A generalised approach to the study and understanding of adaptive evolution

Abstract

Evolutionary theory has made large impacts on our understanding and management of the world, in part because it has been able to incorporate new data and new insights successfully. Nonetheless, there is currently a tension between certain biological phenomena and mainstream evolutionary theory. For example, how does the inheritance of molecular epigenetic changes fit into mainstream evolutionary theory? Is niche construction an evolutionary process? Is local adaptation via habitat choice also adaptive evolution? These examples suggest there is scope (and perhaps even a need) to broaden our views on evolution. We identify three aspects whose incorporation into a single framework would enable a more generalised approach to the understanding and study of adaptive evolution: (i) a broadened view of extended phenotypes; (ii) that traits can respond to each other; and (iii) that inheritance can be non-genetic. We use causal modelling to integrate these three aspects with established views on the variables and mechanisms that drive and allow for adaptive evolution. Our causal model identifies natural selection and non-genetic inheritance of adaptive parental responses as two complementary yet distinct and independent drivers of adaptive evolution. Both drivers are compatible with the Price equation; specifically, non-genetic inheritance of parental responses is captured by an often-neglected component of the Price equation. Our causal model is general and simplified, but can be adjusted flexibly in terms of variables and causal connections, depending on the research question and/or biological system. By revisiting the three examples given above, we show how to use it as a heuristic tool to clarify conceptual issues and to help design empirical research. In contrast to a gene-centric view defining evolution only in terms of genetic change, our generalised approach allows us to see evolution as a change in the whole causal structure, consisting not just of genetic but also of phenotypic and environmental variables.

Keywords: Price equation; adaptation; causal modelling; evolutionary synthesis; evolutionary theory; non-genetic inheritance; plasticity.

Fig. 1

Lewontin's (1974) scheme of evolutionary processes in phenotypic/genotypic spaces (left). The four transition steps (T ₁–T ₄) represent the causal processes that an evolving population undergoes. These causal processes are modelled with a causal graph over relevant variables (right), and each causal connection is quantified by structural equations. Z and X represent the individual phenotypic and genotypic values for the population quantities P and G, respectively. β is the selection gradient, α is an additive genetic effect, and E is an independent noise term. (Note that in the panel on the left the prime indicates different time steps within a generation and different generations are indicated with a numerical subscript for P and G; in the panel on the right the different generations are indicated with a prime, and the numerical subscripts refer to different genes). Combined with the Price equation (see Section IV), the model then induces a formula (in this case, the breeder's equation) that gives the evolutionary change in phenotype between two generations (the dotted line from P ₁ to P ₂). Adapted from Otsuka (2019b ).

Fig. 2

A causal model for evolutionary change involving two traits. Genetic covariance between genes, indicated by a dashed double‐headed arrow, is optional.

Fig. 3

Capturing with a single causal model the different processes that can drive adaptation as identified in Edelaar & Bolnick (2019). Each panel highlights a different aspect of the overall model. The main variables and causal effects involved in each process are indicated in red, and the dependent variable is indicated in italics. Dotted lines are optional. Natural selection (A) is the effect of the phenotype and the environment on fitness [responsiveness of traits and genetic covariance between genes (indicated by a double‐headed arrow) are optional]. Change in the phenotype (B) involves the development of the organism's phenotype due to its genotype, and any additional responsive change due to environmental inputs (i.e. phenotypic plasticity; likely additional, organism‐independent environmental inputs are not depicted). Symmetrically, change in the environment (C) (choice and adjustment of the environment, cf. Edelaar & Bolnick, 2019) involves the development of the organism's environment due to its genotype, and any additional responsive change due to phenotypic inputs (i.e. a form of plasticity; likely additional, organism‐independent environmental inputs are not depicted).

Fig. 4

Adding non‐genetic inheritance to the causal model of Fig. 3 to make it even more general (A). This addition identifies two very distinct forms of adaptive evolution (compare B with C). First, adaptive evolution can be driven by natural selection (B), when traits (classical or extended phenotypes) increase fitness and are transmitted to offspring. Importantly, whether this transmission is achieved genetically or non‐genetically is inconsequential for its occurrence and operation, since fitness depends on phenotypes (and not genotypes). Second (C), adaptive evolution can be driven by the non‐genetic transmission to offspring of parental responses (acquired traits, e.g. parents move into a different habitat and in response acquire a new parental behaviour, which is subsequently copied by offspring), assuming this parental response itself is adaptive. For this, transmission needs to be non‐genetic because changes in phenotypes are not registered in the genotype (as far as we know). Red variables and arrows highlight the key elements of each panel. Dotted arrows indicate causal effects that are optional.