Challenge 3: Agents and targets of selection
General Application — Quantifying the agents and targets of natural selection is essential for understanding local adaptation (Kawecki and Ebert 2004) in any environment, yet is inherently difficult (Endler 1986). Targets of selection may be misidentified or confounded in both phenotypic and genomic approaches due to a poor understanding of the relationships between genotype, phenotype, and environment (Linnen and Hoekstra 2009, Bierne et al. 2011, Hoban et al. 2016). Disentangling selection on single versus multiple correlated traits can be particularly difficult because of genetic, developmental, and functional constraints (Hill and Robertson 1966, Price 1970, Lande and Arnold 1983). The genetic architecture of a phenotype can also complicate genomic tests for local adaptation as polygenic traits may be more difficult to detect in selection scans compared to single locus traits (Hoban et al. 2016). Given the suspected prevalence and importance of polygenic adaptation and that rapid adaptation may involve soft rather than hard selective sweeps, identifying genomic targets of selection may be difficult for many complex phenotypes (Rockman 2011, Messer and Petrov 2013). In addition, large sample sizes and powerful statistical methods may be required to detect what are typically small selection coefficients (Kingsolver et al. 2001), and episodic or age-specific selection may lead to confusion as to when selection has occurred (Grant and Grant 2014). The signatures of past and contemporary selection can be difficult to differentiate (Haller and Hendry 2014) as phenotypes may arise in response to selective pressures in the contemporary environment but also may have arisen under ancestral selective regimes (i.e., are exaptations) or as a consequence of non-adaptive processes (e.g., gene flow). Lastly, in any environment humans can directly or indirectly change factors affecting selection and adaptation such as resource availability, resource distribution, population connectivity, and habitat size.
Human Element — The urban environment is human-built, thus many of the agents of selection are anthropogenic and not previously encountered by organisms or researchers in non-urban environments (Lugo et al. 2018, Alberti 2015). For example, extensive impervious surfaces (e.g., asphalt) within cities can impact local climate because they absorb and radiate solar energy differently than natural substrates (the “urban heat island” effect, Oke 1973), and high concentrations of anthropogenic pollutants in urban habitats could accelerate mutation rates (Yauk et al. 2000, Somers et al. 2004, Johnson and Munshi-South 2017). Understanding these anthropogenic pressures may require cross-disciplinary collaboration (e.g., engineering, physics, chemistry, governance, urban planning; McPhearson et al. 2016). Moreover, teasing apart the relative importance of local adaptation, exaptation, and non-adaptive (e.g., gene flow) origins of urban phenotypes can be particularly challenging in urban environments. For example, as a consequence of human-associated population connectivity, pigeons (Columba livia ) in the Northeastern United States form a large continuous genetic metapopulation spanning city centers separated by over 800 km (Carlen and Munshi-South 2020). In fact, due to human-mediated movement, some organisms have a higher probability, frequency, and distance of dispersal in somewhat predictable ways (e.g., intercity translocations; Gotzet et al. 2015, Bennett et al. 2019). For example, urban areas act as hubs to increase connectivity among populations of the Western black widow spider (Latrodectus hesperus ), including among historically and geographically-distinct populations locally adapted to desert environments (Miles et al. 2018a, 2018b).
Misconceptions — A misconception perpetuated by poorly understood agents and targets of selection is that selection in urban environments is strong primarily as a consequence of humans and human activities as agents. Although rates of phenotypic change have been demonstrated to be elevated in response to some anthropogenic agents (Hendry et al. 2008, Alberti 2015), many studies rely on environmental proxies such as impervious surface cover rather than identifying causal relationships. Researchers may conflate environmental proxies with drivers of selection if the selective agents are unclear, multicollinear, or correlated with general environmental features a problem that plagues adaptation research in any environment (Endler 1986, Mitchell-Olds & Shaw 1987, Kawecki and Ebert 2004). For example, in urban crested anoles (A. cristatellus ), limb length differences can be connected to shifts in structural environment directly related to locomotion (Winchell et al. 2016, 2018), although this trait shift could also be explained by the proxy variable of impervious surface cover correlated with structural environment. In addition, contemporary movement patterns of urban organisms influenced directly and indirectly by human activities can obscure the selective landscape that shaped phenotypes. For example, populations of the mosquito Culex pipiens were presumed to be locally adapted to living in subway stations in London, yet a recent review instead supports exaptive origins of these underground-adapted populations, with adaptive phenotypes previously present in the ancestral populations outside of Europe (Haba and McBride 2022). As in any environment, if we fail to first characterize patterns of gene flow and genetic drift, we may incorrectly conclude local adaptation to urban environments (e.g., Gould and Lewontin 1979, Hoban et al. 2016).
Moving Forward — To address the challenges of understanding novel anthropogenic selective pressures, connecting phenotypes to selective agents and accounting for nonadaptive processes is crucial (Santangelo et al. 2018, Miles et al. 2019). Research that connects adaptive urban phenotypes to selective agents through performance or fitness quantification (e.g., Tüzün and Stoks 2020, Chick et al. 2020) will provide more informative evidence of urban adaptation and reduce the conflation of environmental proxies (e.g., general urban characteristics) with drivers of phenotypic change. Genomic approaches may be particularly valuable to examine adaptive responses while accounting for underlying population structure. For example, Salmón et al. (2021) used genotype-environment association tests to identify adaptation in the great tit (Parus major ) across multiple cities, interpreting results in light of population structure analyses suggesting widespread gene flow across city centers. When populations are highly connected, it can be unclear if adaptive phenotypes arose repeatedly or swept across urban populations, a subtle distinction in the evolutionary mechanism underlying adaptation. Teasing apart these mechanisms is possible: Oziolor et al. (2019) used a model developed by Lee and Coop (2017) to determine how both de novo mutation and adaptive introgression contributed to pollution tolerance in Atlantic killifish (Fundulus heteroclitus ). Lastly, long-term datasets, including building museum resources (see Challenge 2) and research on ancient DNA will provide important context for understanding urban adaptation by addressing temporal variation and timescales in natural selection. For example, in non-urban ecosystems, selection on beak size in Galapagos finches (Geospiza spp.) fluctuates from year to year in variable directions, and by building a multidecadal data set, Grant and Grant (2014) were able to quantify these dynamics.