Developing the genotype‐to‐phenotype relationship in evolutionary theory: A primer of developmental features

For decades, there have been repeated calls for more integration across evolutionary and developmental biology. However, critiques in the literature and recent funding initiatives suggest this integration remains incomplete. We suggest one way forward is to consider how we elaborate the most basic concept of development, the relationship between genotype and phenotype, in traditional models of evolutionary processes. For some questions, when more complex features of development are accounted for, predictions of evolutionary processes shift. We present a primer on concepts of development to clarify confusion in the literature and fuel new questions and approaches. The basic features of development involve expanding a base model of genotype‐to‐phenotype to include the genome, space, and time. A layer of complexity is added by incorporating developmental systems, including signal‐response systems and networks of interactions. The developmental emergence of function, which captures developmental feedbacks and phenotypic performance, offers further model elaborations that explicitly link fitness with developmental systems. Finally, developmental features such as plasticity and developmental niche construction conceptualize the link between a developing phenotype and the external environment, allowing for a fuller inclusion of ecology in evolutionary models. Incorporating aspects of developmental complexity into evolutionary models also accommodates a more pluralistic focus on the causal importance of developmental systems, individual organisms, or agents in generating evolutionary patterns. Thus, by laying out existing concepts of development, and considering how they are used across different fields, we can gain clarity in existing debates around the extended evolutionary synthesis and pursue new directions in evolutionary developmental biology. Finally, we consider how nesting developmental features in traditional models of evolution can highlight areas of evolutionary biology that need more theoretical attention.

Evolutionary biologists seek to understand patterns of adaptation and diversification across the tree of life.Why are some lineages so speciose?How do novel traits arise?When does history constrain adaptation to new conditions?For over a century, biologists have been calling for a greater integration of development and evolution (Baldwin, 1902;Waddington, 1962;West-Eberhard, 2003), with recent calls for an extended evolutionary theory; that incorporates many aspects of development (K.Laland et al., 2014).However, the integration between evolutionary and developmental biology continues to stop short-for example, key textbooks in evolution generally dedicate one chapter to developmental biology (Bergstrom & Dugatkin, 2018), and vice versa (Barresi & Gilbert, 2019).Recent funding initiatives stress the importance of integration across biological disciplines (e.g., NSF "Biology Integration Institutes").Efforts over the last two decades to integrate evolutionary biology with developmental perspectives have reinvigorated the field, but have also greatly expanded the collection of terms and concepts, in addition to creating debates as to how much development is necessary in evolutionary biology.In this review, we present a primer on the diversity of ways that biologists see development, and argue that much of the confusion stems from the fact that there are many ways to model developmental processes, from the highly mechanistic to the phenomenological, and the developmental features one needs to understand evolutionary processes depends on the question.In this review, we consider development as the relationship linking genotype to phenotype in our most basic models of evolution, expanding the textbook metaphor of development as a "map" or "blueprint," as such conceptualizations can bias patterns of inquiry and understanding (Frigg & Hartmann, 2006;Nijhout, 1990;Zuk & Travisano, 2018).
Evolution by natural selection arises from heritable trait variation and differential reproduction, and our models of this process can be powerful predictors of population change.The most basic model of evolution by natural selection, the "Mendelian model," includes a simple link between genotype and phenotype (a "toy model" of development, sensu Frigg & Hartmann, 2006), (Figure 1).When we teach evolution, we often use straightforward examples that conform to this Mendelian model to illustate how evolution works.For instance, imagine a genetic locus with two alleles that result in color variation, perhaps due to expression of a melanin-related gene.Selection on such color variation by predators results in a shift in allele frequency in populations resulting in camouflage.Melanistic peppered moths and background-matching mice represent well-studied examples of evolution in action in the field, where we even know the specific genes involved (Barrett et al., 2019;Cook & Saccheri, 2013).These examples are powerful because they illustrate how evolution by natural selection can follow from a few basic assumptions.However, at the same time, most biologists recognize that such one-to-one mapping between genotypic and phenotypic variation is uncommon for most complex traits (Boyle et al., 2017;Travisano & Shaw, 2013).The Mendelian model also fails to address most macroevolutionary questions-such as the origin of novel traits-since it is limited to a very specific and relatively rare developmental scenario of genotype-to-phenotype mapping, namely, where the presence or absence of one or a few alleles have a large and mostly invariable (e.g., environmentally robust) effect on a phenotype.
One can build on the Mendelian model of evolution by incorporating aspects of development, expanding the map that links genotype to phenotype (Figure 1).Indeed, when models of evolution add aspects of development, we expand the type of evolutionary questions we can address.For example, Draghi and Whitlock (2012) incorporate plasticity and genetic networks into evolutionary simulations and see that plasticity facilitates the accumulation of genetic variation.In another model, Kriegman et al. (2018) incorporate growth and morphogenesis into a soft robotics evolutionary model, which results in more evolvable populations, especially with respect to behavioral development.In each of these models, the Mendelian model of evolution is expanded to include additional features of development, even if it is without much real-life detail.This shifts the types of evolutionary questions we can answer, from thinking about the sorting of existing genetic variation, to the The Mendelian model of evolution considers a population of individuals that harbor genetic variation which is translated into phenotypic variation (through a developmental mapping of genotype to phenotype).Interactions with the environment affect the relative survival and reproduction of each genotype which feeds back to affect their frequency in the next generation.We can make this basic model slightly more realistic by expanding across the genotype (considering variation across all genes) and considering interactions across the loci.
origin of novel traits, the patterns of cryptic genetic variation, and divergence across lineages.Expanding basic evolutionary models with developmental features shows how development shapes possible evolutionary paths (González-Forero, 2022).
In this review, we define development broadly as a process linking genotypes and phenotypes, where the organism undergoes a process of self-construction.We offer a primer of developmental features that have been explored across biology and related fields.We begin by arguing that even a basic expansion of simple models of genotype-to-phenotype translation to involve interactions across the genome (e.g., epistasis) is a way some biologists think of development.We then expand the discussion to increasingly complex features of development: (1) space, time, and growth; (2) signals and responses; (3) trait function and performance; and (4) interactions with the external environment.We see this collection of developmental features as a pluralistic approach to incorporating development in evolutionary theory (Dieckmann & Doebeli, 2005), where theoreticians might choose different conceptualizations of development in their models depending on the types of traits, processes, and questions they are interested in.Throughout, we integrate the concepts argued to be necessary in an "extended evolutionary synthesis" as part of a broader list of developmental features, and suggest that some of the current debate within the field may stem from different concepts of development.Finally, we discuss how considering a range of developmental features may offer new directions in research and theory.By expanding the Mendelian model of evolution with key features of development, we can more explicitly ask when and how development matters in evolution.We couch our primer of developmental features within traditional evolutionary theory.This body of theory, which often includes the Mendelian model as both a pedagogical and theoretical starting point, has been criticized for centering the causal role of genes while ignoring or de-emphasizing the epigenetic context in which genes exist and function (K.Laland et al., 2014;Moczek, 2012).We believe that incorporating increasing developmental realism may help to address some of these critiques, and we also recognize the limitations to our framework and turn to these at the conclusion of the manuscript.

