by Douglas Allchin
Gregor Mendel is a venerable hero of biology (full-size image [190K]). How could he have made a mistake?
The problem is the concept of dominance. Briefly (and I will elaborate further on my provocative claims below), it explains nothing. It is also misleading as a model of gene expression in diploid organisms. Moreover, the term confuses students by priming or reinforcing many misconceptions about genetics. We need to purge the concept of dominance from genetics. Mendel was mistaken. But was Mendel to blame? Ultimately, the entrenched error provides an excellent lesson on the history and nature of science.
Here is a simple question that a student might ask you (or has asked you, or even that you might have asked yourself): how does dominance "work"? That is, by what molecular mechanism does a gene become dominant? Answer: none. There is no mechanism of dominance. The term certainly evokes an image whereby one gene "dominates" another, perhaps by inhibiting transcription of the recessive allele or by actively suppressing its expression in some other way (see, e.g., Levin's popular textbook, Genes VI, page 62). But this does not happen. Genes are not social primates! They do not exhibit interactive behavior or political hierarchies (though the mere term "dominant" elicits such biological connotations). Dominance is not a form of gene regulation (as portrayed in the operon model, for example). In fact, it is not a specific process at all.
What happens, then, when one gene appears to be dominant, another recessive? Just this: somewhere, the unfolding paths of expression of the dominant and recessive alleles diverge. Sometimes, the process is truncated and the alternative phenotype is "incomplete" or is shadowed by the other allele. For example, there may be an early stop codon that ends transcription prematurely, or a missing stop codon that generates an extended mRNA (e.g., blood type A2). Or perhaps a mutation means that an intron is not excised or the exons are rearranged (alcaptonuria #4). Or the protein is translated, but the variant sequence does not fold into the same shape and the protein does not fit in its receptor (insulin, in the case of diabetes) or it does not catalyze reactions (lactose intolerance). Maybe it does not form multimers as would proteins from other alleles (sickle cell anemia). If the protein served in the transmission of neural impulses at the synapse, behavior could be affected, on yet another functional level (e.g., modified serotonin receptors in some cases of depression). At still other times, a mutation might yield a protein that can subvert the function of the "normal" protein. When a single copy of such a gene is sufficient, then the mutant appears dominant, even if the "normal" (now recessive) protein is produced as usual (for example, in osteogenesis imperfecta). In all these cases, the recessive gene is expressed. The consequences of that expression may not be noticeable, however, at the macroscopic or gross organismal level. Still, geneticists can often detect the presence of alternate proteins, and use them to identify heterozygous carriers of a particular allele. It is simply a mistake to imagine that a recessive allele "recedes," is suppressed by the dominant allele, or is not expressed at all. Asking how dominance works thus ultimately provokes skepticism about dominance itself.
In a dominant/recessive gene pair, both alleles are indeed expressed. The appearance of dominance reflects the coupled expression of two alleles, where the contribution of one allele is masked or seems relatively insignificant in the context of its homolog. There is no direct interaction between dominant and recessive genes. And in that sense there is no "dominance" as implied by the nomenclature. Developmental explanations of the classic cases of wrinkle-seeded peas and white-eyed fruit flies provide further vivid examples (see Guilfoile 1997). The wrinkled seed coat of some peas, for example, is an indirect result of starch conversion, sugar concentration, osmotic potential, water imbition and seed drying. But here a single copy of the key starch-conversion enzyme can catalyze the reaction, yielding smooth seed coats. "Wrinkled-seed" appears recessive, even though its non-enzymatic proteins are present. In red-eyed fruit flies, a membrane protein imports pigment precursors into the cell, while in their white-eyed cousins another version of the same protein does not. Though two genes are always expressed, white eyes ("recessive") result if proteins from both fail. These developmental explanations—and thinking in terms of coupled genes—say everything we need to know. The concept of dominance adds nothing.
