns in chloroplasts. The branching pattern of the gene tree is based on rbcL sequences and on the pattern indicated in ref. 37. Each extant taxon was the species in the chloroplast tree is equiv- assigned its observed pattern of chloroplast gene inheritance (based on genetic data only), and ocal because different trees with the same, then the ancestral nodes were assigned patterns by the parsimony algorithm of MacClade (38). minimal number of changes have different Species such as Pelargonium show two or three different patterns but are assigned only one. patterns (Um, UmB, UBU, or Up; see Fig. Interspecific crosses that result in low viability are not included. Genetic data and references are 1 for definitions) at the root. The most in refs. 4, 36, 39-41 with additional data from refs. 42-45. Downloaded by guest on September 25, 2021 Review: Birky Proc. Natl. Acad. Sci. USA 92 (1995) 11335 rounds of replication, which reduces the requirement for repair (58). On the other hand, the loss of organelle fusion would limit the number of genomes available for recombination repair, which might have a small detrimental effect. (iv) Repair of demethylations may occur in sexual dip- loids; but most organelle genomes are unmethylated, or methylation does not affect gene function. (v) The replacement of sexual reproduction by female parthe- nogenesis confers as much as a 2-fold selective advantage in many oogamous organisms by eliminating the cost of pro- ducing males; but the loss of biparental inheritance or recombination of or- ganelles does not confer this advantage because it does not eliminate males. (vi) Sexual reproduction is slower than asexual because of the time required to find mates; but this does not differ for biparental and uniparental inheritance of organelles. (vii) Sexual reproduction has the disadvantage that it can replace more fit heterozygous genotypes by less fit homozygotes; but this is irrelevant to organelles because or- ganelle genomes are rarely heteroplasmic. FIG. 4. Phylogenetic tree of inheritance patterns in mitochondria. The branching pattern is Asexual Reproduction and Uniparental based mainly on 18S rRNA gene sequences (46-48), rbcL sequences (37), and the traditional Inheritance Inhibit the Spread of Cyto- phylogeny of the animals. Ancestral states were reconstructed as in Fig. 3. "Yeasts" = S. cerevisiae plasmic Parasites. Cytoplasmic endosym- and Sch. pombe. Species such as Saccharomyces show two or three different patterns but are assigned only one. The low level of paternal transmission detected in interspecific crosses of bionts are common in many organisms, Drosophila..and Mus is not included, since intraspecific crosses show strictly maternal inheritance and these can be detrimental to their host. (27, 49). Genetics references are as follows: for mycetozoa, refs. 16 and 50-53; for plasmodia, refs. In a sexually reproducing organism in 54 and 55; for Paramecium, ref. 56; for Pythium, ref. 57; for Chlamydomonas, ref. 58; for plants, which the zygote receives cytoplasm from refs. 59 and 60; for fungi, refs. 3, 35, and 61-67; and for animals, refs. 7, 8, 29, 68 and 69. both parents, a cytoplasmic symbiont present in only one of the two parents can yeast) or is superimposed on them, be- which have small sperm, cannot avoid be inherited by all of the offspring and cause it is biologically and evolutionarily producing some uniparental zygotes by spread rapidly in the population when it difficult to reverse. There are only two chance. The production of uniparental replicates faster than the host. But when ways to avoid producing at least a few zygotes by random replication and parti- the zygote inherits cytoplasm from only uniparental zygotes by random replica- tioning is pervasive, probably because of one parent, the spread of detrimental tion. One is to impose stringent replica- the historical accident that it is the prim- cytoplasmic parasites will be limited to the tion and partitioning on both the or- itive state. But most organisms have ad- cytoplasmic descendants of the cell it orig- ganelles and organelle genomes. This ditional mechanisms that result in even inally invaded. Thus, uniparental inheri- would require the acquisition of (i) a fewer biparental zygotes than expected tance could result from selection for mu- cis-acting mechanism to