| FROM THE GENOTYPE-TO-PHENOTYPE MAP TO THE DEVELOPMENT OF AN AGENT
The basic Mendelian model of evolution maps a one-toone correspondence between a single locus and a phenotype.We can begin to incorporate aspects of development into such a model by including additive effects of a combination of loci on a phenotype, along with interactions among these genes, which are widespread in nature and significant drivers of evolutionary dynamics (Barton, 2017;Corbett-Detig et al., 2013).For example, variation in many traits, such as body size or disease susceptibility, results from variation at hundreds of genetic loci, and interactions across these loci, for instance, if the state of one locus masks effects at another (Brem & Kruglyak, 2005;Mackay, 2014).The genotypeto-phenotype "map" considers genetic information across the genome, both additive effects of many loci, and epistatic interactions across these loci.Models of statistical epistasis emerge from classical quantitative genetic theory (Lynch & Walsh, 1998) while physiological or "functional" models of epistasis model gene interactions between specific alleles and quantify the phenotypic effects (Cheverud & Routman, 1995;Hansen, 2013;Hansen & Wagner, 2001).
The mapping of a genotype to a phenotype is possibly the most basic abstraction of development, but it is important to recognize that it is a way to think about development.Incorporating a more complex genotype-to-phenotype map by including genetic interactions within the genome has yielded significant insights in evolutionary theory (Hansen, 2013).Functional epistasis can play an important role in the origin and maintenance of distinct species (Coyne & Orr, 2004;Gavrilets, 2004;Orr et al., 2001), in the evolution of developmental robustness (G.P. Wagner et al., 1997), responses to selection (Carter et al., 2005;Hansen et al., 2006), and the maintenance of cryptic genetic variation (Ghalambor et al., 2007).However, expanding the Mendelian model of inheritance to one that includes genetic interactions in producing phenotypes still falls short in explaining the evolution of many complex traits.For instance, many traits are highly polygenic, but considering only variation in single nucleotide polymorphisms across the genome statistically linked to trait variation consistently underperforms in accounting for patterns of phenotypic variation and evolution.Take human height, for example-an easily measured trait for which we have a huge amount of data linked to genetic background.Across the entire genome, significant variation at 697 genes explains only 16% of the phenotypic variation in height, and 62% of all common genetic variants are associated at least somewhat with variation in height (Boyle et al., 2017).Incorporating other features of development, such as regulatory interactions and the environment, can increase explanatory power in understanding this trait variation.As we add more complex features of development, we shift our focus SNELL-ROOD and EHLMAN | 395 from genes as the causal driver of development to agents constructing their own development.

