Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

Within a decade of James Watson and Francis Crick’s announcement of the double helical structure of DNA, biologists, physicists, and chemists began to ask fundamental questions about how biological molecules had evolved. As proteins and later RNA and DNA sequences became the objects of evolutionary inquiry, the study of molecular evolution coalesced as a new discipline within biology bringing with it significantly new forms of data, inference, modeling, and concepts (Suarez and Barahona 1996; Dietrich 2008).

For many evolutionary biologists, molecular biology represented a mixed blessing. New techniques, such as electrophoresis, were greeted with enthusiasm as they promised access to elusive evidence of genetic variation in natural populations (Beatty 1987a; Lewontin 1991). At the same time, molecular biology was perceived in opposition to evolutionary biology and the “molecular wars” that ensued pitted molecular and organismal evolutionary biologists in a struggle for authority and resources (Wilson 1994; Hagen 1999; Smocovotis 1996; Dietrich 1998). Molecular evolution attempted to span this divide by building a bridge between Neo-Darwinian evolutionary biology, on the one hand, and biochemistry and molecular biology, on the other. However, controversy quickly enveloped molecular evolution itself as some of its early adherents championed the prevalence of neutral mutations that were not subject to natural selection alone, but instead allowed random drift to play a significant role in evolution.

In retrospect, biologist Alexey Kondrashov describes the impact of the introduction of neutrality in dramatic terms:

Once upon a time, the world seemed simple when viewed through the eyes of evolutionary biologists. All genomes were tightly controlled by various forms of natural selection. […] This idyllic world began to crumble in 1968, when Kimura made his modest proposal that most allele substitutions and polymorphisms do not substantially affect an organism’s fitness and are governed, not by positive or balancing selection, but by random drift (Kondrashov 2005)

The supposed tyranny of panselectionism and the valiant challenge of neutrality were themes that Motoo Kimura included in his own account of the neutral theory (Kimura 1983).Footnote 1 To be fair, the neutral theory of molecular evolution was championed by Jack King, Thomas Jukes, and Motoo Kimura beginning in 1968 (King and Jukes 1969; Kimura 1968). Like some biochemists who were developing the first comparisons of molecular similarity and difference, King, Jukes, and Kimura became convinced that not all changes in proteins or DNA were subject to natural selection (Zuckerkandl and Pauling 1965; Dietrich 1994). In the 1950s, neutral mutations had been acknowledged as a possibility by leading evolutionary biologists, such as Theodosius Dobzhansky, but the presence of large numbers of neutral mutants was not taken seriously and so genetic drift, which would describe their evolutionary fate, was not given much weight as an alternative to natural selection (Dobzhansky 1955). In the late 1960s, King, Jukes, and Kimura marshaled emerging molecular evidence to argue that most observed differences in protein sequences and DNA sequences were the result of neutral mutations subject only to random drift. To older evolutionary biologists who had worked hard to establish neo-­Darwinian evolutionary biology with its emphasis on natural selection, the ideas of neutrality and genetic drift were particularly vexing. King and Jukes fanned the flames of dissent by calling their advocacy of neutral mutations and random drift at the molecular level, Non-Darwinian Evolution. The resulting neutralist-selectionist controversy in molecular evolution lasted for at least 20 years pitting King, Jukes, Kimura, and Tomoko Ohta against selectionists such as Theodosius Dobzhanksy, Richard Lewontin, and John Gillespie, to name only a few (Dietrich 1994).

Putting aside the contentious history of molecular evolution, for philosophers its rise allows us to pose fundamental questions about domains, levels, and causal processes in science. The introduction of molecular data greatly expanded the domain of phenomena to which evolutionary biology could be applied. However, that expansion was accompanied by a significant diversification in the kinds of causes used within evolutionary biology that divided its domain between molecular and organismal phenomena. Where natural selection was assumed to predominate at the organism level, genetic drift and selection are claimed to both be important causal factors in molecular evolution. The recognition of drift and neutral molecular evolution weakened the panselectionist perspective that had been typical of most of evolutionary biology and in practice fostered the widespread use of neutral null models leading to a profound methodological reversal in evolutionary biology.