prevent replica- from random replication and partitioning tations that reduce or eliminate the cyto- tion origins from firing more than once in alone; the ubiquity of these mechanisms plasmic contribution from one parent a cell cycle and (ii) a mechanism for must be explained in terms of natural (74). This model has been given a rigorous distinguishing the sister DNA molecules selection. mathematical framework (75), which un- produced by replication and moving them Some Advantages and Disadvantages of fortunately did not include random drift. to opposite sides of the future plane of Sexual Reproduction Do Not Apply to Grun (76) proposed that uniparental division of the cell and organelle. If there Organelle Genes. Among these are the inheritance would be advantageous be- were more than one organelle per cell, following examples. (i) The progeny of cause it reduces the spread of selfish or- there would also have to be mechanisms to sexual reproduction have two parents to ganelle genomes that are detrimental to ensure that each organelle divides once care for them; but organelle genes do not the organism. Examples of such genomes and one daughter organelle is partitioned control parental care, so the loss of bipa- are the suppressive petite mitochondrial to each daughter cell. This combination of rental inheritance or recombination of mutants in yeast and the senescence mu- evolutionary events has never been ob- organelle genes will not affect this some- tations in Neurospora. This advantage of served, and may be effectively impossible. times useful trait. (ii) Sexual reproduction uniparental inheritance would also apply Alternatively, the chance production of leads to the production of resistant spores to detrimental organelle genes as well as uniparental zygotes would be eliminated if that survive harsh conditions in many whole genomes and to detrimental plas- both gametes had such large volumes of organisms; but organelle genes do not mids residing in organelles. The hypothe- cytoplasm and large numbers of or- control spore formation. (iii) Recombina- sis was explored in detail by Hastings (77), ganelles and organelle genomes that the tion in sexual diploids may facilitate the who showed that a detrimental selfish probability of replicating only one be- repair of chromosome damage; but or- organelle genome that is inherited bipa- comes effectively zero (order of magni- ganelle genomes are always present in rentally can increase to an equilibrium tude of rates), as probably hap- many copies per cell, so biparental inher- frequency that significantly reduces the pens in mycelial fusions inAspergillus. But itance provides little additional benefit. population fitness. A mutant nuclear gene most unicellular organisms, which have Moreover, normal genomes can replace that causes uniparental inheritance can small cells, and most oogamous species, damaged genomes by undergoing more increase in frequency and, under some Downloaded by guest on September 25, 2021 11336 Review: Birky Proc. Natl. Acad. Sci. USA 92 (1995) circumstances, the entire population will been proposed. An asexual mutant gives same chromosome and counting genes on become uniparental. This theory required rise to a clone, essentially a new species, different nuclear chromosomes as un- group selection, but two models that in- that is more likely to retain detrimental linked from each other and from organelle voke only individual selection have also and to lose advantageous mu- genes. The complete loss of recombina- been studied (78). tations than are related sexual species. It tion among organelle genes reduces this There are several reasons why limiting is believed to have a higher probability of average recombination frequency by <1% the spread of selfish symbionts or or- extinction and a reduced ability to form (unpublished data). Computer simula- ganelle DNA cannot be a general expla- new species. Species-level selection is fa- tions of directional selection with varying nation of uniparental inheritance. First, a vored by the observation that asexual lin- levels of recombination (e.g., ref. 81) sug- mutation causing uniparental inheritance eages of animals and plants usually rep- gest that a 1% change in recombination is advantageous only when a detrimental resent races of otherwise sexual species, or will have