Gametophytic apomixis: elements and genetic regulation.

Gametophytic apomixis implies formation of unreduced embryo-sacs and capacity of their egg cells for parthenogenetic development. Each of these processes, as well as their union to give matromorphous offspring, are known from several species with sexual reproduction. A comparative survey is given of meiotic modifications, occurrence of apospory, female and male parthenogenesis and aberrant endosperm formation in sexual and apomictic species. 
 
 
 
Knowledge and models concerning the genetic basis of apomictic reproduction are summarized, as well as effects upon apomixis of changes of chromosome number and of hybridization. In the author's opinion, the constituents or elements of apomixis are — with the possible exception of apospory—to a large extent quantitative traits under polygenic control.


Definitions and control of apomictic phenomena
The classical definition of apomixis includes all kinds of asexual reproduction in higher plants that are able to replace sexuality more or less permanently in nature-asexual seed formation (agarnospermy) and vegetative reproduction (GUSTAFSSON 1946). In agamospermy, the embryos are either formed directly from somatic cells (nucellar embryony) or by parthenogenetic development of egg cells in unreduced embryo-sacs (gametophytic apomixis). In the latter case, the embryo-sacs are either formed from primary embryo-sac mother cells (EMC:s) by circumvention of meiosis (diplospory) or from somatic cells in the ovule (apospory).
Apomixis may be obligate or facultative-in the latter case, egg cells in reduced embryo-sacs are usually fertilized, whereas egg cells in unreduced embryo-sacs develop phrthenogenetically. Fertilization of the central nucleus (or polar nuclei) may be essential for seed formation (pseudogamy) or superfluous (autonomous apomixis). Pseudogamy is often combined with facultative apomixis.
According to the opinion of some authors, vegetative reproduction should not be included among the apomictic phenomena (cf. RUTISHAUSER 1967, p. 5-6). Even the inclusion of nucellar embryony has been questioned.
A functional gametophytic apomixis implies various changes of the sexual cycle. The most important ones are the predominant formation of unreduced embryo-sacs, and the highly developed capacity of their egg cells for parthenogenetic development. These two processes are under independent genetic control.
In angiosperms with sexual reproduction, unreduced embryo-sacs and egg cells are often formed, and parthenogenetic development of reduced egg cells sometimes gives rise to haploid offspring. These two phenomena may also, be combined to give diploid offspring of maternal type. Studies of such processes in sexual plants are obviously within the scope of apomixis research. Also must be taken into consideration cases where all (androgenesis) or part of the offspring (semigamy) receives its genetic material only from the male parent. -The reproductive possibilities of sexual and apomictic plants are summarized in Table 1.
Gametophytic apomixis is strongly correlated with the occurrence of hybrid and polyploid complexes. It was formerly thought to be a direct result of hybridization or increased level of ploidy. Today, all scientists in this field agree that apomixis is to some extent genetically regulated. Obviously, apomictic phenomena are also under environmental control, being influenced by factors like light and temperature regimen, and by the choice of pollinator (in facultative apomicts).
Although interesting models have been pro-posed, we don't know with certainty the exact genetic regulation of apomixis in any case, with the possible exception of apospory. It is very important, in my opinion, to keep in mind that the different constituents, or elements (PETROV 1976) of apomixis, lie within the reproductive potentiality of sexual plants. Here, they play a greater part only when the normal sexual process with fusion of reduced gametes is impaired. Where they occur occasionally in sexual plants, such elements cannot be expected to be under simple genetic control.

Origin and advantages of apornixis
A perfectly operating gametophytic apomixis is usually not likely to arise in one single step. Genes promoting effective apomictic reproduction are successively incorporated by mutation or recombination. According to a hypothesis developed for P~r r h e r i i u~?~ by POWERS ( 1945). apomixis depends upon homozygosity for recessive genes influencing its constituents. The sexual ancestors of apomicts are usually cross-fertilizers-preferably self-incompatible or dioecious, because a high level of outcrossing should be necessary to bring together the genes for apomixis. Certain intermediate combinations of such genes are, according to Powers, liable to reduce fertility. As the sexual plants giving rise to apomicts are also often more or less sterile due to hybrid structure or polyploidy, they should usually be perennial plants with good capacities for vegetative reproduction (Table 2). What, if any, are then the evolutionary advantages of acquiring the capacity for apomictic reproduction? According to MAIHER (l943), each population has to find a compromise between immediate fitness-with genetic constancy but difficulties of adaptation to a changing environment-and flexibility-with high genetic variability but with formation of many individuals poorly adapted to the immediate environment. The facultative apomicts have exploited an extremely effective solution to these contrasting demands, one that "enables them to store interspecific variability and, nevertheless, remain constant for ages" (CLAUSEN 1954). STEBBINS (1958) states that adaptations promoting immediate fitness occur in temporary habitats, "Escape from sterility" I . "A blind alley of evolution" (DAR-2. "A regular step of progressive evolu-3. Facultative (as opposed to obligate) apornicts retain their capacity for further evolution tion" (KHOKHLOV 1976) where populations fluctuate greatly in size and formation of several descendants is essential. Such adaptations are apomixis, self-fertilization, and low recombination index (low chromosome number combined with low chiasma frequency).
According to DARLINGTON (l939), agamic complexes have a limited evolutionary potential, losing their capacity for further evolution when the condition of obligate apomixis is finally reached. This is true to the extent that, obviously, apomictic groups have not given rise to new higher taxonomic categories such as families.
On the other extreme stands the view of KHOKH-LOV (1976). He looks upon the development of apomixis as a regular step of progressive evolution. In his opinion, the reduction of the gametophyte in apomicts is only the end point of the successive reduction of the gametophyte during the development of higher plants.
According to the commonly adopted view, species with facultative apomixis retain their capacity for progressive evolution, whereas obligate apomixis should be an irreversible and regressive process. Recent work indicates that traces of sexuality occur even in groups judged to be obligately apomictic (ASKER 1979).
Apomixis obviously involves fixation of heterosis. It is not surprising, then, that induction of apomixis in sexual crops has been considered by breeders. According to SOLNTSEVA (l978), this possibility was discussed by Navashin and Karpechenko as early as in the thirties. During later years, work of this type has been carried out in different materials.
In this paper, especially the constituents (elements) of apomixis are discussed. Their occurrence in sexual taxa is reviewed, and our knowledge concerning their genetic regulation is surveyed.
The elements of apomixis