For some, our late 20th-century molecular insights may seem like valuable updates that extend but do not fundamentally challenge Mendel's model. But even the basic concept of dominance is ill-framed. Consider, for example: does dominance refer to the phenotype or the genotype? —Inheritance patterns or mechanisms of genetic expression? Sometimes we refer to the dominant trait, sometimes to the dominant allele or gene—muddying the genotype/phenotype distinction we try so hard to instill in students. Of course, this occurs because we ourselves are not clear about what dominance means. The problem may be illustrated in how textbooks define dominance. They typically introduce the concept, not in clear statements, but by ostension or exemplification. For example, they use such conditional phrases as "when traits combine in hybrids, the visible trait is called dominant." This way of "defining" dominant traits indirectly underscores that little more than labeling is at work (note also that dominance is rarely characterized as a noun). Yet before long, we begin to attribute the observed phenotype to dominance, as though the descriptive label identified an underlying causal property. Our characterization of dominance soon becomes circular: a trait is dominant when we observe it in the hybrid, and the hybrid exhibits this trait because it is dominant. Dominance can describe a certain types of hybrid results. It explains nothing. It does not explain how only one parental trait becomes visible in the offspring, for instance, nor why this bias should occur for some traits and not for others.
Most problematic, dominance is misleading as a model for the expression of two alleles in diploid organisms. Dominance is not the norm. That is, it is not the most frequent pattern. This was noted at the outset, early in the century, when Mendel's work first received renewed attention. As early as 1907 Hurst observed that incomplete dominance is twice as frequent as complete dominance (Darden 1991, p. 68). A more recent estimate (Rodgers 1991, p. 3) also suggests that fewer than one-third of human clinical genetic conditions follow the dominant-recessive rule. Here is a different kind of "Mendelian ratio," although no deeply systematic study seems to document it. An indirect and informal measure of the prevalence of dominance, though, might be the scarcity of good textbook examples of dominant traits in humans. To illustrate "Mendelian" inheritance, we often appeal to "attached earlobe," "hitchhiker's thumb," "short little finger," "widow's peak," "woolly hair," "crumbly earwax," "tongue-curling," "PTC-tasting." These are trivial. They hardly reflect important dimensions of human genetics. Nearly all the interesting or significant cases have more complex stories. Online Mendelian Inheritance in Man (OMIMtm; 1997), the major reference for human genetics, for example, discontinued classifying traits as dominant and recessive in 1994. Dominance is far from universal.
Nor is dominance basic or foundational by being simplest or exhibiting the fewest assumptions. In the simplest model, all alleles would perform similarly: each allele would be expressed. Hence, if two alleles were present, both would be expressed (as indeed they are). Dominance requires an additional assumption about the relative behavior of the two alleles. "Incomplete dominance" or "codominance" are not exceptions to complete dominance. Rather, dominance is the "exception" to both alleles contributing equally to the phenotype. Our standard thinking needs a gestalt switch, exchanging foreground and background. Mendel's concept of dominance is not the best benchmark.
Mendel in the Classroom
For teachers, the upshot of all these misconceptions about dominance is profound indeed. They underly the most common and frustrating problems encountered in teaching genetics. The notion that some traits are more "powerful"—or even the very term 'dominance'—precipitates multiple student conceptions, all too familiar to any teacher of genetics (see also Donovan 1997):
None of these claims is necessarily true. Some are false. All are misleading. One might wonder, therefore, why students (and, in some cases, prominent biologists historically!) so readily and commonly assume(d) their validity. Moreover, these preconceptions and images are notoriously resilient—difficult for instructors to rectify even when they note and address them explicitly in class. Note, though, how the vernacular meaning of dominance, where one thing "dominates" over another, percolates through every misconception. In Metaphors We Live By, Lakoff and Johnson (1980) describe how our thinking is shaped by the words we choose and their meanings in other contexts. "Dominance," though it may be well defined strictly within genetics, unavoidably carries with it other meanings and non-biological connotations. Students reason about the prevalence of alleles, the interaction of genes, heritability, reproductive fitness, normality and gender using the language available. Instructors should be concerned that these misconceptions will confound discussions outside biology regarding genetic counseling, say, or social policy stemming from the Human Genome Initiative, and revise their teaching and terminology accordingly.
Mendel was right about segregation and recombination (and about independent assortment). Perhaps most importantly—and most counterintuitively—his scheme identifies genetic make-up as dual. We are diploid organisms. We do not have traits, really; we have pairs of traits. Almost as important, Mendel inferred that when we reproduce, genes/traits segregate from their union and are inherited in equal proportions. Certain traits do not gain an edge over others in being inherited. We often say, casually, that we inherit our mother's eyes or our father's stature, say. But this is wrong. As Mendel noted, we inherit genetic material from both parents equally. But then the Mendelian model errs by framing the expression of traits as an either-or question. The principle of dominance is a mistake. But it errs, fortunately, independently of Mendel's other important insights.
