Am. Sci. 75: 44-50. (1987)
The effects of pollen competition.
Mulcahy, D. L., and G. B. Mulcahy.

Figure 1. In this photograph of the stigmatic lobes of a garanium (Geranium maculatum), pollen tubes can be seen extending from three germinating pollen grains on the left down toward the ovary of the plant. Other pollen grains are visible on adjoining lobes. Pollen tubes represent one of the most rapidly growing systems in the plant kingdom. In the geranium, a pollen grain 0.1 mm in diameter produces a pollen tube approximately 1 cm inlength in 20 minutes. The photograph was obtained by rendering the plant tissue transparent, staining it, and then irradiating it with ultraviolet light; the yellow-green light is emitted by cellose, a carbohydrate in the pollen and pollen tubes. (Photo by authors.)

If you were to be asked to provide a superb demonstration of rapid Darwinian selection, our advice would certainly be to choose a microorganism for in that group it is possible to create new adaptations in a matter of days. The sources of this microbial adaptability are well known: haploidy—the possession of a single set of chromosomes—and large population sizes.

The haploidy that characterizes microorganisms is particularly important in the case of newly acquired recessive genes, since these, almost by definition, will be quite rare. In a diploid organism, one possessing two sets of chromosomes, a new recessive gene will almost certainly be paired with a dominant gene and thus not expressed. With haploidy, however, newly acquired genes will be expressed immediately, and deleterious qualities will be eliminated and beneficial ones established through the process of natural selection. In addition to rare alleles, mote complex, multilocus adaptations are also greatly influenced by haploidy. Zamir (1983) has pointed out that if each locus carries two alleles, a complex of 12 coadapted loci can be arranged in all possible combinations with 4,096 haploid individuals. With diploidy, however, 16,777,216 individuals would be required to test all possible combinations.

Large population size, the other key to microbial adaptability, is advantageous in that with it, so-called rare events paradoxically become inevitable. For example if the probability of a particular genotype occurring is one in a million, we would expect to find, on average, ten such genotypes in a population of ten million individuals. The power of large population sizes in combination with haploid genetic systems is only too well known to the human species. We can assume that microorganisms pathogenic to man will evolve resistance to antibiotics very rapidly, and crop pathogens generally need only a few seasons to breach the defenses of newly resistant crop varieties, despite the best efforts of plant breeders.

Considering the effects of haploidy and large population sizes in microbial populations, it is intriguing that the same features are found in the pollen of the angiosperms, the flowering plants that serve as the virtually exclusive basis of our agricultural systems, not to mention much of the vegetation of the earth. In contrast to gymnosperms, which produce naked seeds not enclosed in an ovary, the angiosperms are characterized by a closed ovary. In angiosperm reproduction, pollen grains germinate on stigmatic surfaces within the flower, producing pollen tubes that transport sperm cells to the ovules (Fig. 1). Like microbial organisms, pollen grains are haploid and are produced in vast numbers. Unlike the exclusively haploid microbes, however, pollen grains represent only a very small part of a life cycle. The diploid portion of the angiosperm life cycle, called the sporophyte, predominates, encompassing the stages from zygote through mature plant (Fig 2).

The question then becomes: As a very minor portion of the life cycle, albeit haploid and numerous, can pollen nevertheless provide a significant adaptive mechanism? The renowned British biologist J.B.S. Haldane (1932) was among the first to suggest that it could not. He argued that the pollen grain had "a physiology all its own, influenced by special genes" (pp 123-24), and expressed concern that uncontrolled pollen competition might spread disadvantageous changes. It was then suggested that natural selection should favor nonoverlapping genetic systems—that is, one set of genes expressed in the pollen and another expressed only in the sporophyte. In this way the sporophyte could be protected from possibly detrimental effects of pollen competition.

The concept of "special" pollen genes was reinforced by the discovery that the gene for waxy expressed in the pollen and seed endosperm of corn is not expressed in the sporophytic portion of its life cycle (Bryce and Nelson 1979). Another fact also appeared to support the genetic quiescence, or uniqueness, of pollen genes: when a pollen source is heterozygous, the two gamete types it produces usually seem to be transmitted to the next generation with approximately equal frequency. In fact, this apparently nonselective—that is, random—fertilization serves as the basis for Mendelian genetics. This further suggested that pollen tubes with different genotype's might not in fact differ in growth rate, thus implying that the segregating genes were not expressed in the pollen.