| Space and time
A core feature of the developmental process is that an individual organism emerges over time, as features grow (Figure 2).In the Mendelian model, genotypes "instantly" translate into phenotypes; we know, however, that biological traits take time to develop, as proteins are expressed, structures are built, and traits interact with the environment.We conceive of developmental time in both absolute and relative ways (Reiss, 2003).Fairly straightforward approaches capture "time" as different time steps connecting traits to fitness (Scheiner, 2018) or pauses between an individual's origin, trait expression, and selection (Scheiner, 2013).Other models explicitly incorporate time steps into life history, with selection acting at multiple times (aster modeling, Shaw et al., 2008;Wagenius et al., 2010).Increasingly realistic models of time capture the full complexities of a life cycle as organisms pass through different life stages (e.g., Crouse et al., 1987;Grenfell et al., 1987;Vangroenendael et al., 1988).Time can incorporate underlying mechanisms, such as molecular clocks that tick alongside cell divisions (e.g., Matsuda et al., 2020) or the rate of chemical reactions (e.g., T. L. Wagner et al., 1984).
Over developmental time, an individual is growingbuilding and remodeling parts of itself, whether proteins, cells, or tissues (e.g., Raff & Wray, 1989).In single-celled organisms, development may include gene expression, making proteins, and building structures, such as organelles: the organism, or agent, is building itself.For multicelled organisms, growth additionally occurs through cell divisions.As growth occurs, individual units emerge across space, what we may think of as developmental modules, or traits (intersecting with the literature on defining "trait," e.g., DiFrisco et al., 2020).As these individual modules begin to take up space, physical dynamics emerge.Some concepts of growth and space tend to focus on the relationship of one growing unit to another, within an individual (e.g., allometry; Klingenberg, 1998;Shingleton, 2010).Biomedical engineering considers growth from a physical perspective, as solids and fluids that change over time to fill a threedimensional space (Ambrosi et al., 2011;Kuhl, 2014;Taber, 1995).These models offer a mechanical explanation for a diversity of shapes resulting from differential growth in response to mechanical stress, nutrients, or hormones (Ateshian & Humphrey, 2012;Kuhl, 2014).
Thinking about time and space as a part of development has led to important insights in evolutionary developmental biology.Variation over developmental time can form the basis for subsequent evolutionary diversification, as has long been appreciated in the study of heterochrony in evolutionary developmental biology (Klingenberg, 1998;Raff & Wray, 1989;Smith, 2001).Changes in the timing or duration of growth can result in shifts in body shape (Atchley & Hall, 1991;Klingenberg, 1998) or behavior, such as the retention of some juvenile features in domesticated dogs (Coppinger et al., 1987;Geiger et al., 2017).Similarly, a consideration of space, and the growth of individual units relates to the degree of independence-or modularity-of developing units, which can influence evolutionary divergence (Klingenberg, 2008;G. P. Wagner & Altenberg, 1996) and morphological patterns across species (e.g., Felice & Goswami, 2018).Adding time and space to development means that certain adult phenotypes may be contingent upon earlier developmental steps, or other developing modules, resulting in the evolution of developmental constraints (e.g., Diggle, 2002;Snell-Rood et al., 2015).Similarly, selection can act on the organism at different points of development and the relative maturity of a trait at that point in time can affect its functional performance.