Mendel's concept of dominance further implies that all traits come in pairs (one dominant and one recessive) and that only one element of the pair can have "power." This may seem like an arcane biological quibble. Not so. By portraying nature as structured in simple dualities, the bi-allelic model reinforces cultural tendencies to interpret social issues in bipolar terms (or, conversely, perhaps the pervasive image of two-sided competitions in society makes it more difficult for us to notice the flawed bi-allelic assumption). Further, the notion that only one element of the pair can actually be expressed gives a "naturalness" to casting such issues in either-or terms. Imagine the political implications—from Congress to marriages to sports—of assuming that there can only be two options and that one voice must totally eclipse all others (that is, that equal voice or synergy are "unnatural" alternatives to consider). The more we study the genetic make-up of organisms, the more we appreciate the incredible diversity of alleles, even at a single locus, and their sometimes subtle interactions. By labeling codominance and multiple alleles as "non-Mendelian," for example, we make a clandestine normative claim about what is standard and what exceptional. We should recognize that the concept of dominance carries with it implicit cultural overtones—and endeavor to ensure that science does not participate in promoting or justifying such biases.
An Alternative Mendelian Model
Fortunately, a prospective alternative, dominance-free model of genetics is readily available. The substitute is our current treatment of multiple alleles, exemplified by standard discussion of ABO blood types. First, many possible alleles are acknowledged: A, B and O, commonly, and others (including a remarkable form, cis-AB, that shows dual enzymatic function!; OMIM #110300). However, due to the sexual nature of human reproduction, each individual carries just two alleles, one from each parent. The relevant blood-type phenotypes originally became evident from agglutination of blood mixtures, indicating that red blood cells can have specific antigenic properties. The now well-known safe transfusions between blood types indicate that the O allele does not generate specific antigens, while A and B produce distinct antigens. Hence, an O allele in combination with A or B becomes functionally "invisible" when considering transfusions, by virtue of producing no antigen. (Note, though, that no one sees the need to call A or B "dominant" to O.) Nor is there any confusion when, by contrast, A and B alleles combine: each contributes an antigen to the red blood cells. The hybrid is a hybrid. Furthermore, one can infer possible allelic make-ups from observed blood type, as well as predict possible phenotypes of prospective offspring from two parental genotypes. And as always, in reproduction, alleles segregate and recombine. In this model, we do not establish narrow standards that then require us to refer to supplemental concepts such as incomplete dominance, codominance, multiple alleles, penetrance or expressivity. One set of concepts and language can embrace both the limited domain of dominance and its "exceptions." When dominance dissolves, so, too, do all these secondary concepts. No more "non-Mendelian" genetics. The discourse of blood types illustrates a modified Mendelian model that can be applied uniformly and consistently across all cases. We should welcome the conceptual economy.
Perhaps we can see our way clear of this problem in the future. But, looking backwards, how did we end up in this terrible conceptual tangle? How did the mistaken concept of dominance become so entrenched? And how could Mendel, whom we parade before students as the quintessential scientist, have led us so astray? This case offers further valuable lessons about the process of science.
In unraveling this knot, we must first be careful to tease apart Mendel from the sometimes mythic lore about Mendel (Sapp 1990). Mendel's original paper, for example—so clear by modern standards—had quite different meanings in 1865: we must interpret it in its own historical context. In addition, we should remember that Mendel was looking for "laws of hybridization"; he did not present his results as general principles of heredity, though that is the typical story found in textbooks today.