A . Sexual species
Several mutant genes are known that influence the course of meiosis in sexual species. Such mutations were first described by BEADLE (1930BEADLE ( , 1932BEADLE ( , 1933 in maize and by SATINA and BLAKESLEE (1935) in Datum. Some examples are given in Table 3.
Many of these mutations affect synapsis. According to CATCHESIDE (1977), true asvnapsis -failure of synapsis at zygotene and failure of chiasma formation-is a rare case. More common are desynaptic mutants, where synapsis occurs but no or few chiasmata are formed.
Especially well known is the genetic control of the cytologically diploid behaviour of hexaploid wheat. A gene on chromosome 5B suppresses homoeologous or at the other extreme, promotes homologous, pairing (RILEY and CHAPMAN 1958).
The gene elongate (el) in maize causes a high frequency of unreduced egg cells (RHOADES 1956) and has been used for various experimental purposes. It omits the second division, which should lead to the onset and conservation of a high level of homozygosity .
Mutants are also known that affect post-meiotic development of spores. Especially in fungi, mutations are known that influence recombination in various ways other than meiotic changes.
Unreduced egg cells are rather frequent in sexual plants (HARLAN and DE WET 1975;FRANKE 1975). The formation of unreduced gametes in plants is usually thought to depend on restitution following the first or second meiotic division. But unreduced egg cells may also be derived from embryo-sac mother cells with twice the somatic number, or from occasional aposporous embryo-sacs.
Meiotic disturbances, such as in species hybrids and autotriploids, greatly increase the formation of unreduced gametes through restitution. Often all functional egg cells are unreduced in such cases.-In wide crosses, even where meiosis is not disturbed, sometimes only unreduced egg cells give rise to hybrids. The function of unreduced egg cells is thus dependent on the type of pollination.
Autotriploidssometi mes, perhaps, "future apomicts" -originate by fertilizations of unreduced egg cells. Their occurrence reveals the presence of unreduced gametes even in diploids with perfectly regular meiosis. Function of unreduced gametes is, on the whole, the most important mechanism in the formation of polyploid series.

B . Aponiictic species
We now turn to the modifications of meiosis in apomictic plants. In diplospory, the course of the first nuclear division in the EMC:s may take three different forms, according to GUSTAFSSON (1946). These are a formation of restitution nuclei, pseudohomeotypic division, and mitotic division. The distinction of the pseudohomeotypic division was, however, questioned by RUTISHAUSER (1967).
In Taraxacurn, for instance, stages similar to a normal first division of meiosis are observed, but most chromosomes occur as univalents. The distribution of homologous chromosomes at the end of the first division is interrupted, and all chromosomes are enclosed within a common nuclear membrane. But the second division proceeds normally, giving a dyad of cells one of which gives rise to the embryo-sac.-In Antennaria and Eupatoriuni on the other hand, the EMC grows out directly into a gametophyte without cell divisions. Thus, the first division of the EMC is here of mitotic type.
In all diplosporous apomicts with restitution nuclei so far studied, restitution takes place after the first meiotic division (FDR). However, in occasional formation of unreduced egg cells in sexual plants, both FDR and SDR (restitution after the second division) occur.
As suggested earlier, mutations are also known that induce either FDR or SDR. SDR should lead to homozygosity , whereas FDR preserves heterozygosity and fixes heterosis. These statements are quite true, however, only if no bivalent formation or at least no crossing-over between homologues takes place. The suppression of chromosome pairing, which has been observed in diplosporous plants with restitution nuclei, is maybe another genetically determined meiotic adaption. Inhibition of meiosis in primary EMC:s sometimes occurs even in the case of apospory. In certain Parrlitillu biotypes, meiosis seems to be initiated in primary EMC:s and proceeds to a pachytene-like stage. Here, the divisions stop and the EMC:s gradually degenerate. Meiosis on the male side is much less disturbed in these pseudogamous forms. By~inhibition of meiosis in primary EMC:s the competition between reduced and unreduced embryo-sacs might be eliminated. Constancy of offspring is ensured at the expense of evolutionary rigidity.
In plants with autonomous apomixis, various degrees of degeneration of the male meiosis are observed (some species even lack pollen mother cells). The breakdown of meiosis must depend on the uselessness of pollen for reproduction. The degenerative phenomena are more pronounced in diplosporous apomicts. This may simply be due to the fact that autonomous apomixis is common here, whereas apospory is combined with pseudogamy to a much higher extent.
The occurrence of these meiotic changes in connection with the transition from sexuality to apomixis makes it difficult to conclude to what extent meiotic disturbances were present in the sexual population giving rise to the apomicts (see Table 4). We know almost nothing about the genetic background of the secondary changes of meiosis. The genetics of permanent and casual diplospory is discussed later in this paper.
The course of meiosis in sexuals and facultative apomicts depends also upon environmental influences. These affect in apospory the frequency of reduced egg cells formed after completed meiosis and, in diplospory, the balance between reduction and restitution.

