This page is the second of a series containing the article
Putting the Systems back into Systems Biology by
Athel Cornish-Bowden, published in Perspectives in Biology and Medicine (2006) 49, 475–489.
A PDF file is also available.
But one thing is certain: to understand the whole you must look at the whole.
The forms of hexokinase that predominate in the mammalian muscle and brain are among the enzymes that are often
asserted in textbooks and even in review articles to
control glycolysis. An even more popular candidate for the
enzyme that supposedly fulfills this role is phosphofructokinase, but in any case the claim that any single enzyme
controls a major process is largely a illusion, though it is one that underlies much of the effort in biotechnology
over past decades. The idea, dating from the earliest studies of metabolic regulation in the 1950s and 1960s, is that
any metabolic pathway has a definite
rate-determining step, catalyzed by a
key enzyme, and that once this
enzyme has been identified the techniques of genetic engineering allow it to be overexpressed, thereby increasing the
metabolic flux through the pathway in question to any desired extent.
Heinisch, J. (1986) Isolation and characterisation of the two structural genes coding for phosphofructokinase in yeast, Mol. Gen. Genet. 202, 75–82
Kacser, H. and J. A. Burns (1973) The control of flux. Symp. Soc. Exp. Biol. 27, 65–104
The theoretical reasons why this approach does not work have been known for thirty years (Kacser and Burns, 1973), but even biotechnologists who are impatient with theoretical arguments may pay attention to the experimental observations that it does not work in practice, even in fermenting yeast. There has been widespread agreement in the biochemical literature that the enzyme that controls ethanol production in fermentation is phosphofructokinase. What could be more clear, then, that the rate of ethanol product could be increased by overexpressing phosphofructokinase? Clear or not, the appropriate experiments were first published two decades ago (Heinisch, 1986), however, and showed that there is now no justication for continuing to believe that the naive expectation is the correct one: a 3.5-fold increase in the phosphofructokinase activity in fermenting yeast not only does not lead to a 3.5-fold increase in the rate of ethanol production, it does not have any detectable effect at all on this rate. This result might just reflect some peculiarity of yeast, except that essentially the same observations have now been made in several very different organisms, including potatoes, mice, and bacteria.
Part of the reason why overexpressing a supposedly rate-determining enzyme fails is that organisms have feedback regulatory mechanisms to ensure that metabolic fluxes are suited to the needs of the organism, not to those of an external agency, such as a biotechnologist. This immediately suggests a different way of making organisms satisfy biotechnological ends–using genetic manipulation to eliminate the feedback loops. However, this also fails to have the desired effects, for reasons discussed later.
Cornish-Bowden, A. and M. L. Cárdenas (2001a) Silent genes given voice, Nature 409, 571–572
Yeast also illustrates why another promise of the genomic revolution has proved to be false. It was realized, of
course, that sequencing genomes would reveal the existence of many previously unknown genes—this was, after all, a
major reason for undertaking the sequencing in the first place. It was widely assumed, however, that the function of a
previously unknown gene could be identified, or at least suggested, by doing appropriate genetic experiments to delete
it from the genome and seeing what happened. In practice this rarely works, as a high proportion, maybe as many as 80%,
of genes in yeast are
silent: when one of them is deleted the organism grows and multiplies quite normally, at
the normal rate, and the rates of any metabolic processes that are measured usually prove to be normal as well
(Cornish-Bowden and Cárdenas, 2001a).
1In the context of multienzyme systems the word flux is commonly used for the system property that corresponds to the rate of reaction of a single enzyme. In an unbranched section of a pathway in steady state the flux through each reaction is, of course, the same as the rate of the same reaction, but it is still useful to use different words to make it clear whether the reference is to a property of the system as a whole or to a property of a particular enzyme within it.
Heinrich, R. and T. A. Rapoport (1974) A linear steady-state treatment of enzymatic chains. General properties, control and effector strength. Eur. J. Biochem. 42, 89–95.
