Around 1900 a number of scientists observed “Mendelian ratios” when crossbreeding different varieties of the same species of plant or animal. These ratios can be explained if we assume that an organism contains two factors that determine which character it will display. One factor comes from each parent, and if an organism inherits two different factors, one is always expressed preferentially over the other. Today we are all familiar with these factors, which we know as “genes”, a term introduced to the English language in 1911 by the Danish botanist Wilhelm Johanssen. Each gene comes in a number of alternative forms, known as “alleles”. An organism contains a locus (place) for each gene, which is occupied by two alleles of that gene. This pair of alleles is the “genotype” of the organism at that individual locus. The genotype may be two copies of the same allele (“homozygous” AA or aa) or one copy of each of two different alleles (“heterozygous” Aa or aA). An organism’s overall genotype consists of the pairs of alleles at all the loci.
In the very simplest Mendelian models, like those constructed by Mendel himself, there are only two alleles of each gene. Mendel’s law of segregation says that each parent passes on only one allele to each offspring. Mendel’s law of independent assortment says that the alleles an organism receives at one locus has no effect on which alleles it receives at other loci. Mendel also assumed that for every pair of alleles, one allele is “dominant” and the other “recessive”, so that if an organism has one copy of each allele, it will have the same phenotype as organisms that have two copies of the dominant allele.
When Mendel’s two laws hold, and one allele at each locus is clearly dominant, then a Mendelian model of heredity for two different loci immediately predicts the famous 9:3:3:1 ratios that Mendel observed in his peas and which were rediscovered at the end of the nineteenth century. For every nine offspring with both the dominant phenotypic characters, there will be three that have one recessive phenotypic character, another three that have the other recessive phenotypic character, and one that has both recessive phenotypic characters.
The entity at the heart of Mendelian genetics – the gene – had a very distinctive status. It was not observable, but it was something more than an unobservable entity postulated to explain the data. The Mendelian gene was a tool for predicting the outcome of breeding one organism with another. If organisms are labelled merely with their observable phenotypes, rather than the two unobservable factors that underlie each phenotypes, then it is impossible to calculate the genotypes that will be produced when two organisms interbreed, and hence predict the phenotypes of the offspring. So the gene was not merely postulated to explain why Mendelian genetics made correct predictions, it was an essential tool for making those predictions.
It was natural to hope that the gene would one day be shown to exist as a physical reality within the cells of the organism, and many Mendelians were firmly committed to this idea. But the practical role of the gene in the Mendelian genetics meant that it would remain an important and legitimate idea even if this did not work out. Thomas Hunt Morgan received the Nobel Prize in 1933 for his work demonstrating that genes are distributed along chromosomes. Nevertheless, in his Nobel lecture he noted that, “There is no consensus of opinion amongst geneticists as to what the genes are – whether they are real or purely fictitious – because at the level at which the genetic experiments lie, it does not make the slightest difference whether the gene is a hypothetical unit, or whether the gene is a material particle.”
The historian of genetics Raphael Falk has summed up this situation by saying that the gene of Mendelian genetics – or “classical genetics” as it is often called in its mature form – had two, distinct identities. One identity was as a hypothetical material entity, and some genetic research was directed to confirming the existence of these entities and finding out more about them. But the gene had a second, and more important, identity as an instrumental entity – a tool used to do biology.
The search for the gene as a material entity culminated in the discovery in the late 1950s and early 60s of coding sequences of DNA – sequences that are transcribed into messenger RNAs that are either translated into protein or become functional RNAs. These coding sequences are the genes of contemporary molecular biology – the things that biologists count when they estimate that the human genome contains 20,000 genes or the fruit-fly genome 15,000.
So is the history of the gene a story in which early geneticists glimpsed DNA dimly through the dark glass of breeding experiments until all was revealed by molecular biology? Certainly, many philosophical analyses of genetics have assumed that an older, less adequate theory called Mendelian genetics was reduced to a newer, more adequate theory called molecular genetics, just as Newton’s theory of gravity was reduced to Einstein’s theory of space-time. In that process, the old Mendelian gene was reduced to the new molecular gene.