Mendel's Original Paper
Mendel introduced the term dominirende (translated variously as 'dominating' or 'dominant') in the context of his "Experiments Concerning Plant Hybrids" (1866/1966). He crossed peas with different dichotomous characters, but the hybrids each exhibited only one of the parental characters, "which are transmitted entire, or almost unchanged in the hybridization" (§4; see also §11). That may strike us as significant, but Mendel's contemporaries would have seen this as illustrating the widespread principle of prepotency, whereby certain traits were predisposed to be inherited more strongly in hybrids. Many related this to the sex of the parent. By doing reciprocal crosses, however (as others had), Mendel was able to echo the "interesting fact" that "it is immaterial whether the dominant character belongs to the seed plant or to the pollen plant" (§4). He also argued convincingly that traits are passed on with equal frequency in the gametes; if the frequency of a trait varies in offspring, then other factors, related to expression, must be involved. In correcting these misconceptions, though, Mendel also laid the seed for another. That is, Mendel linked the property of recurring traits to the trait itself. He did not consider fully the role of development or the context of the other hybrid trait. Others would follow his lead in thinking of dominance causally and focusing on the traits. Mendel succeeded and erred simultaneously.
Mendel took further pains to emphasize that the dominant characters "in themselves constitute the characters of the hybrids," with no ostensible contribution from the recessive characters which, though present, are "latent" or "withdraw." The recessive characters do not just partially disappear; they "entirely disappear" (§4). For Mendel (as for others to follow), the dominant characters wholly eclipsed the corresponding recessive characters. "Transitional forms were not observed in any experiment," he stressed again (§5). Why was Mendel so preoccupied with this claim? Well, he knew that in successive generations both the recessive and the dominant traits would re-emerge as "pure" in true-breeding offspring. Again, that may not surprize us, but Mendel's contemporaries widely believed in blending inheritance (recall Fleming Jenkin's scathing criticism of Darwin!). For Mendel, intermediate forms would have suggested that traits had mixed irreversibly, whereas Mendel's mathematical scheme depended on the traits being preserved undiluted. Here, Mendel conflated a lack of dominance (phenotypically) with lack of segregation (genotypically). Nothing about intermediate forms (cases of "incomplete dominance," in modern terms), though, precludes segregation or the "purity of the genes." Pink flowers, for example, do not forever yield a pink lineage; red and white flowers may reappear when "red" and "white" alleles segregate and recombine according to Mendelian principles. It may be that Mendel's recurring binary theme, which helped structure his whole system, facilitated this confusion. Even William Bateson, who helped revive and champion Mendel's work in the early 1900s, was confused by the same issue for a short while (Olby 1987, pp. 414-417). Ultimately, one may appreciate how Mendel's own reasoning may have relied on the concept of dominance (with either-or traits) though logically his conclusions did not(!).
Even more important, though, Mendel himself recognized that not all traits are expressed in dominant and recessive pairs. To his credit, he acknowledged that dominance was not the exclusive norm. Even before introducing dominant traits he noted, for example: "with some of the more striking characters, those, for instance, which relate to the form and size of the leaves, the pubescence of the several parts, etc., the intermediate, indeed, is nearly always to be seen" (§4). Later he commented: "as regards flowering time of the hybrids, . . . the time stands almost exactly between those of the seed and pollen parents" (§8). Mendel certainly found in the years immediately following his work on peas that his results from Pisum did not generalize to Hieracium, or hawkweed (Mendel 1869). For Mendel this merely meant that his law of hybrid development applied only to "those differentiating characters, which admit of easy and certain recognition" (1866, §8). Other characters followed another, possibly different rule or law. Dominance, for Mendel at least, had a limited domain.
Mendel's work became a guide, of course—almost a touchstone—for the pioneers of the new science of genetics at the turn of the century. His particular concept of dominance, though, was not uniformly embraced at the outset. Even Bateson, perhaps Mendel's strongest advocate, could not endorse a principle or law of dominance. Bateson's own work on inheritance in poultry showed that traits "mixed" in hybrids, though the traits still segregated neatly in offspring, according to Mendel's model. "The degree of blending in the heterozygotes," Bateson declared, "has nothing to do with the purity of the gametes" (p. 152). (Note again that intermediate forms are denoted erroneously as a form of "blending inheritance.") Bateson's example of Andalusian fowl—blue-grey hybrids of black and white parents that formed a 1:2:1 ratio in the F2 generation—soon became a classic case, cited in textbooks throughout the century (for example, yours?).