Aposporous embryo-sacs
In apospory, one or more somatic cells in the ovule usually enlarge considerably and become vacuolated, giving rise to a uninuclear embryo-sac. If their development is completed, usually 8-nucleated embryo-sacs of the common Polygonum-type are formed, which are morphologically similar to the meiotic embryo-sacs. In Gramineae (Poaceae), subfamily Panicoideae, the aposporous sacs are usually four-nucleated and monopolar, which facilitates their separation from sexual embryo-sacs and the estimation, of the degree of apospory in squash preparations.-In certain genera of Rosaceae, the complex structure of the archespore makes the distinction between embryo-sacs with diplosporous and aposporous origin difficult.
Aposporous embryo-sacs have been observed in several sexual species. Especially, they have been found within Compositae (Asteraceae)-a list of species is given by DAVIS (1967). For instance, they occur in Cirsium arvense and other Cirsium species (ZABINSKA 1977). Aposporous sacs occur in the intergeneric hybrid Raphanobrassica (ELLERSTR~M and ZAGORCHEVA 1977) and in certain hybrids of Sanguisorba (NORDBORG 1%7). Sometimes, aposporous initials or embryo-sacs degenerate without giving any offspring. Their formation and function is enhanced by, if not caused by, defective functioning of the meiotic sacs.
Apospory , meaning the presence of aposporous embryo-sacs (as contrasted to sexuality = absence of aposporous sacs) is a phenomenon quite independent of the course of meiosis and the disposition to parthenogenetic development of egg cells. As discussed later, it seems to have a simple genetic regulation in certain cases.

A . Haploid formation
Parthenogenetic development of egg cells takes place occasionally in sexual plants, whereas in plants with regular gametophytic apomixis, it is a necessity. The latent tendency towards parthenogenesis in sexual plants is explicitly displayed by the production of haploids in a multitude of plant species. Reviews in this field have been written by, e.g., KIMBER and RILEY (1963) and-concerning maize-by CHASE (1969). The parthenogenetic development of a reduced gamete from a diploid plant by definition gives rise to a monoploid (LANGLET 1927) or simply a haploid, whereas a polyploid plant under the same conditions gives rise to a polyhaploid (KATAYAMA 1934).
Haploids arise by spontaneous parthenogenesis, but their formation is promoted by various cir-cumstances. They are known to occur at increased frequency among twin plants, resulting from polyembryonic seeds. In this case, the haploid embryos may develop from synergids. Special types of pollination can induce parthenogenesis: wide crosses, delayed pollination, use of pollen treated with X-rays, ultraviolet light, or toluidine blue ( K I M B E R and RILEY 1963). Haploids have also been obtained after treatments designed to give polyploids, viz. colchicine treatment and temperature shocks (Table 5). It is usually not possible here to decide if the egg cell or some other cell in the embryo-sac has given rise to the embryo.
Studies of meiotic chromosome behaviour in haploids have revealed pairing affinities not observed in the "diploid" plants, and, thus, contributed to our cytogenetic knowledge of certain groups. By doubling the chromosome number of monohaploids, completely hornozygous ("autodiploid") lines for breeding and experimental purposes are obtained.
Cultivated alfalfa (Medicago sativu) behaves as an autotetraploid, and dihaploids are produced from certain 4x x 2r crosses. Primary haploids are rather infertile. By crossing them as females with 2r Medicago falcaia and repeated back-crosses with dihaploids combined with selection, stable and vigorous diploids containing more than 98 7% cultivated germ-plasm have been obtained ( B I N G -HAM and Mc COY 1979).
From tetraploid Solut~urn tuherosutn (2n =48), it has even been possible to produce monohaploids with 12 chromosomes by two successive cycles of female parthenogenesis, obtained by pollination with S. phureju (BREUKELEN et al. 1975). Such a repeated halving of the chromosome number (or "subhaploid" formation) has earlier been observed in apomictic Poa species (KiELLANDER 1947;AKERBERG and BINCEFORS 1953 interspecific crosses, so-called matromorphic plants are sometimes obtained, even where self-pollination after incomplete castration or contamination with foreign pollen can be excluded. According to EENINK (1974a), matromorphy, being synonymous with diploid parthenogenesis, should not be confused with matrocliny, meaning the formation of true hybrids which resemble the mother much more than the father. We still know too little about the origin, occurrence and frequency of matromorphic plants. Possibly, the formation of such plants is a common phenomenon. Their presence after open pollination is likely to be overlooked. Matromorphy can be more common than haploid parthenogenesis, as especially monoploids are often inviable. The origin of matromorphic plants can be indicated by the use of suitable markers in crosses.
Matromorphy is best known from Brassicu, where it has been extensively studied by EENlNK (1974a. b, c, d , e , 1975a, b). References to other genera are given in Table 6.
Plants of maternal type are derived by asexual seed formation: (1) from EMC:s with twice the somatic chromosome number (pre-meiotic endomitosis has been observed in Brussica according to E E N I N K 1975b), giving rise after meiosis to "unreduced" macrospores and embryo-sacs; (2) by parthenogenetic development of egg cells in embryo-sacs, formed as a result of restitution after 1st or 2nd meiotic division: (3) by parthenogenetic development of reduced egg cells followed by reduplication or fusion.
The last possibility is interesting from the breeder's point of view; it would give rise directly to homozygous diploids. The formation of such diploids has been presupposed by several authors. It seems, however, that we still need a definitive genetical and embryological proof of their origin.
Automixis, fusion of two haploid nuclei in a meiotic embryo-sac to give diploid and homozygous progeny, has been claimed to occur in Rubus (THOMAS 1940). In any case, such an automictic parthenogenesis does not at all play the same role in the plant as in the animal kingdom.
In Brassica, formation on unreduced embryo-sacs has been demonstrated, as well as the capacity for parthenogenetic development of their egg cells. Thus, not only the part processes of apomixis, but also their combination to a kind functional diplospory lie within the scope of reproductive behaviour in sexual plants.
In this connection, a question of terminology should be touched upon. Often plants which have originated by haploid or diploid parthenogenesis are designated as "apomictic haploids, parthogenetic diploids" and so on, irrespective of their own mode of reproduction. This can be misleading, in my opinion. One has to speak of such plants as (poly-)haploids or matromorphic, or else use the more laborious designation "plants that have arisen by . . .".