2 Dividing the Domain of Evolutionary Biology

The rise of molecular evolution does not represent a Kuhnian revolution with a shift to a new scientific paradigm. Molecular evolution is a hybrid discipline borrowing problems, techniques, concepts, and theories from existing fields and combining and adapting them to fit new forms of molecular data. Scientific change in the history of molecular evolution was piecemeal and multi-dimensional in the sense that there was not a single shift in worldview, but a diverse set of drivers that motivated change and created a myriad of distinct novelties. At a very basic level, however, we can say unequivocally that the introduction of molecular data altered the domain of evolutionary biology.

Every field of science has a domain of phenomena that it attempts to explain (Shapere 1977; see Potochnick, this volume for explanation in biology). One of the great virtues of Darwin’s theory of evolution by natural selection is that it explains a very broad range of biological phenomena and is thus heralded as having incredible unifying power (Kitcher 1981). At first glance, the flood of molecular phenomena from the 1960s onward should have only increased the unifying power of evolutionary biology. After all, it now potentially explained a much broader set of phenomena than ever before. However, as the domain of evolutionary phenomena grew, important differences between molecular and organismal evolution began to emerge, as we will see below in the case of the molecular clock.

In today’s post-genomic era, the most common form of data for molecular evolution are DNA sequences. The billions of available base pairs stand in sharp contrast to the early days of molecular evolution when information about DNA content, similarity, or difference had to be indirectly inferred from immunological comparisons, hybridization studies, or protein sequences. Even with these relatively inexact stand-ins for DNA data, molecular evolutionists were able to estimate genetic distances and rates of evolution. These measures allowed them to do unprecedented work in molecular systematics and to postulate the existence of molecular clocks (Morgan 1998).

The term “molecular clock” refers to the approximate rate constancy observed for a type of molecule. Emile Zuckerkandl discovered the phenomena in 1962 when he compared differences among mammalian hemoglobins (Zukerkandl and Pauling 1962, 1965; Zuckerkandl 1963). Zuckerkandl presented his work at the 1962 Wenner-Gren Foundation meeting on molecular anthropology, where luminaries of evolutionary biology, such as Theodosius Dobzhansky, Ernst Mayr, and George G. Simpson, took the opportunity to try to bring him in line with Neo-Darwinism. If evolution is governed predominantly by natural selection and natural selection is modulated by changing environments, then one would expect that the rate of evolution would vary just as the relevant environment varies. Observations of clocklike rates of evolution, thus, presented an anomaly for the Neo-Darwinian perspective since the variable rates at the molecular level seemed to contradict the variable rates of change at the morphological level.

Despite the best efforts of Dobzhanksy, Mayr, and Simpson, they could not contain this anomaly. The result is that, in the late 1960s and 1970s, the perceived division between molecular and organismal levels entered evolutionary biology and began to divide the domain of evolution. What drove the divide was not just the contrast between rate constancy at the molecular level and rate variability at the organismal level that was at issue in the debates over the molecular clock. The molecular – organismal divide was strengthened by the proposal that different causal processes predominated at each level – at the organismal level, selection was the dominant force, while at the molecular level, drift occurred along with selection. In 1971, biochemist Allan Wilson described the molecular – organismal divide as follows:

Molecular evolution proceeds in a rather regular fashion with respect to time. By contrast, organismal evolution is classically considered to be an irregular process, some species (e.g., placental mammals) changing rapidly, while others (e.g., frogs) change slowly. The paradox may be resolved by postulating the random fixation of mutations producing amino-acid substitutions that have no effect, favorable or unfavorable, on protein function and hence no effect on anatomy, physiology or behavior. (Wilson 1971)

The presence of neutral mutations at the molecular level and not the organismal level provided the key difference needed to explain the difference in rates of evolution.

