Theories emphasize that populations can adapt differently depending on the type of environmental fluctuations (see above). The experimental evolution setup in which rapidly reproducing species are let to evolve multiple generations in different kinds of environments is capable of resolving whether evolutionary responses to constant and fluctuating environments are similar and whether adaptations are specific to particular kinds of fluctuations. For example if experimental bacterial strains can be identified from each other (Bennett and Lenski, 1993; Ashrafi et al., 2017), it allows comparing fitness of the strains in competition.

For example one can ask if strains adapted to environmental fluctuations are superior in both constant and in fluctuating environments, in comparison to strains evolved in a constant environment. When this information is projected on tolerance curves obtained at constant environments it unambiguously indicates if fitness in fluctuating environments is reflected in the tolerance curves.

Ketola and Saarinen (2015) did not utilize marker strains and they used fitness surrogates in their experiments. Still, this study provides insights on the value of tolerance curves obtained from constant thermal environments. The idea in Ketola and Saarinen (2015) was to test if fluctuations increased growth or yield of bacterial clones during fluctuations, which could be seen as an adaptation to prevailing conditions. Accordingly, strains evolved in fluctuating regimes had higher growth rate under fluctuations than strains evolved in constant environments. Thus, these experimentally evolved bacteria provided clear evidence for an evolved ability to tolerate fluctuations. However, when thermal tolerance curves were based on several measurements obtained across the range of environments experienced during the process of experimental evolution, no evidence of adaptation to fluctuations could be deducted from the tolerance curve. On the contrary, the strains adapted to fluctuations were outperformed by strains evolved at constant environments. These data clearly demonstrate that evolutionary processes are distinct in fluctuating and constant temperature environments. Recent work on the insect, Drosophila simulans, draws similar conclusions (Manenti et al., 2016).

We propose that experimental evolution experiments that test different and ecologically relevant frequencies of variation are highly needed. Such experiments will resolve if fluctuations that are fast or slow, frequent or infrequent, predictable or not, and have high or low amplitude, have their characteristic adaptations—as theories predict (see above). Such studies will elucidate if tolerance curves from constant environments fail to capture adaptations in certain kinds of fluctuations, as we propose. Based on the reasoning put forward in papers by Schulte et al. (2011) and Sinclair et al. (2016) it is likely that adaptations to fast fluctuations are especially hard to observe from tolerance curves obtained across constant environments. Thus, more work should be done with different kinds of environmental fluctuations (Ketola et al., 2004; Buckling et al., 2007; Manenti et al., 2014), to disentangle critical frequencies, amplitudes, or predictabilities, where traditional tolerance curves may start to fail. Experimental evolution studies have a lot to provide in this context.

There are a large number of quantitative genetic studies exploring the amount of genetic variation in e.g. heat shock survival or in tolerance curve parameters (Scheiner, 1993; Kellermann et al., 2009; Ketola et al., 2014; Kristensen et al., 2015). However, as stated here the adaptive benefit of heritable variation in some proxy of fitness in a constant environment might be minor in fluctuating environments if it is not genetically correlated with fitness in fluctuating environments (Ketola et al., 2014; Manenti et al., 2016). What is crucially missing in most quantitative genetic experiments performed so far is correlating the abovementioned proxies to fitness in fluctuating environments. For example, in a massive experiment we reared half-sib families of D. melanogaster under four different thermal environments, two constant and two fluctuating temperatures, and quantified egg-to-adult viability under these four thermal conditions. Two constant temperatures were used to draw simple linear tolerance curves and to estimate its parameters. Next we resolved if these parameters (elevation and the slope) were genetically correlated with egg-to-adult viability in fluctuating environments. We found that the elevation was under positive selection in fluctuating environments but not the slope. However, interestingly the elevation and the slope together explained a rather small proportion of the variance in egg-to-adult viability in fluctuating environments, supporting the idea that the traditional tolerance curves obtained in constant environments might not be enough for describing performance when environments fluctuate (Ketola et al., 2014).

We advocate that quantitative genetic studies aiming at generating ecologically relevant tolerance curves should use similar designs but investigate more temperatures, or vary other environmental components, and a range of different frequencies and intensities of fluctuations. These should also involve investigating some proxies of tolerance to fast environmental changes, for example including measuring physiological responses to more acute stress like inducible heat shock proteins, metabolic rate, knock down temperatures, or chill coma recovery time. Crucially, these proxies, along with tolerance curve parameters, should also be genetically correlated with lifetime fitness or survival in different kinds of fluctuating “test environments” (Ketola et al., 2014) to resolve which of the proxies are more reliable in predicting tolerances to fluctuations (Rezende et al., 2010). In addition to classic quantitative genetic setups (e.g., full-sib half-sib breeding designs, isofemale lines, or the use of pedigreed populations), using clonal species should also be considered, as this could be an affordable approach to conduct these very demanding experiments.

In addition to experiments, data could also be retrieved from natural populations. Quantitative genetic studies of pedigreed wild populations have revealed selection on tolerance curve characteristics (Nussey, 2005; Charmantier et al., 2008; Debes et al., 2017). Depending if there is between year variation in the amount of environmental fluctuations, field studies on pedigreed populations could allow estimating reaction norms for fitness on the continuum of environmental fluctuations. This approach serves as an important source of information by answering the question whether the same genotypes do well in constant and fluctuating natural environments.

Omics technologies have a lot to offer in relation to pinpointing the physiological and genetic architecture of complex traits in constant and fluctuating environments. For example sequencing of strains/populations from experimental evolution studies evolved in constant and fluctuating environments, or testing performance of gene knock down/knock in strains in both types of environments will provide valuable insight into whether the same or different genes contribute to the tolerance in fluctuating and constant environments (Sørensen et al., 2016; Deatherage et al., 2017). Recent work by Sørensen et al. (2016) is very illuminating in this respect. Exploring gene expression of D. melanogaster flies exposed to different kinds of environments revealed that: “transcriptional responses to mean temperature, acute exposure to heat and fluctuations act through largely independent mechanisms.” This quote from the paper matches the take home message that we wish to convey namely that adaptations to constant and fluctuation environments are partly under distinct genetic control.

Also the ability to use genomic and other omics approaches on non-model species is developing rapidly (Shafer et al., 2015). This provides a range of opportunities to pinpoint genes and mechanism responsible for fitness and plastic responses in the field on a wide range of species. This does not leave laboratory experiments out-of-date but do provide additional opportunities that should be exploited.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.