a negligible effect on the elimi- symbiont or organelle DNA is present. species within genera that also contain nation of detrimental mutations and the Second, the detrimental genes will accu- sexual species, or whole genera, but al- retention of advantageous mutations. mulate while the allele is being fixed, most never whole families or higher-order Moreover, a 1% change in recombination leading to very low fitness and possibly to taxa (79). Molecular data show recent is much less than the intraspecific varia- extinction. Third, the symbiont hypothesis origins for the few asexual animals that tion in nuclear recombination rates: chro- does not explain the many cases of unipa- have been investigated (reviewed in ref. mosomes from different Drosophila rental inheritance in organisms with a 68), except the bdelloid rotifers (Matthew strains vary in recombination frequency substantial contribution of cytoplasm Meselson and David Mark Welch, per- by 13-14% (73). For Drosophila, the ef- from both parents, including organisms sonal communication). In contrast, large fects of eliminating organelle recombina- such as pines and Chlamydomonas, where groups of organisms are characterized by tion altogether can be compensated by an chloroplasts and mitochondria are inher- the loss of biparental inheritance or re- increase in the recombination frequency ited from different parents. Finally, uni- combination, suggesting that it may have for nuclear genes by a factor of 4 x 10-7 parental inheritance may not be necessary little or no effect on rates of extinction and (unpublished data). It appears that elim- to inhibit the spread of selfish organelle speciation. The contrast is particularly inating biparental inheritance of or- DNA or symbionts. Failure of organelles striking in the vertebrates, in which par- ganelles or recombination within the or- to fuse, as in the case of plant chloroplasts, thenogenetic species are found singly or in ganelle genome would have a very small will prevent the spread of symbionts or small genera, while nearly perfect unipa- effect on the amount of linkage disequi- selfish organelle genes in organisms with rental inheritance of mitochondria is librium in a population, so long as the biparental inheritance (36). It will also found in all the sexual species. much larger nuclear genome is sexual and prevent the spread of selfish organelle There are two reasons why the loss of outcrossing with a high recombination fre- DNA molecules, if normal and selfish biparental inheritance or recombination quency. DNA mof&cules have to be in the same may not reduce the effectiveness of selec- Although the contribution of the or- chloroplast or to compete tion as much as the loss of sexual repro- ganelle genome to linkage disequilibrium for replication. duction. First, organelle genotypes, and may be negligible compared to that of the The Efficiency of Natural Selection consequently organelle sex, are largely nuclear genome, that contribution could May Be Only Slightly Reduced by Unipa- irrelevant for some kinds of natural selec- still potentially have serious consequences rental Inheritance. Most of the theories tion. For example, organelle genes do not because the organelle genes play essential about the evolutionary advantages or dis- contribute to a host organism's resistance roles. From time to time the organelle advantages of sexual reproduction focus to parasites, nor do organelle mutations lineages with the fewest detrimental mu- on the fact that biparental inheritance and enable a parasite to overcome host resis- tations will be lost by random drift. This recombination break down linkage dis- tance. Second, because the organelle ge- loss is irreversible in the absence of bipa- equilibria that arise as a result of random nome is much smaller than the nuclear rental inheritance and recombination, so drift, selection, environmental changes, or genome, its contribution to linkage dis- the fitness of the population gradually mutation. This facilitates directional or equilibria is so much smaller as to be declines. This phenomenon, called Mul- stabilizing natural selection under many nearly negligible. The amount of linkage ler's ratchet, will lead eventually to extinc- circumstances (although it can be detri- disequilibrium increases with the number tion (82) if it is