Arrival of the Fittest: Solving Evolution's Greatest Puzzle (33 page)

3
. See McCarthy, Claude, and Copley (1997), Ederer et al. (1997), Nohynek et al. (1996), Copley (2000), and Copley et al. (2012)

4
. See Copley (2000).

5
. See Rehmann and Daugulis (2008).

6
. See van der Meer et al. (1998) and van der Meer (1995).

7
. See Dantas et al. (2008).

8
. See Takiguchi et al. (1989).

9
. See Mommsen and Walsh (1989) and Wright, Felskie, and Anderson (1995).

10
. Plants themselves respire some of the oxygen they produce to build biomass.

11
. Salt-loving bacteria also have other adaptations. See Postgate (1994).

12
. See Steppuhn et al. (2004).

13
. See Bennick (2002).

14
. See McMahon, White, and Sayre (1995).

15
. The reason is that genomes as similar as these, especially in higher organisms, usually encode metabolisms that are also very similar and do not contain very different sets of enzymes.

16
. See Shrestha et al. (2011). They have a mutation that inactivates the enzyme.

17
. See Redfield (1993) and Dubnau (1999).

18
. The genetic material of some viruses is RNA and not DNA, but their life cycle usually involves a DNA intermediate to which similar principles apply.

19
. Excessive DNA can also cause problems when replicated genes or chromosomes need to separate during reproduction.

20
. See Bushman (2002), Loreto, Carareto, and Capy (2008), and Bergthorsson et al. (2003).

21
. The analogy to human races must be taken with a grain of salt. Bacteria do not reproduce sexually like many animals and plants. The notion of a species is not clearly defined for them, and the same holds for even less precise categories such as race.

22
. See Lawrence and Ochman (1998), Blattner et al. (1997), Ochman and Jones (2000), and Pal, Papp, and Lercher (2005).

23
. See Lawrence and Ochman (1998).

24
. Some relevant articles are Smillie et al. (2011), as well as Ochman, Lerat, and Daubin (2005) and Ma and Zeng (2004).

25
. The actual percentage varies, being greater in bacteria and smaller in most multicellular organisms.

26
. See Blattner et al. (1997) and Feist et al. (2007). These may also include enzymes that catalyze chemical reactions outside metabolism, such as enzymes that are involved in transmitting information between cells.

27
. This simple description hides many technical complexities. For example, even similar genes can sometimes encode enzymes that catalyze different reactions, and vice versa. Also, some enzymes can catalyze more than one reaction, some reactions are catalyzed by multiple enzymes, and some enzymes are the products of not just one but multiple genes. In practice, annotating the metabolic reactions in a genome thus involves more than just computerized comparison of genes. See Feist et al. (2009).

28
. This notion of distance is different from the pairwise Hamming distance of two bit strings, which designates the number or fraction of bits at which two strings differ. Specifically, it does not take into account all the reactions that are absent in both metabolisms. Most known metabolisms comprise only a small fraction of the total number of reactions in the known reaction universe. See Ogata et al. (1999). Even if two metabolisms differed in all their reactions, however, there would still exist many reactions that are absent in both metabolisms. For this reason, and because I focus on the proportion of reactions unique to one network, the fraction of shared reactions,
D,
is a more appropriate distance measure than the Hamming distance.

29
. This becomes less surprising if one is aware that the DNA of two such strains may differ in more than one million nucleotides.

30
. I performed this analysis for one bacterial species from each genus, to avoid overrepresenting highly similar species. See Wagner (2009a).

31
. While many colors are caused by pigments, in others a finely textured surface brings forth colors through iridescence, such as in the wing coloration of butterflies. In some coloration phenotypes, such as that of the chameleon, both structural colors and pigment-based colors account for the phenotype.

32
. Although the procedure I described is feasible, it turns out not to be the most efficient way to compute viability. In practice, an approach called flux balance analysis is more useful. It relies on a computational technique called linear programming. For an overview see Price, Reed, and Palsson (2004). Computations like this can determine more than just viability. They also tell us how fast a metabolism works—how speedily it manufactures biomass molecules. In other words, they can tell us whether an organism could go forth and multiply, or whether it would barely hang on to life.

33
. Furthermore, flux balance analysis can also correctly predict growth and nutrient uptake rates under different growth conditions and environments. See Feist et al. (2007), Segre, Vitkup, and Church (2002), Edwards, Ibarra, and Palsson (2001), and Neidhardt (1996). Where predictions and experiment disagree, two principal causes are at work. The first is missing information about a metabolism. The second involves regulatory constraints, where genes for a particular enzyme-catalyzed reaction exist in a genome, but the enzyme is not produced, because the gene is not regulated appropriately. These kinds of constraints are quickly overcome, even in laboratory evolution experiments, and thus do not present a serious obstacle for metabolic innovation. See Fong and Palsson (2004), Fong et al. (2006), Forster et al. (2003), Segre et al. (2002), and Edwards and Palsson (2000).