Where molecular biologists have often been described as making reductionist arguments with regard to organismal phenomena, the division of the domain of evolutionary biology was not understood as implying that the organismal level could be reduced to the molecular level. In fact, quite the opposite was understood by organismal evolutionists, such as Dobzhanksy, Mayr, and Simpson, and molecular evolutionists, such as Kimura, Zuckerkandl, and Wilson. Evolutionary biologists saw the molecular and organismal levels as compatible and complementary. This does not mean the molecular clock and neutral theory were not controversial. It means that expanding the domain of phenomena in evolutionary biology defied easy unification (Smocovitis 1996; Mitchell and Dietrich 2006). A crucial contributor to this divide are the different forms of causal processes that were postulated as operating at different levels by molecular evolutionists.

3 The Causes of Molecular Evolution

The subdivision of evolutionary phenomena into the molecular and the organismal levels supported the causal diversification of evolutionary explanations. Explanations at the organismal level were understood as predominantly selectionist and explanations at the molecular level were understood as predominantly neutralist. This does not mean that selection did not play an important role at the molecular level. Neutralist explanations acknowledged significant roles from both selection and drift. The neutral theory was very controversial, however. In part this controversy was fueled by the difficulty of empirically distinguishing between drift and selection as causes of evolution.

From the outset, advocates of the neutral theory claimed that one of its virtues was that it made testable quantitative predictions (Crow 1969). Frustratingly, those early tests were not definitive. Distinguishing drift and selection was not as easy as it first seemed (Beatty 1987b).

Philosophically, untangling drift and selection requires that we distinguish causal processes from their outcomes and realize that drift and selection are causally active at both the molecular and organismal levels (Dietrich and Millstein 2008). In 2002 and 2005, Roberta Millstein articulated a philosophical distinction between process and outcome as a means of explicating significant differences between biological advocacy of drift and selection. “Causal process” here refers to series of causally connected physical states occurring over time, while “outcome” refers to the effect of that process. Biologists often appeal to both process and outcome without necessarily marking that distinction. Consider Kimura’s definition of drift:

By random genetic drift I mean random fluctuation of gene frequencies in a population caused by random sampling of gametes in reproduction. In any sexually reproducing species, the total number of individuals is not only finite, but also can be regarded as a random sample chosen from a much larger collection of male and female gametes (or ‘gene pool’) produced by the parental generation. The amount of fluctuation in gene frequencies (that is, proportion of various alleles) is expected to be larger, the smaller the population. (Kimura 1983, p. 37)

For Kimura, the “process” of drift is the random sampling of gametes and the “outcome” is the random fluctuation of gene frequencies. The problem is that drift is not the only process that can produce the outcome of random fluctuation of gene frequencies. Selection in a changing environment, for instance, could produce this effect. Unless there is an outcome that is uniquely produced by drift, outcomes are not a reliable means to detect or define drift (Millstein and Dietrich 2008). As a result, Millstein advocates defining drift and selection as causal processes.

Both drift and selection can be understood as sampling processes. Kimura understood the process of drift as random gamete sampling, for instance. John Beatty distinguished drift from selection by arguing that the “random” sampling in a drift process should be thought of as “indiscriminate sampling,” while the sampling process inherent in selection is a discriminate sampling process (Beatty 1984). Put another way, in selection, the heritable physical differences between entities are causally relevant to differences in reproductive success (Millstein 2002, 2005). In drift, heritable physical differences are causally irrelevant to the sampling inherent in the reproductive process.

The distinction between process and outcome is especially important for understanding a distinction the Motoo Kimura and his collaborator, Tomoko Ohta, drew between strictly neutral, effectively neutral, and nearly neutral mutations. By definition, strictly neutral mutations are those where there is no selection operating at all, so drift is the only relevant process. From 1968 on, however, Kimura and Ohta acknowledged that there were probably significant numbers of effectively neutral, also known later as nearly neutral mutations. These mutations behaved very similarly to strictly neutral mutations in that they showed very similar outcomes to strictly neutral mutations that were subject only to drift. In fact, these nearly neutral mutations were subject to very weak selection and to drift. This has led some biologist to claim that “nearly neutral” really means “a little bit selective” (Dover 1997, p. 91). From an outcome point of view, nearly neutral mutants have drift-like outcomes. From a process point of view, nearly neutral mutants are subject to processes of both drift and selection, but unlike most organismal cases nearly neutral molecular mutants are so weakly selected that the process of drift has a much stronger effect (Dietrich and Millstein 2008).