C. Chemical treatments
Various chemical treatments have been tried in order to induce parthenogenesis in plants, but generally without success. On the other hand, parthenocarpy is easily induced by treatment by plant hormones. DEANON (1957) reported a significantly increased frequency of monoploids in maize after treatment with 50 ppm maleic hydrazide. VERMEL and SOLOVOVA (1973a, b) claimed to have induced diploid parthenogenesis in some plant species by treatment with dimethyl sulphoxide (DMSO). According to their opinion, the diploids were homozygous, resulting from haploid parthenogenesis followed by chromosome doubling. The present author (unpublished work) performed DMSO treatments on Ranunculus material, furnished with genetic markers to judge if resulting diploids were homozygous. However, in this self-incompatible material there was n o significant increase of seed set after isolation following DMSO treatment.

D. Pseudogamy and parthenogenesis in apornicts
Haploid and diploid parthenogenesis in sexual plants seem to be combined with pseudogamy. The central nucleus must be fertilized to secure endosperm and seed formation. It is not known if, in this case, the egg cell divides before fertilization.
In pseudogamous animals, the entrance of the sperm initiates parthenogenetic development. Similarly in certain plant species, for instance Ranunrulus auricotnus, the egg cell does not divide until after fertilization of the central nucleus This seems, however, to be an exception among higher plants. In apomictic angiosperms, the parthenogenetic development is usually initiated well before the opening of flowers. At anthesis, the embryo-sacs contain multicellular embryos, instead of egg cells capable of fertilization (POD-DUBNAJA-ARNOLDI 1939in GUSTAFSSON 1946. A regular formation of maternal offspring is secured by an early division of the egg cells. In the case of pseudogamy, however, without fertilization and endosperm formation, the embryo eventually degenerates. According to GWTAFSSON (1947b) "this new rhythm often gets accentuated by the mere increase in chromosome number. Superficially seen, this early egg cell division does not tally with the general effect of polyploidy, which is to delay mitosis". In sexual plants, a prolonged resting stage of the egg cell is desirable to secure fertilization. Polyploidy is thought to eliminate this long resting stage, which would be useful in an apomict .
BENNETT (1977 a.0.) has shown that polyploids have a shorter duration of meiotic divisions than diploids with corresponding DNA amounts. The shorter meiotic divisions may be one cause of early egg cell division in polyploids, and this effect may remain even in diplosporous apomicts where meiosis is mitotisized.
Even in autonomous apomicts, the development of the embryo is sometimes independent of that of the endosperm (Taraxacum), whereas in other cases the egg cells divide after endosperm formation (DOLL 1971).
Fertilization of unreduced egg cells in apomicts-possible, where egg cells do not divide prior to pollination-may add to the formation of polyploid complexes. However, it interferes with the preservation of heterosis and is not desirable to occur in higher frequencies, POWERS (1945) described a Parthenium strain which "polyploidized itself out of existence" by repeated fertilization of unreduced egg cells in the offspring. A similar (RUTISHAUSER 1967).
behaviour was observed by me in offspring from a biotype hybrid in Poteniilla argrntea (ASKER 1970b).
In apomicts where meiosis is not wholly suppressed, polyhaploids are sometimes formed by parthenogenetic development of reduced egg cells. In grasses belonging to the Bothriochloa-Dichanihiuni-complex diploids are regularly formed from apomictic tetraploids in this way (DE WET 1968).