| Developmental systems of signals and responses
As an individual organism develops, whether a single or multicelled individual, the developmental "system" plays a key role in translating the genome into organismal phenotypes, which subsequently feedback to affect development.The idea of a developmental system has emerged independently in a number of fields."Developmental systems theory" (DST) stresses the importance of development as construction, with controls distributed across components of the developing system (Griffiths & Tabery, 2013;Oyama, 1985;Oyama et al., 2001;Robert et al., 2001).Within evolutionary developmental biology, gene regulatory network models attempt to capture some complexity of "developmental programs" playing out over time and space (e.g., Oster & Alberch, 1982).While these approaches (i.e., DST and models of "developmental programs") are often framed in opposition to each other, they both share the view that development systems, themselves, are important units of selection.
Here we highlight three interacting features of this developmental system (Figure 3).The individual organism, or agent, is playing a role in their own development.
The basic unit of a developmental system is an individual signal-response interaction.As proteins, cells, and tissues form and grow, they are responsive, not just through passive, physical responses like bending or folding, but also through active responses, such as expressing genes and building proteins.A signaling system may respond to any range of inputs, including abiotic factors like CO 2 levels, or internal factors like hormone levels or inflammation (e.g., Ip & Davis, 1998;Kim et al., 2010).For example, a muscle cell with androgen receptors will bind circulating testosterone, resulting in a signaling cascade within the cell and downstream effects on gene expression related to muscle fiber size.In evolutionary developmental biology, the concept of the signal-response system forms the basis of a key hypothesis-that regulation of developmental genetic modules may act a basis of evolutionary change.For instance, evolutionary tweaking of transcription factors that regulate segment identity (Hox genes) has led to diversification of limb traits across insects (Heffer & Pick, 2013;Warren et al., 1994).
While we may imagine a single signal-response unit, developmental decisions tend to result from complex networks of signaling systems (Vivekanand & Rebay, 2006).Networks include nodes of interacting pieces, where nodes may have no specific identity (e.g., Draghi & Whitlock, 2012;Mjolsness et al., 1991), or represent specific gene regulatory networks (e.g., Davidson & Erwin, 2006).Responses to a signal can be F I G U R E 3 Schematic representations of signaling systems in development.The basic components of a developmental system consist of a signal-response system, where a cue integrator (e.g., a brain or a cell) responds to developmental signals, whether from the genome or the internal or external environment.These signaling systems exist within complex networks (middle) that play out over time and space (below).Information is transduced by the system within a broader context of noise (e.g., molecules or things in the environment that may interfere with the signal).
both positive (stimulatory) or negative (repressive), and combining such interactions over many individual signal-response units can generate incredible complexity.As these networks play out over time and space, we can see positive or negative feedbacks, robustness and neutral space, thresholds and tipping points, and other complex, nonlinear dynamics (Reed et al., 2017;Zañudo et al., 2017).For example, the testosterone signaling cascade is not a linear unit, but interacts with other cellular signaling systems (Rahman & Christian, 2007).In some cases, we know almost all of the players in a network, for instance, in some metabolic or proteinprotein binding networks (Karr et al., 2012), allowing capture of the entire system in a developmental model.
Incorporating networks into our understanding of developmental processes has led to many insights in evolutionary developmental biology.First, developing networks are generally robust to many environmental and genetic perturbations (Kaneko, 2007;A. Wagner, 2013).Such robustness derives from highly dimensional, nonlinear interactions among many interactors (genes, proteins, tissues, and organs) through space and time (Green et al., 2017;A. Wagner, 2013).Second, networks show that the mapping of underlying genetic variation into phenotypic variation is not a simple, linear developmental process (Vivekanand & Rebay, 2006).Network approaches can translate genetic variation into phenotypic variation in very complex morphological traits, such as teeth (Salazar-Ciudad & Jernvall, 2010).When networks also incorporate environmental signals, underlying patterns of genetic variation may evolve aligned with patterns of environmental variation, which can affect subsequent evolutionary change (Draghi & Whitlock, 2012;D. W. A. Noble et al., 2019).Third, networks of signaling systems can cause developmental bias in evolution, as seen in the diversification of body plans and traits (Davidson & Erwin, 2006;Olson, 2006;Salazar-Ciudad & Jernvall, 2010).Fourth, network approaches reveal several routes to the origins of novel traits.For instance, movement through neutral network space over evolutionary time can shift a lineage into novel phenotypic space (Ciliberti et al., 2007).Phenotypic novelty may more often arise from changes to gene regulatory network connections than from changes to the structural genes themselves, a finding that accords with empirical results (Aguirre et al., 2018;Borenstein & Krakauer, 2008;Levine & Tjian, 2003).
Finally, it is important to note how networks of signal-response systems play out over time and space, resulting in morphological differentiation.For example, as testosterone is expressed in developing male mammals, a common abdominal ground plan acquires traits characteristic of male genitalia.As stem cells divide, their daughters differentiate as they receive information on their location in the body.As opposed to the responsiveness of a network at one point in time, here, we have within-individual inheritance of developmental information, as genetically encoded information becomes an embedded feature of the developmental system (Jablonka, 2002;Oyama et al., 2001;Robert et al., 2001).As lineages of cells divide, they inherit information from previous cell generations in a way analogous to evolutionary lineages (rev.Galperin, 1998), whether that is through epigenetic modifications to DNA or chromatin (Atlasi & Stunnenberg, 2017;Xue & Acar, 2018) or spatial information such as morphogen gradients that trigger axis and segmental identity formation (Briscoe & Small, 2015).As information is passed from one developmental unit to another, patterns begin to form, laying the groundwork for identity formation and the emergence of trait function.Considering the relative timing and spatial location of developmental information gives insights into patterns of phenotypic variation and diversification.For example, information received earlier in development should result in a greater potential for different phenotypic trajectories (Diggle, 2002;Wheeler, 1986); spatially restricted information may cause the differentiation of one module, whereas information spread over the entire developing system may result in a more integrated phenotypic response (e.g., Beldade & Brakefield, 2003).