Kacser, H. and J. A. Burns (1973) The control of flux. Symp. Soc. Exp. Biol. 27, 65–104
Savageau, M. A. (1976) Biochemical systems analysis: a study of function and design in molecular biology, Reading, MA: Addison–Wesley
Escaping the false line of reasoning requires a different way of looking at biological systems that incorporates the idea that they are more than just collections of components. The major conceptual advance was made by Henrik Kacser and Jim Burns (1973) in Edinburgh, with important contributions a little later from Reinhart Heinrich and Tom Rapoport (1974) in Berlin, with similar ideas developed in parallel by Michael Savageau (1976) in Ann Arbor. Instead of a single enzyme that all by itself regulates the flux1of metabolites through a pathway, flux control is a property that is shared unequally among all the enzymes in the metabolic system. As a typical cell contains thousands of different enzymes, this implies that the average share held by any enzyme is less than 0.1% of the total. This in turn suggests that varying the activity of a randomly selected enzyme (or the expression level of its gene) by a small amount should typically have no detectable effect on any flux that we measure. Even a large increase in activity will typically have no measurable effect (as with the 3.5-fold increase in phosphofructokinase activity in fermenting yeast mentioned earlier). These considerations do not make it impossible for all control to be concentrated in one enzyme, at least over a limited range of its activity, but they imply that this will be exceptional.
A minor complication needs to be mentioned. By considering the average share of control we are implicitly assuming that all the shares have positive values, because if large negative shares are allowed there can be more than one enzyme with a large positive share. More detailed analysis shows that although negative shares are not impossible they are usually not large or numerous enough to destroy the argument: they just make it a qualitative argument that is not quantitatively exact.
Fisher, R. A. (1928a) The possible modication of the response of the wild type to recurrent mutations, Amer. Nat. 62, 115–126
Fisher, R. A. (1928b) Two further notes on the origin of dominance, Amer. Nat. 62, 571–574
Kacser H. and J. A. Burns (1981) The molecular basis of dominance, Genetics 97, 639–666
The effects of large decreases in enzyme activity are more complicated, as there are various points to take into
account. Typically the share of flux control held by any enzyme increases when the enzyme activity decreases, and may
approach 100% when the activity approaches zero. If this happens a particular metabolic activity will be eliminated if
the enzyme activity is eliminated, but that does not necessarily imply that the organism will not continue to grow
normally, because organisms have alternative ways of achieving their needs. If the flux in question is just the rate of
growth, the proportion of absolutely essential genes is low; hence the large number of silent genes in yeast. With more
specific fluxes, such as the rate of producing a pigment that is desirable but not essential for life, eliminating a gene
necessary for producing it may have no observable effect on growth rates but it will still usually have an easily
observable effect on the appearance of the organism. Even for essential genes, large decreases in expression levels may
have little or no observable effects. In diploid organisms, for example, the expression level of the protein coded by
the unique copy of a gene in a heterozygote will typically be about half of that in the normal homozygote, but fluxes
that depend on the protein may be so little changed that they appear to be normal. According to Kacser and Burns
(1981), that is why so many mutations are observed to be recessive. Their argument provides a far simpler
interpretation of genetic recessivity than the ideas of Ronald Fisher (1928ab) in terms of
modifier genes, which
were universally accepted as correct for many years, and which are still taught in genetics courses even today, if only
for their historical importance. He considered that dominance and recessivity were evolved properties that were needed
to avoid what would otherwise be the deleterious effects of having low doses of gene products.
Cornish-Bowden, A. and V. Nanjundiah (2006) The basis of dominance, in The biology of genetic dominance, ed. R. A. Veitia, Georgetown, Texas: Landes Bioscience, pp. 116
Shull, G. H. (1909) The
presence and absence hypothesis,
Wright, S. (1929) Fisher’s theory of dominance, Amer. Nat. 63, 274–279
Wright, S. (1934) Physiological and evolutionary theories of dominance, Amer. Nat. 63, 25–53
Reviewing the history today (Cornish-Bowden and Nanjundiah, 2006) it is hard to escape the feeling that Fisher was
catering for people’s need for a difficult and complicated explanation, even though Sewall Wright (1929, 1934) was
already offering one much closer to that of Kacser and Burns. Even earlier, George Shull (1909), who helped to
revolutionize 20th century agriculture with the introduction of hybrid corn, was describing the phenomenon in the
when there is complete dominance of presence over absence, it may mean that already the presence of
the one unit A of the heterozygote is sufficient to result in the maximum reaction, in which case the doubled factor AA
of the positive homozygote can do no more.