Taken together, such considerations led to the view that the genotype of pollen in the flowering plants was, either by chance or because of selective pressures, largely unexpressed in, or at least independent of the genotype of, the sporophyte. That view has now changed. The following paragraphs explain why, and explore important implications of recent findings both for crop improvement and for our broader view of angiosperm success.

How common is pollen competition?

Figure 2. The life cycle of the angiosperm begins when pollen from the anther of the mature plant germinates on the stigma, producing a fast-growing pollen tube that penetrates down the style toward the ovary. when the pollen tube enters an ovule within the ovary, oue of the two sperm cells it carries fuses with the egg, producing a diploid zygote—the first cell of the sporophytic generation. The other sperm cell fuses with the polar nuclei to produce the endosperm, a nutritive material for the embryo. Only the pollen, pollen tube, sperm, egg, pollen nuclei, and embryo sac, shown here in green, are haploid; the rest of the cycle, extending from zygote to mature plant, is diploid. The diploid portion of the life cycle, known as the sporophyte, thus predominates.

Even granting that pollen grains are produced in much greater numbers than are ovules, we are still not justified in presuming that pollen competition is either frequent or intense. Certainly insects devour a large pro portion of these grains; and even when the number of pollen grains successfully reaching a stigma greatly exceeds the number of ovules avail able for fertilization, it is still possible that little or no pollen competition occurs. For example, assume that pollen tubes are able to reach the ovules within one hour of the time the pollen grains land on the stigma—a reasonable estimate for some species. Assume also that the ovary contains 10 ovules. If the first pollinator to visit that flower deposits 10 viable pollen grains on the stigma and the next visit does not come within an hour, the first 10 pollen tubes will experience no competition whatsoever, even if the second visit deposits many hundreds of pollen grains.

Thus, despite all its potential, pollen competition could be quite rare in natural populations. In fact, Bierzychudek (1981) found that the number of seeds produced by several species of spring ephemerals seems to be limited by the amount of pollen placed on their stigmas, suggesting that no pollen competition was occurring. Snow (1982) reached a similar conclusion in her study of pollen competition in Passiflora vitifolia, a passion flower found in Costa Rica. However, in a species of geranium, Geranium maculatum, it is apparent that a significant degree of pollen-tube competition does take place (Mulcahy et al. 1983). In this case, pollen tubes that were successful in reaching unfertilized ovules exhibited growth rates 34 to 41% higher than those of an unselected population of pollen tubes. In the best review of the subject to date, Snow (1986) surveyed the available data and concluded that in 7 species studied, pollen was usually or sometimes a limiting factor in seed production. In 17 other species, pollen was non limiting, and thus the potential for pollen competition exists there. We should now consider a second question which must be answered to clear the way for an understanding of the potential consequences of this competition.

There can be no response to selection without genetic variation, and thus we must next ask to what extent genes are expressed, and therefore exposed to selection, in the pollen. If we consider the gene for waxy in corn, or the genes that code for alcohol dehydrogenase (Schwartz and Osterman 1976) or beta-galactosidase (Fig. 3), also in corn, we see that at least some genes are expressed in pollen. These examples are known to us only because each of the genes in question includes null-alleles, that is, forms which fail to react with stains for the normal gene product. This allows clear visualization of two classes of pollen.

However, such convenient categorization of pollen is quite rare. Most allelic variants at a locus produce active gene products that differ from each other qualitatively rather than quantitatively. For these more typical variants, other means of detection must be used. For example, electrophoresis can be employed to separate molecules on the basis of size, charge, or a combination of the two (Fig. 4). Studies using this technique have indicated that at least 44% of the acid phosphatases present in pollen are the product of genes expressed in the pollen (Miller and Mulcahy 1983). Since this method reveals gene expression only when the loci in question are heterozygous, 44% is a conservative estimate of the fraction of acid phosphatases controlled by genes expressed in the pollen itself.

Figure 3. Early evidence of gene expression in pollen was provided by the gene that codes for beta-galactosidase activity, which can be identified by staining. Here pollen from corn (Zea mays) is seen segregating into two classes. Pollen grains containing beta-galactosidase are green, whereas pollen grains lacking this enzyme are yellow. (Photo courtesy of M. B. SIngh and R. B. Knox.)