In the 1990s, Tomoko Ohta began to champion the Nearly Neutral Theory of molecular evolution, in part because she thought that nearly neutral outcomes could be distinguished from strictly neutral outcomes (Ohta 1992, 2002; Ohta and Gillespie 1996). Francisco Ayala and his coworkers had measured genetic variability in Drosophila and found a large number of relatively rare alleles (Ayala et al., 1974). Ohta realized that the strictly neutral theory would not predict such a high number of these rare alleles. However, if most of the presumed neutral alleles were allowed to be very slightly deleterious alleles, then more rare alleles would be expected. These rare alleles would be slightly deleterious or nearly neutral, so they would still be subject to random drift as a strictly neutral mutant would, but they would also be subject to very weak selection. Ayala’s discovery of rare alleles thus became an outcome associated with near neutrality.

Drift has become strongly linked to molecular evolution because of the neutral theory. The prevalence of drift as a phenomenon or an outcome does not mean that drift alone is the causal process operating to produce those outcomes. While there may be cases of drift processes acting alone or selection processes acting alone, most cases of molecular evolution are best explained in terms of combinations of processes of drift and selection, such as is the case with nearly neutral mutants in natural populations (Dietrich and Millstein 2008).

4 The End of Panselectionism

The neutral theory is commonly juxtaposed to panselectionism – the idea that natural selection is the most important factor in biological evolution and as a result formulating hypotheses of selection and adaptation was the best method for an evolutionary biologist (see Kimura 1983; Provine 1988; Dietrich 2006; for adaptation see also Forber, this volume). Indeed, when Stephen Jay Gould and Richard Lewontin attacked panselectionism and adaptationism in their paper, “The Spandrels of San Marco and the Panglossian Paradigm” (1979), Ernst Mayr responded that drift could not be reliably detected, so starting with drift hypotheses was never going to be a successful method, especially relative to the search for selection in nature (Mayr 1983). Mayr’s confidence was grounded in a history of what he saw as problematic attempts to establish drift as a cause of evolution for morphological traits (see Millstein 2008; Gould 1983). Even at the molecular level, definitive tests of drift had not been established by 1983. That all started to change in 1984, as Martin Kreitman began to apply evolutionary analysis to natural variation in DNA sequences (Kreitman 1983). As DNA sequence data became readily available, statistical tests for selection using DNA data became accepted tools in molecular evolution. Because these tests often used neutral hypotheses, by the 1990s, both proponents and critics of the neutral theory recognized that neutrality, not selection, was a useful starting hypothesis when analyzing DNA sequences in evolutionary biology (Kreitman 2000; Beatty 1987b; Crow 1987).

This methodological shift toward neutrality represents a significant reversal from the selectionist approach. Treating the neutral theory as a null hypothesis spelled the end of methodological panselectionism (Mitchell and Dietrich 2006). According to biologist Roger Selander: “All our work begins with tests of the null hypothesis that variation in allele frequencies generated by random drift is the primary cause of molecular evolutionary change” (Selander 1985, p. 87). Neutrality becomes a starting place because, according to Selander, it is preferable “to begin with the simplest model” and then determine “a baseline for further analysis and interpretation” (Selander 1985, p. 88). Beginning with neutrality as a null hypothesis is not a commitment to neutrality and drift as necessary features of a good evolutionary explanation. Borrowing a distinction regarding adaptationism from Peter Godfrey Smith, we can distinguish empirical, methodological, and explanatory senses of panselectionism. The neutral null model argues against methodological panselectionism, e.g., “The best way for scientists to approach biological systems is to look for evidence of natural selection in some form.” The use of neutral null models requires no prior commitment vis-à-vis empirical panselectionism (“Natural selection is a powerful and ubiquitous force, and there are few constraints, except general and obvious ones, on the biological variation that fuels it.”) or explanatory panselectionism (explaining the effects of selection is “the core intellectual mission of evolutionary theory”) (Godfrey Smith 2001, pp. 335–357).

