Introduction
Phenotypic evolution is biased by the mechanisms that link genetic
variation to phenotypic variation, i.e. the genotype-phenotype map
(Alberch, 1991; Pigliucci, 2010). Mechanisms linking genotype to
phenotype are characterized by both robustness and flexibility. Robust
developmental processes buffer phenotypes from noise such that multiple
transcriptional, biochemical, or cellular network configurations give
rise to similar organismal phenotypes, while flexibility in these
processes allows for phenotypic plasticity in response to environmental
conditions (Marder & Goaillard, 2006; Siegal & Leu, 2014; Wagner,
2011). Though contrasting, both phenomena are ultimately bounded by
phenotypic, developmental, genetic, and mechanistic constraints. The
complex nature of the genotype-phenotype map and the existence of
‘many-to-one’ mappings across hierarchical levels of biological
organization make it challenging to understand how mechanistic
properties facilitate or constrain adaptive evolution. Here, we combine
studies of phenotypic plasticity and parallel adaptation to assess how
mechanistic biases shape evolutionary trajectories across timescales.
Studies of parallel adaptation can reveal biases in mechanisms of
evolution by asking whether similar mechanism are associated with
repeated, independent evolutionary transitions. Similar mechanisms
underlying parallel evolutionary transitions suggest the
genotype-phenotype map is constrained by limits on the possible ways to
construct adaptive phenotypes (Losos, 2011). In contrast, non-shared
mechanisms suggest that mechanistic flexibility may facilitate evolution
by providing multiple alternative ‘solutions’ to a given adaptive
problem and providing alternative paths by which organisms can reach
adaptive peaks (Badyaev & Morrison, 2018; Drion, O’Leary, & Marder,
2015; Grashow, Brookings, & Marder, 2009). Such mechanistic flexibility
may allow organisms to more easily reach the same adaptive peak from
alternative starting points, and additionally buffer organisms from
pleiotropic effects via compensatory mechanisms that rely on the
many-to-one mappings created by these alternative paths.
One potential consequence of the biological robustness generated by
many-to-one mappings is that variation in the function of only a small
number of key genes will alter organismal phenotypes while variation in
most genes has no phenotypic consequences (Yang, Maclean, Park, Zhao, &
Zhang, 2017), either because variation does not propagate to higher
levels of organization and/or because variation is compensatory. In this
case, phenotypic evolution may rely on a limited number of mechanistic
paths, repeatedly targeting those mechanisms that that yield the
greatest phenotypic responses with the smallest pleiotropic costs.
Indeed, compelling examples demonstrate that similar phenotypes share
underlying neural, physiological, molecular, and/or genetic mechanisms,
even across highly divergent taxa (e.g. Insel and Young, 2000; Manceau
et al., 2010; Pankey et al., 2014; Rosenblum et al., 2010). In contrast,
other studies demonstrate flexibility in underlying mechanisms
suggesting that different mechanistic ‘solutions’ can give rise to
shared phenotypes in closely related species, among populations of the
same species, or even among individuals of the same population (e.g.
Abouheif and Wray, 2002; Crawford and Oleksiak, 2007; Drion et al.,
2015; Grashow et al., 2009; Mandic et al., 2018). Nevertheless, even if
mechanistic flexibility is a common feature of robust biological
networks, shared genetic background or patterns of pleiotropy in a
lineage may still direct evolutionary paths toward predictable
mechanistic pathways, especially at long evolutionary timescales (Gompel
& Prud’homme, 2009; Stern & Orgogozo, 2008). Empirical evidence for
both shared and distinct mechanisms underlying parallel phenotypic
evolution leaves open the question of when and why either pattern
dominates.
One feature of biological networks that may channel divergence into
particular paths is environmentally induced plasticity. First, both
plastic and evolutionary processes may rely on those mechanistic paths
that yield the greatest phenotypic responses. In this case, we expect
plastic and evolutionary responses to share underlying mechanisms,
although they may not act in the same direction. Theory predicts that
plastic and evolutionary changes will be in the same direction when
plasticity in a novel environment is adaptative, increasing immediate
survival and allowing time for evolutionary divergence via co-option of
mechanisms involved in environmentally induced responses (Baldwin, 1896;
Ghalambor, McKay, Carroll, & Reznick, 2007; Lande, 2009; West-Eberhard,
2003). In contrast, plastic and evolutionary changes will be in opposite
directions when plasticity in a novel environment is non-adaptive
thereby increasing the strength of selection or when plastic responses
‘overshoot’ adaptive optima and are compensated by selection (Conover,
Duffy, & Hice, 2009; Ghalambor et al., 2007; Grether, 2005; Velotta &
Cheviron, 2018). Importantly, plasticity may facilitate adaptation under
either scenario and empirical studies document both patterns. In
addition, plastic and evolutionary processes may rely on shared
mechanistic paths because the mechanisms mediating phenotypic plasticity
promote the accumulation of cryptic genetic variation that is released
under new environmental conditions, thereby fostering associations
between plasticity and divergence (Draghi & Whitlock, 2012;
Espinosa-Soto, Martin, & Wagner, 2011). Finally, while associations
between phenotypic plasticity and evolution have primarily been
considered in the context of phenotypic change, associations between
developmental plasticity and genetic divergence at the level of
underlying mechanisms could also arise if initially plastic compensatory
or homeostatic responses to maintain, rather than alter, organism level
phenotypes become genetically fixed as adaptation proceeds (Velotta &
Cheviron, 2018), in which case we may also expect the evolution of
plasticity itself.
In the present study, we take advantage of parallel phenotypic evolution
in independent lineages of Trinidadian guppies (Poecilia
reticulata ) to explore patterns of flexibility and constraint in
transcriptional mechanisms mediating repeated adaptation. We compare
whole-brain gene expression patterns based on developmental experience
and genetic background to test four hypotheses: (1) that genes with
significant expression plasticity are more likely to show genetic
divergence in expression, (2) that the direction of plastic responses
predicts the direction of genetic divergence, (3) that gene expression
plasticity itself evolves, and (4) that parallel phenotypic adaptation
across independent lineages relies on shared gene expression changes.