The idea that the Mendelian gene was replaced by a shiny new molecular gene made of DNA fits with some well-known philosophical ideas about the meaning of theoretical terms. Every philosopher is familiar with Hilary Putnam’s account of how words introduced in ignorance can nevertheless refer to the “natural kinds” eventually revealed by science. Water is H2O because, unbeknownst to them, people had been drinking, bathing in and applying the word “water” to H2O. A parallel treatment of genetics would imply that when Mendel talked of “factors” and Johanssen introduced the term “gene” they were both, without knowing it, referring to molecular genes made of DNA. But as we will see, that is not quite right, and it is not quite right in a way that matters.
The Mendelian and the molecular gene also seem a natural fit for the “role and occupant” approach that goes back to the mind/brain identity theories of the 1950s and remains a standard part of the philosophical toolkit. Some concepts can be analysed by identifying the causes and effects of the thing being conceptualised – the “causal role” associated with the concept. Lighting happens in thunderstorms, makes bright flashes in the sky, and has the destructive effects we see as lighting strikes. If some concrete “occupant” is discovered which has the same causes and effects, then it follows that the concept refers to this newly discovered occupant of the causal role. Atmospheric electrostatic discharges happen in thunderstorms, produce bright flashes and have destructive effects when they reach the ground, so they are lightning.
The gene would seem to follow a similar pattern. It was originally identified by a causal role – that of causing Mendelian patterns of inheritance. Later it was discovered that this causal role is played by pieces of DNA passing from parent to offspring. It follows necessarily that Mendelian genes are made of DNA. But it does not follow that the Mendelian gene is the same thing as the molecular gene. As we will now see, the Mendelian gene is alive and well, and living alongside its molecular relative.
These philosophical frameworks for thinking about scientific progress do not take account of the experimental practice of science. The molecular gene can only replace the Mendelian gene if it can take over its practical role in doing genetics. But it cannot do that. Less than 2% of the human genome consists of coding sequences – the sections of DNA that modern, molecular genetics calls genes. But the rest of the DNA still has alleles that occupy loci and behave in a Mendelian fashion. If we insisted that genetic principles could only be applied to molecular genes we would bring much of biology to a halt. It would be equally destructive to demand that molecular biologists treat all DNA sequences as molecular genes.
The alternative to seeing the Mendelian gene as a primitive precursor to the molecular gene is to recognise that the gene started with two identities and that it retains both identities today. There is no ontological mystery here – both identities are anchored in facts about DNA. But there has turned out to be more than one scientifically useful way to think about DNA.
The 1960s saw the unravelling of the genetic code and the basic processes of transcription and translation by which genes give rise to gene products. The two identities of the Mendelian gene, the tool for predicting the results of crossbreeding and the hypothetical material entity, seemed to have converged neatly on a single, well-defined object – the molecular gene. Looking more closely, however, reveals that rather than simply finding the occupant of the Mendelian role, molecular biology substantially revised the role to fit its new tenant.
In classical, Mendelian genetics the gene was the natural unit of three critical biological processes – re plication, mutation, and the production of biomolecules that help to construct the phenotype. The molecular gene, however, is not the unit of replication. It is entire DNA molecules, chromosomes containing many genes and much DNA that is not genes, which are replicated when a cell divides. Nor is the molecular gene the unit of mutation. Single DNA nucleotides mutate, and they do so whether or not they are part of a molecular gene. The only role with respect to which the molecular gene is a natural unit is that of producing a gene product. Seymour Benzer, who played a critical role in these discoveries, proposed that the term “gene” be replaced by three new terms, “recon”, “muton” and “cistron”. In the event, however, the term “gene” was retained in molecular biology to refer to what Benzer termed a “cistron” – a sequence of DNA that is a natural unit in the processes that make other molecules using DNA as a template.
Where does this leave the idea that the Mendelian gene was reduced to the molecular gene, or that classical, Mendelian genetics was reduced to molecular genetics? Philosophers have been debating this idea since 1967, when philosopher of science Kenneth Schaffner pointed out that a successful scientific reduction seemed to be happening right in front of us. A recent development in this debate is the recognition that what needs to be “reduced” if one science is to absorb another is not only the entities and theories of the older science, but also the practical activities that make up research in that field.