| Development of trait function
When does a heart become a heart during development?At the moment cardiac tissue cells begin to differentiate, at the moment it starts to beat, or when it can power a running body?The third category of developmental features gets at the heart of a philosophically fraught, but important, term in biology: "function."It is challenging to define the function of a biological trait, because how the trait is working in an immediate sense may not directly relate to its function in an evolutionary sense (Garson, 2016;Neander, 1991).However, the idea of trait function also allows us to begin to connect trait development to individual fitness, because when a trait is functioning, it is contributing at least somewhat to survival and reproduction.
As developmental responses play out over time and space, developmental modules begin to "function," in whatever way they contribute to the overall performance and eventual fitness of the individual.Amino acids become types of proteins that catalyze reactions, and cells differentiate to form tissues and organs that move resources and waste.How and when the trait begins to function can play a critical role in the opportunity for environmental influences and the integration across developing modules within an individual.For instance, the vertebrate heart begins to beat very early in development, well before an individual is self-sufficient, and the initial circulatory flow plays an important role in subsequent development of the heart, the circulatory system, and the rest of the body (Kemmler et al., 2021;Risau, 1997).
As developmental modules begin to function, they are in many cases subjected to within-individual selectionwhere the performance of that trait feeds back to either reinforce or atrophy developmental variants (Figure 4).This within-individual selection, also termed somatic or developmental selection, represents a distinct feature of development that involves the performance of individual developmental units that may vary over time or space, such as a motor pattern that changes during learning, or muscle cells that vary spatially in how they experience a mechanical force (Frank, 1996;Snell-Rood, 2012;West-Eberhard, 2003).Developmental selection can act as a mechanism of building complex traits, sometimes with information internal to an individual, such as the role of embryonic movement in motor development or spontaneous neural activity in neural development (Penn & Shatz, 1999).Here, the relative performance of developmental variants may depend on a range of factors, such as mechanical stress, oxygen stress, or ion fluxes (Buffelli et al., 2003;Moore, 2003;Risau, 1997).In many cases, the performance of developmental variants depends on interactions with the external environment, which then links this developmental feature to plasticity (see below) and fitness.For example, in the development of acquired immunity, variation in B cell clones is selected upon by both internal interactions (those that react with selfantigens are eliminated), and external interactions (those that match pathogens are amplified and refined).
Within-individual selection is a powerful mechanism for the development of complex phenotypes from relatively "simple" genomes (e.g., building a complex brain through learning or precise antibody-pathogen matches through acquired immunity, Frank, 1996;Matthey-Doret et al., 2020).Incorporating performance cues during development can increase developmental robustness (Matthey-Doret et al., 2020), minimizing the impacts of genetic and environmental perturbations.A genetic mutation that affects the shape of a bone could be "accommodated" through these processes (West-Eberhard, 2003).However, sampling a range of alternate phenotypes necessitates time and energy, thus developmental selection can be costly (Snell-Rood et al., 2018).

| Developmental interactions with the external environment
The majority of our discussion of developmental features so far has focused on "within-individual" features.Of course, individuals are growing and existing within an external environment, which affects their development and eventual fitness.Both evolutionary and developmental biology have long recognized that development is responsive to the external environment (Gilbert & Epel, 2009;West-Eberhard, 2003).Delineating responsiveness to the internal versus the external environment is important because developmental processes within an individual are all under selection together, while the external environment is subject to diverse and often conflicting evolutionary pressures.
F I G U R E 4 Schematic representations of how we think about function in development.Considering trait function links the phenotype (over time) to performance, which affects fitness, for instance, through energy gain.We can also consider how trait performance at one time point feeds back to affect trait development.Through "developmental selection," trait variants that work particularly well are amplified and those that work less well tend to atrophy.

SNELL-ROOD and EHLMAN
| 399 There is a wide range of ways that biologists have thought of responsiveness to the external environment, or phenotypic plasticity, in their models of development and evolution.For instance, we commonly use reaction norms, graphical representations of how the environment affects the translation of a genotype to a range of phenotypes, to capture plasticity in development (Figure 5; Schlichting & Pigliucci, 1998).Behavioral decision rules (Fawcett et al., 2014), developmental thresholds (Leimar et al., 2006), phenotypic variation (Gomez-Mestre & Jovani, 2013), and state-dependent life histories (McNamara & Houston, 1996) also capture environmental responsiveness in development.Including plasticity in evolutionary models impacts persistence and evolution of populations in novel and changing environments: bottlenecks are less severe, preserving underlying genetic variation (Chevin & Lande, 2010;Chevin et al., 2010;Gomez-Mestre & Jovani, 2013).
Bayesian approaches offer a particularly promising tool to integrate both inherited and acquired information throughout development (Dall et al., 2005;Stamps & Frankenhuis, 2016).Here, innate or learned biases form the starting point of a developmental sequence (the "prior"), and information about the external environment and developing traits updates this prior (to the "posterior") as a function of the reliability of information (Stamps & Frankenhuis, 2016).When evolutionary models use Bayesian approaches to capture development, the prior, the sampling process, and the updating process can evolve.These models can predict developmental phenomena as a byproduct of the sampling process: for instance, lack of reliable cues extends the critical period for trait plasticity (Panchanathan & Frankenhuis, 2016).Bayesian models of development have shown how environmental fluctuations shape patterns of plasticity or how environments shape developmental programs (Ehlman et al., 2022;Walasek et al., 2022).
In many organisms, aspects of development play out in very controlled environments that are themselves evolving, an idea coined as "developmental niche construction" (Flynn et al., 2013;K. Laland et al., 2016).In some cases, specific tissues or organs such as the placenta create very particular temperature, nutrient, and oxygen conditions for an individual's development (Blackburn, 2015).In other cases, certain parental behaviors such as incubation, resource provisioning, or placement of eggs in particular locations can bias developmental environments (Badyaev & Uller, 2009;Mousseau & Fox, 1998).Even individual preferences for microhabitats or social environments may bias subsequent development (Saltz & Foley, 2011).Such niche construction may be quantified as correlations between genotypes and developmental environments (K.N. Laland et al., 1996;Saltz & Nuzhdin, 2014), or specific molecular interactions, such as placenta evolution (Barta & Drugan, 2010;Serov et al., 2015).The idea of the "holobiont" also fits here, where microbial aspects of the external environment have become a stable part of the developmental environment (Gilbert & Fusco, 2019;Rosenberg et al., 2007).Developmental niche construction can create complex evolutionary feedbacks as the tissues and behaviors that create developmental environments can themselves evolve (Kerr et al., 1999;Kylafis & Loreau, 2008; similar to "indirect genetic effects" Wolf et al., 1998).Developmental niche construction can also strengthen environment-genotype correlations, speeding adaptation and the evolution of specialization (Edelaar & Bolnick, 2012;Ravigné et al., 2009).It can also affect the maintenance of genetic variation (Hui & Yue, 2005;Silver & Di Paolo, 2006), facilitating the accumulation of cryptic genetic variation relevant to less frequently experienced environments, which may influence evolutionary processes if the environment then changes (Schneider & Meyer, 2017;Snell-Rood et al., 2016).