Others objected to dominance as a universal feature of inheritance. Case after case of intermediate form was cited. For most informed breeders and geneticists, characters that differentiated into only two forms, such as Mendel's tall/dwarf or green/yellow, were relatively rare. In carrying forward the legacy of Mendelism, textbooks in the ensuing decades and beyond have continued to reflect ambivalence towards dominance as a "law" or basic model. For example, an early 1906 text by Lock states that dominance is not universal. Morgan's 1915 synoptic text was equally skeptical:
Whether a character is completely dominant or not appears to be a matter of no special significance. In fact, the failure of many characters to show complete dominance raises a doubt as to whether there is such a condition as complete dominance. (see Darden 1991, p. 72)
Likewise, a 1921 text lauds Mendel's landmark discovery of dominance, then adds ironically, "of course breeding is not so simple as this, and some characteristics do blend or average in the hybrids" (Moon 1921, p. 543). A 1933 zoology text, too, follows its description of dominance with a cautionary note: "dominance and recessiveness do not, however, characterize all cases of inheritance" (Curtis and Gurthrie 1933, p. 184), and then introduces the examples of Andalusian fowl and pink four o'clock flowers. In 1969, we find another text carefully detailing "Mendel's law of dominance," then citing the very same two examples, noting that:
Since Mendel's time, we have found that the law of dominance does not always hold. . . . It is clear that we cannot speak of a "law" of dominance even though dominance occurs frequently. (Kroeber, Wolff and Weaver, 1969, pp. 412-412)
Could more equivocation be found?: dominance is both a law and not a law. By the 1990s dominance and recessiveness had retreated to the status of a "feature" in one standard genetics text (Russell 1992, p. 41).
Despite the ambivalence, dominance continues to be preserved as an essential or core fixture of genetics, consistently introduced before it is dismissed or qualified by any exceptions. Why? Why has dominance persisted as a standard or model, even if in disrepute? Whereas Mendel associated dominance with segregation, we now associate dominance with Mendel himself, as a scientist of mythic proportion (see, e.g., Brannigan 1981; Sapp 1990). Nearly every introductory biology textbook introduces Gregor Mendel with a picture and supplemental comments. They implicitly portray him as an exemplary scientist. He worked alone in an Austrian monastery: scientists modestly seek the truth; they do not ambitiously pursue fame or wealth. Mendel used peas; scientists choose "the right organism for the job." He counted his peas: scientists are quantitative. He counted his peas for many generations over many years: scientists are patient. He counted thousands and thousands of peas: scientists are hard-working. After all this, Mendel was unfairly neglected by his peers, who failed to appreciate the significance of his work, but was later and justly "rediscovered": eventually, scientific truth triumphs over social prejudice and ignorance. Above all, Mendel was right. By all these measures, Mendel is a biological hero, a model for aspiring students. Because dominance was part of Mendel's original scheme and, at the same time, we honor Mendel almost religiously, we do not exclude dominance from basic genetics. Dominance has become entrenched in the romantic lore of science and the pantheon of Mendelism.
So, ultimately, Mendel's mistake is our mistake, too. We have dramatically transformed the concept of dominance from Mendel's originally modest label, especially in giving it broad generality and substance. For example, Mendel only used the adjective, dominirende; we created the noun, dominance, thereby reifying as an abstract property what is fundamentally a relationship between two particular alleles. In addition, we have helped shape the image of Mendel as a scientist. A historian once asked—not rhetorically—"was Mendel a Mendelian?" The answer, of course, was "no." Mendel has been so interpreted and reinterpreted that the principles in textbooks that now bear his name, though derived from Mendel, are not Mendel's alone.
Science may be a self-correcting process, but finding and fixing errors involves work. Solving Mendel's mistake is up to us. Still, this process is problematic because of the "halo" around Mendel and his work. Mendel may not have been canonized by the Church, but he has certainly been canonized in the field of genetics. Our notions of scientists and their work fall into some fairly narrow conventions, as noted above. Foremost in this case, we are not accustomed to acknowledging that scientists, especially great ones, can make mistakes. But even Nobel Prize winners have erred (Darden 1998). Admitting Mendel's flaws in no way detracts from his achievement. Mendel was, after all (like all scientists), human. In fact, to understand the limits and flaws in his work is to make him more real, less a caricature. That could also contribute to a more honest portrait of science for students. Mending Mendel with a bit of reflection can thus also provide a lesson in the process of science.
Acknowledgements. My appreciation to Louise Paquin and Bob Olby for their support in developing the themes presented here.