E . Androgenesis and serniganiy
In special cases, offspring obtained by parthenogenesis has the genotype of the male gamete nucleus. In the case of androgenesis, or male parthenogenesis, the early stages of fertilization take place normally. After the male nucleus has entered the egg cell, the nucleus of the latter degenerates. The pollen nucleus divides in the cytoplasm of the egg cell and finally gives rise to a haploid embryo.
Androgenesis was first observed after a cross Nicotiana digluia x N . tabacum (CLAUSEN and LAMMERTS 1929). In Crepis, GERASSIMOVA (1936) demonstrated the occurrence of androgenesis by X-ray treatment of plants homozygous for a dominant marker and pollination from untreated plants homozygous for the corresponding recessive genes. EHRENSBERGER ( 1948) obtained paternal haploids in Antirrhinum from crosses irradiated egg cells x normal pollen. The reverse combination normal egg cells x irradiated pollen gave rise to maternal haploids. In maize, androgenetic haploid formation occurs spontaneously at a rate about I x (CHASE 1969). Semigamy was first described by BAITAGLIA (1945) in Rudbeckia. Here, sometimes the sperm nucleus does not fuse with the egg nucleus, and they begin to divide independently and simultaneously. In Rudbeckia, the embryo-sac nuclei are unreduced, and in this way chimaeric embryos with both the somatic and half the somatic number arise. In Zephyranthes (SOLNTZEVA 1978) male nuclei have been observed to give rise to diploid tissue after restitutional division.
In a semigamous strain of Gossypium barbadense, about 40 % haploids are formed in the S, progeny (CHAUDHARI 1978). When female plants of this strain are crossed with non-semigamous male parents, about 1 % purely androgenetic haploids appear in the offspring. This technique allows the deliberate production of haploids of different strains of cotton. GERLACH-CRUSE (1970) induced semigamy in Arabidopsis by X-ray irradiation of floral parts, followed by pollination with untreated pollen. As the male nucleus uses the cytoplasm of the egg cell for its development into a sporophyte, the possibility of androgenetic or semigamous haploid formation allows the transfer of genomes to study their effects in different cytoplasms. Examples of androgenesis and semigamy are listed in Table 7

Endosperm formation nucleus must fuse with a central nucleus with
Embryo growth and seed formation in most angiosperms depend on the development of the endosperm. In sexual plants, the polar nuclei of the embryo-sac usually fuse to form a central nucleus, which is fertilized by a sperm nucleus. In a diploid plant, the endosperm normally has the triploid chromosome number. Autonomous endosperm development-without fertilization of the central nucleus or polar nuclei-has only occasionally been observed in sexual plants such as Anemone nemorosa (TRELA 1963) and Triticum aestivum (KANDELAKI 1976). In autonomous apomixis, depending on whether polar nuclei fuse or not, the endosperm will have double the somatic, or the somatic chromosome number.
It is doubtful if pseudogamy precedes autonomous apomixis in the evolution of apomictic taxa. Although in certain cases the genetic control of the parthenogenetic development of nuclei giving rise to endosperm seems to be independent of that of the egg cell, it is not excluded that parthenogenesis of both types of nuclei can be acquired in one single step. Pseudogamy and autonomous apomixis almost never occur in the same agamic complex, with the exceptions of Malus and Poa (DOLL 1971). No transitions are known between pseudogamy and autonomous apomixis: they seem mutually exclusive.
The development and especially the cytology of the endosperm in pseudogamous species offer certain points of interest. Endosperm cytology in apomicts has been studied by NOACK ( When pseudogamous plants are used as mothers in crosses, only the genetic constitution of the endosperm varies, whereas that of the embryo and the maternal tissue is constant. Like in sexual plants, some degree of seed incompatibility (VALENTINE 1960) exists in pseudogamous taxa. For instance, 4~x 2~ crosses in Ranunculus auricomus using tetraploid pseudogamous plants give a very low seed set due to degeneration of the endosperm.
NOGLER (1978) rightly points out that each fertilization of the central nucleus to form endosperm within a pseudogamous biotype is of the same type as that in a sexual 4~x 2~ cross. A reduced male twice the somatic number. Now the genetic constitution of the endosperm is thought to be the cause of seed incompatibility in crosses like 4~x 2~ (VON WANGENHEIM 1967). The relation genome: plasmon in the endosperm is normally 3:l in sexual plants. In a 4~x 2~ cross, this relation is instead 5:2, and this deviation from the normal ratio causes degeneration of the endosperm.-According to RUTISHAUSER (1%9), instead the relation between female and male genomes in the endosperm is decisive for its vigour. SKIEBE (1973) states that endosperms with 3x or a multiple of that number function best.
Special mechanisms for the endosperm formation in pseudogamous plants have been developed to avoid the possible threat of seed incompatibility, according to Nogler. The most radical one is found in grasses of the subfamily Panicoideae. Here, four-nucleated sacs are formed with only one unreduced polar nucleus, which is fertilized by a reduced sperm nucleus to give an endosperm which is triploid in relation to the somatic number. In other cases there occurs fertilization of the central nucleus by two reduced or one unreduced sperm nucleus, or fertilization of polar nuclei (which have not fused) by reduced pollen.
To sum up, endosperm formation is variously modified in apomicts. Many of the changes probably occur after the transition to apomictic reproduction. So far, we know almost nothing of the genetic background of these events. Changes accompanying the transition from sexuality to (gametophytic) apomixis are listed in Table 8.
The genetic regulation of apomixis