| FUTURE DIRECTIONS
In this review, we have offered a brief primer on features of development used across different fields in biology.We believe that some of the recent debate on whether we need more development in evolutionary theory stems from the fact that different biologists use different concepts of development in their approaches.By articulating what we mean by "development," we can clarify some of these disagreements, and work to build a more general theory of evolutionary developmental biology (e.g., Scheiner & Mindell, 2020).Here, we review a few additional future directions opened by explicitly considering a range of ways to broaden the metaphorical map connecting genotype to phenotype in evolutionary models.

| How much does development matter?
Incorporating aspects of development into evolutionary models changes evolutionary dynamics, and offers new insights, but the full breadth of these scenarios remains underexplored.For certain evolutionary questions, such as predicting microevolutionary changes from standing genetic variation, the inclusion of complex developmental features many not appreciably alter predictions.In contrast, for other questions, such as the origin of novel traits or body plans, incorporating more complex features of development may be crucial.With this primer we hope to stimulate a more systematic exploration of the instances in which the inclusion of development in evolutionary models matters versus those in which development plays only a supporting role.We hope that by delineating specific features of development, we spur theoreticians to further formalize these features, mixing and matching developmental features in new ways.Such formalizations have already made some headway in specifying when developmental features in evolutionary models are likely to be important.For instance, using a reaction norm approach to ask questions about evolutionary rescue suggests that the importance of plasticity decreases as environmental predictability decreases (Ashander et al., 2016).It is possible, however, that using a Bayesian or developmental selection model of plasticity (Matthey-Doret et al., 2020) would yield a different answer as learning-like mechanisms are more likely to result in an adaptive plastic response to a novel environment, preventing a population bottleneck (Snell-Rood et al., 2018).How much does developmental detail matter?Many of the conceptualizations reviewed here include precise molecular detail, in contrast to more phenomenological models.By exploring a range of developmental features, we can test when and how certain aspects of development matter for evolutionary questions.

| Predicting macroevolutionary patterns of diversification and speciation
Basic models of evolution exceed at answering microevolutionary questions about allelic changes arising from standing genetic variation from one generation to the next.However, such models often fall short when thinking about macroevolutionary questions: can an expanded view of development help to bridge micro-and macroevolutionary processes (Erwin, 2010;Reznick & Ricklefs, 2009)?For example, building time, space, and signaling into a model's conceptualization of development offers new approaches to questions about speciation, such as incompatibilities between alleles (Orr et al., 2001) or how within-genotype and within-species variation translates into variation across species (e.g., genetic assimilation, Lande, 2009).Including aspects of nutrition and growth allows new ways to model resource-dependent tradeoffs, testing how these may shift across species (e.g., Lorch et al., 2003).We can rethink the classic beak of the finch with models of tissue development (e.g., Kuhl, 2014) that can bridge microevolutionary changes in beak depth with drought, to macroevolutionary shifts in beak shape across species.Verbal evo-devo models can be translated into more quantitative models of evolutionary processes by incorporating these developmental features into evolutionary models.

| Expanding our study of function in development
In reviewing features of development, we note that different subdisciplines tend to focus on different features to varying degrees, and some features are studied more often than others.For instance, molecular biology often focuses on signal-response systems and evolutionary biology often focuses on plasticity and the role of the environment; both fields spend relatively less time on considering the emergence of function over development.While functional approaches to biological traits abound (e.g., functional morphology), we may benefit from more frequent integration of the idea of function into our understanding of development, as function is a critical part of linking traits to performance and fitness.Interesting questions to explore in this area include possible tradeoffs between rapid development and optimal performance, or opportunities in a life cycle for environmental information to affect development (e.g., Snell-Rood et al., 2015).It might also be interesting to consider traits that shift function over developmental time, such as mouthparts that graze in larvae versus puncture in adults-might such functional shifts be more costly or less developmentally integrated?Finally, we might explore new ways to model how an individual effectively and efficiently samples function over developmental time-finding the best phenotype in an infinite phenotypic landscape has parallels to optimizing search SNELL-ROOD and EHLMAN | 401 algorithms in machine learning or phylogenetics (Giribet, 2007;Osugi et al., 2005).