A more general method of testing for gene expression in pollen has been developed by Tanksley and his colleagues (1981). It is well known that pollen is rich in many gene products (Brewbaker 1971). But are these simply loaded into the pollen by the sporophytic pollen source, or are they derived from genes which are transcribed and translated in the pollen itself? Tanksley's method of testing makes use of the fact that some enzymes, termed dimers, are composed of two separate gene products. If the dimeric enzymes from individuals which are homozygous for different alleles are compared, it will be seen that each individual carries a single dimer, one corresponding to the particular locus for which that individual is homozygous. If a hybrid between individuals homozygous for different alleles is produced and analyzed, this heterozygote will be found to contain not only both parental dimers but also a third enzyme. This third enzyme is a hybrid dimer, made up of the combined gene products from each of the two parental alleles.

Tanksley and his co-workers reasoned that if proteins contained in pollen were derived from the diploid pollen source, the pollen of heterozygous plants should, like the sporophytic tissues, contain both of the parental dimers plus the hybrid dimer. On the other hand, if the dimeric proteins contained in the pollen were the result of gene expression in the pollen itself, no hybrid dimers should be found. (Each pollen grain will contain one allele or the other, but not both.) Examining 30 different genes in the tomato (Lycopersicon esculentum), they found that the enzymes present in pollen were exclusively homodimers. The complete absence of hybrid dimers proved that the gene products present were the result of gene expression in the pollen itself.

With several studies indicating that genes are indeed expressed in the pollen, we might reasonably expect that pollen would respond to selection for a variety of qualities. In fact, this is exactly what is reported in the literature. Selecting for pollen tubes that grow rapidly in a particular stylar genotype results in subsequent generations of pollen tubes that exhibit significantly increased growth rates in those styles (Jones 1928; Johnson and Mulcahy 1978; Ottaviano et al. 1983). Thus we conclude that pollen competition not only can but does affect the pollen.

Pollen competition and the sporophyte

To what extent are the genes expressed in the pollen also expressed in the sporophyte? The answer is by no means obvious. In corn, for example, the gene for waxy is expressed only in the pollen and in the endosperm of the seed, as noted above, while the gene for alcohol dehydrogenase is active throughout the plant's life cycle. One way to approach this question is to select in one phase of the life cycle and look for changes in the other. If the genes expressed in the pollen have no members in common with those expressed in the sporophyte, then selecting among one set of genes should have no effect upon the other. When this method was first applied by Ter-Avanesian (1949) and Matthews (see Lewis 1954), they found that reducing the intensity of pollen-tube competition increased the degree of variation exhibited by the subsequent sporophytic generation. Apparently pollen genotypes that grew slowly (and were thus eliminated in normal pollen competition) gave rise to plants that deviated more widely from the norm than did other plants. An obvious explanation for this is that some of the genes that are expressed, and thus exposed to selection during pollen-tube growth, are expressed also in the sporophyte.

Figure 4. Differing gene products are clearly visible in this microelectrophoresis of single pollen grains from the F1 hybrid Cucurbita texana x C. palmata. The pollen grains, seen as a line of dark circular shapes at the top of the figure, were placed on the surface of gel and then separated by a technique called isoelectric focusing, which operates on the basis of charge differences. Some of the pollen contained a protein which migrated only as far as the upper of two horizontal bands of shapes at the bottom of the figure; others, carrying a different gene, released a protein that traveled to the lower band. The gel is 6 cm wide and has been stained to demonstrate acid phosphatase activity. (Photo courtesy of J. Miller.)

However, these early studies included one confounding factor: the intensity of pollen competition was modified by varying the amount of pollen used in pollinations. While certainly effective in increasing or de creasing the intensity of pollen-tube competition, this method often introduces significant variations in the number of seeds produced. In itself, reduction in seed number would be trivial. However, in many cases, the few seeds produced are unusually large. Could the results of Ter-Avanesian and Matthews be explained as artifacts of variation in seed size, better zygote or seedling survival with limited pollinations, or other factors?