Occasional apomixis in sexual plants
As emphasized in the preceding text, formation of unreduced female gametophytes, as well as haploid and diploid parthenogenesis, occurs even in sexual plants. Such phenomena depend on the plant's genetic constitution, but cannot generally be supposed to have a simple genetic background. The degree of apomixis is obviously influenced by environmental factors, the same way as the balance between apomictic and sexual reproduction in facultative apomicts.
Data concerning gene regulation of parthenogenesis in sexual plants are meagre. In maize, genotypic influences of both female and male parents on the frequency of maternal haploidy have been confirmed (SARKAR and COE 1966). Lines with a high frequency of monoploid formation have been isolated in maize (CHASE 1969) and in barley (HAGBERG and HAGBERG 1980). The frequency of matromorphous plants in Brussica crosses is clearly dependent on the genetic constitution of both the female and male parent, according to EENINK (1974b). However, the gene background seems to be complicated.-Probably parthenogenetic development of reduced and unreduced egg cells is under similar genetic control, but the experimental evidence on this point is not clear.
The frequency of unreduced egg cells varies strongly among inbred lines of maize ( A L E X A N D E R and BECKETT 1963). Also in other genera like Sacchuruw and Citrus, it is a well-known fact that some clones or stocks produce unreduced 9 or d gametes to a much greater extent than others (HARLAN and LIE WEI 1975).
BINGHAM and Mc COY (1979) mention a case where formation of unreduced gametes at a high frequency seems to be under simple genetic control in Medicago. Meiotic mutants-including such that give rise to unreduced embryo-sacs-have been accounted for earlier (cf. Table 3). In Table 9, additional evidence on genetic regulation of gametophytic apomixis and its elements in sexuals is summarized. YUDIN ( 1970) conducted parallel crosses between diploid maize lines (2Wx2W) and between their colchicine-induced tetraploid analogues (4x ~4 x ) .
Using appropriate markers, a statistically significant increase in the frequency of reduced parthenogenesis was found in the tetraploids. If parthenogenesis would thus be enhanced by polyploidization, the connection between hybridization and formation of unreduced gametes is more obvious. Especially, interspecific hybridiza-7uhlr 9. Evidence of gene regulation of occasional gametophytic apomixis and its elements in sexual plants (cf. also

Analyses in apomicts
To analyze the gene background of apomictic reproduction in established apomicts is a difficult task. The usual approach is to cross apomictic and related sexual strains and to study the mode of reproduction of the descendants. Several investigations of this type have been performed and extensive analyses of the results have been published, which cannot, however, be accounted for here. For various reasons, the results of such investigations are often difficult to evaluate. Many authors deal only with the inheritance of apomictic reproduction as an entirety, contrasted to sexuality. In this sense, apomictic reproduction is usually recessive to sexuality, but an excess of "apomictic" genomes in a hybrid may shift the mode of reproduction towards apomixis. However, formation of unreduced gametes and parthenogenesis are certainly under independent genetic control. This is indicated by the occur- Table 10. Some models for genetic regulation of apomixis A = apospory, D = diplospory, P = pseudogamy rence, in offspring from crosses between apomicts and sexuals, of plants with haploid parthenogenesis or with fertilization of unreduced egg cells.
The constituents of apomixis should preferably be analyzed separately.
Further, the sexual and apomictic types have not always been very closely related, which renders genetic analyses difficult. Where sexual diploids closely related to the apomicts are not known, a thorough search of natural populations may lead to their discovery (cf. Panicum maximum, PERNES et al. 1975).
A great part of the work in this field has been performed with the giant agamic complexes of Pou and Potentilla, where the conditions are likely to be more complicated than in several other apomictic groups.

Powers' model
Some models proposed for the genetic regulation of gametophytic apomixis deserve to be introduced here (Table 10). Important, although controversial, is the model given by POWERS ( assumed that homozygosity for certain recessive genes is a prerequisite for apomictic reproduction. Three elements of apomixis were discriminated: ( 1 ) reduction as opposed to failure of reduction of chromosome number (controlled by the alleles A a ) , (2) fertilization vs. failure of fertilization of egg cells ( B b ) , and (3) non-development vs. development of the egg cell to an embryo without its being fertilized ( C c ) . Accordingly, apomicts should have the genotype aabbcc; apomixis is likely to arise on the diploid level. Upon this, GUSTAFSSON (1947a) commented, that occurrence vs. failure of fertilization need not be under (simple) genetic control. It depends on the kind of pollen applied, on the time of pollination, and on the environmental conditions of the mother plant. Further, diploid apomicts are rare in nature, ant it seems plausible that apomixis is usually acquired by polyploids. MISHANEC (1950) andGERSTEL et al. ( 1953) only found sexual reproduction among F, plants from crosses between sexual and apomictic diploids of Partheniuriz argentatum. It was concluded, that apomixis is based on at least four recessive genes here: a minimum of two being concerned with meiotic reduction and two more with the requirement of fertilization.