unchecked. How have mental in some situations). This is because of polymorphic genes, and the effect of organisms with uniparental inheritance detrimental mutations may be linked to recombination on selection is larger (80). avoided this "meltdown"? First, the advantageous mutations, a situation The proportion of organelle genes that are ratchet will move slowly or even stop called negative or repulsion linkage dis- polymorphic is similar to or smaller than entirely if many mutations are extremely equilibrium. When this happens, selection that of nuclear genes because the muta- detrimental, as may be the case for animal against the detrimental allele at one locus tion rate is similar or smaller (except in mitochondrial genomes (82). Second, will tend to reduce the frequency of the primate mitochondria) and the effective when selection is soft (i.e., when the ac- advantageous allele at the other locus, and population size is smaller. But the abso- cumulation of mutations does not affect vice versa. Biparental inheritance and re- lute number of polymorphic genes is much the total population size), the ratchet combination will create chromosomes smaller because nuclear genomes have 100 moves but meltdown is delayed. Third, with two or more detrimental alleles to 1000 times as many genes as the or- organelle genomes are subject to intracel- linked to each other, and others with two ganelle genomes in the same organism. It lular and intercellular selection (83), or more advantageous alleles linked to is likely that the complete loss of biparen- which reduces the detrimental mutation each other. Then selection can reduce the tal inheritance (or recombination) of or- rate measured at the level of the organism frequency of the detrimental alleles and, ganelle genes will have the same effect on (84) (but intracellular selection is more independently, increase the frequency of selection as a very modest decrease in effective with biparental than uniparental the advantageous alleles. recombination frequency of nuclear inheritance; ref. 85). Fourth, the ratchet is This is the most general theory of the genes. Another way of looking at the slowed by some but not all forms of ep- evolutionary advantage of sex, applying to problem is to calculate the average recom- istasis (86, 87). These phenomena may all organisms except those that are exclu- bination frequency for all of the genes in slow the movement of the ratchet suffi- sively self-fertilizing. It is essentially a an organism by using known values of ciently for the organelle genomes to be group selection argument, although mod- recombination frequencies per base pair rescued by low levels of biparental inher- els invoking individual selection have also for nuclear and organelle genes on the itance and recombination, by environ- Downloaded by guest on September 25, 2021 Review: Birky Proc. Natl. Acad. Sci. USA 92 (1995) 11337 mental changes that increase their fitness, due to selection on the trait itself. Selec- need to be recast in the form of sets of or by compensating mutations (principally tion for uniparental inheritance might oc- mutually exclusive, exhaustive hypotheses in the nuclear genome because it codes for cur sporadically when an organism was that can be clearly distinguished by labo- most of the organelle proteins). invaded by a detrimental symbiont or ratory experiments or comparative analy- If the maintenance of biparental inher- when mutations produced selfish DNA or ses of natural experiments (92). Fourth, itance and recombination causes only a initiated a nucleocytoplasmic conflict. we need to identify genes that affect the small decrease in the effectiveness of nat- Uniparental lineages produced during transmission of organelle genes and de- ural selection and organisms have alter- these periods would eventually become termine how they act. Finally, here as native ways of escaping extinction by Mul- extinct because of Muller's ratchet or elsewhere in evolutionary biology, we ler's ratchet, then the amount of biparen- other effects of the loss of recombination need accurate measures of the important tal inheritance and recombination may be unless they succeeded in rescuing them- parameters that determine the evolution- determined mainly by other factors. One selves by reacquiring biparental inheri- ary consequences