To answer this question, we adopted a method which was first conceived by Correns (1928), one of the rediscoverers of Mendel's laws. Correns reasoned that the greater the distance traveled by pollen tubes, the greater the opportunity for rapidly growing pollen tubes to surpass more slowly growing tubes. We chose to work with Dianthus chinensis, a commonly cultivated carnation, because its elongated stigma allowed us to pollinate some flowers at the tip of the stigmatic surface and others at the base, thus lengthening or shortening the distance the pollen tubes must travel (Fig. 5). Pollination at the tip should provide greater opportunity for pollen-tube competition than does pollination at the base, but both should result in a full set of normal seeds. This indeed turned out to be the case; seed size did not differ significantly in the two circumstances. When we planted the seeds, moreover, we found that plants resulting from tip pollination germinated more rapidly and grew faster than did plants from base pollination (Mulcahy and Mulcahy 1975). Pollen competition was thus expressed in the next generation even when seed size was held constant.

In a later study of Dianthus chinensis (McKenna and Mulcahy 1983), it was found that plants produced by tip pollination, and therefore under conditions of intense competition, were better able to compete against ryegrass than were plants resulting from base pollination. Similarly, Zamir and Vallejos (1983) found that low temperatures during pollen-tube growth favored greater transmission of pollen genotypes carrying sporophytic markers associated with cold tolerance.

Qualitative characteristics can also be manipulated by pollen selection. In a study using plants which were either sensitive to or tolerant of zinc or copper, it was demonstrated that pollen from zinc- or copper-tolerant plants is itself relatively tolerant of zinc or copper (Searcy and Mulcahy 1985). Furthermore, plants that are heterozygous for tolerance produce two classes of pollen: one sensitive and one tolerant. If these segregating plants are watered with nutrient solutions containing high levels of the heavy metal in question, a large percentage of their pollen grains abort. Applying the pollen that survives this treatment to metal sensitive parents results in a generation exhibiting many more metal-tolerant individuals than expected. Apparently the pollen grains that are killed by zinc or copper are those that carry the gene for sensitivity to the heavy metal. Pollen selection could thus provide a mechanism for developing genotypes tolerant of heavy metals and probably of other stresses as well.

Figure 5. In this dissected specimen of the carnation Dianthus chinensis, the hairlike covering of the white style at the center of the flower represents an unusually extended stigmatic surface receptive to pollen tube penetration. This characteristic allowed pollen competition to be studied directly by varying the area pollinated and thus the distance pollen tubes must travel to reach the ovary. Plants produced by pollination at the tip of the style germinated faster, grew more rapidly, and were better able to compete against ryegrass than plants produced by pollination at the base. (Photo by authors.)

This sharing of qualities by pollen and sporophyte holds some potent practical implications. For example, Ottaviano and his colleagues (1980) have learned that the growth rates of pollen tubes through the styles of corn can be used to predict the total weight of the kernels that would be produced on the resulting hybrid plants with a correlation coefficient of 82.3%.

In addition to this strong circumstantial evidence, which suggests that a substantial portion of the genes expressed in the sporophyte are also expressed in the pollen, several biochemical and molecular investigations point to the same conclusion. For example, in the study described earlier, Tanksley and his co-workers (1981) found that in the tomato about 60% of the sporophytic structural genes are expressed, and thus exposed to selection, in the pollen. One of the 30 genes they studied, however, was expressed only in the pollen.

Recently, Willing and Mascaren has (1984) have developed an even more powerful method of studying gene expression in pollen. Extracting messenger rna from the pollen of spiderwort (Tradescantia paludosa), they used reverse transcriptase to manufacture a copy of all the genes expressed there. This enabled them to learn that there are about 20,000 different genes expressed in pollen, compared to 30,000 in the stem. Comparing the genes expressed in the pollen with those expressed in the stem, they found, as did Tanksley's group, that about 60% of the genes expressed in the sporophytic tissue had a homologue among the genes expressed in the pollen. They also reported that a small proportion of the pollen genes—5 to 10%—were not expressed in the sporophyte. Finally, in a large ongoing study of corn (Zea mays) at the University of Milan, Sari-Gorla and her colleagues (1986) have found a 73% overlap between the sporophytic and pollen genomes.

The conclusion seems inescapable: a large sample of the structural genes expressed in the sporophytic portion of the life cycle are also expressed in the pollen. While this finding is interesting in its own right, its major significance arises from the fact that, since 60% of the structural sporophytic genes will be exposed to selection in the pollen, the sporophyte is also subject to this unique selective system. Does this confer upon the sporophyte a degree of adaptability and versatility comparable to that exhibited by the microbial world? Presumably the possibility exists.

