INTRODUCTION
The Earth’s climate is warming, and the cascading stressors from warming may have irreversible effects on microbes and the ecosystem functions that they drive. Between 2011 and 2020, Earth’s land temperatures increased by 1.59°C, which is the largest rise in temperature in the last 2000 years (
1). Temperature impacts the rates of biological processes (
2) and can result in thermal adaptation or acclimation (
3). The ability of microbes to adapt to environmental change may alter ecosystem function (
4). Because soil microbes drive biogeochemical cycles and mediate atmospheric carbon fluxes (
5,
6), we need to understand the effects of long-term warming on soil microbes. Healthy forest soils are characterized by high concentrations of organic matter and abundant and active microbial communities (
7). The activity of these microbes contributes to new organic matter deposition and soil health (
8). Soils serve as a large carbon sink, and healthy soils absorb more carbon than they release. This reduces the amount of carbon dioxide (CO
2) emitted to the atmosphere and buffers against climate change (
9). Microbial adaptation in response to warming due to climate change could thus impact microbial traits associated with soil carbon cycling (
10).
When microbes adapt, populations acquire new traits that fundamentally change how microbial systems respond to changes in the environment. Microbial adaptation can be defined as irreversible changes in microbial traits that increase fitness. Adaptation does not specify a mechanism and can be used to define changes in traits in response to environmental changes at the community level (
4). Phylogenetic comparative methods (PCMs) can be used to test for evidence of adaptation of microbial traits. PCMs test for differences among species while accounting for phylogenetic relatedness (
11 – 13). In this study, we used phylogenetic generalized least squares (PGLS), which transforms trait data based on phylogenetic distance among species then tests for differences in traits. Such trait transformation allows comparisons among species as if they were independent groups (
11). While PCMs test for evidence of adaptation of microbial traits, they do not test for evolutionary adaptation associated with heritable mutations. Using PGLS, we determined whether long-term soil warming is associated with adaptation of microbial growth traits.
To study the impacts of long-term warming on soils, a 30-year field experiment is ongoing at the Harvard Forest Long-Term Ecological Research (LTER) site in Petersham, Massachusetts. Here, experimental soils are heated 5°C above ambient temperature throughout the year since 1991 to simulate the effects of climate change. Five degrees of warming was chosen as a worst-case scenario for the rise in soil temperatures by the year 2100 (
1); control soils received no warming treatment. Increased rates of decomposition following 30 years of warming has led to 34% loss of soil organic matter and increased flux of CO
2 to the atmosphere in the heated versus control plots (
14). An isolate screen and metagenomic analysis showed that the ability of soil microbes to degrade complex carbohydrates also increased in response to rising temperatures (
6). This was preliminary evidence of adaptation to long-term warming and suggests the potential for adaptation of other microbial traits (
6,
14). Given that microbial growth and activity contribute to soil health, we sought to characterize whether microbial growth traits show responses to long-term soil warming and whether they are adaptive.
We selected Alphaproteobacteria as the focus of our study because they tend to be dominant in soils, and because they showed increased absolute abundance in heated plots compared to control plots in a previous community-level experiment of soil microbes (
15). We hypothesized that (i) growth of Alphaproteobacteria from warmed plots is less temperature-sensitive than that from control plots; (2) optimum growth temperature of Alphaproteobacteria from warmed plots is higher than that from control plots; and (iii) maximum growth temperature of Alphaproteobacteria from warmed plots is higher than that from control plots. Given that microbes in heated soils have been exposed to higher temperatures for 22–23 years at the time of isolation, we expect them to have adapted microbial growth traits that are advantageous in warmer temperatures (
16). However, if warming does not result in adaptation of these microbial growth traits, this would suggest that changes in soil carbon dynamics may be a result of other factors such as nutrient availability, changes in microbial biomass and carbon use efficiency, or thermal acclimation (
14).
To directly measure the adaptation of bacterial growth traits due to chronic warming, we measured growth over time and across temperatures for Alphaproteobacteria isolated from the warmed and control soil plots. We estimated the intrinsic growth rate for each replicate isolate at each temperature (
17). The Ratkowsky 1983 model (
18) and a modified version of Macromolecular Rate Theory (MMRT) (
19,
20) were fitted to data for growth rate over temperature for each isolate to estimate temperature sensitivity of growth, optimum growth temperature, and maximum growth temperature. We chose the Ratkowsky 1983 model because it is a widely accepted model for bacterial growth over temperature and the MMRT model because of its underlying thermodynamic theory and application in soil microbial communities. While we chose to fit both the Ratkowsky 1983 and MMRT models, the objective was to present both fits rather than select one. There lacks a standard approach to modeling such data. Furthermore, fitting both models and estimating several parameters allows for increased versatility in the data set which can be used for future work. Finally, we used phylogenetic comparative methods to test for adaptation of soil microbial growth traits (
11 – 13).