Other models for apomicts
In Scandinavian taxa of Sorbus, the genome A from the apomictic autotetraploid S. aria carries gene(s) for apomixis, according to LIUEFORS ( 1955). The B genome from diploid S . aucuparia carries the corresponding gene(s) for sexuality. Among tetraploid and triploid derivates, AAAA, AAAB and AAB are obligatorily apomictic, AABB are facultative apomicts, ABB either sexual or apomictic, whereas BB is totally sexual. Certain exceptions from this schedule could be due to interchanges between chromosomes from the A-and B-genomes. Thus if genes for apomixis or sexuality are predominant, they give rise to apomictic and sexual reproduction, respectively. Tetraploids where genes for apomixis and sexuality are numerically balanced have capacities both for apomictic and sexual reproduction.
Next, some studies in apomictic fodder grasses belonging to the Panicoideae will be reviewed. BURTON and FORBES ( I 960) studied offspring from crosses between sexual autotetraploid and apomictic tetraploid Paspalum notatum. The sexual to apomictic ratios in F, indicated, in their opi-nion, that apomixis is controlled by a few recessive genes.
In buffelgrass and birdwoodgrass, Penniserum ciliare (or Cenchrus ciliaris), the inheritance of apomixis has been studied by TALIAFERRO and BAS-HAW (1966) and READ (1971). Most cultivars are obligatorily apomictic, but occasional sexual aberrants are obtained. Crosses between such sexual and apomictic plants were performed. Data from offspring by selfing fit a ratio of 13:3 between sexual and apomictic plants. Hybrids between sexual aberrants and obligate apomicts give sexual: apomictic ratios not significantly different from 5:3. The mode of reproduction is proposed to be controlled by a two-gene system where B conditions sexuality and is epistatic to gene A, which controls apomixis. The genotypes of apomictic and sexual plants would then be Aabb and AaBb, respectively.
Even in Panicurn maximum, HANNA et al. (1973) obtained sexual aberrants from tetraploid apomictic strains. In selfed progenies from sexuals, there was a predominance of sexual plants. The data agree rather well with an 11:5 ratio between sexual and apomictic plants. A two-locus control is presupposed where the genotypes aabh, Aabh, and aaBb are apomictic. Offspring from sexual plants with the AaBb genotype contain both sexual and apomictic plants, whereas those with the genotypes AABB, AABh, AAhb, AaBB, and aaBB produce only sexual progeny. -Unfortunately, like the preceding authors H A N N A et al. did not analyze the different components of apomictic reproduction separately.
Concerning diplospory, we have so far no conclusive data to prove a simple gene background for the necessary meiotic changes. They may partly depend on single gene mutations profoundly affecting the reproduction cycle, and partly be under polygenic control.
In the case of parthenogenesis, no case of sim-ple genetic control seems to be known, with the possible exception for barley and maize strains with an excessive haploid formation. Most authors judge the gene background of parthenogenesis to be complicated. In cases where diplospory and apospory seem to occur together and the limit between them is difficult to draw-like in certain Rosaceae-it seems probable that the processes leading to the formation of unreduced gametophytes have a unitary genetic cause.
All species included in Table 10 are pseudogamous and all but one aposporous, which indicates our lack of knowledge concerning diplospory combined with autonomous apomixis.

Effects of polyploidy and hybridization
Sometimes, the apomictic reproduction seems to be stable only on a certain level of ploidy, as euploid or aneuploid changes of the chromosome number may affect the mode of reproduction. In triploid and diplosporous Taraxacum species with 2n =24, eight morphologically different aberrants with 2n=23 were produced by loss of individual chromosomes. Two of these types were partially 1958). In aposporous and pseudogamous Poa pratensis, aberrants which are "triploid" or "haploid" as compared with the original strain have sometimes an increased degree of sexuality (MUNTZING 1940).
The connection between apomixis and polyploidy is often pointed out. In most agamic complexes, apomictic reproduction is restricted to polyploids. However, doubling of the chromosome number is known sometimes to change the mode of reproduction towards sexuality. In Potentilla collina (MUNTZING and MUNTZING 1943), P . crantzii ( MUNTZING 1958), and Hieracium hoppeanum (CHRISTOFF and CHRISTOFF 1948), haploid parthenogenesis frequently occurs after chromosome doubling. Obviously, the inhibition of meiosis is no longer effective.
Diploid facultative apomicts of Potentilla argentea give rise to totally sexual autotetraploids (ASKER 1971). Aposporous initials occur, but probably no mature aposporous embryo-sacs are formed. However, further embryological studies are needed in this case. As stated earlier, polyploids have a shorter meiotic cycle than corresponding diploids (BENNETT 1977). In the present case, this may enhance the formation of meiotic at the expense of the aposporous embryo-sacs.
Sexual (SBRENSEN and GUDJONSSON 1946;S0RENSEN Many agamic complexes are largely of hybrid origin. However, hybridization between apomicts-if possible-sometimes leads to a breakdown of apomixis. This is the case in Poa, where, according to MUNTZING (1940), "apomixis is due to a rather delicate genetic balance. This balance may be upset in various ways, by crosses with other types or merely by a quantitative change in chromosome number either in a plus or minus direction".-In Rubus, crosses between different apomictic microspecies give rise to sexual plants.
Only crosses between closely related individuals give rise to apomictic offspring (GUSTAFSSON 1943).
The breakdown of apomixis in such crosses is thought to depend on its different genetic regulation in the parents. If this is so, the degree of incompatibility between the genetic systems causing apomixis is not always correlated with the general taxonomic distance between the parents. Crosses between apomictic biotypes within Potentilla argentea may give rise to sexual hybrids (ASKER 1970b), whereas the interspecific hybrid P . argentea x P . canescens is obligatorily apomictic (ASKER 1970a). See Table 11.