The primary activity of Mendelian genetics is “genetic analysis”. Genetic analysis has been used to answer an extraordinary range of biological questions. Some famous examples in the early days of Mendelism included how sex is determined, initially in fruit flies, and how plants make the pigments in their flowers. The geneticist uses a carefully designed series of crosses between individuals with different phenotypic characters to dissect those characters. Is the original character we observed, such as having red flowers, really one character or several characters? How are these smaller characters connected to one another? How are they connected to other phenotypic characters? Once the geneticist has a full understanding of the genetics underlying a phenotype the Mendelian genes that have been discovered become tools with which to experiment further, using more carefully designed crosses to put components together or break them apart so as to test theories about underlying mechanisms. This is how many of the fundamental discoveries of molecular biology were made. Seymour Benzer, for example, developed techniques using bacterial viruses that allowed him to recombine genetic material on a massive scale and increase the statistical resolution of genetic analysis to the point where he could “see” the internal structure of an individual gene.
So one way to ask whether Mendelian genetics has been reduced to molecular genetics is to ask whether Mendelian genetic analysis has been reduced to some newer form of analysis based on molecular biology. The answer is that it has not. As Raphael Falk and others have argued, the patterns of reasoning found in classical, Mendelian genetic analysis remain the same even when what is being “hybridised” are not two different varieties of an organism but two pieces of DNA in a test-tube.
The fact that biologists continue to conduct Mendelian genetic analysis would not be very significant if the pieces of DNA identified as genes using these methods were always the pieces of DNA that molecular biologists call genes. But this is not the case. The 98% of the human genome that is not composed of molecular genes – sequences that code for a gene product – contains much of the regulatory machinery that determines how the molecular genes behave. When these other sections of DNA come in two or more forms with different phenotypic effects, they will behave as Mendelian alleles and they can be investigated via genetic analysis. Even if they are not called genes, they are implicitly treated as such. However, such is the flexibility of scientific language that often they are called genes in an appropriate discursive context. For example, when a medical geneticist is seeking the “genes for” a disorder she is looking for Mendelian alleles – the sequences of DNA whose inheritance explains the phenotypic abnormalities observed in patients. Translated into molecular terms these sequences of DNA may well turn out not to be molecular genes.
A nice example of this comes from studies of the gene Lmbr1 in the mouse and it’s homologue on human chromosome seven. This locus sometimes houses an allele that produces abnormal limb development in either mice or humans. But further analysis shows that the molecular gene in which the mutation is located plays no role in the development of the limb abnormalities! Instead, embedded within that gene is a sequence which acts to regulate another molecular gene, “sonic hedgehog” (shh), located about one million DNA nucleotides away on the same chromosome. This regulatory sequence occurs in one of the “introns” of Lmbr1, a piece of the DNA sequence that is spliced out and discarded when the Lmbr1 gene is expressed. The gene shh that this sequence regulates is known to play an important role in the development of limbs. This is why mutations to the sequence cause limb abnormalities. They change when and where shh is active in the embryo. The regulatory element at the original locus is not a molecular gene. But it is the Mendelian allele for this kind of abnormal limb development. Conversely, shh is a paradigmatic molecular gene, but there is no allele of shh which is the Mendelian allele for this kind of abnormal limb development.
What we see in the Lmbr1 case is that when geneticists are hunting for the mutation responsible for the abnormality, the gene takes on its Mendelian identity, but when they turn to analysing the underlying sequence, the gene takes on its molecular identity. In many cases these two identities converge on the same sequence of DNA, but sometimes, as in this case, they do not.
So one sense in which Mendelian genetics does not reduce to molecular genetics is that it was not superseded by molecular genetics but remains alongside it as another, important way of thinking about DNA. As a result, biologists today have two different ways of thinking about genes, as Mendelian alleles and as DNA sequences that template for a gene product. As they move between different kinds of research work they move smoothly between these two ways of thinking – using what Karola Stotz and I have called different “contextually-activated representations” of the gene.
R. Falk Genetic Analysis: A History of Genetic Thinking (Cambridge University Press, 2009)