| Applying theory across fields: Signaling systems
In reviewing the common features of development, we can often make connections across fields and begin to apply insights and questions from one field to another.For example, noting the signal-response system as a basic feature of a developmental feature (Rodbell, 1995), we can potentially link to a range of hypotheses in behavioral ecology, and ask how they might apply to the evolution of developmental systems.Here, we can think of "information" as something "perceived" by a developmental unit that is correlated with something occurring elsewhere within the developing individual (Dall et al., 2005).As development plays out over time and space, information moves between developmental modules (organelles or cells or tissues), coordinating developmental processes.For example, in making "decisions" about next developmental steps, a cell may use a hormone signal from a reproductive organ or a morphogen gradient as a signal of current physiological state or positional information (e.g., Collaer & Hines, 1995;Gurdon & Bourillot, 2001).The idea of "information" has moved from the genome to the developmental system (DST; Oyama et al., 2001;Robert et al., 2001); this internal information is evolving in parallel with the developing individual (i.e., the same unit of selection; Williams, 1992), in contrast to external information which is subject to different evolutionary forces (e.g., Kilner, 2006;Zuk & Kolluru, 1998).
How might such a behavioral ecology "information" perspective on development yield new insights?We know that noise plays a role in the perception and reliability of signals and may favor the evolution of complex and redundant signals (Hebets & Papaj, 2005;Partan, 2017).Similar processes may play out in developmental signaling systems, for instance, in WNT and Notch signaling, where the network structure of the signaling systems may result in noise-filtering through interacting redundancy (Hayward et al., 2008).Concepts from behavior allow predictions for how developmental signaling systems arise over evolutionary time.Signals are likely to evolve from cues that are especially predictive of a given environmental state (Rubi & Stephens, 2016) or more likely to be detected by the receiver (sensory or reception biases; Bradbury & Vehrencamp, 2000;Endler, 1992;Laidre & Johnstone, 2013): is this also true for the molecular genetic signalresponse system?In another example, theory on the "honesty" of signaling systems suggests that reliable signals of a state are more likely when interests of the players are "aligned" and/or when signal investment is costly (Dougherty, 2021;Polnaszek & Stephens, 2014).Do we see a role for "honest" developmental signaling related to cooperative cell interactions, for instance in the context of understanding the emergence of cancer (Merlo et al., 2006).
Analogies with behavior continue in thinking not just about individual signal-response situations, but in terms of memory of a developmental system.As signalresponse systems play out over time, within-individual inheritance invites analogies with memory formation and adaptive forgetting (Ferrari et al., 2010), as in immune "memory" in acquired immunity (Segel & Cohen, 2001;Wilson & Garrett, 2004).Indeed, the degree to which information persists over developmental time may depend on the molecular mechanism, such as chromatin state and cell pluripotency (Gaspar-Maia et al., 2011) and presents opportunities for future theory development, for instance, around questions about the evolution of developmental flexibility (e.g., Aoki & Feldman, 2014;Klein et al., 2002).

| New ways to communicate evolutionary biology
Exploring what we mean by "development" in evolutionary biology can be useful in thinking about how we explain evolution to nonbiologists.The textbook Mendelian model of evolution (Figure 1) is how most students are introduced to evolutionary biology, and for most of the public, it is the end of the conversation.It is no wonder that the "blueprint" model of development persists-it is a "sticky" metaphor, that fits with the toy model of evolution, but it is a poor metaphor for questions about the development and evolution of most complex traits (Heath & Heath, 2007;Nijhout, 1990).Many of the stumbling points in the acceptance of evolutionary processes emerge from misunderstandings of the origins of novel, complex traits, something that incorporating more developmental realism can address (e.g., in refuting intelligent design models; Pennock, 2003).One might even hypothesize that continuing challenges around the nature-nurture debate (Zuk & Spencer, 2020) could be addressed by revisiting how we approach "development" in evolution.Do other approaches to conceptualizing development make it easier to grasp evolutionary theory and biology in general?A Bayesian approach may make it easier to understand how environmental influences factor into the development of behavior, rather than the blackand-white thinking of nature-nurture or heritability estimates (Stamps & Frankenhuis, 2016), while a developmental systems approach may be a more intuitive way to introduce macroevolutionary concepts (Griffiths & Tabery, 2013).Similarly, as discussed more below, an agential perspective may unite the diverse perspectives here, and prove a more effective way to communicate developmental evolutionary processes.The efficacy of these different metaphors of development and evolution could be tested in the classroom (Kampourakis & Minelli, 2014).