of uniparental and bi- is the presence of detrimental cytoplasmic tance and recombination. parental inheritance, such as recombina- parasites or selfish organelle DNA, as tion frequencies and rates and selection discussed above. Another is the evolution coefficients of mutations. of oogamy and of extraembryonic tissues, CONCLUSIONS both of which result in the production of Uniparental and biparental inheritance I thank Barbara Sears, Deborah Charles- substantial numbers of uniparental zy- are not simple alternative traits; organelle worth, and Brian Charlesworth for many help- gotes. Selective silencing in Chlamydomo- transmission is really a quantitative trait ful suggestions and comments on an earlier nas and some other organisms might have that is affected by many different molec- version of the manuscript. evolved as a mechanism for utilizing or- ular and cellular processes at all stages of ganelle DNA as a source of nucleotides sexual reproduction. No single mechanism 1. Baur, E. (1909) Zeit. Vererbungsl. 1, 330- during periods of starvation (58). It has explains all cases of uniparental inheri- 351. also been proposed that organelle genes tance, and no single evolutionary hypoth- 2. Correns, C. (1909) Zeit. Vererbungsl. 1, themselves might instigate uniparental in- esis can explain the great variation in 291-329. heritance and thereby enhance their own extent to which organelle genes are inher- 3. Birky, C. W., Jr. (1994) J. Hered. 85, 355- A ited from both parents. Some generaliza- 366. fitness. mutant organelle genome could 4. Kirk, J. T. 0. & Tilney-Bassett, R. A. E. increase in frequency by causing the deg- tions can be made: nearly all of the or- (1978) The Plastids (Elsevier/North-Hol- radation of organelle DNA from the op- ganisms that have been studied produce at land, Amsterdam). posite mating type, as was proposed to least some uniparental zygotes, and or- 5. Thrailkill, K M., Birky, C. W., Jr., Lucke- explain the origin of uniparental inheri- ganelle genes fail to recombine in the mann, G. & Wolf, K (1980) Genetics 96, tance by selective silencing in Chlamydo- biparental zygotes in many cases. Conse- 237-262. monas (88). Such mutants have been used quently, the benefits and costs of sexual 6. Tilney-Bassett, R. A. E. & Birky, C. W. J. as the starting point of models to explain reproduction are greatly reduced or elim- (1981) Theor. Appl. Genet. 60, 43-53. the evolution of two mating types as well inated in most organisms. The evolution- 7. Skibinski, D. 0. F., Gallagher, C. & Bey- as of uniparental inheritance (89, 90). ary history of organelle sex is full of non, C. M. (1994) Genetics 138, 801-809. It has also been shown that any differ- reversals and parallel changes. This sug- 8. Zouros, E., Ball, A. O., Saavedra, C. & ence between sexes or mating types that gests that it is not consistently strongly Freeman, K. R. (1994) Proc. Natl. Acad. detrimental) for species Sci. USA 91, 7463-7467. favors transmission of organelle genes advantageous (or 9. Gyllensten, U., Wharton, D. & Wilson, from that sex (e.g., anisogamy with more (and maybe not for individuals either). A. C. (1985) J. Hered. 76, 321-324. organelle replication in the female germ Consequently, the amount of uniparental 10. Gyllensten, U., Wharton, D., Josefsson, A. line to produce large eggs), plus organelle inheritance and recombination is probably & Wilson, A. C. (1991) Nature (London) variation favoring replication in one sex or determined largely by some combination 352, 255-257. mating type, will result in selection for of (i) chance events (mutation and drift, 11. Harris, E. H. (1989) The Chlamydomonas organelle genomes that replicate better in extinction), (ii) changes in selection coef- Sourcebook (Academic, New York). the sex or mating type with stronger trans- ficients because of the presence or ab- 12. Birky, C. W., Jr. (1978) Annu. Rev. Genet. mission (91). But this scenario seems un- sence of cytoplasmic parasites, (iii) selec- 12, 471-512. likely to play a major role in the evolution tion on other features (e.g., oogamy), and 13. Birky, C. W., Jr., Acton, A. R., Dietrich, of uniparental inheritance. There is no (iv) nucleocytoplasmic conflict. The R. & Carver, M. (1982) in Mitochondrial direct evidence for organelle variants that forces acting on the evolutionary history Genes, eds. Slonimski, P., Borst, P. & at in different of inheritance may be as di- Attardi, G. (Cold Spring Harbor Lab. replicate different rates uniparental Press, Plainview, NY), pp. 333-348. sexes or mating types. This hypothesis verse as those acting on sexual reproduc- 14. Wolfe, K. (1987) in Gene Structure in does not explain paternal inheritance in tion and as difficult to unravel. In neither Eukaryotic Microbes, ed. Kinghorn, J. R. oogamous organisms such as geraniums, case is there likely to be a single selective (IRL, Oxford), pp. 69-91. alfalfa, or conifers; neither does it explain force of overriding importance. 