Synthetic Biology of Phenotypic Adaptation in Vertebrates: The Next Frontier

Abstract

For over the last 2 decades, positively selected amino acid sites have been inferred almost exclusively by showing that the number of nonsynonymous substitutions per nonsynonymous site (dn) is greater than that of synonymous substitutions per synonymous site (ds). However, virtually none of these statistical results have been experimentally tested and remain as hypotheses. To perform such experimental tests, we must connect genotype and phenotype and relate the phenotypic changes to the environmental and behavioral changes of the organism. The genotype–phenotype relationship can be established only by synthesizing and manipulating “proper” ancestral phenotypes, whereas the actual functions of adaptive mutations can be learned by studying their chemical roles in phenotypic changes.

The molecular basis of phenotypic adaptation in higher eukaryotes is a central unanswered question in evolutionary biology. Using microorganisms, various evolutionary hypotheses have been tested experimentally, and adaptive mutations have been identified (e.g., Dean and Thornton 2007; Romero and Arnold 2009). However, for higher eukaryotes, particularly for vertebrates, it is extremely difficult to even establish that certain phenotypic (or functional) changes have undergone adaptive evolution (Morgan 1903; Gould and Lewontin 1979; Lewontin 1979). Because of this difficulty, only a part of adaptive evolution, molecular adaptation, has often been studied using the criterion that dn is significantly greater than ds (Hughes and Nei 1988; Nei and Kumar 2000). This approach has gained popularity, and statistical results are accumulating rapidly (Hughes 2008). However, the condition of “dn > ds (or ω = dn/ds > 1)” is an untested assumption, and these statistical results contain significant proportions of false positives and false negatives (Yokoyama, Tada, et al. 2008; Nozawa et al. 2009a). The reliabilities of such statistical procedures and results have been intensely debated, but their consensus is that the statistical results provide biological hypotheses (Nozawa et al. 2009a, 2009b; Yang et al. 2009; Nei et al. 2010).

To test these statistical results and thus hypotheses on molecular adaptation, we have to consider relevant phenotypes that are subjected to natural selection. Then, we need to 1) establish the genotype–phenotype relationship and 2) relate the phenotypic changes to the ecological environments and behaviors of the organism. Here, I shall use vertebrate vision as a model system to explore how one can study the molecular basis of phenotypic adaptation.

Vertebrate Visual Pigments

Each visual pigment consists of an opsin, encoded by a specific opsin gene, and an actual light-sensitive molecule called retinal (vitamin A₁ or vitamin A₂), derived from diet (Lythgoe 1979). Vitamin A₁ and vitamin A₂ in solution have a wavelength of maximal absorption (λ_max) at about 380 and 390 nanometer (nm), respectively. However, by interacting with various opsins, these retinals achieve λ_maxs between 360 nm (UV) and 630 nm (far red) (Loew and Dartnall 1976; Loew and Lythgoe 1978; Yokoyama 2008). This phenomenon is known as “spectral tuning.” Therefore, the mechanism of phenotypic adaptation of vertebrate vision must be understood by studying how different species have tuned their visual pigments to adapt to their environments.

For vertebrate visual pigments, the genotype–phenotype relationships can be studied rather easily using the in vitro assay. For invertebrate pigments, with some exceptions (e.g., Wakakuwa 2010), such a rapid assay is not available and more time-consuming heterologous expression methods based on P-element-mediated germ line transformation are required (e.g., Knox et al. 2003; Salcedo 2009). Applying the in vitro assay to bovine rod-specific rhodopsin, Doi et al. (1990) and the subsequent series of articles from the H. Gobind Khorana’s laboratory and others elucidated the biochemical properties of key sites of visual pigments. Unfortunately, most mutations considered in these analyses did not occur in nature and are irrelevant in studying the molecular mechanism of phenotypic adaptation.

Problems with Traditional Mutagenesis Experiments

In performing mutagenesis experiments, biologists introduce mutations into present-day molecules. Unfortunately, this approach often leads to erroneous genotype–phenotype relationships! To appreciate this point, we may consider the following four sets of examples.

The first example involves a visual pigment found in a rod photoreceptor (called RH1 pigment) of the conger eel (Conger myriaster) and its ancestral pigment (Ancestor H) with λ_maxs of 486 nm and 501 nm, respectively (fig. 1A). When S292A was introduced into the present-day conger pigment, the λ_max was increased by 12 nm. Hence, A292S seems to explain the conger pigment evolution, but A292S in the Ancestor H does not decrease the λ_max (fig. 1A). This example shows that forward (A292S) and reverse (S292A) mutations do not shift the λ_max in the opposite direction by the same amount.

Fig. 1.