| The concept of agency allows an integrated expansion of traditional models
We have couched this primer of developmental features within traditional models of evolution: how can we expand the mapping of genotype to phenotype with different concepts of development?This has the advantages that many existing ways that teach or model evolution can be expanded with different features of development that incorporate more biological realism and result in more insights with respect to macroevolutionary processes.However, constraining ourselves to this framing also comes with some limitations which parallel other critiques of the metaphor of the genotype-to-phenotype map (e.g., Milocco & Salazar-Ciudad, 2019).Some of the calls for revising evolutionary theory have focused on causal linkages between genotypes and phenotypes that are bidirectional, whereas the mapping metaphor may bias our thinking to unidirectional thinking.Incorporating feedbacks from the environment and the developmental process itself add some bidirectionality to this model (e.g., Figure 5), but also result in a bit of a mess to the metaphor.
One way forward is to layer the concept of an agent on top of this process."Biological agency" emphasizes how "living systems have evolved to be robust, responsive, flexible, self-synthesizing and self-regulating" and regulate their "structures and activities in response to the conditions they encounter" (Sultan et al., 2022).Within this framework, ideas about environmental responsiveness can be formalized in models of adaptive developmental plasticity, those about robustness and selfregulation might be modeled using gene regulatory networks, and self-synthesizing aspects of biological agents might be encapsulated in evolutionary models of growth and developmental selection.In short, many of the elements that are discussed in the biological agency framework can begin to be modeled using features of development outlined here.This suggests that the avenue offered here-of highlighting how common features of development might expand simpler models of evolutionary patterns and processes, may offer a bridge between disparate literatures and terms around which debate has simmered for the last decade.In conceptual terms, we may think of the many arrows between environment, phenotype, genotype, and performance (e.g., Figure 5), encapsulated within an individual organism or agent (Figure 6), shifting the emphasis in evolution from genes to organisms, including their physiology to behavior (e.g., Moczek, 2012;D. Noble et al., 2014;Walsh, 2014).
This perspective (Figure 6) also gives us a chance to evaluate what the "expanded Mendelian model" is still missing.For instance, the developmental features reviewed here include some of the components of the extended evolutionary synthesis, such as plasticity and niche construction (Figure 5), but it is more challenging to capture concepts related to developmental bias and inclusive inheritance (K.N. Laland et al., 2015).In addition, we primarily focus on aspects of development that affect the genotype-to-phenotype relationship within a generation (Figures 1 and 6), but of course, many aspects of development play out across generations.This suggests that we may especially need new modeling frameworks for these concepts in evolution.In other words, some aspects of the extended evolutionary synthesis may simply elaborate existing models with F I G U R E 6 Layering agency on top of developmental features that involve interaction with the environment.Incorporating complex interactions between the environment, phenotypes, and development shifts the focus of evolutionary processes from genes, to the entire agents that play a role in their own development and performance.As the traditional model of evolution becomes more complex when features of development are added, an agential perspective can simplify the conceptual framework.Instead of a pool of "genes" as the subject of evolution, we shift our attention to a population of agents, and how they change over generations (Figure 1).

SNELL-ROOD and EHLMAN
| 403 new developmental features, as we discuss here, but others may require more fundamental changes to the field (K.N. Laland et al., 2015;Lewens, 2019).

| CONCLUSIONS
In this manuscript, we have reviewed basic features of development that broaden the metaphorical map linking genotype to phenotype.We have drawn from decades of research around development and its interaction with evolution.We have outlined key features of development, in order of increasing complexity, that can be incorporated into the Mendelian model of evolutionary processes, potentially altering evolutionary dynamics and providing approaches to address macroevolutionary questions (Snell-Rood & Ehlman, 2021).These features are not meant as hard categories as they all interact and overlap when considering the development of a complex trait.Instead, this collection of features is intended to facilitate crosstalk amongst fields and generate consideration of the components of development that we have (or have not) incorporated into our own research and theory.Linking all of these developmental features within an individual organism or "agent" provides a way of conceptualizing the resulting mess of causal arrows in development, which reduce the causal primacy of the "gene" (D.Noble, 2008;A. Wagner, 1996A. Wagner, , 1999)).
For new students in evolutionary developmental biology, we hope this primer can clarify some of the confusion in the field around incorporating development into evolutionary biology-we have many, diverse concepts of development, from maps of genetic interactions, to plasticity and developmental niche construction.This review can help clarify when more complex features of development are necessary in models of evolutionary processes, and thus quell some of the debate on the need to incorporate "more" development into evolutionary biology.At the same time, this approach may highlight where the expanded Mendelian model falls short, and where the development of novel theoretical approaches may be needed to advance our integration of development with evolutionary biology.

F
I G U R E 2 Schematic representations of time, growth, and space in development.Concepts of time consider traits or phenotypes at different time points, whether this is different periods of embryogenesis, or different life stages across an organism's life cycle.Growth and space add a spatial element with the consideration that proteins, cells, tissues, and traits are growing over time.With this, spatial elements come potential interactions within a developing individual.

F
I G U R E 5 Schematic representations of how we consider the external environment in development.Plasticity refers to situations where the development of a phenotype depends somewhat on information from the environment; the environment may also affect how a trait performs, which can feed back to affect development through learning or developmental selection.When an individual's traits affect their developmental environment, termed "developmental niche construction," complex evolutionary feedbacks can emerge.