Angiosperms and gymnosperms compared

It has recently been discovered that in the Monterey pine (Pinus radiata), a gymnosperm and an important timber crop in Australia and New Zealand, 60% of the genes expressed in the sporophyte are also expressed in the pollen (G. Moran, pers. com.). On the basis of these data, it thus appears that the degree of overlap exhibited by the genomes of pollen and sporophyte in the gymnosperms equals that in the angiosperms.

Does this indicate that the sporophytes of both groups would benefit equally from the fact that 60% of the sporophytic genome is exposed to haploid screening? Not really. We know that the intensity of pollen selection is positively correlated with at least three separate factors: the distance competing pollen tubes must travel, the extent to which pollen grains outnumber available eggs, and the temporal pattern of pollen deposition. In the gymnosperms, pollen grains are passively transported to within a few pollen diameters of the developing egg. In the angiosperms, however, the closed carpels impose upon competing pollen tubes a journey of perhaps several hundred or even several thousand pollen diameters. This requirement for extensive growth provides ample opportunity for pollen competition to occur.

In angiosperms, moreover, insects are the primary pollen vector, whereas in the gymnosperms, pollen is dispersed by the wind. It is generally accepted that insects are by far the more efficient vector, depositing greater numbers of pollen grains for each available egg than does the wind (Whitehead 1984). In fact, a study of loblolly pine (Pinus taeda) showed that ovules received an average of only 3.97 pollen grains each (Bramlett et al. 1985). This might seem to represent somewhat of an excess. However, each pine ovule contains several functional eggs, so pollen competition might be quite rare. In the gymnosperms, as illustrated by this one species, each of the eggs might be fertilized, but only one results in a mature embryo. Much of the "quality control" thus seems to take place at the embryo stage rather than through pollen.

The temporal pattern of pollen deposition also differs with wind and insect vectors. In wind deposition, pollen grains tend to arrive singly and over long periods of time. The opposite is true in insect pollination, where large clumps of pollen are deposited simultaneously. The influence of these differences becomes obvious if we compare pollen competition to a marathon race. No one will deny that the marathon is a grueling test of a runner. Nevertheless, the race would be meaningless if contestants were assigned random starting times, with the first starting eight hours before the last. Thus in three major determinants of pollen competition—distance traveled, pollen quantity, and temporal pattern of pollen deposition—angiosperms exhibit a greater intensity of pollen selection than gymnosperms.

The rise of the angiosperms

Earlier suggestions that the effects of pollen competition were either neutral or negative now appear to have little substance. Haldane's concerns about unfavorable effects of pollen competition, while logical, seem unnecessary. The supposition that genes expressed in pollen would not be expressed in the sporophyte, based on the example of the gene waxy in corn, has proved incorrect. Even the question of the seemingly random transmission of alleles through pollen, not discussed here, has been resolved by the finding that this randomness may be more apparent than real (Mulcahy and Kaplan 1979).

This readjustment in our view of the effectiveness of pollen competition has interesting implications for the current picture of the rise of the angiosperms. Elsewhere it has been proposed (Mulcahy 1979) that the rise of the angiosperms could per haps have been aided by this difference in intensity of pollen competition between the angiosperms and the gymnosperm-like ancestors from which they descended. While we don't wish to exaggerate the significance of pollen-tube selection, this hypothesis has been supported by Crepet's recent finding (1984) that the inception of advanced insect pollination either immediately preceded or coincided with a major radiation of the flowering plants. Crepet further points out that paleontological data are consistent with neontological evidence suggesting an important role for advanced insect pollinators in establishing contemporary angiosperm diversity.

We suggest that one aspect of this role may have been that, through closed carpels and insect pollination, the angiosperms were endowed with a greatly enhanced pollen competition. This would confer upon 60% of the structural genes expressed in the sporophytic portion of the angiosperm life cycle a microbial-like system of screening. It would also allow, as has been empirically demonstrated, the facile development of characteristics such as in creased stress tolerance, improved competitive ability, and enhanced sporophytic vigor. This, in turn, would result in increased versatility and adaptability in the angiosperms. Could this have contributed to the rise of the angiosperms? We suggest that the possibility is certainly a real one.