Apomixis versus sexuality
It seems that the ideas concerning the inheritance of apomixis need some revaluation. To look upon apomixis as an antipode to sexuality is not correct, in my opinion. Sexual plants can form "matromorphous" offspring, usually by parthenogenetic development of egg cells in unreduced embryo-sacs, formed by some kind of diplospory.
These capacities for apomictic reproduction are usually realized only to a limited extent. Under special circumstances-such as meiotic disturbances due to hybridization or polyploidy-only apomixis allows seed formation. This may be compared with cases where vegetative reproduction becomes the only way to form offspring, due to sterility or absence of male plants.
A simple genetic regulation, perhaps by a few recessive genes, has sometimes been proposed for diplospory and parthenogenesis. Even if such a regulation cannot be excluded in certain cases, the above-mentioned two processes are certainly to a great extent under polygenic control.
When fertility is impaired, the presence of a high dosage of polygenes promoting apomictic reproduction must be favourable. Such gene combi- nations can arise as transgressions and be one cause of the connection between apomixis and hybridization. The presence of recessive genes with a strong effect upon the constituents of apomixis-and each taken separately strongly reducing fitness in homozygous state-is thus not a prerequisite in the population for the origin of apomictic reproduction. On the other hand, mutations enhancing apomixis will have a great selective value for instance in a sterile hybrid, which has a certain level of apomictic seed production but is mainly restricted to vegetative reproduction for its survival.
Concerning apospory, the occurrence and function of aposporous embryo-sacs seem to be independent of the sexual process. Apospory seems to have a simple genetic background, at least in certain cases, but superimposed are effects of several genes influencing the expression of the capacity for apomictic reproduction.
In studies on the balance between apomictic and sexual reproduction, the dependence on a level of ploidy is a factor to be taken into consideration.
For reasons still not clearly understood, apomictic reproduction is often impaired after changes of chromosomes number. Further, hybridization between facultative apomicts may give sexual offspring. This could support the theory of recessive genes for apomixis-different gene backgrounds in the parents would combine to create heterozygosity for the dominant genes for "sexuality". However, alternative explanations are possible. We need more experimental evidence in this area, like in the case of the genetic background for anomalous endosperm formation in apomicts.

Models for the origin of gametophytic apomixis
Short-time and long-time advantages of apomictic reproduction were discussed in the beginning of this paper. Here were also discussed the pre-requisites of certain plant populations for the development of apomixis. It is possible to construct simple models for its origin. POWERS (1945). based on his conception of a few recessive genes for apomixis, presented a hypothesis for the origin of apomixis in sexual populations. Even later authors have treated this question.
It must be kept in mind that different kinds of gametophytic apomixis have a different origin and genetic regulation. The background may differ, for instance, ( 1) in dioecious and hermaphroditic plant populations and (2) in the case of diplospory (often with autonomous endosperm development. more or less obligate and combined with the occurrence of triploid apomicts) as contrasted to apospory (often with pseudogamy, facultative apomixis. and rareness of triploids). Theories concerning the origin of apomixis will be the subject of a future paper.
Mathematical models of the evolution of mating systems, including gametophytic apomixis. have been proposed by MAYNARD SMITH (1977) and LLOYD (1977). Models for genetic variability in parthenogenetic animal populations have been constructed by ASHER ), NACE et al. (1970), and TEMPLETON and ROTHMAN (1973. Concerning plant populations, a model for estimating the level of apomixis was presented by MARSHALL and BROWN (1974). and a single-locus model for the maintenance of genetic variation, by MARSHALL and WEIR ( 1979).

Introduction of apomixis in sexual crops
Finally, the possibility of inducing permanent apomixis in sexual crops to fix heterosis will be regarded. Apomixis may be induced in any of the following ways: (1) by selection of and crossing between lines with high frequencies of haploid formation and with high frequencies of formation of unreduced embryo-sacs; (2) by combination of suitable induced or spontaneous mutations with strong effects upon the reproductive system to give an "artificial" apomict (of course, (1) and (2) may be combined in various ways); (3) crosses between distantly related sexual forms followed by selection for apomixis in the hybrid offspring: (4) crosses with related apomictic species (for references, see ASKER 1979).
Work along these lines is under way, but difficulties have to be overcome, for instance the reduced fertility occurring in (4), and lack of good screening techniques for reproductive mutants in (2). So far, only preliminary results have been obtained. With improved basic knowledge of the regulation of apomixis, future progress in this field will be achieved.