34
. A notable exception would be endosymbionts, organisms that live inside other organisms and benefit from the constant environment their hosts provide. An example of a long-standing endosymbiosis that has endured for many millions of years is found in the bacterial genus
Buchnera,
an endosymbiont of aphids. See Moran, McCutcheon, and Nakabachi (2008), as well as chapter 6.

35
. See Feist et al. (2007).

36
. A (hypothetical) metabolism viable on all possible fuels certainly has a phenotype, but it can no longer experience a fuel innovation, such that the number of possible innovations must be strictly smaller than that of phenotypes.

37
. A classical work exploring spaces of many dimensions is Abbott (2002). A more contemporary exploration can be found in Stewart (2001).

38
. In mathematical language, our three-dimensional space and the metabolic library—a space of metabolic genotypes—are both metric spaces, because a notion of distance exists in both of them. See Searcoid (2007). Mathematicians also study nonmetric spaces, but their properties are more difficult to understand intuitively, precisely because they lack a notion of distance.

39
. (4 × 10
⁹
[yr]) × (365 [d/yr]) × (8.64 × 10
⁴
[s/d]) × (5 × 10
³⁰
)
= 6.3 × 10
⁴⁷
combinations.

40
. By conventional measures of scientific output, such as citations of scientific publications per capita, it is arguably the world leader. See Cole and Phelan (1999). For what it’s worth, Switzerland has also produced more Nobel laureates per capita than even the United States. “List of Nobel laureates by country per capita,” Wikipedia, http://en.wikipedia.org/wiki/List_of_Nobel_laureates_by_country_per_capita.

41
. More precisely, it is the ability to synthesize the carbon backbone of all these molecules from the carbon atoms and from the energy stored in this sugar.

42
. Other researchers focused on a different question, namely whether all reactions in a metabolism are essential, and found that they are not, which leads to the same conclusion. See Edwards and Palsson (2000), as well as Fong and Palsson (2004).

43
. This work was carried out in collaboration with Areejit Samal and Olivier Martin. See Samal et al. (2010).

44
. The work I discuss here is summarized in Rodrigues and Wagner (2009), as well as in Samal et al. (2010) and Rodrigues and Wagner (2011).

45
. See Rodrigues and Wagner (2009).

46
. I note that in our analyses we started from viable networks of different numbers of reactions, and kept the number of reactions in a network approximately constant during a random walk. Each such walk thus explored a “slice” through the hypercube of the metabolic library.

47
. See Rodrigues and Wagner (2009).

CHAPTER FOUR: SHAPELY BEAUTIES

1
. Fletcher, Hew, and Davies (2001) present an overview of antifreeze proteins in fish.

2
. Some enzymes can catalyze multiple reactions, and are often called “promiscuous” for that reason. See O’Brien and Herschlag (1999). Conversely, some reactions are catalyzed by multiple enzymes.

3
. See Zhao et al. (2001). Charcot-Marie-Tooth disease can also be caused by mutations in other genes.

4
. Other factors matter as well, such as the electric charge of amino acids, but because most of what I say holds for these factors as well, I will let shape stand for them. Binding of molecules with shape complementarity involves specific interactions and attractive forces between molecules, such as hydrogen bonding. See Branden and Tooze (1999).

5
. The technically more precise term for a single amino acid chain is
polypeptide
. A protein can consist of one polypeptide or multiple polypeptides.

6
. The words I use here to describe protein folding are anthropomorphic, but the process is purely physical, no less than how iron filings align in a magnetic field, but more complicated than that, because multiple conflicting attractive and repulsive amino acid interactions are at work.

7
. More precisely, these elements of a protein’s structure are called an α-helix and a pleated β sheet. A pleated β sheet forms from parts of the amino acid chain that are not necessarily contiguous in the chain. These parts are also called β-strands. In the figure, they correspond to the nearly straight ribbons terminated by arrowheads. See also Branden and Tooze (1999).

8
. What you see is actually only about half of the entire amino acid string, also called the N-terminal domain. The entire sucrase molecule is a complex of two amino acid strings. See Sim et al. (2010). The large size of a protein may provide it with stability to thermal motion, specificity for its target molecule, high rate of catalysis, and the ability to regulate its activity. Although one can synthesize catalytic peptides, much smaller enzymes consisting of few amino acids, such peptides do not have these properties of more complex enzymes. See Tanaka, Fuller, and Barbas (2005).

9
. This jiggling is also the basis for enzyme promiscuity, the phenomenon in which some enzymes can catalyze multiple chemical reactions. Some of the oscillating shapes they form can bind molecules other than their main targets, and help these molecules react. They may not be very good at these side jobs, but good enough to accelerate the rate at which these other reactions proceed. See, for example, O’Brien and Herschlag (1999). In the evolution of enzymes early in the history of life, some enzymes probably were highly promiscuous. They catalyzed multiple reactions, each at a low rate, and became specialized later for one reaction that they could catalyze efficiently. See Kacser and Beeby (1984). Because no one enzyme can catalyze all reactions necessary to sustain modern life, promiscuity does not eliminate the need to understand how enzymes with new catalytic abilities arose in the first place.

10
. See Szegezdi et al. (2006).

11
. From the closing paragraph of Darwin (1859).