15. Rowlands, R. T. & Turner, G. (1974) Mol. uniparental inheritance from one mating How can we evaluate these hypotheses Gen. Genet. 133, 151-161. type when the mating types are similar in about the evolution of uniparental inher- 16. Kawano, S., Takano, H., Imai, J., Mori, K. size and physiology. itance? First, we need to accept the like- & Kuroiwa, T. (1993) Genetics 133, 213- Why Does the Evolution of Uniparental lihood that no one hypothesis is sufficient 224. Inheritance Show so Much Homoplasy? to explain the diversity of patterns of 17. Chiu, W.-L. & Sears, B. B. (1985) Mol. This discussion suggests several possible organelle gene inheritance. Second, we Gen. Genet. 198, 525-528. explanations for the frequent changes need more sophisticated phylogenetic 18. Masoud, S.A., Johnson, L. B. & So- to inheri- of the history of uniparental and rensen, E. L. (1990) Theor. Appl. Genet. from uniparental biparental analyses 79, 49-55. tance and back again, seen in the phylo- biparental inheritance, including estima- 19. Medgyesy, P., Fejes, E. & Maliga, P. genetic trees of Figs. 3 and 4. This kind of tions of ages of uniparental and biparental (1985) Proc. Natl. Acad. Sci. USA 82, instability might be expected from a trait lineages. Third, we need to ask the right 6960-6964. that is of little evolutionary consequence questions. Some of the hypothetical mod- 20. Thanh, N. D. & Medgyesy, P. (1989) Plant itself, and so is largely a byproduct of els for the evolution of uniparental inher- Mol. Biol. 12, 87-93. selection for other aspects of reproduc- itance are so detailed that they are un- 21. Rawson, P. D. & Hibish, T. J. (1995) Mol. tion. Alternatively, the instability might be likely to be correct in all aspects. These Biol. Evol. 12, 893-901. Downloaded by guest on September 25, 2021 11338 Review: Birky Proc. Natl. Acad. Sci. USA 92 (1995) 22. Stewart, D. T., Saavedra, C., Stanwood, 46. Cavalier-Smith, T. (1993) Microbiol. Rev. 70. Ursprung, H. & Schabtach, E. (1965) J. R., Ball, A. 0. & Zouros, E. (1995) Mol. 57, 953-994. Exp. Zool. 159, 379-384. Biol. Evol. 12, 735-747. 47. Knoll, A. H. (1992) Science 256, 622-627. 71. Anderson, W. A. (1968)J. Ultrastruct. Res. 23. Moses, M. J. (1961) J. Biophys. Biochem. 48. Sogin, M. L., Gunderson, J. H., Elwood, 24, 311-321. Cytol. 10, 301-333. H. J., Alonso, R. A. & Peattie, D.A. 72. Kondrashov, A. S. (1993) J. Hered. 84, 24. Sager, R. & Kitchin, R. (1975) Science (1989) Science 243, 75-77. 372-387. 189, 426-433. 49. Kondo, R., Satta, Y., Matsuura, E. T., 73. Brooks, L. D. (1988) in The Evolution of 25. Tilney-Bassett, R. A. E. (1970) Ishiwa, H., Takahata, N. & Chigusa, S. I. Sex, eds. Michod, R. E. & Levin, B. R. 25, 89-103. (1990) Genetics 126, 657-663. (Sinauer, Sunderland, MA), pp. 87-105. 26. Hiraoka, J.-i. & Hirao, Y.-h. (1988) Ga- 50. Kawano, S., Anderson, R. W., Nanba, T. 74. Coleman, A. W. (1982) J. Theor. Biol. 97, mete Res. 19, 369-380. & Kuroiwa, T. (1987) J. Gen. Microbiol. 367-369. 27. Kaneda, H., Hayashi, J.-I., Takahama, S., 133, 3175-3182. 75. Law, R. & Hutson, V. (1992) Philos. Taya, C., Fischer Lindahl, K. & 51. Meland, S., Johansen, S., Johansen, T., Trans. R. Soc. London B 248, 69-77. 76. Grun, P. (1976) Cytoplasmic Genetics and Yonekawa, H. (1995) Proc. Natl. Acad. Haugli, K. & Haugli, F. (1991) Curr. Evolution (Columbia Univ. Press, New Sci. USA 92, 4542-4546. Genet. 19, 55-58. York). 28. Szollosi, D. (1965) J. Exp. Zool. 159, 367- 52. Mirfakhrai, M., Tanaka, Y. & Yanagi- 77. Hastings, I. M. (1992) Genet. Res. 59, 215- 378. sawa, K. (1990) Genetics 124, 607-613. 225. 29. Meusel, M. S. & Moritz, R. F. A. (1993) 53. Silliker, M. E. & Collins, 0. R. (1988) 78. Hoekstra, R. F. (1990) in Organizational Curr. Genet. 