DISCUSSION
We expected Alphaproteobacteria isolated from warmed plots would have (i) lower temperature sensitivities of growth; (ii) higher optimum growth temperatures; and (iii) higher maximum growth temperatures compared to isolates from control plots. Our results showed evidence of adaptation of optimum growth temperature quantified by the Ratkowsky 1983 model, but not for other measured traits. Evidence of adaptation of
Topt estimated by the Ratkowsky 1983 model affirm observations from previous studies, where increased optimum growth temperature is associated with warmer soils (
50,
51). However, the lack of differences observed in other microbial growth traits estimated by both the Ratkowsky 1983 and MMRT models may be due to the shape of the temperature response curve, model fitting, or the magnitude and duration of warming, for example. Evidence for this conclusion lies in the observation that the Ratkowsky 1983 model fit was better than the MMRT model for optimum growth temperature.
The difference in evidence of adaptation for optimum growth temperature quantified by the Ratkowsky 1983 and modified MMRT models may be due to a difference in fits. The residual standard errors for the Ratkowsky 1983 fitted model on each isolate are two to three orders of magnitude lower than those of the MMRT fitted models (Table S2). Adequate fitting of MMRT requires a data set to at least capture the optimum growth temperature. Although our data set includes
Topt, it is considerably limited at lower temperatures and lacks growth rate data at the temperature minima. This limitation may be associated with the less accurate MMRT fits, resulting in inaccurate estimations of
Topt. Alster et al. (
52) suggested increasing the number of temperature points for adequate MMRT model fitting, which could be applied in future studies. The difference in evidence of adaptation when
Topt was estimated by the Ratkowsky 1983 and MMRT models suggests that microbial trait estimations may depend on model fits.
There are several differences between the Ratkowsky 1983 model and Macromolecular Rate Theory. Ratkowsky 1983 is an empirically determined model of growth rate over temperature for each isolate (
18). MMRT is based on thermodynamic theory and is not empirically determined. It accounts for changes in the temperature response in the absence of enzyme denaturation at temperatures above the optimum temperature through changes in heat capacity. The residual standard errors indicate that Ratkowsky 1983 is a more appropriate fit for our data of growth rate over temperature compared to MMRT. However, we are particularly interested in MMRT due to its underlying thermodynamic theory, as well as its application in soil ecosystems (
19,
20,
52).
The lack of evidence of adaptation of other microbial growth traits demonstrates the limitations of inferring microbial growth traits based on a single temperature point. Traits such as temperature sensitivity of growth are more nuanced and may be impacted by thermal niche breadth. Thermal niche breadth is the range of temperatures that permits microbial growth. Previous studies observed that changes in the range between minimum and maximum growth temperatures depended on soil incubation temperatures (
53,
54). This suggests that the relationship between growth rate and temperature may also vary between minimum and maximum growth temperatures. The rate at which growth rate changes across temperatures, or the steepness of the temperature response curve, may be impacted by the environment, thus altering thermal niche breadth. Challenges in quantifying change in microbial growth rate over temperature may result if such environmental factors are not fully accounted for. This concept of a thermal niche breadth may have an associated fitness cost, as seen with other microorganisms (
55). Therefore, it may be challenging to identify microbial growth trait adaptation without also considering changing thermal niche breadths.
Pagel’s λ was intermediate (0 < λ < 1) for all microbial growth traits, which suggests that the distribution of traits was not as expected under Brownian Motion. There are multiple explanations for such results. One explanation is that climate warming may be associated with selection of intermediate phenotypes (i.e., stabilizing selection) instead of extremes. This may have resulted in constrained trait evolution. Additionally, changes in evolutionary rate over time may have also resulted in non-Brownian Motion distribution of traits (
51). It is possible that discontinuous substrate availability over the decades of experimental warming could have caused a difference in growth rate, and possibly evolutionary rate, over time (
14). Phylogenetic signal is also often quantified by Blomberg’s
K, which is a variance ratio and has the advantage of being able to be greater than one. However, our data were not suitable for Blomberg’s
K estimations as it resulted in a singular matrix.