The λ_max shifts caused by forward mutations in the ancestral pigments and reverse mutations in the present-day pigments. (A) The conger and its ancestral RH1 pigments (Ancestor H) (Yokoyama, Tada, et al. 2008). (B) The scabbardfish, bluefin killifish, lampfish, and the ancestral SWS1 pigments (Ancestors I and J) (Tada et al. 2009). The dotted arrow shows the transition from UV sensitivity to violet sensitivity detected by QM/MM analyses. (C) The wallaby, elephant, bovine, and their corresponding ancestral pigments (Ancestors K, L, and M) (Shozo Yokoyama and Takashi Tada, unpublished data). (D) The human MWS and LWS pigments and the LWS pigments in the mammalian ancestor (Ancestor N) (Asenjo et al. 1994; Yokoyama, Yang, et al. 2008). The wavelengths below the ancestral pigments show their λ_maxs, and the numbers in the parentheses next to arrows are λ_max shifts caused by specific amino acid changes. The black, dark blue, light blue, green, and red colors indicate various RH1, UV-sensitive SWS1, violet-sensitive SWS1, MWS, and LWS pigments, respectively. The numbers after P denote the λ_maxs of visual pigments.

The second example considers the short wavelength-sensitive type 1 (SWS1) pigments of scabbardfish (Lepidopus caudatus), bluefin killifish (Lucania goodei), lampfish (Stenobrachius leucopsarus), and the pigments in their common ancestor (Ancestor J) as well as in the vertebrate ancestor (Ancestor I) (fig. 1B). The magnitudes of the λ_max shifts that are generated by the F86 insertion into the scabbardfish pigment (Δλ = −60 nm) and by the F86 deletion from the Ancestor I (Δλ = 19 nm) differ significantly. The effects of identical F86 deletions from the Ancestor I (Δλ = 19 nm), bluefin killifish pigment (Δλ = 37 nm), and lampfish pigment (Δλ = 58 nm) also show dramatic differences.

The third example includes the SWS1 pigments of wallaby (Macropus eugenii), elephant (Loxodonta africana), bovine (Bos taurus), and their respective ancestors K, L, and M that have evolved sequentially (Shozo Yokoyama and Takashi Tada, unpublished results, fig. 1C). F86Y in the Ancestor K (Δλ = 57 nm) and Y86F in the wallaby pigment (Δλ = −59 nm) cause virtually identical magnitudes of λ_max shift in the opposite directions, but the forward and reverse mutations between the Ancestor L and the elephant pigment (Δλ = 14 nm vs. −52 nm) and between the Ancestor M and the bovine pigment (Δλ = 48 nm vs. −71 nm) cause highly different λ_max shift in the opposite directions. Moreover, the same mutations (F86Y or Y86F) in different pigments result in significantly different magnitudes of λ_max shifts.

The fourth example considers the middle wavelength-sensitive (MWS) human (P530) and long wavelength-sensitive (LWS) human (P560) pigments in human and their ancestral pigment (Ancestor N, fig. 1D). The total effect of S180A, Y277F, and T285A (or S180A/Y277F/T285A) in human (P560) (Δλ = −33 nm) and that of the reverse changes in human (P530) (Δλ = 23 nm) differ by 10 nm (Asenjo et al. 1994). In this example, four additional amino acid changes are required to explain the λ_max shift from human (P530) to human (P560) (Asenjo et al. 1994). Therefore, the λ_max difference between the two human pigments can be explained by two different mechanisms!

These and many other examples reveal that not only can the same mutations in different pigments cause different magnitudes of λ_max shifts but also forward and reverse mutations can shift the λ_maxs in the same directions (Yokoyama 2012). These inconsistent results can occur through epistatic interactions among the critical mutation(s) with other amino acids in the background. Such epistatic effects certainly are not limited to visual pigments. For example, the transcription factors, glucocorticoid receptor (GR) and the mineralocorticoid receptor (MR), evolved from the ancestral corticoid receptor (AncCR). The MR controls electrolyte homeostasis, kidney, and colon function, whereas GR regulates such functions as the stress response and glucose homeostasis. The function of the ancestral receptor has been retained in MR and the new function of GR evolved from that of AncCR (Bridgham et al. 2006). When S106P/L111Q are introduced into AncCR (Bridgham et al. 2006) and into a more recent ancestral receptor (Ortlund et al. 2007), the function of GR is achieved. However, when the same two mutations are introduced into the present-day human MR, the mutant receptor becomes nonfunctional (Ortlund et al. 2007), again showing that identical amino acid changes in different proteins can produce very different phenotypes.