24, 539-543. Mol. Gen. Genet. 213, 370-378. Constraints on the Dynamics ofEvolution, 30. Mogensen, H. L. (1995) Am. J. Bot., in 54. Creasey, A. M., Ranford-Cartwright, eds. Maynard Smith, J. & Vida, G. press. L. C., Moore, D. J., Williamson, D. H., (Manchester Univ. Press, Manchester, 31. Szmidt, A. E., Alden, T. & Hallgren, J.-E. Wilson, R. J. M., Walliker, D. & Carter, U.K.), pp. 269-278. (1987) Plant Mol. Biol. 9, 59-64. R. (1993) Curr. Genet. 23, 360-364. 79. Maynard Smith, J. (1978) The Evolution of 32. Neale, D. B. & Sederoff, R. R. (1989) 55. Vaidya, A. B., Morrisey, J., Plowe, C. V., Sex (Cambridge Univ. Press, Cambridge, Theor. Appl. Genet. 77, 212-216. Kaslow, D. C. & Wellems, T. E. (1993) U.K.). 33. Gillham, N. W. (1994) Organelle Genes Mol. Cell. Biol. 13, 7349-7357. 80. Martin, F. G. & Cockerham, C. C. (1960) and Genomes (Oxford Univ. Press, New 56. Adoutte, A. & Beisson, J. (1970) Mol. in Biometrical Genetics, ed. Kempthorne, York). Gen. Genet. 108, 70-77. 0. (Pergamon, New York), pp. 35-45. 34. Rowlands, R. T. & Turner, G. (1973) Mol. 57. Martin, F. N. (1989) Curr. Genet. 16, 373- 81. Pamilo, P., Nei, M. & Li, W.-H. (1987) Gen. Genet. 126, 201-216. 374. Genet. Res. 49, 135-146. 35. Rowlands, R. T. & Turner, G. (1976) 58. Sears, B. B. & VanWinkle-Swift, K. 82. Lynch, M., Burger, R., Butcher, D. & Genet. Res. 28, 281-290. (1994) J. Hered. 85, 366-376. Gabriel, W. (1993) J. Hered. 84, 339-344. 36. Sears, B. B. (1980) Plasmid 4, 233-255. 59. Neale, D. B., Marshall, K. & Sederoff, 83. Backer, J. S. & Birky, C. W. J. (1985) 37. Chase, M. W., Soltis, D. E., Olmstead, R. R. (1989) Proc. Natl. Acad. Sci. USA 86, Curr. Genet. 9, 627-640. R. G., Morgan, D., Les, D. H., et al. (1993) 9347-9349. 84. Birky, C. W., Jr. (1991) in Evolution at the Ann. Mo. Bot. Gard. 80, 528-580. 60. Wagner, D. B., Dong, J., Carlson, M. R. & Molecular Level, eds. Selander, R. K., 38. Maddison, W. P. & Maddison, D. R. Yuanchuk, A. D. (1991) Theor. Appl. Clark, A. G. & Whittam, T. S. (Sinauer, (1992) MacClade: Analysis of Phylogeny Genet. 82, 510-514. Sunderland, MA), pp. 112-134. and Chracter Evolution (Sinauer, Sunder- 61. Belcour, L. & Begel, 0. (1977) Mol. Gen. 85. Walsh, J. B. (1993) J. Hered. 84, 415-418. land, MA). Genet. 153, 11-21. 86. Butcher, D. (1995) Genetics 141,431-437. 39. Hagemann, R. (1992) in Cell Organelles, 62. Borkhardt, B. & Olson, L. W. (1983) Curr. 87. Kondrashov, A. S. (1994) Genetics 137, ed. Hermann, R. (Springer, New York), Genet. 7, 403-404. 311-318. pp. 65-96. 63. E. 88. Charlesworth, B. (1983) Nature (London) Caten, C. (1972) J. Gen. Microbiol. 72, 304, 211. 40. Reboud, X. & Zeyl, C. (1994) Heredity 72, 221-229. 89. Hoekstra, R. F. (1987) in The Evolution of 132-140. 64. Casselton, L. A. & Condit, A. (1972) J. Sex and Its Consequences, ed. Stearns, 41. Tilney-Bassett, R. A. E. & Abdel-Wahab, Gen. Microbiol. 72, 521-527. S. C. (Birkhauser, Basel), pp. 59-91. 0. A. L. (1979) in Maternal Effects in De- 65. Economou, A., Lees, V., Pukkila, P. J., 90. Hurst, L. D. & Hamilton, W. D. (1992) velopment, eds. Newth, D. R. & Balls, M. Zolan, M. E. & Casselton, L.A. (1987) Proc. R. Soc. London B 247, 189-194. (Cambridge Univ. Press, Cambridge, Curr. Genet. 11, 513-519. 91. Godelle, B. & Reboud, X. (1995) Proc. R. U.K.), 29-45. 66. Fukuda, M., Harad, Y., Imahori, S., Fu- Soc. London B 259, 27-33. 42. Dally, A. M. & Second, G. (1990) Theor. kumasa-Nakai, Y. & Hayashi, Y. (1995) 92. Platt, J. R. (1965) Science 146, 347-353. Appl. Genet. 80, 209-222. Curr. Genet. 27, 550-554. 93. Lewis, I. F. (1925)Am. J. Bot. 12,351-357. 43. Lee, P.-C. (1988) (M.S. thesis, Michigan 67. Yang, X. & Griffiths, A. J. F. (1992) Ge- 94. Kuroiwa, T., Kawano, S., Watanabe, M. & State University, East Lansing). netics 134, 1055-1062. Hori, T. (1991) Protoplasma 163, 102-113. 44. Mason, R. J., Holsinger, K. E. & Jansen, 68. Avise, J. C. (1994) Molecular Markers, 95. Ogawa, S. (1988) Bot. Gaz. (Chicago) 149, R. K. (1994) J. Hered. 85, 171-173. Natural History and Evolution (Chapman 25-29. 45. Steinborn, R., Linke, B., Nothnagel, T. & & Hall, New York). 96. Smith, G. M. (1950) The Fresh-Water Al- Borner, T. (1995) Theor. Appl. Genet., in 69. Magoulas, A. & Zouros, E. (1993) Mol. gae of the United States (McGraw-Hill, press. Biol. Evol. 10, 319-325. New York). Downloaded by guest on September 25, 2021