Soil microbial growth tends to be limited by substrate availability, so evidence of adaptation from PGLS tests may have been occluded by high levels of nutrient availability in the laboratory growth conditions of these experiments. Kamble et al. (
56) observed that bacterial and fungal growth in soils was carbon limited (
56). In a community-level experiment in a boreal forest, Ekblad et al. (
57) observed that soil microbial biomass was limited by carbon but not nitrogen availability (
57). Although these experiments were conducted in soils on the community level, it is possible that the carbon and nutrient-rich media used in this study may obscure the effect of nutrient availability and substrate-specific growth dynamics of microbes in warming soils. Studying microbial growth under lower nutrient conditions may provide a different perspective on how warming impacts microbial growth traits.
Thermal adaptation of increasing growth with temperature has been observed for other organisms in response to climate warming. Among other microorganisms, growth rate of pathogenic fungi,
Mycosphaerella graminicola was observed to be associated with increasing temperatures (
58). Globally distributed plant pathogens were also found to locally adapt to their environments, resulting in significantly different optimum growth temperatures (
59). Thermal adaptation is also often investigated more broadly among other ectotherms. Villeneuve et al. (
60) observed that growth of
Urosalpinx cinerea (Atlantic oyster drill) was positively associated with spawning temperature (
60). Studying thermal adaptation is highly relevant as the effects of the climate crisis increase. However, doing so is challenging among organisms with longer generation times, which highlights the importance of utilizing techniques beyond lab and field-based experiments and suggests a benefit to studying adaptation among organisms with short generation times and large populations like microbes. It is also possible that the organismic adaptation to temperature appears overly significant due to the difficulty in publishing negative or non-significant results.
Change in microbial growth traits is just one example of how warming may impact soil microbes. Increasing temperatures are also associated with evolutionary selection of organisms with smaller genome sizes, as seen in fire-affected soils (
61). Evidence of adaptation for other metabolic processes, such as respiration, has also been observed (
62,
63). Differences in microbial growth traits between isolates from warmed and control soils may be due to reasons other than adaptation. Such differences may be due to depletion of labile carbon (
14,
64), changing microbial community structure (
14,
65), microbial physiology (
66), and species sorting and functional diversity (
51).
While thermal adaptation of microbial traits has been observed in other studies, our results demonstrate that measuring growth potential may be impacted by additional factors. We used laboratory settings to quantify microbial growth traits, which may be an inaccurate representation of field conditions. Under these conditions, results of our study suggest that warming has not resulted in adaptation of temperature sensitivity of growth and maximum growth temperature quantified by the Ratkowsky 1983 model and temperature inflection point and optimum growth temperature quantified by MMRT. However, optimum growth temperature estimated by Ratkowsky 1983 showed some evidence of adaptation. As temperatures increase, changes in soil microbial growth rate may affect rates of atmospheric carbon cycling. Future exploration of whether growth strategies explain microbial adaptation to warming will help predict changes in microbial community and ecosystem function and allow us to better understand soil microbial responses to warming.
ACKNOWLEDGMENTS
The authors are grateful to all the people who contributed to the isolation, genotyping, sequencing, and annotation of isolates in this study, including (but probably not limited to) Erin Bergeron, Andrew Billings, Isabella Bushko, Gina Chaput, Mallory Choudoir, Emily Clark, Luiz Domeignoz-Horta, Alon Efroni, Spencer Moore, Samantha Murphy, Grace Pold, Damayanti Rodriguez-Ramos, Rachel Simoes, Abigail Sondrini, Bianca Surjawan, and Wing Yin Tam.
This project was supported by a grant from the National Science Foundation (No. DEB-1749206) to K.M.D. The soil warming experiments at Harvard Forest are maintained with support from the National Science Foundation (NSF) Long-Term Ecological Research Program (DEB-1832110) and a Long-Term Research in Environmental Biology grant (DEB-1456610). The authors are grateful to Serita Frey and Mel Knorr for their support of our soil sampling. The work conducted by the U.S. Department of Energy Joint Genome Institute (
https://ror.org/04xm1d337), a DOE Office of Science User Facility, is supported by the Office of Science of the U.S. Department of Energy operated under Contract No. DE-AC02-05CH11231.