The Solution—Synthetic Biology

To evaluate the genotype–phenotype relationship correctly, it is necessary to evaluate the epistatic interactions of critical mutations and those in the backgrounds correctly. For that purpose, we need to recapitulate the evolutionary process and manipulate ancestral molecules and phenotypes (Yokoyama and Radlwimmer 2001; Shi and Yokoyama 2003; Bridgham et al. 2006; Ortlund et al. 2007; Yokoyama, Tada, et al. 2008). Therefore, only the mutagenesis experiments using synthesized ancestral pigments can reveal correct genotype–phenotype relationships (fig. 1A, C, and D). The importance of manipulating ancestral molecules has also been stressed by Harms and Thornton (2010).

For the usage of ancestral molecules, two comments may be in order. First, amino acid sequences of ancestral molecules were first inferred using statistical methods, and necessary mutations are introduced by site-directed mutagenesis. The uncertainties of the recreated ancestral molecules may be expressed as low posterior probabilities in the Bayesian method (Yang 2007). The effects of “ambiguous” amino acids on the function of the reconstructed molecules should be checked by replacing them either individually by those with the second highest probabilities or collectively by constructing chimeric molecules and evaluating their functions (e.g., Yokoyama and Radlwimmer 2001; Shi and Yokoyama 2003). Second, it is not sufficient to use any ancestral molecules; rather only “proper” ancestral molecules must be manipulated. We saw that the true effect of F86 deletion in scabbardfish SWS1 pigment evolution could not be studied by manipulating Ancestor I (fig. 1B); instead, a more immediate ancestor (Ancestor J) must be used (see Quantum Chemistry).

Neutral Mutations Can Become Adaptive

In the 1970s, evolutionary biology was dominated by the controversy between the so-called “neutralists” and “selectionists” (see Nei et al. 2010). The grouping of mutations into “neutral” and “adaptive” classes, however, may be less productive than previously thought because such classifications do not reveal how natural selection actually works. The λ_max of the conger RH1 pigment was decreased by three amino acid substitutions A292S/P194R/N195A, but none of them shift the λ_max individually (fig. 1A). Here, the neutral mutations become adaptive when they come together under a certain environment. Indeed, it appears that “neutral mutations prepare the ground for later evolutionary adaptation” (Wagner 2008).

Quantum Chemistry

Using synthetic biological methods, we may eventually be able to identify all adaptive mutations. However, we still cannot tell how these mutations regulate the functional differentiation of proteins. For visual pigments, the precise roles of amino acid changes in phenotypic adaptation can be studied using quantum mechanical/molecular mechanical (QM/MM) analyses, in which each visual pigment is treated to consist of a mixture of two groups: pigments with protonated Schiff base nitrogen-linked retinals (PSBR group) and those with unprotonated forms (SBR group). The shift in the λ_max of a visual pigment is caused by the changes in the ground-state energies of SBR group (E_SBR) and PSBR group (E_PSBR). When ΔE ( = E_PSBR − E_SBR) is positive, the SBR group is energetically more stable than the PSBR group, and the pigment is UV sensitive, whereas when ΔE is negative, the pigment consists mostly of PSBR group and is violet sensitive.

We have seen that F86 deletion in Ancestor I did not explain the λ_max shift of the scabbardfish pigment (fig. 1B). The QM/MM analyses show that Ancestor I with and without the F86 deletion (ΔE = 5.8 and 6.2 kcal/mol, respectively) are both UV sensitive. The more recent Ancestor J is also UV sensitive (ΔE = 3.2 kcal/mol). However, the F86 deletion makes this pigment violet sensitive (ΔE = −6.2 kcal/mol). Therefore, two phenotypically identical UV-sensitive ancestral pigments can respond to the identical mutation very differently, showing that the genotype–phenotype relationship can be evaluated correctly only by manipulating a “proper” ancestral phenotype.

How is the λ_max of Ancestor J regulated? Its SBR and PSBR groups consist of five and three H bonds near their retinals, respectively, making the SBR group energetically more stable than the PSBR group and, therefore, Ancestor J is UV sensitive. On the other hand, having two and three H bonds at the corresponding regions of its SBR- and PSBR groups, the scabbardfish pigment is now violet sensitive (Tada et al. 2009). Furthermore, by studying the switches between UV- and violet sensitivities at several major stages of vertebrate evolution, it has been shown that the evolution of various SWS1 pigments is mediated by the “H-bond network” connected by amino acid sites 86, 90, 113, 114, 118, 295, and two water molecules (Altun et al. 2011).

Conclusion

At present, positive Darwinian selection of molecules is studied mostly by using comparative sequence analyses. By engineering and manipulating “proper” ancestral molecules, we can study molecular adaptation and phenotypic adaptation together. The synthetic biological approach will reveal the molecular basis of phenotypic adaption that cannot be imagined using the current molecular evolution methods. It is hoped that more evolutionary biologists will find such experimental approaches necessary in studying how Darwinian natural selection actually operates.