That said, the rise of statistical tests of selection do have consequences for both empirical and explanatory panselectionism. At the height of the neutralist-­selectionist controversy in the 1970s, Christopher Wills offered a defense of “naïve panselectionism” where he claimed that “virtually any change in amino acid composition of any protein molecule produces a molecule of slightly different properties and therefore of slightly different selective value from the original” (Wills 1973, p. 23). No biologist today would defend such an extreme claim. In fact, Martin Kreitman has argued that “Kimura’s theory of neutrally evolving mutations is the backbone for evolutionary analysis of DNA sequence variation and change” because a “substantial fraction” of the genome is best modeled as selectively neutral, because selective neutrality is a “useful null hypothesis,” and because “statistical analysis of (potentially) neutral variation in a gene (or other region of the genome) can be informative about selection acting at linked sites” (Kreitman 2000, pp. 541–542; Hudson et al. 1987). Kreitman’s approach embraces the idea that neutral and selected variation is to be expected in a natural population, and that the function of statistical tests of selection is to locate and characterize selected sites in nucleotide sequences. Such a view argues against the universality of selection at the molecular level and rejects empirical panselectionism. Moreover, the adoption of these tests argues against explanatory adaptationism as well. The existence of significant proportion of neutral sites makes it untenable to argue that the primary explanatory goal of molecular ­evolution can be that of explaining selection.

Kondrashov was right when he argued that the idyll of selection has crumbled, at least, if he meant the idyll of panselectionism. At the molecular level, neutral variation is accepted and is the effect of random drift. Although neutral null hypotheses are now common starting places in molecular analysis, panneutralism has not replaced panselectionism. Instead, molecular evolution is understood in terms of the complex interplay of drift and selection acting upon sites ranging from strictly neutral to strongly selected.

5 Conclusion

The availability of genomic data and the success of statistical tests of selection that use genomic data have contributed to increased engagement with molecular evolution and tests of selection in particular. This increased activity has produced a shift in the place of molecular evolution within the evolutionary biology curriculum. Consider the subsequent chapters of Douglas Futuyma’s textbooks on evolution. In the editions spanning 1986–2009, molecular evolution has always had its own chapter toward the end of the book. But, in later editions, material on molecular evolution has been increasingly incorporated into other chapters. In Futuyma’s 1986 Evolutionary Biology, for instance, the chapter on drift has three pages on the neutral theory. In the 1998 edition of the same textbook, the neutral theory section had expanded to seven pages with additional pages on coalescents. By the 2005 textbook, Kimura has been added to the introductory chapter, molecular clocks and molecular systematics to the classification chapter, and statistical test of selection to the chapter on evolutionary genetics. By the 2009 edition, even the chapter on adaptation has examples of the molecular evolution of crystallin proteins while the molecular evolution chapter has steadily increased in size (Futuyma 1986, 1998, 2005, 2009). Even if an instructor chooses to skip the chapter on molecular evolution, evolution at the molecular level has become unavoidable in more general chapters on selection, classification, and of course drift.

Molecular evolution offers an opportunity to explore the basic structure of evolutionary biology. Comparison between molecular and organismal evolution shed light on how the domain of evolutionary biology is defined, how it has grown, and how it has been subdivided. The different roles of drift and selection at the organismal and molecular levels raise important questions about the nature of causal processes and their reliable detection. Lastly, the success of statistical tests using genomic data reveal an important methodological and empirical reversal in evolutionary biology that marks the end of panselectionism and the beginning of a more pluralistic approach to evolution in terms of both neutrality and selection. Articulating how processes of drift and selection operate at the molecular level and how molecular and organism phenomena can be integrated are challenges now facing evolutionary biology as a result.